Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ 4a2c0db0

History | View | Annotate | Download (398.4 kB)

1
#
2
#
3

    
4
# Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011 Google Inc.
5
#
6
# This program is free software; you can redistribute it and/or modify
7
# it under the terms of the GNU General Public License as published by
8
# the Free Software Foundation; either version 2 of the License, or
9
# (at your option) any later version.
10
#
11
# This program is distributed in the hope that it will be useful, but
12
# WITHOUT ANY WARRANTY; without even the implied warranty of
13
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14
# General Public License for more details.
15
#
16
# You should have received a copy of the GNU General Public License
17
# along with this program; if not, write to the Free Software
18
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
19
# 02110-1301, USA.
20

    
21

    
22
"""Module implementing the master-side code."""
23

    
24
# pylint: disable-msg=W0201,C0302
25

    
26
# W0201 since most LU attributes are defined in CheckPrereq or similar
27
# functions
28

    
29
# C0302: since we have waaaay to many lines in this module
30

    
31
import os
32
import os.path
33
import time
34
import re
35
import platform
36
import logging
37
import copy
38
import OpenSSL
39
import socket
40
import tempfile
41
import shutil
42
import itertools
43

    
44
from ganeti import ssh
45
from ganeti import utils
46
from ganeti import errors
47
from ganeti import hypervisor
48
from ganeti import locking
49
from ganeti import constants
50
from ganeti import objects
51
from ganeti import serializer
52
from ganeti import ssconf
53
from ganeti import uidpool
54
from ganeti import compat
55
from ganeti import masterd
56
from ganeti import netutils
57
from ganeti import query
58
from ganeti import qlang
59
from ganeti import opcodes
60

    
61
import ganeti.masterd.instance # pylint: disable-msg=W0611
62

    
63

    
64
def _SupportsOob(cfg, node):
65
  """Tells if node supports OOB.
66

67
  @type cfg: L{config.ConfigWriter}
68
  @param cfg: The cluster configuration
69
  @type node: L{objects.Node}
70
  @param node: The node
71
  @return: The OOB script if supported or an empty string otherwise
72

73
  """
74
  return cfg.GetNdParams(node)[constants.ND_OOB_PROGRAM]
75

    
76

    
77
# End types
78
class LogicalUnit(object):
79
  """Logical Unit base class.
80

81
  Subclasses must follow these rules:
82
    - implement ExpandNames
83
    - implement CheckPrereq (except when tasklets are used)
84
    - implement Exec (except when tasklets are used)
85
    - implement BuildHooksEnv
86
    - redefine HPATH and HTYPE
87
    - optionally redefine their run requirements:
88
        REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
89

90
  Note that all commands require root permissions.
91

92
  @ivar dry_run_result: the value (if any) that will be returned to the caller
93
      in dry-run mode (signalled by opcode dry_run parameter)
94

95
  """
96
  HPATH = None
97
  HTYPE = None
98
  REQ_BGL = True
99

    
100
  def __init__(self, processor, op, context, rpc):
101
    """Constructor for LogicalUnit.
102

103
    This needs to be overridden in derived classes in order to check op
104
    validity.
105

106
    """
107
    self.proc = processor
108
    self.op = op
109
    self.cfg = context.cfg
110
    self.context = context
111
    self.rpc = rpc
112
    # Dicts used to declare locking needs to mcpu
113
    self.needed_locks = None
114
    self.acquired_locks = {}
115
    self.share_locks = dict.fromkeys(locking.LEVELS, 0)
116
    self.add_locks = {}
117
    self.remove_locks = {}
118
    # Used to force good behavior when calling helper functions
119
    self.recalculate_locks = {}
120
    self.__ssh = None
121
    # logging
122
    self.Log = processor.Log # pylint: disable-msg=C0103
123
    self.LogWarning = processor.LogWarning # pylint: disable-msg=C0103
124
    self.LogInfo = processor.LogInfo # pylint: disable-msg=C0103
125
    self.LogStep = processor.LogStep # pylint: disable-msg=C0103
126
    # support for dry-run
127
    self.dry_run_result = None
128
    # support for generic debug attribute
129
    if (not hasattr(self.op, "debug_level") or
130
        not isinstance(self.op.debug_level, int)):
131
      self.op.debug_level = 0
132

    
133
    # Tasklets
134
    self.tasklets = None
135

    
136
    # Validate opcode parameters and set defaults
137
    self.op.Validate(True)
138

    
139
    self.CheckArguments()
140

    
141
  def __GetSSH(self):
142
    """Returns the SshRunner object
143

144
    """
145
    if not self.__ssh:
146
      self.__ssh = ssh.SshRunner(self.cfg.GetClusterName())
147
    return self.__ssh
148

    
149
  ssh = property(fget=__GetSSH)
150

    
151
  def CheckArguments(self):
152
    """Check syntactic validity for the opcode arguments.
153

154
    This method is for doing a simple syntactic check and ensure
155
    validity of opcode parameters, without any cluster-related
156
    checks. While the same can be accomplished in ExpandNames and/or
157
    CheckPrereq, doing these separate is better because:
158

159
      - ExpandNames is left as as purely a lock-related function
160
      - CheckPrereq is run after we have acquired locks (and possible
161
        waited for them)
162

163
    The function is allowed to change the self.op attribute so that
164
    later methods can no longer worry about missing parameters.
165

166
    """
167
    pass
168

    
169
  def ExpandNames(self):
170
    """Expand names for this LU.
171

172
    This method is called before starting to execute the opcode, and it should
173
    update all the parameters of the opcode to their canonical form (e.g. a
174
    short node name must be fully expanded after this method has successfully
175
    completed). This way locking, hooks, logging, etc. can work correctly.
176

177
    LUs which implement this method must also populate the self.needed_locks
178
    member, as a dict with lock levels as keys, and a list of needed lock names
179
    as values. Rules:
180

181
      - use an empty dict if you don't need any lock
182
      - if you don't need any lock at a particular level omit that level
183
      - don't put anything for the BGL level
184
      - if you want all locks at a level use locking.ALL_SET as a value
185

186
    If you need to share locks (rather than acquire them exclusively) at one
187
    level you can modify self.share_locks, setting a true value (usually 1) for
188
    that level. By default locks are not shared.
189

190
    This function can also define a list of tasklets, which then will be
191
    executed in order instead of the usual LU-level CheckPrereq and Exec
192
    functions, if those are not defined by the LU.
193

194
    Examples::
195

196
      # Acquire all nodes and one instance
197
      self.needed_locks = {
198
        locking.LEVEL_NODE: locking.ALL_SET,
199
        locking.LEVEL_INSTANCE: ['instance1.example.com'],
200
      }
201
      # Acquire just two nodes
202
      self.needed_locks = {
203
        locking.LEVEL_NODE: ['node1.example.com', 'node2.example.com'],
204
      }
205
      # Acquire no locks
206
      self.needed_locks = {} # No, you can't leave it to the default value None
207

208
    """
209
    # The implementation of this method is mandatory only if the new LU is
210
    # concurrent, so that old LUs don't need to be changed all at the same
211
    # time.
212
    if self.REQ_BGL:
213
      self.needed_locks = {} # Exclusive LUs don't need locks.
214
    else:
215
      raise NotImplementedError
216

    
217
  def DeclareLocks(self, level):
218
    """Declare LU locking needs for a level
219

220
    While most LUs can just declare their locking needs at ExpandNames time,
221
    sometimes there's the need to calculate some locks after having acquired
222
    the ones before. This function is called just before acquiring locks at a
223
    particular level, but after acquiring the ones at lower levels, and permits
224
    such calculations. It can be used to modify self.needed_locks, and by
225
    default it does nothing.
226

227
    This function is only called if you have something already set in
228
    self.needed_locks for the level.
229

230
    @param level: Locking level which is going to be locked
231
    @type level: member of ganeti.locking.LEVELS
232

233
    """
234

    
235
  def CheckPrereq(self):
236
    """Check prerequisites for this LU.
237

238
    This method should check that the prerequisites for the execution
239
    of this LU are fulfilled. It can do internode communication, but
240
    it should be idempotent - no cluster or system changes are
241
    allowed.
242

243
    The method should raise errors.OpPrereqError in case something is
244
    not fulfilled. Its return value is ignored.
245

246
    This method should also update all the parameters of the opcode to
247
    their canonical form if it hasn't been done by ExpandNames before.
248

249
    """
250
    if self.tasklets is not None:
251
      for (idx, tl) in enumerate(self.tasklets):
252
        logging.debug("Checking prerequisites for tasklet %s/%s",
253
                      idx + 1, len(self.tasklets))
254
        tl.CheckPrereq()
255
    else:
256
      pass
257

    
258
  def Exec(self, feedback_fn):
259
    """Execute the LU.
260

261
    This method should implement the actual work. It should raise
262
    errors.OpExecError for failures that are somewhat dealt with in
263
    code, or expected.
264

265
    """
266
    if self.tasklets is not None:
267
      for (idx, tl) in enumerate(self.tasklets):
268
        logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets))
269
        tl.Exec(feedback_fn)
270
    else:
271
      raise NotImplementedError
272

    
273
  def BuildHooksEnv(self):
274
    """Build hooks environment for this LU.
275

276
    This method should return a three-node tuple consisting of: a dict
277
    containing the environment that will be used for running the
278
    specific hook for this LU, a list of node names on which the hook
279
    should run before the execution, and a list of node names on which
280
    the hook should run after the execution.
281

282
    The keys of the dict must not have 'GANETI_' prefixed as this will
283
    be handled in the hooks runner. Also note additional keys will be
284
    added by the hooks runner. If the LU doesn't define any
285
    environment, an empty dict (and not None) should be returned.
286

287
    No nodes should be returned as an empty list (and not None).
288

289
    Note that if the HPATH for a LU class is None, this function will
290
    not be called.
291

292
    """
293
    raise NotImplementedError
294

    
295
  def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
296
    """Notify the LU about the results of its hooks.
297

298
    This method is called every time a hooks phase is executed, and notifies
299
    the Logical Unit about the hooks' result. The LU can then use it to alter
300
    its result based on the hooks.  By default the method does nothing and the
301
    previous result is passed back unchanged but any LU can define it if it
302
    wants to use the local cluster hook-scripts somehow.
303

304
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
305
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
306
    @param hook_results: the results of the multi-node hooks rpc call
307
    @param feedback_fn: function used send feedback back to the caller
308
    @param lu_result: the previous Exec result this LU had, or None
309
        in the PRE phase
310
    @return: the new Exec result, based on the previous result
311
        and hook results
312

313
    """
314
    # API must be kept, thus we ignore the unused argument and could
315
    # be a function warnings
316
    # pylint: disable-msg=W0613,R0201
317
    return lu_result
318

    
319
  def _ExpandAndLockInstance(self):
320
    """Helper function to expand and lock an instance.
321

322
    Many LUs that work on an instance take its name in self.op.instance_name
323
    and need to expand it and then declare the expanded name for locking. This
324
    function does it, and then updates self.op.instance_name to the expanded
325
    name. It also initializes needed_locks as a dict, if this hasn't been done
326
    before.
327

328
    """
329
    if self.needed_locks is None:
330
      self.needed_locks = {}
331
    else:
332
      assert locking.LEVEL_INSTANCE not in self.needed_locks, \
333
        "_ExpandAndLockInstance called with instance-level locks set"
334
    self.op.instance_name = _ExpandInstanceName(self.cfg,
335
                                                self.op.instance_name)
336
    self.needed_locks[locking.LEVEL_INSTANCE] = self.op.instance_name
337

    
338
  def _LockInstancesNodes(self, primary_only=False):
339
    """Helper function to declare instances' nodes for locking.
340

341
    This function should be called after locking one or more instances to lock
342
    their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
343
    with all primary or secondary nodes for instances already locked and
344
    present in self.needed_locks[locking.LEVEL_INSTANCE].
345

346
    It should be called from DeclareLocks, and for safety only works if
347
    self.recalculate_locks[locking.LEVEL_NODE] is set.
348

349
    In the future it may grow parameters to just lock some instance's nodes, or
350
    to just lock primaries or secondary nodes, if needed.
351

352
    If should be called in DeclareLocks in a way similar to::
353

354
      if level == locking.LEVEL_NODE:
355
        self._LockInstancesNodes()
356

357
    @type primary_only: boolean
358
    @param primary_only: only lock primary nodes of locked instances
359

360
    """
361
    assert locking.LEVEL_NODE in self.recalculate_locks, \
362
      "_LockInstancesNodes helper function called with no nodes to recalculate"
363

    
364
    # TODO: check if we're really been called with the instance locks held
365

    
366
    # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
367
    # future we might want to have different behaviors depending on the value
368
    # of self.recalculate_locks[locking.LEVEL_NODE]
369
    wanted_nodes = []
370
    for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
371
      instance = self.context.cfg.GetInstanceInfo(instance_name)
372
      wanted_nodes.append(instance.primary_node)
373
      if not primary_only:
374
        wanted_nodes.extend(instance.secondary_nodes)
375

    
376
    if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE:
377
      self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
378
    elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND:
379
      self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
380

    
381
    del self.recalculate_locks[locking.LEVEL_NODE]
382

    
383

    
384
class NoHooksLU(LogicalUnit): # pylint: disable-msg=W0223
385
  """Simple LU which runs no hooks.
386

387
  This LU is intended as a parent for other LogicalUnits which will
388
  run no hooks, in order to reduce duplicate code.
389

390
  """
391
  HPATH = None
392
  HTYPE = None
393

    
394
  def BuildHooksEnv(self):
395
    """Empty BuildHooksEnv for NoHooksLu.
396

397
    This just raises an error.
398

399
    """
400
    assert False, "BuildHooksEnv called for NoHooksLUs"
401

    
402

    
403
class Tasklet:
404
  """Tasklet base class.
405

406
  Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
407
  they can mix legacy code with tasklets. Locking needs to be done in the LU,
408
  tasklets know nothing about locks.
409

410
  Subclasses must follow these rules:
411
    - Implement CheckPrereq
412
    - Implement Exec
413

414
  """
415
  def __init__(self, lu):
416
    self.lu = lu
417

    
418
    # Shortcuts
419
    self.cfg = lu.cfg
420
    self.rpc = lu.rpc
421

    
422
  def CheckPrereq(self):
423
    """Check prerequisites for this tasklets.
424

425
    This method should check whether the prerequisites for the execution of
426
    this tasklet are fulfilled. It can do internode communication, but it
427
    should be idempotent - no cluster or system changes are allowed.
428

429
    The method should raise errors.OpPrereqError in case something is not
430
    fulfilled. Its return value is ignored.
431

432
    This method should also update all parameters to their canonical form if it
433
    hasn't been done before.
434

435
    """
436
    pass
437

    
438
  def Exec(self, feedback_fn):
439
    """Execute the tasklet.
440

441
    This method should implement the actual work. It should raise
442
    errors.OpExecError for failures that are somewhat dealt with in code, or
443
    expected.
444

445
    """
446
    raise NotImplementedError
447

    
448

    
449
class _QueryBase:
450
  """Base for query utility classes.
451

452
  """
453
  #: Attribute holding field definitions
454
  FIELDS = None
455

    
456
  def __init__(self, names, fields, use_locking):
457
    """Initializes this class.
458

459
    """
460
    self.names = names
461
    self.use_locking = use_locking
462

    
463
    self.query = query.Query(self.FIELDS, fields)
464
    self.requested_data = self.query.RequestedData()
465

    
466
    self.do_locking = None
467
    self.wanted = None
468

    
469
  def _GetNames(self, lu, all_names, lock_level):
470
    """Helper function to determine names asked for in the query.
471

472
    """
473
    if self.do_locking:
474
      names = lu.acquired_locks[lock_level]
475
    else:
476
      names = all_names
477

    
478
    if self.wanted == locking.ALL_SET:
479
      assert not self.names
480
      # caller didn't specify names, so ordering is not important
481
      return utils.NiceSort(names)
482

    
483
    # caller specified names and we must keep the same order
484
    assert self.names
485
    assert not self.do_locking or lu.acquired_locks[lock_level]
486

    
487
    missing = set(self.wanted).difference(names)
488
    if missing:
489
      raise errors.OpExecError("Some items were removed before retrieving"
490
                               " their data: %s" % missing)
491

    
492
    # Return expanded names
493
    return self.wanted
494

    
495
  @classmethod
496
  def FieldsQuery(cls, fields):
497
    """Returns list of available fields.
498

499
    @return: List of L{objects.QueryFieldDefinition}
500

501
    """
502
    return query.QueryFields(cls.FIELDS, fields)
503

    
504
  def ExpandNames(self, lu):
505
    """Expand names for this query.
506

507
    See L{LogicalUnit.ExpandNames}.
508

509
    """
510
    raise NotImplementedError()
511

    
512
  def DeclareLocks(self, lu, level):
513
    """Declare locks for this query.
514

515
    See L{LogicalUnit.DeclareLocks}.
516

517
    """
518
    raise NotImplementedError()
519

    
520
  def _GetQueryData(self, lu):
521
    """Collects all data for this query.
522

523
    @return: Query data object
524

525
    """
526
    raise NotImplementedError()
527

    
528
  def NewStyleQuery(self, lu):
529
    """Collect data and execute query.
530

531
    """
532
    return query.GetQueryResponse(self.query, self._GetQueryData(lu))
533

    
534
  def OldStyleQuery(self, lu):
535
    """Collect data and execute query.
536

537
    """
538
    return self.query.OldStyleQuery(self._GetQueryData(lu))
539

    
540

    
541
def _GetWantedNodes(lu, nodes):
542
  """Returns list of checked and expanded node names.
543

544
  @type lu: L{LogicalUnit}
545
  @param lu: the logical unit on whose behalf we execute
546
  @type nodes: list
547
  @param nodes: list of node names or None for all nodes
548
  @rtype: list
549
  @return: the list of nodes, sorted
550
  @raise errors.ProgrammerError: if the nodes parameter is wrong type
551

552
  """
553
  if nodes:
554
    return [_ExpandNodeName(lu.cfg, name) for name in nodes]
555

    
556
  return utils.NiceSort(lu.cfg.GetNodeList())
557

    
558

    
559
def _GetWantedInstances(lu, instances):
560
  """Returns list of checked and expanded instance names.
561

562
  @type lu: L{LogicalUnit}
563
  @param lu: the logical unit on whose behalf we execute
564
  @type instances: list
565
  @param instances: list of instance names or None for all instances
566
  @rtype: list
567
  @return: the list of instances, sorted
568
  @raise errors.OpPrereqError: if the instances parameter is wrong type
569
  @raise errors.OpPrereqError: if any of the passed instances is not found
570

571
  """
572
  if instances:
573
    wanted = [_ExpandInstanceName(lu.cfg, name) for name in instances]
574
  else:
575
    wanted = utils.NiceSort(lu.cfg.GetInstanceList())
576
  return wanted
577

    
578

    
579
def _GetUpdatedParams(old_params, update_dict,
580
                      use_default=True, use_none=False):
581
  """Return the new version of a parameter dictionary.
582

583
  @type old_params: dict
584
  @param old_params: old parameters
585
  @type update_dict: dict
586
  @param update_dict: dict containing new parameter values, or
587
      constants.VALUE_DEFAULT to reset the parameter to its default
588
      value
589
  @param use_default: boolean
590
  @type use_default: whether to recognise L{constants.VALUE_DEFAULT}
591
      values as 'to be deleted' values
592
  @param use_none: boolean
593
  @type use_none: whether to recognise C{None} values as 'to be
594
      deleted' values
595
  @rtype: dict
596
  @return: the new parameter dictionary
597

598
  """
599
  params_copy = copy.deepcopy(old_params)
600
  for key, val in update_dict.iteritems():
601
    if ((use_default and val == constants.VALUE_DEFAULT) or
602
        (use_none and val is None)):
603
      try:
604
        del params_copy[key]
605
      except KeyError:
606
        pass
607
    else:
608
      params_copy[key] = val
609
  return params_copy
610

    
611

    
612
def _CheckOutputFields(static, dynamic, selected):
613
  """Checks whether all selected fields are valid.
614

615
  @type static: L{utils.FieldSet}
616
  @param static: static fields set
617
  @type dynamic: L{utils.FieldSet}
618
  @param dynamic: dynamic fields set
619

620
  """
621
  f = utils.FieldSet()
622
  f.Extend(static)
623
  f.Extend(dynamic)
624

    
625
  delta = f.NonMatching(selected)
626
  if delta:
627
    raise errors.OpPrereqError("Unknown output fields selected: %s"
628
                               % ",".join(delta), errors.ECODE_INVAL)
629

    
630

    
631
def _CheckGlobalHvParams(params):
632
  """Validates that given hypervisor params are not global ones.
633

634
  This will ensure that instances don't get customised versions of
635
  global params.
636

637
  """
638
  used_globals = constants.HVC_GLOBALS.intersection(params)
639
  if used_globals:
640
    msg = ("The following hypervisor parameters are global and cannot"
641
           " be customized at instance level, please modify them at"
642
           " cluster level: %s" % utils.CommaJoin(used_globals))
643
    raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
644

    
645

    
646
def _CheckNodeOnline(lu, node, msg=None):
647
  """Ensure that a given node is online.
648

649
  @param lu: the LU on behalf of which we make the check
650
  @param node: the node to check
651
  @param msg: if passed, should be a message to replace the default one
652
  @raise errors.OpPrereqError: if the node is offline
653

654
  """
655
  if msg is None:
656
    msg = "Can't use offline node"
657
  if lu.cfg.GetNodeInfo(node).offline:
658
    raise errors.OpPrereqError("%s: %s" % (msg, node), errors.ECODE_STATE)
659

    
660

    
661
def _CheckNodeNotDrained(lu, node):
662
  """Ensure that a given node is not drained.
663

664
  @param lu: the LU on behalf of which we make the check
665
  @param node: the node to check
666
  @raise errors.OpPrereqError: if the node is drained
667

668
  """
669
  if lu.cfg.GetNodeInfo(node).drained:
670
    raise errors.OpPrereqError("Can't use drained node %s" % node,
671
                               errors.ECODE_STATE)
672

    
673

    
674
def _CheckNodeVmCapable(lu, node):
675
  """Ensure that a given node is vm capable.
676

677
  @param lu: the LU on behalf of which we make the check
678
  @param node: the node to check
679
  @raise errors.OpPrereqError: if the node is not vm capable
680

681
  """
682
  if not lu.cfg.GetNodeInfo(node).vm_capable:
683
    raise errors.OpPrereqError("Can't use non-vm_capable node %s" % node,
684
                               errors.ECODE_STATE)
685

    
686

    
687
def _CheckNodeHasOS(lu, node, os_name, force_variant):
688
  """Ensure that a node supports a given OS.
689

690
  @param lu: the LU on behalf of which we make the check
691
  @param node: the node to check
692
  @param os_name: the OS to query about
693
  @param force_variant: whether to ignore variant errors
694
  @raise errors.OpPrereqError: if the node is not supporting the OS
695

696
  """
697
  result = lu.rpc.call_os_get(node, os_name)
698
  result.Raise("OS '%s' not in supported OS list for node %s" %
699
               (os_name, node),
700
               prereq=True, ecode=errors.ECODE_INVAL)
701
  if not force_variant:
702
    _CheckOSVariant(result.payload, os_name)
703

    
704

    
705
def _CheckNodeHasSecondaryIP(lu, node, secondary_ip, prereq):
706
  """Ensure that a node has the given secondary ip.
707

708
  @type lu: L{LogicalUnit}
709
  @param lu: the LU on behalf of which we make the check
710
  @type node: string
711
  @param node: the node to check
712
  @type secondary_ip: string
713
  @param secondary_ip: the ip to check
714
  @type prereq: boolean
715
  @param prereq: whether to throw a prerequisite or an execute error
716
  @raise errors.OpPrereqError: if the node doesn't have the ip, and prereq=True
717
  @raise errors.OpExecError: if the node doesn't have the ip, and prereq=False
718

719
  """
720
  result = lu.rpc.call_node_has_ip_address(node, secondary_ip)
721
  result.Raise("Failure checking secondary ip on node %s" % node,
722
               prereq=prereq, ecode=errors.ECODE_ENVIRON)
723
  if not result.payload:
724
    msg = ("Node claims it doesn't have the secondary ip you gave (%s),"
725
           " please fix and re-run this command" % secondary_ip)
726
    if prereq:
727
      raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
728
    else:
729
      raise errors.OpExecError(msg)
730

    
731

    
732
def _GetClusterDomainSecret():
733
  """Reads the cluster domain secret.
734

735
  """
736
  return utils.ReadOneLineFile(constants.CLUSTER_DOMAIN_SECRET_FILE,
737
                               strict=True)
738

    
739

    
740
def _CheckInstanceDown(lu, instance, reason):
741
  """Ensure that an instance is not running."""
742
  if instance.admin_up:
743
    raise errors.OpPrereqError("Instance %s is marked to be up, %s" %
744
                               (instance.name, reason), errors.ECODE_STATE)
745

    
746
  pnode = instance.primary_node
747
  ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
748
  ins_l.Raise("Can't contact node %s for instance information" % pnode,
749
              prereq=True, ecode=errors.ECODE_ENVIRON)
750

    
751
  if instance.name in ins_l.payload:
752
    raise errors.OpPrereqError("Instance %s is running, %s" %
753
                               (instance.name, reason), errors.ECODE_STATE)
754

    
755

    
756
def _ExpandItemName(fn, name, kind):
757
  """Expand an item name.
758

759
  @param fn: the function to use for expansion
760
  @param name: requested item name
761
  @param kind: text description ('Node' or 'Instance')
762
  @return: the resolved (full) name
763
  @raise errors.OpPrereqError: if the item is not found
764

765
  """
766
  full_name = fn(name)
767
  if full_name is None:
768
    raise errors.OpPrereqError("%s '%s' not known" % (kind, name),
769
                               errors.ECODE_NOENT)
770
  return full_name
771

    
772

    
773
def _ExpandNodeName(cfg, name):
774
  """Wrapper over L{_ExpandItemName} for nodes."""
775
  return _ExpandItemName(cfg.ExpandNodeName, name, "Node")
776

    
777

    
778
def _ExpandInstanceName(cfg, name):
779
  """Wrapper over L{_ExpandItemName} for instance."""
780
  return _ExpandItemName(cfg.ExpandInstanceName, name, "Instance")
781

    
782

    
783
def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
784
                          memory, vcpus, nics, disk_template, disks,
785
                          bep, hvp, hypervisor_name):
786
  """Builds instance related env variables for hooks
787

788
  This builds the hook environment from individual variables.
789

790
  @type name: string
791
  @param name: the name of the instance
792
  @type primary_node: string
793
  @param primary_node: the name of the instance's primary node
794
  @type secondary_nodes: list
795
  @param secondary_nodes: list of secondary nodes as strings
796
  @type os_type: string
797
  @param os_type: the name of the instance's OS
798
  @type status: boolean
799
  @param status: the should_run status of the instance
800
  @type memory: string
801
  @param memory: the memory size of the instance
802
  @type vcpus: string
803
  @param vcpus: the count of VCPUs the instance has
804
  @type nics: list
805
  @param nics: list of tuples (ip, mac, mode, link) representing
806
      the NICs the instance has
807
  @type disk_template: string
808
  @param disk_template: the disk template of the instance
809
  @type disks: list
810
  @param disks: the list of (size, mode) pairs
811
  @type bep: dict
812
  @param bep: the backend parameters for the instance
813
  @type hvp: dict
814
  @param hvp: the hypervisor parameters for the instance
815
  @type hypervisor_name: string
816
  @param hypervisor_name: the hypervisor for the instance
817
  @rtype: dict
818
  @return: the hook environment for this instance
819

820
  """
821
  if status:
822
    str_status = "up"
823
  else:
824
    str_status = "down"
825
  env = {
826
    "OP_TARGET": name,
827
    "INSTANCE_NAME": name,
828
    "INSTANCE_PRIMARY": primary_node,
829
    "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
830
    "INSTANCE_OS_TYPE": os_type,
831
    "INSTANCE_STATUS": str_status,
832
    "INSTANCE_MEMORY": memory,
833
    "INSTANCE_VCPUS": vcpus,
834
    "INSTANCE_DISK_TEMPLATE": disk_template,
835
    "INSTANCE_HYPERVISOR": hypervisor_name,
836
  }
837

    
838
  if nics:
839
    nic_count = len(nics)
840
    for idx, (ip, mac, mode, link) in enumerate(nics):
841
      if ip is None:
842
        ip = ""
843
      env["INSTANCE_NIC%d_IP" % idx] = ip
844
      env["INSTANCE_NIC%d_MAC" % idx] = mac
845
      env["INSTANCE_NIC%d_MODE" % idx] = mode
846
      env["INSTANCE_NIC%d_LINK" % idx] = link
847
      if mode == constants.NIC_MODE_BRIDGED:
848
        env["INSTANCE_NIC%d_BRIDGE" % idx] = link
849
  else:
850
    nic_count = 0
851

    
852
  env["INSTANCE_NIC_COUNT"] = nic_count
853

    
854
  if disks:
855
    disk_count = len(disks)
856
    for idx, (size, mode) in enumerate(disks):
857
      env["INSTANCE_DISK%d_SIZE" % idx] = size
858
      env["INSTANCE_DISK%d_MODE" % idx] = mode
859
  else:
860
    disk_count = 0
861

    
862
  env["INSTANCE_DISK_COUNT"] = disk_count
863

    
864
  for source, kind in [(bep, "BE"), (hvp, "HV")]:
865
    for key, value in source.items():
866
      env["INSTANCE_%s_%s" % (kind, key)] = value
867

    
868
  return env
869

    
870

    
871
def _NICListToTuple(lu, nics):
872
  """Build a list of nic information tuples.
873

874
  This list is suitable to be passed to _BuildInstanceHookEnv or as a return
875
  value in LUInstanceQueryData.
876

877
  @type lu:  L{LogicalUnit}
878
  @param lu: the logical unit on whose behalf we execute
879
  @type nics: list of L{objects.NIC}
880
  @param nics: list of nics to convert to hooks tuples
881

882
  """
883
  hooks_nics = []
884
  cluster = lu.cfg.GetClusterInfo()
885
  for nic in nics:
886
    ip = nic.ip
887
    mac = nic.mac
888
    filled_params = cluster.SimpleFillNIC(nic.nicparams)
889
    mode = filled_params[constants.NIC_MODE]
890
    link = filled_params[constants.NIC_LINK]
891
    hooks_nics.append((ip, mac, mode, link))
892
  return hooks_nics
893

    
894

    
895
def _BuildInstanceHookEnvByObject(lu, instance, override=None):
896
  """Builds instance related env variables for hooks from an object.
897

898
  @type lu: L{LogicalUnit}
899
  @param lu: the logical unit on whose behalf we execute
900
  @type instance: L{objects.Instance}
901
  @param instance: the instance for which we should build the
902
      environment
903
  @type override: dict
904
  @param override: dictionary with key/values that will override
905
      our values
906
  @rtype: dict
907
  @return: the hook environment dictionary
908

909
  """
910
  cluster = lu.cfg.GetClusterInfo()
911
  bep = cluster.FillBE(instance)
912
  hvp = cluster.FillHV(instance)
913
  args = {
914
    'name': instance.name,
915
    'primary_node': instance.primary_node,
916
    'secondary_nodes': instance.secondary_nodes,
917
    'os_type': instance.os,
918
    'status': instance.admin_up,
919
    'memory': bep[constants.BE_MEMORY],
920
    'vcpus': bep[constants.BE_VCPUS],
921
    'nics': _NICListToTuple(lu, instance.nics),
922
    'disk_template': instance.disk_template,
923
    'disks': [(disk.size, disk.mode) for disk in instance.disks],
924
    'bep': bep,
925
    'hvp': hvp,
926
    'hypervisor_name': instance.hypervisor,
927
  }
928
  if override:
929
    args.update(override)
930
  return _BuildInstanceHookEnv(**args) # pylint: disable-msg=W0142
931

    
932

    
933
def _AdjustCandidatePool(lu, exceptions):
934
  """Adjust the candidate pool after node operations.
935

936
  """
937
  mod_list = lu.cfg.MaintainCandidatePool(exceptions)
938
  if mod_list:
939
    lu.LogInfo("Promoted nodes to master candidate role: %s",
940
               utils.CommaJoin(node.name for node in mod_list))
941
    for name in mod_list:
942
      lu.context.ReaddNode(name)
943
  mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
944
  if mc_now > mc_max:
945
    lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
946
               (mc_now, mc_max))
947

    
948

    
949
def _DecideSelfPromotion(lu, exceptions=None):
950
  """Decide whether I should promote myself as a master candidate.
951

952
  """
953
  cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
954
  mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
955
  # the new node will increase mc_max with one, so:
956
  mc_should = min(mc_should + 1, cp_size)
957
  return mc_now < mc_should
958

    
959

    
960
def _CheckNicsBridgesExist(lu, target_nics, target_node):
961
  """Check that the brigdes needed by a list of nics exist.
962

963
  """
964
  cluster = lu.cfg.GetClusterInfo()
965
  paramslist = [cluster.SimpleFillNIC(nic.nicparams) for nic in target_nics]
966
  brlist = [params[constants.NIC_LINK] for params in paramslist
967
            if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
968
  if brlist:
969
    result = lu.rpc.call_bridges_exist(target_node, brlist)
970
    result.Raise("Error checking bridges on destination node '%s'" %
971
                 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
972

    
973

    
974
def _CheckInstanceBridgesExist(lu, instance, node=None):
975
  """Check that the brigdes needed by an instance exist.
976

977
  """
978
  if node is None:
979
    node = instance.primary_node
980
  _CheckNicsBridgesExist(lu, instance.nics, node)
981

    
982

    
983
def _CheckOSVariant(os_obj, name):
984
  """Check whether an OS name conforms to the os variants specification.
985

986
  @type os_obj: L{objects.OS}
987
  @param os_obj: OS object to check
988
  @type name: string
989
  @param name: OS name passed by the user, to check for validity
990

991
  """
992
  if not os_obj.supported_variants:
993
    return
994
  variant = objects.OS.GetVariant(name)
995
  if not variant:
996
    raise errors.OpPrereqError("OS name must include a variant",
997
                               errors.ECODE_INVAL)
998

    
999
  if variant not in os_obj.supported_variants:
1000
    raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
1001

    
1002

    
1003
def _GetNodeInstancesInner(cfg, fn):
1004
  return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
1005

    
1006

    
1007
def _GetNodeInstances(cfg, node_name):
1008
  """Returns a list of all primary and secondary instances on a node.
1009

1010
  """
1011

    
1012
  return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
1013

    
1014

    
1015
def _GetNodePrimaryInstances(cfg, node_name):
1016
  """Returns primary instances on a node.
1017

1018
  """
1019
  return _GetNodeInstancesInner(cfg,
1020
                                lambda inst: node_name == inst.primary_node)
1021

    
1022

    
1023
def _GetNodeSecondaryInstances(cfg, node_name):
1024
  """Returns secondary instances on a node.
1025

1026
  """
1027
  return _GetNodeInstancesInner(cfg,
1028
                                lambda inst: node_name in inst.secondary_nodes)
1029

    
1030

    
1031
def _GetStorageTypeArgs(cfg, storage_type):
1032
  """Returns the arguments for a storage type.
1033

1034
  """
1035
  # Special case for file storage
1036
  if storage_type == constants.ST_FILE:
1037
    # storage.FileStorage wants a list of storage directories
1038
    return [[cfg.GetFileStorageDir()]]
1039

    
1040
  return []
1041

    
1042

    
1043
def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq):
1044
  faulty = []
1045

    
1046
  for dev in instance.disks:
1047
    cfg.SetDiskID(dev, node_name)
1048

    
1049
  result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks)
1050
  result.Raise("Failed to get disk status from node %s" % node_name,
1051
               prereq=prereq, ecode=errors.ECODE_ENVIRON)
1052

    
1053
  for idx, bdev_status in enumerate(result.payload):
1054
    if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
1055
      faulty.append(idx)
1056

    
1057
  return faulty
1058

    
1059

    
1060
def _CheckIAllocatorOrNode(lu, iallocator_slot, node_slot):
1061
  """Check the sanity of iallocator and node arguments and use the
1062
  cluster-wide iallocator if appropriate.
1063

1064
  Check that at most one of (iallocator, node) is specified. If none is
1065
  specified, then the LU's opcode's iallocator slot is filled with the
1066
  cluster-wide default iallocator.
1067

1068
  @type iallocator_slot: string
1069
  @param iallocator_slot: the name of the opcode iallocator slot
1070
  @type node_slot: string
1071
  @param node_slot: the name of the opcode target node slot
1072

1073
  """
1074
  node = getattr(lu.op, node_slot, None)
1075
  iallocator = getattr(lu.op, iallocator_slot, None)
1076

    
1077
  if node is not None and iallocator is not None:
1078
    raise errors.OpPrereqError("Do not specify both, iallocator and node.",
1079
                               errors.ECODE_INVAL)
1080
  elif node is None and iallocator is None:
1081
    default_iallocator = lu.cfg.GetDefaultIAllocator()
1082
    if default_iallocator:
1083
      setattr(lu.op, iallocator_slot, default_iallocator)
1084
    else:
1085
      raise errors.OpPrereqError("No iallocator or node given and no"
1086
                                 " cluster-wide default iallocator found."
1087
                                 " Please specify either an iallocator or a"
1088
                                 " node, or set a cluster-wide default"
1089
                                 " iallocator.")
1090

    
1091

    
1092
class LUClusterPostInit(LogicalUnit):
1093
  """Logical unit for running hooks after cluster initialization.
1094

1095
  """
1096
  HPATH = "cluster-init"
1097
  HTYPE = constants.HTYPE_CLUSTER
1098

    
1099
  def BuildHooksEnv(self):
1100
    """Build hooks env.
1101

1102
    """
1103
    env = {"OP_TARGET": self.cfg.GetClusterName()}
1104
    mn = self.cfg.GetMasterNode()
1105
    return env, [], [mn]
1106

    
1107
  def Exec(self, feedback_fn):
1108
    """Nothing to do.
1109

1110
    """
1111
    return True
1112

    
1113

    
1114
class LUClusterDestroy(LogicalUnit):
1115
  """Logical unit for destroying the cluster.
1116

1117
  """
1118
  HPATH = "cluster-destroy"
1119
  HTYPE = constants.HTYPE_CLUSTER
1120

    
1121
  def BuildHooksEnv(self):
1122
    """Build hooks env.
1123

1124
    """
1125
    env = {"OP_TARGET": self.cfg.GetClusterName()}
1126
    return env, [], []
1127

    
1128
  def CheckPrereq(self):
1129
    """Check prerequisites.
1130

1131
    This checks whether the cluster is empty.
1132

1133
    Any errors are signaled by raising errors.OpPrereqError.
1134

1135
    """
1136
    master = self.cfg.GetMasterNode()
1137

    
1138
    nodelist = self.cfg.GetNodeList()
1139
    if len(nodelist) != 1 or nodelist[0] != master:
1140
      raise errors.OpPrereqError("There are still %d node(s) in"
1141
                                 " this cluster." % (len(nodelist) - 1),
1142
                                 errors.ECODE_INVAL)
1143
    instancelist = self.cfg.GetInstanceList()
1144
    if instancelist:
1145
      raise errors.OpPrereqError("There are still %d instance(s) in"
1146
                                 " this cluster." % len(instancelist),
1147
                                 errors.ECODE_INVAL)
1148

    
1149
  def Exec(self, feedback_fn):
1150
    """Destroys the cluster.
1151

1152
    """
1153
    master = self.cfg.GetMasterNode()
1154

    
1155
    # Run post hooks on master node before it's removed
1156
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
1157
    try:
1158
      hm.RunPhase(constants.HOOKS_PHASE_POST, [master])
1159
    except:
1160
      # pylint: disable-msg=W0702
1161
      self.LogWarning("Errors occurred running hooks on %s" % master)
1162

    
1163
    result = self.rpc.call_node_stop_master(master, False)
1164
    result.Raise("Could not disable the master role")
1165

    
1166
    return master
1167

    
1168

    
1169
def _VerifyCertificate(filename):
1170
  """Verifies a certificate for LUClusterVerify.
1171

1172
  @type filename: string
1173
  @param filename: Path to PEM file
1174

1175
  """
1176
  try:
1177
    cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
1178
                                           utils.ReadFile(filename))
1179
  except Exception, err: # pylint: disable-msg=W0703
1180
    return (LUClusterVerify.ETYPE_ERROR,
1181
            "Failed to load X509 certificate %s: %s" % (filename, err))
1182

    
1183
  (errcode, msg) = \
1184
    utils.VerifyX509Certificate(cert, constants.SSL_CERT_EXPIRATION_WARN,
1185
                                constants.SSL_CERT_EXPIRATION_ERROR)
1186

    
1187
  if msg:
1188
    fnamemsg = "While verifying %s: %s" % (filename, msg)
1189
  else:
1190
    fnamemsg = None
1191

    
1192
  if errcode is None:
1193
    return (None, fnamemsg)
1194
  elif errcode == utils.CERT_WARNING:
1195
    return (LUClusterVerify.ETYPE_WARNING, fnamemsg)
1196
  elif errcode == utils.CERT_ERROR:
1197
    return (LUClusterVerify.ETYPE_ERROR, fnamemsg)
1198

    
1199
  raise errors.ProgrammerError("Unhandled certificate error code %r" % errcode)
1200

    
1201

    
1202
class LUClusterVerify(LogicalUnit):
1203
  """Verifies the cluster status.
1204

1205
  """
1206
  HPATH = "cluster-verify"
1207
  HTYPE = constants.HTYPE_CLUSTER
1208
  REQ_BGL = False
1209

    
1210
  TCLUSTER = "cluster"
1211
  TNODE = "node"
1212
  TINSTANCE = "instance"
1213

    
1214
  ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
1215
  ECLUSTERCERT = (TCLUSTER, "ECLUSTERCERT")
1216
  EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
1217
  EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
1218
  EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
1219
  EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
1220
  EINSTANCEFAULTYDISK = (TINSTANCE, "EINSTANCEFAULTYDISK")
1221
  EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE")
1222
  EINSTANCESPLITGROUPS = (TINSTANCE, "EINSTANCESPLITGROUPS")
1223
  ENODEDRBD = (TNODE, "ENODEDRBD")
1224
  ENODEDRBDHELPER = (TNODE, "ENODEDRBDHELPER")
1225
  ENODEFILECHECK = (TNODE, "ENODEFILECHECK")
1226
  ENODEHOOKS = (TNODE, "ENODEHOOKS")
1227
  ENODEHV = (TNODE, "ENODEHV")
1228
  ENODELVM = (TNODE, "ENODELVM")
1229
  ENODEN1 = (TNODE, "ENODEN1")
1230
  ENODENET = (TNODE, "ENODENET")
1231
  ENODEOS = (TNODE, "ENODEOS")
1232
  ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE")
1233
  ENODEORPHANLV = (TNODE, "ENODEORPHANLV")
1234
  ENODERPC = (TNODE, "ENODERPC")
1235
  ENODESSH = (TNODE, "ENODESSH")
1236
  ENODEVERSION = (TNODE, "ENODEVERSION")
1237
  ENODESETUP = (TNODE, "ENODESETUP")
1238
  ENODETIME = (TNODE, "ENODETIME")
1239
  ENODEOOBPATH = (TNODE, "ENODEOOBPATH")
1240

    
1241
  ETYPE_FIELD = "code"
1242
  ETYPE_ERROR = "ERROR"
1243
  ETYPE_WARNING = "WARNING"
1244

    
1245
  _HOOKS_INDENT_RE = re.compile("^", re.M)
1246

    
1247
  class NodeImage(object):
1248
    """A class representing the logical and physical status of a node.
1249

1250
    @type name: string
1251
    @ivar name: the node name to which this object refers
1252
    @ivar volumes: a structure as returned from
1253
        L{ganeti.backend.GetVolumeList} (runtime)
1254
    @ivar instances: a list of running instances (runtime)
1255
    @ivar pinst: list of configured primary instances (config)
1256
    @ivar sinst: list of configured secondary instances (config)
1257
    @ivar sbp: diction of {secondary-node: list of instances} of all peers
1258
        of this node (config)
1259
    @ivar mfree: free memory, as reported by hypervisor (runtime)
1260
    @ivar dfree: free disk, as reported by the node (runtime)
1261
    @ivar offline: the offline status (config)
1262
    @type rpc_fail: boolean
1263
    @ivar rpc_fail: whether the RPC verify call was successfull (overall,
1264
        not whether the individual keys were correct) (runtime)
1265
    @type lvm_fail: boolean
1266
    @ivar lvm_fail: whether the RPC call didn't return valid LVM data
1267
    @type hyp_fail: boolean
1268
    @ivar hyp_fail: whether the RPC call didn't return the instance list
1269
    @type ghost: boolean
1270
    @ivar ghost: whether this is a known node or not (config)
1271
    @type os_fail: boolean
1272
    @ivar os_fail: whether the RPC call didn't return valid OS data
1273
    @type oslist: list
1274
    @ivar oslist: list of OSes as diagnosed by DiagnoseOS
1275
    @type vm_capable: boolean
1276
    @ivar vm_capable: whether the node can host instances
1277

1278
    """
1279
    def __init__(self, offline=False, name=None, vm_capable=True):
1280
      self.name = name
1281
      self.volumes = {}
1282
      self.instances = []
1283
      self.pinst = []
1284
      self.sinst = []
1285
      self.sbp = {}
1286
      self.mfree = 0
1287
      self.dfree = 0
1288
      self.offline = offline
1289
      self.vm_capable = vm_capable
1290
      self.rpc_fail = False
1291
      self.lvm_fail = False
1292
      self.hyp_fail = False
1293
      self.ghost = False
1294
      self.os_fail = False
1295
      self.oslist = {}
1296

    
1297
  def ExpandNames(self):
1298
    self.needed_locks = {
1299
      locking.LEVEL_NODE: locking.ALL_SET,
1300
      locking.LEVEL_INSTANCE: locking.ALL_SET,
1301
    }
1302
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
1303

    
1304
  def _Error(self, ecode, item, msg, *args, **kwargs):
1305
    """Format an error message.
1306

1307
    Based on the opcode's error_codes parameter, either format a
1308
    parseable error code, or a simpler error string.
1309

1310
    This must be called only from Exec and functions called from Exec.
1311

1312
    """
1313
    ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
1314
    itype, etxt = ecode
1315
    # first complete the msg
1316
    if args:
1317
      msg = msg % args
1318
    # then format the whole message
1319
    if self.op.error_codes:
1320
      msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
1321
    else:
1322
      if item:
1323
        item = " " + item
1324
      else:
1325
        item = ""
1326
      msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
1327
    # and finally report it via the feedback_fn
1328
    self._feedback_fn("  - %s" % msg)
1329

    
1330
  def _ErrorIf(self, cond, *args, **kwargs):
1331
    """Log an error message if the passed condition is True.
1332

1333
    """
1334
    cond = bool(cond) or self.op.debug_simulate_errors
1335
    if cond:
1336
      self._Error(*args, **kwargs)
1337
    # do not mark the operation as failed for WARN cases only
1338
    if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
1339
      self.bad = self.bad or cond
1340

    
1341
  def _VerifyNode(self, ninfo, nresult):
1342
    """Perform some basic validation on data returned from a node.
1343

1344
      - check the result data structure is well formed and has all the
1345
        mandatory fields
1346
      - check ganeti version
1347

1348
    @type ninfo: L{objects.Node}
1349
    @param ninfo: the node to check
1350
    @param nresult: the results from the node
1351
    @rtype: boolean
1352
    @return: whether overall this call was successful (and we can expect
1353
         reasonable values in the respose)
1354

1355
    """
1356
    node = ninfo.name
1357
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1358

    
1359
    # main result, nresult should be a non-empty dict
1360
    test = not nresult or not isinstance(nresult, dict)
1361
    _ErrorIf(test, self.ENODERPC, node,
1362
                  "unable to verify node: no data returned")
1363
    if test:
1364
      return False
1365

    
1366
    # compares ganeti version
1367
    local_version = constants.PROTOCOL_VERSION
1368
    remote_version = nresult.get("version", None)
1369
    test = not (remote_version and
1370
                isinstance(remote_version, (list, tuple)) and
1371
                len(remote_version) == 2)
1372
    _ErrorIf(test, self.ENODERPC, node,
1373
             "connection to node returned invalid data")
1374
    if test:
1375
      return False
1376

    
1377
    test = local_version != remote_version[0]
1378
    _ErrorIf(test, self.ENODEVERSION, node,
1379
             "incompatible protocol versions: master %s,"
1380
             " node %s", local_version, remote_version[0])
1381
    if test:
1382
      return False
1383

    
1384
    # node seems compatible, we can actually try to look into its results
1385

    
1386
    # full package version
1387
    self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
1388
                  self.ENODEVERSION, node,
1389
                  "software version mismatch: master %s, node %s",
1390
                  constants.RELEASE_VERSION, remote_version[1],
1391
                  code=self.ETYPE_WARNING)
1392

    
1393
    hyp_result = nresult.get(constants.NV_HYPERVISOR, None)
1394
    if ninfo.vm_capable and isinstance(hyp_result, dict):
1395
      for hv_name, hv_result in hyp_result.iteritems():
1396
        test = hv_result is not None
1397
        _ErrorIf(test, self.ENODEHV, node,
1398
                 "hypervisor %s verify failure: '%s'", hv_name, hv_result)
1399

    
1400
    hvp_result = nresult.get(constants.NV_HVPARAMS, None)
1401
    if ninfo.vm_capable and isinstance(hvp_result, list):
1402
      for item, hv_name, hv_result in hvp_result:
1403
        _ErrorIf(True, self.ENODEHV, node,
1404
                 "hypervisor %s parameter verify failure (source %s): %s",
1405
                 hv_name, item, hv_result)
1406

    
1407
    test = nresult.get(constants.NV_NODESETUP,
1408
                           ["Missing NODESETUP results"])
1409
    _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s",
1410
             "; ".join(test))
1411

    
1412
    return True
1413

    
1414
  def _VerifyNodeTime(self, ninfo, nresult,
1415
                      nvinfo_starttime, nvinfo_endtime):
1416
    """Check the node time.
1417

1418
    @type ninfo: L{objects.Node}
1419
    @param ninfo: the node to check
1420
    @param nresult: the remote results for the node
1421
    @param nvinfo_starttime: the start time of the RPC call
1422
    @param nvinfo_endtime: the end time of the RPC call
1423

1424
    """
1425
    node = ninfo.name
1426
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1427

    
1428
    ntime = nresult.get(constants.NV_TIME, None)
1429
    try:
1430
      ntime_merged = utils.MergeTime(ntime)
1431
    except (ValueError, TypeError):
1432
      _ErrorIf(True, self.ENODETIME, node, "Node returned invalid time")
1433
      return
1434

    
1435
    if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
1436
      ntime_diff = "%.01fs" % abs(nvinfo_starttime - ntime_merged)
1437
    elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
1438
      ntime_diff = "%.01fs" % abs(ntime_merged - nvinfo_endtime)
1439
    else:
1440
      ntime_diff = None
1441

    
1442
    _ErrorIf(ntime_diff is not None, self.ENODETIME, node,
1443
             "Node time diverges by at least %s from master node time",
1444
             ntime_diff)
1445

    
1446
  def _VerifyNodeLVM(self, ninfo, nresult, vg_name):
1447
    """Check the node time.
1448

1449
    @type ninfo: L{objects.Node}
1450
    @param ninfo: the node to check
1451
    @param nresult: the remote results for the node
1452
    @param vg_name: the configured VG name
1453

1454
    """
1455
    if vg_name is None:
1456
      return
1457

    
1458
    node = ninfo.name
1459
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1460

    
1461
    # checks vg existence and size > 20G
1462
    vglist = nresult.get(constants.NV_VGLIST, None)
1463
    test = not vglist
1464
    _ErrorIf(test, self.ENODELVM, node, "unable to check volume groups")
1465
    if not test:
1466
      vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
1467
                                            constants.MIN_VG_SIZE)
1468
      _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
1469

    
1470
    # check pv names
1471
    pvlist = nresult.get(constants.NV_PVLIST, None)
1472
    test = pvlist is None
1473
    _ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node")
1474
    if not test:
1475
      # check that ':' is not present in PV names, since it's a
1476
      # special character for lvcreate (denotes the range of PEs to
1477
      # use on the PV)
1478
      for _, pvname, owner_vg in pvlist:
1479
        test = ":" in pvname
1480
        _ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV"
1481
                 " '%s' of VG '%s'", pvname, owner_vg)
1482

    
1483
  def _VerifyNodeNetwork(self, ninfo, nresult):
1484
    """Check the node time.
1485

1486
    @type ninfo: L{objects.Node}
1487
    @param ninfo: the node to check
1488
    @param nresult: the remote results for the node
1489

1490
    """
1491
    node = ninfo.name
1492
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1493

    
1494
    test = constants.NV_NODELIST not in nresult
1495
    _ErrorIf(test, self.ENODESSH, node,
1496
             "node hasn't returned node ssh connectivity data")
1497
    if not test:
1498
      if nresult[constants.NV_NODELIST]:
1499
        for a_node, a_msg in nresult[constants.NV_NODELIST].items():
1500
          _ErrorIf(True, self.ENODESSH, node,
1501
                   "ssh communication with node '%s': %s", a_node, a_msg)
1502

    
1503
    test = constants.NV_NODENETTEST not in nresult
1504
    _ErrorIf(test, self.ENODENET, node,
1505
             "node hasn't returned node tcp connectivity data")
1506
    if not test:
1507
      if nresult[constants.NV_NODENETTEST]:
1508
        nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys())
1509
        for anode in nlist:
1510
          _ErrorIf(True, self.ENODENET, node,
1511
                   "tcp communication with node '%s': %s",
1512
                   anode, nresult[constants.NV_NODENETTEST][anode])
1513

    
1514
    test = constants.NV_MASTERIP not in nresult
1515
    _ErrorIf(test, self.ENODENET, node,
1516
             "node hasn't returned node master IP reachability data")
1517
    if not test:
1518
      if not nresult[constants.NV_MASTERIP]:
1519
        if node == self.master_node:
1520
          msg = "the master node cannot reach the master IP (not configured?)"
1521
        else:
1522
          msg = "cannot reach the master IP"
1523
        _ErrorIf(True, self.ENODENET, node, msg)
1524

    
1525
  def _VerifyInstance(self, instance, instanceconfig, node_image,
1526
                      diskstatus):
1527
    """Verify an instance.
1528

1529
    This function checks to see if the required block devices are
1530
    available on the instance's node.
1531

1532
    """
1533
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1534
    node_current = instanceconfig.primary_node
1535

    
1536
    node_vol_should = {}
1537
    instanceconfig.MapLVsByNode(node_vol_should)
1538

    
1539
    for node in node_vol_should:
1540
      n_img = node_image[node]
1541
      if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1542
        # ignore missing volumes on offline or broken nodes
1543
        continue
1544
      for volume in node_vol_should[node]:
1545
        test = volume not in n_img.volumes
1546
        _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
1547
                 "volume %s missing on node %s", volume, node)
1548

    
1549
    if instanceconfig.admin_up:
1550
      pri_img = node_image[node_current]
1551
      test = instance not in pri_img.instances and not pri_img.offline
1552
      _ErrorIf(test, self.EINSTANCEDOWN, instance,
1553
               "instance not running on its primary node %s",
1554
               node_current)
1555

    
1556
    for node, n_img in node_image.items():
1557
      if node != node_current:
1558
        test = instance in n_img.instances
1559
        _ErrorIf(test, self.EINSTANCEWRONGNODE, instance,
1560
                 "instance should not run on node %s", node)
1561

    
1562
    diskdata = [(nname, success, status, idx)
1563
                for (nname, disks) in diskstatus.items()
1564
                for idx, (success, status) in enumerate(disks)]
1565

    
1566
    for nname, success, bdev_status, idx in diskdata:
1567
      # the 'ghost node' construction in Exec() ensures that we have a
1568
      # node here
1569
      snode = node_image[nname]
1570
      bad_snode = snode.ghost or snode.offline
1571
      _ErrorIf(instanceconfig.admin_up and not success and not bad_snode,
1572
               self.EINSTANCEFAULTYDISK, instance,
1573
               "couldn't retrieve status for disk/%s on %s: %s",
1574
               idx, nname, bdev_status)
1575
      _ErrorIf((instanceconfig.admin_up and success and
1576
                bdev_status.ldisk_status == constants.LDS_FAULTY),
1577
               self.EINSTANCEFAULTYDISK, instance,
1578
               "disk/%s on %s is faulty", idx, nname)
1579

    
1580
  def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):
1581
    """Verify if there are any unknown volumes in the cluster.
1582

1583
    The .os, .swap and backup volumes are ignored. All other volumes are
1584
    reported as unknown.
1585

1586
    @type reserved: L{ganeti.utils.FieldSet}
1587
    @param reserved: a FieldSet of reserved volume names
1588

1589
    """
1590
    for node, n_img in node_image.items():
1591
      if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1592
        # skip non-healthy nodes
1593
        continue
1594
      for volume in n_img.volumes:
1595
        test = ((node not in node_vol_should or
1596
                volume not in node_vol_should[node]) and
1597
                not reserved.Matches(volume))
1598
        self._ErrorIf(test, self.ENODEORPHANLV, node,
1599
                      "volume %s is unknown", volume)
1600

    
1601
  def _VerifyOrphanInstances(self, instancelist, node_image):
1602
    """Verify the list of running instances.
1603

1604
    This checks what instances are running but unknown to the cluster.
1605

1606
    """
1607
    for node, n_img in node_image.items():
1608
      for o_inst in n_img.instances:
1609
        test = o_inst not in instancelist
1610
        self._ErrorIf(test, self.ENODEORPHANINSTANCE, node,
1611
                      "instance %s on node %s should not exist", o_inst, node)
1612

    
1613
  def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
1614
    """Verify N+1 Memory Resilience.
1615

1616
    Check that if one single node dies we can still start all the
1617
    instances it was primary for.
1618

1619
    """
1620
    for node, n_img in node_image.items():
1621
      # This code checks that every node which is now listed as
1622
      # secondary has enough memory to host all instances it is
1623
      # supposed to should a single other node in the cluster fail.
1624
      # FIXME: not ready for failover to an arbitrary node
1625
      # FIXME: does not support file-backed instances
1626
      # WARNING: we currently take into account down instances as well
1627
      # as up ones, considering that even if they're down someone
1628
      # might want to start them even in the event of a node failure.
1629
      if n_img.offline:
1630
        # we're skipping offline nodes from the N+1 warning, since
1631
        # most likely we don't have good memory infromation from them;
1632
        # we already list instances living on such nodes, and that's
1633
        # enough warning
1634
        continue
1635
      for prinode, instances in n_img.sbp.items():
1636
        needed_mem = 0
1637
        for instance in instances:
1638
          bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
1639
          if bep[constants.BE_AUTO_BALANCE]:
1640
            needed_mem += bep[constants.BE_MEMORY]
1641
        test = n_img.mfree < needed_mem
1642
        self._ErrorIf(test, self.ENODEN1, node,
1643
                      "not enough memory to accomodate instance failovers"
1644
                      " should node %s fail (%dMiB needed, %dMiB available)",
1645
                      prinode, needed_mem, n_img.mfree)
1646

    
1647
  def _VerifyNodeFiles(self, ninfo, nresult, file_list, local_cksum,
1648
                       master_files):
1649
    """Verifies and computes the node required file checksums.
1650

1651
    @type ninfo: L{objects.Node}
1652
    @param ninfo: the node to check
1653
    @param nresult: the remote results for the node
1654
    @param file_list: required list of files
1655
    @param local_cksum: dictionary of local files and their checksums
1656
    @param master_files: list of files that only masters should have
1657

1658
    """
1659
    node = ninfo.name
1660
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1661

    
1662
    remote_cksum = nresult.get(constants.NV_FILELIST, None)
1663
    test = not isinstance(remote_cksum, dict)
1664
    _ErrorIf(test, self.ENODEFILECHECK, node,
1665
             "node hasn't returned file checksum data")
1666
    if test:
1667
      return
1668

    
1669
    for file_name in file_list:
1670
      node_is_mc = ninfo.master_candidate
1671
      must_have = (file_name not in master_files) or node_is_mc
1672
      # missing
1673
      test1 = file_name not in remote_cksum
1674
      # invalid checksum
1675
      test2 = not test1 and remote_cksum[file_name] != local_cksum[file_name]
1676
      # existing and good
1677
      test3 = not test1 and remote_cksum[file_name] == local_cksum[file_name]
1678
      _ErrorIf(test1 and must_have, self.ENODEFILECHECK, node,
1679
               "file '%s' missing", file_name)
1680
      _ErrorIf(test2 and must_have, self.ENODEFILECHECK, node,
1681
               "file '%s' has wrong checksum", file_name)
1682
      # not candidate and this is not a must-have file
1683
      _ErrorIf(test2 and not must_have, self.ENODEFILECHECK, node,
1684
               "file '%s' should not exist on non master"
1685
               " candidates (and the file is outdated)", file_name)
1686
      # all good, except non-master/non-must have combination
1687
      _ErrorIf(test3 and not must_have, self.ENODEFILECHECK, node,
1688
               "file '%s' should not exist"
1689
               " on non master candidates", file_name)
1690

    
1691
  def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper,
1692
                      drbd_map):
1693
    """Verifies and the node DRBD status.
1694

1695
    @type ninfo: L{objects.Node}
1696
    @param ninfo: the node to check
1697
    @param nresult: the remote results for the node
1698
    @param instanceinfo: the dict of instances
1699
    @param drbd_helper: the configured DRBD usermode helper
1700
    @param drbd_map: the DRBD map as returned by
1701
        L{ganeti.config.ConfigWriter.ComputeDRBDMap}
1702

1703
    """
1704
    node = ninfo.name
1705
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1706

    
1707
    if drbd_helper:
1708
      helper_result = nresult.get(constants.NV_DRBDHELPER, None)
1709
      test = (helper_result == None)
1710
      _ErrorIf(test, self.ENODEDRBDHELPER, node,
1711
               "no drbd usermode helper returned")
1712
      if helper_result:
1713
        status, payload = helper_result
1714
        test = not status
1715
        _ErrorIf(test, self.ENODEDRBDHELPER, node,
1716
                 "drbd usermode helper check unsuccessful: %s", payload)
1717
        test = status and (payload != drbd_helper)
1718
        _ErrorIf(test, self.ENODEDRBDHELPER, node,
1719
                 "wrong drbd usermode helper: %s", payload)
1720

    
1721
    # compute the DRBD minors
1722
    node_drbd = {}
1723
    for minor, instance in drbd_map[node].items():
1724
      test = instance not in instanceinfo
1725
      _ErrorIf(test, self.ECLUSTERCFG, None,
1726
               "ghost instance '%s' in temporary DRBD map", instance)
1727
        # ghost instance should not be running, but otherwise we
1728
        # don't give double warnings (both ghost instance and
1729
        # unallocated minor in use)
1730
      if test:
1731
        node_drbd[minor] = (instance, False)
1732
      else:
1733
        instance = instanceinfo[instance]
1734
        node_drbd[minor] = (instance.name, instance.admin_up)
1735

    
1736
    # and now check them
1737
    used_minors = nresult.get(constants.NV_DRBDLIST, [])
1738
    test = not isinstance(used_minors, (tuple, list))
1739
    _ErrorIf(test, self.ENODEDRBD, node,
1740
             "cannot parse drbd status file: %s", str(used_minors))
1741
    if test:
1742
      # we cannot check drbd status
1743
      return
1744

    
1745
    for minor, (iname, must_exist) in node_drbd.items():
1746
      test = minor not in used_minors and must_exist
1747
      _ErrorIf(test, self.ENODEDRBD, node,
1748
               "drbd minor %d of instance %s is not active", minor, iname)
1749
    for minor in used_minors:
1750
      test = minor not in node_drbd
1751
      _ErrorIf(test, self.ENODEDRBD, node,
1752
               "unallocated drbd minor %d is in use", minor)
1753

    
1754
  def _UpdateNodeOS(self, ninfo, nresult, nimg):
1755
    """Builds the node OS structures.
1756

1757
    @type ninfo: L{objects.Node}
1758
    @param ninfo: the node to check
1759
    @param nresult: the remote results for the node
1760
    @param nimg: the node image object
1761

1762
    """
1763
    node = ninfo.name
1764
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1765

    
1766
    remote_os = nresult.get(constants.NV_OSLIST, None)
1767
    test = (not isinstance(remote_os, list) or
1768
            not compat.all(isinstance(v, list) and len(v) == 7
1769
                           for v in remote_os))
1770

    
1771
    _ErrorIf(test, self.ENODEOS, node,
1772
             "node hasn't returned valid OS data")
1773

    
1774
    nimg.os_fail = test
1775

    
1776
    if test:
1777
      return
1778

    
1779
    os_dict = {}
1780

    
1781
    for (name, os_path, status, diagnose,
1782
         variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
1783

    
1784
      if name not in os_dict:
1785
        os_dict[name] = []
1786

    
1787
      # parameters is a list of lists instead of list of tuples due to
1788
      # JSON lacking a real tuple type, fix it:
1789
      parameters = [tuple(v) for v in parameters]
1790
      os_dict[name].append((os_path, status, diagnose,
1791
                            set(variants), set(parameters), set(api_ver)))
1792

    
1793
    nimg.oslist = os_dict
1794

    
1795
  def _VerifyNodeOS(self, ninfo, nimg, base):
1796
    """Verifies the node OS list.
1797

1798
    @type ninfo: L{objects.Node}
1799
    @param ninfo: the node to check
1800
    @param nimg: the node image object
1801
    @param base: the 'template' node we match against (e.g. from the master)
1802

1803
    """
1804
    node = ninfo.name
1805
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1806

    
1807
    assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
1808

    
1809
    for os_name, os_data in nimg.oslist.items():
1810
      assert os_data, "Empty OS status for OS %s?!" % os_name
1811
      f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
1812
      _ErrorIf(not f_status, self.ENODEOS, node,
1813
               "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
1814
      _ErrorIf(len(os_data) > 1, self.ENODEOS, node,
1815
               "OS '%s' has multiple entries (first one shadows the rest): %s",
1816
               os_name, utils.CommaJoin([v[0] for v in os_data]))
1817
      # this will catched in backend too
1818
      _ErrorIf(compat.any(v >= constants.OS_API_V15 for v in f_api)
1819
               and not f_var, self.ENODEOS, node,
1820
               "OS %s with API at least %d does not declare any variant",
1821
               os_name, constants.OS_API_V15)
1822
      # comparisons with the 'base' image
1823
      test = os_name not in base.oslist
1824
      _ErrorIf(test, self.ENODEOS, node,
1825
               "Extra OS %s not present on reference node (%s)",
1826
               os_name, base.name)
1827
      if test:
1828
        continue
1829
      assert base.oslist[os_name], "Base node has empty OS status?"
1830
      _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
1831
      if not b_status:
1832
        # base OS is invalid, skipping
1833
        continue
1834
      for kind, a, b in [("API version", f_api, b_api),
1835
                         ("variants list", f_var, b_var),
1836
                         ("parameters", f_param, b_param)]:
1837
        _ErrorIf(a != b, self.ENODEOS, node,
1838
                 "OS %s %s differs from reference node %s: %s vs. %s",
1839
                 kind, os_name, base.name,
1840
                 utils.CommaJoin(a), utils.CommaJoin(b))
1841

    
1842
    # check any missing OSes
1843
    missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
1844
    _ErrorIf(missing, self.ENODEOS, node,
1845
             "OSes present on reference node %s but missing on this node: %s",
1846
             base.name, utils.CommaJoin(missing))
1847

    
1848
  def _VerifyOob(self, ninfo, nresult):
1849
    """Verifies out of band functionality of a node.
1850

1851
    @type ninfo: L{objects.Node}
1852
    @param ninfo: the node to check
1853
    @param nresult: the remote results for the node
1854

1855
    """
1856
    node = ninfo.name
1857
    # We just have to verify the paths on master and/or master candidates
1858
    # as the oob helper is invoked on the master
1859
    if ((ninfo.master_candidate or ninfo.master_capable) and
1860
        constants.NV_OOB_PATHS in nresult):
1861
      for path_result in nresult[constants.NV_OOB_PATHS]:
1862
        self._ErrorIf(path_result, self.ENODEOOBPATH, node, path_result)
1863

    
1864
  def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
1865
    """Verifies and updates the node volume data.
1866

1867
    This function will update a L{NodeImage}'s internal structures
1868
    with data from the remote call.
1869

1870
    @type ninfo: L{objects.Node}
1871
    @param ninfo: the node to check
1872
    @param nresult: the remote results for the node
1873
    @param nimg: the node image object
1874
    @param vg_name: the configured VG name
1875

1876
    """
1877
    node = ninfo.name
1878
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1879

    
1880
    nimg.lvm_fail = True
1881
    lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
1882
    if vg_name is None:
1883
      pass
1884
    elif isinstance(lvdata, basestring):
1885
      _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
1886
               utils.SafeEncode(lvdata))
1887
    elif not isinstance(lvdata, dict):
1888
      _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
1889
    else:
1890
      nimg.volumes = lvdata
1891
      nimg.lvm_fail = False
1892

    
1893
  def _UpdateNodeInstances(self, ninfo, nresult, nimg):
1894
    """Verifies and updates the node instance list.
1895

1896
    If the listing was successful, then updates this node's instance
1897
    list. Otherwise, it marks the RPC call as failed for the instance
1898
    list key.
1899

1900
    @type ninfo: L{objects.Node}
1901
    @param ninfo: the node to check
1902
    @param nresult: the remote results for the node
1903
    @param nimg: the node image object
1904

1905
    """
1906
    idata = nresult.get(constants.NV_INSTANCELIST, None)
1907
    test = not isinstance(idata, list)
1908
    self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed"
1909
                  " (instancelist): %s", utils.SafeEncode(str(idata)))
1910
    if test:
1911
      nimg.hyp_fail = True
1912
    else:
1913
      nimg.instances = idata
1914

    
1915
  def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
1916
    """Verifies and computes a node information map
1917

1918
    @type ninfo: L{objects.Node}
1919
    @param ninfo: the node to check
1920
    @param nresult: the remote results for the node
1921
    @param nimg: the node image object
1922
    @param vg_name: the configured VG name
1923

1924
    """
1925
    node = ninfo.name
1926
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1927

    
1928
    # try to read free memory (from the hypervisor)
1929
    hv_info = nresult.get(constants.NV_HVINFO, None)
1930
    test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
1931
    _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
1932
    if not test:
1933
      try:
1934
        nimg.mfree = int(hv_info["memory_free"])
1935
      except (ValueError, TypeError):
1936
        _ErrorIf(True, self.ENODERPC, node,
1937
                 "node returned invalid nodeinfo, check hypervisor")
1938

    
1939
    # FIXME: devise a free space model for file based instances as well
1940
    if vg_name is not None:
1941
      test = (constants.NV_VGLIST not in nresult or
1942
              vg_name not in nresult[constants.NV_VGLIST])
1943
      _ErrorIf(test, self.ENODELVM, node,
1944
               "node didn't return data for the volume group '%s'"
1945
               " - it is either missing or broken", vg_name)
1946
      if not test:
1947
        try:
1948
          nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
1949
        except (ValueError, TypeError):
1950
          _ErrorIf(True, self.ENODERPC, node,
1951
                   "node returned invalid LVM info, check LVM status")
1952

    
1953
  def _CollectDiskInfo(self, nodelist, node_image, instanceinfo):
1954
    """Gets per-disk status information for all instances.
1955

1956
    @type nodelist: list of strings
1957
    @param nodelist: Node names
1958
    @type node_image: dict of (name, L{objects.Node})
1959
    @param node_image: Node objects
1960
    @type instanceinfo: dict of (name, L{objects.Instance})
1961
    @param instanceinfo: Instance objects
1962
    @rtype: {instance: {node: [(succes, payload)]}}
1963
    @return: a dictionary of per-instance dictionaries with nodes as
1964
        keys and disk information as values; the disk information is a
1965
        list of tuples (success, payload)
1966

1967
    """
1968
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1969

    
1970
    node_disks = {}
1971
    node_disks_devonly = {}
1972
    diskless_instances = set()
1973
    diskless = constants.DT_DISKLESS
1974

    
1975
    for nname in nodelist:
1976
      node_instances = list(itertools.chain(node_image[nname].pinst,
1977
                                            node_image[nname].sinst))
1978
      diskless_instances.update(inst for inst in node_instances
1979
                                if instanceinfo[inst].disk_template == diskless)
1980
      disks = [(inst, disk)
1981
               for inst in node_instances
1982
               for disk in instanceinfo[inst].disks]
1983

    
1984
      if not disks:
1985
        # No need to collect data
1986
        continue
1987

    
1988
      node_disks[nname] = disks
1989

    
1990
      # Creating copies as SetDiskID below will modify the objects and that can
1991
      # lead to incorrect data returned from nodes
1992
      devonly = [dev.Copy() for (_, dev) in disks]
1993

    
1994
      for dev in devonly:
1995
        self.cfg.SetDiskID(dev, nname)
1996

    
1997
      node_disks_devonly[nname] = devonly
1998

    
1999
    assert len(node_disks) == len(node_disks_devonly)
2000

    
2001
    # Collect data from all nodes with disks
2002
    result = self.rpc.call_blockdev_getmirrorstatus_multi(node_disks.keys(),
2003
                                                          node_disks_devonly)
2004

    
2005
    assert len(result) == len(node_disks)
2006

    
2007
    instdisk = {}
2008

    
2009
    for (nname, nres) in result.items():
2010
      disks = node_disks[nname]
2011

    
2012
      if nres.offline:
2013
        # No data from this node
2014
        data = len(disks) * [(False, "node offline")]
2015
      else:
2016
        msg = nres.fail_msg
2017
        _ErrorIf(msg, self.ENODERPC, nname,
2018
                 "while getting disk information: %s", msg)
2019
        if msg:
2020
          # No data from this node
2021
          data = len(disks) * [(False, msg)]
2022
        else:
2023
          data = []
2024
          for idx, i in enumerate(nres.payload):
2025
            if isinstance(i, (tuple, list)) and len(i) == 2:
2026
              data.append(i)
2027
            else:
2028
              logging.warning("Invalid result from node %s, entry %d: %s",
2029
                              nname, idx, i)
2030
              data.append((False, "Invalid result from the remote node"))
2031

    
2032
      for ((inst, _), status) in zip(disks, data):
2033
        instdisk.setdefault(inst, {}).setdefault(nname, []).append(status)
2034

    
2035
    # Add empty entries for diskless instances.
2036
    for inst in diskless_instances:
2037
      assert inst not in instdisk
2038
      instdisk[inst] = {}
2039

    
2040
    assert compat.all(len(statuses) == len(instanceinfo[inst].disks) and
2041
                      len(nnames) <= len(instanceinfo[inst].all_nodes) and
2042
                      compat.all(isinstance(s, (tuple, list)) and
2043
                                 len(s) == 2 for s in statuses)
2044
                      for inst, nnames in instdisk.items()
2045
                      for nname, statuses in nnames.items())
2046
    assert set(instdisk) == set(instanceinfo), "instdisk consistency failure"
2047

    
2048
    return instdisk
2049

    
2050
  def _VerifyHVP(self, hvp_data):
2051
    """Verifies locally the syntax of the hypervisor parameters.
2052

2053
    """
2054
    for item, hv_name, hv_params in hvp_data:
2055
      msg = ("hypervisor %s parameters syntax check (source %s): %%s" %
2056
             (item, hv_name))
2057
      try:
2058
        hv_class = hypervisor.GetHypervisor(hv_name)
2059
        utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2060
        hv_class.CheckParameterSyntax(hv_params)
2061
      except errors.GenericError, err:
2062
        self._ErrorIf(True, self.ECLUSTERCFG, None, msg % str(err))
2063

    
2064

    
2065
  def BuildHooksEnv(self):
2066
    """Build hooks env.
2067

2068
    Cluster-Verify hooks just ran in the post phase and their failure makes
2069
    the output be logged in the verify output and the verification to fail.
2070

2071
    """
2072
    all_nodes = self.cfg.GetNodeList()
2073
    env = {
2074
      "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
2075
      }
2076
    for node in self.cfg.GetAllNodesInfo().values():
2077
      env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
2078

    
2079
    return env, [], all_nodes
2080

    
2081
  def Exec(self, feedback_fn):
2082
    """Verify integrity of cluster, performing various test on nodes.
2083

2084
    """
2085
    # This method has too many local variables. pylint: disable-msg=R0914
2086
    self.bad = False
2087
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
2088
    verbose = self.op.verbose
2089
    self._feedback_fn = feedback_fn
2090
    feedback_fn("* Verifying global settings")
2091
    for msg in self.cfg.VerifyConfig():
2092
      _ErrorIf(True, self.ECLUSTERCFG, None, msg)
2093

    
2094
    # Check the cluster certificates
2095
    for cert_filename in constants.ALL_CERT_FILES:
2096
      (errcode, msg) = _VerifyCertificate(cert_filename)
2097
      _ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
2098

    
2099
    vg_name = self.cfg.GetVGName()
2100
    drbd_helper = self.cfg.GetDRBDHelper()
2101
    hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
2102
    cluster = self.cfg.GetClusterInfo()
2103
    nodelist = utils.NiceSort(self.cfg.GetNodeList())
2104
    nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
2105
    nodeinfo_byname = dict(zip(nodelist, nodeinfo))
2106
    instancelist = utils.NiceSort(self.cfg.GetInstanceList())
2107
    instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
2108
                        for iname in instancelist)
2109
    groupinfo = self.cfg.GetAllNodeGroupsInfo()
2110
    i_non_redundant = [] # Non redundant instances
2111
    i_non_a_balanced = [] # Non auto-balanced instances
2112
    n_offline = 0 # Count of offline nodes
2113
    n_drained = 0 # Count of nodes being drained
2114
    node_vol_should = {}
2115

    
2116
    # FIXME: verify OS list
2117
    # do local checksums
2118
    master_files = [constants.CLUSTER_CONF_FILE]
2119
    master_node = self.master_node = self.cfg.GetMasterNode()
2120
    master_ip = self.cfg.GetMasterIP()
2121

    
2122
    file_names = ssconf.SimpleStore().GetFileList()
2123
    file_names.extend(constants.ALL_CERT_FILES)
2124
    file_names.extend(master_files)
2125
    if cluster.modify_etc_hosts:
2126
      file_names.append(constants.ETC_HOSTS)
2127

    
2128
    local_checksums = utils.FingerprintFiles(file_names)
2129

    
2130
    # Compute the set of hypervisor parameters
2131
    hvp_data = []
2132
    for hv_name in hypervisors:
2133
      hvp_data.append(("cluster", hv_name, cluster.GetHVDefaults(hv_name)))
2134
    for os_name, os_hvp in cluster.os_hvp.items():
2135
      for hv_name, hv_params in os_hvp.items():
2136
        if not hv_params:
2137
          continue
2138
        full_params = cluster.GetHVDefaults(hv_name, os_name=os_name)
2139
        hvp_data.append(("os %s" % os_name, hv_name, full_params))
2140
    # TODO: collapse identical parameter values in a single one
2141
    for instance in instanceinfo.values():
2142
      if not instance.hvparams:
2143
        continue
2144
      hvp_data.append(("instance %s" % instance.name, instance.hypervisor,
2145
                       cluster.FillHV(instance)))
2146
    # and verify them locally
2147
    self._VerifyHVP(hvp_data)
2148

    
2149
    feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
2150
    node_verify_param = {
2151
      constants.NV_FILELIST: file_names,
2152
      constants.NV_NODELIST: [node.name for node in nodeinfo
2153
                              if not node.offline],
2154
      constants.NV_HYPERVISOR: hypervisors,
2155
      constants.NV_HVPARAMS: hvp_data,
2156
      constants.NV_NODENETTEST: [(node.name, node.primary_ip,
2157
                                  node.secondary_ip) for node in nodeinfo
2158
                                 if not node.offline],
2159
      constants.NV_INSTANCELIST: hypervisors,
2160
      constants.NV_VERSION: None,
2161
      constants.NV_HVINFO: self.cfg.GetHypervisorType(),
2162
      constants.NV_NODESETUP: None,
2163
      constants.NV_TIME: None,
2164
      constants.NV_MASTERIP: (master_node, master_ip),
2165
      constants.NV_OSLIST: None,
2166
      constants.NV_VMNODES: self.cfg.GetNonVmCapableNodeList(),
2167
      }
2168

    
2169
    if vg_name is not None:
2170
      node_verify_param[constants.NV_VGLIST] = None
2171
      node_verify_param[constants.NV_LVLIST] = vg_name
2172
      node_verify_param[constants.NV_PVLIST] = [vg_name]
2173
      node_verify_param[constants.NV_DRBDLIST] = None
2174

    
2175
    if drbd_helper:
2176
      node_verify_param[constants.NV_DRBDHELPER] = drbd_helper
2177

    
2178
    # Build our expected cluster state
2179
    node_image = dict((node.name, self.NodeImage(offline=node.offline,
2180
                                                 name=node.name,
2181
                                                 vm_capable=node.vm_capable))
2182
                      for node in nodeinfo)
2183

    
2184
    # Gather OOB paths
2185
    oob_paths = []
2186
    for node in nodeinfo:
2187
      path = _SupportsOob(self.cfg, node)
2188
      if path and path not in oob_paths:
2189
        oob_paths.append(path)
2190

    
2191
    if oob_paths:
2192
      node_verify_param[constants.NV_OOB_PATHS] = oob_paths
2193

    
2194
    for instance in instancelist:
2195
      inst_config = instanceinfo[instance]
2196

    
2197
      for nname in inst_config.all_nodes:
2198
        if nname not in node_image:
2199
          # ghost node
2200
          gnode = self.NodeImage(name=nname)
2201
          gnode.ghost = True
2202
          node_image[nname] = gnode
2203

    
2204
      inst_config.MapLVsByNode(node_vol_should)
2205

    
2206
      pnode = inst_config.primary_node
2207
      node_image[pnode].pinst.append(instance)
2208

    
2209
      for snode in inst_config.secondary_nodes:
2210
        nimg = node_image[snode]
2211
        nimg.sinst.append(instance)
2212
        if pnode not in nimg.sbp:
2213
          nimg.sbp[pnode] = []
2214
        nimg.sbp[pnode].append(instance)
2215

    
2216
    # At this point, we have the in-memory data structures complete,
2217
    # except for the runtime information, which we'll gather next
2218

    
2219
    # Due to the way our RPC system works, exact response times cannot be
2220
    # guaranteed (e.g. a broken node could run into a timeout). By keeping the
2221
    # time before and after executing the request, we can at least have a time
2222
    # window.
2223
    nvinfo_starttime = time.time()
2224
    all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
2225
                                           self.cfg.GetClusterName())
2226
    nvinfo_endtime = time.time()
2227

    
2228
    all_drbd_map = self.cfg.ComputeDRBDMap()
2229

    
2230
    feedback_fn("* Gathering disk information (%s nodes)" % len(nodelist))
2231
    instdisk = self._CollectDiskInfo(nodelist, node_image, instanceinfo)
2232

    
2233
    feedback_fn("* Verifying node status")
2234

    
2235
    refos_img = None
2236

    
2237
    for node_i in nodeinfo:
2238
      node = node_i.name
2239
      nimg = node_image[node]
2240

    
2241
      if node_i.offline:
2242
        if verbose:
2243
          feedback_fn("* Skipping offline node %s" % (node,))
2244
        n_offline += 1
2245
        continue
2246

    
2247
      if node == master_node:
2248
        ntype = "master"
2249
      elif node_i.master_candidate:
2250
        ntype = "master candidate"
2251
      elif node_i.drained:
2252
        ntype = "drained"
2253
        n_drained += 1
2254
      else:
2255
        ntype = "regular"
2256
      if verbose:
2257
        feedback_fn("* Verifying node %s (%s)" % (node, ntype))
2258

    
2259
      msg = all_nvinfo[node].fail_msg
2260
      _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
2261
      if msg:
2262
        nimg.rpc_fail = True
2263
        continue
2264

    
2265
      nresult = all_nvinfo[node].payload
2266

    
2267
      nimg.call_ok = self._VerifyNode(node_i, nresult)
2268
      self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
2269
      self._VerifyNodeNetwork(node_i, nresult)
2270
      self._VerifyNodeFiles(node_i, nresult, file_names, local_checksums,
2271
                            master_files)
2272

    
2273
      self._VerifyOob(node_i, nresult)
2274

    
2275
      if nimg.vm_capable:
2276
        self._VerifyNodeLVM(node_i, nresult, vg_name)
2277
        self._VerifyNodeDrbd(node_i, nresult, instanceinfo, drbd_helper,
2278
                             all_drbd_map)
2279

    
2280
        self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
2281
        self._UpdateNodeInstances(node_i, nresult, nimg)
2282
        self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
2283
        self._UpdateNodeOS(node_i, nresult, nimg)
2284
        if not nimg.os_fail:
2285
          if refos_img is None:
2286
            refos_img = nimg
2287
          self._VerifyNodeOS(node_i, nimg, refos_img)
2288

    
2289
    feedback_fn("* Verifying instance status")
2290
    for instance in instancelist:
2291
      if verbose:
2292
        feedback_fn("* Verifying instance %s" % instance)
2293
      inst_config = instanceinfo[instance]
2294
      self._VerifyInstance(instance, inst_config, node_image,
2295
                           instdisk[instance])
2296
      inst_nodes_offline = []
2297

    
2298
      pnode = inst_config.primary_node
2299
      pnode_img = node_image[pnode]
2300
      _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
2301
               self.ENODERPC, pnode, "instance %s, connection to"
2302
               " primary node failed", instance)
2303

    
2304
      _ErrorIf(pnode_img.offline, self.EINSTANCEBADNODE, instance,
2305
               "instance lives on offline node %s", inst_config.primary_node)
2306

    
2307
      # If the instance is non-redundant we cannot survive losing its primary
2308
      # node, so we are not N+1 compliant. On the other hand we have no disk
2309
      # templates with more than one secondary so that situation is not well
2310
      # supported either.
2311
      # FIXME: does not support file-backed instances
2312
      if not inst_config.secondary_nodes:
2313
        i_non_redundant.append(instance)
2314

    
2315
      _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT,
2316
               instance, "instance has multiple secondary nodes: %s",
2317
               utils.CommaJoin(inst_config.secondary_nodes),
2318
               code=self.ETYPE_WARNING)
2319

    
2320
      if inst_config.disk_template in constants.DTS_NET_MIRROR:
2321
        pnode = inst_config.primary_node
2322
        instance_nodes = utils.NiceSort(inst_config.all_nodes)
2323
        instance_groups = {}
2324

    
2325
        for node in instance_nodes:
2326
          instance_groups.setdefault(nodeinfo_byname[node].group,
2327
                                     []).append(node)
2328

    
2329
        pretty_list = [
2330
          "%s (group %s)" % (utils.CommaJoin(nodes), groupinfo[group].name)
2331
          # Sort so that we always list the primary node first.
2332
          for group, nodes in sorted(instance_groups.items(),
2333
                                     key=lambda (_, nodes): pnode in nodes,
2334
                                     reverse=True)]
2335

    
2336
        self._ErrorIf(len(instance_groups) > 1, self.EINSTANCESPLITGROUPS,
2337
                      instance, "instance has primary and secondary nodes in"
2338
                      " different groups: %s", utils.CommaJoin(pretty_list),
2339
                      code=self.ETYPE_WARNING)
2340

    
2341
      if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
2342
        i_non_a_balanced.append(instance)
2343

    
2344
      for snode in inst_config.secondary_nodes:
2345
        s_img = node_image[snode]
2346
        _ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode,
2347
                 "instance %s, connection to secondary node failed", instance)
2348

    
2349
        if s_img.offline:
2350
          inst_nodes_offline.append(snode)
2351

    
2352
      # warn that the instance lives on offline nodes
2353
      _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
2354
               "instance has offline secondary node(s) %s",
2355
               utils.CommaJoin(inst_nodes_offline))
2356
      # ... or ghost/non-vm_capable nodes
2357
      for node in inst_config.all_nodes:
2358
        _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
2359
                 "instance lives on ghost node %s", node)
2360
        _ErrorIf(not node_image[node].vm_capable, self.EINSTANCEBADNODE,
2361
                 instance, "instance lives on non-vm_capable node %s", node)
2362

    
2363
    feedback_fn("* Verifying orphan volumes")
2364
    reserved = utils.FieldSet(*cluster.reserved_lvs)
2365
    self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
2366

    
2367
    feedback_fn("* Verifying orphan instances")
2368
    self._VerifyOrphanInstances(instancelist, node_image)
2369

    
2370
    if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
2371
      feedback_fn("* Verifying N+1 Memory redundancy")
2372
      self._VerifyNPlusOneMemory(node_image, instanceinfo)
2373

    
2374
    feedback_fn("* Other Notes")
2375
    if i_non_redundant:
2376
      feedback_fn("  - NOTICE: %d non-redundant instance(s) found."
2377
                  % len(i_non_redundant))
2378

    
2379
    if i_non_a_balanced:
2380
      feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
2381
                  % len(i_non_a_balanced))
2382

    
2383
    if n_offline:
2384
      feedback_fn("  - NOTICE: %d offline node(s) found." % n_offline)
2385

    
2386
    if n_drained:
2387
      feedback_fn("  - NOTICE: %d drained node(s) found." % n_drained)
2388

    
2389
    return not self.bad
2390

    
2391
  def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
2392
    """Analyze the post-hooks' result
2393

2394
    This method analyses the hook result, handles it, and sends some
2395
    nicely-formatted feedback back to the user.
2396

2397
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
2398
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
2399
    @param hooks_results: the results of the multi-node hooks rpc call
2400
    @param feedback_fn: function used send feedback back to the caller
2401
    @param lu_result: previous Exec result
2402
    @return: the new Exec result, based on the previous result
2403
        and hook results
2404

2405
    """
2406
    # We only really run POST phase hooks, and are only interested in
2407
    # their results
2408
    if phase == constants.HOOKS_PHASE_POST:
2409
      # Used to change hooks' output to proper indentation
2410
      feedback_fn("* Hooks Results")
2411
      assert hooks_results, "invalid result from hooks"
2412

    
2413
      for node_name in hooks_results:
2414
        res = hooks_results[node_name]
2415
        msg = res.fail_msg
2416
        test = msg and not res.offline
2417
        self._ErrorIf(test, self.ENODEHOOKS, node_name,
2418
                      "Communication failure in hooks execution: %s", msg)
2419
        if res.offline or msg:
2420
          # No need to investigate payload if node is offline or gave an error.
2421
          # override manually lu_result here as _ErrorIf only
2422
          # overrides self.bad
2423
          lu_result = 1
2424
          continue
2425
        for script, hkr, output in res.payload:
2426
          test = hkr == constants.HKR_FAIL
2427
          self._ErrorIf(test, self.ENODEHOOKS, node_name,
2428
                        "Script %s failed, output:", script)
2429
          if test:
2430
            output = self._HOOKS_INDENT_RE.sub('      ', output)
2431
            feedback_fn("%s" % output)
2432
            lu_result = 0
2433

    
2434
      return lu_result
2435

    
2436

    
2437
class LUClusterVerifyDisks(NoHooksLU):
2438
  """Verifies the cluster disks status.
2439

2440
  """
2441
  REQ_BGL = False
2442

    
2443
  def ExpandNames(self):
2444
    self.needed_locks = {
2445
      locking.LEVEL_NODE: locking.ALL_SET,
2446
      locking.LEVEL_INSTANCE: locking.ALL_SET,
2447
    }
2448
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
2449

    
2450
  def Exec(self, feedback_fn):
2451
    """Verify integrity of cluster disks.
2452

2453
    @rtype: tuple of three items
2454
    @return: a tuple of (dict of node-to-node_error, list of instances
2455
        which need activate-disks, dict of instance: (node, volume) for
2456
        missing volumes
2457

2458
    """
2459
    result = res_nodes, res_instances, res_missing = {}, [], {}
2460

    
2461
    nodes = utils.NiceSort(self.cfg.GetVmCapableNodeList())
2462
    instances = self.cfg.GetAllInstancesInfo().values()
2463

    
2464
    nv_dict = {}
2465
    for inst in instances:
2466
      inst_lvs = {}
2467
      if not inst.admin_up:
2468
        continue
2469
      inst.MapLVsByNode(inst_lvs)
2470
      # transform { iname: {node: [vol,],},} to {(node, vol): iname}
2471
      for node, vol_list in inst_lvs.iteritems():
2472
        for vol in vol_list:
2473
          nv_dict[(node, vol)] = inst
2474

    
2475
    if not nv_dict:
2476
      return result
2477

    
2478
    node_lvs = self.rpc.call_lv_list(nodes, [])
2479
    for node, node_res in node_lvs.items():
2480
      if node_res.offline:
2481
        continue
2482
      msg = node_res.fail_msg
2483
      if msg:
2484
        logging.warning("Error enumerating LVs on node %s: %s", node, msg)
2485
        res_nodes[node] = msg
2486
        continue
2487

    
2488
      lvs = node_res.payload
2489
      for lv_name, (_, _, lv_online) in lvs.items():
2490
        inst = nv_dict.pop((node, lv_name), None)
2491
        if (not lv_online and inst is not None
2492
            and inst.name not in res_instances):
2493
          res_instances.append(inst.name)
2494

    
2495
    # any leftover items in nv_dict are missing LVs, let's arrange the
2496
    # data better
2497
    for key, inst in nv_dict.iteritems():
2498
      if inst.name not in res_missing:
2499
        res_missing[inst.name] = []
2500
      res_missing[inst.name].append(key)
2501

    
2502
    return result
2503

    
2504

    
2505
class LUClusterRepairDiskSizes(NoHooksLU):
2506
  """Verifies the cluster disks sizes.
2507

2508
  """
2509
  REQ_BGL = False
2510

    
2511
  def ExpandNames(self):
2512
    if self.op.instances:
2513
      self.wanted_names = []
2514
      for name in self.op.instances:
2515
        full_name = _ExpandInstanceName(self.cfg, name)
2516
        self.wanted_names.append(full_name)
2517
      self.needed_locks = {
2518
        locking.LEVEL_NODE: [],
2519
        locking.LEVEL_INSTANCE: self.wanted_names,
2520
        }
2521
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2522
    else:
2523
      self.wanted_names = None
2524
      self.needed_locks = {
2525
        locking.LEVEL_NODE: locking.ALL_SET,
2526
        locking.LEVEL_INSTANCE: locking.ALL_SET,
2527
        }
2528
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
2529

    
2530
  def DeclareLocks(self, level):
2531
    if level == locking.LEVEL_NODE and self.wanted_names is not None:
2532
      self._LockInstancesNodes(primary_only=True)
2533

    
2534
  def CheckPrereq(self):
2535
    """Check prerequisites.
2536

2537
    This only checks the optional instance list against the existing names.
2538

2539
    """
2540
    if self.wanted_names is None:
2541
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
2542

    
2543
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
2544
                             in self.wanted_names]
2545

    
2546
  def _EnsureChildSizes(self, disk):
2547
    """Ensure children of the disk have the needed disk size.
2548

2549
    This is valid mainly for DRBD8 and fixes an issue where the
2550
    children have smaller disk size.
2551

2552
    @param disk: an L{ganeti.objects.Disk} object
2553

2554
    """
2555
    if disk.dev_type == constants.LD_DRBD8:
2556
      assert disk.children, "Empty children for DRBD8?"
2557
      fchild = disk.children[0]
2558
      mismatch = fchild.size < disk.size
2559
      if mismatch:
2560
        self.LogInfo("Child disk has size %d, parent %d, fixing",
2561
                     fchild.size, disk.size)
2562
        fchild.size = disk.size
2563

    
2564
      # and we recurse on this child only, not on the metadev
2565
      return self._EnsureChildSizes(fchild) or mismatch
2566
    else:
2567
      return False
2568

    
2569
  def Exec(self, feedback_fn):
2570
    """Verify the size of cluster disks.
2571

2572
    """
2573
    # TODO: check child disks too
2574
    # TODO: check differences in size between primary/secondary nodes
2575
    per_node_disks = {}
2576
    for instance in self.wanted_instances:
2577
      pnode = instance.primary_node
2578
      if pnode not in per_node_disks:
2579
        per_node_disks[pnode] = []
2580
      for idx, disk in enumerate(instance.disks):
2581
        per_node_disks[pnode].append((instance, idx, disk))
2582

    
2583
    changed = []
2584
    for node, dskl in per_node_disks.items():
2585
      newl = [v[2].Copy() for v in dskl]
2586
      for dsk in newl:
2587
        self.cfg.SetDiskID(dsk, node)
2588
      result = self.rpc.call_blockdev_getsize(node, newl)
2589
      if result.fail_msg:
2590
        self.LogWarning("Failure in blockdev_getsize call to node"
2591
                        " %s, ignoring", node)
2592
        continue
2593
      if len(result.payload) != len(dskl):
2594
        logging.warning("Invalid result from node %s: len(dksl)=%d,"
2595
                        " result.payload=%s", node, len(dskl), result.payload)
2596
        self.LogWarning("Invalid result from node %s, ignoring node results",
2597
                        node)
2598
        continue
2599
      for ((instance, idx, disk), size) in zip(dskl, result.payload):
2600
        if size is None:
2601
          self.LogWarning("Disk %d of instance %s did not return size"
2602
                          " information, ignoring", idx, instance.name)
2603
          continue
2604
        if not isinstance(size, (int, long)):
2605
          self.LogWarning("Disk %d of instance %s did not return valid"
2606
                          " size information, ignoring", idx, instance.name)
2607
          continue
2608
        size = size >> 20
2609
        if size != disk.size:
2610
          self.LogInfo("Disk %d of instance %s has mismatched size,"
2611
                       " correcting: recorded %d, actual %d", idx,
2612
                       instance.name, disk.size, size)
2613
          disk.size = size
2614
          self.cfg.Update(instance, feedback_fn)
2615
          changed.append((instance.name, idx, size))
2616
        if self._EnsureChildSizes(disk):
2617
          self.cfg.Update(instance, feedback_fn)
2618
          changed.append((instance.name, idx, disk.size))
2619
    return changed
2620

    
2621

    
2622
class LUClusterRename(LogicalUnit):
2623
  """Rename the cluster.
2624

2625
  """
2626
  HPATH = "cluster-rename"
2627
  HTYPE = constants.HTYPE_CLUSTER
2628

    
2629
  def BuildHooksEnv(self):
2630
    """Build hooks env.
2631

2632
    """
2633
    env = {
2634
      "OP_TARGET": self.cfg.GetClusterName(),
2635
      "NEW_NAME": self.op.name,
2636
      }
2637
    mn = self.cfg.GetMasterNode()
2638
    all_nodes = self.cfg.GetNodeList()
2639
    return env, [mn], all_nodes
2640

    
2641
  def CheckPrereq(self):
2642
    """Verify that the passed name is a valid one.
2643

2644
    """
2645
    hostname = netutils.GetHostname(name=self.op.name,
2646
                                    family=self.cfg.GetPrimaryIPFamily())
2647

    
2648
    new_name = hostname.name
2649
    self.ip = new_ip = hostname.ip
2650
    old_name = self.cfg.GetClusterName()
2651
    old_ip = self.cfg.GetMasterIP()
2652
    if new_name == old_name and new_ip == old_ip:
2653
      raise errors.OpPrereqError("Neither the name nor the IP address of the"
2654
                                 " cluster has changed",
2655
                                 errors.ECODE_INVAL)
2656
    if new_ip != old_ip:
2657
      if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
2658
        raise errors.OpPrereqError("The given cluster IP address (%s) is"
2659
                                   " reachable on the network" %
2660
                                   new_ip, errors.ECODE_NOTUNIQUE)
2661

    
2662
    self.op.name = new_name
2663

    
2664
  def Exec(self, feedback_fn):
2665
    """Rename the cluster.
2666

2667
    """
2668
    clustername = self.op.name
2669
    ip = self.ip
2670

    
2671
    # shutdown the master IP
2672
    master = self.cfg.GetMasterNode()
2673
    result = self.rpc.call_node_stop_master(master, False)
2674
    result.Raise("Could not disable the master role")
2675

    
2676
    try:
2677
      cluster = self.cfg.GetClusterInfo()
2678
      cluster.cluster_name = clustername
2679
      cluster.master_ip = ip
2680
      self.cfg.Update(cluster, feedback_fn)
2681

    
2682
      # update the known hosts file
2683
      ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
2684
      node_list = self.cfg.GetOnlineNodeList()
2685
      try:
2686
        node_list.remove(master)
2687
      except ValueError:
2688
        pass
2689
      _UploadHelper(self, node_list, constants.SSH_KNOWN_HOSTS_FILE)
2690
    finally:
2691
      result = self.rpc.call_node_start_master(master, False, False)
2692
      msg = result.fail_msg
2693
      if msg:
2694
        self.LogWarning("Could not re-enable the master role on"
2695
                        " the master, please restart manually: %s", msg)
2696

    
2697
    return clustername
2698

    
2699

    
2700
class LUClusterSetParams(LogicalUnit):
2701
  """Change the parameters of the cluster.
2702

2703
  """
2704
  HPATH = "cluster-modify"
2705
  HTYPE = constants.HTYPE_CLUSTER
2706
  REQ_BGL = False
2707

    
2708
  def CheckArguments(self):
2709
    """Check parameters
2710

2711
    """
2712
    if self.op.uid_pool:
2713
      uidpool.CheckUidPool(self.op.uid_pool)
2714

    
2715
    if self.op.add_uids:
2716
      uidpool.CheckUidPool(self.op.add_uids)
2717

    
2718
    if self.op.remove_uids:
2719
      uidpool.CheckUidPool(self.op.remove_uids)
2720

    
2721
  def ExpandNames(self):
2722
    # FIXME: in the future maybe other cluster params won't require checking on
2723
    # all nodes to be modified.
2724
    self.needed_locks = {
2725
      locking.LEVEL_NODE: locking.ALL_SET,
2726
    }
2727
    self.share_locks[locking.LEVEL_NODE] = 1
2728

    
2729
  def BuildHooksEnv(self):
2730
    """Build hooks env.
2731

2732
    """
2733
    env = {
2734
      "OP_TARGET": self.cfg.GetClusterName(),
2735
      "NEW_VG_NAME": self.op.vg_name,
2736
      }
2737
    mn = self.cfg.GetMasterNode()
2738
    return env, [mn], [mn]
2739

    
2740
  def CheckPrereq(self):
2741
    """Check prerequisites.
2742

2743
    This checks whether the given params don't conflict and
2744
    if the given volume group is valid.
2745

2746
    """
2747
    if self.op.vg_name is not None and not self.op.vg_name:
2748
      if self.cfg.HasAnyDiskOfType(constants.LD_LV):
2749
        raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based"
2750
                                   " instances exist", errors.ECODE_INVAL)
2751

    
2752
    if self.op.drbd_helper is not None and not self.op.drbd_helper:
2753
      if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8):
2754
        raise errors.OpPrereqError("Cannot disable drbd helper while"
2755
                                   " drbd-based instances exist",
2756
                                   errors.ECODE_INVAL)
2757

    
2758
    node_list = self.acquired_locks[locking.LEVEL_NODE]
2759

    
2760
    # if vg_name not None, checks given volume group on all nodes
2761
    if self.op.vg_name:
2762
      vglist = self.rpc.call_vg_list(node_list)
2763
      for node in node_list:
2764
        msg = vglist[node].fail_msg
2765
        if msg:
2766
          # ignoring down node
2767
          self.LogWarning("Error while gathering data on node %s"
2768
                          " (ignoring node): %s", node, msg)
2769
          continue
2770
        vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
2771
                                              self.op.vg_name,
2772
                                              constants.MIN_VG_SIZE)
2773
        if vgstatus:
2774
          raise errors.OpPrereqError("Error on node '%s': %s" %
2775
                                     (node, vgstatus), errors.ECODE_ENVIRON)
2776

    
2777
    if self.op.drbd_helper:
2778
      # checks given drbd helper on all nodes
2779
      helpers = self.rpc.call_drbd_helper(node_list)
2780
      for node in node_list:
2781
        ninfo = self.cfg.GetNodeInfo(node)
2782
        if ninfo.offline:
2783
          self.LogInfo("Not checking drbd helper on offline node %s", node)
2784
          continue
2785
        msg = helpers[node].fail_msg
2786
        if msg:
2787
          raise errors.OpPrereqError("Error checking drbd helper on node"
2788
                                     " '%s': %s" % (node, msg),
2789
                                     errors.ECODE_ENVIRON)
2790
        node_helper = helpers[node].payload
2791
        if node_helper != self.op.drbd_helper:
2792
          raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" %
2793
                                     (node, node_helper), errors.ECODE_ENVIRON)
2794

    
2795
    self.cluster = cluster = self.cfg.GetClusterInfo()
2796
    # validate params changes
2797
    if self.op.beparams:
2798
      utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
2799
      self.new_beparams = cluster.SimpleFillBE(self.op.beparams)
2800

    
2801
    if self.op.ndparams:
2802
      utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
2803
      self.new_ndparams = cluster.SimpleFillND(self.op.ndparams)
2804

    
2805
      # TODO: we need a more general way to handle resetting
2806
      # cluster-level parameters to default values
2807
      if self.new_ndparams["oob_program"] == "":
2808
        self.new_ndparams["oob_program"] = \
2809
            constants.NDC_DEFAULTS[constants.ND_OOB_PROGRAM]
2810

    
2811
    if self.op.nicparams:
2812
      utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
2813
      self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams)
2814
      objects.NIC.CheckParameterSyntax(self.new_nicparams)
2815
      nic_errors = []
2816

    
2817
      # check all instances for consistency
2818
      for instance in self.cfg.GetAllInstancesInfo().values():
2819
        for nic_idx, nic in enumerate(instance.nics):
2820
          params_copy = copy.deepcopy(nic.nicparams)
2821
          params_filled = objects.FillDict(self.new_nicparams, params_copy)
2822

    
2823
          # check parameter syntax
2824
          try:
2825
            objects.NIC.CheckParameterSyntax(params_filled)
2826
          except errors.ConfigurationError, err:
2827
            nic_errors.append("Instance %s, nic/%d: %s" %
2828
                              (instance.name, nic_idx, err))
2829

    
2830
          # if we're moving instances to routed, check that they have an ip
2831
          target_mode = params_filled[constants.NIC_MODE]
2832
          if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
2833
            nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
2834
                              (instance.name, nic_idx))
2835
      if nic_errors:
2836
        raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
2837
                                   "\n".join(nic_errors))
2838

    
2839
    # hypervisor list/parameters
2840
    self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
2841
    if self.op.hvparams:
2842
      for hv_name, hv_dict in self.op.hvparams.items():
2843
        if hv_name not in self.new_hvparams:
2844
          self.new_hvparams[hv_name] = hv_dict
2845
        else:
2846
          self.new_hvparams[hv_name].update(hv_dict)
2847

    
2848
    # os hypervisor parameters
2849
    self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
2850
    if self.op.os_hvp:
2851
      for os_name, hvs in self.op.os_hvp.items():
2852
        if os_name not in self.new_os_hvp:
2853
          self.new_os_hvp[os_name] = hvs
2854
        else:
2855
          for hv_name, hv_dict in hvs.items():
2856
            if hv_name not in self.new_os_hvp[os_name]:
2857
              self.new_os_hvp[os_name][hv_name] = hv_dict
2858
            else:
2859
              self.new_os_hvp[os_name][hv_name].update(hv_dict)
2860

    
2861
    # os parameters
2862
    self.new_osp = objects.FillDict(cluster.osparams, {})
2863
    if self.op.osparams:
2864
      for os_name, osp in self.op.osparams.items():
2865
        if os_name not in self.new_osp:
2866
          self.new_osp[os_name] = {}
2867

    
2868
        self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp,
2869
                                                  use_none=True)
2870

    
2871
        if not self.new_osp[os_name]:
2872
          # we removed all parameters
2873
          del self.new_osp[os_name]
2874
        else:
2875
          # check the parameter validity (remote check)
2876
          _CheckOSParams(self, False, [self.cfg.GetMasterNode()],
2877
                         os_name, self.new_osp[os_name])
2878

    
2879
    # changes to the hypervisor list
2880
    if self.op.enabled_hypervisors is not None:
2881
      self.hv_list = self.op.enabled_hypervisors
2882
      for hv in self.hv_list:
2883
        # if the hypervisor doesn't already exist in the cluster
2884
        # hvparams, we initialize it to empty, and then (in both
2885
        # cases) we make sure to fill the defaults, as we might not
2886
        # have a complete defaults list if the hypervisor wasn't
2887
        # enabled before
2888
        if hv not in new_hvp:
2889
          new_hvp[hv] = {}
2890
        new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
2891
        utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
2892
    else:
2893
      self.hv_list = cluster.enabled_hypervisors
2894

    
2895
    if self.op.hvparams or self.op.enabled_hypervisors is not None:
2896
      # either the enabled list has changed, or the parameters have, validate
2897
      for hv_name, hv_params in self.new_hvparams.items():
2898
        if ((self.op.hvparams and hv_name in self.op.hvparams) or
2899
            (self.op.enabled_hypervisors and
2900
             hv_name in self.op.enabled_hypervisors)):
2901
          # either this is a new hypervisor, or its parameters have changed
2902
          hv_class = hypervisor.GetHypervisor(hv_name)
2903
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2904
          hv_class.CheckParameterSyntax(hv_params)
2905
          _CheckHVParams(self, node_list, hv_name, hv_params)
2906

    
2907
    if self.op.os_hvp:
2908
      # no need to check any newly-enabled hypervisors, since the
2909
      # defaults have already been checked in the above code-block
2910
      for os_name, os_hvp in self.new_os_hvp.items():
2911
        for hv_name, hv_params in os_hvp.items():
2912
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2913
          # we need to fill in the new os_hvp on top of the actual hv_p
2914
          cluster_defaults = self.new_hvparams.get(hv_name, {})
2915
          new_osp = objects.FillDict(cluster_defaults, hv_params)
2916
          hv_class = hypervisor.GetHypervisor(hv_name)
2917
          hv_class.CheckParameterSyntax(new_osp)
2918
          _CheckHVParams(self, node_list, hv_name, new_osp)
2919

    
2920
    if self.op.default_iallocator:
2921
      alloc_script = utils.FindFile(self.op.default_iallocator,
2922
                                    constants.IALLOCATOR_SEARCH_PATH,
2923
                                    os.path.isfile)
2924
      if alloc_script is None:
2925
        raise errors.OpPrereqError("Invalid default iallocator script '%s'"
2926
                                   " specified" % self.op.default_iallocator,
2927
                                   errors.ECODE_INVAL)
2928

    
2929
  def Exec(self, feedback_fn):
2930
    """Change the parameters of the cluster.
2931

2932
    """
2933
    if self.op.vg_name is not None:
2934
      new_volume = self.op.vg_name
2935
      if not new_volume:
2936
        new_volume = None
2937
      if new_volume != self.cfg.GetVGName():
2938
        self.cfg.SetVGName(new_volume)
2939
      else:
2940
        feedback_fn("Cluster LVM configuration already in desired"
2941
                    " state, not changing")
2942
    if self.op.drbd_helper is not None:
2943
      new_helper = self.op.drbd_helper
2944
      if not new_helper:
2945
        new_helper = None
2946
      if new_helper != self.cfg.GetDRBDHelper():
2947
        self.cfg.SetDRBDHelper(new_helper)
2948
      else:
2949
        feedback_fn("Cluster DRBD helper already in desired state,"
2950
                    " not changing")
2951
    if self.op.hvparams:
2952
      self.cluster.hvparams = self.new_hvparams
2953
    if self.op.os_hvp:
2954
      self.cluster.os_hvp = self.new_os_hvp
2955
    if self.op.enabled_hypervisors is not None:
2956
      self.cluster.hvparams = self.new_hvparams
2957
      self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
2958
    if self.op.beparams:
2959
      self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
2960
    if self.op.nicparams:
2961
      self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
2962
    if self.op.osparams:
2963
      self.cluster.osparams = self.new_osp
2964
    if self.op.ndparams:
2965
      self.cluster.ndparams = self.new_ndparams
2966

    
2967
    if self.op.candidate_pool_size is not None:
2968
      self.cluster.candidate_pool_size = self.op.candidate_pool_size
2969
      # we need to update the pool size here, otherwise the save will fail
2970
      _AdjustCandidatePool(self, [])
2971

    
2972
    if self.op.maintain_node_health is not None:
2973
      self.cluster.maintain_node_health = self.op.maintain_node_health
2974

    
2975
    if self.op.prealloc_wipe_disks is not None:
2976
      self.cluster.prealloc_wipe_disks = self.op.prealloc_wipe_disks
2977

    
2978
    if self.op.add_uids is not None:
2979
      uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
2980

    
2981
    if self.op.remove_uids is not None:
2982
      uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
2983

    
2984
    if self.op.uid_pool is not None:
2985
      self.cluster.uid_pool = self.op.uid_pool
2986

    
2987
    if self.op.default_iallocator is not None:
2988
      self.cluster.default_iallocator = self.op.default_iallocator
2989

    
2990
    if self.op.reserved_lvs is not None:
2991
      self.cluster.reserved_lvs = self.op.reserved_lvs
2992

    
2993
    def helper_os(aname, mods, desc):
2994
      desc += " OS list"
2995
      lst = getattr(self.cluster, aname)
2996
      for key, val in mods:
2997
        if key == constants.DDM_ADD:
2998
          if val in lst:
2999
            feedback_fn("OS %s already in %s, ignoring" % (val, desc))
3000
          else:
3001
            lst.append(val)
3002
        elif key == constants.DDM_REMOVE:
3003
          if val in lst:
3004
            lst.remove(val)
3005
          else:
3006
            feedback_fn("OS %s not found in %s, ignoring" % (val, desc))
3007
        else:
3008
          raise errors.ProgrammerError("Invalid modification '%s'" % key)
3009

    
3010
    if self.op.hidden_os:
3011
      helper_os("hidden_os", self.op.hidden_os, "hidden")
3012

    
3013
    if self.op.blacklisted_os:
3014
      helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted")
3015

    
3016
    if self.op.master_netdev:
3017
      master = self.cfg.GetMasterNode()
3018
      feedback_fn("Shutting down master ip on the current netdev (%s)" %
3019
                  self.cluster.master_netdev)
3020
      result = self.rpc.call_node_stop_master(master, False)
3021
      result.Raise("Could not disable the master ip")
3022
      feedback_fn("Changing master_netdev from %s to %s" %
3023
                  (self.cluster.master_netdev, self.op.master_netdev))
3024
      self.cluster.master_netdev = self.op.master_netdev
3025

    
3026
    self.cfg.Update(self.cluster, feedback_fn)
3027

    
3028
    if self.op.master_netdev:
3029
      feedback_fn("Starting the master ip on the new master netdev (%s)" %
3030
                  self.op.master_netdev)
3031
      result = self.rpc.call_node_start_master(master, False, False)
3032
      if result.fail_msg:
3033
        self.LogWarning("Could not re-enable the master ip on"
3034
                        " the master, please restart manually: %s",
3035
                        result.fail_msg)
3036

    
3037

    
3038
def _UploadHelper(lu, nodes, fname):
3039
  """Helper for uploading a file and showing warnings.
3040

3041
  """
3042
  if os.path.exists(fname):
3043
    result = lu.rpc.call_upload_file(nodes, fname)
3044
    for to_node, to_result in result.items():
3045
      msg = to_result.fail_msg
3046
      if msg:
3047
        msg = ("Copy of file %s to node %s failed: %s" %
3048
               (fname, to_node, msg))
3049
        lu.proc.LogWarning(msg)
3050

    
3051

    
3052
def _RedistributeAncillaryFiles(lu, additional_nodes=None, additional_vm=True):
3053
  """Distribute additional files which are part of the cluster configuration.
3054

3055
  ConfigWriter takes care of distributing the config and ssconf files, but
3056
  there are more files which should be distributed to all nodes. This function
3057
  makes sure those are copied.
3058

3059
  @param lu: calling logical unit
3060
  @param additional_nodes: list of nodes not in the config to distribute to
3061
  @type additional_vm: boolean
3062
  @param additional_vm: whether the additional nodes are vm-capable or not
3063

3064
  """
3065
  # 1. Gather target nodes
3066
  myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
3067
  dist_nodes = lu.cfg.GetOnlineNodeList()
3068
  nvm_nodes = lu.cfg.GetNonVmCapableNodeList()
3069
  vm_nodes = [name for name in dist_nodes if name not in nvm_nodes]
3070
  if additional_nodes is not None:
3071
    dist_nodes.extend(additional_nodes)
3072
    if additional_vm:
3073
      vm_nodes.extend(additional_nodes)
3074
  if myself.name in dist_nodes:
3075
    dist_nodes.remove(myself.name)
3076
  if myself.name in vm_nodes:
3077
    vm_nodes.remove(myself.name)
3078

    
3079
  # 2. Gather files to distribute
3080
  dist_files = set([constants.ETC_HOSTS,
3081
                    constants.SSH_KNOWN_HOSTS_FILE,
3082
                    constants.RAPI_CERT_FILE,
3083
                    constants.RAPI_USERS_FILE,
3084
                    constants.CONFD_HMAC_KEY,
3085
                    constants.CLUSTER_DOMAIN_SECRET_FILE,
3086
                   ])
3087

    
3088
  vm_files = set()
3089
  enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
3090
  for hv_name in enabled_hypervisors:
3091
    hv_class = hypervisor.GetHypervisor(hv_name)
3092
    vm_files.update(hv_class.GetAncillaryFiles())
3093

    
3094
  # 3. Perform the files upload
3095
  for fname in dist_files:
3096
    _UploadHelper(lu, dist_nodes, fname)
3097
  for fname in vm_files:
3098
    _UploadHelper(lu, vm_nodes, fname)
3099

    
3100

    
3101
class LUClusterRedistConf(NoHooksLU):
3102
  """Force the redistribution of cluster configuration.
3103

3104
  This is a very simple LU.
3105

3106
  """
3107
  REQ_BGL = False
3108

    
3109
  def ExpandNames(self):
3110
    self.needed_locks = {
3111
      locking.LEVEL_NODE: locking.ALL_SET,
3112
    }
3113
    self.share_locks[locking.LEVEL_NODE] = 1
3114

    
3115
  def Exec(self, feedback_fn):
3116
    """Redistribute the configuration.
3117

3118
    """
3119
    self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
3120
    _RedistributeAncillaryFiles(self)
3121

    
3122

    
3123
def _WaitForSync(lu, instance, disks=None, oneshot=False):
3124
  """Sleep and poll for an instance's disk to sync.
3125

3126
  """
3127
  if not instance.disks or disks is not None and not disks:
3128
    return True
3129

    
3130
  disks = _ExpandCheckDisks(instance, disks)
3131

    
3132
  if not oneshot:
3133
    lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
3134

    
3135
  node = instance.primary_node
3136

    
3137
  for dev in disks:
3138
    lu.cfg.SetDiskID(dev, node)
3139

    
3140
  # TODO: Convert to utils.Retry
3141

    
3142
  retries = 0
3143
  degr_retries = 10 # in seconds, as we sleep 1 second each time
3144
  while True:
3145
    max_time = 0
3146
    done = True
3147
    cumul_degraded = False
3148
    rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
3149
    msg = rstats.fail_msg
3150
    if msg:
3151
      lu.LogWarning("Can't get any data from node %s: %s", node, msg)
3152
      retries += 1
3153
      if retries >= 10:
3154
        raise errors.RemoteError("Can't contact node %s for mirror data,"
3155
                                 " aborting." % node)
3156
      time.sleep(6)
3157
      continue
3158
    rstats = rstats.payload
3159
    retries = 0
3160
    for i, mstat in enumerate(rstats):
3161
      if mstat is None:
3162
        lu.LogWarning("Can't compute data for node %s/%s",
3163
                           node, disks[i].iv_name)
3164
        continue
3165

    
3166
      cumul_degraded = (cumul_degraded or
3167
                        (mstat.is_degraded and mstat.sync_percent is None))
3168
      if mstat.sync_percent is not None:
3169
        done = False
3170
        if mstat.estimated_time is not None:
3171
          rem_time = ("%s remaining (estimated)" %
3172
                      utils.FormatSeconds(mstat.estimated_time))
3173
          max_time = mstat.estimated_time
3174
        else:
3175
          rem_time = "no time estimate"
3176
        lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
3177
                        (disks[i].iv_name, mstat.sync_percent, rem_time))
3178

    
3179
    # if we're done but degraded, let's do a few small retries, to
3180
    # make sure we see a stable and not transient situation; therefore
3181
    # we force restart of the loop
3182
    if (done or oneshot) and cumul_degraded and degr_retries > 0:
3183
      logging.info("Degraded disks found, %d retries left", degr_retries)
3184
      degr_retries -= 1
3185
      time.sleep(1)
3186
      continue
3187

    
3188
    if done or oneshot:
3189
      break
3190

    
3191
    time.sleep(min(60, max_time))
3192

    
3193
  if done:
3194
    lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
3195
  return not cumul_degraded
3196

    
3197

    
3198
def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
3199
  """Check that mirrors are not degraded.
3200

3201
  The ldisk parameter, if True, will change the test from the
3202
  is_degraded attribute (which represents overall non-ok status for
3203
  the device(s)) to the ldisk (representing the local storage status).
3204

3205
  """
3206
  lu.cfg.SetDiskID(dev, node)
3207

    
3208
  result = True
3209

    
3210
  if on_primary or dev.AssembleOnSecondary():
3211
    rstats = lu.rpc.call_blockdev_find(node, dev)
3212
    msg = rstats.fail_msg
3213
    if msg:
3214
      lu.LogWarning("Can't find disk on node %s: %s", node, msg)
3215
      result = False
3216
    elif not rstats.payload:
3217
      lu.LogWarning("Can't find disk on node %s", node)
3218
      result = False
3219
    else:
3220
      if ldisk:
3221
        result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
3222
      else:
3223
        result = result and not rstats.payload.is_degraded
3224

    
3225
  if dev.children:
3226
    for child in dev.children:
3227
      result = result and _CheckDiskConsistency(lu, child, node, on_primary)
3228

    
3229
  return result
3230

    
3231

    
3232
class LUOobCommand(NoHooksLU):
3233
  """Logical unit for OOB handling.
3234

3235
  """
3236
  REG_BGL = False
3237

    
3238
  def CheckPrereq(self):
3239
    """Check prerequisites.
3240

3241
    This checks:
3242
     - the node exists in the configuration
3243
     - OOB is supported
3244

3245
    Any errors are signaled by raising errors.OpPrereqError.
3246

3247
    """
3248
    self.nodes = []
3249
    for node_name in self.op.node_names:
3250
      node = self.cfg.GetNodeInfo(node_name)
3251

    
3252
      if node is None:
3253
        raise errors.OpPrereqError("Node %s not found" % node_name,
3254
                                   errors.ECODE_NOENT)
3255
      else:
3256
        self.nodes.append(node)
3257

    
3258
      if (self.op.command == constants.OOB_POWER_OFF and not node.offline):
3259
        raise errors.OpPrereqError(("Cannot power off node %s because it is"
3260
                                    " not marked offline") % node_name,
3261
                                   errors.ECODE_STATE)
3262

    
3263
  def ExpandNames(self):
3264
    """Gather locks we need.
3265

3266
    """
3267
    if self.op.node_names:
3268
      self.op.node_names = [_ExpandNodeName(self.cfg, name)
3269
                            for name in self.op.node_names]
3270
    else:
3271
      self.op.node_names = self.cfg.GetNodeList()
3272

    
3273
    self.needed_locks = {
3274
      locking.LEVEL_NODE: self.op.node_names,
3275
      }
3276

    
3277
  def Exec(self, feedback_fn):
3278
    """Execute OOB and return result if we expect any.
3279

3280
    """
3281
    master_node = self.cfg.GetMasterNode()
3282
    ret = []
3283

    
3284
    for node in self.nodes:
3285
      node_entry = [(constants.RS_NORMAL, node.name)]
3286
      ret.append(node_entry)
3287

    
3288
      oob_program = _SupportsOob(self.cfg, node)
3289

    
3290
      if not oob_program:
3291
        node_entry.append((constants.RS_UNAVAIL, None))
3292
        continue
3293

    
3294
      logging.info("Executing out-of-band command '%s' using '%s' on %s",
3295
                   self.op.command, oob_program, node.name)
3296
      result = self.rpc.call_run_oob(master_node, oob_program,
3297
                                     self.op.command, node.name,
3298
                                     self.op.timeout)
3299

    
3300
      if result.fail_msg:
3301
        self.LogWarning("On node '%s' out-of-band RPC failed with: %s",
3302
                        node.name, result.fail_msg)
3303
        node_entry.append((constants.RS_NODATA, None))
3304
      else:
3305
        try:
3306
          self._CheckPayload(result)
3307
        except errors.OpExecError, err:
3308
          self.LogWarning("The payload returned by '%s' is not valid: %s",
3309
                          node.name, err)
3310
          node_entry.append((constants.RS_NODATA, None))
3311
        else:
3312
          if self.op.command == constants.OOB_HEALTH:
3313
            # For health we should log important events
3314
            for item, status in result.payload:
3315
              if status in [constants.OOB_STATUS_WARNING,
3316
                            constants.OOB_STATUS_CRITICAL]:
3317
                self.LogWarning("On node '%s' item '%s' has status '%s'",
3318
                                node.name, item, status)
3319

    
3320
          if self.op.command == constants.OOB_POWER_ON:
3321
            node.powered = True
3322
          elif self.op.command == constants.OOB_POWER_OFF:
3323
            node.powered = False
3324
          elif self.op.command == constants.OOB_POWER_STATUS:
3325
            powered = result.payload[constants.OOB_POWER_STATUS_POWERED]
3326
            if powered != node.powered:
3327
              logging.warning(("Recorded power state (%s) of node '%s' does not"
3328
                               " match actual power state (%s)"), node.powered,
3329
                              node.name, powered)
3330

    
3331
          # For configuration changing commands we should update the node
3332
          if self.op.command in (constants.OOB_POWER_ON,
3333
                                 constants.OOB_POWER_OFF):
3334
            self.cfg.Update(node, feedback_fn)
3335

    
3336
          node_entry.append((constants.RS_NORMAL, result.payload))
3337

    
3338
    return ret
3339

    
3340
  def _CheckPayload(self, result):
3341
    """Checks if the payload is valid.
3342

3343
    @param result: RPC result
3344
    @raises errors.OpExecError: If payload is not valid
3345

3346
    """
3347
    errs = []
3348
    if self.op.command == constants.OOB_HEALTH:
3349
      if not isinstance(result.payload, list):
3350
        errs.append("command 'health' is expected to return a list but got %s" %
3351
                    type(result.payload))
3352
      else:
3353
        for item, status in result.payload:
3354
          if status not in constants.OOB_STATUSES:
3355
            errs.append("health item '%s' has invalid status '%s'" %
3356
                        (item, status))
3357

    
3358
    if self.op.command == constants.OOB_POWER_STATUS:
3359
      if not isinstance(result.payload, dict):
3360
        errs.append("power-status is expected to return a dict but got %s" %
3361
                    type(result.payload))
3362

    
3363
    if self.op.command in [
3364
        constants.OOB_POWER_ON,
3365
        constants.OOB_POWER_OFF,
3366
        constants.OOB_POWER_CYCLE,
3367
        ]:
3368
      if result.payload is not None:
3369
        errs.append("%s is expected to not return payload but got '%s'" %
3370
                    (self.op.command, result.payload))
3371

    
3372
    if errs:
3373
      raise errors.OpExecError("Check of out-of-band payload failed due to %s" %
3374
                               utils.CommaJoin(errs))
3375

    
3376

    
3377

    
3378
class LUOsDiagnose(NoHooksLU):
3379
  """Logical unit for OS diagnose/query.
3380

3381
  """
3382
  REQ_BGL = False
3383
  _HID = "hidden"
3384
  _BLK = "blacklisted"
3385
  _VLD = "valid"
3386
  _FIELDS_STATIC = utils.FieldSet()
3387
  _FIELDS_DYNAMIC = utils.FieldSet("name", _VLD, "node_status", "variants",
3388
                                   "parameters", "api_versions", _HID, _BLK)
3389

    
3390
  def CheckArguments(self):
3391
    if self.op.names:
3392
      raise errors.OpPrereqError("Selective OS query not supported",
3393
                                 errors.ECODE_INVAL)
3394

    
3395
    _CheckOutputFields(static=self._FIELDS_STATIC,
3396
                       dynamic=self._FIELDS_DYNAMIC,
3397
                       selected=self.op.output_fields)
3398

    
3399
  def ExpandNames(self):
3400
    # Lock all nodes, in shared mode
3401
    # Temporary removal of locks, should be reverted later
3402
    # TODO: reintroduce locks when they are lighter-weight
3403
    self.needed_locks = {}
3404
    #self.share_locks[locking.LEVEL_NODE] = 1
3405
    #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3406

    
3407
  @staticmethod
3408
  def _DiagnoseByOS(rlist):
3409
    """Remaps a per-node return list into an a per-os per-node dictionary
3410

3411
    @param rlist: a map with node names as keys and OS objects as values
3412

3413
    @rtype: dict
3414
    @return: a dictionary with osnames as keys and as value another
3415
        map, with nodes as keys and tuples of (path, status, diagnose,
3416
        variants, parameters, api_versions) as values, eg::
3417

3418
          {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
3419
                                     (/srv/..., False, "invalid api")],
3420
                           "node2": [(/srv/..., True, "", [], [])]}
3421
          }
3422

3423
    """
3424
    all_os = {}
3425
    # we build here the list of nodes that didn't fail the RPC (at RPC
3426
    # level), so that nodes with a non-responding node daemon don't
3427
    # make all OSes invalid
3428
    good_nodes = [node_name for node_name in rlist
3429
                  if not rlist[node_name].fail_msg]
3430
    for node_name, nr in rlist.items():
3431
      if nr.fail_msg or not nr.payload:
3432
        continue
3433
      for (name, path, status, diagnose, variants,
3434
           params, api_versions) in nr.payload:
3435
        if name not in all_os:
3436
          # build a list of nodes for this os containing empty lists
3437
          # for each node in node_list
3438
          all_os[name] = {}
3439
          for nname in good_nodes:
3440
            all_os[name][nname] = []
3441
        # convert params from [name, help] to (name, help)
3442
        params = [tuple(v) for v in params]
3443
        all_os[name][node_name].append((path, status, diagnose,
3444
                                        variants, params, api_versions))
3445
    return all_os
3446

    
3447
  def Exec(self, feedback_fn):
3448
    """Compute the list of OSes.
3449

3450
    """
3451
    valid_nodes = [node.name
3452
                   for node in self.cfg.GetAllNodesInfo().values()
3453
                   if not node.offline and node.vm_capable]
3454
    node_data = self.rpc.call_os_diagnose(valid_nodes)
3455
    pol = self._DiagnoseByOS(node_data)
3456
    output = []
3457
    cluster = self.cfg.GetClusterInfo()
3458

    
3459
    for os_name in utils.NiceSort(pol.keys()):
3460
      os_data = pol[os_name]
3461
      row = []
3462
      valid = True
3463
      (variants, params, api_versions) = null_state = (set(), set(), set())
3464
      for idx, osl in enumerate(os_data.values()):
3465
        valid = bool(valid and osl and osl[0][1])
3466
        if not valid:
3467
          (variants, params, api_versions) = null_state
3468
          break
3469
        node_variants, node_params, node_api = osl[0][3:6]
3470
        if idx == 0: # first entry
3471
          variants = set(node_variants)
3472
          params = set(node_params)
3473
          api_versions = set(node_api)
3474
        else: # keep consistency
3475
          variants.intersection_update(node_variants)
3476
          params.intersection_update(node_params)
3477
          api_versions.intersection_update(node_api)
3478

    
3479
      is_hid = os_name in cluster.hidden_os
3480
      is_blk = os_name in cluster.blacklisted_os
3481
      if ((self._HID not in self.op.output_fields and is_hid) or
3482
          (self._BLK not in self.op.output_fields and is_blk) or
3483
          (self._VLD not in self.op.output_fields and not valid)):
3484
        continue
3485

    
3486
      for field in self.op.output_fields:
3487
        if field == "name":
3488
          val = os_name
3489
        elif field == self._VLD:
3490
          val = valid
3491
        elif field == "node_status":
3492
          # this is just a copy of the dict
3493
          val = {}
3494
          for node_name, nos_list in os_data.items():
3495
            val[node_name] = nos_list
3496
        elif field == "variants":
3497
          val = utils.NiceSort(list(variants))
3498
        elif field == "parameters":
3499
          val = list(params)
3500
        elif field == "api_versions":
3501
          val = list(api_versions)
3502
        elif field == self._HID:
3503
          val = is_hid
3504
        elif field == self._BLK:
3505
          val = is_blk
3506
        else:
3507
          raise errors.ParameterError(field)
3508
        row.append(val)
3509
      output.append(row)
3510

    
3511
    return output
3512

    
3513

    
3514
class LUNodeRemove(LogicalUnit):
3515
  """Logical unit for removing a node.
3516

3517
  """
3518
  HPATH = "node-remove"
3519
  HTYPE = constants.HTYPE_NODE
3520

    
3521
  def BuildHooksEnv(self):
3522
    """Build hooks env.
3523

3524
    This doesn't run on the target node in the pre phase as a failed
3525
    node would then be impossible to remove.
3526

3527
    """
3528
    env = {
3529
      "OP_TARGET": self.op.node_name,
3530
      "NODE_NAME": self.op.node_name,
3531
      }
3532
    all_nodes = self.cfg.GetNodeList()
3533
    try:
3534
      all_nodes.remove(self.op.node_name)
3535
    except ValueError:
3536
      logging.warning("Node %s which is about to be removed not found"
3537
                      " in the all nodes list", self.op.node_name)
3538
    return env, all_nodes, all_nodes
3539

    
3540
  def CheckPrereq(self):
3541
    """Check prerequisites.
3542

3543
    This checks:
3544
     - the node exists in the configuration
3545
     - it does not have primary or secondary instances
3546
     - it's not the master
3547

3548
    Any errors are signaled by raising errors.OpPrereqError.
3549

3550
    """
3551
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3552
    node = self.cfg.GetNodeInfo(self.op.node_name)
3553
    assert node is not None
3554

    
3555
    instance_list = self.cfg.GetInstanceList()
3556

    
3557
    masternode = self.cfg.GetMasterNode()
3558
    if node.name == masternode:
3559
      raise errors.OpPrereqError("Node is the master node,"
3560
                                 " you need to failover first.",
3561
                                 errors.ECODE_INVAL)
3562

    
3563
    for instance_name in instance_list:
3564
      instance = self.cfg.GetInstanceInfo(instance_name)
3565
      if node.name in instance.all_nodes:
3566
        raise errors.OpPrereqError("Instance %s is still running on the node,"
3567
                                   " please remove first." % instance_name,
3568
                                   errors.ECODE_INVAL)
3569
    self.op.node_name = node.name
3570
    self.node = node
3571

    
3572
  def Exec(self, feedback_fn):
3573
    """Removes the node from the cluster.
3574

3575
    """
3576
    node = self.node
3577
    logging.info("Stopping the node daemon and removing configs from node %s",
3578
                 node.name)
3579

    
3580
    modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
3581

    
3582
    # Promote nodes to master candidate as needed
3583
    _AdjustCandidatePool(self, exceptions=[node.name])
3584
    self.context.RemoveNode(node.name)
3585

    
3586
    # Run post hooks on the node before it's removed
3587
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
3588
    try:
3589
      hm.RunPhase(constants.HOOKS_PHASE_POST, [node.name])
3590
    except:
3591
      # pylint: disable-msg=W0702
3592
      self.LogWarning("Errors occurred running hooks on %s" % node.name)
3593

    
3594
    result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
3595
    msg = result.fail_msg
3596
    if msg:
3597
      self.LogWarning("Errors encountered on the remote node while leaving"
3598
                      " the cluster: %s", msg)
3599

    
3600
    # Remove node from our /etc/hosts
3601
    if self.cfg.GetClusterInfo().modify_etc_hosts:
3602
      master_node = self.cfg.GetMasterNode()
3603
      result = self.rpc.call_etc_hosts_modify(master_node,
3604
                                              constants.ETC_HOSTS_REMOVE,
3605
                                              node.name, None)
3606
      result.Raise("Can't update hosts file with new host data")
3607
      _RedistributeAncillaryFiles(self)
3608

    
3609

    
3610
class _NodeQuery(_QueryBase):
3611
  FIELDS = query.NODE_FIELDS
3612

    
3613
  def ExpandNames(self, lu):
3614
    lu.needed_locks = {}
3615
    lu.share_locks[locking.LEVEL_NODE] = 1
3616

    
3617
    if self.names:
3618
      self.wanted = _GetWantedNodes(lu, self.names)
3619
    else:
3620
      self.wanted = locking.ALL_SET
3621

    
3622
    self.do_locking = (self.use_locking and
3623
                       query.NQ_LIVE in self.requested_data)
3624

    
3625
    if self.do_locking:
3626
      # if we don't request only static fields, we need to lock the nodes
3627
      lu.needed_locks[locking.LEVEL_NODE] = self.wanted
3628

    
3629
  def DeclareLocks(self, lu, level):
3630
    pass
3631

    
3632
  def _GetQueryData(self, lu):
3633
    """Computes the list of nodes and their attributes.
3634

3635
    """
3636
    all_info = lu.cfg.GetAllNodesInfo()
3637

    
3638
    nodenames = self._GetNames(lu, all_info.keys(), locking.LEVEL_NODE)
3639

    
3640
    # Gather data as requested
3641
    if query.NQ_LIVE in self.requested_data:
3642
      # filter out non-vm_capable nodes
3643
      toquery_nodes = [name for name in nodenames if all_info[name].vm_capable]
3644

    
3645
      node_data = lu.rpc.call_node_info(toquery_nodes, lu.cfg.GetVGName(),
3646
                                        lu.cfg.GetHypervisorType())
3647
      live_data = dict((name, nresult.payload)
3648
                       for (name, nresult) in node_data.items()
3649
                       if not nresult.fail_msg and nresult.payload)
3650
    else:
3651
      live_data = None
3652

    
3653
    if query.NQ_INST in self.requested_data:
3654
      node_to_primary = dict([(name, set()) for name in nodenames])
3655
      node_to_secondary = dict([(name, set()) for name in nodenames])
3656

    
3657
      inst_data = lu.cfg.GetAllInstancesInfo()
3658

    
3659
      for inst in inst_data.values():
3660
        if inst.primary_node in node_to_primary:
3661
          node_to_primary[inst.primary_node].add(inst.name)
3662
        for secnode in inst.secondary_nodes:
3663
          if secnode in node_to_secondary:
3664
            node_to_secondary[secnode].add(inst.name)
3665
    else:
3666
      node_to_primary = None
3667
      node_to_secondary = None
3668

    
3669
    if query.NQ_OOB in self.requested_data:
3670
      oob_support = dict((name, bool(_SupportsOob(lu.cfg, node)))
3671
                         for name, node in all_info.iteritems())
3672
    else:
3673
      oob_support = None
3674

    
3675
    if query.NQ_GROUP in self.requested_data:
3676
      groups = lu.cfg.GetAllNodeGroupsInfo()
3677
    else:
3678
      groups = {}
3679

    
3680
    return query.NodeQueryData([all_info[name] for name in nodenames],
3681
                               live_data, lu.cfg.GetMasterNode(),
3682
                               node_to_primary, node_to_secondary, groups,
3683
                               oob_support, lu.cfg.GetClusterInfo())
3684

    
3685

    
3686
class LUNodeQuery(NoHooksLU):
3687
  """Logical unit for querying nodes.
3688

3689
  """
3690
  # pylint: disable-msg=W0142
3691
  REQ_BGL = False
3692

    
3693
  def CheckArguments(self):
3694
    self.nq = _NodeQuery(self.op.names, self.op.output_fields,
3695
                         self.op.use_locking)
3696

    
3697
  def ExpandNames(self):
3698
    self.nq.ExpandNames(self)
3699

    
3700
  def Exec(self, feedback_fn):
3701
    return self.nq.OldStyleQuery(self)
3702

    
3703

    
3704
class LUNodeQueryvols(NoHooksLU):
3705
  """Logical unit for getting volumes on node(s).
3706

3707
  """
3708
  REQ_BGL = False
3709
  _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
3710
  _FIELDS_STATIC = utils.FieldSet("node")
3711

    
3712
  def CheckArguments(self):
3713
    _CheckOutputFields(static=self._FIELDS_STATIC,
3714
                       dynamic=self._FIELDS_DYNAMIC,
3715
                       selected=self.op.output_fields)
3716

    
3717
  def ExpandNames(self):
3718
    self.needed_locks = {}
3719
    self.share_locks[locking.LEVEL_NODE] = 1
3720
    if not self.op.nodes:
3721
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3722
    else:
3723
      self.needed_locks[locking.LEVEL_NODE] = \
3724
        _GetWantedNodes(self, self.op.nodes)
3725

    
3726
  def Exec(self, feedback_fn):
3727
    """Computes the list of nodes and their attributes.
3728

3729
    """
3730
    nodenames = self.acquired_locks[locking.LEVEL_NODE]
3731
    volumes = self.rpc.call_node_volumes(nodenames)
3732

    
3733
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
3734
             in self.cfg.GetInstanceList()]
3735

    
3736
    lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
3737

    
3738
    output = []
3739
    for node in nodenames:
3740
      nresult = volumes[node]
3741
      if nresult.offline:
3742
        continue
3743
      msg = nresult.fail_msg
3744
      if msg:
3745
        self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
3746
        continue
3747

    
3748
      node_vols = nresult.payload[:]
3749
      node_vols.sort(key=lambda vol: vol['dev'])
3750

    
3751
      for vol in node_vols:
3752
        node_output = []
3753
        for field in self.op.output_fields:
3754
          if field == "node":
3755
            val = node
3756
          elif field == "phys":
3757
            val = vol['dev']
3758
          elif field == "vg":
3759
            val = vol['vg']
3760
          elif field == "name":
3761
            val = vol['name']
3762
          elif field == "size":
3763
            val = int(float(vol['size']))
3764
          elif field == "instance":
3765
            for inst in ilist:
3766
              if node not in lv_by_node[inst]:
3767
                continue
3768
              if vol['name'] in lv_by_node[inst][node]:
3769
                val = inst.name
3770
                break
3771
            else:
3772
              val = '-'
3773
          else:
3774
            raise errors.ParameterError(field)
3775
          node_output.append(str(val))
3776

    
3777
        output.append(node_output)
3778

    
3779
    return output
3780

    
3781

    
3782
class LUNodeQueryStorage(NoHooksLU):
3783
  """Logical unit for getting information on storage units on node(s).
3784

3785
  """
3786
  _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
3787
  REQ_BGL = False
3788

    
3789
  def CheckArguments(self):
3790
    _CheckOutputFields(static=self._FIELDS_STATIC,
3791
                       dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
3792
                       selected=self.op.output_fields)
3793

    
3794
  def ExpandNames(self):
3795
    self.needed_locks = {}
3796
    self.share_locks[locking.LEVEL_NODE] = 1
3797

    
3798
    if self.op.nodes:
3799
      self.needed_locks[locking.LEVEL_NODE] = \
3800
        _GetWantedNodes(self, self.op.nodes)
3801
    else:
3802
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3803

    
3804
  def Exec(self, feedback_fn):
3805
    """Computes the list of nodes and their attributes.
3806

3807
    """
3808
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
3809

    
3810
    # Always get name to sort by
3811
    if constants.SF_NAME in self.op.output_fields:
3812
      fields = self.op.output_fields[:]
3813
    else:
3814
      fields = [constants.SF_NAME] + self.op.output_fields
3815

    
3816
    # Never ask for node or type as it's only known to the LU
3817
    for extra in [constants.SF_NODE, constants.SF_TYPE]:
3818
      while extra in fields:
3819
        fields.remove(extra)
3820

    
3821
    field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
3822
    name_idx = field_idx[constants.SF_NAME]
3823

    
3824
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3825
    data = self.rpc.call_storage_list(self.nodes,
3826
                                      self.op.storage_type, st_args,
3827
                                      self.op.name, fields)
3828

    
3829
    result = []
3830

    
3831
    for node in utils.NiceSort(self.nodes):
3832
      nresult = data[node]
3833
      if nresult.offline:
3834
        continue
3835

    
3836
      msg = nresult.fail_msg
3837
      if msg:
3838
        self.LogWarning("Can't get storage data from node %s: %s", node, msg)
3839
        continue
3840

    
3841
      rows = dict([(row[name_idx], row) for row in nresult.payload])
3842

    
3843
      for name in utils.NiceSort(rows.keys()):
3844
        row = rows[name]
3845

    
3846
        out = []
3847

    
3848
        for field in self.op.output_fields:
3849
          if field == constants.SF_NODE:
3850
            val = node
3851
          elif field == constants.SF_TYPE:
3852
            val = self.op.storage_type
3853
          elif field in field_idx:
3854
            val = row[field_idx[field]]
3855
          else:
3856
            raise errors.ParameterError(field)
3857

    
3858
          out.append(val)
3859

    
3860
        result.append(out)
3861

    
3862
    return result
3863

    
3864

    
3865
class _InstanceQuery(_QueryBase):
3866
  FIELDS = query.INSTANCE_FIELDS
3867

    
3868
  def ExpandNames(self, lu):
3869
    lu.needed_locks = {}
3870
    lu.share_locks[locking.LEVEL_INSTANCE] = 1
3871
    lu.share_locks[locking.LEVEL_NODE] = 1
3872

    
3873
    if self.names:
3874
      self.wanted = _GetWantedInstances(lu, self.names)
3875
    else:
3876
      self.wanted = locking.ALL_SET
3877

    
3878
    self.do_locking = (self.use_locking and
3879
                       query.IQ_LIVE in self.requested_data)
3880
    if self.do_locking:
3881
      lu.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
3882
      lu.needed_locks[locking.LEVEL_NODE] = []
3883
      lu.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3884

    
3885
  def DeclareLocks(self, lu, level):
3886
    if level == locking.LEVEL_NODE and self.do_locking:
3887
      lu._LockInstancesNodes() # pylint: disable-msg=W0212
3888

    
3889
  def _GetQueryData(self, lu):
3890
    """Computes the list of instances and their attributes.
3891

3892
    """
3893
    cluster = lu.cfg.GetClusterInfo()
3894
    all_info = lu.cfg.GetAllInstancesInfo()
3895

    
3896
    instance_names = self._GetNames(lu, all_info.keys(), locking.LEVEL_INSTANCE)
3897

    
3898
    instance_list = [all_info[name] for name in instance_names]
3899
    nodes = frozenset(itertools.chain(*(inst.all_nodes
3900
                                        for inst in instance_list)))
3901
    hv_list = list(set([inst.hypervisor for inst in instance_list]))
3902
    bad_nodes = []
3903
    offline_nodes = []
3904
    wrongnode_inst = set()
3905

    
3906
    # Gather data as requested
3907
    if self.requested_data & set([query.IQ_LIVE, query.IQ_CONSOLE]):
3908
      live_data = {}
3909
      node_data = lu.rpc.call_all_instances_info(nodes, hv_list)
3910
      for name in nodes:
3911
        result = node_data[name]
3912
        if result.offline:
3913
          # offline nodes will be in both lists
3914
          assert result.fail_msg
3915
          offline_nodes.append(name)
3916
        if result.fail_msg:
3917
          bad_nodes.append(name)
3918
        elif result.payload:
3919
          for inst in result.payload:
3920
            if inst in all_info:
3921
              if all_info[inst].primary_node == name:
3922
                live_data.update(result.payload)
3923
              else:
3924
                wrongnode_inst.add(inst)
3925
            else:
3926
              # orphan instance; we don't list it here as we don't
3927
              # handle this case yet in the output of instance listing
3928
              logging.warning("Orphan instance '%s' found on node %s",
3929
                              inst, name)
3930
        # else no instance is alive
3931
    else:
3932
      live_data = {}
3933

    
3934
    if query.IQ_DISKUSAGE in self.requested_data:
3935
      disk_usage = dict((inst.name,
3936
                         _ComputeDiskSize(inst.disk_template,
3937
                                          [{"size": disk.size}
3938
                                           for disk in inst.disks]))
3939
                        for inst in instance_list)
3940
    else:
3941
      disk_usage = None
3942

    
3943
    if query.IQ_CONSOLE in self.requested_data:
3944
      consinfo = {}
3945
      for inst in instance_list:
3946
        if inst.name in live_data:
3947
          # Instance is running
3948
          consinfo[inst.name] = _GetInstanceConsole(cluster, inst)
3949
        else:
3950
          consinfo[inst.name] = None
3951
      assert set(consinfo.keys()) == set(instance_names)
3952
    else:
3953
      consinfo = None
3954

    
3955
    return query.InstanceQueryData(instance_list, lu.cfg.GetClusterInfo(),
3956
                                   disk_usage, offline_nodes, bad_nodes,
3957
                                   live_data, wrongnode_inst, consinfo)
3958

    
3959

    
3960
class LUQuery(NoHooksLU):
3961
  """Query for resources/items of a certain kind.
3962

3963
  """
3964
  # pylint: disable-msg=W0142
3965
  REQ_BGL = False
3966

    
3967
  def CheckArguments(self):
3968
    qcls = _GetQueryImplementation(self.op.what)
3969
    names = qlang.ReadSimpleFilter("name", self.op.filter)
3970

    
3971
    self.impl = qcls(names, self.op.fields, False)
3972

    
3973
  def ExpandNames(self):
3974
    self.impl.ExpandNames(self)
3975

    
3976
  def DeclareLocks(self, level):
3977
    self.impl.DeclareLocks(self, level)
3978

    
3979
  def Exec(self, feedback_fn):
3980
    return self.impl.NewStyleQuery(self)
3981

    
3982

    
3983
class LUQueryFields(NoHooksLU):
3984
  """Query for resources/items of a certain kind.
3985

3986
  """
3987
  # pylint: disable-msg=W0142
3988
  REQ_BGL = False
3989

    
3990
  def CheckArguments(self):
3991
    self.qcls = _GetQueryImplementation(self.op.what)
3992

    
3993
  def ExpandNames(self):
3994
    self.needed_locks = {}
3995

    
3996
  def Exec(self, feedback_fn):
3997
    return self.qcls.FieldsQuery(self.op.fields)
3998

    
3999

    
4000
class LUNodeModifyStorage(NoHooksLU):
4001
  """Logical unit for modifying a storage volume on a node.
4002

4003
  """
4004
  REQ_BGL = False
4005

    
4006
  def CheckArguments(self):
4007
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4008

    
4009
    storage_type = self.op.storage_type
4010

    
4011
    try:
4012
      modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
4013
    except KeyError:
4014
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
4015
                                 " modified" % storage_type,
4016
                                 errors.ECODE_INVAL)
4017

    
4018
    diff = set(self.op.changes.keys()) - modifiable
4019
    if diff:
4020
      raise errors.OpPrereqError("The following fields can not be modified for"
4021
                                 " storage units of type '%s': %r" %
4022
                                 (storage_type, list(diff)),
4023
                                 errors.ECODE_INVAL)
4024

    
4025
  def ExpandNames(self):
4026
    self.needed_locks = {
4027
      locking.LEVEL_NODE: self.op.node_name,
4028
      }
4029

    
4030
  def Exec(self, feedback_fn):
4031
    """Computes the list of nodes and their attributes.
4032

4033
    """
4034
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
4035
    result = self.rpc.call_storage_modify(self.op.node_name,
4036
                                          self.op.storage_type, st_args,
4037
                                          self.op.name, self.op.changes)
4038
    result.Raise("Failed to modify storage unit '%s' on %s" %
4039
                 (self.op.name, self.op.node_name))
4040

    
4041

    
4042
class LUNodeAdd(LogicalUnit):
4043
  """Logical unit for adding node to the cluster.
4044

4045
  """
4046
  HPATH = "node-add"
4047
  HTYPE = constants.HTYPE_NODE
4048
  _NFLAGS = ["master_capable", "vm_capable"]
4049

    
4050
  def CheckArguments(self):
4051
    self.primary_ip_family = self.cfg.GetPrimaryIPFamily()
4052
    # validate/normalize the node name
4053
    self.hostname = netutils.GetHostname(name=self.op.node_name,
4054
                                         family=self.primary_ip_family)
4055
    self.op.node_name = self.hostname.name
4056
    if self.op.readd and self.op.group:
4057
      raise errors.OpPrereqError("Cannot pass a node group when a node is"
4058
                                 " being readded", errors.ECODE_INVAL)
4059

    
4060
  def BuildHooksEnv(self):
4061
    """Build hooks env.
4062

4063
    This will run on all nodes before, and on all nodes + the new node after.
4064

4065
    """
4066
    env = {
4067
      "OP_TARGET": self.op.node_name,
4068
      "NODE_NAME": self.op.node_name,
4069
      "NODE_PIP": self.op.primary_ip,
4070
      "NODE_SIP": self.op.secondary_ip,
4071
      "MASTER_CAPABLE": str(self.op.master_capable),
4072
      "VM_CAPABLE": str(self.op.vm_capable),
4073
      }
4074
    nodes_0 = self.cfg.GetNodeList()
4075
    nodes_1 = nodes_0 + [self.op.node_name, ]
4076
    return env, nodes_0, nodes_1
4077

    
4078
  def CheckPrereq(self):
4079
    """Check prerequisites.
4080

4081
    This checks:
4082
     - the new node is not already in the config
4083
     - it is resolvable
4084
     - its parameters (single/dual homed) matches the cluster
4085

4086
    Any errors are signaled by raising errors.OpPrereqError.
4087

4088
    """
4089
    cfg = self.cfg
4090
    hostname = self.hostname
4091
    node = hostname.name
4092
    primary_ip = self.op.primary_ip = hostname.ip
4093
    if self.op.secondary_ip is None:
4094
      if self.primary_ip_family == netutils.IP6Address.family:
4095
        raise errors.OpPrereqError("When using a IPv6 primary address, a valid"
4096
                                   " IPv4 address must be given as secondary",
4097
                                   errors.ECODE_INVAL)
4098
      self.op.secondary_ip = primary_ip
4099

    
4100
    secondary_ip = self.op.secondary_ip
4101
    if not netutils.IP4Address.IsValid(secondary_ip):
4102
      raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
4103
                                 " address" % secondary_ip, errors.ECODE_INVAL)
4104

    
4105
    node_list = cfg.GetNodeList()
4106
    if not self.op.readd and node in node_list:
4107
      raise errors.OpPrereqError("Node %s is already in the configuration" %
4108
                                 node, errors.ECODE_EXISTS)
4109
    elif self.op.readd and node not in node_list:
4110
      raise errors.OpPrereqError("Node %s is not in the configuration" % node,
4111
                                 errors.ECODE_NOENT)
4112

    
4113
    self.changed_primary_ip = False
4114

    
4115
    for existing_node_name in node_list:
4116
      existing_node = cfg.GetNodeInfo(existing_node_name)
4117

    
4118
      if self.op.readd and node == existing_node_name:
4119
        if existing_node.secondary_ip != secondary_ip:
4120
          raise errors.OpPrereqError("Readded node doesn't have the same IP"
4121
                                     " address configuration as before",
4122
                                     errors.ECODE_INVAL)
4123
        if existing_node.primary_ip != primary_ip:
4124
          self.changed_primary_ip = True
4125

    
4126
        continue
4127

    
4128
      if (existing_node.primary_ip == primary_ip or
4129
          existing_node.secondary_ip == primary_ip or
4130
          existing_node.primary_ip == secondary_ip or
4131
          existing_node.secondary_ip == secondary_ip):
4132
        raise errors.OpPrereqError("New node ip address(es) conflict with"
4133
                                   " existing node %s" % existing_node.name,
4134
                                   errors.ECODE_NOTUNIQUE)
4135

    
4136
    # After this 'if' block, None is no longer a valid value for the
4137
    # _capable op attributes
4138
    if self.op.readd:
4139
      old_node = self.cfg.GetNodeInfo(node)
4140
      assert old_node is not None, "Can't retrieve locked node %s" % node
4141
      for attr in self._NFLAGS:
4142
        if getattr(self.op, attr) is None:
4143
          setattr(self.op, attr, getattr(old_node, attr))
4144
    else:
4145
      for attr in self._NFLAGS:
4146
        if getattr(self.op, attr) is None:
4147
          setattr(self.op, attr, True)
4148

    
4149
    if self.op.readd and not self.op.vm_capable:
4150
      pri, sec = cfg.GetNodeInstances(node)
4151
      if pri or sec:
4152
        raise errors.OpPrereqError("Node %s being re-added with vm_capable"
4153
                                   " flag set to false, but it already holds"
4154
                                   " instances" % node,
4155
                                   errors.ECODE_STATE)
4156

    
4157
    # check that the type of the node (single versus dual homed) is the
4158
    # same as for the master
4159
    myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
4160
    master_singlehomed = myself.secondary_ip == myself.primary_ip
4161
    newbie_singlehomed = secondary_ip == primary_ip
4162
    if master_singlehomed != newbie_singlehomed:
4163
      if master_singlehomed:
4164
        raise errors.OpPrereqError("The master has no secondary ip but the"
4165
                                   " new node has one",
4166
                                   errors.ECODE_INVAL)
4167
      else:
4168
        raise errors.OpPrereqError("The master has a secondary ip but the"
4169
                                   " new node doesn't have one",
4170
                                   errors.ECODE_INVAL)
4171

    
4172
    # checks reachability
4173
    if not netutils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
4174
      raise errors.OpPrereqError("Node not reachable by ping",
4175
                                 errors.ECODE_ENVIRON)
4176

    
4177
    if not newbie_singlehomed:
4178
      # check reachability from my secondary ip to newbie's secondary ip
4179
      if not netutils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
4180
                           source=myself.secondary_ip):
4181
        raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
4182
                                   " based ping to node daemon port",
4183
                                   errors.ECODE_ENVIRON)
4184

    
4185
    if self.op.readd:
4186
      exceptions = [node]
4187
    else:
4188
      exceptions = []
4189

    
4190
    if self.op.master_capable:
4191
      self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
4192
    else:
4193
      self.master_candidate = False
4194

    
4195
    if self.op.readd:
4196
      self.new_node = old_node
4197
    else:
4198
      node_group = cfg.LookupNodeGroup(self.op.group)
4199
      self.new_node = objects.Node(name=node,
4200
                                   primary_ip=primary_ip,
4201
                                   secondary_ip=secondary_ip,
4202
                                   master_candidate=self.master_candidate,
4203
                                   offline=False, drained=False,
4204
                                   group=node_group)
4205

    
4206
    if self.op.ndparams:
4207
      utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
4208

    
4209
  def Exec(self, feedback_fn):
4210
    """Adds the new node to the cluster.
4211

4212
    """
4213
    new_node = self.new_node
4214
    node = new_node.name
4215

    
4216
    # We adding a new node so we assume it's powered
4217
    new_node.powered = True
4218

    
4219
    # for re-adds, reset the offline/drained/master-candidate flags;
4220
    # we need to reset here, otherwise offline would prevent RPC calls
4221
    # later in the procedure; this also means that if the re-add
4222
    # fails, we are left with a non-offlined, broken node
4223
    if self.op.readd:
4224
      new_node.drained = new_node.offline = False # pylint: disable-msg=W0201
4225
      self.LogInfo("Readding a node, the offline/drained flags were reset")
4226
      # if we demote the node, we do cleanup later in the procedure
4227
      new_node.master_candidate = self.master_candidate
4228
      if self.changed_primary_ip:
4229
        new_node.primary_ip = self.op.primary_ip
4230

    
4231
    # copy the master/vm_capable flags
4232
    for attr in self._NFLAGS:
4233
      setattr(new_node, attr, getattr(self.op, attr))
4234

    
4235
    # notify the user about any possible mc promotion
4236
    if new_node.master_candidate:
4237
      self.LogInfo("Node will be a master candidate")
4238

    
4239
    if self.op.ndparams:
4240
      new_node.ndparams = self.op.ndparams
4241
    else:
4242
      new_node.ndparams = {}
4243

    
4244
    # check connectivity
4245
    result = self.rpc.call_version([node])[node]
4246
    result.Raise("Can't get version information from node %s" % node)
4247
    if constants.PROTOCOL_VERSION == result.payload:
4248
      logging.info("Communication to node %s fine, sw version %s match",
4249
                   node, result.payload)
4250
    else:
4251
      raise errors.OpExecError("Version mismatch master version %s,"
4252
                               " node version %s" %
4253
                               (constants.PROTOCOL_VERSION, result.payload))
4254

    
4255
    # Add node to our /etc/hosts, and add key to known_hosts
4256
    if self.cfg.GetClusterInfo().modify_etc_hosts:
4257
      master_node = self.cfg.GetMasterNode()
4258
      result = self.rpc.call_etc_hosts_modify(master_node,
4259
                                              constants.ETC_HOSTS_ADD,
4260
                                              self.hostname.name,
4261
                                              self.hostname.ip)
4262
      result.Raise("Can't update hosts file with new host data")
4263

    
4264
    if new_node.secondary_ip != new_node.primary_ip:
4265
      _CheckNodeHasSecondaryIP(self, new_node.name, new_node.secondary_ip,
4266
                               False)
4267

    
4268
    node_verify_list = [self.cfg.GetMasterNode()]
4269
    node_verify_param = {
4270
      constants.NV_NODELIST: [node],
4271
      # TODO: do a node-net-test as well?
4272
    }
4273

    
4274
    result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
4275
                                       self.cfg.GetClusterName())
4276
    for verifier in node_verify_list:
4277
      result[verifier].Raise("Cannot communicate with node %s" % verifier)
4278
      nl_payload = result[verifier].payload[constants.NV_NODELIST]
4279
      if nl_payload:
4280
        for failed in nl_payload:
4281
          feedback_fn("ssh/hostname verification failed"
4282
                      " (checking from %s): %s" %
4283
                      (verifier, nl_payload[failed]))
4284
        raise errors.OpExecError("ssh/hostname verification failed.")
4285

    
4286
    if self.op.readd:
4287
      _RedistributeAncillaryFiles(self)
4288
      self.context.ReaddNode(new_node)
4289
      # make sure we redistribute the config
4290
      self.cfg.Update(new_node, feedback_fn)
4291
      # and make sure the new node will not have old files around
4292
      if not new_node.master_candidate:
4293
        result = self.rpc.call_node_demote_from_mc(new_node.name)
4294
        msg = result.fail_msg
4295
        if msg:
4296
          self.LogWarning("Node failed to demote itself from master"
4297
                          " candidate status: %s" % msg)
4298
    else:
4299
      _RedistributeAncillaryFiles(self, additional_nodes=[node],
4300
                                  additional_vm=self.op.vm_capable)
4301
      self.context.AddNode(new_node, self.proc.GetECId())
4302

    
4303

    
4304
class LUNodeSetParams(LogicalUnit):
4305
  """Modifies the parameters of a node.
4306

4307
  @cvar _F2R: a dictionary from tuples of flags (mc, drained, offline)
4308
      to the node role (as _ROLE_*)
4309
  @cvar _R2F: a dictionary from node role to tuples of flags
4310
  @cvar _FLAGS: a list of attribute names corresponding to the flags
4311

4312
  """
4313
  HPATH = "node-modify"
4314
  HTYPE = constants.HTYPE_NODE
4315
  REQ_BGL = False
4316
  (_ROLE_CANDIDATE, _ROLE_DRAINED, _ROLE_OFFLINE, _ROLE_REGULAR) = range(4)
4317
  _F2R = {
4318
    (True, False, False): _ROLE_CANDIDATE,
4319
    (False, True, False): _ROLE_DRAINED,
4320
    (False, False, True): _ROLE_OFFLINE,
4321
    (False, False, False): _ROLE_REGULAR,
4322
    }
4323
  _R2F = dict((v, k) for k, v in _F2R.items())
4324
  _FLAGS = ["master_candidate", "drained", "offline"]
4325

    
4326
  def CheckArguments(self):
4327
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4328
    all_mods = [self.op.offline, self.op.master_candidate, self.op.drained,
4329
                self.op.master_capable, self.op.vm_capable,
4330
                self.op.secondary_ip, self.op.ndparams]
4331
    if all_mods.count(None) == len(all_mods):
4332
      raise errors.OpPrereqError("Please pass at least one modification",
4333
                                 errors.ECODE_INVAL)
4334
    if all_mods.count(True) > 1:
4335
      raise errors.OpPrereqError("Can't set the node into more than one"
4336
                                 " state at the same time",
4337
                                 errors.ECODE_INVAL)
4338

    
4339
    # Boolean value that tells us whether we might be demoting from MC
4340
    self.might_demote = (self.op.master_candidate == False or
4341
                         self.op.offline == True or
4342
                         self.op.drained == True or
4343
                         self.op.master_capable == False)
4344

    
4345
    if self.op.secondary_ip:
4346
      if not netutils.IP4Address.IsValid(self.op.secondary_ip):
4347
        raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
4348
                                   " address" % self.op.secondary_ip,
4349
                                   errors.ECODE_INVAL)
4350

    
4351
    self.lock_all = self.op.auto_promote and self.might_demote
4352
    self.lock_instances = self.op.secondary_ip is not None
4353

    
4354
  def ExpandNames(self):
4355
    if self.lock_all:
4356
      self.needed_locks = {locking.LEVEL_NODE: locking.ALL_SET}
4357
    else:
4358
      self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
4359

    
4360
    if self.lock_instances:
4361
      self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
4362

    
4363
  def DeclareLocks(self, level):
4364
    # If we have locked all instances, before waiting to lock nodes, release
4365
    # all the ones living on nodes unrelated to the current operation.
4366
    if level == locking.LEVEL_NODE and self.lock_instances:
4367
      instances_release = []
4368
      instances_keep = []
4369
      self.affected_instances = []
4370
      if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
4371
        for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
4372
          instance = self.context.cfg.GetInstanceInfo(instance_name)
4373
          i_mirrored = instance.disk_template in constants.DTS_NET_MIRROR
4374
          if i_mirrored and self.op.node_name in instance.all_nodes:
4375
            instances_keep.append(instance_name)
4376
            self.affected_instances.append(instance)
4377
          else:
4378
            instances_release.append(instance_name)
4379
        if instances_release:
4380
          self.context.glm.release(locking.LEVEL_INSTANCE, instances_release)
4381
          self.acquired_locks[locking.LEVEL_INSTANCE] = instances_keep
4382

    
4383
  def BuildHooksEnv(self):
4384
    """Build hooks env.
4385

4386
    This runs on the master node.
4387

4388
    """
4389
    env = {
4390
      "OP_TARGET": self.op.node_name,
4391
      "MASTER_CANDIDATE": str(self.op.master_candidate),
4392
      "OFFLINE": str(self.op.offline),
4393
      "DRAINED": str(self.op.drained),
4394
      "MASTER_CAPABLE": str(self.op.master_capable),
4395
      "VM_CAPABLE": str(self.op.vm_capable),
4396
      }
4397
    nl = [self.cfg.GetMasterNode(),
4398
          self.op.node_name]
4399
    return env, nl, nl
4400

    
4401
  def CheckPrereq(self):
4402
    """Check prerequisites.
4403

4404
    This only checks the instance list against the existing names.
4405

4406
    """
4407
    node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
4408

    
4409
    if (self.op.master_candidate is not None or
4410
        self.op.drained is not None or
4411
        self.op.offline is not None):
4412
      # we can't change the master's node flags
4413
      if self.op.node_name == self.cfg.GetMasterNode():
4414
        raise errors.OpPrereqError("The master role can be changed"
4415
                                   " only via master-failover",
4416
                                   errors.ECODE_INVAL)
4417

    
4418
    if self.op.master_candidate and not node.master_capable:
4419
      raise errors.OpPrereqError("Node %s is not master capable, cannot make"
4420
                                 " it a master candidate" % node.name,
4421
                                 errors.ECODE_STATE)
4422

    
4423
    if self.op.vm_capable == False:
4424
      (ipri, isec) = self.cfg.GetNodeInstances(self.op.node_name)
4425
      if ipri or isec:
4426
        raise errors.OpPrereqError("Node %s hosts instances, cannot unset"
4427
                                   " the vm_capable flag" % node.name,
4428
                                   errors.ECODE_STATE)
4429

    
4430
    if node.master_candidate and self.might_demote and not self.lock_all:
4431
      assert not self.op.auto_promote, "auto_promote set but lock_all not"
4432
      # check if after removing the current node, we're missing master
4433
      # candidates
4434
      (mc_remaining, mc_should, _) = \
4435
          self.cfg.GetMasterCandidateStats(exceptions=[node.name])
4436
      if mc_remaining < mc_should:
4437
        raise errors.OpPrereqError("Not enough master candidates, please"
4438
                                   " pass auto promote option to allow"
4439
                                   " promotion", errors.ECODE_STATE)
4440

    
4441
    self.old_flags = old_flags = (node.master_candidate,
4442
                                  node.drained, node.offline)
4443
    assert old_flags in self._F2R, "Un-handled old flags  %s" % str(old_flags)
4444
    self.old_role = old_role = self._F2R[old_flags]
4445

    
4446
    # Check for ineffective changes
4447
    for attr in self._FLAGS:
4448
      if (getattr(self.op, attr) == False and getattr(node, attr) == False):
4449
        self.LogInfo("Ignoring request to unset flag %s, already unset", attr)
4450
        setattr(self.op, attr, None)
4451

    
4452
    # Past this point, any flag change to False means a transition
4453
    # away from the respective state, as only real changes are kept
4454

    
4455
    # TODO: We might query the real power state if it supports OOB
4456
    if _SupportsOob(self.cfg, node):
4457
      if self.op.offline is False and not (node.powered or
4458
                                           self.op.powered == True):
4459
        raise errors.OpPrereqError(("Please power on node %s first before you"
4460
                                    " can reset offline state") %
4461
                                   self.op.node_name)
4462
    elif self.op.powered is not None:
4463
      raise errors.OpPrereqError(("Unable to change powered state for node %s"
4464
                                  " which does not support out-of-band"
4465
                                  " handling") % self.op.node_name)
4466

    
4467
    # If we're being deofflined/drained, we'll MC ourself if needed
4468
    if (self.op.drained == False or self.op.offline == False or
4469
        (self.op.master_capable and not node.master_capable)):
4470
      if _DecideSelfPromotion(self):
4471
        self.op.master_candidate = True
4472
        self.LogInfo("Auto-promoting node to master candidate")
4473

    
4474
    # If we're no longer master capable, we'll demote ourselves from MC
4475
    if self.op.master_capable == False and node.master_candidate:
4476
      self.LogInfo("Demoting from master candidate")
4477
      self.op.master_candidate = False
4478

    
4479
    # Compute new role
4480
    assert [getattr(self.op, attr) for attr in self._FLAGS].count(True) <= 1
4481
    if self.op.master_candidate:
4482
      new_role = self._ROLE_CANDIDATE
4483
    elif self.op.drained:
4484
      new_role = self._ROLE_DRAINED
4485
    elif self.op.offline:
4486
      new_role = self._ROLE_OFFLINE
4487
    elif False in [self.op.master_candidate, self.op.drained, self.op.offline]:
4488
      # False is still in new flags, which means we're un-setting (the
4489
      # only) True flag
4490
      new_role = self._ROLE_REGULAR
4491
    else: # no new flags, nothing, keep old role
4492
      new_role = old_role
4493

    
4494
    self.new_role = new_role
4495

    
4496
    if old_role == self._ROLE_OFFLINE and new_role != old_role:
4497
      # Trying to transition out of offline status
4498
      result = self.rpc.call_version([node.name])[node.name]
4499
      if result.fail_msg:
4500
        raise errors.OpPrereqError("Node %s is being de-offlined but fails"
4501
                                   " to report its version: %s" %
4502
                                   (node.name, result.fail_msg),
4503
                                   errors.ECODE_STATE)
4504
      else:
4505
        self.LogWarning("Transitioning node from offline to online state"
4506
                        " without using re-add. Please make sure the node"
4507
                        " is healthy!")
4508

    
4509
    if self.op.secondary_ip:
4510
      # Ok even without locking, because this can't be changed by any LU
4511
      master = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
4512
      master_singlehomed = master.secondary_ip == master.primary_ip
4513
      if master_singlehomed and self.op.secondary_ip:
4514
        raise errors.OpPrereqError("Cannot change the secondary ip on a single"
4515
                                   " homed cluster", errors.ECODE_INVAL)
4516

    
4517
      if node.offline:
4518
        if self.affected_instances:
4519
          raise errors.OpPrereqError("Cannot change secondary ip: offline"
4520
                                     " node has instances (%s) configured"
4521
                                     " to use it" % self.affected_instances)
4522
      else:
4523
        # On online nodes, check that no instances are running, and that
4524
        # the node has the new ip and we can reach it.
4525
        for instance in self.affected_instances:
4526
          _CheckInstanceDown(self, instance, "cannot change secondary ip")
4527

    
4528
        _CheckNodeHasSecondaryIP(self, node.name, self.op.secondary_ip, True)
4529
        if master.name != node.name:
4530
          # check reachability from master secondary ip to new secondary ip
4531
          if not netutils.TcpPing(self.op.secondary_ip,
4532
                                  constants.DEFAULT_NODED_PORT,
4533
                                  source=master.secondary_ip):
4534
            raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
4535
                                       " based ping to node daemon port",
4536
                                       errors.ECODE_ENVIRON)
4537

    
4538
    if self.op.ndparams:
4539
      new_ndparams = _GetUpdatedParams(self.node.ndparams, self.op.ndparams)
4540
      utils.ForceDictType(new_ndparams, constants.NDS_PARAMETER_TYPES)
4541
      self.new_ndparams = new_ndparams
4542

    
4543
  def Exec(self, feedback_fn):
4544
    """Modifies a node.
4545

4546
    """
4547
    node = self.node
4548
    old_role = self.old_role
4549
    new_role = self.new_role
4550

    
4551
    result = []
4552

    
4553
    if self.op.ndparams:
4554
      node.ndparams = self.new_ndparams
4555

    
4556
    if self.op.powered is not None:
4557
      node.powered = self.op.powered
4558

    
4559
    for attr in ["master_capable", "vm_capable"]:
4560
      val = getattr(self.op, attr)
4561
      if val is not None:
4562
        setattr(node, attr, val)
4563
        result.append((attr, str(val)))
4564

    
4565
    if new_role != old_role:
4566
      # Tell the node to demote itself, if no longer MC and not offline
4567
      if old_role == self._ROLE_CANDIDATE and new_role != self._ROLE_OFFLINE:
4568
        msg = self.rpc.call_node_demote_from_mc(node.name).fail_msg
4569
        if msg:
4570
          self.LogWarning("Node failed to demote itself: %s", msg)
4571

    
4572
      new_flags = self._R2F[new_role]
4573
      for of, nf, desc in zip(self.old_flags, new_flags, self._FLAGS):
4574
        if of != nf:
4575
          result.append((desc, str(nf)))
4576
      (node.master_candidate, node.drained, node.offline) = new_flags
4577

    
4578
      # we locked all nodes, we adjust the CP before updating this node
4579
      if self.lock_all:
4580
        _AdjustCandidatePool(self, [node.name])
4581

    
4582
    if self.op.secondary_ip:
4583
      node.secondary_ip = self.op.secondary_ip
4584
      result.append(("secondary_ip", self.op.secondary_ip))
4585

    
4586
    # this will trigger configuration file update, if needed
4587
    self.cfg.Update(node, feedback_fn)
4588

    
4589
    # this will trigger job queue propagation or cleanup if the mc
4590
    # flag changed
4591
    if [old_role, new_role].count(self._ROLE_CANDIDATE) == 1:
4592
      self.context.ReaddNode(node)
4593

    
4594
    return result
4595

    
4596

    
4597
class LUNodePowercycle(NoHooksLU):
4598
  """Powercycles a node.
4599

4600
  """
4601
  REQ_BGL = False
4602

    
4603
  def CheckArguments(self):
4604
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4605
    if self.op.node_name == self.cfg.GetMasterNode() and not self.op.force:
4606
      raise errors.OpPrereqError("The node is the master and the force"
4607
                                 " parameter was not set",
4608
                                 errors.ECODE_INVAL)
4609

    
4610
  def ExpandNames(self):
4611
    """Locking for PowercycleNode.
4612

4613
    This is a last-resort option and shouldn't block on other
4614
    jobs. Therefore, we grab no locks.
4615

4616
    """
4617
    self.needed_locks = {}
4618

    
4619
  def Exec(self, feedback_fn):
4620
    """Reboots a node.
4621

4622
    """
4623
    result = self.rpc.call_node_powercycle(self.op.node_name,
4624
                                           self.cfg.GetHypervisorType())
4625
    result.Raise("Failed to schedule the reboot")
4626
    return result.payload
4627

    
4628

    
4629
class LUClusterQuery(NoHooksLU):
4630
  """Query cluster configuration.
4631

4632
  """
4633
  REQ_BGL = False
4634

    
4635
  def ExpandNames(self):
4636
    self.needed_locks = {}
4637

    
4638
  def Exec(self, feedback_fn):
4639
    """Return cluster config.
4640

4641
    """
4642
    cluster = self.cfg.GetClusterInfo()
4643
    os_hvp = {}
4644

    
4645
    # Filter just for enabled hypervisors
4646
    for os_name, hv_dict in cluster.os_hvp.items():
4647
      os_hvp[os_name] = {}
4648
      for hv_name, hv_params in hv_dict.items():
4649
        if hv_name in cluster.enabled_hypervisors:
4650
          os_hvp[os_name][hv_name] = hv_params
4651

    
4652
    # Convert ip_family to ip_version
4653
    primary_ip_version = constants.IP4_VERSION
4654
    if cluster.primary_ip_family == netutils.IP6Address.family:
4655
      primary_ip_version = constants.IP6_VERSION
4656

    
4657
    result = {
4658
      "software_version": constants.RELEASE_VERSION,
4659
      "protocol_version": constants.PROTOCOL_VERSION,
4660
      "config_version": constants.CONFIG_VERSION,
4661
      "os_api_version": max(constants.OS_API_VERSIONS),
4662
      "export_version": constants.EXPORT_VERSION,
4663
      "architecture": (platform.architecture()[0], platform.machine()),
4664
      "name": cluster.cluster_name,
4665
      "master": cluster.master_node,
4666
      "default_hypervisor": cluster.enabled_hypervisors[0],
4667
      "enabled_hypervisors": cluster.enabled_hypervisors,
4668
      "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
4669
                        for hypervisor_name in cluster.enabled_hypervisors]),
4670
      "os_hvp": os_hvp,
4671
      "beparams": cluster.beparams,
4672
      "osparams": cluster.osparams,
4673
      "nicparams": cluster.nicparams,
4674
      "ndparams": cluster.ndparams,
4675
      "candidate_pool_size": cluster.candidate_pool_size,
4676
      "master_netdev": cluster.master_netdev,
4677
      "volume_group_name": cluster.volume_group_name,
4678
      "drbd_usermode_helper": cluster.drbd_usermode_helper,
4679
      "file_storage_dir": cluster.file_storage_dir,
4680
      "maintain_node_health": cluster.maintain_node_health,
4681
      "ctime": cluster.ctime,
4682
      "mtime": cluster.mtime,
4683
      "uuid": cluster.uuid,
4684
      "tags": list(cluster.GetTags()),
4685
      "uid_pool": cluster.uid_pool,
4686
      "default_iallocator": cluster.default_iallocator,
4687
      "reserved_lvs": cluster.reserved_lvs,
4688
      "primary_ip_version": primary_ip_version,
4689
      "prealloc_wipe_disks": cluster.prealloc_wipe_disks,
4690
      "hidden_os": cluster.hidden_os,
4691
      "blacklisted_os": cluster.blacklisted_os,
4692
      }
4693

    
4694
    return result
4695

    
4696

    
4697
class LUClusterConfigQuery(NoHooksLU):
4698
  """Return configuration values.
4699

4700
  """
4701
  REQ_BGL = False
4702
  _FIELDS_DYNAMIC = utils.FieldSet()
4703
  _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag",
4704
                                  "watcher_pause", "volume_group_name")
4705

    
4706
  def CheckArguments(self):
4707
    _CheckOutputFields(static=self._FIELDS_STATIC,
4708
                       dynamic=self._FIELDS_DYNAMIC,
4709
                       selected=self.op.output_fields)
4710

    
4711
  def ExpandNames(self):
4712
    self.needed_locks = {}
4713

    
4714
  def Exec(self, feedback_fn):
4715
    """Dump a representation of the cluster config to the standard output.
4716

4717
    """
4718
    values = []
4719
    for field in self.op.output_fields:
4720
      if field == "cluster_name":
4721
        entry = self.cfg.GetClusterName()
4722
      elif field == "master_node":
4723
        entry = self.cfg.GetMasterNode()
4724
      elif field == "drain_flag":
4725
        entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
4726
      elif field == "watcher_pause":
4727
        entry = utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
4728
      elif field == "volume_group_name":
4729
        entry = self.cfg.GetVGName()
4730
      else:
4731
        raise errors.ParameterError(field)
4732
      values.append(entry)
4733
    return values
4734

    
4735

    
4736
class LUInstanceActivateDisks(NoHooksLU):
4737
  """Bring up an instance's disks.
4738

4739
  """
4740
  REQ_BGL = False
4741

    
4742
  def ExpandNames(self):
4743
    self._ExpandAndLockInstance()
4744
    self.needed_locks[locking.LEVEL_NODE] = []
4745
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4746

    
4747
  def DeclareLocks(self, level):
4748
    if level == locking.LEVEL_NODE:
4749
      self._LockInstancesNodes()
4750

    
4751
  def CheckPrereq(self):
4752
    """Check prerequisites.
4753

4754
    This checks that the instance is in the cluster.
4755

4756
    """
4757
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4758
    assert self.instance is not None, \
4759
      "Cannot retrieve locked instance %s" % self.op.instance_name
4760
    _CheckNodeOnline(self, self.instance.primary_node)
4761

    
4762
  def Exec(self, feedback_fn):
4763
    """Activate the disks.
4764

4765
    """
4766
    disks_ok, disks_info = \
4767
              _AssembleInstanceDisks(self, self.instance,
4768
                                     ignore_size=self.op.ignore_size)
4769
    if not disks_ok:
4770
      raise errors.OpExecError("Cannot activate block devices")
4771

    
4772
    return disks_info
4773

    
4774

    
4775
def _AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False,
4776
                           ignore_size=False):
4777
  """Prepare the block devices for an instance.
4778

4779
  This sets up the block devices on all nodes.
4780

4781
  @type lu: L{LogicalUnit}
4782
  @param lu: the logical unit on whose behalf we execute
4783
  @type instance: L{objects.Instance}
4784
  @param instance: the instance for whose disks we assemble
4785
  @type disks: list of L{objects.Disk} or None
4786
  @param disks: which disks to assemble (or all, if None)
4787
  @type ignore_secondaries: boolean
4788
  @param ignore_secondaries: if true, errors on secondary nodes
4789
      won't result in an error return from the function
4790
  @type ignore_size: boolean
4791
  @param ignore_size: if true, the current known size of the disk
4792
      will not be used during the disk activation, useful for cases
4793
      when the size is wrong
4794
  @return: False if the operation failed, otherwise a list of
4795
      (host, instance_visible_name, node_visible_name)
4796
      with the mapping from node devices to instance devices
4797

4798
  """
4799
  device_info = []
4800
  disks_ok = True
4801
  iname = instance.name
4802
  disks = _ExpandCheckDisks(instance, disks)
4803

    
4804
  # With the two passes mechanism we try to reduce the window of
4805
  # opportunity for the race condition of switching DRBD to primary
4806
  # before handshaking occured, but we do not eliminate it
4807

    
4808
  # The proper fix would be to wait (with some limits) until the
4809
  # connection has been made and drbd transitions from WFConnection
4810
  # into any other network-connected state (Connected, SyncTarget,
4811
  # SyncSource, etc.)
4812

    
4813
  # 1st pass, assemble on all nodes in secondary mode
4814
  for idx, inst_disk in enumerate(disks):
4815
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
4816
      if ignore_size:
4817
        node_disk = node_disk.Copy()
4818
        node_disk.UnsetSize()
4819
      lu.cfg.SetDiskID(node_disk, node)
4820
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False, idx)
4821
      msg = result.fail_msg
4822
      if msg:
4823
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
4824
                           " (is_primary=False, pass=1): %s",
4825
                           inst_disk.iv_name, node, msg)
4826
        if not ignore_secondaries:
4827
          disks_ok = False
4828

    
4829
  # FIXME: race condition on drbd migration to primary
4830

    
4831
  # 2nd pass, do only the primary node
4832
  for idx, inst_disk in enumerate(disks):
4833
    dev_path = None
4834

    
4835
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
4836
      if node != instance.primary_node:
4837
        continue
4838
      if ignore_size:
4839
        node_disk = node_disk.Copy()
4840
        node_disk.UnsetSize()
4841
      lu.cfg.SetDiskID(node_disk, node)
4842
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True, idx)
4843
      msg = result.fail_msg
4844
      if msg:
4845
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
4846
                           " (is_primary=True, pass=2): %s",
4847
                           inst_disk.iv_name, node, msg)
4848
        disks_ok = False
4849
      else:
4850
        dev_path = result.payload
4851

    
4852
    device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
4853

    
4854
  # leave the disks configured for the primary node
4855
  # this is a workaround that would be fixed better by
4856
  # improving the logical/physical id handling
4857
  for disk in disks:
4858
    lu.cfg.SetDiskID(disk, instance.primary_node)
4859

    
4860
  return disks_ok, device_info
4861

    
4862

    
4863
def _StartInstanceDisks(lu, instance, force):
4864
  """Start the disks of an instance.
4865

4866
  """
4867
  disks_ok, _ = _AssembleInstanceDisks(lu, instance,
4868
                                           ignore_secondaries=force)
4869
  if not disks_ok:
4870
    _ShutdownInstanceDisks(lu, instance)
4871
    if force is not None and not force:
4872
      lu.proc.LogWarning("", hint="If the message above refers to a"
4873
                         " secondary node,"
4874
                         " you can retry the operation using '--force'.")
4875
    raise errors.OpExecError("Disk consistency error")
4876

    
4877

    
4878
class LUInstanceDeactivateDisks(NoHooksLU):
4879
  """Shutdown an instance's disks.
4880

4881
  """
4882
  REQ_BGL = False
4883

    
4884
  def ExpandNames(self):
4885
    self._ExpandAndLockInstance()
4886
    self.needed_locks[locking.LEVEL_NODE] = []
4887
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4888

    
4889
  def DeclareLocks(self, level):
4890
    if level == locking.LEVEL_NODE:
4891
      self._LockInstancesNodes()
4892

    
4893
  def CheckPrereq(self):
4894
    """Check prerequisites.
4895

4896
    This checks that the instance is in the cluster.
4897

4898
    """
4899
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4900
    assert self.instance is not None, \
4901
      "Cannot retrieve locked instance %s" % self.op.instance_name
4902

    
4903
  def Exec(self, feedback_fn):
4904
    """Deactivate the disks
4905

4906
    """
4907
    instance = self.instance
4908
    if self.op.force:
4909
      _ShutdownInstanceDisks(self, instance)
4910
    else:
4911
      _SafeShutdownInstanceDisks(self, instance)
4912

    
4913

    
4914
def _SafeShutdownInstanceDisks(lu, instance, disks=None):
4915
  """Shutdown block devices of an instance.
4916

4917
  This function checks if an instance is running, before calling
4918
  _ShutdownInstanceDisks.
4919

4920
  """
4921
  _CheckInstanceDown(lu, instance, "cannot shutdown disks")
4922
  _ShutdownInstanceDisks(lu, instance, disks=disks)
4923

    
4924

    
4925
def _ExpandCheckDisks(instance, disks):
4926
  """Return the instance disks selected by the disks list
4927

4928
  @type disks: list of L{objects.Disk} or None
4929
  @param disks: selected disks
4930
  @rtype: list of L{objects.Disk}
4931
  @return: selected instance disks to act on
4932

4933
  """
4934
  if disks is None:
4935
    return instance.disks
4936
  else:
4937
    if not set(disks).issubset(instance.disks):
4938
      raise errors.ProgrammerError("Can only act on disks belonging to the"
4939
                                   " target instance")
4940
    return disks
4941

    
4942

    
4943
def _ShutdownInstanceDisks(lu, instance, disks=None, ignore_primary=False):
4944
  """Shutdown block devices of an instance.
4945

4946
  This does the shutdown on all nodes of the instance.
4947

4948
  If the ignore_primary is false, errors on the primary node are
4949
  ignored.
4950

4951
  """
4952
  all_result = True
4953
  disks = _ExpandCheckDisks(instance, disks)
4954

    
4955
  for disk in disks:
4956
    for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
4957
      lu.cfg.SetDiskID(top_disk, node)
4958
      result = lu.rpc.call_blockdev_shutdown(node, top_disk)
4959
      msg = result.fail_msg
4960
      if msg:
4961
        lu.LogWarning("Could not shutdown block device %s on node %s: %s",
4962
                      disk.iv_name, node, msg)
4963
        if ((node == instance.primary_node and not ignore_primary) or
4964
            (node != instance.primary_node and not result.offline)):
4965
          all_result = False
4966
  return all_result
4967

    
4968

    
4969
def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
4970
  """Checks if a node has enough free memory.
4971

4972
  This function check if a given node has the needed amount of free
4973
  memory. In case the node has less memory or we cannot get the
4974
  information from the node, this function raise an OpPrereqError
4975
  exception.
4976

4977
  @type lu: C{LogicalUnit}
4978
  @param lu: a logical unit from which we get configuration data
4979
  @type node: C{str}
4980
  @param node: the node to check
4981
  @type reason: C{str}
4982
  @param reason: string to use in the error message
4983
  @type requested: C{int}
4984
  @param requested: the amount of memory in MiB to check for
4985
  @type hypervisor_name: C{str}
4986
  @param hypervisor_name: the hypervisor to ask for memory stats
4987
  @raise errors.OpPrereqError: if the node doesn't have enough memory, or
4988
      we cannot check the node
4989

4990
  """
4991
  nodeinfo = lu.rpc.call_node_info([node], None, hypervisor_name)
4992
  nodeinfo[node].Raise("Can't get data from node %s" % node,
4993
                       prereq=True, ecode=errors.ECODE_ENVIRON)
4994
  free_mem = nodeinfo[node].payload.get('memory_free', None)
4995
  if not isinstance(free_mem, int):
4996
    raise errors.OpPrereqError("Can't compute free memory on node %s, result"
4997
                               " was '%s'" % (node, free_mem),
4998
                               errors.ECODE_ENVIRON)
4999
  if requested > free_mem:
5000
    raise errors.OpPrereqError("Not enough memory on node %s for %s:"
5001
                               " needed %s MiB, available %s MiB" %
5002
                               (node, reason, requested, free_mem),
5003
                               errors.ECODE_NORES)
5004

    
5005

    
5006
def _CheckNodesFreeDiskPerVG(lu, nodenames, req_sizes):
5007
  """Checks if nodes have enough free disk space in the all VGs.
5008

5009
  This function check if all given nodes have the needed amount of
5010
  free disk. In case any node has less disk or we cannot get the
5011
  information from the node, this function raise an OpPrereqError
5012
  exception.
5013

5014
  @type lu: C{LogicalUnit}
5015
  @param lu: a logical unit from which we get configuration data
5016
  @type nodenames: C{list}
5017
  @param nodenames: the list of node names to check
5018
  @type req_sizes: C{dict}
5019
  @param req_sizes: the hash of vg and corresponding amount of disk in
5020
      MiB to check for
5021
  @raise errors.OpPrereqError: if the node doesn't have enough disk,
5022
      or we cannot check the node
5023

5024
  """
5025
  for vg, req_size in req_sizes.items():
5026
    _CheckNodesFreeDiskOnVG(lu, nodenames, vg, req_size)
5027

    
5028

    
5029
def _CheckNodesFreeDiskOnVG(lu, nodenames, vg, requested):
5030
  """Checks if nodes have enough free disk space in the specified VG.
5031

5032
  This function check if all given nodes have the needed amount of
5033
  free disk. In case any node has less disk or we cannot get the
5034
  information from the node, this function raise an OpPrereqError
5035
  exception.
5036

5037
  @type lu: C{LogicalUnit}
5038
  @param lu: a logical unit from which we get configuration data
5039
  @type nodenames: C{list}
5040
  @param nodenames: the list of node names to check
5041
  @type vg: C{str}
5042
  @param vg: the volume group to check
5043
  @type requested: C{int}
5044
  @param requested: the amount of disk in MiB to check for
5045
  @raise errors.OpPrereqError: if the node doesn't have enough disk,
5046
      or we cannot check the node
5047

5048
  """
5049
  nodeinfo = lu.rpc.call_node_info(nodenames, vg, None)
5050
  for node in nodenames:
5051
    info = nodeinfo[node]
5052
    info.Raise("Cannot get current information from node %s" % node,
5053
               prereq=True, ecode=errors.ECODE_ENVIRON)
5054
    vg_free = info.payload.get("vg_free", None)
5055
    if not isinstance(vg_free, int):
5056
      raise errors.OpPrereqError("Can't compute free disk space on node"
5057
                                 " %s for vg %s, result was '%s'" %
5058
                                 (node, vg, vg_free), errors.ECODE_ENVIRON)
5059
    if requested > vg_free:
5060
      raise errors.OpPrereqError("Not enough disk space on target node %s"
5061
                                 " vg %s: required %d MiB, available %d MiB" %
5062
                                 (node, vg, requested, vg_free),
5063
                                 errors.ECODE_NORES)
5064

    
5065

    
5066
class LUInstanceStartup(LogicalUnit):
5067
  """Starts an instance.
5068

5069
  """
5070
  HPATH = "instance-start"
5071
  HTYPE = constants.HTYPE_INSTANCE
5072
  REQ_BGL = False
5073

    
5074
  def CheckArguments(self):
5075
    # extra beparams
5076
    if self.op.beparams:
5077
      # fill the beparams dict
5078
      utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
5079

    
5080
  def ExpandNames(self):
5081
    self._ExpandAndLockInstance()
5082

    
5083
  def BuildHooksEnv(self):
5084
    """Build hooks env.
5085

5086
    This runs on master, primary and secondary nodes of the instance.
5087

5088
    """
5089
    env = {
5090
      "FORCE": self.op.force,
5091
      }
5092
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5093
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5094
    return env, nl, nl
5095

    
5096
  def CheckPrereq(self):
5097
    """Check prerequisites.
5098

5099
    This checks that the instance is in the cluster.
5100

5101
    """
5102
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5103
    assert self.instance is not None, \
5104
      "Cannot retrieve locked instance %s" % self.op.instance_name
5105

    
5106
    # extra hvparams
5107
    if self.op.hvparams:
5108
      # check hypervisor parameter syntax (locally)
5109
      cluster = self.cfg.GetClusterInfo()
5110
      utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
5111
      filled_hvp = cluster.FillHV(instance)
5112
      filled_hvp.update(self.op.hvparams)
5113
      hv_type = hypervisor.GetHypervisor(instance.hypervisor)
5114
      hv_type.CheckParameterSyntax(filled_hvp)
5115
      _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
5116

    
5117
    self.primary_offline = self.cfg.GetNodeInfo(instance.primary_node).offline
5118

    
5119
    if self.primary_offline and self.op.ignore_offline_nodes:
5120
      self.proc.LogWarning("Ignoring offline primary node")
5121

    
5122
      if self.op.hvparams or self.op.beparams:
5123
        self.proc.LogWarning("Overridden parameters are ignored")
5124
    else:
5125
      _CheckNodeOnline(self, instance.primary_node)
5126

    
5127
      bep = self.cfg.GetClusterInfo().FillBE(instance)
5128

    
5129
      # check bridges existence
5130
      _CheckInstanceBridgesExist(self, instance)
5131

    
5132
      remote_info = self.rpc.call_instance_info(instance.primary_node,
5133
                                                instance.name,
5134
                                                instance.hypervisor)
5135
      remote_info.Raise("Error checking node %s" % instance.primary_node,
5136
                        prereq=True, ecode=errors.ECODE_ENVIRON)
5137
      if not remote_info.payload: # not running already
5138
        _CheckNodeFreeMemory(self, instance.primary_node,
5139
                             "starting instance %s" % instance.name,
5140
                             bep[constants.BE_MEMORY], instance.hypervisor)
5141

    
5142
  def Exec(self, feedback_fn):
5143
    """Start the instance.
5144

5145
    """
5146
    instance = self.instance
5147
    force = self.op.force
5148

    
5149
    self.cfg.MarkInstanceUp(instance.name)
5150

    
5151
    if self.primary_offline:
5152
      assert self.op.ignore_offline_nodes
5153
      self.proc.LogInfo("Primary node offline, marked instance as started")
5154
    else:
5155
      node_current = instance.primary_node
5156

    
5157
      _StartInstanceDisks(self, instance, force)
5158

    
5159
      result = self.rpc.call_instance_start(node_current, instance,
5160
                                            self.op.hvparams, self.op.beparams)
5161
      msg = result.fail_msg
5162
      if msg:
5163
        _ShutdownInstanceDisks(self, instance)
5164
        raise errors.OpExecError("Could not start instance: %s" % msg)
5165

    
5166

    
5167
class LUInstanceReboot(LogicalUnit):
5168
  """Reboot an instance.
5169

5170
  """
5171
  HPATH = "instance-reboot"
5172
  HTYPE = constants.HTYPE_INSTANCE
5173
  REQ_BGL = False
5174

    
5175
  def ExpandNames(self):
5176
    self._ExpandAndLockInstance()
5177

    
5178
  def BuildHooksEnv(self):
5179
    """Build hooks env.
5180

5181
    This runs on master, primary and secondary nodes of the instance.
5182

5183
    """
5184
    env = {
5185
      "IGNORE_SECONDARIES": self.op.ignore_secondaries,
5186
      "REBOOT_TYPE": self.op.reboot_type,
5187
      "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5188
      }
5189
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5190
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5191
    return env, nl, nl
5192

    
5193
  def CheckPrereq(self):
5194
    """Check prerequisites.
5195

5196
    This checks that the instance is in the cluster.
5197

5198
    """
5199
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5200
    assert self.instance is not None, \
5201
      "Cannot retrieve locked instance %s" % self.op.instance_name
5202

    
5203
    _CheckNodeOnline(self, instance.primary_node)
5204

    
5205
    # check bridges existence
5206
    _CheckInstanceBridgesExist(self, instance)
5207

    
5208
  def Exec(self, feedback_fn):
5209
    """Reboot the instance.
5210

5211
    """
5212
    instance = self.instance
5213
    ignore_secondaries = self.op.ignore_secondaries
5214
    reboot_type = self.op.reboot_type
5215

    
5216
    node_current = instance.primary_node
5217

    
5218
    if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
5219
                       constants.INSTANCE_REBOOT_HARD]:
5220
      for disk in instance.disks:
5221
        self.cfg.SetDiskID(disk, node_current)
5222
      result = self.rpc.call_instance_reboot(node_current, instance,
5223
                                             reboot_type,
5224
                                             self.op.shutdown_timeout)
5225
      result.Raise("Could not reboot instance")
5226
    else:
5227
      result = self.rpc.call_instance_shutdown(node_current, instance,
5228
                                               self.op.shutdown_timeout)
5229
      result.Raise("Could not shutdown instance for full reboot")
5230
      _ShutdownInstanceDisks(self, instance)
5231
      _StartInstanceDisks(self, instance, ignore_secondaries)
5232
      result = self.rpc.call_instance_start(node_current, instance, None, None)
5233
      msg = result.fail_msg
5234
      if msg:
5235
        _ShutdownInstanceDisks(self, instance)
5236
        raise errors.OpExecError("Could not start instance for"
5237
                                 " full reboot: %s" % msg)
5238

    
5239
    self.cfg.MarkInstanceUp(instance.name)
5240

    
5241

    
5242
class LUInstanceShutdown(LogicalUnit):
5243
  """Shutdown an instance.
5244

5245
  """
5246
  HPATH = "instance-stop"
5247
  HTYPE = constants.HTYPE_INSTANCE
5248
  REQ_BGL = False
5249

    
5250
  def ExpandNames(self):
5251
    self._ExpandAndLockInstance()
5252

    
5253
  def BuildHooksEnv(self):
5254
    """Build hooks env.
5255

5256
    This runs on master, primary and secondary nodes of the instance.
5257

5258
    """
5259
    env = _BuildInstanceHookEnvByObject(self, self.instance)
5260
    env["TIMEOUT"] = self.op.timeout
5261
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5262
    return env, nl, nl
5263

    
5264
  def CheckPrereq(self):
5265
    """Check prerequisites.
5266

5267
    This checks that the instance is in the cluster.
5268

5269
    """
5270
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5271
    assert self.instance is not None, \
5272
      "Cannot retrieve locked instance %s" % self.op.instance_name
5273

    
5274
    self.primary_offline = \
5275
      self.cfg.GetNodeInfo(self.instance.primary_node).offline
5276

    
5277
    if self.primary_offline and self.op.ignore_offline_nodes:
5278
      self.proc.LogWarning("Ignoring offline primary node")
5279
    else:
5280
      _CheckNodeOnline(self, self.instance.primary_node)
5281

    
5282
  def Exec(self, feedback_fn):
5283
    """Shutdown the instance.
5284

5285
    """
5286
    instance = self.instance
5287
    node_current = instance.primary_node
5288
    timeout = self.op.timeout
5289

    
5290
    self.cfg.MarkInstanceDown(instance.name)
5291

    
5292
    if self.primary_offline:
5293
      assert self.op.ignore_offline_nodes
5294
      self.proc.LogInfo("Primary node offline, marked instance as stopped")
5295
    else:
5296
      result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
5297
      msg = result.fail_msg
5298
      if msg:
5299
        self.proc.LogWarning("Could not shutdown instance: %s" % msg)
5300

    
5301
      _ShutdownInstanceDisks(self, instance)
5302

    
5303

    
5304
class LUInstanceReinstall(LogicalUnit):
5305
  """Reinstall an instance.
5306

5307
  """
5308
  HPATH = "instance-reinstall"
5309
  HTYPE = constants.HTYPE_INSTANCE
5310
  REQ_BGL = False
5311

    
5312
  def ExpandNames(self):
5313
    self._ExpandAndLockInstance()
5314

    
5315
  def BuildHooksEnv(self):
5316
    """Build hooks env.
5317

5318
    This runs on master, primary and secondary nodes of the instance.
5319

5320
    """
5321
    env = _BuildInstanceHookEnvByObject(self, self.instance)
5322
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5323
    return env, nl, nl
5324

    
5325
  def CheckPrereq(self):
5326
    """Check prerequisites.
5327

5328
    This checks that the instance is in the cluster and is not running.
5329

5330
    """
5331
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5332
    assert instance is not None, \
5333
      "Cannot retrieve locked instance %s" % self.op.instance_name
5334
    _CheckNodeOnline(self, instance.primary_node, "Instance primary node"
5335
                     " offline, cannot reinstall")
5336
    for node in instance.secondary_nodes:
5337
      _CheckNodeOnline(self, node, "Instance secondary node offline,"
5338
                       " cannot reinstall")
5339

    
5340
    if instance.disk_template == constants.DT_DISKLESS:
5341
      raise errors.OpPrereqError("Instance '%s' has no disks" %
5342
                                 self.op.instance_name,
5343
                                 errors.ECODE_INVAL)
5344
    _CheckInstanceDown(self, instance, "cannot reinstall")
5345

    
5346
    if self.op.os_type is not None:
5347
      # OS verification
5348
      pnode = _ExpandNodeName(self.cfg, instance.primary_node)
5349
      _CheckNodeHasOS(self, pnode, self.op.os_type, self.op.force_variant)
5350
      instance_os = self.op.os_type
5351
    else:
5352
      instance_os = instance.os
5353

    
5354
    nodelist = list(instance.all_nodes)
5355

    
5356
    if self.op.osparams:
5357
      i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
5358
      _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
5359
      self.os_inst = i_osdict # the new dict (without defaults)
5360
    else:
5361
      self.os_inst = None
5362

    
5363
    self.instance = instance
5364

    
5365
  def Exec(self, feedback_fn):
5366
    """Reinstall the instance.
5367

5368
    """
5369
    inst = self.instance
5370

    
5371
    if self.op.os_type is not None:
5372
      feedback_fn("Changing OS to '%s'..." % self.op.os_type)
5373
      inst.os = self.op.os_type
5374
      # Write to configuration
5375
      self.cfg.Update(inst, feedback_fn)
5376

    
5377
    _StartInstanceDisks(self, inst, None)
5378
    try:
5379
      feedback_fn("Running the instance OS create scripts...")
5380
      # FIXME: pass debug option from opcode to backend
5381
      result = self.rpc.call_instance_os_add(inst.primary_node, inst, True,
5382
                                             self.op.debug_level,
5383
                                             osparams=self.os_inst)
5384
      result.Raise("Could not install OS for instance %s on node %s" %
5385
                   (inst.name, inst.primary_node))
5386
    finally:
5387
      _ShutdownInstanceDisks(self, inst)
5388

    
5389

    
5390
class LUInstanceRecreateDisks(LogicalUnit):
5391
  """Recreate an instance's missing disks.
5392

5393
  """
5394
  HPATH = "instance-recreate-disks"
5395
  HTYPE = constants.HTYPE_INSTANCE
5396
  REQ_BGL = False
5397

    
5398
  def ExpandNames(self):
5399
    self._ExpandAndLockInstance()
5400

    
5401
  def BuildHooksEnv(self):
5402
    """Build hooks env.
5403

5404
    This runs on master, primary and secondary nodes of the instance.
5405

5406
    """
5407
    env = _BuildInstanceHookEnvByObject(self, self.instance)
5408
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5409
    return env, nl, nl
5410

    
5411
  def CheckPrereq(self):
5412
    """Check prerequisites.
5413

5414
    This checks that the instance is in the cluster and is not running.
5415

5416
    """
5417
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5418
    assert instance is not None, \
5419
      "Cannot retrieve locked instance %s" % self.op.instance_name
5420
    _CheckNodeOnline(self, instance.primary_node)
5421

    
5422
    if instance.disk_template == constants.DT_DISKLESS:
5423
      raise errors.OpPrereqError("Instance '%s' has no disks" %
5424
                                 self.op.instance_name, errors.ECODE_INVAL)
5425
    _CheckInstanceDown(self, instance, "cannot recreate disks")
5426

    
5427
    if not self.op.disks:
5428
      self.op.disks = range(len(instance.disks))
5429
    else:
5430
      for idx in self.op.disks:
5431
        if idx >= len(instance.disks):
5432
          raise errors.OpPrereqError("Invalid disk index passed '%s'" % idx,
5433
                                     errors.ECODE_INVAL)
5434

    
5435
    self.instance = instance
5436

    
5437
  def Exec(self, feedback_fn):
5438
    """Recreate the disks.
5439

5440
    """
5441
    to_skip = []
5442
    for idx, _ in enumerate(self.instance.disks):
5443
      if idx not in self.op.disks: # disk idx has not been passed in
5444
        to_skip.append(idx)
5445
        continue
5446

    
5447
    _CreateDisks(self, self.instance, to_skip=to_skip)
5448

    
5449

    
5450
class LUInstanceRename(LogicalUnit):
5451
  """Rename an instance.
5452

5453
  """
5454
  HPATH = "instance-rename"
5455
  HTYPE = constants.HTYPE_INSTANCE
5456

    
5457
  def CheckArguments(self):
5458
    """Check arguments.
5459

5460
    """
5461
    if self.op.ip_check and not self.op.name_check:
5462
      # TODO: make the ip check more flexible and not depend on the name check
5463
      raise errors.OpPrereqError("Cannot do ip check without a name check",
5464
                                 errors.ECODE_INVAL)
5465

    
5466
  def BuildHooksEnv(self):
5467
    """Build hooks env.
5468

5469
    This runs on master, primary and secondary nodes of the instance.
5470

5471
    """
5472
    env = _BuildInstanceHookEnvByObject(self, self.instance)
5473
    env["INSTANCE_NEW_NAME"] = self.op.new_name
5474
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5475
    return env, nl, nl
5476

    
5477
  def CheckPrereq(self):
5478
    """Check prerequisites.
5479

5480
    This checks that the instance is in the cluster and is not running.
5481

5482
    """
5483
    self.op.instance_name = _ExpandInstanceName(self.cfg,
5484
                                                self.op.instance_name)
5485
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5486
    assert instance is not None
5487
    _CheckNodeOnline(self, instance.primary_node)
5488
    _CheckInstanceDown(self, instance, "cannot rename")
5489
    self.instance = instance
5490

    
5491
    new_name = self.op.new_name
5492
    if self.op.name_check:
5493
      hostname = netutils.GetHostname(name=new_name)
5494
      self.LogInfo("Resolved given name '%s' to '%s'", new_name,
5495
                   hostname.name)
5496
      new_name = self.op.new_name = hostname.name
5497
      if (self.op.ip_check and
5498
          netutils.TcpPing(hostname.ip, constants.DEFAULT_NODED_PORT)):
5499
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
5500
                                   (hostname.ip, new_name),
5501
                                   errors.ECODE_NOTUNIQUE)
5502

    
5503
    instance_list = self.cfg.GetInstanceList()
5504
    if new_name in instance_list and new_name != instance.name:
5505
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
5506
                                 new_name, errors.ECODE_EXISTS)
5507

    
5508
  def Exec(self, feedback_fn):
5509
    """Rename the instance.
5510

5511
    """
5512
    inst = self.instance
5513
    old_name = inst.name
5514

    
5515
    rename_file_storage = False
5516
    if (inst.disk_template == constants.DT_FILE and
5517
        self.op.new_name != inst.name):
5518
      old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
5519
      rename_file_storage = True
5520

    
5521
    self.cfg.RenameInstance(inst.name, self.op.new_name)
5522
    # Change the instance lock. This is definitely safe while we hold the BGL
5523
    self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
5524
    self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
5525

    
5526
    # re-read the instance from the configuration after rename
5527
    inst = self.cfg.GetInstanceInfo(self.op.new_name)
5528

    
5529
    if rename_file_storage:
5530
      new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
5531
      result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
5532
                                                     old_file_storage_dir,
5533
                                                     new_file_storage_dir)
5534
      result.Raise("Could not rename on node %s directory '%s' to '%s'"
5535
                   " (but the instance has been renamed in Ganeti)" %
5536
                   (inst.primary_node, old_file_storage_dir,
5537
                    new_file_storage_dir))
5538

    
5539
    _StartInstanceDisks(self, inst, None)
5540
    try:
5541
      result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
5542
                                                 old_name, self.op.debug_level)
5543
      msg = result.fail_msg
5544
      if msg:
5545
        msg = ("Could not run OS rename script for instance %s on node %s"
5546
               " (but the instance has been renamed in Ganeti): %s" %
5547
               (inst.name, inst.primary_node, msg))
5548
        self.proc.LogWarning(msg)
5549
    finally:
5550
      _ShutdownInstanceDisks(self, inst)
5551

    
5552
    return inst.name
5553

    
5554

    
5555
class LUInstanceRemove(LogicalUnit):
5556
  """Remove an instance.
5557

5558
  """
5559
  HPATH = "instance-remove"
5560
  HTYPE = constants.HTYPE_INSTANCE
5561
  REQ_BGL = False
5562

    
5563
  def ExpandNames(self):
5564
    self._ExpandAndLockInstance()
5565
    self.needed_locks[locking.LEVEL_NODE] = []
5566
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5567

    
5568
  def DeclareLocks(self, level):
5569
    if level == locking.LEVEL_NODE:
5570
      self._LockInstancesNodes()
5571

    
5572
  def BuildHooksEnv(self):
5573
    """Build hooks env.
5574

5575
    This runs on master, primary and secondary nodes of the instance.
5576

5577
    """
5578
    env = _BuildInstanceHookEnvByObject(self, self.instance)
5579
    env["SHUTDOWN_TIMEOUT"] = self.op.shutdown_timeout
5580
    nl = [self.cfg.GetMasterNode()]
5581
    nl_post = list(self.instance.all_nodes) + nl
5582
    return env, nl, nl_post
5583

    
5584
  def CheckPrereq(self):
5585
    """Check prerequisites.
5586

5587
    This checks that the instance is in the cluster.
5588

5589
    """
5590
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5591
    assert self.instance is not None, \
5592
      "Cannot retrieve locked instance %s" % self.op.instance_name
5593

    
5594
  def Exec(self, feedback_fn):
5595
    """Remove the instance.
5596

5597
    """
5598
    instance = self.instance
5599
    logging.info("Shutting down instance %s on node %s",
5600
                 instance.name, instance.primary_node)
5601

    
5602
    result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
5603
                                             self.op.shutdown_timeout)
5604
    msg = result.fail_msg
5605
    if msg:
5606
      if self.op.ignore_failures:
5607
        feedback_fn("Warning: can't shutdown instance: %s" % msg)
5608
      else:
5609
        raise errors.OpExecError("Could not shutdown instance %s on"
5610
                                 " node %s: %s" %
5611
                                 (instance.name, instance.primary_node, msg))
5612

    
5613
    _RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures)
5614

    
5615

    
5616
def _RemoveInstance(lu, feedback_fn, instance, ignore_failures):
5617
  """Utility function to remove an instance.
5618

5619
  """
5620
  logging.info("Removing block devices for instance %s", instance.name)
5621

    
5622
  if not _RemoveDisks(lu, instance):
5623
    if not ignore_failures:
5624
      raise errors.OpExecError("Can't remove instance's disks")
5625
    feedback_fn("Warning: can't remove instance's disks")
5626

    
5627
  logging.info("Removing instance %s out of cluster config", instance.name)
5628

    
5629
  lu.cfg.RemoveInstance(instance.name)
5630

    
5631
  assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \
5632
    "Instance lock removal conflict"
5633

    
5634
  # Remove lock for the instance
5635
  lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name
5636

    
5637

    
5638
class LUInstanceQuery(NoHooksLU):
5639
  """Logical unit for querying instances.
5640

5641
  """
5642
  # pylint: disable-msg=W0142
5643
  REQ_BGL = False
5644

    
5645
  def CheckArguments(self):
5646
    self.iq = _InstanceQuery(self.op.names, self.op.output_fields,
5647
                             self.op.use_locking)
5648

    
5649
  def ExpandNames(self):
5650
    self.iq.ExpandNames(self)
5651

    
5652
  def DeclareLocks(self, level):
5653
    self.iq.DeclareLocks(self, level)
5654

    
5655
  def Exec(self, feedback_fn):
5656
    return self.iq.OldStyleQuery(self)
5657

    
5658

    
5659
class LUInstanceFailover(LogicalUnit):
5660
  """Failover an instance.
5661

5662
  """
5663
  HPATH = "instance-failover"
5664
  HTYPE = constants.HTYPE_INSTANCE
5665
  REQ_BGL = False
5666

    
5667
  def ExpandNames(self):
5668
    self._ExpandAndLockInstance()
5669
    self.needed_locks[locking.LEVEL_NODE] = []
5670
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5671

    
5672
  def DeclareLocks(self, level):
5673
    if level == locking.LEVEL_NODE:
5674
      self._LockInstancesNodes()
5675

    
5676
  def BuildHooksEnv(self):
5677
    """Build hooks env.
5678

5679
    This runs on master, primary and secondary nodes of the instance.
5680

5681
    """
5682
    instance = self.instance
5683
    source_node = instance.primary_node
5684
    target_node = instance.secondary_nodes[0]
5685
    env = {
5686
      "IGNORE_CONSISTENCY": self.op.ignore_consistency,
5687
      "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5688
      "OLD_PRIMARY": source_node,
5689
      "OLD_SECONDARY": target_node,
5690
      "NEW_PRIMARY": target_node,
5691
      "NEW_SECONDARY": source_node,
5692
      }
5693
    env.update(_BuildInstanceHookEnvByObject(self, instance))
5694
    nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5695
    nl_post = list(nl)
5696
    nl_post.append(source_node)
5697
    return env, nl, nl_post
5698

    
5699
  def CheckPrereq(self):
5700
    """Check prerequisites.
5701

5702
    This checks that the instance is in the cluster.
5703

5704
    """
5705
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5706
    assert self.instance is not None, \
5707
      "Cannot retrieve locked instance %s" % self.op.instance_name
5708

    
5709
    bep = self.cfg.GetClusterInfo().FillBE(instance)
5710
    if instance.disk_template not in constants.DTS_NET_MIRROR:
5711
      raise errors.OpPrereqError("Instance's disk layout is not"
5712
                                 " network mirrored, cannot failover.",
5713
                                 errors.ECODE_STATE)
5714

    
5715
    secondary_nodes = instance.secondary_nodes
5716
    if not secondary_nodes:
5717
      raise errors.ProgrammerError("no secondary node but using "
5718
                                   "a mirrored disk template")
5719

    
5720
    target_node = secondary_nodes[0]
5721
    _CheckNodeOnline(self, target_node)
5722
    _CheckNodeNotDrained(self, target_node)
5723
    if instance.admin_up:
5724
      # check memory requirements on the secondary node
5725
      _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5726
                           instance.name, bep[constants.BE_MEMORY],
5727
                           instance.hypervisor)
5728
    else:
5729
      self.LogInfo("Not checking memory on the secondary node as"
5730
                   " instance will not be started")
5731

    
5732
    # check bridge existance
5733
    _CheckInstanceBridgesExist(self, instance, node=target_node)
5734

    
5735
  def Exec(self, feedback_fn):
5736
    """Failover an instance.
5737

5738
    The failover is done by shutting it down on its present node and
5739
    starting it on the secondary.
5740

5741
    """
5742
    instance = self.instance
5743
    primary_node = self.cfg.GetNodeInfo(instance.primary_node)
5744

    
5745
    source_node = instance.primary_node
5746
    target_node = instance.secondary_nodes[0]
5747

    
5748
    if instance.admin_up:
5749
      feedback_fn("* checking disk consistency between source and target")
5750
      for dev in instance.disks:
5751
        # for drbd, these are drbd over lvm
5752
        if not _CheckDiskConsistency(self, dev, target_node, False):
5753
          if not self.op.ignore_consistency:
5754
            raise errors.OpExecError("Disk %s is degraded on target node,"
5755
                                     " aborting failover." % dev.iv_name)
5756
    else:
5757
      feedback_fn("* not checking disk consistency as instance is not running")
5758

    
5759
    feedback_fn("* shutting down instance on source node")
5760
    logging.info("Shutting down instance %s on node %s",
5761
                 instance.name, source_node)
5762

    
5763
    result = self.rpc.call_instance_shutdown(source_node, instance,
5764
                                             self.op.shutdown_timeout)
5765
    msg = result.fail_msg
5766
    if msg:
5767
      if self.op.ignore_consistency or primary_node.offline:
5768
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
5769
                             " Proceeding anyway. Please make sure node"
5770
                             " %s is down. Error details: %s",
5771
                             instance.name, source_node, source_node, msg)
5772
      else:
5773
        raise errors.OpExecError("Could not shutdown instance %s on"
5774
                                 " node %s: %s" %
5775
                                 (instance.name, source_node, msg))
5776

    
5777
    feedback_fn("* deactivating the instance's disks on source node")
5778
    if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
5779
      raise errors.OpExecError("Can't shut down the instance's disks.")
5780

    
5781
    instance.primary_node = target_node
5782
    # distribute new instance config to the other nodes
5783
    self.cfg.Update(instance, feedback_fn)
5784

    
5785
    # Only start the instance if it's marked as up
5786
    if instance.admin_up:
5787
      feedback_fn("* activating the instance's disks on target node")
5788
      logging.info("Starting instance %s on node %s",
5789
                   instance.name, target_node)
5790

    
5791
      disks_ok, _ = _AssembleInstanceDisks(self, instance,
5792
                                           ignore_secondaries=True)
5793
      if not disks_ok:
5794
        _ShutdownInstanceDisks(self, instance)
5795
        raise errors.OpExecError("Can't activate the instance's disks")
5796

    
5797
      feedback_fn("* starting the instance on the target node")
5798
      result = self.rpc.call_instance_start(target_node, instance, None, None)
5799
      msg = result.fail_msg
5800
      if msg:
5801
        _ShutdownInstanceDisks(self, instance)
5802
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
5803
                                 (instance.name, target_node, msg))
5804

    
5805

    
5806
class LUInstanceMigrate(LogicalUnit):
5807
  """Migrate an instance.
5808

5809
  This is migration without shutting down, compared to the failover,
5810
  which is done with shutdown.
5811

5812
  """
5813
  HPATH = "instance-migrate"
5814
  HTYPE = constants.HTYPE_INSTANCE
5815
  REQ_BGL = False
5816

    
5817
  def ExpandNames(self):
5818
    self._ExpandAndLockInstance()
5819

    
5820
    self.needed_locks[locking.LEVEL_NODE] = []
5821
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5822

    
5823
    self._migrater = TLMigrateInstance(self, self.op.instance_name,
5824
                                       self.op.cleanup)
5825
    self.tasklets = [self._migrater]
5826

    
5827
  def DeclareLocks(self, level):
5828
    if level == locking.LEVEL_NODE:
5829
      self._LockInstancesNodes()
5830

    
5831
  def BuildHooksEnv(self):
5832
    """Build hooks env.
5833

5834
    This runs on master, primary and secondary nodes of the instance.
5835

5836
    """
5837
    instance = self._migrater.instance
5838
    source_node = instance.primary_node
5839
    target_node = instance.secondary_nodes[0]
5840
    env = _BuildInstanceHookEnvByObject(self, instance)
5841
    env["MIGRATE_LIVE"] = self._migrater.live
5842
    env["MIGRATE_CLEANUP"] = self.op.cleanup
5843
    env.update({
5844
        "OLD_PRIMARY": source_node,
5845
        "OLD_SECONDARY": target_node,
5846
        "NEW_PRIMARY": target_node,
5847
        "NEW_SECONDARY": source_node,
5848
        })
5849
    nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5850
    nl_post = list(nl)
5851
    nl_post.append(source_node)
5852
    return env, nl, nl_post
5853

    
5854

    
5855
class LUInstanceMove(LogicalUnit):
5856
  """Move an instance by data-copying.
5857

5858
  """
5859
  HPATH = "instance-move"
5860
  HTYPE = constants.HTYPE_INSTANCE
5861
  REQ_BGL = False
5862

    
5863
  def ExpandNames(self):
5864
    self._ExpandAndLockInstance()
5865
    target_node = _ExpandNodeName(self.cfg, self.op.target_node)
5866
    self.op.target_node = target_node
5867
    self.needed_locks[locking.LEVEL_NODE] = [target_node]
5868
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5869

    
5870
  def DeclareLocks(self, level):
5871
    if level == locking.LEVEL_NODE:
5872
      self._LockInstancesNodes(primary_only=True)
5873

    
5874
  def BuildHooksEnv(self):
5875
    """Build hooks env.
5876

5877
    This runs on master, primary and secondary nodes of the instance.
5878

5879
    """
5880
    env = {
5881
      "TARGET_NODE": self.op.target_node,
5882
      "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5883
      }
5884
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5885
    nl = [self.cfg.GetMasterNode()] + [self.instance.primary_node,
5886
                                       self.op.target_node]
5887
    return env, nl, nl
5888

    
5889
  def CheckPrereq(self):
5890
    """Check prerequisites.
5891

5892
    This checks that the instance is in the cluster.
5893

5894
    """
5895
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5896
    assert self.instance is not None, \
5897
      "Cannot retrieve locked instance %s" % self.op.instance_name
5898

    
5899
    node = self.cfg.GetNodeInfo(self.op.target_node)
5900
    assert node is not None, \
5901
      "Cannot retrieve locked node %s" % self.op.target_node
5902

    
5903
    self.target_node = target_node = node.name
5904

    
5905
    if target_node == instance.primary_node:
5906
      raise errors.OpPrereqError("Instance %s is already on the node %s" %
5907
                                 (instance.name, target_node),
5908
                                 errors.ECODE_STATE)
5909

    
5910
    bep = self.cfg.GetClusterInfo().FillBE(instance)
5911

    
5912
    for idx, dsk in enumerate(instance.disks):
5913
      if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
5914
        raise errors.OpPrereqError("Instance disk %d has a complex layout,"
5915
                                   " cannot copy" % idx, errors.ECODE_STATE)
5916

    
5917
    _CheckNodeOnline(self, target_node)
5918
    _CheckNodeNotDrained(self, target_node)
5919
    _CheckNodeVmCapable(self, target_node)
5920

    
5921
    if instance.admin_up:
5922
      # check memory requirements on the secondary node
5923
      _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5924
                           instance.name, bep[constants.BE_MEMORY],
5925
                           instance.hypervisor)
5926
    else:
5927
      self.LogInfo("Not checking memory on the secondary node as"
5928
                   " instance will not be started")
5929

    
5930
    # check bridge existance
5931
    _CheckInstanceBridgesExist(self, instance, node=target_node)
5932

    
5933
  def Exec(self, feedback_fn):
5934
    """Move an instance.
5935

5936
    The move is done by shutting it down on its present node, copying
5937
    the data over (slow) and starting it on the new node.
5938

5939
    """
5940
    instance = self.instance
5941

    
5942
    source_node = instance.primary_node
5943
    target_node = self.target_node
5944

    
5945
    self.LogInfo("Shutting down instance %s on source node %s",
5946
                 instance.name, source_node)
5947

    
5948
    result = self.rpc.call_instance_shutdown(source_node, instance,
5949
                                             self.op.shutdown_timeout)
5950
    msg = result.fail_msg
5951
    if msg:
5952
      if self.op.ignore_consistency:
5953
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
5954
                             " Proceeding anyway. Please make sure node"
5955
                             " %s is down. Error details: %s",
5956
                             instance.name, source_node, source_node, msg)
5957
      else:
5958
        raise errors.OpExecError("Could not shutdown instance %s on"
5959
                                 " node %s: %s" %
5960
                                 (instance.name, source_node, msg))
5961

    
5962
    # create the target disks
5963
    try:
5964
      _CreateDisks(self, instance, target_node=target_node)
5965
    except errors.OpExecError:
5966
      self.LogWarning("Device creation failed, reverting...")
5967
      try:
5968
        _RemoveDisks(self, instance, target_node=target_node)
5969
      finally:
5970
        self.cfg.ReleaseDRBDMinors(instance.name)
5971
        raise
5972

    
5973
    cluster_name = self.cfg.GetClusterInfo().cluster_name
5974

    
5975
    errs = []
5976
    # activate, get path, copy the data over
5977
    for idx, disk in enumerate(instance.disks):
5978
      self.LogInfo("Copying data for disk %d", idx)
5979
      result = self.rpc.call_blockdev_assemble(target_node, disk,
5980
                                               instance.name, True, idx)
5981
      if result.fail_msg:
5982
        self.LogWarning("Can't assemble newly created disk %d: %s",
5983
                        idx, result.fail_msg)
5984
        errs.append(result.fail_msg)
5985
        break
5986
      dev_path = result.payload
5987
      result = self.rpc.call_blockdev_export(source_node, disk,
5988
                                             target_node, dev_path,
5989
                                             cluster_name)
5990
      if result.fail_msg:
5991
        self.LogWarning("Can't copy data over for disk %d: %s",
5992
                        idx, result.fail_msg)
5993
        errs.append(result.fail_msg)
5994
        break
5995

    
5996
    if errs:
5997
      self.LogWarning("Some disks failed to copy, aborting")
5998
      try:
5999
        _RemoveDisks(self, instance, target_node=target_node)
6000
      finally:
6001
        self.cfg.ReleaseDRBDMinors(instance.name)
6002
        raise errors.OpExecError("Errors during disk copy: %s" %
6003
                                 (",".join(errs),))
6004

    
6005
    instance.primary_node = target_node
6006
    self.cfg.Update(instance, feedback_fn)
6007

    
6008
    self.LogInfo("Removing the disks on the original node")
6009
    _RemoveDisks(self, instance, target_node=source_node)
6010

    
6011
    # Only start the instance if it's marked as up
6012
    if instance.admin_up:
6013
      self.LogInfo("Starting instance %s on node %s",
6014
                   instance.name, target_node)
6015

    
6016
      disks_ok, _ = _AssembleInstanceDisks(self, instance,
6017
                                           ignore_secondaries=True)
6018
      if not disks_ok:
6019
        _ShutdownInstanceDisks(self, instance)
6020
        raise errors.OpExecError("Can't activate the instance's disks")
6021

    
6022
      result = self.rpc.call_instance_start(target_node, instance, None, None)
6023
      msg = result.fail_msg
6024
      if msg:
6025
        _ShutdownInstanceDisks(self, instance)
6026
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
6027
                                 (instance.name, target_node, msg))
6028

    
6029

    
6030
class LUNodeMigrate(LogicalUnit):
6031
  """Migrate all instances from a node.
6032

6033
  """
6034
  HPATH = "node-migrate"
6035
  HTYPE = constants.HTYPE_NODE
6036
  REQ_BGL = False
6037

    
6038
  def ExpandNames(self):
6039
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
6040

    
6041
    self.needed_locks = {
6042
      locking.LEVEL_NODE: [self.op.node_name],
6043
      }
6044

    
6045
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
6046

    
6047
    # Create tasklets for migrating instances for all instances on this node
6048
    names = []
6049
    tasklets = []
6050

    
6051
    for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name):
6052
      logging.debug("Migrating instance %s", inst.name)
6053
      names.append(inst.name)
6054

    
6055
      tasklets.append(TLMigrateInstance(self, inst.name, False))
6056

    
6057
    self.tasklets = tasklets
6058

    
6059
    # Declare instance locks
6060
    self.needed_locks[locking.LEVEL_INSTANCE] = names
6061

    
6062
  def DeclareLocks(self, level):
6063
    if level == locking.LEVEL_NODE:
6064
      self._LockInstancesNodes()
6065

    
6066
  def BuildHooksEnv(self):
6067
    """Build hooks env.
6068

6069
    This runs on the master, the primary and all the secondaries.
6070

6071
    """
6072
    env = {
6073
      "NODE_NAME": self.op.node_name,
6074
      }
6075

    
6076
    nl = [self.cfg.GetMasterNode()]
6077

    
6078
    return (env, nl, nl)
6079

    
6080

    
6081
class TLMigrateInstance(Tasklet):
6082
  """Tasklet class for instance migration.
6083

6084
  @type live: boolean
6085
  @ivar live: whether the migration will be done live or non-live;
6086
      this variable is initalized only after CheckPrereq has run
6087

6088
  """
6089
  def __init__(self, lu, instance_name, cleanup):
6090
    """Initializes this class.
6091

6092
    """
6093
    Tasklet.__init__(self, lu)
6094

    
6095
    # Parameters
6096
    self.instance_name = instance_name
6097
    self.cleanup = cleanup
6098
    self.live = False # will be overridden later
6099

    
6100
  def CheckPrereq(self):
6101
    """Check prerequisites.
6102

6103
    This checks that the instance is in the cluster.
6104

6105
    """
6106
    instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name)
6107
    instance = self.cfg.GetInstanceInfo(instance_name)
6108
    assert instance is not None
6109

    
6110
    if instance.disk_template != constants.DT_DRBD8:
6111
      raise errors.OpPrereqError("Instance's disk layout is not"
6112
                                 " drbd8, cannot migrate.", errors.ECODE_STATE)
6113

    
6114
    secondary_nodes = instance.secondary_nodes
6115
    if not secondary_nodes:
6116
      raise errors.ConfigurationError("No secondary node but using"
6117
                                      " drbd8 disk template")
6118

    
6119
    i_be = self.cfg.GetClusterInfo().FillBE(instance)
6120

    
6121
    target_node = secondary_nodes[0]
6122
    # check memory requirements on the secondary node
6123
    _CheckNodeFreeMemory(self.lu, target_node, "migrating instance %s" %
6124
                         instance.name, i_be[constants.BE_MEMORY],
6125
                         instance.hypervisor)
6126

    
6127
    # check bridge existance
6128
    _CheckInstanceBridgesExist(self.lu, instance, node=target_node)
6129

    
6130
    if not self.cleanup:
6131
      _CheckNodeNotDrained(self.lu, target_node)
6132
      result = self.rpc.call_instance_migratable(instance.primary_node,
6133
                                                 instance)
6134
      result.Raise("Can't migrate, please use failover",
6135
                   prereq=True, ecode=errors.ECODE_STATE)
6136

    
6137
    self.instance = instance
6138

    
6139
    if self.lu.op.live is not None and self.lu.op.mode is not None:
6140
      raise errors.OpPrereqError("Only one of the 'live' and 'mode'"
6141
                                 " parameters are accepted",
6142
                                 errors.ECODE_INVAL)
6143
    if self.lu.op.live is not None:
6144
      if self.lu.op.live:
6145
        self.lu.op.mode = constants.HT_MIGRATION_LIVE
6146
      else:
6147
        self.lu.op.mode = constants.HT_MIGRATION_NONLIVE
6148
      # reset the 'live' parameter to None so that repeated
6149
      # invocations of CheckPrereq do not raise an exception
6150
      self.lu.op.live = None
6151
    elif self.lu.op.mode is None:
6152
      # read the default value from the hypervisor
6153
      i_hv = self.cfg.GetClusterInfo().FillHV(instance, skip_globals=False)
6154
      self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE]
6155

    
6156
    self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE
6157

    
6158
  def _WaitUntilSync(self):
6159
    """Poll with custom rpc for disk sync.
6160

6161
    This uses our own step-based rpc call.
6162

6163
    """
6164
    self.feedback_fn("* wait until resync is done")
6165
    all_done = False
6166
    while not all_done:
6167
      all_done = True
6168
      result = self.rpc.call_drbd_wait_sync(self.all_nodes,
6169
                                            self.nodes_ip,
6170
                                            self.instance.disks)
6171
      min_percent = 100
6172
      for node, nres in result.items():
6173
        nres.Raise("Cannot resync disks on node %s" % node)
6174
        node_done, node_percent = nres.payload
6175
        all_done = all_done and node_done
6176
        if node_percent is not None:
6177
          min_percent = min(min_percent, node_percent)
6178
      if not all_done:
6179
        if min_percent < 100:
6180
          self.feedback_fn("   - progress: %.1f%%" % min_percent)
6181
        time.sleep(2)
6182

    
6183
  def _EnsureSecondary(self, node):
6184
    """Demote a node to secondary.
6185

6186
    """
6187
    self.feedback_fn("* switching node %s to secondary mode" % node)
6188

    
6189
    for dev in self.instance.disks:
6190
      self.cfg.SetDiskID(dev, node)
6191

    
6192
    result = self.rpc.call_blockdev_close(node, self.instance.name,
6193
                                          self.instance.disks)
6194
    result.Raise("Cannot change disk to secondary on node %s" % node)
6195

    
6196
  def _GoStandalone(self):
6197
    """Disconnect from the network.
6198

6199
    """
6200
    self.feedback_fn("* changing into standalone mode")
6201
    result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
6202
                                               self.instance.disks)
6203
    for node, nres in result.items():
6204
      nres.Raise("Cannot disconnect disks node %s" % node)
6205

    
6206
  def _GoReconnect(self, multimaster):
6207
    """Reconnect to the network.
6208

6209
    """
6210
    if multimaster:
6211
      msg = "dual-master"
6212
    else:
6213
      msg = "single-master"
6214
    self.feedback_fn("* changing disks into %s mode" % msg)
6215
    result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
6216
                                           self.instance.disks,
6217
                                           self.instance.name, multimaster)
6218
    for node, nres in result.items():
6219
      nres.Raise("Cannot change disks config on node %s" % node)
6220

    
6221
  def _ExecCleanup(self):
6222
    """Try to cleanup after a failed migration.
6223

6224
    The cleanup is done by:
6225
      - check that the instance is running only on one node
6226
        (and update the config if needed)
6227
      - change disks on its secondary node to secondary
6228
      - wait until disks are fully synchronized
6229
      - disconnect from the network
6230
      - change disks into single-master mode
6231
      - wait again until disks are fully synchronized
6232

6233
    """
6234
    instance = self.instance
6235
    target_node = self.target_node
6236
    source_node = self.source_node
6237

    
6238
    # check running on only one node
6239
    self.feedback_fn("* checking where the instance actually runs"
6240
                     " (if this hangs, the hypervisor might be in"
6241
                     " a bad state)")
6242
    ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
6243
    for node, result in ins_l.items():
6244
      result.Raise("Can't contact node %s" % node)
6245

    
6246
    runningon_source = instance.name in ins_l[source_node].payload
6247
    runningon_target = instance.name in ins_l[target_node].payload
6248

    
6249
    if runningon_source and runningon_target:
6250
      raise errors.OpExecError("Instance seems to be running on two nodes,"
6251
                               " or the hypervisor is confused. You will have"
6252
                               " to ensure manually that it runs only on one"
6253
                               " and restart this operation.")
6254

    
6255
    if not (runningon_source or runningon_target):
6256
      raise errors.OpExecError("Instance does not seem to be running at all."
6257
                               " In this case, it's safer to repair by"
6258
                               " running 'gnt-instance stop' to ensure disk"
6259
                               " shutdown, and then restarting it.")
6260

    
6261
    if runningon_target:
6262
      # the migration has actually succeeded, we need to update the config
6263
      self.feedback_fn("* instance running on secondary node (%s),"
6264
                       " updating config" % target_node)
6265
      instance.primary_node = target_node
6266
      self.cfg.Update(instance, self.feedback_fn)
6267
      demoted_node = source_node
6268
    else:
6269
      self.feedback_fn("* instance confirmed to be running on its"
6270
                       " primary node (%s)" % source_node)
6271
      demoted_node = target_node
6272

    
6273
    self._EnsureSecondary(demoted_node)
6274
    try:
6275
      self._WaitUntilSync()
6276
    except errors.OpExecError:
6277
      # we ignore here errors, since if the device is standalone, it
6278
      # won't be able to sync
6279
      pass
6280
    self._GoStandalone()
6281
    self._GoReconnect(False)
6282
    self._WaitUntilSync()
6283

    
6284
    self.feedback_fn("* done")
6285

    
6286
  def _RevertDiskStatus(self):
6287
    """Try to revert the disk status after a failed migration.
6288

6289
    """
6290
    target_node = self.target_node
6291
    try:
6292
      self._EnsureSecondary(target_node)
6293
      self._GoStandalone()
6294
      self._GoReconnect(False)
6295
      self._WaitUntilSync()
6296
    except errors.OpExecError, err:
6297
      self.lu.LogWarning("Migration failed and I can't reconnect the"
6298
                         " drives: error '%s'\n"
6299
                         "Please look and recover the instance status" %
6300
                         str(err))
6301

    
6302
  def _AbortMigration(self):
6303
    """Call the hypervisor code to abort a started migration.
6304

6305
    """
6306
    instance = self.instance
6307
    target_node = self.target_node
6308
    migration_info = self.migration_info
6309

    
6310
    abort_result = self.rpc.call_finalize_migration(target_node,
6311
                                                    instance,
6312
                                                    migration_info,
6313
                                                    False)
6314
    abort_msg = abort_result.fail_msg
6315
    if abort_msg:
6316
      logging.error("Aborting migration failed on target node %s: %s",
6317
                    target_node, abort_msg)
6318
      # Don't raise an exception here, as we stil have to try to revert the
6319
      # disk status, even if this step failed.
6320

    
6321
  def _ExecMigration(self):
6322
    """Migrate an instance.
6323

6324
    The migrate is done by:
6325
      - change the disks into dual-master mode
6326
      - wait until disks are fully synchronized again
6327
      - migrate the instance
6328
      - change disks on the new secondary node (the old primary) to secondary
6329
      - wait until disks are fully synchronized
6330
      - change disks into single-master mode
6331

6332
    """
6333
    instance = self.instance
6334
    target_node = self.target_node
6335
    source_node = self.source_node
6336

    
6337
    self.feedback_fn("* checking disk consistency between source and target")
6338
    for dev in instance.disks:
6339
      if not _CheckDiskConsistency(self.lu, dev, target_node, False):
6340
        raise errors.OpExecError("Disk %s is degraded or not fully"
6341
                                 " synchronized on target node,"
6342
                                 " aborting migrate." % dev.iv_name)
6343

    
6344
    # First get the migration information from the remote node
6345
    result = self.rpc.call_migration_info(source_node, instance)
6346
    msg = result.fail_msg
6347
    if msg:
6348
      log_err = ("Failed fetching source migration information from %s: %s" %
6349
                 (source_node, msg))
6350
      logging.error(log_err)
6351
      raise errors.OpExecError(log_err)
6352

    
6353
    self.migration_info = migration_info = result.payload
6354

    
6355
    # Then switch the disks to master/master mode
6356
    self._EnsureSecondary(target_node)
6357
    self._GoStandalone()
6358
    self._GoReconnect(True)
6359
    self._WaitUntilSync()
6360

    
6361
    self.feedback_fn("* preparing %s to accept the instance" % target_node)
6362
    result = self.rpc.call_accept_instance(target_node,
6363
                                           instance,
6364
                                           migration_info,
6365
                                           self.nodes_ip[target_node])
6366

    
6367
    msg = result.fail_msg
6368
    if msg:
6369
      logging.error("Instance pre-migration failed, trying to revert"
6370
                    " disk status: %s", msg)
6371
      self.feedback_fn("Pre-migration failed, aborting")
6372
      self._AbortMigration()
6373
      self._RevertDiskStatus()
6374
      raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
6375
                               (instance.name, msg))
6376

    
6377
    self.feedback_fn("* migrating instance to %s" % target_node)
6378
    time.sleep(10)
6379
    result = self.rpc.call_instance_migrate(source_node, instance,
6380
                                            self.nodes_ip[target_node],
6381
                                            self.live)
6382
    msg = result.fail_msg
6383
    if msg:
6384
      logging.error("Instance migration failed, trying to revert"
6385
                    " disk status: %s", msg)
6386
      self.feedback_fn("Migration failed, aborting")
6387
      self._AbortMigration()
6388
      self._RevertDiskStatus()
6389
      raise errors.OpExecError("Could not migrate instance %s: %s" %
6390
                               (instance.name, msg))
6391
    time.sleep(10)
6392

    
6393
    instance.primary_node = target_node
6394
    # distribute new instance config to the other nodes
6395
    self.cfg.Update(instance, self.feedback_fn)
6396

    
6397
    result = self.rpc.call_finalize_migration(target_node,
6398
                                              instance,
6399
                                              migration_info,
6400
                                              True)
6401
    msg = result.fail_msg
6402
    if msg:
6403
      logging.error("Instance migration succeeded, but finalization failed:"
6404
                    " %s", msg)
6405
      raise errors.OpExecError("Could not finalize instance migration: %s" %
6406
                               msg)
6407

    
6408
    self._EnsureSecondary(source_node)
6409
    self._WaitUntilSync()
6410
    self._GoStandalone()
6411
    self._GoReconnect(False)
6412
    self._WaitUntilSync()
6413

    
6414
    self.feedback_fn("* done")
6415

    
6416
  def Exec(self, feedback_fn):
6417
    """Perform the migration.
6418

6419
    """
6420
    feedback_fn("Migrating instance %s" % self.instance.name)
6421

    
6422
    self.feedback_fn = feedback_fn
6423

    
6424
    self.source_node = self.instance.primary_node
6425
    self.target_node = self.instance.secondary_nodes[0]
6426
    self.all_nodes = [self.source_node, self.target_node]
6427
    self.nodes_ip = {
6428
      self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
6429
      self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
6430
      }
6431

    
6432
    if self.cleanup:
6433
      return self._ExecCleanup()
6434
    else:
6435
      return self._ExecMigration()
6436

    
6437

    
6438
def _CreateBlockDev(lu, node, instance, device, force_create,
6439
                    info, force_open):
6440
  """Create a tree of block devices on a given node.
6441

6442
  If this device type has to be created on secondaries, create it and
6443
  all its children.
6444

6445
  If not, just recurse to children keeping the same 'force' value.
6446

6447
  @param lu: the lu on whose behalf we execute
6448
  @param node: the node on which to create the device
6449
  @type instance: L{objects.Instance}
6450
  @param instance: the instance which owns the device
6451
  @type device: L{objects.Disk}
6452
  @param device: the device to create
6453
  @type force_create: boolean
6454
  @param force_create: whether to force creation of this device; this
6455
      will be change to True whenever we find a device which has
6456
      CreateOnSecondary() attribute
6457
  @param info: the extra 'metadata' we should attach to the device
6458
      (this will be represented as a LVM tag)
6459
  @type force_open: boolean
6460
  @param force_open: this parameter will be passes to the
6461
      L{backend.BlockdevCreate} function where it specifies
6462
      whether we run on primary or not, and it affects both
6463
      the child assembly and the device own Open() execution
6464

6465
  """
6466
  if device.CreateOnSecondary():
6467
    force_create = True
6468

    
6469
  if device.children:
6470
    for child in device.children:
6471
      _CreateBlockDev(lu, node, instance, child, force_create,
6472
                      info, force_open)
6473

    
6474
  if not force_create:
6475
    return
6476

    
6477
  _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
6478

    
6479

    
6480
def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
6481
  """Create a single block device on a given node.
6482

6483
  This will not recurse over children of the device, so they must be
6484
  created in advance.
6485

6486
  @param lu: the lu on whose behalf we execute
6487
  @param node: the node on which to create the device
6488
  @type instance: L{objects.Instance}
6489
  @param instance: the instance which owns the device
6490
  @type device: L{objects.Disk}
6491
  @param device: the device to create
6492
  @param info: the extra 'metadata' we should attach to the device
6493
      (this will be represented as a LVM tag)
6494
  @type force_open: boolean
6495
  @param force_open: this parameter will be passes to the
6496
      L{backend.BlockdevCreate} function where it specifies
6497
      whether we run on primary or not, and it affects both
6498
      the child assembly and the device own Open() execution
6499

6500
  """
6501
  lu.cfg.SetDiskID(device, node)
6502
  result = lu.rpc.call_blockdev_create(node, device, device.size,
6503
                                       instance.name, force_open, info)
6504
  result.Raise("Can't create block device %s on"
6505
               " node %s for instance %s" % (device, node, instance.name))
6506
  if device.physical_id is None:
6507
    device.physical_id = result.payload
6508

    
6509

    
6510
def _GenerateUniqueNames(lu, exts):
6511
  """Generate a suitable LV name.
6512

6513
  This will generate a logical volume name for the given instance.
6514

6515
  """
6516
  results = []
6517
  for val in exts:
6518
    new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
6519
    results.append("%s%s" % (new_id, val))
6520
  return results
6521

    
6522

    
6523
def _GenerateDRBD8Branch(lu, primary, secondary, size, vgname, names, iv_name,
6524
                         p_minor, s_minor):
6525
  """Generate a drbd8 device complete with its children.
6526

6527
  """
6528
  port = lu.cfg.AllocatePort()
6529
  shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
6530
  dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
6531
                          logical_id=(vgname, names[0]))
6532
  dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
6533
                          logical_id=(vgname, names[1]))
6534
  drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
6535
                          logical_id=(primary, secondary, port,
6536
                                      p_minor, s_minor,
6537
                                      shared_secret),
6538
                          children=[dev_data, dev_meta],
6539
                          iv_name=iv_name)
6540
  return drbd_dev
6541

    
6542

    
6543
def _GenerateDiskTemplate(lu, template_name,
6544
                          instance_name, primary_node,
6545
                          secondary_nodes, disk_info,
6546
                          file_storage_dir, file_driver,
6547
                          base_index, feedback_fn):
6548
  """Generate the entire disk layout for a given template type.
6549

6550
  """
6551
  #TODO: compute space requirements
6552

    
6553
  vgname = lu.cfg.GetVGName()
6554
  disk_count = len(disk_info)
6555
  disks = []
6556
  if template_name == constants.DT_DISKLESS:
6557
    pass
6558
  elif template_name == constants.DT_PLAIN:
6559
    if len(secondary_nodes) != 0:
6560
      raise errors.ProgrammerError("Wrong template configuration")
6561

    
6562
    names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
6563
                                      for i in range(disk_count)])
6564
    for idx, disk in enumerate(disk_info):
6565
      disk_index = idx + base_index
6566
      vg = disk.get("vg", vgname)
6567
      feedback_fn("* disk %i, vg %s, name %s" % (idx, vg, names[idx]))
6568
      disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
6569
                              logical_id=(vg, names[idx]),
6570
                              iv_name="disk/%d" % disk_index,
6571
                              mode=disk["mode"])
6572
      disks.append(disk_dev)
6573
  elif template_name == constants.DT_DRBD8:
6574
    if len(secondary_nodes) != 1:
6575
      raise errors.ProgrammerError("Wrong template configuration")
6576
    remote_node = secondary_nodes[0]
6577
    minors = lu.cfg.AllocateDRBDMinor(
6578
      [primary_node, remote_node] * len(disk_info), instance_name)
6579

    
6580
    names = []
6581
    for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
6582
                                               for i in range(disk_count)]):
6583
      names.append(lv_prefix + "_data")
6584
      names.append(lv_prefix + "_meta")
6585
    for idx, disk in enumerate(disk_info):
6586
      disk_index = idx + base_index
6587
      vg = disk.get("vg", vgname)
6588
      disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
6589
                                      disk["size"], vg, names[idx*2:idx*2+2],
6590
                                      "disk/%d" % disk_index,
6591
                                      minors[idx*2], minors[idx*2+1])
6592
      disk_dev.mode = disk["mode"]
6593
      disks.append(disk_dev)
6594
  elif template_name == constants.DT_FILE:
6595
    if len(secondary_nodes) != 0:
6596
      raise errors.ProgrammerError("Wrong template configuration")
6597

    
6598
    opcodes.RequireFileStorage()
6599

    
6600
    for idx, disk in enumerate(disk_info):
6601
      disk_index = idx + base_index
6602
      disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
6603
                              iv_name="disk/%d" % disk_index,
6604
                              logical_id=(file_driver,
6605
                                          "%s/disk%d" % (file_storage_dir,
6606
                                                         disk_index)),
6607
                              mode=disk["mode"])
6608
      disks.append(disk_dev)
6609
  else:
6610
    raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
6611
  return disks
6612

    
6613

    
6614
def _GetInstanceInfoText(instance):
6615
  """Compute that text that should be added to the disk's metadata.
6616

6617
  """
6618
  return "originstname+%s" % instance.name
6619

    
6620

    
6621
def _CalcEta(time_taken, written, total_size):
6622
  """Calculates the ETA based on size written and total size.
6623

6624
  @param time_taken: The time taken so far
6625
  @param written: amount written so far
6626
  @param total_size: The total size of data to be written
6627
  @return: The remaining time in seconds
6628

6629
  """
6630
  avg_time = time_taken / float(written)
6631
  return (total_size - written) * avg_time
6632

    
6633

    
6634
def _WipeDisks(lu, instance):
6635
  """Wipes instance disks.
6636

6637
  @type lu: L{LogicalUnit}
6638
  @param lu: the logical unit on whose behalf we execute
6639
  @type instance: L{objects.Instance}
6640
  @param instance: the instance whose disks we should create
6641
  @return: the success of the wipe
6642

6643
  """
6644
  node = instance.primary_node
6645

    
6646
  for device in instance.disks:
6647
    lu.cfg.SetDiskID(device, node)
6648

    
6649
  logging.info("Pause sync of instance %s disks", instance.name)
6650
  result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, True)
6651

    
6652
  for idx, success in enumerate(result.payload):
6653
    if not success:
6654
      logging.warn("pause-sync of instance %s for disks %d failed",
6655
                   instance.name, idx)
6656

    
6657
  try:
6658
    for idx, device in enumerate(instance.disks):
6659
      lu.LogInfo("* Wiping disk %d", idx)
6660
      logging.info("Wiping disk %d for instance %s, node %s",
6661
                   idx, instance.name, node)
6662

    
6663
      # The wipe size is MIN_WIPE_CHUNK_PERCENT % of the instance disk but
6664
      # MAX_WIPE_CHUNK at max
6665
      wipe_chunk_size = min(constants.MAX_WIPE_CHUNK, device.size / 100.0 *
6666
                            constants.MIN_WIPE_CHUNK_PERCENT)
6667

    
6668
      offset = 0
6669
      size = device.size
6670
      last_output = 0
6671
      start_time = time.time()
6672

    
6673
      while offset < size:
6674
        wipe_size = min(wipe_chunk_size, size - offset)
6675
        result = lu.rpc.call_blockdev_wipe(node, device, offset, wipe_size)
6676
        result.Raise("Could not wipe disk %d at offset %d for size %d" %
6677
                     (idx, offset, wipe_size))
6678
        now = time.time()
6679
        offset += wipe_size
6680
        if now - last_output >= 60:
6681
          eta = _CalcEta(now - start_time, offset, size)
6682
          lu.LogInfo(" - done: %.1f%% ETA: %s" %
6683
                     (offset / float(size) * 100, utils.FormatSeconds(eta)))
6684
          last_output = now
6685
  finally:
6686
    logging.info("Resume sync of instance %s disks", instance.name)
6687

    
6688
    result = lu.rpc.call_blockdev_pause_resume_sync(node, instance.disks, False)
6689

    
6690
    for idx, success in enumerate(result.payload):
6691
      if not success:
6692
        lu.LogWarning("Warning: Resume sync of disk %d failed. Please have a"
6693
                      " look at the status and troubleshoot the issue.", idx)
6694
        logging.warn("resume-sync of instance %s for disks %d failed",
6695
                     instance.name, idx)
6696

    
6697

    
6698
def _CreateDisks(lu, instance, to_skip=None, target_node=None):
6699
  """Create all disks for an instance.
6700

6701
  This abstracts away some work from AddInstance.
6702

6703
  @type lu: L{LogicalUnit}
6704
  @param lu: the logical unit on whose behalf we execute
6705
  @type instance: L{objects.Instance}
6706
  @param instance: the instance whose disks we should create
6707
  @type to_skip: list
6708
  @param to_skip: list of indices to skip
6709
  @type target_node: string
6710
  @param target_node: if passed, overrides the target node for creation
6711
  @rtype: boolean
6712
  @return: the success of the creation
6713

6714
  """
6715
  info = _GetInstanceInfoText(instance)
6716
  if target_node is None:
6717
    pnode = instance.primary_node
6718
    all_nodes = instance.all_nodes
6719
  else:
6720
    pnode = target_node
6721
    all_nodes = [pnode]
6722

    
6723
  if instance.disk_template == constants.DT_FILE:
6724
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
6725
    result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
6726

    
6727
    result.Raise("Failed to create directory '%s' on"
6728
                 " node %s" % (file_storage_dir, pnode))
6729

    
6730
  # Note: this needs to be kept in sync with adding of disks in
6731
  # LUInstanceSetParams
6732
  for idx, device in enumerate(instance.disks):
6733
    if to_skip and idx in to_skip:
6734
      continue
6735
    logging.info("Creating volume %s for instance %s",
6736
                 device.iv_name, instance.name)
6737
    #HARDCODE
6738
    for node in all_nodes:
6739
      f_create = node == pnode
6740
      _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
6741

    
6742

    
6743
def _RemoveDisks(lu, instance, target_node=None):
6744
  """Remove all disks for an instance.
6745

6746
  This abstracts away some work from `AddInstance()` and
6747
  `RemoveInstance()`. Note that in case some of the devices couldn't
6748
  be removed, the removal will continue with the other ones (compare
6749
  with `_CreateDisks()`).
6750

6751
  @type lu: L{LogicalUnit}
6752
  @param lu: the logical unit on whose behalf we execute
6753
  @type instance: L{objects.Instance}
6754
  @param instance: the instance whose disks we should remove
6755
  @type target_node: string
6756
  @param target_node: used to override the node on which to remove the disks
6757
  @rtype: boolean
6758
  @return: the success of the removal
6759

6760
  """
6761
  logging.info("Removing block devices for instance %s", instance.name)
6762

    
6763
  all_result = True
6764
  for device in instance.disks:
6765
    if target_node:
6766
      edata = [(target_node, device)]
6767
    else:
6768
      edata = device.ComputeNodeTree(instance.primary_node)
6769
    for node, disk in edata:
6770
      lu.cfg.SetDiskID(disk, node)
6771
      msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
6772
      if msg:
6773
        lu.LogWarning("Could not remove block device %s on node %s,"
6774
                      " continuing anyway: %s", device.iv_name, node, msg)
6775
        all_result = False
6776

    
6777
  if instance.disk_template == constants.DT_FILE:
6778
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
6779
    if target_node:
6780
      tgt = target_node
6781
    else:
6782
      tgt = instance.primary_node
6783
    result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
6784
    if result.fail_msg:
6785
      lu.LogWarning("Could not remove directory '%s' on node %s: %s",
6786
                    file_storage_dir, instance.primary_node, result.fail_msg)
6787
      all_result = False
6788

    
6789
  return all_result
6790

    
6791

    
6792
def _ComputeDiskSizePerVG(disk_template, disks):
6793
  """Compute disk size requirements in the volume group
6794

6795
  """
6796
  def _compute(disks, payload):
6797
    """Universal algorithm
6798

6799
    """
6800
    vgs = {}
6801
    for disk in disks:
6802
      vgs[disk["vg"]] = vgs.get("vg", 0) + disk["size"] + payload
6803

    
6804
    return vgs
6805

    
6806
  # Required free disk space as a function of disk and swap space
6807
  req_size_dict = {
6808
    constants.DT_DISKLESS: {},
6809
    constants.DT_PLAIN: _compute(disks, 0),
6810
    # 128 MB are added for drbd metadata for each disk
6811
    constants.DT_DRBD8: _compute(disks, 128),
6812
    constants.DT_FILE: {},
6813
  }
6814

    
6815
  if disk_template not in req_size_dict:
6816
    raise errors.ProgrammerError("Disk template '%s' size requirement"
6817
                                 " is unknown" %  disk_template)
6818

    
6819
  return req_size_dict[disk_template]
6820

    
6821

    
6822
def _ComputeDiskSize(disk_template, disks):
6823
  """Compute disk size requirements in the volume group
6824

6825
  """
6826
  # Required free disk space as a function of disk and swap space
6827
  req_size_dict = {
6828
    constants.DT_DISKLESS: None,
6829
    constants.DT_PLAIN: sum(d["size"] for d in disks),
6830
    # 128 MB are added for drbd metadata for each disk
6831
    constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
6832
    constants.DT_FILE: None,
6833
  }
6834

    
6835
  if disk_template not in req_size_dict:
6836
    raise errors.ProgrammerError("Disk template '%s' size requirement"
6837
                                 " is unknown" %  disk_template)
6838

    
6839
  return req_size_dict[disk_template]
6840

    
6841

    
6842
def _FilterVmNodes(lu, nodenames):
6843
  """Filters out non-vm_capable nodes from a list.
6844

6845
  @type lu: L{LogicalUnit}
6846
  @param lu: the logical unit for which we check
6847
  @type nodenames: list
6848
  @param nodenames: the list of nodes on which we should check
6849
  @rtype: list
6850
  @return: the list of vm-capable nodes
6851

6852
  """
6853
  vm_nodes = frozenset(lu.cfg.GetNonVmCapableNodeList())
6854
  return [name for name in nodenames if name not in vm_nodes]
6855

    
6856

    
6857
def _CheckHVParams(lu, nodenames, hvname, hvparams):
6858
  """Hypervisor parameter validation.
6859

6860
  This function abstract the hypervisor parameter validation to be
6861
  used in both instance create and instance modify.
6862

6863
  @type lu: L{LogicalUnit}
6864
  @param lu: the logical unit for which we check
6865
  @type nodenames: list
6866
  @param nodenames: the list of nodes on which we should check
6867
  @type hvname: string
6868
  @param hvname: the name of the hypervisor we should use
6869
  @type hvparams: dict
6870
  @param hvparams: the parameters which we need to check
6871
  @raise errors.OpPrereqError: if the parameters are not valid
6872

6873
  """
6874
  nodenames = _FilterVmNodes(lu, nodenames)
6875
  hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
6876
                                                  hvname,
6877
                                                  hvparams)
6878
  for node in nodenames:
6879
    info = hvinfo[node]
6880
    if info.offline:
6881
      continue
6882
    info.Raise("Hypervisor parameter validation failed on node %s" % node)
6883

    
6884

    
6885
def _CheckOSParams(lu, required, nodenames, osname, osparams):
6886
  """OS parameters validation.
6887

6888
  @type lu: L{LogicalUnit}
6889
  @param lu: the logical unit for which we check
6890
  @type required: boolean
6891
  @param required: whether the validation should fail if the OS is not
6892
      found
6893
  @type nodenames: list
6894
  @param nodenames: the list of nodes on which we should check
6895
  @type osname: string
6896
  @param osname: the name of the hypervisor we should use
6897
  @type osparams: dict
6898
  @param osparams: the parameters which we need to check
6899
  @raise errors.OpPrereqError: if the parameters are not valid
6900

6901
  """
6902
  nodenames = _FilterVmNodes(lu, nodenames)
6903
  result = lu.rpc.call_os_validate(required, nodenames, osname,
6904
                                   [constants.OS_VALIDATE_PARAMETERS],
6905
                                   osparams)
6906
  for node, nres in result.items():
6907
    # we don't check for offline cases since this should be run only
6908
    # against the master node and/or an instance's nodes
6909
    nres.Raise("OS Parameters validation failed on node %s" % node)
6910
    if not nres.payload:
6911
      lu.LogInfo("OS %s not found on node %s, validation skipped",
6912
                 osname, node)
6913

    
6914

    
6915
class LUInstanceCreate(LogicalUnit):
6916
  """Create an instance.
6917

6918
  """
6919
  HPATH = "instance-add"
6920
  HTYPE = constants.HTYPE_INSTANCE
6921
  REQ_BGL = False
6922

    
6923
  def CheckArguments(self):
6924
    """Check arguments.
6925

6926
    """
6927
    # do not require name_check to ease forward/backward compatibility
6928
    # for tools
6929
    if self.op.no_install and self.op.start:
6930
      self.LogInfo("No-installation mode selected, disabling startup")
6931
      self.op.start = False
6932
    # validate/normalize the instance name
6933
    self.op.instance_name = \
6934
      netutils.Hostname.GetNormalizedName(self.op.instance_name)
6935

    
6936
    if self.op.ip_check and not self.op.name_check:
6937
      # TODO: make the ip check more flexible and not depend on the name check
6938
      raise errors.OpPrereqError("Cannot do ip check without a name check",
6939
                                 errors.ECODE_INVAL)
6940

    
6941
    # check nics' parameter names
6942
    for nic in self.op.nics:
6943
      utils.ForceDictType(nic, constants.INIC_PARAMS_TYPES)
6944

    
6945
    # check disks. parameter names and consistent adopt/no-adopt strategy
6946
    has_adopt = has_no_adopt = False
6947
    for disk in self.op.disks:
6948
      utils.ForceDictType(disk, constants.IDISK_PARAMS_TYPES)
6949
      if "adopt" in disk:
6950
        has_adopt = True
6951
      else:
6952
        has_no_adopt = True
6953
    if has_adopt and has_no_adopt:
6954
      raise errors.OpPrereqError("Either all disks are adopted or none is",
6955
                                 errors.ECODE_INVAL)
6956
    if has_adopt:
6957
      if self.op.disk_template not in constants.DTS_MAY_ADOPT:
6958
        raise errors.OpPrereqError("Disk adoption is not supported for the"
6959
                                   " '%s' disk template" %
6960
                                   self.op.disk_template,
6961
                                   errors.ECODE_INVAL)
6962
      if self.op.iallocator is not None:
6963
        raise errors.OpPrereqError("Disk adoption not allowed with an"
6964
                                   " iallocator script", errors.ECODE_INVAL)
6965
      if self.op.mode == constants.INSTANCE_IMPORT:
6966
        raise errors.OpPrereqError("Disk adoption not allowed for"
6967
                                   " instance import", errors.ECODE_INVAL)
6968

    
6969
    self.adopt_disks = has_adopt
6970

    
6971
    # instance name verification
6972
    if self.op.name_check:
6973
      self.hostname1 = netutils.GetHostname(name=self.op.instance_name)
6974
      self.op.instance_name = self.hostname1.name
6975
      # used in CheckPrereq for ip ping check
6976
      self.check_ip = self.hostname1.ip
6977
    else:
6978
      self.check_ip = None
6979

    
6980
    # file storage checks
6981
    if (self.op.file_driver and
6982
        not self.op.file_driver in constants.FILE_DRIVER):
6983
      raise errors.OpPrereqError("Invalid file driver name '%s'" %
6984
                                 self.op.file_driver, errors.ECODE_INVAL)
6985

    
6986
    if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
6987
      raise errors.OpPrereqError("File storage directory path not absolute",
6988
                                 errors.ECODE_INVAL)
6989

    
6990
    ### Node/iallocator related checks
6991
    _CheckIAllocatorOrNode(self, "iallocator", "pnode")
6992

    
6993
    if self.op.pnode is not None:
6994
      if self.op.disk_template in constants.DTS_NET_MIRROR:
6995
        if self.op.snode is None:
6996
          raise errors.OpPrereqError("The networked disk templates need"
6997
                                     " a mirror node", errors.ECODE_INVAL)
6998
      elif self.op.snode:
6999
        self.LogWarning("Secondary node will be ignored on non-mirrored disk"
7000
                        " template")
7001
        self.op.snode = None
7002

    
7003
    self._cds = _GetClusterDomainSecret()
7004

    
7005
    if self.op.mode == constants.INSTANCE_IMPORT:
7006
      # On import force_variant must be True, because if we forced it at
7007
      # initial install, our only chance when importing it back is that it
7008
      # works again!
7009
      self.op.force_variant = True
7010

    
7011
      if self.op.no_install:
7012
        self.LogInfo("No-installation mode has no effect during import")
7013

    
7014
    elif self.op.mode == constants.INSTANCE_CREATE:
7015
      if self.op.os_type is None:
7016
        raise errors.OpPrereqError("No guest OS specified",
7017
                                   errors.ECODE_INVAL)
7018
      if self.op.os_type in self.cfg.GetClusterInfo().blacklisted_os:
7019
        raise errors.OpPrereqError("Guest OS '%s' is not allowed for"
7020
                                   " installation" % self.op.os_type,
7021
                                   errors.ECODE_STATE)
7022
      if self.op.disk_template is None:
7023
        raise errors.OpPrereqError("No disk template specified",
7024
                                   errors.ECODE_INVAL)
7025

    
7026
    elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
7027
      # Check handshake to ensure both clusters have the same domain secret
7028
      src_handshake = self.op.source_handshake
7029
      if not src_handshake:
7030
        raise errors.OpPrereqError("Missing source handshake",
7031
                                   errors.ECODE_INVAL)
7032

    
7033
      errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds,
7034
                                                           src_handshake)
7035
      if errmsg:
7036
        raise errors.OpPrereqError("Invalid handshake: %s" % errmsg,
7037
                                   errors.ECODE_INVAL)
7038

    
7039
      # Load and check source CA
7040
      self.source_x509_ca_pem = self.op.source_x509_ca
7041
      if not self.source_x509_ca_pem:
7042
        raise errors.OpPrereqError("Missing source X509 CA",
7043
                                   errors.ECODE_INVAL)
7044

    
7045
      try:
7046
        (cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem,
7047
                                                    self._cds)
7048
      except OpenSSL.crypto.Error, err:
7049
        raise errors.OpPrereqError("Unable to load source X509 CA (%s)" %
7050
                                   (err, ), errors.ECODE_INVAL)
7051

    
7052
      (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
7053
      if errcode is not None:
7054
        raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ),
7055
                                   errors.ECODE_INVAL)
7056

    
7057
      self.source_x509_ca = cert
7058

    
7059
      src_instance_name = self.op.source_instance_name
7060
      if not src_instance_name:
7061
        raise errors.OpPrereqError("Missing source instance name",
7062
                                   errors.ECODE_INVAL)
7063

    
7064
      self.source_instance_name = \
7065
          netutils.GetHostname(name=src_instance_name).name
7066

    
7067
    else:
7068
      raise errors.OpPrereqError("Invalid instance creation mode %r" %
7069
                                 self.op.mode, errors.ECODE_INVAL)
7070

    
7071
  def ExpandNames(self):
7072
    """ExpandNames for CreateInstance.
7073

7074
    Figure out the right locks for instance creation.
7075

7076
    """
7077
    self.needed_locks = {}
7078

    
7079
    instance_name = self.op.instance_name
7080
    # this is just a preventive check, but someone might still add this
7081
    # instance in the meantime, and creation will fail at lock-add time
7082
    if instance_name in self.cfg.GetInstanceList():
7083
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
7084
                                 instance_name, errors.ECODE_EXISTS)
7085

    
7086
    self.add_locks[locking.LEVEL_INSTANCE] = instance_name
7087

    
7088
    if self.op.iallocator:
7089
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7090
    else:
7091
      self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode)
7092
      nodelist = [self.op.pnode]
7093
      if self.op.snode is not None:
7094
        self.op.snode = _ExpandNodeName(self.cfg, self.op.snode)
7095
        nodelist.append(self.op.snode)
7096
      self.needed_locks[locking.LEVEL_NODE] = nodelist
7097

    
7098
    # in case of import lock the source node too
7099
    if self.op.mode == constants.INSTANCE_IMPORT:
7100
      src_node = self.op.src_node
7101
      src_path = self.op.src_path
7102

    
7103
      if src_path is None:
7104
        self.op.src_path = src_path = self.op.instance_name
7105

    
7106
      if src_node is None:
7107
        self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7108
        self.op.src_node = None
7109
        if os.path.isabs(src_path):
7110
          raise errors.OpPrereqError("Importing an instance from an absolute"
7111
                                     " path requires a source node option.",
7112
                                     errors.ECODE_INVAL)
7113
      else:
7114
        self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node)
7115
        if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
7116
          self.needed_locks[locking.LEVEL_NODE].append(src_node)
7117
        if not os.path.isabs(src_path):
7118
          self.op.src_path = src_path = \
7119
            utils.PathJoin(constants.EXPORT_DIR, src_path)
7120

    
7121
  def _RunAllocator(self):
7122
    """Run the allocator based on input opcode.
7123

7124
    """
7125
    nics = [n.ToDict() for n in self.nics]
7126
    ial = IAllocator(self.cfg, self.rpc,
7127
                     mode=constants.IALLOCATOR_MODE_ALLOC,
7128
                     name=self.op.instance_name,
7129
                     disk_template=self.op.disk_template,
7130
                     tags=[],
7131
                     os=self.op.os_type,
7132
                     vcpus=self.be_full[constants.BE_VCPUS],
7133
                     mem_size=self.be_full[constants.BE_MEMORY],
7134
                     disks=self.disks,
7135
                     nics=nics,
7136
                     hypervisor=self.op.hypervisor,
7137
                     )
7138

    
7139
    ial.Run(self.op.iallocator)
7140

    
7141
    if not ial.success:
7142
      raise errors.OpPrereqError("Can't compute nodes using"
7143
                                 " iallocator '%s': %s" %
7144
                                 (self.op.iallocator, ial.info),
7145
                                 errors.ECODE_NORES)
7146
    if len(ial.result) != ial.required_nodes:
7147
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
7148
                                 " of nodes (%s), required %s" %
7149
                                 (self.op.iallocator, len(ial.result),
7150
                                  ial.required_nodes), errors.ECODE_FAULT)
7151
    self.op.pnode = ial.result[0]
7152
    self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
7153
                 self.op.instance_name, self.op.iallocator,
7154
                 utils.CommaJoin(ial.result))
7155
    if ial.required_nodes == 2:
7156
      self.op.snode = ial.result[1]
7157

    
7158
  def BuildHooksEnv(self):
7159
    """Build hooks env.
7160

7161
    This runs on master, primary and secondary nodes of the instance.
7162

7163
    """
7164
    env = {
7165
      "ADD_MODE": self.op.mode,
7166
      }
7167
    if self.op.mode == constants.INSTANCE_IMPORT:
7168
      env["SRC_NODE"] = self.op.src_node
7169
      env["SRC_PATH"] = self.op.src_path
7170
      env["SRC_IMAGES"] = self.src_images
7171

    
7172
    env.update(_BuildInstanceHookEnv(
7173
      name=self.op.instance_name,
7174
      primary_node=self.op.pnode,
7175
      secondary_nodes=self.secondaries,
7176
      status=self.op.start,
7177
      os_type=self.op.os_type,
7178
      memory=self.be_full[constants.BE_MEMORY],
7179
      vcpus=self.be_full[constants.BE_VCPUS],
7180
      nics=_NICListToTuple(self, self.nics),
7181
      disk_template=self.op.disk_template,
7182
      disks=[(d["size"], d["mode"]) for d in self.disks],
7183
      bep=self.be_full,
7184
      hvp=self.hv_full,
7185
      hypervisor_name=self.op.hypervisor,
7186
    ))
7187

    
7188
    nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
7189
          self.secondaries)
7190
    return env, nl, nl
7191

    
7192
  def _ReadExportInfo(self):
7193
    """Reads the export information from disk.
7194

7195
    It will override the opcode source node and path with the actual
7196
    information, if these two were not specified before.
7197

7198
    @return: the export information
7199

7200
    """
7201
    assert self.op.mode == constants.INSTANCE_IMPORT
7202

    
7203
    src_node = self.op.src_node
7204
    src_path = self.op.src_path
7205

    
7206
    if src_node is None:
7207
      locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
7208
      exp_list = self.rpc.call_export_list(locked_nodes)
7209
      found = False
7210
      for node in exp_list:
7211
        if exp_list[node].fail_msg:
7212
          continue
7213
        if src_path in exp_list[node].payload:
7214
          found = True
7215
          self.op.src_node = src_node = node
7216
          self.op.src_path = src_path = utils.PathJoin(constants.EXPORT_DIR,
7217
                                                       src_path)
7218
          break
7219
      if not found:
7220
        raise errors.OpPrereqError("No export found for relative path %s" %
7221
                                    src_path, errors.ECODE_INVAL)
7222

    
7223
    _CheckNodeOnline(self, src_node)
7224
    result = self.rpc.call_export_info(src_node, src_path)
7225
    result.Raise("No export or invalid export found in dir %s" % src_path)
7226

    
7227
    export_info = objects.SerializableConfigParser.Loads(str(result.payload))
7228
    if not export_info.has_section(constants.INISECT_EXP):
7229
      raise errors.ProgrammerError("Corrupted export config",
7230
                                   errors.ECODE_ENVIRON)
7231

    
7232
    ei_version = export_info.get(constants.INISECT_EXP, "version")
7233
    if (int(ei_version) != constants.EXPORT_VERSION):
7234
      raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
7235
                                 (ei_version, constants.EXPORT_VERSION),
7236
                                 errors.ECODE_ENVIRON)
7237
    return export_info
7238

    
7239
  def _ReadExportParams(self, einfo):
7240
    """Use export parameters as defaults.
7241

7242
    In case the opcode doesn't specify (as in override) some instance
7243
    parameters, then try to use them from the export information, if
7244
    that declares them.
7245

7246
    """
7247
    self.op.os_type = einfo.get(constants.INISECT_EXP, "os")
7248

    
7249
    if self.op.disk_template is None:
7250
      if einfo.has_option(constants.INISECT_INS, "disk_template"):
7251
        self.op.disk_template = einfo.get(constants.INISECT_INS,
7252
                                          "disk_template")
7253
      else:
7254
        raise errors.OpPrereqError("No disk template specified and the export"
7255
                                   " is missing the disk_template information",
7256
                                   errors.ECODE_INVAL)
7257

    
7258
    if not self.op.disks:
7259
      if einfo.has_option(constants.INISECT_INS, "disk_count"):
7260
        disks = []
7261
        # TODO: import the disk iv_name too
7262
        for idx in range(einfo.getint(constants.INISECT_INS, "disk_count")):
7263
          disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
7264
          disks.append({"size": disk_sz})
7265
        self.op.disks = disks
7266
      else:
7267
        raise errors.OpPrereqError("No disk info specified and the export"
7268
                                   " is missing the disk information",
7269
                                   errors.ECODE_INVAL)
7270

    
7271
    if (not self.op.nics and
7272
        einfo.has_option(constants.INISECT_INS, "nic_count")):
7273
      nics = []
7274
      for idx in range(einfo.getint(constants.INISECT_INS, "nic_count")):
7275
        ndict = {}
7276
        for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]:
7277
          v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
7278
          ndict[name] = v
7279
        nics.append(ndict)
7280
      self.op.nics = nics
7281

    
7282
    if (self.op.hypervisor is None and
7283
        einfo.has_option(constants.INISECT_INS, "hypervisor")):
7284
      self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor")
7285
    if einfo.has_section(constants.INISECT_HYP):
7286
      # use the export parameters but do not override the ones
7287
      # specified by the user
7288
      for name, value in einfo.items(constants.INISECT_HYP):
7289
        if name not in self.op.hvparams:
7290
          self.op.hvparams[name] = value
7291

    
7292
    if einfo.has_section(constants.INISECT_BEP):
7293
      # use the parameters, without overriding
7294
      for name, value in einfo.items(constants.INISECT_BEP):
7295
        if name not in self.op.beparams:
7296
          self.op.beparams[name] = value
7297
    else:
7298
      # try to read the parameters old style, from the main section
7299
      for name in constants.BES_PARAMETERS:
7300
        if (name not in self.op.beparams and
7301
            einfo.has_option(constants.INISECT_INS, name)):
7302
          self.op.beparams[name] = einfo.get(constants.INISECT_INS, name)
7303

    
7304
    if einfo.has_section(constants.INISECT_OSP):
7305
      # use the parameters, without overriding
7306
      for name, value in einfo.items(constants.INISECT_OSP):
7307
        if name not in self.op.osparams:
7308
          self.op.osparams[name] = value
7309

    
7310
  def _RevertToDefaults(self, cluster):
7311
    """Revert the instance parameters to the default values.
7312

7313
    """
7314
    # hvparams
7315
    hv_defs = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, {})
7316
    for name in self.op.hvparams.keys():
7317
      if name in hv_defs and hv_defs[name] == self.op.hvparams[name]:
7318
        del self.op.hvparams[name]
7319
    # beparams
7320
    be_defs = cluster.SimpleFillBE({})
7321
    for name in self.op.beparams.keys():
7322
      if name in be_defs and be_defs[name] == self.op.beparams[name]:
7323
        del self.op.beparams[name]
7324
    # nic params
7325
    nic_defs = cluster.SimpleFillNIC({})
7326
    for nic in self.op.nics:
7327
      for name in constants.NICS_PARAMETERS:
7328
        if name in nic and name in nic_defs and nic[name] == nic_defs[name]:
7329
          del nic[name]
7330
    # osparams
7331
    os_defs = cluster.SimpleFillOS(self.op.os_type, {})
7332
    for name in self.op.osparams.keys():
7333
      if name in os_defs and os_defs[name] == self.op.osparams[name]:
7334
        del self.op.osparams[name]
7335

    
7336
  def CheckPrereq(self):
7337
    """Check prerequisites.
7338

7339
    """
7340
    if self.op.mode == constants.INSTANCE_IMPORT:
7341
      export_info = self._ReadExportInfo()
7342
      self._ReadExportParams(export_info)
7343

    
7344
    if (not self.cfg.GetVGName() and
7345
        self.op.disk_template not in constants.DTS_NOT_LVM):
7346
      raise errors.OpPrereqError("Cluster does not support lvm-based"
7347
                                 " instances", errors.ECODE_STATE)
7348

    
7349
    if self.op.hypervisor is None:
7350
      self.op.hypervisor = self.cfg.GetHypervisorType()
7351

    
7352
    cluster = self.cfg.GetClusterInfo()
7353
    enabled_hvs = cluster.enabled_hypervisors
7354
    if self.op.hypervisor not in enabled_hvs:
7355
      raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
7356
                                 " cluster (%s)" % (self.op.hypervisor,
7357
                                  ",".join(enabled_hvs)),
7358
                                 errors.ECODE_STATE)
7359

    
7360
    # check hypervisor parameter syntax (locally)
7361
    utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
7362
    filled_hvp = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type,
7363
                                      self.op.hvparams)
7364
    hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
7365
    hv_type.CheckParameterSyntax(filled_hvp)
7366
    self.hv_full = filled_hvp
7367
    # check that we don't specify global parameters on an instance
7368
    _CheckGlobalHvParams(self.op.hvparams)
7369

    
7370
    # fill and remember the beparams dict
7371
    utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
7372
    self.be_full = cluster.SimpleFillBE(self.op.beparams)
7373

    
7374
    # build os parameters
7375
    self.os_full = cluster.SimpleFillOS(self.op.os_type, self.op.osparams)
7376

    
7377
    # now that hvp/bep are in final format, let's reset to defaults,
7378
    # if told to do so
7379
    if self.op.identify_defaults:
7380
      self._RevertToDefaults(cluster)
7381

    
7382
    # NIC buildup
7383
    self.nics = []
7384
    for idx, nic in enumerate(self.op.nics):
7385
      nic_mode_req = nic.get("mode", None)
7386
      nic_mode = nic_mode_req
7387
      if nic_mode is None:
7388
        nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
7389

    
7390
      # in routed mode, for the first nic, the default ip is 'auto'
7391
      if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
7392
        default_ip_mode = constants.VALUE_AUTO
7393
      else:
7394
        default_ip_mode = constants.VALUE_NONE
7395

    
7396
      # ip validity checks
7397
      ip = nic.get("ip", default_ip_mode)
7398
      if ip is None or ip.lower() == constants.VALUE_NONE:
7399
        nic_ip = None
7400
      elif ip.lower() == constants.VALUE_AUTO:
7401
        if not self.op.name_check:
7402
          raise errors.OpPrereqError("IP address set to auto but name checks"
7403
                                     " have been skipped",
7404
                                     errors.ECODE_INVAL)
7405
        nic_ip = self.hostname1.ip
7406
      else:
7407
        if not netutils.IPAddress.IsValid(ip):
7408
          raise errors.OpPrereqError("Invalid IP address '%s'" % ip,
7409
                                     errors.ECODE_INVAL)
7410
        nic_ip = ip
7411

    
7412
      # TODO: check the ip address for uniqueness
7413
      if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
7414
        raise errors.OpPrereqError("Routed nic mode requires an ip address",
7415
                                   errors.ECODE_INVAL)
7416

    
7417
      # MAC address verification
7418
      mac = nic.get("mac", constants.VALUE_AUTO)
7419
      if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7420
        mac = utils.NormalizeAndValidateMac(mac)
7421

    
7422
        try:
7423
          self.cfg.ReserveMAC(mac, self.proc.GetECId())
7424
        except errors.ReservationError:
7425
          raise errors.OpPrereqError("MAC address %s already in use"
7426
                                     " in cluster" % mac,
7427
                                     errors.ECODE_NOTUNIQUE)
7428

    
7429
      # bridge verification
7430
      bridge = nic.get("bridge", None)
7431
      link = nic.get("link", None)
7432
      if bridge and link:
7433
        raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
7434
                                   " at the same time", errors.ECODE_INVAL)
7435
      elif bridge and nic_mode == constants.NIC_MODE_ROUTED:
7436
        raise errors.OpPrereqError("Cannot pass 'bridge' on a routed nic",
7437
                                   errors.ECODE_INVAL)
7438
      elif bridge:
7439
        link = bridge
7440

    
7441
      nicparams = {}
7442
      if nic_mode_req:
7443
        nicparams[constants.NIC_MODE] = nic_mode_req
7444
      if link:
7445
        nicparams[constants.NIC_LINK] = link
7446

    
7447
      check_params = cluster.SimpleFillNIC(nicparams)
7448
      objects.NIC.CheckParameterSyntax(check_params)
7449
      self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
7450

    
7451
    # disk checks/pre-build
7452
    self.disks = []
7453
    for disk in self.op.disks:
7454
      mode = disk.get("mode", constants.DISK_RDWR)
7455
      if mode not in constants.DISK_ACCESS_SET:
7456
        raise errors.OpPrereqError("Invalid disk access mode '%s'" %
7457
                                   mode, errors.ECODE_INVAL)
7458
      size = disk.get("size", None)
7459
      if size is None:
7460
        raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
7461
      try:
7462
        size = int(size)
7463
      except (TypeError, ValueError):
7464
        raise errors.OpPrereqError("Invalid disk size '%s'" % size,
7465
                                   errors.ECODE_INVAL)
7466
      vg = disk.get("vg", self.cfg.GetVGName())
7467
      new_disk = {"size": size, "mode": mode, "vg": vg}
7468
      if "adopt" in disk:
7469
        new_disk["adopt"] = disk["adopt"]
7470
      self.disks.append(new_disk)
7471

    
7472
    if self.op.mode == constants.INSTANCE_IMPORT:
7473

    
7474
      # Check that the new instance doesn't have less disks than the export
7475
      instance_disks = len(self.disks)
7476
      export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
7477
      if instance_disks < export_disks:
7478
        raise errors.OpPrereqError("Not enough disks to import."
7479
                                   " (instance: %d, export: %d)" %
7480
                                   (instance_disks, export_disks),
7481
                                   errors.ECODE_INVAL)
7482

    
7483
      disk_images = []
7484
      for idx in range(export_disks):
7485
        option = 'disk%d_dump' % idx
7486
        if export_info.has_option(constants.INISECT_INS, option):
7487
          # FIXME: are the old os-es, disk sizes, etc. useful?
7488
          export_name = export_info.get(constants.INISECT_INS, option)
7489
          image = utils.PathJoin(self.op.src_path, export_name)
7490
          disk_images.append(image)
7491
        else:
7492
          disk_images.append(False)
7493

    
7494
      self.src_images = disk_images
7495

    
7496
      old_name = export_info.get(constants.INISECT_INS, 'name')
7497
      try:
7498
        exp_nic_count = export_info.getint(constants.INISECT_INS, 'nic_count')
7499
      except (TypeError, ValueError), err:
7500
        raise errors.OpPrereqError("Invalid export file, nic_count is not"
7501
                                   " an integer: %s" % str(err),
7502
                                   errors.ECODE_STATE)
7503
      if self.op.instance_name == old_name:
7504
        for idx, nic in enumerate(self.nics):
7505
          if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
7506
            nic_mac_ini = 'nic%d_mac' % idx
7507
            nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
7508

    
7509
    # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
7510

    
7511
    # ip ping checks (we use the same ip that was resolved in ExpandNames)
7512
    if self.op.ip_check:
7513
      if netutils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
7514
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
7515
                                   (self.check_ip, self.op.instance_name),
7516
                                   errors.ECODE_NOTUNIQUE)
7517

    
7518
    #### mac address generation
7519
    # By generating here the mac address both the allocator and the hooks get
7520
    # the real final mac address rather than the 'auto' or 'generate' value.
7521
    # There is a race condition between the generation and the instance object
7522
    # creation, which means that we know the mac is valid now, but we're not
7523
    # sure it will be when we actually add the instance. If things go bad
7524
    # adding the instance will abort because of a duplicate mac, and the
7525
    # creation job will fail.
7526
    for nic in self.nics:
7527
      if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7528
        nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
7529

    
7530
    #### allocator run
7531

    
7532
    if self.op.iallocator is not None:
7533
      self._RunAllocator()
7534

    
7535
    #### node related checks
7536

    
7537
    # check primary node
7538
    self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
7539
    assert self.pnode is not None, \
7540
      "Cannot retrieve locked node %s" % self.op.pnode
7541
    if pnode.offline:
7542
      raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
7543
                                 pnode.name, errors.ECODE_STATE)
7544
    if pnode.drained:
7545
      raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
7546
                                 pnode.name, errors.ECODE_STATE)
7547
    if not pnode.vm_capable:
7548
      raise errors.OpPrereqError("Cannot use non-vm_capable primary node"
7549
                                 " '%s'" % pnode.name, errors.ECODE_STATE)
7550

    
7551
    self.secondaries = []
7552

    
7553
    # mirror node verification
7554
    if self.op.disk_template in constants.DTS_NET_MIRROR:
7555
      if self.op.snode == pnode.name:
7556
        raise errors.OpPrereqError("The secondary node cannot be the"
7557
                                   " primary node.", errors.ECODE_INVAL)
7558
      _CheckNodeOnline(self, self.op.snode)
7559
      _CheckNodeNotDrained(self, self.op.snode)
7560
      _CheckNodeVmCapable(self, self.op.snode)
7561
      self.secondaries.append(self.op.snode)
7562

    
7563
    nodenames = [pnode.name] + self.secondaries
7564

    
7565
    if not self.adopt_disks:
7566
      # Check lv size requirements, if not adopting
7567
      req_sizes = _ComputeDiskSizePerVG(self.op.disk_template, self.disks)
7568
      _CheckNodesFreeDiskPerVG(self, nodenames, req_sizes)
7569

    
7570
    else: # instead, we must check the adoption data
7571
      all_lvs = set([i["vg"] + "/" + i["adopt"] for i in self.disks])
7572
      if len(all_lvs) != len(self.disks):
7573
        raise errors.OpPrereqError("Duplicate volume names given for adoption",
7574
                                   errors.ECODE_INVAL)
7575
      for lv_name in all_lvs:
7576
        try:
7577
          # FIXME: lv_name here is "vg/lv" need to ensure that other calls
7578
          # to ReserveLV uses the same syntax
7579
          self.cfg.ReserveLV(lv_name, self.proc.GetECId())
7580
        except errors.ReservationError:
7581
          raise errors.OpPrereqError("LV named %s used by another instance" %
7582
                                     lv_name, errors.ECODE_NOTUNIQUE)
7583

    
7584
      vg_names = self.rpc.call_vg_list([pnode.name])[pnode.name]
7585
      vg_names.Raise("Cannot get VG information from node %s" % pnode.name)
7586

    
7587
      node_lvs = self.rpc.call_lv_list([pnode.name],
7588
                                       vg_names.payload.keys())[pnode.name]
7589
      node_lvs.Raise("Cannot get LV information from node %s" % pnode.name)
7590
      node_lvs = node_lvs.payload
7591

    
7592
      delta = all_lvs.difference(node_lvs.keys())
7593
      if delta:
7594
        raise errors.OpPrereqError("Missing logical volume(s): %s" %
7595
                                   utils.CommaJoin(delta),
7596
                                   errors.ECODE_INVAL)
7597
      online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]]
7598
      if online_lvs:
7599
        raise errors.OpPrereqError("Online logical volumes found, cannot"
7600
                                   " adopt: %s" % utils.CommaJoin(online_lvs),
7601
                                   errors.ECODE_STATE)
7602
      # update the size of disk based on what is found
7603
      for dsk in self.disks:
7604
        dsk["size"] = int(float(node_lvs[dsk["vg"] + "/" + dsk["adopt"]][0]))
7605

    
7606
    _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
7607

    
7608
    _CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant)
7609
    # check OS parameters (remotely)
7610
    _CheckOSParams(self, True, nodenames, self.op.os_type, self.os_full)
7611

    
7612
    _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
7613

    
7614
    # memory check on primary node
7615
    if self.op.start:
7616
      _CheckNodeFreeMemory(self, self.pnode.name,
7617
                           "creating instance %s" % self.op.instance_name,
7618
                           self.be_full[constants.BE_MEMORY],
7619
                           self.op.hypervisor)
7620

    
7621
    self.dry_run_result = list(nodenames)
7622

    
7623
  def Exec(self, feedback_fn):
7624
    """Create and add the instance to the cluster.
7625

7626
    """
7627
    instance = self.op.instance_name
7628
    pnode_name = self.pnode.name
7629

    
7630
    ht_kind = self.op.hypervisor
7631
    if ht_kind in constants.HTS_REQ_PORT:
7632
      network_port = self.cfg.AllocatePort()
7633
    else:
7634
      network_port = None
7635

    
7636
    if constants.ENABLE_FILE_STORAGE:
7637
      # this is needed because os.path.join does not accept None arguments
7638
      if self.op.file_storage_dir is None:
7639
        string_file_storage_dir = ""
7640
      else:
7641
        string_file_storage_dir = self.op.file_storage_dir
7642

    
7643
      # build the full file storage dir path
7644
      file_storage_dir = utils.PathJoin(self.cfg.GetFileStorageDir(),
7645
                                        string_file_storage_dir, instance)
7646
    else:
7647
      file_storage_dir = ""
7648

    
7649
    disks = _GenerateDiskTemplate(self,
7650
                                  self.op.disk_template,
7651
                                  instance, pnode_name,
7652
                                  self.secondaries,
7653
                                  self.disks,
7654
                                  file_storage_dir,
7655
                                  self.op.file_driver,
7656
                                  0,
7657
                                  feedback_fn)
7658

    
7659
    iobj = objects.Instance(name=instance, os=self.op.os_type,
7660
                            primary_node=pnode_name,
7661
                            nics=self.nics, disks=disks,
7662
                            disk_template=self.op.disk_template,
7663
                            admin_up=False,
7664
                            network_port=network_port,
7665
                            beparams=self.op.beparams,
7666
                            hvparams=self.op.hvparams,
7667
                            hypervisor=self.op.hypervisor,
7668
                            osparams=self.op.osparams,
7669
                            )
7670

    
7671
    if self.adopt_disks:
7672
      # rename LVs to the newly-generated names; we need to construct
7673
      # 'fake' LV disks with the old data, plus the new unique_id
7674
      tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks]
7675
      rename_to = []
7676
      for t_dsk, a_dsk in zip (tmp_disks, self.disks):
7677
        rename_to.append(t_dsk.logical_id)
7678
        t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk["adopt"])
7679
        self.cfg.SetDiskID(t_dsk, pnode_name)
7680
      result = self.rpc.call_blockdev_rename(pnode_name,
7681
                                             zip(tmp_disks, rename_to))
7682
      result.Raise("Failed to rename adoped LVs")
7683
    else:
7684
      feedback_fn("* creating instance disks...")
7685
      try:
7686
        _CreateDisks(self, iobj)
7687
      except errors.OpExecError:
7688
        self.LogWarning("Device creation failed, reverting...")
7689
        try:
7690
          _RemoveDisks(self, iobj)
7691
        finally:
7692
          self.cfg.ReleaseDRBDMinors(instance)
7693
          raise
7694

    
7695
    feedback_fn("adding instance %s to cluster config" % instance)
7696

    
7697
    self.cfg.AddInstance(iobj, self.proc.GetECId())
7698

    
7699
    # Declare that we don't want to remove the instance lock anymore, as we've
7700
    # added the instance to the config
7701
    del self.remove_locks[locking.LEVEL_INSTANCE]
7702
    # Unlock all the nodes
7703
    if self.op.mode == constants.INSTANCE_IMPORT:
7704
      nodes_keep = [self.op.src_node]
7705
      nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
7706
                       if node != self.op.src_node]
7707
      self.context.glm.release(locking.LEVEL_NODE, nodes_release)
7708
      self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
7709
    else:
7710
      self.context.glm.release(locking.LEVEL_NODE)
7711
      del self.acquired_locks[locking.LEVEL_NODE]
7712

    
7713
    disk_abort = False
7714
    if not self.adopt_disks and self.cfg.GetClusterInfo().prealloc_wipe_disks:
7715
      feedback_fn("* wiping instance disks...")
7716
      try:
7717
        _WipeDisks(self, iobj)
7718
      except errors.OpExecError, err:
7719
        logging.exception("Wiping disks failed")
7720
        self.LogWarning("Wiping instance disks failed (%s)", err)
7721
        disk_abort = True
7722

    
7723
    if disk_abort:
7724
      # Something is already wrong with the disks, don't do anything else
7725
      pass
7726
    elif self.op.wait_for_sync:
7727
      disk_abort = not _WaitForSync(self, iobj)
7728
    elif iobj.disk_template in constants.DTS_NET_MIRROR:
7729
      # make sure the disks are not degraded (still sync-ing is ok)
7730
      time.sleep(15)
7731
      feedback_fn("* checking mirrors status")
7732
      disk_abort = not _WaitForSync(self, iobj, oneshot=True)
7733
    else:
7734
      disk_abort = False
7735

    
7736
    if disk_abort:
7737
      _RemoveDisks(self, iobj)
7738
      self.cfg.RemoveInstance(iobj.name)
7739
      # Make sure the instance lock gets removed
7740
      self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
7741
      raise errors.OpExecError("There are some degraded disks for"
7742
                               " this instance")
7743

    
7744
    if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks:
7745
      if self.op.mode == constants.INSTANCE_CREATE:
7746
        if not self.op.no_install:
7747
          feedback_fn("* running the instance OS create scripts...")
7748
          # FIXME: pass debug option from opcode to backend
7749
          result = self.rpc.call_instance_os_add(pnode_name, iobj, False,
7750
                                                 self.op.debug_level)
7751
          result.Raise("Could not add os for instance %s"
7752
                       " on node %s" % (instance, pnode_name))
7753

    
7754
      elif self.op.mode == constants.INSTANCE_IMPORT:
7755
        feedback_fn("* running the instance OS import scripts...")
7756

    
7757
        transfers = []
7758

    
7759
        for idx, image in enumerate(self.src_images):
7760
          if not image:
7761
            continue
7762

    
7763
          # FIXME: pass debug option from opcode to backend
7764
          dt = masterd.instance.DiskTransfer("disk/%s" % idx,
7765
                                             constants.IEIO_FILE, (image, ),
7766
                                             constants.IEIO_SCRIPT,
7767
                                             (iobj.disks[idx], idx),
7768
                                             None)
7769
          transfers.append(dt)
7770

    
7771
        import_result = \
7772
          masterd.instance.TransferInstanceData(self, feedback_fn,
7773
                                                self.op.src_node, pnode_name,
7774
                                                self.pnode.secondary_ip,
7775
                                                iobj, transfers)
7776
        if not compat.all(import_result):
7777
          self.LogWarning("Some disks for instance %s on node %s were not"
7778
                          " imported successfully" % (instance, pnode_name))
7779

    
7780
      elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
7781
        feedback_fn("* preparing remote import...")
7782
        # The source cluster will stop the instance before attempting to make a
7783
        # connection. In some cases stopping an instance can take a long time,
7784
        # hence the shutdown timeout is added to the connection timeout.
7785
        connect_timeout = (constants.RIE_CONNECT_TIMEOUT +
7786
                           self.op.source_shutdown_timeout)
7787
        timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
7788

    
7789
        assert iobj.primary_node == self.pnode.name
7790
        disk_results = \
7791
          masterd.instance.RemoteImport(self, feedback_fn, iobj, self.pnode,
7792
                                        self.source_x509_ca,
7793
                                        self._cds, timeouts)
7794
        if not compat.all(disk_results):
7795
          # TODO: Should the instance still be started, even if some disks
7796
          # failed to import (valid for local imports, too)?
7797
          self.LogWarning("Some disks for instance %s on node %s were not"
7798
                          " imported successfully" % (instance, pnode_name))
7799

    
7800
        # Run rename script on newly imported instance
7801
        assert iobj.name == instance
7802
        feedback_fn("Running rename script for %s" % instance)
7803
        result = self.rpc.call_instance_run_rename(pnode_name, iobj,
7804
                                                   self.source_instance_name,
7805
                                                   self.op.debug_level)
7806
        if result.fail_msg:
7807
          self.LogWarning("Failed to run rename script for %s on node"
7808
                          " %s: %s" % (instance, pnode_name, result.fail_msg))
7809

    
7810
      else:
7811
        # also checked in the prereq part
7812
        raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
7813
                                     % self.op.mode)
7814

    
7815
    if self.op.start:
7816
      iobj.admin_up = True
7817
      self.cfg.Update(iobj, feedback_fn)
7818
      logging.info("Starting instance %s on node %s", instance, pnode_name)
7819
      feedback_fn("* starting instance...")
7820
      result = self.rpc.call_instance_start(pnode_name, iobj, None, None)
7821
      result.Raise("Could not start instance")
7822

    
7823
    return list(iobj.all_nodes)
7824

    
7825

    
7826
class LUInstanceConsole(NoHooksLU):
7827
  """Connect to an instance's console.
7828

7829
  This is somewhat special in that it returns the command line that
7830
  you need to run on the master node in order to connect to the
7831
  console.
7832

7833
  """
7834
  REQ_BGL = False
7835

    
7836
  def ExpandNames(self):
7837
    self._ExpandAndLockInstance()
7838

    
7839
  def CheckPrereq(self):
7840
    """Check prerequisites.
7841

7842
    This checks that the instance is in the cluster.
7843

7844
    """
7845
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7846
    assert self.instance is not None, \
7847
      "Cannot retrieve locked instance %s" % self.op.instance_name
7848
    _CheckNodeOnline(self, self.instance.primary_node)
7849

    
7850
  def Exec(self, feedback_fn):
7851
    """Connect to the console of an instance
7852

7853
    """
7854
    instance = self.instance
7855
    node = instance.primary_node
7856

    
7857
    node_insts = self.rpc.call_instance_list([node],
7858
                                             [instance.hypervisor])[node]
7859
    node_insts.Raise("Can't get node information from %s" % node)
7860

    
7861
    if instance.name not in node_insts.payload:
7862
      if instance.admin_up:
7863
        state = "ERROR_down"
7864
      else:
7865
        state = "ADMIN_down"
7866
      raise errors.OpExecError("Instance %s is not running (state %s)" %
7867
                               (instance.name, state))
7868

    
7869
    logging.debug("Connecting to console of %s on %s", instance.name, node)
7870

    
7871
    return _GetInstanceConsole(self.cfg.GetClusterInfo(), instance)
7872

    
7873

    
7874
def _GetInstanceConsole(cluster, instance):
7875
  """Returns console information for an instance.
7876

7877
  @type cluster: L{objects.Cluster}
7878
  @type instance: L{objects.Instance}
7879
  @rtype: dict
7880

7881
  """
7882
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
7883
  # beparams and hvparams are passed separately, to avoid editing the
7884
  # instance and then saving the defaults in the instance itself.
7885
  hvparams = cluster.FillHV(instance)
7886
  beparams = cluster.FillBE(instance)
7887
  console = hyper.GetInstanceConsole(instance, hvparams, beparams)
7888

    
7889
  assert console.instance == instance.name
7890
  assert console.Validate()
7891

    
7892
  return console.ToDict()
7893

    
7894

    
7895
class LUInstanceReplaceDisks(LogicalUnit):
7896
  """Replace the disks of an instance.
7897

7898
  """
7899
  HPATH = "mirrors-replace"
7900
  HTYPE = constants.HTYPE_INSTANCE
7901
  REQ_BGL = False
7902

    
7903
  def CheckArguments(self):
7904
    TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
7905
                                  self.op.iallocator)
7906

    
7907
  def ExpandNames(self):
7908
    self._ExpandAndLockInstance()
7909

    
7910
    if self.op.iallocator is not None:
7911
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7912

    
7913
    elif self.op.remote_node is not None:
7914
      remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
7915
      self.op.remote_node = remote_node
7916

    
7917
      # Warning: do not remove the locking of the new secondary here
7918
      # unless DRBD8.AddChildren is changed to work in parallel;
7919
      # currently it doesn't since parallel invocations of
7920
      # FindUnusedMinor will conflict
7921
      self.needed_locks[locking.LEVEL_NODE] = [remote_node]
7922
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
7923

    
7924
    else:
7925
      self.needed_locks[locking.LEVEL_NODE] = []
7926
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7927

    
7928
    self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
7929
                                   self.op.iallocator, self.op.remote_node,
7930
                                   self.op.disks, False, self.op.early_release)
7931

    
7932
    self.tasklets = [self.replacer]
7933

    
7934
  def DeclareLocks(self, level):
7935
    # If we're not already locking all nodes in the set we have to declare the
7936
    # instance's primary/secondary nodes.
7937
    if (level == locking.LEVEL_NODE and
7938
        self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
7939
      self._LockInstancesNodes()
7940

    
7941
  def BuildHooksEnv(self):
7942
    """Build hooks env.
7943

7944
    This runs on the master, the primary and all the secondaries.
7945

7946
    """
7947
    instance = self.replacer.instance
7948
    env = {
7949
      "MODE": self.op.mode,
7950
      "NEW_SECONDARY": self.op.remote_node,
7951
      "OLD_SECONDARY": instance.secondary_nodes[0],
7952
      }
7953
    env.update(_BuildInstanceHookEnvByObject(self, instance))
7954
    nl = [
7955
      self.cfg.GetMasterNode(),
7956
      instance.primary_node,
7957
      ]
7958
    if self.op.remote_node is not None:
7959
      nl.append(self.op.remote_node)
7960
    return env, nl, nl
7961

    
7962

    
7963
class TLReplaceDisks(Tasklet):
7964
  """Replaces disks for an instance.
7965

7966
  Note: Locking is not within the scope of this class.
7967

7968
  """
7969
  def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
7970
               disks, delay_iallocator, early_release):
7971
    """Initializes this class.
7972

7973
    """
7974
    Tasklet.__init__(self, lu)
7975

    
7976
    # Parameters
7977
    self.instance_name = instance_name
7978
    self.mode = mode
7979
    self.iallocator_name = iallocator_name
7980
    self.remote_node = remote_node
7981
    self.disks = disks
7982
    self.delay_iallocator = delay_iallocator
7983
    self.early_release = early_release
7984

    
7985
    # Runtime data
7986
    self.instance = None
7987
    self.new_node = None
7988
    self.target_node = None
7989
    self.other_node = None
7990
    self.remote_node_info = None
7991
    self.node_secondary_ip = None
7992

    
7993
  @staticmethod
7994
  def CheckArguments(mode, remote_node, iallocator):
7995
    """Helper function for users of this class.
7996

7997
    """
7998
    # check for valid parameter combination
7999
    if mode == constants.REPLACE_DISK_CHG:
8000
      if remote_node is None and iallocator is None:
8001
        raise errors.OpPrereqError("When changing the secondary either an"
8002
                                   " iallocator script must be used or the"
8003
                                   " new node given", errors.ECODE_INVAL)
8004

    
8005
      if remote_node is not None and iallocator is not None:
8006
        raise errors.OpPrereqError("Give either the iallocator or the new"
8007
                                   " secondary, not both", errors.ECODE_INVAL)
8008

    
8009
    elif remote_node is not None or iallocator is not None:
8010
      # Not replacing the secondary
8011
      raise errors.OpPrereqError("The iallocator and new node options can"
8012
                                 " only be used when changing the"
8013
                                 " secondary node", errors.ECODE_INVAL)
8014

    
8015
  @staticmethod
8016
  def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
8017
    """Compute a new secondary node using an IAllocator.
8018

8019
    """
8020
    ial = IAllocator(lu.cfg, lu.rpc,
8021
                     mode=constants.IALLOCATOR_MODE_RELOC,
8022
                     name=instance_name,
8023
                     relocate_from=relocate_from)
8024

    
8025
    ial.Run(iallocator_name)
8026

    
8027
    if not ial.success:
8028
      raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
8029
                                 " %s" % (iallocator_name, ial.info),
8030
                                 errors.ECODE_NORES)
8031

    
8032
    if len(ial.result) != ial.required_nodes:
8033
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
8034
                                 " of nodes (%s), required %s" %
8035
                                 (iallocator_name,
8036
                                  len(ial.result), ial.required_nodes),
8037
                                 errors.ECODE_FAULT)
8038

    
8039
    remote_node_name = ial.result[0]
8040

    
8041
    lu.LogInfo("Selected new secondary for instance '%s': %s",
8042
               instance_name, remote_node_name)
8043

    
8044
    return remote_node_name
8045

    
8046
  def _FindFaultyDisks(self, node_name):
8047
    return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
8048
                                    node_name, True)
8049

    
8050
  def CheckPrereq(self):
8051
    """Check prerequisites.
8052

8053
    This checks that the instance is in the cluster.
8054

8055
    """
8056
    self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
8057
    assert instance is not None, \
8058
      "Cannot retrieve locked instance %s" % self.instance_name
8059

    
8060
    if instance.disk_template != constants.DT_DRBD8:
8061
      raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
8062
                                 " instances", errors.ECODE_INVAL)
8063

    
8064
    if len(instance.secondary_nodes) != 1:
8065
      raise errors.OpPrereqError("The instance has a strange layout,"
8066
                                 " expected one secondary but found %d" %
8067
                                 len(instance.secondary_nodes),
8068
                                 errors.ECODE_FAULT)
8069

    
8070
    if not self.delay_iallocator:
8071
      self._CheckPrereq2()
8072

    
8073
  def _CheckPrereq2(self):
8074
    """Check prerequisites, second part.
8075

8076
    This function should always be part of CheckPrereq. It was separated and is
8077
    now called from Exec because during node evacuation iallocator was only
8078
    called with an unmodified cluster model, not taking planned changes into
8079
    account.
8080

8081
    """
8082
    instance = self.instance
8083
    secondary_node = instance.secondary_nodes[0]
8084

    
8085
    if self.iallocator_name is None:
8086
      remote_node = self.remote_node
8087
    else:
8088
      remote_node = self._RunAllocator(self.lu, self.iallocator_name,
8089
                                       instance.name, instance.secondary_nodes)
8090

    
8091
    if remote_node is not None:
8092
      self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
8093
      assert self.remote_node_info is not None, \
8094
        "Cannot retrieve locked node %s" % remote_node
8095
    else:
8096
      self.remote_node_info = None
8097

    
8098
    if remote_node == self.instance.primary_node:
8099
      raise errors.OpPrereqError("The specified node is the primary node of"
8100
                                 " the instance.", errors.ECODE_INVAL)
8101

    
8102
    if remote_node == secondary_node:
8103
      raise errors.OpPrereqError("The specified node is already the"
8104
                                 " secondary node of the instance.",
8105
                                 errors.ECODE_INVAL)
8106

    
8107
    if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
8108
                                    constants.REPLACE_DISK_CHG):
8109
      raise errors.OpPrereqError("Cannot specify disks to be replaced",
8110
                                 errors.ECODE_INVAL)
8111

    
8112
    if self.mode == constants.REPLACE_DISK_AUTO:
8113
      faulty_primary = self._FindFaultyDisks(instance.primary_node)
8114
      faulty_secondary = self._FindFaultyDisks(secondary_node)
8115

    
8116
      if faulty_primary and faulty_secondary:
8117
        raise errors.OpPrereqError("Instance %s has faulty disks on more than"
8118
                                   " one node and can not be repaired"
8119
                                   " automatically" % self.instance_name,
8120
                                   errors.ECODE_STATE)
8121

    
8122
      if faulty_primary:
8123
        self.disks = faulty_primary
8124
        self.target_node = instance.primary_node
8125
        self.other_node = secondary_node
8126
        check_nodes = [self.target_node, self.other_node]
8127
      elif faulty_secondary:
8128
        self.disks = faulty_secondary
8129
        self.target_node = secondary_node
8130
        self.other_node = instance.primary_node
8131
        check_nodes = [self.target_node, self.other_node]
8132
      else:
8133
        self.disks = []
8134
        check_nodes = []
8135

    
8136
    else:
8137
      # Non-automatic modes
8138
      if self.mode == constants.REPLACE_DISK_PRI:
8139
        self.target_node = instance.primary_node
8140
        self.other_node = secondary_node
8141
        check_nodes = [self.target_node, self.other_node]
8142

    
8143
      elif self.mode == constants.REPLACE_DISK_SEC:
8144
        self.target_node = secondary_node
8145
        self.other_node = instance.primary_node
8146
        check_nodes = [self.target_node, self.other_node]
8147

    
8148
      elif self.mode == constants.REPLACE_DISK_CHG:
8149
        self.new_node = remote_node
8150
        self.other_node = instance.primary_node
8151
        self.target_node = secondary_node
8152
        check_nodes = [self.new_node, self.other_node]
8153

    
8154
        _CheckNodeNotDrained(self.lu, remote_node)
8155
        _CheckNodeVmCapable(self.lu, remote_node)
8156

    
8157
        old_node_info = self.cfg.GetNodeInfo(secondary_node)
8158
        assert old_node_info is not None
8159
        if old_node_info.offline and not self.early_release:
8160
          # doesn't make sense to delay the release
8161
          self.early_release = True
8162
          self.lu.LogInfo("Old secondary %s is offline, automatically enabling"
8163
                          " early-release mode", secondary_node)
8164

    
8165
      else:
8166
        raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
8167
                                     self.mode)
8168

    
8169
      # If not specified all disks should be replaced
8170
      if not self.disks:
8171
        self.disks = range(len(self.instance.disks))
8172

    
8173
    for node in check_nodes:
8174
      _CheckNodeOnline(self.lu, node)
8175

    
8176
    # Check whether disks are valid
8177
    for disk_idx in self.disks:
8178
      instance.FindDisk(disk_idx)
8179

    
8180
    # Get secondary node IP addresses
8181
    node_2nd_ip = {}
8182

    
8183
    for node_name in [self.target_node, self.other_node, self.new_node]:
8184
      if node_name is not None:
8185
        node_2nd_ip[node_name] = self.cfg.GetNodeInfo(node_name).secondary_ip
8186

    
8187
    self.node_secondary_ip = node_2nd_ip
8188

    
8189
  def Exec(self, feedback_fn):
8190
    """Execute disk replacement.
8191

8192
    This dispatches the disk replacement to the appropriate handler.
8193

8194
    """
8195
    if self.delay_iallocator:
8196
      self._CheckPrereq2()
8197

    
8198
    if not self.disks:
8199
      feedback_fn("No disks need replacement")
8200
      return
8201

    
8202
    feedback_fn("Replacing disk(s) %s for %s" %
8203
                (utils.CommaJoin(self.disks), self.instance.name))
8204

    
8205
    activate_disks = (not self.instance.admin_up)
8206

    
8207
    # Activate the instance disks if we're replacing them on a down instance
8208
    if activate_disks:
8209
      _StartInstanceDisks(self.lu, self.instance, True)
8210

    
8211
    try:
8212
      # Should we replace the secondary node?
8213
      if self.new_node is not None:
8214
        fn = self._ExecDrbd8Secondary
8215
      else:
8216
        fn = self._ExecDrbd8DiskOnly
8217

    
8218
      return fn(feedback_fn)
8219

    
8220
    finally:
8221
      # Deactivate the instance disks if we're replacing them on a
8222
      # down instance
8223
      if activate_disks:
8224
        _SafeShutdownInstanceDisks(self.lu, self.instance)
8225

    
8226
  def _CheckVolumeGroup(self, nodes):
8227
    self.lu.LogInfo("Checking volume groups")
8228

    
8229
    vgname = self.cfg.GetVGName()
8230

    
8231
    # Make sure volume group exists on all involved nodes
8232
    results = self.rpc.call_vg_list(nodes)
8233
    if not results:
8234
      raise errors.OpExecError("Can't list volume groups on the nodes")
8235

    
8236
    for node in nodes:
8237
      res = results[node]
8238
      res.Raise("Error checking node %s" % node)
8239
      if vgname not in res.payload:
8240
        raise errors.OpExecError("Volume group '%s' not found on node %s" %
8241
                                 (vgname, node))
8242

    
8243
  def _CheckDisksExistence(self, nodes):
8244
    # Check disk existence
8245
    for idx, dev in enumerate(self.instance.disks):
8246
      if idx not in self.disks:
8247
        continue
8248

    
8249
      for node in nodes:
8250
        self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
8251
        self.cfg.SetDiskID(dev, node)
8252

    
8253
        result = self.rpc.call_blockdev_find(node, dev)
8254

    
8255
        msg = result.fail_msg
8256
        if msg or not result.payload:
8257
          if not msg:
8258
            msg = "disk not found"
8259
          raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
8260
                                   (idx, node, msg))
8261

    
8262
  def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
8263
    for idx, dev in enumerate(self.instance.disks):
8264
      if idx not in self.disks:
8265
        continue
8266

    
8267
      self.lu.LogInfo("Checking disk/%d consistency on node %s" %
8268
                      (idx, node_name))
8269

    
8270
      if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
8271
                                   ldisk=ldisk):
8272
        raise errors.OpExecError("Node %s has degraded storage, unsafe to"
8273
                                 " replace disks for instance %s" %
8274
                                 (node_name, self.instance.name))
8275

    
8276
  def _CreateNewStorage(self, node_name):
8277
    vgname = self.cfg.GetVGName()
8278
    iv_names = {}
8279

    
8280
    for idx, dev in enumerate(self.instance.disks):
8281
      if idx not in self.disks:
8282
        continue
8283

    
8284
      self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
8285

    
8286
      self.cfg.SetDiskID(dev, node_name)
8287

    
8288
      lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
8289
      names = _GenerateUniqueNames(self.lu, lv_names)
8290

    
8291
      lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
8292
                             logical_id=(vgname, names[0]))
8293
      lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
8294
                             logical_id=(vgname, names[1]))
8295

    
8296
      new_lvs = [lv_data, lv_meta]
8297
      old_lvs = dev.children
8298
      iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
8299

    
8300
      # we pass force_create=True to force the LVM creation
8301
      for new_lv in new_lvs:
8302
        _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
8303
                        _GetInstanceInfoText(self.instance), False)
8304

    
8305
    return iv_names
8306

    
8307
  def _CheckDevices(self, node_name, iv_names):
8308
    for name, (dev, _, _) in iv_names.iteritems():
8309
      self.cfg.SetDiskID(dev, node_name)
8310

    
8311
      result = self.rpc.call_blockdev_find(node_name, dev)
8312

    
8313
      msg = result.fail_msg
8314
      if msg or not result.payload:
8315
        if not msg:
8316
          msg = "disk not found"
8317
        raise errors.OpExecError("Can't find DRBD device %s: %s" %
8318
                                 (name, msg))
8319

    
8320
      if result.payload.is_degraded:
8321
        raise errors.OpExecError("DRBD device %s is degraded!" % name)
8322

    
8323
  def _RemoveOldStorage(self, node_name, iv_names):
8324
    for name, (_, old_lvs, _) in iv_names.iteritems():
8325
      self.lu.LogInfo("Remove logical volumes for %s" % name)
8326

    
8327
      for lv in old_lvs:
8328
        self.cfg.SetDiskID(lv, node_name)
8329

    
8330
        msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
8331
        if msg:
8332
          self.lu.LogWarning("Can't remove old LV: %s" % msg,
8333
                             hint="remove unused LVs manually")
8334

    
8335
  def _ReleaseNodeLock(self, node_name):
8336
    """Releases the lock for a given node."""
8337
    self.lu.context.glm.release(locking.LEVEL_NODE, node_name)
8338

    
8339
  def _ExecDrbd8DiskOnly(self, feedback_fn):
8340
    """Replace a disk on the primary or secondary for DRBD 8.
8341

8342
    The algorithm for replace is quite complicated:
8343

8344
      1. for each disk to be replaced:
8345

8346
        1. create new LVs on the target node with unique names
8347
        1. detach old LVs from the drbd device
8348
        1. rename old LVs to name_replaced.<time_t>
8349
        1. rename new LVs to old LVs
8350
        1. attach the new LVs (with the old names now) to the drbd device
8351

8352
      1. wait for sync across all devices
8353

8354
      1. for each modified disk:
8355

8356
        1. remove old LVs (which have the name name_replaces.<time_t>)
8357

8358
    Failures are not very well handled.
8359

8360
    """
8361
    steps_total = 6
8362

    
8363
    # Step: check device activation
8364
    self.lu.LogStep(1, steps_total, "Check device existence")
8365
    self._CheckDisksExistence([self.other_node, self.target_node])
8366
    self._CheckVolumeGroup([self.target_node, self.other_node])
8367

    
8368
    # Step: check other node consistency
8369
    self.lu.LogStep(2, steps_total, "Check peer consistency")
8370
    self._CheckDisksConsistency(self.other_node,
8371
                                self.other_node == self.instance.primary_node,
8372
                                False)
8373

    
8374
    # Step: create new storage
8375
    self.lu.LogStep(3, steps_total, "Allocate new storage")
8376
    iv_names = self._CreateNewStorage(self.target_node)
8377

    
8378
    # Step: for each lv, detach+rename*2+attach
8379
    self.lu.LogStep(4, steps_total, "Changing drbd configuration")
8380
    for dev, old_lvs, new_lvs in iv_names.itervalues():
8381
      self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
8382

    
8383
      result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
8384
                                                     old_lvs)
8385
      result.Raise("Can't detach drbd from local storage on node"
8386
                   " %s for device %s" % (self.target_node, dev.iv_name))
8387
      #dev.children = []
8388
      #cfg.Update(instance)
8389

    
8390
      # ok, we created the new LVs, so now we know we have the needed
8391
      # storage; as such, we proceed on the target node to rename
8392
      # old_lv to _old, and new_lv to old_lv; note that we rename LVs
8393
      # using the assumption that logical_id == physical_id (which in
8394
      # turn is the unique_id on that node)
8395

    
8396
      # FIXME(iustin): use a better name for the replaced LVs
8397
      temp_suffix = int(time.time())
8398
      ren_fn = lambda d, suff: (d.physical_id[0],
8399
                                d.physical_id[1] + "_replaced-%s" % suff)
8400

    
8401
      # Build the rename list based on what LVs exist on the node
8402
      rename_old_to_new = []
8403
      for to_ren in old_lvs:
8404
        result = self.rpc.call_blockdev_find(self.target_node, to_ren)
8405
        if not result.fail_msg and result.payload:
8406
          # device exists
8407
          rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
8408

    
8409
      self.lu.LogInfo("Renaming the old LVs on the target node")
8410
      result = self.rpc.call_blockdev_rename(self.target_node,
8411
                                             rename_old_to_new)
8412
      result.Raise("Can't rename old LVs on node %s" % self.target_node)
8413

    
8414
      # Now we rename the new LVs to the old LVs
8415
      self.lu.LogInfo("Renaming the new LVs on the target node")
8416
      rename_new_to_old = [(new, old.physical_id)
8417
                           for old, new in zip(old_lvs, new_lvs)]
8418
      result = self.rpc.call_blockdev_rename(self.target_node,
8419
                                             rename_new_to_old)
8420
      result.Raise("Can't rename new LVs on node %s" % self.target_node)
8421

    
8422
      for old, new in zip(old_lvs, new_lvs):
8423
        new.logical_id = old.logical_id
8424
        self.cfg.SetDiskID(new, self.target_node)
8425

    
8426
      for disk in old_lvs:
8427
        disk.logical_id = ren_fn(disk, temp_suffix)
8428
        self.cfg.SetDiskID(disk, self.target_node)
8429

    
8430
      # Now that the new lvs have the old name, we can add them to the device
8431
      self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
8432
      result = self.rpc.call_blockdev_addchildren(self.target_node, dev,
8433
                                                  new_lvs)
8434
      msg = result.fail_msg
8435
      if msg:
8436
        for new_lv in new_lvs:
8437
          msg2 = self.rpc.call_blockdev_remove(self.target_node,
8438
                                               new_lv).fail_msg
8439
          if msg2:
8440
            self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
8441
                               hint=("cleanup manually the unused logical"
8442
                                     "volumes"))
8443
        raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
8444

    
8445
      dev.children = new_lvs
8446

    
8447
      self.cfg.Update(self.instance, feedback_fn)
8448

    
8449
    cstep = 5
8450
    if self.early_release:
8451
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
8452
      cstep += 1
8453
      self._RemoveOldStorage(self.target_node, iv_names)
8454
      # WARNING: we release both node locks here, do not do other RPCs
8455
      # than WaitForSync to the primary node
8456
      self._ReleaseNodeLock([self.target_node, self.other_node])
8457

    
8458
    # Wait for sync
8459
    # This can fail as the old devices are degraded and _WaitForSync
8460
    # does a combined result over all disks, so we don't check its return value
8461
    self.lu.LogStep(cstep, steps_total, "Sync devices")
8462
    cstep += 1
8463
    _WaitForSync(self.lu, self.instance)
8464

    
8465
    # Check all devices manually
8466
    self._CheckDevices(self.instance.primary_node, iv_names)
8467

    
8468
    # Step: remove old storage
8469
    if not self.early_release:
8470
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
8471
      cstep += 1
8472
      self._RemoveOldStorage(self.target_node, iv_names)
8473

    
8474
  def _ExecDrbd8Secondary(self, feedback_fn):
8475
    """Replace the secondary node for DRBD 8.
8476

8477
    The algorithm for replace is quite complicated:
8478
      - for all disks of the instance:
8479
        - create new LVs on the new node with same names
8480
        - shutdown the drbd device on the old secondary
8481
        - disconnect the drbd network on the primary
8482
        - create the drbd device on the new secondary
8483
        - network attach the drbd on the primary, using an artifice:
8484
          the drbd code for Attach() will connect to the network if it
8485
          finds a device which is connected to the good local disks but
8486
          not network enabled
8487
      - wait for sync across all devices
8488
      - remove all disks from the old secondary
8489

8490
    Failures are not very well handled.
8491

8492
    """
8493
    steps_total = 6
8494

    
8495
    # Step: check device activation
8496
    self.lu.LogStep(1, steps_total, "Check device existence")
8497
    self._CheckDisksExistence([self.instance.primary_node])
8498
    self._CheckVolumeGroup([self.instance.primary_node])
8499

    
8500
    # Step: check other node consistency
8501
    self.lu.LogStep(2, steps_total, "Check peer consistency")
8502
    self._CheckDisksConsistency(self.instance.primary_node, True, True)
8503

    
8504
    # Step: create new storage
8505
    self.lu.LogStep(3, steps_total, "Allocate new storage")
8506
    for idx, dev in enumerate(self.instance.disks):
8507
      self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
8508
                      (self.new_node, idx))
8509
      # we pass force_create=True to force LVM creation
8510
      for new_lv in dev.children:
8511
        _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
8512
                        _GetInstanceInfoText(self.instance), False)
8513

    
8514
    # Step 4: dbrd minors and drbd setups changes
8515
    # after this, we must manually remove the drbd minors on both the
8516
    # error and the success paths
8517
    self.lu.LogStep(4, steps_total, "Changing drbd configuration")
8518
    minors = self.cfg.AllocateDRBDMinor([self.new_node
8519
                                         for dev in self.instance.disks],
8520
                                        self.instance.name)
8521
    logging.debug("Allocated minors %r", minors)
8522

    
8523
    iv_names = {}
8524
    for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
8525
      self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
8526
                      (self.new_node, idx))
8527
      # create new devices on new_node; note that we create two IDs:
8528
      # one without port, so the drbd will be activated without
8529
      # networking information on the new node at this stage, and one
8530
      # with network, for the latter activation in step 4
8531
      (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
8532
      if self.instance.primary_node == o_node1:
8533
        p_minor = o_minor1
8534
      else:
8535
        assert self.instance.primary_node == o_node2, "Three-node instance?"
8536
        p_minor = o_minor2
8537

    
8538
      new_alone_id = (self.instance.primary_node, self.new_node, None,
8539
                      p_minor, new_minor, o_secret)
8540
      new_net_id = (self.instance.primary_node, self.new_node, o_port,
8541
                    p_minor, new_minor, o_secret)
8542

    
8543
      iv_names[idx] = (dev, dev.children, new_net_id)
8544
      logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
8545
                    new_net_id)
8546
      new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
8547
                              logical_id=new_alone_id,
8548
                              children=dev.children,
8549
                              size=dev.size)
8550
      try:
8551
        _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
8552
                              _GetInstanceInfoText(self.instance), False)
8553
      except errors.GenericError:
8554
        self.cfg.ReleaseDRBDMinors(self.instance.name)
8555
        raise
8556

    
8557
    # We have new devices, shutdown the drbd on the old secondary
8558
    for idx, dev in enumerate(self.instance.disks):
8559
      self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
8560
      self.cfg.SetDiskID(dev, self.target_node)
8561
      msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
8562
      if msg:
8563
        self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
8564
                           "node: %s" % (idx, msg),
8565
                           hint=("Please cleanup this device manually as"
8566
                                 " soon as possible"))
8567

    
8568
    self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
8569
    result = self.rpc.call_drbd_disconnect_net([self.instance.primary_node],
8570
                                               self.node_secondary_ip,
8571
                                               self.instance.disks)\
8572
                                              [self.instance.primary_node]
8573

    
8574
    msg = result.fail_msg
8575
    if msg:
8576
      # detaches didn't succeed (unlikely)
8577
      self.cfg.ReleaseDRBDMinors(self.instance.name)
8578
      raise errors.OpExecError("Can't detach the disks from the network on"
8579
                               " old node: %s" % (msg,))
8580

    
8581
    # if we managed to detach at least one, we update all the disks of
8582
    # the instance to point to the new secondary
8583
    self.lu.LogInfo("Updating instance configuration")
8584
    for dev, _, new_logical_id in iv_names.itervalues():
8585
      dev.logical_id = new_logical_id
8586
      self.cfg.SetDiskID(dev, self.instance.primary_node)
8587

    
8588
    self.cfg.Update(self.instance, feedback_fn)
8589

    
8590
    # and now perform the drbd attach
8591
    self.lu.LogInfo("Attaching primary drbds to new secondary"
8592
                    " (standalone => connected)")
8593
    result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
8594
                                            self.new_node],
8595
                                           self.node_secondary_ip,
8596
                                           self.instance.disks,
8597
                                           self.instance.name,
8598
                                           False)
8599
    for to_node, to_result in result.items():
8600
      msg = to_result.fail_msg
8601
      if msg:
8602
        self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
8603
                           to_node, msg,
8604
                           hint=("please do a gnt-instance info to see the"
8605
                                 " status of disks"))
8606
    cstep = 5
8607
    if self.early_release:
8608
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
8609
      cstep += 1
8610
      self._RemoveOldStorage(self.target_node, iv_names)
8611
      # WARNING: we release all node locks here, do not do other RPCs
8612
      # than WaitForSync to the primary node
8613
      self._ReleaseNodeLock([self.instance.primary_node,
8614
                             self.target_node,
8615
                             self.new_node])
8616

    
8617
    # Wait for sync
8618
    # This can fail as the old devices are degraded and _WaitForSync
8619
    # does a combined result over all disks, so we don't check its return value
8620
    self.lu.LogStep(cstep, steps_total, "Sync devices")
8621
    cstep += 1
8622
    _WaitForSync(self.lu, self.instance)
8623

    
8624
    # Check all devices manually
8625
    self._CheckDevices(self.instance.primary_node, iv_names)
8626

    
8627
    # Step: remove old storage
8628
    if not self.early_release:
8629
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
8630
      self._RemoveOldStorage(self.target_node, iv_names)
8631

    
8632

    
8633
class LURepairNodeStorage(NoHooksLU):
8634
  """Repairs the volume group on a node.
8635

8636
  """
8637
  REQ_BGL = False
8638

    
8639
  def CheckArguments(self):
8640
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
8641

    
8642
    storage_type = self.op.storage_type
8643

    
8644
    if (constants.SO_FIX_CONSISTENCY not in
8645
        constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
8646
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
8647
                                 " repaired" % storage_type,
8648
                                 errors.ECODE_INVAL)
8649

    
8650
  def ExpandNames(self):
8651
    self.needed_locks = {
8652
      locking.LEVEL_NODE: [self.op.node_name],
8653
      }
8654

    
8655
  def _CheckFaultyDisks(self, instance, node_name):
8656
    """Ensure faulty disks abort the opcode or at least warn."""
8657
    try:
8658
      if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
8659
                                  node_name, True):
8660
        raise errors.OpPrereqError("Instance '%s' has faulty disks on"
8661
                                   " node '%s'" % (instance.name, node_name),
8662
                                   errors.ECODE_STATE)
8663
    except errors.OpPrereqError, err:
8664
      if self.op.ignore_consistency:
8665
        self.proc.LogWarning(str(err.args[0]))
8666
      else:
8667
        raise
8668

    
8669
  def CheckPrereq(self):
8670
    """Check prerequisites.
8671

8672
    """
8673
    # Check whether any instance on this node has faulty disks
8674
    for inst in _GetNodeInstances(self.cfg, self.op.node_name):
8675
      if not inst.admin_up:
8676
        continue
8677
      check_nodes = set(inst.all_nodes)
8678
      check_nodes.discard(self.op.node_name)
8679
      for inst_node_name in check_nodes:
8680
        self._CheckFaultyDisks(inst, inst_node_name)
8681

    
8682
  def Exec(self, feedback_fn):
8683
    feedback_fn("Repairing storage unit '%s' on %s ..." %
8684
                (self.op.name, self.op.node_name))
8685

    
8686
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
8687
    result = self.rpc.call_storage_execute(self.op.node_name,
8688
                                           self.op.storage_type, st_args,
8689
                                           self.op.name,
8690
                                           constants.SO_FIX_CONSISTENCY)
8691
    result.Raise("Failed to repair storage unit '%s' on %s" %
8692
                 (self.op.name, self.op.node_name))
8693

    
8694

    
8695
class LUNodeEvacStrategy(NoHooksLU):
8696
  """Computes the node evacuation strategy.
8697

8698
  """
8699
  REQ_BGL = False
8700

    
8701
  def CheckArguments(self):
8702
    _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
8703

    
8704
  def ExpandNames(self):
8705
    self.op.nodes = _GetWantedNodes(self, self.op.nodes)
8706
    self.needed_locks = locks = {}
8707
    if self.op.remote_node is None:
8708
      locks[locking.LEVEL_NODE] = locking.ALL_SET
8709
    else:
8710
      self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8711
      locks[locking.LEVEL_NODE] = self.op.nodes + [self.op.remote_node]
8712

    
8713
  def Exec(self, feedback_fn):
8714
    if self.op.remote_node is not None:
8715
      instances = []
8716
      for node in self.op.nodes:
8717
        instances.extend(_GetNodeSecondaryInstances(self.cfg, node))
8718
      result = []
8719
      for i in instances:
8720
        if i.primary_node == self.op.remote_node:
8721
          raise errors.OpPrereqError("Node %s is the primary node of"
8722
                                     " instance %s, cannot use it as"
8723
                                     " secondary" %
8724
                                     (self.op.remote_node, i.name),
8725
                                     errors.ECODE_INVAL)
8726
        result.append([i.name, self.op.remote_node])
8727
    else:
8728
      ial = IAllocator(self.cfg, self.rpc,
8729
                       mode=constants.IALLOCATOR_MODE_MEVAC,
8730
                       evac_nodes=self.op.nodes)
8731
      ial.Run(self.op.iallocator, validate=True)
8732
      if not ial.success:
8733
        raise errors.OpExecError("No valid evacuation solution: %s" % ial.info,
8734
                                 errors.ECODE_NORES)
8735
      result = ial.result
8736
    return result
8737

    
8738

    
8739
class LUInstanceGrowDisk(LogicalUnit):
8740
  """Grow a disk of an instance.
8741

8742
  """
8743
  HPATH = "disk-grow"
8744
  HTYPE = constants.HTYPE_INSTANCE
8745
  REQ_BGL = False
8746

    
8747
  def ExpandNames(self):
8748
    self._ExpandAndLockInstance()
8749
    self.needed_locks[locking.LEVEL_NODE] = []
8750
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8751

    
8752
  def DeclareLocks(self, level):
8753
    if level == locking.LEVEL_NODE:
8754
      self._LockInstancesNodes()
8755

    
8756
  def BuildHooksEnv(self):
8757
    """Build hooks env.
8758

8759
    This runs on the master, the primary and all the secondaries.
8760

8761
    """
8762
    env = {
8763
      "DISK": self.op.disk,
8764
      "AMOUNT": self.op.amount,
8765
      }
8766
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
8767
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
8768
    return env, nl, nl
8769

    
8770
  def CheckPrereq(self):
8771
    """Check prerequisites.
8772

8773
    This checks that the instance is in the cluster.
8774

8775
    """
8776
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
8777
    assert instance is not None, \
8778
      "Cannot retrieve locked instance %s" % self.op.instance_name
8779
    nodenames = list(instance.all_nodes)
8780
    for node in nodenames:
8781
      _CheckNodeOnline(self, node)
8782

    
8783
    self.instance = instance
8784

    
8785
    if instance.disk_template not in constants.DTS_GROWABLE:
8786
      raise errors.OpPrereqError("Instance's disk layout does not support"
8787
                                 " growing.", errors.ECODE_INVAL)
8788

    
8789
    self.disk = instance.FindDisk(self.op.disk)
8790

    
8791
    if instance.disk_template != constants.DT_FILE:
8792
      # TODO: check the free disk space for file, when that feature
8793
      # will be supported
8794
      _CheckNodesFreeDiskPerVG(self, nodenames,
8795
                               self.disk.ComputeGrowth(self.op.amount))
8796

    
8797
  def Exec(self, feedback_fn):
8798
    """Execute disk grow.
8799

8800
    """
8801
    instance = self.instance
8802
    disk = self.disk
8803

    
8804
    disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk])
8805
    if not disks_ok:
8806
      raise errors.OpExecError("Cannot activate block device to grow")
8807

    
8808
    for node in instance.all_nodes:
8809
      self.cfg.SetDiskID(disk, node)
8810
      result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
8811
      result.Raise("Grow request failed to node %s" % node)
8812

    
8813
      # TODO: Rewrite code to work properly
8814
      # DRBD goes into sync mode for a short amount of time after executing the
8815
      # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
8816
      # calling "resize" in sync mode fails. Sleeping for a short amount of
8817
      # time is a work-around.
8818
      time.sleep(5)
8819

    
8820
    disk.RecordGrow(self.op.amount)
8821
    self.cfg.Update(instance, feedback_fn)
8822
    if self.op.wait_for_sync:
8823
      disk_abort = not _WaitForSync(self, instance, disks=[disk])
8824
      if disk_abort:
8825
        self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
8826
                             " status.\nPlease check the instance.")
8827
      if not instance.admin_up:
8828
        _SafeShutdownInstanceDisks(self, instance, disks=[disk])
8829
    elif not instance.admin_up:
8830
      self.proc.LogWarning("Not shutting down the disk even if the instance is"
8831
                           " not supposed to be running because no wait for"
8832
                           " sync mode was requested.")
8833

    
8834

    
8835
class LUInstanceQueryData(NoHooksLU):
8836
  """Query runtime instance data.
8837

8838
  """
8839
  REQ_BGL = False
8840

    
8841
  def ExpandNames(self):
8842
    self.needed_locks = {}
8843

    
8844
    # Use locking if requested or when non-static information is wanted
8845
    if not (self.op.static or self.op.use_locking):
8846
      self.LogWarning("Non-static data requested, locks need to be acquired")
8847
      self.op.use_locking = True
8848

    
8849
    if self.op.instances or not self.op.use_locking:
8850
      # Expand instance names right here
8851
      self.wanted_names = _GetWantedInstances(self, self.op.instances)
8852
    else:
8853
      # Will use acquired locks
8854
      self.wanted_names = None
8855

    
8856
    if self.op.use_locking:
8857
      self.share_locks = dict.fromkeys(locking.LEVELS, 1)
8858

    
8859
      if self.wanted_names is None:
8860
        self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
8861
      else:
8862
        self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
8863

    
8864
      self.needed_locks[locking.LEVEL_NODE] = []
8865
      self.share_locks = dict.fromkeys(locking.LEVELS, 1)
8866
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8867

    
8868
  def DeclareLocks(self, level):
8869
    if self.op.use_locking and level == locking.LEVEL_NODE:
8870
      self._LockInstancesNodes()
8871

    
8872
  def CheckPrereq(self):
8873
    """Check prerequisites.
8874

8875
    This only checks the optional instance list against the existing names.
8876

8877
    """
8878
    if self.wanted_names is None:
8879
      assert self.op.use_locking, "Locking was not used"
8880
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
8881

    
8882
    self.wanted_instances = [self.cfg.GetInstanceInfo(name)
8883
                             for name in self.wanted_names]
8884

    
8885
  def _ComputeBlockdevStatus(self, node, instance_name, dev):
8886
    """Returns the status of a block device
8887

8888
    """
8889
    if self.op.static or not node:
8890
      return None
8891

    
8892
    self.cfg.SetDiskID(dev, node)
8893

    
8894
    result = self.rpc.call_blockdev_find(node, dev)
8895
    if result.offline:
8896
      return None
8897

    
8898
    result.Raise("Can't compute disk status for %s" % instance_name)
8899

    
8900
    status = result.payload
8901
    if status is None:
8902
      return None
8903

    
8904
    return (status.dev_path, status.major, status.minor,
8905
            status.sync_percent, status.estimated_time,
8906
            status.is_degraded, status.ldisk_status)
8907

    
8908
  def _ComputeDiskStatus(self, instance, snode, dev):
8909
    """Compute block device status.
8910

8911
    """
8912
    if dev.dev_type in constants.LDS_DRBD:
8913
      # we change the snode then (otherwise we use the one passed in)
8914
      if dev.logical_id[0] == instance.primary_node:
8915
        snode = dev.logical_id[1]
8916
      else:
8917
        snode = dev.logical_id[0]
8918

    
8919
    dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
8920
                                              instance.name, dev)
8921
    dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
8922

    
8923
    if dev.children:
8924
      dev_children = [self._ComputeDiskStatus(instance, snode, child)
8925
                      for child in dev.children]
8926
    else:
8927
      dev_children = []
8928

    
8929
    return {
8930
      "iv_name": dev.iv_name,
8931
      "dev_type": dev.dev_type,
8932
      "logical_id": dev.logical_id,
8933
      "physical_id": dev.physical_id,
8934
      "pstatus": dev_pstatus,
8935
      "sstatus": dev_sstatus,
8936
      "children": dev_children,
8937
      "mode": dev.mode,
8938
      "size": dev.size,
8939
      }
8940

    
8941
  def Exec(self, feedback_fn):
8942
    """Gather and return data"""
8943
    result = {}
8944

    
8945
    cluster = self.cfg.GetClusterInfo()
8946

    
8947
    for instance in self.wanted_instances:
8948
      if not self.op.static:
8949
        remote_info = self.rpc.call_instance_info(instance.primary_node,
8950
                                                  instance.name,
8951
                                                  instance.hypervisor)
8952
        remote_info.Raise("Error checking node %s" % instance.primary_node)
8953
        remote_info = remote_info.payload
8954
        if remote_info and "state" in remote_info:
8955
          remote_state = "up"
8956
        else:
8957
          remote_state = "down"
8958
      else:
8959
        remote_state = None
8960
      if instance.admin_up:
8961
        config_state = "up"
8962
      else:
8963
        config_state = "down"
8964

    
8965
      disks = [self._ComputeDiskStatus(instance, None, device)
8966
               for device in instance.disks]
8967

    
8968
      result[instance.name] = {
8969
        "name": instance.name,
8970
        "config_state": config_state,
8971
        "run_state": remote_state,
8972
        "pnode": instance.primary_node,
8973
        "snodes": instance.secondary_nodes,
8974
        "os": instance.os,
8975
        # this happens to be the same format used for hooks
8976
        "nics": _NICListToTuple(self, instance.nics),
8977
        "disk_template": instance.disk_template,
8978
        "disks": disks,
8979
        "hypervisor": instance.hypervisor,
8980
        "network_port": instance.network_port,
8981
        "hv_instance": instance.hvparams,
8982
        "hv_actual": cluster.FillHV(instance, skip_globals=True),
8983
        "be_instance": instance.beparams,
8984
        "be_actual": cluster.FillBE(instance),
8985
        "os_instance": instance.osparams,
8986
        "os_actual": cluster.SimpleFillOS(instance.os, instance.osparams),
8987
        "serial_no": instance.serial_no,
8988
        "mtime": instance.mtime,
8989
        "ctime": instance.ctime,
8990
        "uuid": instance.uuid,
8991
        }
8992

    
8993
    return result
8994

    
8995

    
8996
class LUInstanceSetParams(LogicalUnit):
8997
  """Modifies an instances's parameters.
8998

8999
  """
9000
  HPATH = "instance-modify"
9001
  HTYPE = constants.HTYPE_INSTANCE
9002
  REQ_BGL = False
9003

    
9004
  def CheckArguments(self):
9005
    if not (self.op.nics or self.op.disks or self.op.disk_template or
9006
            self.op.hvparams or self.op.beparams or self.op.os_name):
9007
      raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
9008

    
9009
    if self.op.hvparams:
9010
      _CheckGlobalHvParams(self.op.hvparams)
9011

    
9012
    # Disk validation
9013
    disk_addremove = 0
9014
    for disk_op, disk_dict in self.op.disks:
9015
      utils.ForceDictType(disk_dict, constants.IDISK_PARAMS_TYPES)
9016
      if disk_op == constants.DDM_REMOVE:
9017
        disk_addremove += 1
9018
        continue
9019
      elif disk_op == constants.DDM_ADD:
9020
        disk_addremove += 1
9021
      else:
9022
        if not isinstance(disk_op, int):
9023
          raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL)
9024
        if not isinstance(disk_dict, dict):
9025
          msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
9026
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
9027

    
9028
      if disk_op == constants.DDM_ADD:
9029
        mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
9030
        if mode not in constants.DISK_ACCESS_SET:
9031
          raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
9032
                                     errors.ECODE_INVAL)
9033
        size = disk_dict.get('size', None)
9034
        if size is None:
9035
          raise errors.OpPrereqError("Required disk parameter size missing",
9036
                                     errors.ECODE_INVAL)
9037
        try:
9038
          size = int(size)
9039
        except (TypeError, ValueError), err:
9040
          raise errors.OpPrereqError("Invalid disk size parameter: %s" %
9041
                                     str(err), errors.ECODE_INVAL)
9042
        disk_dict['size'] = size
9043
      else:
9044
        # modification of disk
9045
        if 'size' in disk_dict:
9046
          raise errors.OpPrereqError("Disk size change not possible, use"
9047
                                     " grow-disk", errors.ECODE_INVAL)
9048

    
9049
    if disk_addremove > 1:
9050
      raise errors.OpPrereqError("Only one disk add or remove operation"
9051
                                 " supported at a time", errors.ECODE_INVAL)
9052

    
9053
    if self.op.disks and self.op.disk_template is not None:
9054
      raise errors.OpPrereqError("Disk template conversion and other disk"
9055
                                 " changes not supported at the same time",
9056
                                 errors.ECODE_INVAL)
9057

    
9058
    if (self.op.disk_template and
9059
        self.op.disk_template in constants.DTS_NET_MIRROR and
9060
        self.op.remote_node is None):
9061
      raise errors.OpPrereqError("Changing the disk template to a mirrored"
9062
                                 " one requires specifying a secondary node",
9063
                                 errors.ECODE_INVAL)
9064

    
9065
    # NIC validation
9066
    nic_addremove = 0
9067
    for nic_op, nic_dict in self.op.nics:
9068
      utils.ForceDictType(nic_dict, constants.INIC_PARAMS_TYPES)
9069
      if nic_op == constants.DDM_REMOVE:
9070
        nic_addremove += 1
9071
        continue
9072
      elif nic_op == constants.DDM_ADD:
9073
        nic_addremove += 1
9074
      else:
9075
        if not isinstance(nic_op, int):
9076
          raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL)
9077
        if not isinstance(nic_dict, dict):
9078
          msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
9079
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
9080

    
9081
      # nic_dict should be a dict
9082
      nic_ip = nic_dict.get('ip', None)
9083
      if nic_ip is not None:
9084
        if nic_ip.lower() == constants.VALUE_NONE:
9085
          nic_dict['ip'] = None
9086
        else:
9087
          if not netutils.IPAddress.IsValid(nic_ip):
9088
            raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip,
9089
                                       errors.ECODE_INVAL)
9090

    
9091
      nic_bridge = nic_dict.get('bridge', None)
9092
      nic_link = nic_dict.get('link', None)
9093
      if nic_bridge and nic_link:
9094
        raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
9095
                                   " at the same time", errors.ECODE_INVAL)
9096
      elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
9097
        nic_dict['bridge'] = None
9098
      elif nic_link and nic_link.lower() == constants.VALUE_NONE:
9099
        nic_dict['link'] = None
9100

    
9101
      if nic_op == constants.DDM_ADD:
9102
        nic_mac = nic_dict.get('mac', None)
9103
        if nic_mac is None:
9104
          nic_dict['mac'] = constants.VALUE_AUTO
9105

    
9106
      if 'mac' in nic_dict:
9107
        nic_mac = nic_dict['mac']
9108
        if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
9109
          nic_mac = utils.NormalizeAndValidateMac(nic_mac)
9110

    
9111
        if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
9112
          raise errors.OpPrereqError("'auto' is not a valid MAC address when"
9113
                                     " modifying an existing nic",
9114
                                     errors.ECODE_INVAL)
9115

    
9116
    if nic_addremove > 1:
9117
      raise errors.OpPrereqError("Only one NIC add or remove operation"
9118
                                 " supported at a time", errors.ECODE_INVAL)
9119

    
9120
  def ExpandNames(self):
9121
    self._ExpandAndLockInstance()
9122
    self.needed_locks[locking.LEVEL_NODE] = []
9123
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
9124

    
9125
  def DeclareLocks(self, level):
9126
    if level == locking.LEVEL_NODE:
9127
      self._LockInstancesNodes()
9128
      if self.op.disk_template and self.op.remote_node:
9129
        self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
9130
        self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)
9131

    
9132
  def BuildHooksEnv(self):
9133
    """Build hooks env.
9134

9135
    This runs on the master, primary and secondaries.
9136

9137
    """
9138
    args = dict()
9139
    if constants.BE_MEMORY in self.be_new:
9140
      args['memory'] = self.be_new[constants.BE_MEMORY]
9141
    if constants.BE_VCPUS in self.be_new:
9142
      args['vcpus'] = self.be_new[constants.BE_VCPUS]
9143
    # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
9144
    # information at all.
9145
    if self.op.nics:
9146
      args['nics'] = []
9147
      nic_override = dict(self.op.nics)
9148
      for idx, nic in enumerate(self.instance.nics):
9149
        if idx in nic_override:
9150
          this_nic_override = nic_override[idx]
9151
        else:
9152
          this_nic_override = {}
9153
        if 'ip' in this_nic_override:
9154
          ip = this_nic_override['ip']
9155
        else:
9156
          ip = nic.ip
9157
        if 'mac' in this_nic_override:
9158
          mac = this_nic_override['mac']
9159
        else:
9160
          mac = nic.mac
9161
        if idx in self.nic_pnew:
9162
          nicparams = self.nic_pnew[idx]
9163
        else:
9164
          nicparams = self.cluster.SimpleFillNIC(nic.nicparams)
9165
        mode = nicparams[constants.NIC_MODE]
9166
        link = nicparams[constants.NIC_LINK]
9167
        args['nics'].append((ip, mac, mode, link))
9168
      if constants.DDM_ADD in nic_override:
9169
        ip = nic_override[constants.DDM_ADD].get('ip', None)
9170
        mac = nic_override[constants.DDM_ADD]['mac']
9171
        nicparams = self.nic_pnew[constants.DDM_ADD]
9172
        mode = nicparams[constants.NIC_MODE]
9173
        link = nicparams[constants.NIC_LINK]
9174
        args['nics'].append((ip, mac, mode, link))
9175
      elif constants.DDM_REMOVE in nic_override:
9176
        del args['nics'][-1]
9177

    
9178
    env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
9179
    if self.op.disk_template:
9180
      env["NEW_DISK_TEMPLATE"] = self.op.disk_template
9181
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
9182
    return env, nl, nl
9183

    
9184
  def CheckPrereq(self):
9185
    """Check prerequisites.
9186

9187
    This only checks the instance list against the existing names.
9188

9189
    """
9190
    # checking the new params on the primary/secondary nodes
9191

    
9192
    instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
9193
    cluster = self.cluster = self.cfg.GetClusterInfo()
9194
    assert self.instance is not None, \
9195
      "Cannot retrieve locked instance %s" % self.op.instance_name
9196
    pnode = instance.primary_node
9197
    nodelist = list(instance.all_nodes)
9198

    
9199
    # OS change
9200
    if self.op.os_name and not self.op.force:
9201
      _CheckNodeHasOS(self, instance.primary_node, self.op.os_name,
9202
                      self.op.force_variant)
9203
      instance_os = self.op.os_name
9204
    else:
9205
      instance_os = instance.os
9206

    
9207
    if self.op.disk_template:
9208
      if instance.disk_template == self.op.disk_template:
9209
        raise errors.OpPrereqError("Instance already has disk template %s" %
9210
                                   instance.disk_template, errors.ECODE_INVAL)
9211

    
9212
      if (instance.disk_template,
9213
          self.op.disk_template) not in self._DISK_CONVERSIONS:
9214
        raise errors.OpPrereqError("Unsupported disk template conversion from"
9215
                                   " %s to %s" % (instance.disk_template,
9216
                                                  self.op.disk_template),
9217
                                   errors.ECODE_INVAL)
9218
      _CheckInstanceDown(self, instance, "cannot change disk template")
9219
      if self.op.disk_template in constants.DTS_NET_MIRROR:
9220
        if self.op.remote_node == pnode:
9221
          raise errors.OpPrereqError("Given new secondary node %s is the same"
9222
                                     " as the primary node of the instance" %
9223
                                     self.op.remote_node, errors.ECODE_STATE)
9224
        _CheckNodeOnline(self, self.op.remote_node)
9225
        _CheckNodeNotDrained(self, self.op.remote_node)
9226
        # FIXME: here we assume that the old instance type is DT_PLAIN
9227
        assert instance.disk_template == constants.DT_PLAIN
9228
        disks = [{"size": d.size, "vg": d.logical_id[0]}
9229
                 for d in instance.disks]
9230
        required = _ComputeDiskSizePerVG(self.op.disk_template, disks)
9231
        _CheckNodesFreeDiskPerVG(self, [self.op.remote_node], required)
9232

    
9233
    # hvparams processing
9234
    if self.op.hvparams:
9235
      hv_type = instance.hypervisor
9236
      i_hvdict = _GetUpdatedParams(instance.hvparams, self.op.hvparams)
9237
      utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES)
9238
      hv_new = cluster.SimpleFillHV(hv_type, instance.os, i_hvdict)
9239

    
9240
      # local check
9241
      hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new)
9242
      _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
9243
      self.hv_new = hv_new # the new actual values
9244
      self.hv_inst = i_hvdict # the new dict (without defaults)
9245
    else:
9246
      self.hv_new = self.hv_inst = {}
9247

    
9248
    # beparams processing
9249
    if self.op.beparams:
9250
      i_bedict = _GetUpdatedParams(instance.beparams, self.op.beparams,
9251
                                   use_none=True)
9252
      utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
9253
      be_new = cluster.SimpleFillBE(i_bedict)
9254
      self.be_new = be_new # the new actual values
9255
      self.be_inst = i_bedict # the new dict (without defaults)
9256
    else:
9257
      self.be_new = self.be_inst = {}
9258

    
9259
    # osparams processing
9260
    if self.op.osparams:
9261
      i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
9262
      _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
9263
      self.os_inst = i_osdict # the new dict (without defaults)
9264
    else:
9265
      self.os_inst = {}
9266

    
9267
    self.warn = []
9268

    
9269
    if constants.BE_MEMORY in self.op.beparams and not self.op.force:
9270
      mem_check_list = [pnode]
9271
      if be_new[constants.BE_AUTO_BALANCE]:
9272
        # either we changed auto_balance to yes or it was from before
9273
        mem_check_list.extend(instance.secondary_nodes)
9274
      instance_info = self.rpc.call_instance_info(pnode, instance.name,
9275
                                                  instance.hypervisor)
9276
      nodeinfo = self.rpc.call_node_info(mem_check_list, None,
9277
                                         instance.hypervisor)
9278
      pninfo = nodeinfo[pnode]
9279
      msg = pninfo.fail_msg
9280
      if msg:
9281
        # Assume the primary node is unreachable and go ahead
9282
        self.warn.append("Can't get info from primary node %s: %s" %
9283
                         (pnode,  msg))
9284
      elif not isinstance(pninfo.payload.get('memory_free', None), int):
9285
        self.warn.append("Node data from primary node %s doesn't contain"
9286
                         " free memory information" % pnode)
9287
      elif instance_info.fail_msg:
9288
        self.warn.append("Can't get instance runtime information: %s" %
9289
                        instance_info.fail_msg)
9290
      else:
9291
        if instance_info.payload:
9292
          current_mem = int(instance_info.payload['memory'])
9293
        else:
9294
          # Assume instance not running
9295
          # (there is a slight race condition here, but it's not very probable,
9296
          # and we have no other way to check)
9297
          current_mem = 0
9298
        miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
9299
                    pninfo.payload['memory_free'])
9300
        if miss_mem > 0:
9301
          raise errors.OpPrereqError("This change will prevent the instance"
9302
                                     " from starting, due to %d MB of memory"
9303
                                     " missing on its primary node" % miss_mem,
9304
                                     errors.ECODE_NORES)
9305

    
9306
      if be_new[constants.BE_AUTO_BALANCE]:
9307
        for node, nres in nodeinfo.items():
9308
          if node not in instance.secondary_nodes:
9309
            continue
9310
          msg = nres.fail_msg
9311
          if msg:
9312
            self.warn.append("Can't get info from secondary node %s: %s" %
9313
                             (node, msg))
9314
          elif not isinstance(nres.payload.get('memory_free', None), int):
9315
            self.warn.append("Secondary node %s didn't return free"
9316
                             " memory information" % node)
9317
          elif be_new[constants.BE_MEMORY] > nres.payload['memory_free']:
9318
            self.warn.append("Not enough memory to failover instance to"
9319
                             " secondary node %s" % node)
9320

    
9321
    # NIC processing
9322
    self.nic_pnew = {}
9323
    self.nic_pinst = {}
9324
    for nic_op, nic_dict in self.op.nics:
9325
      if nic_op == constants.DDM_REMOVE:
9326
        if not instance.nics:
9327
          raise errors.OpPrereqError("Instance has no NICs, cannot remove",
9328
                                     errors.ECODE_INVAL)
9329
        continue
9330
      if nic_op != constants.DDM_ADD:
9331
        # an existing nic
9332
        if not instance.nics:
9333
          raise errors.OpPrereqError("Invalid NIC index %s, instance has"
9334
                                     " no NICs" % nic_op,
9335
                                     errors.ECODE_INVAL)
9336
        if nic_op < 0 or nic_op >= len(instance.nics):
9337
          raise errors.OpPrereqError("Invalid NIC index %s, valid values"
9338
                                     " are 0 to %d" %
9339
                                     (nic_op, len(instance.nics) - 1),
9340
                                     errors.ECODE_INVAL)
9341
        old_nic_params = instance.nics[nic_op].nicparams
9342
        old_nic_ip = instance.nics[nic_op].ip
9343
      else:
9344
        old_nic_params = {}
9345
        old_nic_ip = None
9346

    
9347
      update_params_dict = dict([(key, nic_dict[key])
9348
                                 for key in constants.NICS_PARAMETERS
9349
                                 if key in nic_dict])
9350

    
9351
      if 'bridge' in nic_dict:
9352
        update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
9353

    
9354
      new_nic_params = _GetUpdatedParams(old_nic_params,
9355
                                         update_params_dict)
9356
      utils.ForceDictType(new_nic_params, constants.NICS_PARAMETER_TYPES)
9357
      new_filled_nic_params = cluster.SimpleFillNIC(new_nic_params)
9358
      objects.NIC.CheckParameterSyntax(new_filled_nic_params)
9359
      self.nic_pinst[nic_op] = new_nic_params
9360
      self.nic_pnew[nic_op] = new_filled_nic_params
9361
      new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
9362

    
9363
      if new_nic_mode == constants.NIC_MODE_BRIDGED:
9364
        nic_bridge = new_filled_nic_params[constants.NIC_LINK]
9365
        msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
9366
        if msg:
9367
          msg = "Error checking bridges on node %s: %s" % (pnode, msg)
9368
          if self.op.force:
9369
            self.warn.append(msg)
9370
          else:
9371
            raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
9372
      if new_nic_mode == constants.NIC_MODE_ROUTED:
9373
        if 'ip' in nic_dict:
9374
          nic_ip = nic_dict['ip']
9375
        else:
9376
          nic_ip = old_nic_ip
9377
        if nic_ip is None:
9378
          raise errors.OpPrereqError('Cannot set the nic ip to None'
9379
                                     ' on a routed nic', errors.ECODE_INVAL)
9380
      if 'mac' in nic_dict:
9381
        nic_mac = nic_dict['mac']
9382
        if nic_mac is None:
9383
          raise errors.OpPrereqError('Cannot set the nic mac to None',
9384
                                     errors.ECODE_INVAL)
9385
        elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
9386
          # otherwise generate the mac
9387
          nic_dict['mac'] = self.cfg.GenerateMAC(self.proc.GetECId())
9388
        else:
9389
          # or validate/reserve the current one
9390
          try:
9391
            self.cfg.ReserveMAC(nic_mac, self.proc.GetECId())
9392
          except errors.ReservationError:
9393
            raise errors.OpPrereqError("MAC address %s already in use"
9394
                                       " in cluster" % nic_mac,
9395
                                       errors.ECODE_NOTUNIQUE)
9396

    
9397
    # DISK processing
9398
    if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
9399
      raise errors.OpPrereqError("Disk operations not supported for"
9400
                                 " diskless instances",
9401
                                 errors.ECODE_INVAL)
9402
    for disk_op, _ in self.op.disks:
9403
      if disk_op == constants.DDM_REMOVE:
9404
        if len(instance.disks) == 1:
9405
          raise errors.OpPrereqError("Cannot remove the last disk of"
9406
                                     " an instance", errors.ECODE_INVAL)
9407
        _CheckInstanceDown(self, instance, "cannot remove disks")
9408

    
9409
      if (disk_op == constants.DDM_ADD and
9410
          len(instance.disks) >= constants.MAX_DISKS):
9411
        raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
9412
                                   " add more" % constants.MAX_DISKS,
9413
                                   errors.ECODE_STATE)
9414
      if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
9415
        # an existing disk
9416
        if disk_op < 0 or disk_op >= len(instance.disks):
9417
          raise errors.OpPrereqError("Invalid disk index %s, valid values"
9418
                                     " are 0 to %d" %
9419
                                     (disk_op, len(instance.disks)),
9420
                                     errors.ECODE_INVAL)
9421

    
9422
    return
9423

    
9424
  def _ConvertPlainToDrbd(self, feedback_fn):
9425
    """Converts an instance from plain to drbd.
9426

9427
    """
9428
    feedback_fn("Converting template to drbd")
9429
    instance = self.instance
9430
    pnode = instance.primary_node
9431
    snode = self.op.remote_node
9432

    
9433
    # create a fake disk info for _GenerateDiskTemplate
9434
    disk_info = [{"size": d.size, "mode": d.mode} for d in instance.disks]
9435
    new_disks = _GenerateDiskTemplate(self, self.op.disk_template,
9436
                                      instance.name, pnode, [snode],
9437
                                      disk_info, None, None, 0, feedback_fn)
9438
    info = _GetInstanceInfoText(instance)
9439
    feedback_fn("Creating aditional volumes...")
9440
    # first, create the missing data and meta devices
9441
    for disk in new_disks:
9442
      # unfortunately this is... not too nice
9443
      _CreateSingleBlockDev(self, pnode, instance, disk.children[1],
9444
                            info, True)
9445
      for child in disk.children:
9446
        _CreateSingleBlockDev(self, snode, instance, child, info, True)
9447
    # at this stage, all new LVs have been created, we can rename the
9448
    # old ones
9449
    feedback_fn("Renaming original volumes...")
9450
    rename_list = [(o, n.children[0].logical_id)
9451
                   for (o, n) in zip(instance.disks, new_disks)]
9452
    result = self.rpc.call_blockdev_rename(pnode, rename_list)
9453
    result.Raise("Failed to rename original LVs")
9454

    
9455
    feedback_fn("Initializing DRBD devices...")
9456
    # all child devices are in place, we can now create the DRBD devices
9457
    for disk in new_disks:
9458
      for node in [pnode, snode]:
9459
        f_create = node == pnode
9460
        _CreateSingleBlockDev(self, node, instance, disk, info, f_create)
9461

    
9462
    # at this point, the instance has been modified
9463
    instance.disk_template = constants.DT_DRBD8
9464
    instance.disks = new_disks
9465
    self.cfg.Update(instance, feedback_fn)
9466

    
9467
    # disks are created, waiting for sync
9468
    disk_abort = not _WaitForSync(self, instance)
9469
    if disk_abort:
9470
      raise errors.OpExecError("There are some degraded disks for"
9471
                               " this instance, please cleanup manually")
9472

    
9473
  def _ConvertDrbdToPlain(self, feedback_fn):
9474
    """Converts an instance from drbd to plain.
9475

9476
    """
9477
    instance = self.instance
9478
    assert len(instance.secondary_nodes) == 1
9479
    pnode = instance.primary_node
9480
    snode = instance.secondary_nodes[0]
9481
    feedback_fn("Converting template to plain")
9482

    
9483
    old_disks = instance.disks
9484
    new_disks = [d.children[0] for d in old_disks]
9485

    
9486
    # copy over size and mode
9487
    for parent, child in zip(old_disks, new_disks):
9488
      child.size = parent.size
9489
      child.mode = parent.mode
9490

    
9491
    # update instance structure
9492
    instance.disks = new_disks
9493
    instance.disk_template = constants.DT_PLAIN
9494
    self.cfg.Update(instance, feedback_fn)
9495

    
9496
    feedback_fn("Removing volumes on the secondary node...")
9497
    for disk in old_disks:
9498
      self.cfg.SetDiskID(disk, snode)
9499
      msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg
9500
      if msg:
9501
        self.LogWarning("Could not remove block device %s on node %s,"
9502
                        " continuing anyway: %s", disk.iv_name, snode, msg)
9503

    
9504
    feedback_fn("Removing unneeded volumes on the primary node...")
9505
    for idx, disk in enumerate(old_disks):
9506
      meta = disk.children[1]
9507
      self.cfg.SetDiskID(meta, pnode)
9508
      msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg
9509
      if msg:
9510
        self.LogWarning("Could not remove metadata for disk %d on node %s,"
9511
                        " continuing anyway: %s", idx, pnode, msg)
9512

    
9513
  def Exec(self, feedback_fn):
9514
    """Modifies an instance.
9515

9516
    All parameters take effect only at the next restart of the instance.
9517

9518
    """
9519
    # Process here the warnings from CheckPrereq, as we don't have a
9520
    # feedback_fn there.
9521
    for warn in self.warn:
9522
      feedback_fn("WARNING: %s" % warn)
9523

    
9524
    result = []
9525
    instance = self.instance
9526
    # disk changes
9527
    for disk_op, disk_dict in self.op.disks:
9528
      if disk_op == constants.DDM_REMOVE:
9529
        # remove the last disk
9530
        device = instance.disks.pop()
9531
        device_idx = len(instance.disks)
9532
        for node, disk in device.ComputeNodeTree(instance.primary_node):
9533
          self.cfg.SetDiskID(disk, node)
9534
          msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
9535
          if msg:
9536
            self.LogWarning("Could not remove disk/%d on node %s: %s,"
9537
                            " continuing anyway", device_idx, node, msg)
9538
        result.append(("disk/%d" % device_idx, "remove"))
9539
      elif disk_op == constants.DDM_ADD:
9540
        # add a new disk
9541
        if instance.disk_template == constants.DT_FILE:
9542
          file_driver, file_path = instance.disks[0].logical_id
9543
          file_path = os.path.dirname(file_path)
9544
        else:
9545
          file_driver = file_path = None
9546
        disk_idx_base = len(instance.disks)
9547
        new_disk = _GenerateDiskTemplate(self,
9548
                                         instance.disk_template,
9549
                                         instance.name, instance.primary_node,
9550
                                         instance.secondary_nodes,
9551
                                         [disk_dict],
9552
                                         file_path,
9553
                                         file_driver,
9554
                                         disk_idx_base, feedback_fn)[0]
9555
        instance.disks.append(new_disk)
9556
        info = _GetInstanceInfoText(instance)
9557

    
9558
        logging.info("Creating volume %s for instance %s",
9559
                     new_disk.iv_name, instance.name)
9560
        # Note: this needs to be kept in sync with _CreateDisks
9561
        #HARDCODE
9562
        for node in instance.all_nodes:
9563
          f_create = node == instance.primary_node
9564
          try:
9565
            _CreateBlockDev(self, node, instance, new_disk,
9566
                            f_create, info, f_create)
9567
          except errors.OpExecError, err:
9568
            self.LogWarning("Failed to create volume %s (%s) on"
9569
                            " node %s: %s",
9570
                            new_disk.iv_name, new_disk, node, err)
9571
        result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
9572
                       (new_disk.size, new_disk.mode)))
9573
      else:
9574
        # change a given disk
9575
        instance.disks[disk_op].mode = disk_dict['mode']
9576
        result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
9577

    
9578
    if self.op.disk_template:
9579
      r_shut = _ShutdownInstanceDisks(self, instance)
9580
      if not r_shut:
9581
        raise errors.OpExecError("Cannot shutdown instance disks, unable to"
9582
                                 " proceed with disk template conversion")
9583
      mode = (instance.disk_template, self.op.disk_template)
9584
      try:
9585
        self._DISK_CONVERSIONS[mode](self, feedback_fn)
9586
      except:
9587
        self.cfg.ReleaseDRBDMinors(instance.name)
9588
        raise
9589
      result.append(("disk_template", self.op.disk_template))
9590

    
9591
    # NIC changes
9592
    for nic_op, nic_dict in self.op.nics:
9593
      if nic_op == constants.DDM_REMOVE:
9594
        # remove the last nic
9595
        del instance.nics[-1]
9596
        result.append(("nic.%d" % len(instance.nics), "remove"))
9597
      elif nic_op == constants.DDM_ADD:
9598
        # mac and bridge should be set, by now
9599
        mac = nic_dict['mac']
9600
        ip = nic_dict.get('ip', None)
9601
        nicparams = self.nic_pinst[constants.DDM_ADD]
9602
        new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
9603
        instance.nics.append(new_nic)
9604
        result.append(("nic.%d" % (len(instance.nics) - 1),
9605
                       "add:mac=%s,ip=%s,mode=%s,link=%s" %
9606
                       (new_nic.mac, new_nic.ip,
9607
                        self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
9608
                        self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
9609
                       )))
9610
      else:
9611
        for key in 'mac', 'ip':
9612
          if key in nic_dict:
9613
            setattr(instance.nics[nic_op], key, nic_dict[key])
9614
        if nic_op in self.nic_pinst:
9615
          instance.nics[nic_op].nicparams = self.nic_pinst[nic_op]
9616
        for key, val in nic_dict.iteritems():
9617
          result.append(("nic.%s/%d" % (key, nic_op), val))
9618

    
9619
    # hvparams changes
9620
    if self.op.hvparams:
9621
      instance.hvparams = self.hv_inst
9622
      for key, val in self.op.hvparams.iteritems():
9623
        result.append(("hv/%s" % key, val))
9624

    
9625
    # beparams changes
9626
    if self.op.beparams:
9627
      instance.beparams = self.be_inst
9628
      for key, val in self.op.beparams.iteritems():
9629
        result.append(("be/%s" % key, val))
9630

    
9631
    # OS change
9632
    if self.op.os_name:
9633
      instance.os = self.op.os_name
9634

    
9635
    # osparams changes
9636
    if self.op.osparams:
9637
      instance.osparams = self.os_inst
9638
      for key, val in self.op.osparams.iteritems():
9639
        result.append(("os/%s" % key, val))
9640

    
9641
    self.cfg.Update(instance, feedback_fn)
9642

    
9643
    return result
9644

    
9645
  _DISK_CONVERSIONS = {
9646
    (constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd,
9647
    (constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain,
9648
    }
9649

    
9650

    
9651
class LUBackupQuery(NoHooksLU):
9652
  """Query the exports list
9653

9654
  """
9655
  REQ_BGL = False
9656

    
9657
  def ExpandNames(self):
9658
    self.needed_locks = {}
9659
    self.share_locks[locking.LEVEL_NODE] = 1
9660
    if not self.op.nodes:
9661
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9662
    else:
9663
      self.needed_locks[locking.LEVEL_NODE] = \
9664
        _GetWantedNodes(self, self.op.nodes)
9665

    
9666
  def Exec(self, feedback_fn):
9667
    """Compute the list of all the exported system images.
9668

9669
    @rtype: dict
9670
    @return: a dictionary with the structure node->(export-list)
9671
        where export-list is a list of the instances exported on
9672
        that node.
9673

9674
    """
9675
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
9676
    rpcresult = self.rpc.call_export_list(self.nodes)
9677
    result = {}
9678
    for node in rpcresult:
9679
      if rpcresult[node].fail_msg:
9680
        result[node] = False
9681
      else:
9682
        result[node] = rpcresult[node].payload
9683

    
9684
    return result
9685

    
9686

    
9687
class LUBackupPrepare(NoHooksLU):
9688
  """Prepares an instance for an export and returns useful information.
9689

9690
  """
9691
  REQ_BGL = False
9692

    
9693
  def ExpandNames(self):
9694
    self._ExpandAndLockInstance()
9695

    
9696
  def CheckPrereq(self):
9697
    """Check prerequisites.
9698

9699
    """
9700
    instance_name = self.op.instance_name
9701

    
9702
    self.instance = self.cfg.GetInstanceInfo(instance_name)
9703
    assert self.instance is not None, \
9704
          "Cannot retrieve locked instance %s" % self.op.instance_name
9705
    _CheckNodeOnline(self, self.instance.primary_node)
9706

    
9707
    self._cds = _GetClusterDomainSecret()
9708

    
9709
  def Exec(self, feedback_fn):
9710
    """Prepares an instance for an export.
9711

9712
    """
9713
    instance = self.instance
9714

    
9715
    if self.op.mode == constants.EXPORT_MODE_REMOTE:
9716
      salt = utils.GenerateSecret(8)
9717

    
9718
      feedback_fn("Generating X509 certificate on %s" % instance.primary_node)
9719
      result = self.rpc.call_x509_cert_create(instance.primary_node,
9720
                                              constants.RIE_CERT_VALIDITY)
9721
      result.Raise("Can't create X509 key and certificate on %s" % result.node)
9722

    
9723
      (name, cert_pem) = result.payload
9724

    
9725
      cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
9726
                                             cert_pem)
9727

    
9728
      return {
9729
        "handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds),
9730
        "x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt),
9731
                          salt),
9732
        "x509_ca": utils.SignX509Certificate(cert, self._cds, salt),
9733
        }
9734

    
9735
    return None
9736

    
9737

    
9738
class LUBackupExport(LogicalUnit):
9739
  """Export an instance to an image in the cluster.
9740

9741
  """
9742
  HPATH = "instance-export"
9743
  HTYPE = constants.HTYPE_INSTANCE
9744
  REQ_BGL = False
9745

    
9746
  def CheckArguments(self):
9747
    """Check the arguments.
9748

9749
    """
9750
    self.x509_key_name = self.op.x509_key_name
9751
    self.dest_x509_ca_pem = self.op.destination_x509_ca
9752

    
9753
    if self.op.mode == constants.EXPORT_MODE_REMOTE:
9754
      if not self.x509_key_name:
9755
        raise errors.OpPrereqError("Missing X509 key name for encryption",
9756
                                   errors.ECODE_INVAL)
9757

    
9758
      if not self.dest_x509_ca_pem:
9759
        raise errors.OpPrereqError("Missing destination X509 CA",
9760
                                   errors.ECODE_INVAL)
9761

    
9762
  def ExpandNames(self):
9763
    self._ExpandAndLockInstance()
9764

    
9765
    # Lock all nodes for local exports
9766
    if self.op.mode == constants.EXPORT_MODE_LOCAL:
9767
      # FIXME: lock only instance primary and destination node
9768
      #
9769
      # Sad but true, for now we have do lock all nodes, as we don't know where
9770
      # the previous export might be, and in this LU we search for it and
9771
      # remove it from its current node. In the future we could fix this by:
9772
      #  - making a tasklet to search (share-lock all), then create the
9773
      #    new one, then one to remove, after
9774
      #  - removing the removal operation altogether
9775
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9776

    
9777
  def DeclareLocks(self, level):
9778
    """Last minute lock declaration."""
9779
    # All nodes are locked anyway, so nothing to do here.
9780

    
9781
  def BuildHooksEnv(self):
9782
    """Build hooks env.
9783

9784
    This will run on the master, primary node and target node.
9785

9786
    """
9787
    env = {
9788
      "EXPORT_MODE": self.op.mode,
9789
      "EXPORT_NODE": self.op.target_node,
9790
      "EXPORT_DO_SHUTDOWN": self.op.shutdown,
9791
      "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
9792
      # TODO: Generic function for boolean env variables
9793
      "REMOVE_INSTANCE": str(bool(self.op.remove_instance)),
9794
      }
9795

    
9796
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
9797

    
9798
    nl = [self.cfg.GetMasterNode(), self.instance.primary_node]
9799

    
9800
    if self.op.mode == constants.EXPORT_MODE_LOCAL:
9801
      nl.append(self.op.target_node)
9802

    
9803
    return env, nl, nl
9804

    
9805
  def CheckPrereq(self):
9806
    """Check prerequisites.
9807

9808
    This checks that the instance and node names are valid.
9809

9810
    """
9811
    instance_name = self.op.instance_name
9812

    
9813
    self.instance = self.cfg.GetInstanceInfo(instance_name)
9814
    assert self.instance is not None, \
9815
          "Cannot retrieve locked instance %s" % self.op.instance_name
9816
    _CheckNodeOnline(self, self.instance.primary_node)
9817

    
9818
    if (self.op.remove_instance and self.instance.admin_up and
9819
        not self.op.shutdown):
9820
      raise errors.OpPrereqError("Can not remove instance without shutting it"
9821
                                 " down before")
9822

    
9823
    if self.op.mode == constants.EXPORT_MODE_LOCAL:
9824
      self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
9825
      self.dst_node = self.cfg.GetNodeInfo(self.op.target_node)
9826
      assert self.dst_node is not None
9827

    
9828
      _CheckNodeOnline(self, self.dst_node.name)
9829
      _CheckNodeNotDrained(self, self.dst_node.name)
9830

    
9831
      self._cds = None
9832
      self.dest_disk_info = None
9833
      self.dest_x509_ca = None
9834

    
9835
    elif self.op.mode == constants.EXPORT_MODE_REMOTE:
9836
      self.dst_node = None
9837

    
9838
      if len(self.op.target_node) != len(self.instance.disks):
9839
        raise errors.OpPrereqError(("Received destination information for %s"
9840
                                    " disks, but instance %s has %s disks") %
9841
                                   (len(self.op.target_node), instance_name,
9842
                                    len(self.instance.disks)),
9843
                                   errors.ECODE_INVAL)
9844

    
9845
      cds = _GetClusterDomainSecret()
9846

    
9847
      # Check X509 key name
9848
      try:
9849
        (key_name, hmac_digest, hmac_salt) = self.x509_key_name
9850
      except (TypeError, ValueError), err:
9851
        raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err)
9852

    
9853
      if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt):
9854
        raise errors.OpPrereqError("HMAC for X509 key name is wrong",
9855
                                   errors.ECODE_INVAL)
9856

    
9857
      # Load and verify CA
9858
      try:
9859
        (cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds)
9860
      except OpenSSL.crypto.Error, err:
9861
        raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" %
9862
                                   (err, ), errors.ECODE_INVAL)
9863

    
9864
      (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
9865
      if errcode is not None:
9866
        raise errors.OpPrereqError("Invalid destination X509 CA (%s)" %
9867
                                   (msg, ), errors.ECODE_INVAL)
9868

    
9869
      self.dest_x509_ca = cert
9870

    
9871
      # Verify target information
9872
      disk_info = []
9873
      for idx, disk_data in enumerate(self.op.target_node):
9874
        try:
9875
          (host, port, magic) = \
9876
            masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data)
9877
        except errors.GenericError, err:
9878
          raise errors.OpPrereqError("Target info for disk %s: %s" %
9879
                                     (idx, err), errors.ECODE_INVAL)
9880

    
9881
        disk_info.append((host, port, magic))
9882

    
9883
      assert len(disk_info) == len(self.op.target_node)
9884
      self.dest_disk_info = disk_info
9885

    
9886
    else:
9887
      raise errors.ProgrammerError("Unhandled export mode %r" %
9888
                                   self.op.mode)
9889

    
9890
    # instance disk type verification
9891
    # TODO: Implement export support for file-based disks
9892
    for disk in self.instance.disks:
9893
      if disk.dev_type == constants.LD_FILE:
9894
        raise errors.OpPrereqError("Export not supported for instances with"
9895
                                   " file-based disks", errors.ECODE_INVAL)
9896

    
9897
  def _CleanupExports(self, feedback_fn):
9898
    """Removes exports of current instance from all other nodes.
9899

9900
    If an instance in a cluster with nodes A..D was exported to node C, its
9901
    exports will be removed from the nodes A, B and D.
9902

9903
    """
9904
    assert self.op.mode != constants.EXPORT_MODE_REMOTE
9905

    
9906
    nodelist = self.cfg.GetNodeList()
9907
    nodelist.remove(self.dst_node.name)
9908

    
9909
    # on one-node clusters nodelist will be empty after the removal
9910
    # if we proceed the backup would be removed because OpBackupQuery
9911
    # substitutes an empty list with the full cluster node list.
9912
    iname = self.instance.name
9913
    if nodelist:
9914
      feedback_fn("Removing old exports for instance %s" % iname)
9915
      exportlist = self.rpc.call_export_list(nodelist)
9916
      for node in exportlist:
9917
        if exportlist[node].fail_msg:
9918
          continue
9919
        if iname in exportlist[node].payload:
9920
          msg = self.rpc.call_export_remove(node, iname).fail_msg
9921
          if msg:
9922
            self.LogWarning("Could not remove older export for instance %s"
9923
                            " on node %s: %s", iname, node, msg)
9924

    
9925
  def Exec(self, feedback_fn):
9926
    """Export an instance to an image in the cluster.
9927

9928
    """
9929
    assert self.op.mode in constants.EXPORT_MODES
9930

    
9931
    instance = self.instance
9932
    src_node = instance.primary_node
9933

    
9934
    if self.op.shutdown:
9935
      # shutdown the instance, but not the disks
9936
      feedback_fn("Shutting down instance %s" % instance.name)
9937
      result = self.rpc.call_instance_shutdown(src_node, instance,
9938
                                               self.op.shutdown_timeout)
9939
      # TODO: Maybe ignore failures if ignore_remove_failures is set
9940
      result.Raise("Could not shutdown instance %s on"
9941
                   " node %s" % (instance.name, src_node))
9942

    
9943
    # set the disks ID correctly since call_instance_start needs the
9944
    # correct drbd minor to create the symlinks
9945
    for disk in instance.disks:
9946
      self.cfg.SetDiskID(disk, src_node)
9947

    
9948
    activate_disks = (not instance.admin_up)
9949

    
9950
    if activate_disks:
9951
      # Activate the instance disks if we'exporting a stopped instance
9952
      feedback_fn("Activating disks for %s" % instance.name)
9953
      _StartInstanceDisks(self, instance, None)
9954

    
9955
    try:
9956
      helper = masterd.instance.ExportInstanceHelper(self, feedback_fn,
9957
                                                     instance)
9958

    
9959
      helper.CreateSnapshots()
9960
      try:
9961
        if (self.op.shutdown and instance.admin_up and
9962
            not self.op.remove_instance):
9963
          assert not activate_disks
9964
          feedback_fn("Starting instance %s" % instance.name)
9965
          result = self.rpc.call_instance_start(src_node, instance, None, None)
9966
          msg = result.fail_msg
9967
          if msg:
9968
            feedback_fn("Failed to start instance: %s" % msg)
9969
            _ShutdownInstanceDisks(self, instance)
9970
            raise errors.OpExecError("Could not start instance: %s" % msg)
9971

    
9972
        if self.op.mode == constants.EXPORT_MODE_LOCAL:
9973
          (fin_resu, dresults) = helper.LocalExport(self.dst_node)
9974
        elif self.op.mode == constants.EXPORT_MODE_REMOTE:
9975
          connect_timeout = constants.RIE_CONNECT_TIMEOUT
9976
          timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
9977

    
9978
          (key_name, _, _) = self.x509_key_name
9979

    
9980
          dest_ca_pem = \
9981
            OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM,
9982
                                            self.dest_x509_ca)
9983

    
9984
          (fin_resu, dresults) = helper.RemoteExport(self.dest_disk_info,
9985
                                                     key_name, dest_ca_pem,
9986
                                                     timeouts)
9987
      finally:
9988
        helper.Cleanup()
9989

    
9990
      # Check for backwards compatibility
9991
      assert len(dresults) == len(instance.disks)
9992
      assert compat.all(isinstance(i, bool) for i in dresults), \
9993
             "Not all results are boolean: %r" % dresults
9994

    
9995
    finally:
9996
      if activate_disks:
9997
        feedback_fn("Deactivating disks for %s" % instance.name)
9998
        _ShutdownInstanceDisks(self, instance)
9999

    
10000
    if not (compat.all(dresults) and fin_resu):
10001
      failures = []
10002
      if not fin_resu:
10003
        failures.append("export finalization")
10004
      if not compat.all(dresults):
10005
        fdsk = utils.CommaJoin(idx for (idx, dsk) in enumerate(dresults)
10006
                               if not dsk)
10007
        failures.append("disk export: disk(s) %s" % fdsk)
10008

    
10009
      raise errors.OpExecError("Export failed, errors in %s" %
10010
                               utils.CommaJoin(failures))
10011

    
10012
    # At this point, the export was successful, we can cleanup/finish
10013

    
10014
    # Remove instance if requested
10015
    if self.op.remove_instance:
10016
      feedback_fn("Removing instance %s" % instance.name)
10017
      _RemoveInstance(self, feedback_fn, instance,
10018
                      self.op.ignore_remove_failures)
10019

    
10020
    if self.op.mode == constants.EXPORT_MODE_LOCAL:
10021
      self._CleanupExports(feedback_fn)
10022

    
10023
    return fin_resu, dresults
10024

    
10025

    
10026
class LUBackupRemove(NoHooksLU):
10027
  """Remove exports related to the named instance.
10028

10029
  """
10030
  REQ_BGL = False
10031

    
10032
  def ExpandNames(self):
10033
    self.needed_locks = {}
10034
    # We need all nodes to be locked in order for RemoveExport to work, but we
10035
    # don't need to lock the instance itself, as nothing will happen to it (and
10036
    # we can remove exports also for a removed instance)
10037
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
10038

    
10039
  def Exec(self, feedback_fn):
10040
    """Remove any export.
10041

10042
    """
10043
    instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
10044
    # If the instance was not found we'll try with the name that was passed in.
10045
    # This will only work if it was an FQDN, though.
10046
    fqdn_warn = False
10047
    if not instance_name:
10048
      fqdn_warn = True
10049
      instance_name = self.op.instance_name
10050

    
10051
    locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
10052
    exportlist = self.rpc.call_export_list(locked_nodes)
10053
    found = False
10054
    for node in exportlist:
10055
      msg = exportlist[node].fail_msg
10056
      if msg:
10057
        self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
10058
        continue
10059
      if instance_name in exportlist[node].payload:
10060
        found = True
10061
        result = self.rpc.call_export_remove(node, instance_name)
10062
        msg = result.fail_msg
10063
        if msg:
10064
          logging.error("Could not remove export for instance %s"
10065
                        " on node %s: %s", instance_name, node, msg)
10066

    
10067
    if fqdn_warn and not found:
10068
      feedback_fn("Export not found. If trying to remove an export belonging"
10069
                  " to a deleted instance please use its Fully Qualified"
10070
                  " Domain Name.")
10071

    
10072

    
10073
class LUGroupAdd(LogicalUnit):
10074
  """Logical unit for creating node groups.
10075

10076
  """
10077
  HPATH = "group-add"
10078
  HTYPE = constants.HTYPE_GROUP
10079
  REQ_BGL = False
10080

    
10081
  def ExpandNames(self):
10082
    # We need the new group's UUID here so that we can create and acquire the
10083
    # corresponding lock. Later, in Exec(), we'll indicate to cfg.AddNodeGroup
10084
    # that it should not check whether the UUID exists in the configuration.
10085
    self.group_uuid = self.cfg.GenerateUniqueID(self.proc.GetECId())
10086
    self.needed_locks = {}
10087
    self.add_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
10088

    
10089
  def CheckPrereq(self):
10090
    """Check prerequisites.
10091

10092
    This checks that the given group name is not an existing node group
10093
    already.
10094

10095
    """
10096
    try:
10097
      existing_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
10098
    except errors.OpPrereqError:
10099
      pass
10100
    else:
10101
      raise errors.OpPrereqError("Desired group name '%s' already exists as a"
10102
                                 " node group (UUID: %s)" %
10103
                                 (self.op.group_name, existing_uuid),
10104
                                 errors.ECODE_EXISTS)
10105

    
10106
    if self.op.ndparams:
10107
      utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
10108

    
10109
  def BuildHooksEnv(self):
10110
    """Build hooks env.
10111

10112
    """
10113
    env = {
10114
      "GROUP_NAME": self.op.group_name,
10115
      }
10116
    mn = self.cfg.GetMasterNode()
10117
    return env, [mn], [mn]
10118

    
10119
  def Exec(self, feedback_fn):
10120
    """Add the node group to the cluster.
10121

10122
    """
10123
    group_obj = objects.NodeGroup(name=self.op.group_name, members=[],
10124
                                  uuid=self.group_uuid,
10125
                                  alloc_policy=self.op.alloc_policy,
10126
                                  ndparams=self.op.ndparams)
10127

    
10128
    self.cfg.AddNodeGroup(group_obj, self.proc.GetECId(), check_uuid=False)
10129
    del self.remove_locks[locking.LEVEL_NODEGROUP]
10130

    
10131

    
10132
class LUGroupAssignNodes(NoHooksLU):
10133
  """Logical unit for assigning nodes to groups.
10134

10135
  """
10136
  REQ_BGL = False
10137

    
10138
  def ExpandNames(self):
10139
    # These raise errors.OpPrereqError on their own:
10140
    self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
10141
    self.op.nodes = _GetWantedNodes(self, self.op.nodes)
10142

    
10143
    # We want to lock all the affected nodes and groups. We have readily
10144
    # available the list of nodes, and the *destination* group. To gather the
10145
    # list of "source" groups, we need to fetch node information.
10146
    self.node_data = self.cfg.GetAllNodesInfo()
10147
    affected_groups = set(self.node_data[node].group for node in self.op.nodes)
10148
    affected_groups.add(self.group_uuid)
10149

    
10150
    self.needed_locks = {
10151
      locking.LEVEL_NODEGROUP: list(affected_groups),
10152
      locking.LEVEL_NODE: self.op.nodes,
10153
      }
10154

    
10155
  def CheckPrereq(self):
10156
    """Check prerequisites.
10157

10158
    """
10159
    self.group = self.cfg.GetNodeGroup(self.group_uuid)
10160
    instance_data = self.cfg.GetAllInstancesInfo()
10161

    
10162
    if self.group is None:
10163
      raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
10164
                               (self.op.group_name, self.group_uuid))
10165

    
10166
    (new_splits, previous_splits) = \
10167
      self.CheckAssignmentForSplitInstances([(node, self.group_uuid)
10168
                                             for node in self.op.nodes],
10169
                                            self.node_data, instance_data)
10170

    
10171
    if new_splits:
10172
      fmt_new_splits = utils.CommaJoin(utils.NiceSort(new_splits))
10173

    
10174
      if not self.op.force:
10175
        raise errors.OpExecError("The following instances get split by this"
10176
                                 " change and --force was not given: %s" %
10177
                                 fmt_new_splits)
10178
      else:
10179
        self.LogWarning("This operation will split the following instances: %s",
10180
                        fmt_new_splits)
10181

    
10182
        if previous_splits:
10183
          self.LogWarning("In addition, these already-split instances continue"
10184
                          " to be spit across groups: %s",
10185
                          utils.CommaJoin(utils.NiceSort(previous_splits)))
10186

    
10187
  def Exec(self, feedback_fn):
10188
    """Assign nodes to a new group.
10189

10190
    """
10191
    for node in self.op.nodes:
10192
      self.node_data[node].group = self.group_uuid
10193

    
10194
    self.cfg.Update(self.group, feedback_fn) # Saves all modified nodes.
10195

    
10196
  @staticmethod
10197
  def CheckAssignmentForSplitInstances(changes, node_data, instance_data):
10198
    """Check for split instances after a node assignment.
10199

10200
    This method considers a series of node assignments as an atomic operation,
10201
    and returns information about split instances after applying the set of
10202
    changes.
10203

10204
    In particular, it returns information about newly split instances, and
10205
    instances that were already split, and remain so after the change.
10206

10207
    Only instances whose disk template is listed in constants.DTS_NET_MIRROR are
10208
    considered.
10209

10210
    @type changes: list of (node_name, new_group_uuid) pairs.
10211
    @param changes: list of node assignments to consider.
10212
    @param node_data: a dict with data for all nodes
10213
    @param instance_data: a dict with all instances to consider
10214
    @rtype: a two-tuple
10215
    @return: a list of instances that were previously okay and result split as a
10216
      consequence of this change, and a list of instances that were previously
10217
      split and this change does not fix.
10218

10219
    """
10220
    changed_nodes = dict((node, group) for node, group in changes
10221
                         if node_data[node].group != group)
10222

    
10223
    all_split_instances = set()
10224
    previously_split_instances = set()
10225

    
10226
    def InstanceNodes(instance):
10227
      return [instance.primary_node] + list(instance.secondary_nodes)
10228

    
10229
    for inst in instance_data.values():
10230
      if inst.disk_template not in constants.DTS_NET_MIRROR:
10231
        continue
10232

    
10233
      instance_nodes = InstanceNodes(inst)
10234

    
10235
      if len(set(node_data[node].group for node in instance_nodes)) > 1:
10236
        previously_split_instances.add(inst.name)
10237

    
10238
      if len(set(changed_nodes.get(node, node_data[node].group)
10239
                 for node in instance_nodes)) > 1:
10240
        all_split_instances.add(inst.name)
10241

    
10242
    return (list(all_split_instances - previously_split_instances),
10243
            list(previously_split_instances & all_split_instances))
10244

    
10245

    
10246
class _GroupQuery(_QueryBase):
10247

    
10248
  FIELDS = query.GROUP_FIELDS
10249

    
10250
  def ExpandNames(self, lu):
10251
    lu.needed_locks = {}
10252

    
10253
    self._all_groups = lu.cfg.GetAllNodeGroupsInfo()
10254
    name_to_uuid = dict((g.name, g.uuid) for g in self._all_groups.values())
10255

    
10256
    if not self.names:
10257
      self.wanted = [name_to_uuid[name]
10258
                     for name in utils.NiceSort(name_to_uuid.keys())]
10259
    else:
10260
      # Accept names to be either names or UUIDs.
10261
      missing = []
10262
      self.wanted = []
10263
      all_uuid = frozenset(self._all_groups.keys())
10264

    
10265
      for name in self.names:
10266
        if name in all_uuid:
10267
          self.wanted.append(name)
10268
        elif name in name_to_uuid:
10269
          self.wanted.append(name_to_uuid[name])
10270
        else:
10271
          missing.append(name)
10272

    
10273
      if missing:
10274
        raise errors.OpPrereqError("Some groups do not exist: %s" %
10275
                                   utils.CommaJoin(missing),
10276
                                   errors.ECODE_NOENT)
10277

    
10278
  def DeclareLocks(self, lu, level):
10279
    pass
10280

    
10281
  def _GetQueryData(self, lu):
10282
    """Computes the list of node groups and their attributes.
10283

10284
    """
10285
    do_nodes = query.GQ_NODE in self.requested_data
10286
    do_instances = query.GQ_INST in self.requested_data
10287

    
10288
    group_to_nodes = None
10289
    group_to_instances = None
10290

    
10291
    # For GQ_NODE, we need to map group->[nodes], and group->[instances] for
10292
    # GQ_INST. The former is attainable with just GetAllNodesInfo(), but for the
10293
    # latter GetAllInstancesInfo() is not enough, for we have to go through
10294
    # instance->node. Hence, we will need to process nodes even if we only need
10295
    # instance information.
10296
    if do_nodes or do_instances:
10297
      all_nodes = lu.cfg.GetAllNodesInfo()
10298
      group_to_nodes = dict((uuid, []) for uuid in self.wanted)
10299
      node_to_group = {}
10300

    
10301
      for node in all_nodes.values():
10302
        if node.group in group_to_nodes:
10303
          group_to_nodes[node.group].append(node.name)
10304
          node_to_group[node.name] = node.group
10305

    
10306
      if do_instances:
10307
        all_instances = lu.cfg.GetAllInstancesInfo()
10308
        group_to_instances = dict((uuid, []) for uuid in self.wanted)
10309

    
10310
        for instance in all_instances.values():
10311
          node = instance.primary_node
10312
          if node in node_to_group:
10313
            group_to_instances[node_to_group[node]].append(instance.name)
10314

    
10315
        if not do_nodes:
10316
          # Do not pass on node information if it was not requested.
10317
          group_to_nodes = None
10318

    
10319
    return query.GroupQueryData([self._all_groups[uuid]
10320
                                 for uuid in self.wanted],
10321
                                group_to_nodes, group_to_instances)
10322

    
10323

    
10324
class LUGroupQuery(NoHooksLU):
10325
  """Logical unit for querying node groups.
10326

10327
  """
10328
  REQ_BGL = False
10329

    
10330
  def CheckArguments(self):
10331
    self.gq = _GroupQuery(self.op.names, self.op.output_fields, False)
10332

    
10333
  def ExpandNames(self):
10334
    self.gq.ExpandNames(self)
10335

    
10336
  def Exec(self, feedback_fn):
10337
    return self.gq.OldStyleQuery(self)
10338

    
10339

    
10340
class LUGroupSetParams(LogicalUnit):
10341
  """Modifies the parameters of a node group.
10342

10343
  """
10344
  HPATH = "group-modify"
10345
  HTYPE = constants.HTYPE_GROUP
10346
  REQ_BGL = False
10347

    
10348
  def CheckArguments(self):
10349
    all_changes = [
10350
      self.op.ndparams,
10351
      self.op.alloc_policy,
10352
      ]
10353

    
10354
    if all_changes.count(None) == len(all_changes):
10355
      raise errors.OpPrereqError("Please pass at least one modification",
10356
                                 errors.ECODE_INVAL)
10357

    
10358
  def ExpandNames(self):
10359
    # This raises errors.OpPrereqError on its own:
10360
    self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
10361

    
10362
    self.needed_locks = {
10363
      locking.LEVEL_NODEGROUP: [self.group_uuid],
10364
      }
10365

    
10366
  def CheckPrereq(self):
10367
    """Check prerequisites.
10368

10369
    """
10370
    self.group = self.cfg.GetNodeGroup(self.group_uuid)
10371

    
10372
    if self.group is None:
10373
      raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
10374
                               (self.op.group_name, self.group_uuid))
10375

    
10376
    if self.op.ndparams:
10377
      new_ndparams = _GetUpdatedParams(self.group.ndparams, self.op.ndparams)
10378
      utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
10379
      self.new_ndparams = new_ndparams
10380

    
10381
  def BuildHooksEnv(self):
10382
    """Build hooks env.
10383

10384
    """
10385
    env = {
10386
      "GROUP_NAME": self.op.group_name,
10387
      "NEW_ALLOC_POLICY": self.op.alloc_policy,
10388
      }
10389
    mn = self.cfg.GetMasterNode()
10390
    return env, [mn], [mn]
10391

    
10392
  def Exec(self, feedback_fn):
10393
    """Modifies the node group.
10394

10395
    """
10396
    result = []
10397

    
10398
    if self.op.ndparams:
10399
      self.group.ndparams = self.new_ndparams
10400
      result.append(("ndparams", str(self.group.ndparams)))
10401

    
10402
    if self.op.alloc_policy:
10403
      self.group.alloc_policy = self.op.alloc_policy
10404

    
10405
    self.cfg.Update(self.group, feedback_fn)
10406
    return result
10407

    
10408

    
10409

    
10410
class LUGroupRemove(LogicalUnit):
10411
  HPATH = "group-remove"
10412
  HTYPE = constants.HTYPE_GROUP
10413
  REQ_BGL = False
10414

    
10415
  def ExpandNames(self):
10416
    # This will raises errors.OpPrereqError on its own:
10417
    self.group_uuid = self.cfg.LookupNodeGroup(self.op.group_name)
10418
    self.needed_locks = {
10419
      locking.LEVEL_NODEGROUP: [self.group_uuid],
10420
      }
10421

    
10422
  def CheckPrereq(self):
10423
    """Check prerequisites.
10424

10425
    This checks that the given group name exists as a node group, that is
10426
    empty (i.e., contains no nodes), and that is not the last group of the
10427
    cluster.
10428

10429
    """
10430
    # Verify that the group is empty.
10431
    group_nodes = [node.name
10432
                   for node in self.cfg.GetAllNodesInfo().values()
10433
                   if node.group == self.group_uuid]
10434

    
10435
    if group_nodes:
10436
      raise errors.OpPrereqError("Group '%s' not empty, has the following"
10437
                                 " nodes: %s" %
10438
                                 (self.op.group_name,
10439
                                  utils.CommaJoin(utils.NiceSort(group_nodes))),
10440
                                 errors.ECODE_STATE)
10441

    
10442
    # Verify the cluster would not be left group-less.
10443
    if len(self.cfg.GetNodeGroupList()) == 1:
10444
      raise errors.OpPrereqError("Group '%s' is the only group,"
10445
                                 " cannot be removed" %
10446
                                 self.op.group_name,
10447
                                 errors.ECODE_STATE)
10448

    
10449
  def BuildHooksEnv(self):
10450
    """Build hooks env.
10451

10452
    """
10453
    env = {
10454
      "GROUP_NAME": self.op.group_name,
10455
      }
10456
    mn = self.cfg.GetMasterNode()
10457
    return env, [mn], [mn]
10458

    
10459
  def Exec(self, feedback_fn):
10460
    """Remove the node group.
10461

10462
    """
10463
    try:
10464
      self.cfg.RemoveNodeGroup(self.group_uuid)
10465
    except errors.ConfigurationError:
10466
      raise errors.OpExecError("Group '%s' with UUID %s disappeared" %
10467
                               (self.op.group_name, self.group_uuid))
10468

    
10469
    self.remove_locks[locking.LEVEL_NODEGROUP] = self.group_uuid
10470

    
10471

    
10472
class LUGroupRename(LogicalUnit):
10473
  HPATH = "group-rename"
10474
  HTYPE = constants.HTYPE_GROUP
10475
  REQ_BGL = False
10476

    
10477
  def ExpandNames(self):
10478
    # This raises errors.OpPrereqError on its own:
10479
    self.group_uuid = self.cfg.LookupNodeGroup(self.op.old_name)
10480

    
10481
    self.needed_locks = {
10482
      locking.LEVEL_NODEGROUP: [self.group_uuid],
10483
      }
10484

    
10485
  def CheckPrereq(self):
10486
    """Check prerequisites.
10487

10488
    This checks that the given old_name exists as a node group, and that
10489
    new_name doesn't.
10490

10491
    """
10492
    try:
10493
      new_name_uuid = self.cfg.LookupNodeGroup(self.op.new_name)
10494
    except errors.OpPrereqError:
10495
      pass
10496
    else:
10497
      raise errors.OpPrereqError("Desired new name '%s' clashes with existing"
10498
                                 " node group (UUID: %s)" %
10499
                                 (self.op.new_name, new_name_uuid),
10500
                                 errors.ECODE_EXISTS)
10501

    
10502
  def BuildHooksEnv(self):
10503
    """Build hooks env.
10504

10505
    """
10506
    env = {
10507
      "OLD_NAME": self.op.old_name,
10508
      "NEW_NAME": self.op.new_name,
10509
      }
10510

    
10511
    mn = self.cfg.GetMasterNode()
10512
    all_nodes = self.cfg.GetAllNodesInfo()
10513
    run_nodes = [mn]
10514
    all_nodes.pop(mn, None)
10515

    
10516
    for node in all_nodes.values():
10517
      if node.group == self.group_uuid:
10518
        run_nodes.append(node.name)
10519

    
10520
    return env, run_nodes, run_nodes
10521

    
10522
  def Exec(self, feedback_fn):
10523
    """Rename the node group.
10524

10525
    """
10526
    group = self.cfg.GetNodeGroup(self.group_uuid)
10527

    
10528
    if group is None:
10529
      raise errors.OpExecError("Could not retrieve group '%s' (UUID: %s)" %
10530
                               (self.op.old_name, self.group_uuid))
10531

    
10532
    group.name = self.op.new_name
10533
    self.cfg.Update(group, feedback_fn)
10534

    
10535
    return self.op.new_name
10536

    
10537

    
10538
class TagsLU(NoHooksLU): # pylint: disable-msg=W0223
10539
  """Generic tags LU.
10540

10541
  This is an abstract class which is the parent of all the other tags LUs.
10542

10543
  """
10544

    
10545
  def ExpandNames(self):
10546
    self.needed_locks = {}
10547
    if self.op.kind == constants.TAG_NODE:
10548
      self.op.name = _ExpandNodeName(self.cfg, self.op.name)
10549
      self.needed_locks[locking.LEVEL_NODE] = self.op.name
10550
    elif self.op.kind == constants.TAG_INSTANCE:
10551
      self.op.name = _ExpandInstanceName(self.cfg, self.op.name)
10552
      self.needed_locks[locking.LEVEL_INSTANCE] = self.op.name
10553

    
10554
    # FIXME: Acquire BGL for cluster tag operations (as of this writing it's
10555
    # not possible to acquire the BGL based on opcode parameters)
10556

    
10557
  def CheckPrereq(self):
10558
    """Check prerequisites.
10559

10560
    """
10561
    if self.op.kind == constants.TAG_CLUSTER:
10562
      self.target = self.cfg.GetClusterInfo()
10563
    elif self.op.kind == constants.TAG_NODE:
10564
      self.target = self.cfg.GetNodeInfo(self.op.name)
10565
    elif self.op.kind == constants.TAG_INSTANCE:
10566
      self.target = self.cfg.GetInstanceInfo(self.op.name)
10567
    else:
10568
      raise errors.OpPrereqError("Wrong tag type requested (%s)" %
10569
                                 str(self.op.kind), errors.ECODE_INVAL)
10570

    
10571

    
10572
class LUTagsGet(TagsLU):
10573
  """Returns the tags of a given object.
10574

10575
  """
10576
  REQ_BGL = False
10577

    
10578
  def ExpandNames(self):
10579
    TagsLU.ExpandNames(self)
10580

    
10581
    # Share locks as this is only a read operation
10582
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
10583

    
10584
  def Exec(self, feedback_fn):
10585
    """Returns the tag list.
10586

10587
    """
10588
    return list(self.target.GetTags())
10589

    
10590

    
10591
class LUTagsSearch(NoHooksLU):
10592
  """Searches the tags for a given pattern.
10593

10594
  """
10595
  REQ_BGL = False
10596

    
10597
  def ExpandNames(self):
10598
    self.needed_locks = {}
10599

    
10600
  def CheckPrereq(self):
10601
    """Check prerequisites.
10602

10603
    This checks the pattern passed for validity by compiling it.
10604

10605
    """
10606
    try:
10607
      self.re = re.compile(self.op.pattern)
10608
    except re.error, err:
10609
      raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
10610
                                 (self.op.pattern, err), errors.ECODE_INVAL)
10611

    
10612
  def Exec(self, feedback_fn):
10613
    """Returns the tag list.
10614

10615
    """
10616
    cfg = self.cfg
10617
    tgts = [("/cluster", cfg.GetClusterInfo())]
10618
    ilist = cfg.GetAllInstancesInfo().values()
10619
    tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
10620
    nlist = cfg.GetAllNodesInfo().values()
10621
    tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
10622
    results = []
10623
    for path, target in tgts:
10624
      for tag in target.GetTags():
10625
        if self.re.search(tag):
10626
          results.append((path, tag))
10627
    return results
10628

    
10629

    
10630
class LUTagsSet(TagsLU):
10631
  """Sets a tag on a given object.
10632

10633
  """
10634
  REQ_BGL = False
10635

    
10636
  def CheckPrereq(self):
10637
    """Check prerequisites.
10638

10639
    This checks the type and length of the tag name and value.
10640

10641
    """
10642
    TagsLU.CheckPrereq(self)
10643
    for tag in self.op.tags:
10644
      objects.TaggableObject.ValidateTag(tag)
10645

    
10646
  def Exec(self, feedback_fn):
10647
    """Sets the tag.
10648

10649
    """
10650
    try:
10651
      for tag in self.op.tags:
10652
        self.target.AddTag(tag)
10653
    except errors.TagError, err:
10654
      raise errors.OpExecError("Error while setting tag: %s" % str(err))
10655
    self.cfg.Update(self.target, feedback_fn)
10656

    
10657

    
10658
class LUTagsDel(TagsLU):
10659
  """Delete a list of tags from a given object.
10660

10661
  """
10662
  REQ_BGL = False
10663

    
10664
  def CheckPrereq(self):
10665
    """Check prerequisites.
10666

10667
    This checks that we have the given tag.
10668

10669
    """
10670
    TagsLU.CheckPrereq(self)
10671
    for tag in self.op.tags:
10672
      objects.TaggableObject.ValidateTag(tag)
10673
    del_tags = frozenset(self.op.tags)
10674
    cur_tags = self.target.GetTags()
10675

    
10676
    diff_tags = del_tags - cur_tags
10677
    if diff_tags:
10678
      diff_names = ("'%s'" % i for i in sorted(diff_tags))
10679
      raise errors.OpPrereqError("Tag(s) %s not found" %
10680
                                 (utils.CommaJoin(diff_names), ),
10681
                                 errors.ECODE_NOENT)
10682

    
10683
  def Exec(self, feedback_fn):
10684
    """Remove the tag from the object.
10685

10686
    """
10687
    for tag in self.op.tags:
10688
      self.target.RemoveTag(tag)
10689
    self.cfg.Update(self.target, feedback_fn)
10690

    
10691

    
10692
class LUTestDelay(NoHooksLU):
10693
  """Sleep for a specified amount of time.
10694

10695
  This LU sleeps on the master and/or nodes for a specified amount of
10696
  time.
10697

10698
  """
10699
  REQ_BGL = False
10700

    
10701
  def ExpandNames(self):
10702
    """Expand names and set required locks.
10703

10704
    This expands the node list, if any.
10705

10706
    """
10707
    self.needed_locks = {}
10708
    if self.op.on_nodes:
10709
      # _GetWantedNodes can be used here, but is not always appropriate to use
10710
      # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
10711
      # more information.
10712
      self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
10713
      self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
10714

    
10715
  def _TestDelay(self):
10716
    """Do the actual sleep.
10717

10718
    """
10719
    if self.op.on_master:
10720
      if not utils.TestDelay(self.op.duration):
10721
        raise errors.OpExecError("Error during master delay test")
10722
    if self.op.on_nodes:
10723
      result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
10724
      for node, node_result in result.items():
10725
        node_result.Raise("Failure during rpc call to node %s" % node)
10726

    
10727
  def Exec(self, feedback_fn):
10728
    """Execute the test delay opcode, with the wanted repetitions.
10729

10730
    """
10731
    if self.op.repeat == 0:
10732
      self._TestDelay()
10733
    else:
10734
      top_value = self.op.repeat - 1
10735
      for i in range(self.op.repeat):
10736
        self.LogInfo("Test delay iteration %d/%d" % (i, top_value))
10737
        self._TestDelay()
10738

    
10739

    
10740
class LUTestJqueue(NoHooksLU):
10741
  """Utility LU to test some aspects of the job queue.
10742

10743
  """
10744
  REQ_BGL = False
10745

    
10746
  # Must be lower than default timeout for WaitForJobChange to see whether it
10747
  # notices changed jobs
10748
  _CLIENT_CONNECT_TIMEOUT = 20.0
10749
  _CLIENT_CONFIRM_TIMEOUT = 60.0
10750

    
10751
  @classmethod
10752
  def _NotifyUsingSocket(cls, cb, errcls):
10753
    """Opens a Unix socket and waits for another program to connect.
10754

10755
    @type cb: callable
10756
    @param cb: Callback to send socket name to client
10757
    @type errcls: class
10758
    @param errcls: Exception class to use for errors
10759

10760
    """
10761
    # Using a temporary directory as there's no easy way to create temporary
10762
    # sockets without writing a custom loop around tempfile.mktemp and
10763
    # socket.bind
10764
    tmpdir = tempfile.mkdtemp()
10765
    try:
10766
      tmpsock = utils.PathJoin(tmpdir, "sock")
10767

    
10768
      logging.debug("Creating temporary socket at %s", tmpsock)
10769
      sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
10770
      try:
10771
        sock.bind(tmpsock)
10772
        sock.listen(1)
10773

    
10774
        # Send details to client
10775
        cb(tmpsock)
10776

    
10777
        # Wait for client to connect before continuing
10778
        sock.settimeout(cls._CLIENT_CONNECT_TIMEOUT)
10779
        try:
10780
          (conn, _) = sock.accept()
10781
        except socket.error, err:
10782
          raise errcls("Client didn't connect in time (%s)" % err)
10783
      finally:
10784
        sock.close()
10785
    finally:
10786
      # Remove as soon as client is connected
10787
      shutil.rmtree(tmpdir)
10788

    
10789
    # Wait for client to close
10790
    try:
10791
      try:
10792
        # pylint: disable-msg=E1101
10793
        # Instance of '_socketobject' has no ... member
10794
        conn.settimeout(cls._CLIENT_CONFIRM_TIMEOUT)
10795
        conn.recv(1)
10796
      except socket.error, err:
10797
        raise errcls("Client failed to confirm notification (%s)" % err)
10798
    finally:
10799
      conn.close()
10800

    
10801
  def _SendNotification(self, test, arg, sockname):
10802
    """Sends a notification to the client.
10803

10804
    @type test: string
10805
    @param test: Test name
10806
    @param arg: Test argument (depends on test)
10807
    @type sockname: string
10808
    @param sockname: Socket path
10809

10810
    """
10811
    self.Log(constants.ELOG_JQUEUE_TEST, (sockname, test, arg))
10812

    
10813
  def _Notify(self, prereq, test, arg):
10814
    """Notifies the client of a test.
10815

10816
    @type prereq: bool
10817
    @param prereq: Whether this is a prereq-phase test
10818
    @type test: string
10819
    @param test: Test name
10820
    @param arg: Test argument (depends on test)
10821

10822
    """
10823
    if prereq:
10824
      errcls = errors.OpPrereqError
10825
    else:
10826
      errcls = errors.OpExecError
10827

    
10828
    return self._NotifyUsingSocket(compat.partial(self._SendNotification,
10829
                                                  test, arg),
10830
                                   errcls)
10831

    
10832
  def CheckArguments(self):
10833
    self.checkargs_calls = getattr(self, "checkargs_calls", 0) + 1
10834
    self.expandnames_calls = 0
10835

    
10836
  def ExpandNames(self):
10837
    checkargs_calls = getattr(self, "checkargs_calls", 0)
10838
    if checkargs_calls < 1:
10839
      raise errors.ProgrammerError("CheckArguments was not called")
10840

    
10841
    self.expandnames_calls += 1
10842

    
10843
    if self.op.notify_waitlock:
10844
      self._Notify(True, constants.JQT_EXPANDNAMES, None)
10845

    
10846
    self.LogInfo("Expanding names")
10847

    
10848
    # Get lock on master node (just to get a lock, not for a particular reason)
10849
    self.needed_locks = {
10850
      locking.LEVEL_NODE: self.cfg.GetMasterNode(),
10851
      }
10852

    
10853
  def Exec(self, feedback_fn):
10854
    if self.expandnames_calls < 1:
10855
      raise errors.ProgrammerError("ExpandNames was not called")
10856

    
10857
    if self.op.notify_exec:
10858
      self._Notify(False, constants.JQT_EXEC, None)
10859

    
10860
    self.LogInfo("Executing")
10861

    
10862
    if self.op.log_messages:
10863
      self._Notify(False, constants.JQT_STARTMSG, len(self.op.log_messages))
10864
      for idx, msg in enumerate(self.op.log_messages):
10865
        self.LogInfo("Sending log message %s", idx + 1)
10866
        feedback_fn(constants.JQT_MSGPREFIX + msg)
10867
        # Report how many test messages have been sent
10868
        self._Notify(False, constants.JQT_LOGMSG, idx + 1)
10869

    
10870
    if self.op.fail:
10871
      raise errors.OpExecError("Opcode failure was requested")
10872

    
10873
    return True
10874

    
10875

    
10876
class IAllocator(object):
10877
  """IAllocator framework.
10878

10879
  An IAllocator instance has three sets of attributes:
10880
    - cfg that is needed to query the cluster
10881
    - input data (all members of the _KEYS class attribute are required)
10882
    - four buffer attributes (in|out_data|text), that represent the
10883
      input (to the external script) in text and data structure format,
10884
      and the output from it, again in two formats
10885
    - the result variables from the script (success, info, nodes) for
10886
      easy usage
10887

10888
  """
10889
  # pylint: disable-msg=R0902
10890
  # lots of instance attributes
10891
  _ALLO_KEYS = [
10892
    "name", "mem_size", "disks", "disk_template",
10893
    "os", "tags", "nics", "vcpus", "hypervisor",
10894
    ]
10895
  _RELO_KEYS = [
10896
    "name", "relocate_from",
10897
    ]
10898
  _EVAC_KEYS = [
10899
    "evac_nodes",
10900
    ]
10901

    
10902
  def __init__(self, cfg, rpc, mode, **kwargs):
10903
    self.cfg = cfg
10904
    self.rpc = rpc
10905
    # init buffer variables
10906
    self.in_text = self.out_text = self.in_data = self.out_data = None
10907
    # init all input fields so that pylint is happy
10908
    self.mode = mode
10909
    self.mem_size = self.disks = self.disk_template = None
10910
    self.os = self.tags = self.nics = self.vcpus = None
10911
    self.hypervisor = None
10912
    self.relocate_from = None
10913
    self.name = None
10914
    self.evac_nodes = None
10915
    # computed fields
10916
    self.required_nodes = None
10917
    # init result fields
10918
    self.success = self.info = self.result = None
10919
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
10920
      keyset = self._ALLO_KEYS
10921
      fn = self._AddNewInstance
10922
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
10923
      keyset = self._RELO_KEYS
10924
      fn = self._AddRelocateInstance
10925
    elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
10926
      keyset = self._EVAC_KEYS
10927
      fn = self._AddEvacuateNodes
10928
    else:
10929
      raise errors.ProgrammerError("Unknown mode '%s' passed to the"
10930
                                   " IAllocator" % self.mode)
10931
    for key in kwargs:
10932
      if key not in keyset:
10933
        raise errors.ProgrammerError("Invalid input parameter '%s' to"
10934
                                     " IAllocator" % key)
10935
      setattr(self, key, kwargs[key])
10936

    
10937
    for key in keyset:
10938
      if key not in kwargs:
10939
        raise errors.ProgrammerError("Missing input parameter '%s' to"
10940
                                     " IAllocator" % key)
10941
    self._BuildInputData(fn)
10942

    
10943
  def _ComputeClusterData(self):
10944
    """Compute the generic allocator input data.
10945

10946
    This is the data that is independent of the actual operation.
10947

10948
    """
10949
    cfg = self.cfg
10950
    cluster_info = cfg.GetClusterInfo()
10951
    # cluster data
10952
    data = {
10953
      "version": constants.IALLOCATOR_VERSION,
10954
      "cluster_name": cfg.GetClusterName(),
10955
      "cluster_tags": list(cluster_info.GetTags()),
10956
      "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
10957
      # we don't have job IDs
10958
      }
10959
    ninfo = cfg.GetAllNodesInfo()
10960
    iinfo = cfg.GetAllInstancesInfo().values()
10961
    i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
10962

    
10963
    # node data
10964
    node_list = [n.name for n in ninfo.values() if n.vm_capable]
10965

    
10966
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
10967
      hypervisor_name = self.hypervisor
10968
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
10969
      hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
10970
    elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
10971
      hypervisor_name = cluster_info.enabled_hypervisors[0]
10972

    
10973
    node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(),
10974
                                        hypervisor_name)
10975
    node_iinfo = \
10976
      self.rpc.call_all_instances_info(node_list,
10977
                                       cluster_info.enabled_hypervisors)
10978

    
10979
    data["nodegroups"] = self._ComputeNodeGroupData(cfg)
10980

    
10981
    config_ndata = self._ComputeBasicNodeData(ninfo)
10982
    data["nodes"] = self._ComputeDynamicNodeData(ninfo, node_data, node_iinfo,
10983
                                                 i_list, config_ndata)
10984
    assert len(data["nodes"]) == len(ninfo), \
10985
        "Incomplete node data computed"
10986

    
10987
    data["instances"] = self._ComputeInstanceData(cluster_info, i_list)
10988

    
10989
    self.in_data = data
10990

    
10991
  @staticmethod
10992
  def _ComputeNodeGroupData(cfg):
10993
    """Compute node groups data.
10994

10995
    """
10996
    ng = {}
10997
    for guuid, gdata in cfg.GetAllNodeGroupsInfo().items():
10998
      ng[guuid] = {
10999
        "name": gdata.name,
11000
        "alloc_policy": gdata.alloc_policy,
11001
        }
11002
    return ng
11003

    
11004
  @staticmethod
11005
  def _ComputeBasicNodeData(node_cfg):
11006
    """Compute global node data.
11007

11008
    @rtype: dict
11009
    @returns: a dict of name: (node dict, node config)
11010

11011
    """
11012
    node_results = {}
11013
    for ninfo in node_cfg.values():
11014
      # fill in static (config-based) values
11015
      pnr = {
11016
        "tags": list(ninfo.GetTags()),
11017
        "primary_ip": ninfo.primary_ip,
11018
        "secondary_ip": ninfo.secondary_ip,
11019
        "offline": ninfo.offline,
11020
        "drained": ninfo.drained,
11021
        "master_candidate": ninfo.master_candidate,
11022
        "group": ninfo.group,
11023
        "master_capable": ninfo.master_capable,
11024
        "vm_capable": ninfo.vm_capable,
11025
        }
11026

    
11027
      node_results[ninfo.name] = pnr
11028

    
11029
    return node_results
11030

    
11031
  @staticmethod
11032
  def _ComputeDynamicNodeData(node_cfg, node_data, node_iinfo, i_list,
11033
                              node_results):
11034
    """Compute global node data.
11035

11036
    @param node_results: the basic node structures as filled from the config
11037

11038
    """
11039
    # make a copy of the current dict
11040
    node_results = dict(node_results)
11041
    for nname, nresult in node_data.items():
11042
      assert nname in node_results, "Missing basic data for node %s" % nname
11043
      ninfo = node_cfg[nname]
11044

    
11045
      if not (ninfo.offline or ninfo.drained):
11046
        nresult.Raise("Can't get data for node %s" % nname)
11047
        node_iinfo[nname].Raise("Can't get node instance info from node %s" %
11048
                                nname)
11049
        remote_info = nresult.payload
11050

    
11051
        for attr in ['memory_total', 'memory_free', 'memory_dom0',
11052
                     'vg_size', 'vg_free', 'cpu_total']:
11053
          if attr not in remote_info:
11054
            raise errors.OpExecError("Node '%s' didn't return attribute"
11055
                                     " '%s'" % (nname, attr))
11056
          if not isinstance(remote_info[attr], int):
11057
            raise errors.OpExecError("Node '%s' returned invalid value"
11058
                                     " for '%s': %s" %
11059
                                     (nname, attr, remote_info[attr]))
11060
        # compute memory used by primary instances
11061
        i_p_mem = i_p_up_mem = 0
11062
        for iinfo, beinfo in i_list:
11063
          if iinfo.primary_node == nname:
11064
            i_p_mem += beinfo[constants.BE_MEMORY]
11065
            if iinfo.name not in node_iinfo[nname].payload:
11066
              i_used_mem = 0
11067
            else:
11068
              i_used_mem = int(node_iinfo[nname].payload[iinfo.name]['memory'])
11069
            i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
11070
            remote_info['memory_free'] -= max(0, i_mem_diff)
11071

    
11072
            if iinfo.admin_up:
11073
              i_p_up_mem += beinfo[constants.BE_MEMORY]
11074

    
11075
        # compute memory used by instances
11076
        pnr_dyn = {
11077
          "total_memory": remote_info['memory_total'],
11078
          "reserved_memory": remote_info['memory_dom0'],
11079
          "free_memory": remote_info['memory_free'],
11080
          "total_disk": remote_info['vg_size'],
11081
          "free_disk": remote_info['vg_free'],
11082
          "total_cpus": remote_info['cpu_total'],
11083
          "i_pri_memory": i_p_mem,
11084
          "i_pri_up_memory": i_p_up_mem,
11085
          }
11086
        pnr_dyn.update(node_results[nname])
11087
        node_results[nname] = pnr_dyn
11088

    
11089
    return node_results
11090

    
11091
  @staticmethod
11092
  def _ComputeInstanceData(cluster_info, i_list):
11093
    """Compute global instance data.
11094

11095
    """
11096
    instance_data = {}
11097
    for iinfo, beinfo in i_list:
11098
      nic_data = []
11099
      for nic in iinfo.nics:
11100
        filled_params = cluster_info.SimpleFillNIC(nic.nicparams)
11101
        nic_dict = {"mac": nic.mac,
11102
                    "ip": nic.ip,
11103
                    "mode": filled_params[constants.NIC_MODE],
11104
                    "link": filled_params[constants.NIC_LINK],
11105
                   }
11106
        if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
11107
          nic_dict["bridge"] = filled_params[constants.NIC_LINK]
11108
        nic_data.append(nic_dict)
11109
      pir = {
11110
        "tags": list(iinfo.GetTags()),
11111
        "admin_up": iinfo.admin_up,
11112
        "vcpus": beinfo[constants.BE_VCPUS],
11113
        "memory": beinfo[constants.BE_MEMORY],
11114
        "os": iinfo.os,
11115
        "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
11116
        "nics": nic_data,
11117
        "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
11118
        "disk_template": iinfo.disk_template,
11119
        "hypervisor": iinfo.hypervisor,
11120
        }
11121
      pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
11122
                                                 pir["disks"])
11123
      instance_data[iinfo.name] = pir
11124

    
11125
    return instance_data
11126

    
11127
  def _AddNewInstance(self):
11128
    """Add new instance data to allocator structure.
11129

11130
    This in combination with _AllocatorGetClusterData will create the
11131
    correct structure needed as input for the allocator.
11132

11133
    The checks for the completeness of the opcode must have already been
11134
    done.
11135

11136
    """
11137
    disk_space = _ComputeDiskSize(self.disk_template, self.disks)
11138

    
11139
    if self.disk_template in constants.DTS_NET_MIRROR:
11140
      self.required_nodes = 2
11141
    else:
11142
      self.required_nodes = 1
11143
    request = {
11144
      "name": self.name,
11145
      "disk_template": self.disk_template,
11146
      "tags": self.tags,
11147
      "os": self.os,
11148
      "vcpus": self.vcpus,
11149
      "memory": self.mem_size,
11150
      "disks": self.disks,
11151
      "disk_space_total": disk_space,
11152
      "nics": self.nics,
11153
      "required_nodes": self.required_nodes,
11154
      }
11155
    return request
11156

    
11157
  def _AddRelocateInstance(self):
11158
    """Add relocate instance data to allocator structure.
11159

11160
    This in combination with _IAllocatorGetClusterData will create the
11161
    correct structure needed as input for the allocator.
11162

11163
    The checks for the completeness of the opcode must have already been
11164
    done.
11165

11166
    """
11167
    instance = self.cfg.GetInstanceInfo(self.name)
11168
    if instance is None:
11169
      raise errors.ProgrammerError("Unknown instance '%s' passed to"
11170
                                   " IAllocator" % self.name)
11171

    
11172
    if instance.disk_template not in constants.DTS_NET_MIRROR:
11173
      raise errors.OpPrereqError("Can't relocate non-mirrored instances",
11174
                                 errors.ECODE_INVAL)
11175

    
11176
    if len(instance.secondary_nodes) != 1:
11177
      raise errors.OpPrereqError("Instance has not exactly one secondary node",
11178
                                 errors.ECODE_STATE)
11179

    
11180
    self.required_nodes = 1
11181
    disk_sizes = [{'size': disk.size} for disk in instance.disks]
11182
    disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
11183

    
11184
    request = {
11185
      "name": self.name,
11186
      "disk_space_total": disk_space,
11187
      "required_nodes": self.required_nodes,
11188
      "relocate_from": self.relocate_from,
11189
      }
11190
    return request
11191

    
11192
  def _AddEvacuateNodes(self):
11193
    """Add evacuate nodes data to allocator structure.
11194

11195
    """
11196
    request = {
11197
      "evac_nodes": self.evac_nodes
11198
      }
11199
    return request
11200

    
11201
  def _BuildInputData(self, fn):
11202
    """Build input data structures.
11203

11204
    """
11205
    self._ComputeClusterData()
11206

    
11207
    request = fn()
11208
    request["type"] = self.mode
11209
    self.in_data["request"] = request
11210

    
11211
    self.in_text = serializer.Dump(self.in_data)
11212

    
11213
  def Run(self, name, validate=True, call_fn=None):
11214
    """Run an instance allocator and return the results.
11215

11216
    """
11217
    if call_fn is None:
11218
      call_fn = self.rpc.call_iallocator_runner
11219

    
11220
    result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
11221
    result.Raise("Failure while running the iallocator script")
11222

    
11223
    self.out_text = result.payload
11224
    if validate:
11225
      self._ValidateResult()
11226

    
11227
  def _ValidateResult(self):
11228
    """Process the allocator results.
11229

11230
    This will process and if successful save the result in
11231
    self.out_data and the other parameters.
11232

11233
    """
11234
    try:
11235
      rdict = serializer.Load(self.out_text)
11236
    except Exception, err:
11237
      raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
11238

    
11239
    if not isinstance(rdict, dict):
11240
      raise errors.OpExecError("Can't parse iallocator results: not a dict")
11241

    
11242
    # TODO: remove backwards compatiblity in later versions
11243
    if "nodes" in rdict and "result" not in rdict:
11244
      rdict["result"] = rdict["nodes"]
11245
      del rdict["nodes"]
11246

    
11247
    for key in "success", "info", "result":
11248
      if key not in rdict:
11249
        raise errors.OpExecError("Can't parse iallocator results:"
11250
                                 " missing key '%s'" % key)
11251
      setattr(self, key, rdict[key])
11252

    
11253
    if not isinstance(rdict["result"], list):
11254
      raise errors.OpExecError("Can't parse iallocator results: 'result' key"
11255
                               " is not a list")
11256
    self.out_data = rdict
11257

    
11258

    
11259
class LUTestAllocator(NoHooksLU):
11260
  """Run allocator tests.
11261

11262
  This LU runs the allocator tests
11263

11264
  """
11265
  def CheckPrereq(self):
11266
    """Check prerequisites.
11267

11268
    This checks the opcode parameters depending on the director and mode test.
11269

11270
    """
11271
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
11272
      for attr in ["mem_size", "disks", "disk_template",
11273
                   "os", "tags", "nics", "vcpus"]:
11274
        if not hasattr(self.op, attr):
11275
          raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
11276
                                     attr, errors.ECODE_INVAL)
11277
      iname = self.cfg.ExpandInstanceName(self.op.name)
11278
      if iname is not None:
11279
        raise errors.OpPrereqError("Instance '%s' already in the cluster" %
11280
                                   iname, errors.ECODE_EXISTS)
11281
      if not isinstance(self.op.nics, list):
11282
        raise errors.OpPrereqError("Invalid parameter 'nics'",
11283
                                   errors.ECODE_INVAL)
11284
      if not isinstance(self.op.disks, list):
11285
        raise errors.OpPrereqError("Invalid parameter 'disks'",
11286
                                   errors.ECODE_INVAL)
11287
      for row in self.op.disks:
11288
        if (not isinstance(row, dict) or
11289
            "size" not in row or
11290
            not isinstance(row["size"], int) or
11291
            "mode" not in row or
11292
            row["mode"] not in ['r', 'w']):
11293
          raise errors.OpPrereqError("Invalid contents of the 'disks'"
11294
                                     " parameter", errors.ECODE_INVAL)
11295
      if self.op.hypervisor is None:
11296
        self.op.hypervisor = self.cfg.GetHypervisorType()
11297
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
11298
      fname = _ExpandInstanceName(self.cfg, self.op.name)
11299
      self.op.name = fname
11300
      self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
11301
    elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
11302
      if not hasattr(self.op, "evac_nodes"):
11303
        raise errors.OpPrereqError("Missing attribute 'evac_nodes' on"
11304
                                   " opcode input", errors.ECODE_INVAL)
11305
    else:
11306
      raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
11307
                                 self.op.mode, errors.ECODE_INVAL)
11308

    
11309
    if self.op.direction == constants.IALLOCATOR_DIR_OUT:
11310
      if self.op.allocator is None:
11311
        raise errors.OpPrereqError("Missing allocator name",
11312
                                   errors.ECODE_INVAL)
11313
    elif self.op.direction != constants.IALLOCATOR_DIR_IN:
11314
      raise errors.OpPrereqError("Wrong allocator test '%s'" %
11315
                                 self.op.direction, errors.ECODE_INVAL)
11316

    
11317
  def Exec(self, feedback_fn):
11318
    """Run the allocator test.
11319

11320
    """
11321
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
11322
      ial = IAllocator(self.cfg, self.rpc,
11323
                       mode=self.op.mode,
11324
                       name=self.op.name,
11325
                       mem_size=self.op.mem_size,
11326
                       disks=self.op.disks,
11327
                       disk_template=self.op.disk_template,
11328
                       os=self.op.os,
11329
                       tags=self.op.tags,
11330
                       nics=self.op.nics,
11331
                       vcpus=self.op.vcpus,
11332
                       hypervisor=self.op.hypervisor,
11333
                       )
11334
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
11335
      ial = IAllocator(self.cfg, self.rpc,
11336
                       mode=self.op.mode,
11337
                       name=self.op.name,
11338
                       relocate_from=list(self.relocate_from),
11339
                       )
11340
    elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
11341
      ial = IAllocator(self.cfg, self.rpc,
11342
                       mode=self.op.mode,
11343
                       evac_nodes=self.op.evac_nodes)
11344
    else:
11345
      raise errors.ProgrammerError("Uncatched mode %s in"
11346
                                   " LUTestAllocator.Exec", self.op.mode)
11347

    
11348
    if self.op.direction == constants.IALLOCATOR_DIR_IN:
11349
      result = ial.in_text
11350
    else:
11351
      ial.Run(self.op.allocator, validate=False)
11352
      result = ial.out_text
11353
    return result
11354

    
11355

    
11356
#: Query type implementations
11357
_QUERY_IMPL = {
11358
  constants.QR_INSTANCE: _InstanceQuery,
11359
  constants.QR_NODE: _NodeQuery,
11360
  constants.QR_GROUP: _GroupQuery,
11361
  }
11362

    
11363

    
11364
def _GetQueryImplementation(name):
11365
  """Returns the implemtnation for a query type.
11366

11367
  @param name: Query type, must be one of L{constants.QR_OP_QUERY}
11368

11369
  """
11370
  try:
11371
    return _QUERY_IMPL[name]
11372
  except KeyError:
11373
    raise errors.OpPrereqError("Unknown query resource '%s'" % name,
11374
                               errors.ECODE_INVAL)