Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ c190e817

History | View | Annotate | Download (372 kB)

1
#
2
#
3

    
4
# Copyright (C) 2006, 2007, 2008, 2009, 2010 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

    
43
from ganeti import ssh
44
from ganeti import utils
45
from ganeti import errors
46
from ganeti import hypervisor
47
from ganeti import locking
48
from ganeti import constants
49
from ganeti import objects
50
from ganeti import serializer
51
from ganeti import ssconf
52
from ganeti import uidpool
53
from ganeti import compat
54
from ganeti import masterd
55
from ganeti import netutils
56
from ganeti import ht
57

    
58
import ganeti.masterd.instance # pylint: disable-msg=W0611
59

    
60
# Common opcode attributes
61

    
62
#: output fields for a query operation
63
_POutputFields = ("output_fields", ht.NoDefault, ht.TListOf(ht.TNonEmptyString))
64

    
65

    
66
#: the shutdown timeout
67
_PShutdownTimeout = ("shutdown_timeout", constants.DEFAULT_SHUTDOWN_TIMEOUT,
68
                     ht.TPositiveInt)
69

    
70
#: the force parameter
71
_PForce = ("force", False, ht.TBool)
72

    
73
#: a required instance name (for single-instance LUs)
74
_PInstanceName = ("instance_name", ht.NoDefault, ht.TNonEmptyString)
75

    
76
#: Whether to ignore offline nodes
77
_PIgnoreOfflineNodes = ("ignore_offline_nodes", False, ht.TBool)
78

    
79
#: a required node name (for single-node LUs)
80
_PNodeName = ("node_name", ht.NoDefault, ht.TNonEmptyString)
81

    
82
#: the migration type (live/non-live)
83
_PMigrationMode = ("mode", None,
84
                   ht.TOr(ht.TNone, ht.TElemOf(constants.HT_MIGRATION_MODES)))
85

    
86
#: the obsolete 'live' mode (boolean)
87
_PMigrationLive = ("live", None, ht.TMaybeBool)
88

    
89

    
90
# End types
91
class LogicalUnit(object):
92
  """Logical Unit base class.
93

94
  Subclasses must follow these rules:
95
    - implement ExpandNames
96
    - implement CheckPrereq (except when tasklets are used)
97
    - implement Exec (except when tasklets are used)
98
    - implement BuildHooksEnv
99
    - redefine HPATH and HTYPE
100
    - optionally redefine their run requirements:
101
        REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
102

103
  Note that all commands require root permissions.
104

105
  @ivar dry_run_result: the value (if any) that will be returned to the caller
106
      in dry-run mode (signalled by opcode dry_run parameter)
107
  @cvar _OP_PARAMS: a list of opcode attributes, their defaults values
108
      they should get if not already defined, and types they must match
109

110
  """
111
  HPATH = None
112
  HTYPE = None
113
  _OP_PARAMS = []
114
  REQ_BGL = True
115

    
116
  def __init__(self, processor, op, context, rpc):
117
    """Constructor for LogicalUnit.
118

119
    This needs to be overridden in derived classes in order to check op
120
    validity.
121

122
    """
123
    self.proc = processor
124
    self.op = op
125
    self.cfg = context.cfg
126
    self.context = context
127
    self.rpc = rpc
128
    # Dicts used to declare locking needs to mcpu
129
    self.needed_locks = None
130
    self.acquired_locks = {}
131
    self.share_locks = dict.fromkeys(locking.LEVELS, 0)
132
    self.add_locks = {}
133
    self.remove_locks = {}
134
    # Used to force good behavior when calling helper functions
135
    self.recalculate_locks = {}
136
    self.__ssh = None
137
    # logging
138
    self.Log = processor.Log # pylint: disable-msg=C0103
139
    self.LogWarning = processor.LogWarning # pylint: disable-msg=C0103
140
    self.LogInfo = processor.LogInfo # pylint: disable-msg=C0103
141
    self.LogStep = processor.LogStep # pylint: disable-msg=C0103
142
    # support for dry-run
143
    self.dry_run_result = None
144
    # support for generic debug attribute
145
    if (not hasattr(self.op, "debug_level") or
146
        not isinstance(self.op.debug_level, int)):
147
      self.op.debug_level = 0
148

    
149
    # Tasklets
150
    self.tasklets = None
151

    
152
    # The new kind-of-type-system
153
    op_id = self.op.OP_ID
154
    for attr_name, aval, test in self._OP_PARAMS:
155
      if not hasattr(op, attr_name):
156
        if aval == ht.NoDefault:
157
          raise errors.OpPrereqError("Required parameter '%s.%s' missing" %
158
                                     (op_id, attr_name), errors.ECODE_INVAL)
159
        else:
160
          if callable(aval):
161
            dval = aval()
162
          else:
163
            dval = aval
164
          setattr(self.op, attr_name, dval)
165
      attr_val = getattr(op, attr_name)
166
      if test == ht.NoType:
167
        # no tests here
168
        continue
169
      if not callable(test):
170
        raise errors.ProgrammerError("Validation for parameter '%s.%s' failed,"
171
                                     " given type is not a proper type (%s)" %
172
                                     (op_id, attr_name, test))
173
      if not test(attr_val):
174
        logging.error("OpCode %s, parameter %s, has invalid type %s/value %s",
175
                      self.op.OP_ID, attr_name, type(attr_val), attr_val)
176
        raise errors.OpPrereqError("Parameter '%s.%s' fails validation" %
177
                                   (op_id, attr_name), errors.ECODE_INVAL)
178

    
179
    self.CheckArguments()
180

    
181
  def __GetSSH(self):
182
    """Returns the SshRunner object
183

184
    """
185
    if not self.__ssh:
186
      self.__ssh = ssh.SshRunner(self.cfg.GetClusterName())
187
    return self.__ssh
188

    
189
  ssh = property(fget=__GetSSH)
190

    
191
  def CheckArguments(self):
192
    """Check syntactic validity for the opcode arguments.
193

194
    This method is for doing a simple syntactic check and ensure
195
    validity of opcode parameters, without any cluster-related
196
    checks. While the same can be accomplished in ExpandNames and/or
197
    CheckPrereq, doing these separate is better because:
198

199
      - ExpandNames is left as as purely a lock-related function
200
      - CheckPrereq is run after we have acquired locks (and possible
201
        waited for them)
202

203
    The function is allowed to change the self.op attribute so that
204
    later methods can no longer worry about missing parameters.
205

206
    """
207
    pass
208

    
209
  def ExpandNames(self):
210
    """Expand names for this LU.
211

212
    This method is called before starting to execute the opcode, and it should
213
    update all the parameters of the opcode to their canonical form (e.g. a
214
    short node name must be fully expanded after this method has successfully
215
    completed). This way locking, hooks, logging, ecc. can work correctly.
216

217
    LUs which implement this method must also populate the self.needed_locks
218
    member, as a dict with lock levels as keys, and a list of needed lock names
219
    as values. Rules:
220

221
      - use an empty dict if you don't need any lock
222
      - if you don't need any lock at a particular level omit that level
223
      - don't put anything for the BGL level
224
      - if you want all locks at a level use locking.ALL_SET as a value
225

226
    If you need to share locks (rather than acquire them exclusively) at one
227
    level you can modify self.share_locks, setting a true value (usually 1) for
228
    that level. By default locks are not shared.
229

230
    This function can also define a list of tasklets, which then will be
231
    executed in order instead of the usual LU-level CheckPrereq and Exec
232
    functions, if those are not defined by the LU.
233

234
    Examples::
235

236
      # Acquire all nodes and one instance
237
      self.needed_locks = {
238
        locking.LEVEL_NODE: locking.ALL_SET,
239
        locking.LEVEL_INSTANCE: ['instance1.example.com'],
240
      }
241
      # Acquire just two nodes
242
      self.needed_locks = {
243
        locking.LEVEL_NODE: ['node1.example.com', 'node2.example.com'],
244
      }
245
      # Acquire no locks
246
      self.needed_locks = {} # No, you can't leave it to the default value None
247

248
    """
249
    # The implementation of this method is mandatory only if the new LU is
250
    # concurrent, so that old LUs don't need to be changed all at the same
251
    # time.
252
    if self.REQ_BGL:
253
      self.needed_locks = {} # Exclusive LUs don't need locks.
254
    else:
255
      raise NotImplementedError
256

    
257
  def DeclareLocks(self, level):
258
    """Declare LU locking needs for a level
259

260
    While most LUs can just declare their locking needs at ExpandNames time,
261
    sometimes there's the need to calculate some locks after having acquired
262
    the ones before. This function is called just before acquiring locks at a
263
    particular level, but after acquiring the ones at lower levels, and permits
264
    such calculations. It can be used to modify self.needed_locks, and by
265
    default it does nothing.
266

267
    This function is only called if you have something already set in
268
    self.needed_locks for the level.
269

270
    @param level: Locking level which is going to be locked
271
    @type level: member of ganeti.locking.LEVELS
272

273
    """
274

    
275
  def CheckPrereq(self):
276
    """Check prerequisites for this LU.
277

278
    This method should check that the prerequisites for the execution
279
    of this LU are fulfilled. It can do internode communication, but
280
    it should be idempotent - no cluster or system changes are
281
    allowed.
282

283
    The method should raise errors.OpPrereqError in case something is
284
    not fulfilled. Its return value is ignored.
285

286
    This method should also update all the parameters of the opcode to
287
    their canonical form if it hasn't been done by ExpandNames before.
288

289
    """
290
    if self.tasklets is not None:
291
      for (idx, tl) in enumerate(self.tasklets):
292
        logging.debug("Checking prerequisites for tasklet %s/%s",
293
                      idx + 1, len(self.tasklets))
294
        tl.CheckPrereq()
295
    else:
296
      pass
297

    
298
  def Exec(self, feedback_fn):
299
    """Execute the LU.
300

301
    This method should implement the actual work. It should raise
302
    errors.OpExecError for failures that are somewhat dealt with in
303
    code, or expected.
304

305
    """
306
    if self.tasklets is not None:
307
      for (idx, tl) in enumerate(self.tasklets):
308
        logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets))
309
        tl.Exec(feedback_fn)
310
    else:
311
      raise NotImplementedError
312

    
313
  def BuildHooksEnv(self):
314
    """Build hooks environment for this LU.
315

316
    This method should return a three-node tuple consisting of: a dict
317
    containing the environment that will be used for running the
318
    specific hook for this LU, a list of node names on which the hook
319
    should run before the execution, and a list of node names on which
320
    the hook should run after the execution.
321

322
    The keys of the dict must not have 'GANETI_' prefixed as this will
323
    be handled in the hooks runner. Also note additional keys will be
324
    added by the hooks runner. If the LU doesn't define any
325
    environment, an empty dict (and not None) should be returned.
326

327
    No nodes should be returned as an empty list (and not None).
328

329
    Note that if the HPATH for a LU class is None, this function will
330
    not be called.
331

332
    """
333
    raise NotImplementedError
334

    
335
  def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
336
    """Notify the LU about the results of its hooks.
337

338
    This method is called every time a hooks phase is executed, and notifies
339
    the Logical Unit about the hooks' result. The LU can then use it to alter
340
    its result based on the hooks.  By default the method does nothing and the
341
    previous result is passed back unchanged but any LU can define it if it
342
    wants to use the local cluster hook-scripts somehow.
343

344
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
345
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
346
    @param hook_results: the results of the multi-node hooks rpc call
347
    @param feedback_fn: function used send feedback back to the caller
348
    @param lu_result: the previous Exec result this LU had, or None
349
        in the PRE phase
350
    @return: the new Exec result, based on the previous result
351
        and hook results
352

353
    """
354
    # API must be kept, thus we ignore the unused argument and could
355
    # be a function warnings
356
    # pylint: disable-msg=W0613,R0201
357
    return lu_result
358

    
359
  def _ExpandAndLockInstance(self):
360
    """Helper function to expand and lock an instance.
361

362
    Many LUs that work on an instance take its name in self.op.instance_name
363
    and need to expand it and then declare the expanded name for locking. This
364
    function does it, and then updates self.op.instance_name to the expanded
365
    name. It also initializes needed_locks as a dict, if this hasn't been done
366
    before.
367

368
    """
369
    if self.needed_locks is None:
370
      self.needed_locks = {}
371
    else:
372
      assert locking.LEVEL_INSTANCE not in self.needed_locks, \
373
        "_ExpandAndLockInstance called with instance-level locks set"
374
    self.op.instance_name = _ExpandInstanceName(self.cfg,
375
                                                self.op.instance_name)
376
    self.needed_locks[locking.LEVEL_INSTANCE] = self.op.instance_name
377

    
378
  def _LockInstancesNodes(self, primary_only=False):
379
    """Helper function to declare instances' nodes for locking.
380

381
    This function should be called after locking one or more instances to lock
382
    their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
383
    with all primary or secondary nodes for instances already locked and
384
    present in self.needed_locks[locking.LEVEL_INSTANCE].
385

386
    It should be called from DeclareLocks, and for safety only works if
387
    self.recalculate_locks[locking.LEVEL_NODE] is set.
388

389
    In the future it may grow parameters to just lock some instance's nodes, or
390
    to just lock primaries or secondary nodes, if needed.
391

392
    If should be called in DeclareLocks in a way similar to::
393

394
      if level == locking.LEVEL_NODE:
395
        self._LockInstancesNodes()
396

397
    @type primary_only: boolean
398
    @param primary_only: only lock primary nodes of locked instances
399

400
    """
401
    assert locking.LEVEL_NODE in self.recalculate_locks, \
402
      "_LockInstancesNodes helper function called with no nodes to recalculate"
403

    
404
    # TODO: check if we're really been called with the instance locks held
405

    
406
    # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
407
    # future we might want to have different behaviors depending on the value
408
    # of self.recalculate_locks[locking.LEVEL_NODE]
409
    wanted_nodes = []
410
    for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
411
      instance = self.context.cfg.GetInstanceInfo(instance_name)
412
      wanted_nodes.append(instance.primary_node)
413
      if not primary_only:
414
        wanted_nodes.extend(instance.secondary_nodes)
415

    
416
    if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE:
417
      self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
418
    elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND:
419
      self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
420

    
421
    del self.recalculate_locks[locking.LEVEL_NODE]
422

    
423

    
424
class NoHooksLU(LogicalUnit): # pylint: disable-msg=W0223
425
  """Simple LU which runs no hooks.
426

427
  This LU is intended as a parent for other LogicalUnits which will
428
  run no hooks, in order to reduce duplicate code.
429

430
  """
431
  HPATH = None
432
  HTYPE = None
433

    
434
  def BuildHooksEnv(self):
435
    """Empty BuildHooksEnv for NoHooksLu.
436

437
    This just raises an error.
438

439
    """
440
    assert False, "BuildHooksEnv called for NoHooksLUs"
441

    
442

    
443
class Tasklet:
444
  """Tasklet base class.
445

446
  Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
447
  they can mix legacy code with tasklets. Locking needs to be done in the LU,
448
  tasklets know nothing about locks.
449

450
  Subclasses must follow these rules:
451
    - Implement CheckPrereq
452
    - Implement Exec
453

454
  """
455
  def __init__(self, lu):
456
    self.lu = lu
457

    
458
    # Shortcuts
459
    self.cfg = lu.cfg
460
    self.rpc = lu.rpc
461

    
462
  def CheckPrereq(self):
463
    """Check prerequisites for this tasklets.
464

465
    This method should check whether the prerequisites for the execution of
466
    this tasklet are fulfilled. It can do internode communication, but it
467
    should be idempotent - no cluster or system changes are allowed.
468

469
    The method should raise errors.OpPrereqError in case something is not
470
    fulfilled. Its return value is ignored.
471

472
    This method should also update all parameters to their canonical form if it
473
    hasn't been done before.
474

475
    """
476
    pass
477

    
478
  def Exec(self, feedback_fn):
479
    """Execute the tasklet.
480

481
    This method should implement the actual work. It should raise
482
    errors.OpExecError for failures that are somewhat dealt with in code, or
483
    expected.
484

485
    """
486
    raise NotImplementedError
487

    
488

    
489
def _GetWantedNodes(lu, nodes):
490
  """Returns list of checked and expanded node names.
491

492
  @type lu: L{LogicalUnit}
493
  @param lu: the logical unit on whose behalf we execute
494
  @type nodes: list
495
  @param nodes: list of node names or None for all nodes
496
  @rtype: list
497
  @return: the list of nodes, sorted
498
  @raise errors.ProgrammerError: if the nodes parameter is wrong type
499

500
  """
501
  if not nodes:
502
    raise errors.ProgrammerError("_GetWantedNodes should only be called with a"
503
      " non-empty list of nodes whose name is to be expanded.")
504

    
505
  wanted = [_ExpandNodeName(lu.cfg, name) for name in nodes]
506
  return utils.NiceSort(wanted)
507

    
508

    
509
def _GetWantedInstances(lu, instances):
510
  """Returns list of checked and expanded instance names.
511

512
  @type lu: L{LogicalUnit}
513
  @param lu: the logical unit on whose behalf we execute
514
  @type instances: list
515
  @param instances: list of instance names or None for all instances
516
  @rtype: list
517
  @return: the list of instances, sorted
518
  @raise errors.OpPrereqError: if the instances parameter is wrong type
519
  @raise errors.OpPrereqError: if any of the passed instances is not found
520

521
  """
522
  if instances:
523
    wanted = [_ExpandInstanceName(lu.cfg, name) for name in instances]
524
  else:
525
    wanted = utils.NiceSort(lu.cfg.GetInstanceList())
526
  return wanted
527

    
528

    
529
def _GetUpdatedParams(old_params, update_dict,
530
                      use_default=True, use_none=False):
531
  """Return the new version of a parameter dictionary.
532

533
  @type old_params: dict
534
  @param old_params: old parameters
535
  @type update_dict: dict
536
  @param update_dict: dict containing new parameter values, or
537
      constants.VALUE_DEFAULT to reset the parameter to its default
538
      value
539
  @param use_default: boolean
540
  @type use_default: whether to recognise L{constants.VALUE_DEFAULT}
541
      values as 'to be deleted' values
542
  @param use_none: boolean
543
  @type use_none: whether to recognise C{None} values as 'to be
544
      deleted' values
545
  @rtype: dict
546
  @return: the new parameter dictionary
547

548
  """
549
  params_copy = copy.deepcopy(old_params)
550
  for key, val in update_dict.iteritems():
551
    if ((use_default and val == constants.VALUE_DEFAULT) or
552
        (use_none and val is None)):
553
      try:
554
        del params_copy[key]
555
      except KeyError:
556
        pass
557
    else:
558
      params_copy[key] = val
559
  return params_copy
560

    
561

    
562
def _CheckOutputFields(static, dynamic, selected):
563
  """Checks whether all selected fields are valid.
564

565
  @type static: L{utils.FieldSet}
566
  @param static: static fields set
567
  @type dynamic: L{utils.FieldSet}
568
  @param dynamic: dynamic fields set
569

570
  """
571
  f = utils.FieldSet()
572
  f.Extend(static)
573
  f.Extend(dynamic)
574

    
575
  delta = f.NonMatching(selected)
576
  if delta:
577
    raise errors.OpPrereqError("Unknown output fields selected: %s"
578
                               % ",".join(delta), errors.ECODE_INVAL)
579

    
580

    
581
def _CheckGlobalHvParams(params):
582
  """Validates that given hypervisor params are not global ones.
583

584
  This will ensure that instances don't get customised versions of
585
  global params.
586

587
  """
588
  used_globals = constants.HVC_GLOBALS.intersection(params)
589
  if used_globals:
590
    msg = ("The following hypervisor parameters are global and cannot"
591
           " be customized at instance level, please modify them at"
592
           " cluster level: %s" % utils.CommaJoin(used_globals))
593
    raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
594

    
595

    
596
def _CheckNodeOnline(lu, node):
597
  """Ensure that a given node is online.
598

599
  @param lu: the LU on behalf of which we make the check
600
  @param node: the node to check
601
  @raise errors.OpPrereqError: if the node is offline
602

603
  """
604
  if lu.cfg.GetNodeInfo(node).offline:
605
    raise errors.OpPrereqError("Can't use offline node %s" % node,
606
                               errors.ECODE_INVAL)
607

    
608

    
609
def _CheckNodeNotDrained(lu, node):
610
  """Ensure that a given node is not drained.
611

612
  @param lu: the LU on behalf of which we make the check
613
  @param node: the node to check
614
  @raise errors.OpPrereqError: if the node is drained
615

616
  """
617
  if lu.cfg.GetNodeInfo(node).drained:
618
    raise errors.OpPrereqError("Can't use drained node %s" % node,
619
                               errors.ECODE_INVAL)
620

    
621

    
622
def _CheckNodeHasOS(lu, node, os_name, force_variant):
623
  """Ensure that a node supports a given OS.
624

625
  @param lu: the LU on behalf of which we make the check
626
  @param node: the node to check
627
  @param os_name: the OS to query about
628
  @param force_variant: whether to ignore variant errors
629
  @raise errors.OpPrereqError: if the node is not supporting the OS
630

631
  """
632
  result = lu.rpc.call_os_get(node, os_name)
633
  result.Raise("OS '%s' not in supported OS list for node %s" %
634
               (os_name, node),
635
               prereq=True, ecode=errors.ECODE_INVAL)
636
  if not force_variant:
637
    _CheckOSVariant(result.payload, os_name)
638

    
639

    
640
def _RequireFileStorage():
641
  """Checks that file storage is enabled.
642

643
  @raise errors.OpPrereqError: when file storage is disabled
644

645
  """
646
  if not constants.ENABLE_FILE_STORAGE:
647
    raise errors.OpPrereqError("File storage disabled at configure time",
648
                               errors.ECODE_INVAL)
649

    
650

    
651
def _CheckDiskTemplate(template):
652
  """Ensure a given disk template is valid.
653

654
  """
655
  if template not in constants.DISK_TEMPLATES:
656
    msg = ("Invalid disk template name '%s', valid templates are: %s" %
657
           (template, utils.CommaJoin(constants.DISK_TEMPLATES)))
658
    raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
659
  if template == constants.DT_FILE:
660
    _RequireFileStorage()
661
  return True
662

    
663

    
664
def _CheckStorageType(storage_type):
665
  """Ensure a given storage type is valid.
666

667
  """
668
  if storage_type not in constants.VALID_STORAGE_TYPES:
669
    raise errors.OpPrereqError("Unknown storage type: %s" % storage_type,
670
                               errors.ECODE_INVAL)
671
  if storage_type == constants.ST_FILE:
672
    _RequireFileStorage()
673
  return True
674

    
675

    
676
def _GetClusterDomainSecret():
677
  """Reads the cluster domain secret.
678

679
  """
680
  return utils.ReadOneLineFile(constants.CLUSTER_DOMAIN_SECRET_FILE,
681
                               strict=True)
682

    
683

    
684
def _CheckInstanceDown(lu, instance, reason):
685
  """Ensure that an instance is not running."""
686
  if instance.admin_up:
687
    raise errors.OpPrereqError("Instance %s is marked to be up, %s" %
688
                               (instance.name, reason), errors.ECODE_STATE)
689

    
690
  pnode = instance.primary_node
691
  ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
692
  ins_l.Raise("Can't contact node %s for instance information" % pnode,
693
              prereq=True, ecode=errors.ECODE_ENVIRON)
694

    
695
  if instance.name in ins_l.payload:
696
    raise errors.OpPrereqError("Instance %s is running, %s" %
697
                               (instance.name, reason), errors.ECODE_STATE)
698

    
699

    
700
def _ExpandItemName(fn, name, kind):
701
  """Expand an item name.
702

703
  @param fn: the function to use for expansion
704
  @param name: requested item name
705
  @param kind: text description ('Node' or 'Instance')
706
  @return: the resolved (full) name
707
  @raise errors.OpPrereqError: if the item is not found
708

709
  """
710
  full_name = fn(name)
711
  if full_name is None:
712
    raise errors.OpPrereqError("%s '%s' not known" % (kind, name),
713
                               errors.ECODE_NOENT)
714
  return full_name
715

    
716

    
717
def _ExpandNodeName(cfg, name):
718
  """Wrapper over L{_ExpandItemName} for nodes."""
719
  return _ExpandItemName(cfg.ExpandNodeName, name, "Node")
720

    
721

    
722
def _ExpandInstanceName(cfg, name):
723
  """Wrapper over L{_ExpandItemName} for instance."""
724
  return _ExpandItemName(cfg.ExpandInstanceName, name, "Instance")
725

    
726

    
727
def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
728
                          memory, vcpus, nics, disk_template, disks,
729
                          bep, hvp, hypervisor_name):
730
  """Builds instance related env variables for hooks
731

732
  This builds the hook environment from individual variables.
733

734
  @type name: string
735
  @param name: the name of the instance
736
  @type primary_node: string
737
  @param primary_node: the name of the instance's primary node
738
  @type secondary_nodes: list
739
  @param secondary_nodes: list of secondary nodes as strings
740
  @type os_type: string
741
  @param os_type: the name of the instance's OS
742
  @type status: boolean
743
  @param status: the should_run status of the instance
744
  @type memory: string
745
  @param memory: the memory size of the instance
746
  @type vcpus: string
747
  @param vcpus: the count of VCPUs the instance has
748
  @type nics: list
749
  @param nics: list of tuples (ip, mac, mode, link) representing
750
      the NICs the instance has
751
  @type disk_template: string
752
  @param disk_template: the disk template of the instance
753
  @type disks: list
754
  @param disks: the list of (size, mode) pairs
755
  @type bep: dict
756
  @param bep: the backend parameters for the instance
757
  @type hvp: dict
758
  @param hvp: the hypervisor parameters for the instance
759
  @type hypervisor_name: string
760
  @param hypervisor_name: the hypervisor for the instance
761
  @rtype: dict
762
  @return: the hook environment for this instance
763

764
  """
765
  if status:
766
    str_status = "up"
767
  else:
768
    str_status = "down"
769
  env = {
770
    "OP_TARGET": name,
771
    "INSTANCE_NAME": name,
772
    "INSTANCE_PRIMARY": primary_node,
773
    "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
774
    "INSTANCE_OS_TYPE": os_type,
775
    "INSTANCE_STATUS": str_status,
776
    "INSTANCE_MEMORY": memory,
777
    "INSTANCE_VCPUS": vcpus,
778
    "INSTANCE_DISK_TEMPLATE": disk_template,
779
    "INSTANCE_HYPERVISOR": hypervisor_name,
780
  }
781

    
782
  if nics:
783
    nic_count = len(nics)
784
    for idx, (ip, mac, mode, link) in enumerate(nics):
785
      if ip is None:
786
        ip = ""
787
      env["INSTANCE_NIC%d_IP" % idx] = ip
788
      env["INSTANCE_NIC%d_MAC" % idx] = mac
789
      env["INSTANCE_NIC%d_MODE" % idx] = mode
790
      env["INSTANCE_NIC%d_LINK" % idx] = link
791
      if mode == constants.NIC_MODE_BRIDGED:
792
        env["INSTANCE_NIC%d_BRIDGE" % idx] = link
793
  else:
794
    nic_count = 0
795

    
796
  env["INSTANCE_NIC_COUNT"] = nic_count
797

    
798
  if disks:
799
    disk_count = len(disks)
800
    for idx, (size, mode) in enumerate(disks):
801
      env["INSTANCE_DISK%d_SIZE" % idx] = size
802
      env["INSTANCE_DISK%d_MODE" % idx] = mode
803
  else:
804
    disk_count = 0
805

    
806
  env["INSTANCE_DISK_COUNT"] = disk_count
807

    
808
  for source, kind in [(bep, "BE"), (hvp, "HV")]:
809
    for key, value in source.items():
810
      env["INSTANCE_%s_%s" % (kind, key)] = value
811

    
812
  return env
813

    
814

    
815
def _NICListToTuple(lu, nics):
816
  """Build a list of nic information tuples.
817

818
  This list is suitable to be passed to _BuildInstanceHookEnv or as a return
819
  value in LUQueryInstanceData.
820

821
  @type lu:  L{LogicalUnit}
822
  @param lu: the logical unit on whose behalf we execute
823
  @type nics: list of L{objects.NIC}
824
  @param nics: list of nics to convert to hooks tuples
825

826
  """
827
  hooks_nics = []
828
  cluster = lu.cfg.GetClusterInfo()
829
  for nic in nics:
830
    ip = nic.ip
831
    mac = nic.mac
832
    filled_params = cluster.SimpleFillNIC(nic.nicparams)
833
    mode = filled_params[constants.NIC_MODE]
834
    link = filled_params[constants.NIC_LINK]
835
    hooks_nics.append((ip, mac, mode, link))
836
  return hooks_nics
837

    
838

    
839
def _BuildInstanceHookEnvByObject(lu, instance, override=None):
840
  """Builds instance related env variables for hooks from an object.
841

842
  @type lu: L{LogicalUnit}
843
  @param lu: the logical unit on whose behalf we execute
844
  @type instance: L{objects.Instance}
845
  @param instance: the instance for which we should build the
846
      environment
847
  @type override: dict
848
  @param override: dictionary with key/values that will override
849
      our values
850
  @rtype: dict
851
  @return: the hook environment dictionary
852

853
  """
854
  cluster = lu.cfg.GetClusterInfo()
855
  bep = cluster.FillBE(instance)
856
  hvp = cluster.FillHV(instance)
857
  args = {
858
    'name': instance.name,
859
    'primary_node': instance.primary_node,
860
    'secondary_nodes': instance.secondary_nodes,
861
    'os_type': instance.os,
862
    'status': instance.admin_up,
863
    'memory': bep[constants.BE_MEMORY],
864
    'vcpus': bep[constants.BE_VCPUS],
865
    'nics': _NICListToTuple(lu, instance.nics),
866
    'disk_template': instance.disk_template,
867
    'disks': [(disk.size, disk.mode) for disk in instance.disks],
868
    'bep': bep,
869
    'hvp': hvp,
870
    'hypervisor_name': instance.hypervisor,
871
  }
872
  if override:
873
    args.update(override)
874
  return _BuildInstanceHookEnv(**args) # pylint: disable-msg=W0142
875

    
876

    
877
def _AdjustCandidatePool(lu, exceptions):
878
  """Adjust the candidate pool after node operations.
879

880
  """
881
  mod_list = lu.cfg.MaintainCandidatePool(exceptions)
882
  if mod_list:
883
    lu.LogInfo("Promoted nodes to master candidate role: %s",
884
               utils.CommaJoin(node.name for node in mod_list))
885
    for name in mod_list:
886
      lu.context.ReaddNode(name)
887
  mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
888
  if mc_now > mc_max:
889
    lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
890
               (mc_now, mc_max))
891

    
892

    
893
def _DecideSelfPromotion(lu, exceptions=None):
894
  """Decide whether I should promote myself as a master candidate.
895

896
  """
897
  cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
898
  mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
899
  # the new node will increase mc_max with one, so:
900
  mc_should = min(mc_should + 1, cp_size)
901
  return mc_now < mc_should
902

    
903

    
904
def _CheckNicsBridgesExist(lu, target_nics, target_node):
905
  """Check that the brigdes needed by a list of nics exist.
906

907
  """
908
  cluster = lu.cfg.GetClusterInfo()
909
  paramslist = [cluster.SimpleFillNIC(nic.nicparams) for nic in target_nics]
910
  brlist = [params[constants.NIC_LINK] for params in paramslist
911
            if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
912
  if brlist:
913
    result = lu.rpc.call_bridges_exist(target_node, brlist)
914
    result.Raise("Error checking bridges on destination node '%s'" %
915
                 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
916

    
917

    
918
def _CheckInstanceBridgesExist(lu, instance, node=None):
919
  """Check that the brigdes needed by an instance exist.
920

921
  """
922
  if node is None:
923
    node = instance.primary_node
924
  _CheckNicsBridgesExist(lu, instance.nics, node)
925

    
926

    
927
def _CheckOSVariant(os_obj, name):
928
  """Check whether an OS name conforms to the os variants specification.
929

930
  @type os_obj: L{objects.OS}
931
  @param os_obj: OS object to check
932
  @type name: string
933
  @param name: OS name passed by the user, to check for validity
934

935
  """
936
  if not os_obj.supported_variants:
937
    return
938
  variant = objects.OS.GetVariant(name)
939
  if not variant:
940
    raise errors.OpPrereqError("OS name must include a variant",
941
                               errors.ECODE_INVAL)
942

    
943
  if variant not in os_obj.supported_variants:
944
    raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
945

    
946

    
947
def _GetNodeInstancesInner(cfg, fn):
948
  return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
949

    
950

    
951
def _GetNodeInstances(cfg, node_name):
952
  """Returns a list of all primary and secondary instances on a node.
953

954
  """
955

    
956
  return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
957

    
958

    
959
def _GetNodePrimaryInstances(cfg, node_name):
960
  """Returns primary instances on a node.
961

962
  """
963
  return _GetNodeInstancesInner(cfg,
964
                                lambda inst: node_name == inst.primary_node)
965

    
966

    
967
def _GetNodeSecondaryInstances(cfg, node_name):
968
  """Returns secondary instances on a node.
969

970
  """
971
  return _GetNodeInstancesInner(cfg,
972
                                lambda inst: node_name in inst.secondary_nodes)
973

    
974

    
975
def _GetStorageTypeArgs(cfg, storage_type):
976
  """Returns the arguments for a storage type.
977

978
  """
979
  # Special case for file storage
980
  if storage_type == constants.ST_FILE:
981
    # storage.FileStorage wants a list of storage directories
982
    return [[cfg.GetFileStorageDir()]]
983

    
984
  return []
985

    
986

    
987
def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq):
988
  faulty = []
989

    
990
  for dev in instance.disks:
991
    cfg.SetDiskID(dev, node_name)
992

    
993
  result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks)
994
  result.Raise("Failed to get disk status from node %s" % node_name,
995
               prereq=prereq, ecode=errors.ECODE_ENVIRON)
996

    
997
  for idx, bdev_status in enumerate(result.payload):
998
    if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
999
      faulty.append(idx)
1000

    
1001
  return faulty
1002

    
1003

    
1004
def _CheckIAllocatorOrNode(lu, iallocator_slot, node_slot):
1005
  """Check the sanity of iallocator and node arguments and use the
1006
  cluster-wide iallocator if appropriate.
1007

1008
  Check that at most one of (iallocator, node) is specified. If none is
1009
  specified, then the LU's opcode's iallocator slot is filled with the
1010
  cluster-wide default iallocator.
1011

1012
  @type iallocator_slot: string
1013
  @param iallocator_slot: the name of the opcode iallocator slot
1014
  @type node_slot: string
1015
  @param node_slot: the name of the opcode target node slot
1016

1017
  """
1018
  node = getattr(lu.op, node_slot, None)
1019
  iallocator = getattr(lu.op, iallocator_slot, None)
1020

    
1021
  if node is not None and iallocator is not None:
1022
    raise errors.OpPrereqError("Do not specify both, iallocator and node.",
1023
                               errors.ECODE_INVAL)
1024
  elif node is None and iallocator is None:
1025
    default_iallocator = lu.cfg.GetDefaultIAllocator()
1026
    if default_iallocator:
1027
      setattr(lu.op, iallocator_slot, default_iallocator)
1028
    else:
1029
      raise errors.OpPrereqError("No iallocator or node given and no"
1030
                                 " cluster-wide default iallocator found."
1031
                                 " Please specify either an iallocator or a"
1032
                                 " node, or set a cluster-wide default"
1033
                                 " iallocator.")
1034

    
1035

    
1036
class LUPostInitCluster(LogicalUnit):
1037
  """Logical unit for running hooks after cluster initialization.
1038

1039
  """
1040
  HPATH = "cluster-init"
1041
  HTYPE = constants.HTYPE_CLUSTER
1042

    
1043
  def BuildHooksEnv(self):
1044
    """Build hooks env.
1045

1046
    """
1047
    env = {"OP_TARGET": self.cfg.GetClusterName()}
1048
    mn = self.cfg.GetMasterNode()
1049
    return env, [], [mn]
1050

    
1051
  def Exec(self, feedback_fn):
1052
    """Nothing to do.
1053

1054
    """
1055
    return True
1056

    
1057

    
1058
class LUDestroyCluster(LogicalUnit):
1059
  """Logical unit for destroying the cluster.
1060

1061
  """
1062
  HPATH = "cluster-destroy"
1063
  HTYPE = constants.HTYPE_CLUSTER
1064

    
1065
  def BuildHooksEnv(self):
1066
    """Build hooks env.
1067

1068
    """
1069
    env = {"OP_TARGET": self.cfg.GetClusterName()}
1070
    return env, [], []
1071

    
1072
  def CheckPrereq(self):
1073
    """Check prerequisites.
1074

1075
    This checks whether the cluster is empty.
1076

1077
    Any errors are signaled by raising errors.OpPrereqError.
1078

1079
    """
1080
    master = self.cfg.GetMasterNode()
1081

    
1082
    nodelist = self.cfg.GetNodeList()
1083
    if len(nodelist) != 1 or nodelist[0] != master:
1084
      raise errors.OpPrereqError("There are still %d node(s) in"
1085
                                 " this cluster." % (len(nodelist) - 1),
1086
                                 errors.ECODE_INVAL)
1087
    instancelist = self.cfg.GetInstanceList()
1088
    if instancelist:
1089
      raise errors.OpPrereqError("There are still %d instance(s) in"
1090
                                 " this cluster." % len(instancelist),
1091
                                 errors.ECODE_INVAL)
1092

    
1093
  def Exec(self, feedback_fn):
1094
    """Destroys the cluster.
1095

1096
    """
1097
    master = self.cfg.GetMasterNode()
1098

    
1099
    # Run post hooks on master node before it's removed
1100
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
1101
    try:
1102
      hm.RunPhase(constants.HOOKS_PHASE_POST, [master])
1103
    except:
1104
      # pylint: disable-msg=W0702
1105
      self.LogWarning("Errors occurred running hooks on %s" % master)
1106

    
1107
    result = self.rpc.call_node_stop_master(master, False)
1108
    result.Raise("Could not disable the master role")
1109

    
1110
    return master
1111

    
1112

    
1113
def _VerifyCertificate(filename):
1114
  """Verifies a certificate for LUVerifyCluster.
1115

1116
  @type filename: string
1117
  @param filename: Path to PEM file
1118

1119
  """
1120
  try:
1121
    cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
1122
                                           utils.ReadFile(filename))
1123
  except Exception, err: # pylint: disable-msg=W0703
1124
    return (LUVerifyCluster.ETYPE_ERROR,
1125
            "Failed to load X509 certificate %s: %s" % (filename, err))
1126

    
1127
  (errcode, msg) = \
1128
    utils.VerifyX509Certificate(cert, constants.SSL_CERT_EXPIRATION_WARN,
1129
                                constants.SSL_CERT_EXPIRATION_ERROR)
1130

    
1131
  if msg:
1132
    fnamemsg = "While verifying %s: %s" % (filename, msg)
1133
  else:
1134
    fnamemsg = None
1135

    
1136
  if errcode is None:
1137
    return (None, fnamemsg)
1138
  elif errcode == utils.CERT_WARNING:
1139
    return (LUVerifyCluster.ETYPE_WARNING, fnamemsg)
1140
  elif errcode == utils.CERT_ERROR:
1141
    return (LUVerifyCluster.ETYPE_ERROR, fnamemsg)
1142

    
1143
  raise errors.ProgrammerError("Unhandled certificate error code %r" % errcode)
1144

    
1145

    
1146
class LUVerifyCluster(LogicalUnit):
1147
  """Verifies the cluster status.
1148

1149
  """
1150
  HPATH = "cluster-verify"
1151
  HTYPE = constants.HTYPE_CLUSTER
1152
  _OP_PARAMS = [
1153
    ("skip_checks", ht.EmptyList,
1154
     ht.TListOf(ht.TElemOf(constants.VERIFY_OPTIONAL_CHECKS))),
1155
    ("verbose", False, ht.TBool),
1156
    ("error_codes", False, ht.TBool),
1157
    ("debug_simulate_errors", False, ht.TBool),
1158
    ]
1159
  REQ_BGL = False
1160

    
1161
  TCLUSTER = "cluster"
1162
  TNODE = "node"
1163
  TINSTANCE = "instance"
1164

    
1165
  ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
1166
  ECLUSTERCERT = (TCLUSTER, "ECLUSTERCERT")
1167
  EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
1168
  EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
1169
  EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
1170
  EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
1171
  EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
1172
  EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE")
1173
  ENODEDRBD = (TNODE, "ENODEDRBD")
1174
  ENODEDRBDHELPER = (TNODE, "ENODEDRBDHELPER")
1175
  ENODEFILECHECK = (TNODE, "ENODEFILECHECK")
1176
  ENODEHOOKS = (TNODE, "ENODEHOOKS")
1177
  ENODEHV = (TNODE, "ENODEHV")
1178
  ENODELVM = (TNODE, "ENODELVM")
1179
  ENODEN1 = (TNODE, "ENODEN1")
1180
  ENODENET = (TNODE, "ENODENET")
1181
  ENODEOS = (TNODE, "ENODEOS")
1182
  ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE")
1183
  ENODEORPHANLV = (TNODE, "ENODEORPHANLV")
1184
  ENODERPC = (TNODE, "ENODERPC")
1185
  ENODESSH = (TNODE, "ENODESSH")
1186
  ENODEVERSION = (TNODE, "ENODEVERSION")
1187
  ENODESETUP = (TNODE, "ENODESETUP")
1188
  ENODETIME = (TNODE, "ENODETIME")
1189

    
1190
  ETYPE_FIELD = "code"
1191
  ETYPE_ERROR = "ERROR"
1192
  ETYPE_WARNING = "WARNING"
1193

    
1194
  class NodeImage(object):
1195
    """A class representing the logical and physical status of a node.
1196

1197
    @type name: string
1198
    @ivar name: the node name to which this object refers
1199
    @ivar volumes: a structure as returned from
1200
        L{ganeti.backend.GetVolumeList} (runtime)
1201
    @ivar instances: a list of running instances (runtime)
1202
    @ivar pinst: list of configured primary instances (config)
1203
    @ivar sinst: list of configured secondary instances (config)
1204
    @ivar sbp: diction of {secondary-node: list of instances} of all peers
1205
        of this node (config)
1206
    @ivar mfree: free memory, as reported by hypervisor (runtime)
1207
    @ivar dfree: free disk, as reported by the node (runtime)
1208
    @ivar offline: the offline status (config)
1209
    @type rpc_fail: boolean
1210
    @ivar rpc_fail: whether the RPC verify call was successfull (overall,
1211
        not whether the individual keys were correct) (runtime)
1212
    @type lvm_fail: boolean
1213
    @ivar lvm_fail: whether the RPC call didn't return valid LVM data
1214
    @type hyp_fail: boolean
1215
    @ivar hyp_fail: whether the RPC call didn't return the instance list
1216
    @type ghost: boolean
1217
    @ivar ghost: whether this is a known node or not (config)
1218
    @type os_fail: boolean
1219
    @ivar os_fail: whether the RPC call didn't return valid OS data
1220
    @type oslist: list
1221
    @ivar oslist: list of OSes as diagnosed by DiagnoseOS
1222

1223
    """
1224
    def __init__(self, offline=False, name=None):
1225
      self.name = name
1226
      self.volumes = {}
1227
      self.instances = []
1228
      self.pinst = []
1229
      self.sinst = []
1230
      self.sbp = {}
1231
      self.mfree = 0
1232
      self.dfree = 0
1233
      self.offline = offline
1234
      self.rpc_fail = False
1235
      self.lvm_fail = False
1236
      self.hyp_fail = False
1237
      self.ghost = False
1238
      self.os_fail = False
1239
      self.oslist = {}
1240

    
1241
  def ExpandNames(self):
1242
    self.needed_locks = {
1243
      locking.LEVEL_NODE: locking.ALL_SET,
1244
      locking.LEVEL_INSTANCE: locking.ALL_SET,
1245
    }
1246
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
1247

    
1248
  def _Error(self, ecode, item, msg, *args, **kwargs):
1249
    """Format an error message.
1250

1251
    Based on the opcode's error_codes parameter, either format a
1252
    parseable error code, or a simpler error string.
1253

1254
    This must be called only from Exec and functions called from Exec.
1255

1256
    """
1257
    ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
1258
    itype, etxt = ecode
1259
    # first complete the msg
1260
    if args:
1261
      msg = msg % args
1262
    # then format the whole message
1263
    if self.op.error_codes:
1264
      msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
1265
    else:
1266
      if item:
1267
        item = " " + item
1268
      else:
1269
        item = ""
1270
      msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
1271
    # and finally report it via the feedback_fn
1272
    self._feedback_fn("  - %s" % msg)
1273

    
1274
  def _ErrorIf(self, cond, *args, **kwargs):
1275
    """Log an error message if the passed condition is True.
1276

1277
    """
1278
    cond = bool(cond) or self.op.debug_simulate_errors
1279
    if cond:
1280
      self._Error(*args, **kwargs)
1281
    # do not mark the operation as failed for WARN cases only
1282
    if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
1283
      self.bad = self.bad or cond
1284

    
1285
  def _VerifyNode(self, ninfo, nresult):
1286
    """Perform some basic validation on data returned from a node.
1287

1288
      - check the result data structure is well formed and has all the
1289
        mandatory fields
1290
      - check ganeti version
1291

1292
    @type ninfo: L{objects.Node}
1293
    @param ninfo: the node to check
1294
    @param nresult: the results from the node
1295
    @rtype: boolean
1296
    @return: whether overall this call was successful (and we can expect
1297
         reasonable values in the respose)
1298

1299
    """
1300
    node = ninfo.name
1301
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1302

    
1303
    # main result, nresult should be a non-empty dict
1304
    test = not nresult or not isinstance(nresult, dict)
1305
    _ErrorIf(test, self.ENODERPC, node,
1306
                  "unable to verify node: no data returned")
1307
    if test:
1308
      return False
1309

    
1310
    # compares ganeti version
1311
    local_version = constants.PROTOCOL_VERSION
1312
    remote_version = nresult.get("version", None)
1313
    test = not (remote_version and
1314
                isinstance(remote_version, (list, tuple)) and
1315
                len(remote_version) == 2)
1316
    _ErrorIf(test, self.ENODERPC, node,
1317
             "connection to node returned invalid data")
1318
    if test:
1319
      return False
1320

    
1321
    test = local_version != remote_version[0]
1322
    _ErrorIf(test, self.ENODEVERSION, node,
1323
             "incompatible protocol versions: master %s,"
1324
             " node %s", local_version, remote_version[0])
1325
    if test:
1326
      return False
1327

    
1328
    # node seems compatible, we can actually try to look into its results
1329

    
1330
    # full package version
1331
    self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
1332
                  self.ENODEVERSION, node,
1333
                  "software version mismatch: master %s, node %s",
1334
                  constants.RELEASE_VERSION, remote_version[1],
1335
                  code=self.ETYPE_WARNING)
1336

    
1337
    hyp_result = nresult.get(constants.NV_HYPERVISOR, None)
1338
    if isinstance(hyp_result, dict):
1339
      for hv_name, hv_result in hyp_result.iteritems():
1340
        test = hv_result is not None
1341
        _ErrorIf(test, self.ENODEHV, node,
1342
                 "hypervisor %s verify failure: '%s'", hv_name, hv_result)
1343

    
1344

    
1345
    test = nresult.get(constants.NV_NODESETUP,
1346
                           ["Missing NODESETUP results"])
1347
    _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s",
1348
             "; ".join(test))
1349

    
1350
    return True
1351

    
1352
  def _VerifyNodeTime(self, ninfo, nresult,
1353
                      nvinfo_starttime, nvinfo_endtime):
1354
    """Check the node time.
1355

1356
    @type ninfo: L{objects.Node}
1357
    @param ninfo: the node to check
1358
    @param nresult: the remote results for the node
1359
    @param nvinfo_starttime: the start time of the RPC call
1360
    @param nvinfo_endtime: the end time of the RPC call
1361

1362
    """
1363
    node = ninfo.name
1364
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1365

    
1366
    ntime = nresult.get(constants.NV_TIME, None)
1367
    try:
1368
      ntime_merged = utils.MergeTime(ntime)
1369
    except (ValueError, TypeError):
1370
      _ErrorIf(True, self.ENODETIME, node, "Node returned invalid time")
1371
      return
1372

    
1373
    if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
1374
      ntime_diff = "%.01fs" % abs(nvinfo_starttime - ntime_merged)
1375
    elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
1376
      ntime_diff = "%.01fs" % abs(ntime_merged - nvinfo_endtime)
1377
    else:
1378
      ntime_diff = None
1379

    
1380
    _ErrorIf(ntime_diff is not None, self.ENODETIME, node,
1381
             "Node time diverges by at least %s from master node time",
1382
             ntime_diff)
1383

    
1384
  def _VerifyNodeLVM(self, ninfo, nresult, vg_name):
1385
    """Check the node time.
1386

1387
    @type ninfo: L{objects.Node}
1388
    @param ninfo: the node to check
1389
    @param nresult: the remote results for the node
1390
    @param vg_name: the configured VG name
1391

1392
    """
1393
    if vg_name is None:
1394
      return
1395

    
1396
    node = ninfo.name
1397
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1398

    
1399
    # checks vg existence and size > 20G
1400
    vglist = nresult.get(constants.NV_VGLIST, None)
1401
    test = not vglist
1402
    _ErrorIf(test, self.ENODELVM, node, "unable to check volume groups")
1403
    if not test:
1404
      vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
1405
                                            constants.MIN_VG_SIZE)
1406
      _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
1407

    
1408
    # check pv names
1409
    pvlist = nresult.get(constants.NV_PVLIST, None)
1410
    test = pvlist is None
1411
    _ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node")
1412
    if not test:
1413
      # check that ':' is not present in PV names, since it's a
1414
      # special character for lvcreate (denotes the range of PEs to
1415
      # use on the PV)
1416
      for _, pvname, owner_vg in pvlist:
1417
        test = ":" in pvname
1418
        _ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV"
1419
                 " '%s' of VG '%s'", pvname, owner_vg)
1420

    
1421
  def _VerifyNodeNetwork(self, ninfo, nresult):
1422
    """Check the node time.
1423

1424
    @type ninfo: L{objects.Node}
1425
    @param ninfo: the node to check
1426
    @param nresult: the remote results for the node
1427

1428
    """
1429
    node = ninfo.name
1430
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1431

    
1432
    test = constants.NV_NODELIST not in nresult
1433
    _ErrorIf(test, self.ENODESSH, node,
1434
             "node hasn't returned node ssh connectivity data")
1435
    if not test:
1436
      if nresult[constants.NV_NODELIST]:
1437
        for a_node, a_msg in nresult[constants.NV_NODELIST].items():
1438
          _ErrorIf(True, self.ENODESSH, node,
1439
                   "ssh communication with node '%s': %s", a_node, a_msg)
1440

    
1441
    test = constants.NV_NODENETTEST not in nresult
1442
    _ErrorIf(test, self.ENODENET, node,
1443
             "node hasn't returned node tcp connectivity data")
1444
    if not test:
1445
      if nresult[constants.NV_NODENETTEST]:
1446
        nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys())
1447
        for anode in nlist:
1448
          _ErrorIf(True, self.ENODENET, node,
1449
                   "tcp communication with node '%s': %s",
1450
                   anode, nresult[constants.NV_NODENETTEST][anode])
1451

    
1452
    test = constants.NV_MASTERIP not in nresult
1453
    _ErrorIf(test, self.ENODENET, node,
1454
             "node hasn't returned node master IP reachability data")
1455
    if not test:
1456
      if not nresult[constants.NV_MASTERIP]:
1457
        if node == self.master_node:
1458
          msg = "the master node cannot reach the master IP (not configured?)"
1459
        else:
1460
          msg = "cannot reach the master IP"
1461
        _ErrorIf(True, self.ENODENET, node, msg)
1462

    
1463

    
1464
  def _VerifyInstance(self, instance, instanceconfig, node_image):
1465
    """Verify an instance.
1466

1467
    This function checks to see if the required block devices are
1468
    available on the instance's node.
1469

1470
    """
1471
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1472
    node_current = instanceconfig.primary_node
1473

    
1474
    node_vol_should = {}
1475
    instanceconfig.MapLVsByNode(node_vol_should)
1476

    
1477
    for node in node_vol_should:
1478
      n_img = node_image[node]
1479
      if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1480
        # ignore missing volumes on offline or broken nodes
1481
        continue
1482
      for volume in node_vol_should[node]:
1483
        test = volume not in n_img.volumes
1484
        _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
1485
                 "volume %s missing on node %s", volume, node)
1486

    
1487
    if instanceconfig.admin_up:
1488
      pri_img = node_image[node_current]
1489
      test = instance not in pri_img.instances and not pri_img.offline
1490
      _ErrorIf(test, self.EINSTANCEDOWN, instance,
1491
               "instance not running on its primary node %s",
1492
               node_current)
1493

    
1494
    for node, n_img in node_image.items():
1495
      if (not node == node_current):
1496
        test = instance in n_img.instances
1497
        _ErrorIf(test, self.EINSTANCEWRONGNODE, instance,
1498
                 "instance should not run on node %s", node)
1499

    
1500
  def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):
1501
    """Verify if there are any unknown volumes in the cluster.
1502

1503
    The .os, .swap and backup volumes are ignored. All other volumes are
1504
    reported as unknown.
1505

1506
    @type reserved: L{ganeti.utils.FieldSet}
1507
    @param reserved: a FieldSet of reserved volume names
1508

1509
    """
1510
    for node, n_img in node_image.items():
1511
      if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1512
        # skip non-healthy nodes
1513
        continue
1514
      for volume in n_img.volumes:
1515
        test = ((node not in node_vol_should or
1516
                volume not in node_vol_should[node]) and
1517
                not reserved.Matches(volume))
1518
        self._ErrorIf(test, self.ENODEORPHANLV, node,
1519
                      "volume %s is unknown", volume)
1520

    
1521
  def _VerifyOrphanInstances(self, instancelist, node_image):
1522
    """Verify the list of running instances.
1523

1524
    This checks what instances are running but unknown to the cluster.
1525

1526
    """
1527
    for node, n_img in node_image.items():
1528
      for o_inst in n_img.instances:
1529
        test = o_inst not in instancelist
1530
        self._ErrorIf(test, self.ENODEORPHANINSTANCE, node,
1531
                      "instance %s on node %s should not exist", o_inst, node)
1532

    
1533
  def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
1534
    """Verify N+1 Memory Resilience.
1535

1536
    Check that if one single node dies we can still start all the
1537
    instances it was primary for.
1538

1539
    """
1540
    for node, n_img in node_image.items():
1541
      # This code checks that every node which is now listed as
1542
      # secondary has enough memory to host all instances it is
1543
      # supposed to should a single other node in the cluster fail.
1544
      # FIXME: not ready for failover to an arbitrary node
1545
      # FIXME: does not support file-backed instances
1546
      # WARNING: we currently take into account down instances as well
1547
      # as up ones, considering that even if they're down someone
1548
      # might want to start them even in the event of a node failure.
1549
      for prinode, instances in n_img.sbp.items():
1550
        needed_mem = 0
1551
        for instance in instances:
1552
          bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
1553
          if bep[constants.BE_AUTO_BALANCE]:
1554
            needed_mem += bep[constants.BE_MEMORY]
1555
        test = n_img.mfree < needed_mem
1556
        self._ErrorIf(test, self.ENODEN1, node,
1557
                      "not enough memory on to accommodate"
1558
                      " failovers should peer node %s fail", prinode)
1559

    
1560
  def _VerifyNodeFiles(self, ninfo, nresult, file_list, local_cksum,
1561
                       master_files):
1562
    """Verifies and computes the node required file checksums.
1563

1564
    @type ninfo: L{objects.Node}
1565
    @param ninfo: the node to check
1566
    @param nresult: the remote results for the node
1567
    @param file_list: required list of files
1568
    @param local_cksum: dictionary of local files and their checksums
1569
    @param master_files: list of files that only masters should have
1570

1571
    """
1572
    node = ninfo.name
1573
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1574

    
1575
    remote_cksum = nresult.get(constants.NV_FILELIST, None)
1576
    test = not isinstance(remote_cksum, dict)
1577
    _ErrorIf(test, self.ENODEFILECHECK, node,
1578
             "node hasn't returned file checksum data")
1579
    if test:
1580
      return
1581

    
1582
    for file_name in file_list:
1583
      node_is_mc = ninfo.master_candidate
1584
      must_have = (file_name not in master_files) or node_is_mc
1585
      # missing
1586
      test1 = file_name not in remote_cksum
1587
      # invalid checksum
1588
      test2 = not test1 and remote_cksum[file_name] != local_cksum[file_name]
1589
      # existing and good
1590
      test3 = not test1 and remote_cksum[file_name] == local_cksum[file_name]
1591
      _ErrorIf(test1 and must_have, self.ENODEFILECHECK, node,
1592
               "file '%s' missing", file_name)
1593
      _ErrorIf(test2 and must_have, self.ENODEFILECHECK, node,
1594
               "file '%s' has wrong checksum", file_name)
1595
      # not candidate and this is not a must-have file
1596
      _ErrorIf(test2 and not must_have, self.ENODEFILECHECK, node,
1597
               "file '%s' should not exist on non master"
1598
               " candidates (and the file is outdated)", file_name)
1599
      # all good, except non-master/non-must have combination
1600
      _ErrorIf(test3 and not must_have, self.ENODEFILECHECK, node,
1601
               "file '%s' should not exist"
1602
               " on non master candidates", file_name)
1603

    
1604
  def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper,
1605
                      drbd_map):
1606
    """Verifies and the node DRBD status.
1607

1608
    @type ninfo: L{objects.Node}
1609
    @param ninfo: the node to check
1610
    @param nresult: the remote results for the node
1611
    @param instanceinfo: the dict of instances
1612
    @param drbd_helper: the configured DRBD usermode helper
1613
    @param drbd_map: the DRBD map as returned by
1614
        L{ganeti.config.ConfigWriter.ComputeDRBDMap}
1615

1616
    """
1617
    node = ninfo.name
1618
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1619

    
1620
    if drbd_helper:
1621
      helper_result = nresult.get(constants.NV_DRBDHELPER, None)
1622
      test = (helper_result == None)
1623
      _ErrorIf(test, self.ENODEDRBDHELPER, node,
1624
               "no drbd usermode helper returned")
1625
      if helper_result:
1626
        status, payload = helper_result
1627
        test = not status
1628
        _ErrorIf(test, self.ENODEDRBDHELPER, node,
1629
                 "drbd usermode helper check unsuccessful: %s", payload)
1630
        test = status and (payload != drbd_helper)
1631
        _ErrorIf(test, self.ENODEDRBDHELPER, node,
1632
                 "wrong drbd usermode helper: %s", payload)
1633

    
1634
    # compute the DRBD minors
1635
    node_drbd = {}
1636
    for minor, instance in drbd_map[node].items():
1637
      test = instance not in instanceinfo
1638
      _ErrorIf(test, self.ECLUSTERCFG, None,
1639
               "ghost instance '%s' in temporary DRBD map", instance)
1640
        # ghost instance should not be running, but otherwise we
1641
        # don't give double warnings (both ghost instance and
1642
        # unallocated minor in use)
1643
      if test:
1644
        node_drbd[minor] = (instance, False)
1645
      else:
1646
        instance = instanceinfo[instance]
1647
        node_drbd[minor] = (instance.name, instance.admin_up)
1648

    
1649
    # and now check them
1650
    used_minors = nresult.get(constants.NV_DRBDLIST, [])
1651
    test = not isinstance(used_minors, (tuple, list))
1652
    _ErrorIf(test, self.ENODEDRBD, node,
1653
             "cannot parse drbd status file: %s", str(used_minors))
1654
    if test:
1655
      # we cannot check drbd status
1656
      return
1657

    
1658
    for minor, (iname, must_exist) in node_drbd.items():
1659
      test = minor not in used_minors and must_exist
1660
      _ErrorIf(test, self.ENODEDRBD, node,
1661
               "drbd minor %d of instance %s is not active", minor, iname)
1662
    for minor in used_minors:
1663
      test = minor not in node_drbd
1664
      _ErrorIf(test, self.ENODEDRBD, node,
1665
               "unallocated drbd minor %d is in use", minor)
1666

    
1667
  def _UpdateNodeOS(self, ninfo, nresult, nimg):
1668
    """Builds the node OS structures.
1669

1670
    @type ninfo: L{objects.Node}
1671
    @param ninfo: the node to check
1672
    @param nresult: the remote results for the node
1673
    @param nimg: the node image object
1674

1675
    """
1676
    node = ninfo.name
1677
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1678

    
1679
    remote_os = nresult.get(constants.NV_OSLIST, None)
1680
    test = (not isinstance(remote_os, list) or
1681
            not compat.all(isinstance(v, list) and len(v) == 7
1682
                           for v in remote_os))
1683

    
1684
    _ErrorIf(test, self.ENODEOS, node,
1685
             "node hasn't returned valid OS data")
1686

    
1687
    nimg.os_fail = test
1688

    
1689
    if test:
1690
      return
1691

    
1692
    os_dict = {}
1693

    
1694
    for (name, os_path, status, diagnose,
1695
         variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
1696

    
1697
      if name not in os_dict:
1698
        os_dict[name] = []
1699

    
1700
      # parameters is a list of lists instead of list of tuples due to
1701
      # JSON lacking a real tuple type, fix it:
1702
      parameters = [tuple(v) for v in parameters]
1703
      os_dict[name].append((os_path, status, diagnose,
1704
                            set(variants), set(parameters), set(api_ver)))
1705

    
1706
    nimg.oslist = os_dict
1707

    
1708
  def _VerifyNodeOS(self, ninfo, nimg, base):
1709
    """Verifies the node OS list.
1710

1711
    @type ninfo: L{objects.Node}
1712
    @param ninfo: the node to check
1713
    @param nimg: the node image object
1714
    @param base: the 'template' node we match against (e.g. from the master)
1715

1716
    """
1717
    node = ninfo.name
1718
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1719

    
1720
    assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
1721

    
1722
    for os_name, os_data in nimg.oslist.items():
1723
      assert os_data, "Empty OS status for OS %s?!" % os_name
1724
      f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
1725
      _ErrorIf(not f_status, self.ENODEOS, node,
1726
               "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
1727
      _ErrorIf(len(os_data) > 1, self.ENODEOS, node,
1728
               "OS '%s' has multiple entries (first one shadows the rest): %s",
1729
               os_name, utils.CommaJoin([v[0] for v in os_data]))
1730
      # this will catched in backend too
1731
      _ErrorIf(compat.any(v >= constants.OS_API_V15 for v in f_api)
1732
               and not f_var, self.ENODEOS, node,
1733
               "OS %s with API at least %d does not declare any variant",
1734
               os_name, constants.OS_API_V15)
1735
      # comparisons with the 'base' image
1736
      test = os_name not in base.oslist
1737
      _ErrorIf(test, self.ENODEOS, node,
1738
               "Extra OS %s not present on reference node (%s)",
1739
               os_name, base.name)
1740
      if test:
1741
        continue
1742
      assert base.oslist[os_name], "Base node has empty OS status?"
1743
      _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
1744
      if not b_status:
1745
        # base OS is invalid, skipping
1746
        continue
1747
      for kind, a, b in [("API version", f_api, b_api),
1748
                         ("variants list", f_var, b_var),
1749
                         ("parameters", f_param, b_param)]:
1750
        _ErrorIf(a != b, self.ENODEOS, node,
1751
                 "OS %s %s differs from reference node %s: %s vs. %s",
1752
                 kind, os_name, base.name,
1753
                 utils.CommaJoin(a), utils.CommaJoin(b))
1754

    
1755
    # check any missing OSes
1756
    missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
1757
    _ErrorIf(missing, self.ENODEOS, node,
1758
             "OSes present on reference node %s but missing on this node: %s",
1759
             base.name, utils.CommaJoin(missing))
1760

    
1761
  def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
1762
    """Verifies and updates the node volume data.
1763

1764
    This function will update a L{NodeImage}'s internal structures
1765
    with data from the remote call.
1766

1767
    @type ninfo: L{objects.Node}
1768
    @param ninfo: the node to check
1769
    @param nresult: the remote results for the node
1770
    @param nimg: the node image object
1771
    @param vg_name: the configured VG name
1772

1773
    """
1774
    node = ninfo.name
1775
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1776

    
1777
    nimg.lvm_fail = True
1778
    lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
1779
    if vg_name is None:
1780
      pass
1781
    elif isinstance(lvdata, basestring):
1782
      _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
1783
               utils.SafeEncode(lvdata))
1784
    elif not isinstance(lvdata, dict):
1785
      _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
1786
    else:
1787
      nimg.volumes = lvdata
1788
      nimg.lvm_fail = False
1789

    
1790
  def _UpdateNodeInstances(self, ninfo, nresult, nimg):
1791
    """Verifies and updates the node instance list.
1792

1793
    If the listing was successful, then updates this node's instance
1794
    list. Otherwise, it marks the RPC call as failed for the instance
1795
    list key.
1796

1797
    @type ninfo: L{objects.Node}
1798
    @param ninfo: the node to check
1799
    @param nresult: the remote results for the node
1800
    @param nimg: the node image object
1801

1802
    """
1803
    idata = nresult.get(constants.NV_INSTANCELIST, None)
1804
    test = not isinstance(idata, list)
1805
    self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed"
1806
                  " (instancelist): %s", utils.SafeEncode(str(idata)))
1807
    if test:
1808
      nimg.hyp_fail = True
1809
    else:
1810
      nimg.instances = idata
1811

    
1812
  def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
1813
    """Verifies and computes a node information map
1814

1815
    @type ninfo: L{objects.Node}
1816
    @param ninfo: the node to check
1817
    @param nresult: the remote results for the node
1818
    @param nimg: the node image object
1819
    @param vg_name: the configured VG name
1820

1821
    """
1822
    node = ninfo.name
1823
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1824

    
1825
    # try to read free memory (from the hypervisor)
1826
    hv_info = nresult.get(constants.NV_HVINFO, None)
1827
    test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
1828
    _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
1829
    if not test:
1830
      try:
1831
        nimg.mfree = int(hv_info["memory_free"])
1832
      except (ValueError, TypeError):
1833
        _ErrorIf(True, self.ENODERPC, node,
1834
                 "node returned invalid nodeinfo, check hypervisor")
1835

    
1836
    # FIXME: devise a free space model for file based instances as well
1837
    if vg_name is not None:
1838
      test = (constants.NV_VGLIST not in nresult or
1839
              vg_name not in nresult[constants.NV_VGLIST])
1840
      _ErrorIf(test, self.ENODELVM, node,
1841
               "node didn't return data for the volume group '%s'"
1842
               " - it is either missing or broken", vg_name)
1843
      if not test:
1844
        try:
1845
          nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
1846
        except (ValueError, TypeError):
1847
          _ErrorIf(True, self.ENODERPC, node,
1848
                   "node returned invalid LVM info, check LVM status")
1849

    
1850
  def BuildHooksEnv(self):
1851
    """Build hooks env.
1852

1853
    Cluster-Verify hooks just ran in the post phase and their failure makes
1854
    the output be logged in the verify output and the verification to fail.
1855

1856
    """
1857
    all_nodes = self.cfg.GetNodeList()
1858
    env = {
1859
      "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
1860
      }
1861
    for node in self.cfg.GetAllNodesInfo().values():
1862
      env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
1863

    
1864
    return env, [], all_nodes
1865

    
1866
  def Exec(self, feedback_fn):
1867
    """Verify integrity of cluster, performing various test on nodes.
1868

1869
    """
1870
    self.bad = False
1871
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1872
    verbose = self.op.verbose
1873
    self._feedback_fn = feedback_fn
1874
    feedback_fn("* Verifying global settings")
1875
    for msg in self.cfg.VerifyConfig():
1876
      _ErrorIf(True, self.ECLUSTERCFG, None, msg)
1877

    
1878
    # Check the cluster certificates
1879
    for cert_filename in constants.ALL_CERT_FILES:
1880
      (errcode, msg) = _VerifyCertificate(cert_filename)
1881
      _ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
1882

    
1883
    vg_name = self.cfg.GetVGName()
1884
    drbd_helper = self.cfg.GetDRBDHelper()
1885
    hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
1886
    cluster = self.cfg.GetClusterInfo()
1887
    nodelist = utils.NiceSort(self.cfg.GetNodeList())
1888
    nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
1889
    instancelist = utils.NiceSort(self.cfg.GetInstanceList())
1890
    instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
1891
                        for iname in instancelist)
1892
    i_non_redundant = [] # Non redundant instances
1893
    i_non_a_balanced = [] # Non auto-balanced instances
1894
    n_offline = 0 # Count of offline nodes
1895
    n_drained = 0 # Count of nodes being drained
1896
    node_vol_should = {}
1897

    
1898
    # FIXME: verify OS list
1899
    # do local checksums
1900
    master_files = [constants.CLUSTER_CONF_FILE]
1901
    master_node = self.master_node = self.cfg.GetMasterNode()
1902
    master_ip = self.cfg.GetMasterIP()
1903

    
1904
    file_names = ssconf.SimpleStore().GetFileList()
1905
    file_names.extend(constants.ALL_CERT_FILES)
1906
    file_names.extend(master_files)
1907
    if cluster.modify_etc_hosts:
1908
      file_names.append(constants.ETC_HOSTS)
1909

    
1910
    local_checksums = utils.FingerprintFiles(file_names)
1911

    
1912
    feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
1913
    node_verify_param = {
1914
      constants.NV_FILELIST: file_names,
1915
      constants.NV_NODELIST: [node.name for node in nodeinfo
1916
                              if not node.offline],
1917
      constants.NV_HYPERVISOR: hypervisors,
1918
      constants.NV_NODENETTEST: [(node.name, node.primary_ip,
1919
                                  node.secondary_ip) for node in nodeinfo
1920
                                 if not node.offline],
1921
      constants.NV_INSTANCELIST: hypervisors,
1922
      constants.NV_VERSION: None,
1923
      constants.NV_HVINFO: self.cfg.GetHypervisorType(),
1924
      constants.NV_NODESETUP: None,
1925
      constants.NV_TIME: None,
1926
      constants.NV_MASTERIP: (master_node, master_ip),
1927
      constants.NV_OSLIST: None,
1928
      }
1929

    
1930
    if vg_name is not None:
1931
      node_verify_param[constants.NV_VGLIST] = None
1932
      node_verify_param[constants.NV_LVLIST] = vg_name
1933
      node_verify_param[constants.NV_PVLIST] = [vg_name]
1934
      node_verify_param[constants.NV_DRBDLIST] = None
1935

    
1936
    if drbd_helper:
1937
      node_verify_param[constants.NV_DRBDHELPER] = drbd_helper
1938

    
1939
    # Build our expected cluster state
1940
    node_image = dict((node.name, self.NodeImage(offline=node.offline,
1941
                                                 name=node.name))
1942
                      for node in nodeinfo)
1943

    
1944
    for instance in instancelist:
1945
      inst_config = instanceinfo[instance]
1946

    
1947
      for nname in inst_config.all_nodes:
1948
        if nname not in node_image:
1949
          # ghost node
1950
          gnode = self.NodeImage(name=nname)
1951
          gnode.ghost = True
1952
          node_image[nname] = gnode
1953

    
1954
      inst_config.MapLVsByNode(node_vol_should)
1955

    
1956
      pnode = inst_config.primary_node
1957
      node_image[pnode].pinst.append(instance)
1958

    
1959
      for snode in inst_config.secondary_nodes:
1960
        nimg = node_image[snode]
1961
        nimg.sinst.append(instance)
1962
        if pnode not in nimg.sbp:
1963
          nimg.sbp[pnode] = []
1964
        nimg.sbp[pnode].append(instance)
1965

    
1966
    # At this point, we have the in-memory data structures complete,
1967
    # except for the runtime information, which we'll gather next
1968

    
1969
    # Due to the way our RPC system works, exact response times cannot be
1970
    # guaranteed (e.g. a broken node could run into a timeout). By keeping the
1971
    # time before and after executing the request, we can at least have a time
1972
    # window.
1973
    nvinfo_starttime = time.time()
1974
    all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
1975
                                           self.cfg.GetClusterName())
1976
    nvinfo_endtime = time.time()
1977

    
1978
    all_drbd_map = self.cfg.ComputeDRBDMap()
1979

    
1980
    feedback_fn("* Verifying node status")
1981

    
1982
    refos_img = None
1983

    
1984
    for node_i in nodeinfo:
1985
      node = node_i.name
1986
      nimg = node_image[node]
1987

    
1988
      if node_i.offline:
1989
        if verbose:
1990
          feedback_fn("* Skipping offline node %s" % (node,))
1991
        n_offline += 1
1992
        continue
1993

    
1994
      if node == master_node:
1995
        ntype = "master"
1996
      elif node_i.master_candidate:
1997
        ntype = "master candidate"
1998
      elif node_i.drained:
1999
        ntype = "drained"
2000
        n_drained += 1
2001
      else:
2002
        ntype = "regular"
2003
      if verbose:
2004
        feedback_fn("* Verifying node %s (%s)" % (node, ntype))
2005

    
2006
      msg = all_nvinfo[node].fail_msg
2007
      _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
2008
      if msg:
2009
        nimg.rpc_fail = True
2010
        continue
2011

    
2012
      nresult = all_nvinfo[node].payload
2013

    
2014
      nimg.call_ok = self._VerifyNode(node_i, nresult)
2015
      self._VerifyNodeNetwork(node_i, nresult)
2016
      self._VerifyNodeLVM(node_i, nresult, vg_name)
2017
      self._VerifyNodeFiles(node_i, nresult, file_names, local_checksums,
2018
                            master_files)
2019
      self._VerifyNodeDrbd(node_i, nresult, instanceinfo, drbd_helper,
2020
                           all_drbd_map)
2021
      self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
2022

    
2023
      self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
2024
      self._UpdateNodeInstances(node_i, nresult, nimg)
2025
      self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
2026
      self._UpdateNodeOS(node_i, nresult, nimg)
2027
      if not nimg.os_fail:
2028
        if refos_img is None:
2029
          refos_img = nimg
2030
        self._VerifyNodeOS(node_i, nimg, refos_img)
2031

    
2032
    feedback_fn("* Verifying instance status")
2033
    for instance in instancelist:
2034
      if verbose:
2035
        feedback_fn("* Verifying instance %s" % instance)
2036
      inst_config = instanceinfo[instance]
2037
      self._VerifyInstance(instance, inst_config, node_image)
2038
      inst_nodes_offline = []
2039

    
2040
      pnode = inst_config.primary_node
2041
      pnode_img = node_image[pnode]
2042
      _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
2043
               self.ENODERPC, pnode, "instance %s, connection to"
2044
               " primary node failed", instance)
2045

    
2046
      if pnode_img.offline:
2047
        inst_nodes_offline.append(pnode)
2048

    
2049
      # If the instance is non-redundant we cannot survive losing its primary
2050
      # node, so we are not N+1 compliant. On the other hand we have no disk
2051
      # templates with more than one secondary so that situation is not well
2052
      # supported either.
2053
      # FIXME: does not support file-backed instances
2054
      if not inst_config.secondary_nodes:
2055
        i_non_redundant.append(instance)
2056
      _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT,
2057
               instance, "instance has multiple secondary nodes: %s",
2058
               utils.CommaJoin(inst_config.secondary_nodes),
2059
               code=self.ETYPE_WARNING)
2060

    
2061
      if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
2062
        i_non_a_balanced.append(instance)
2063

    
2064
      for snode in inst_config.secondary_nodes:
2065
        s_img = node_image[snode]
2066
        _ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode,
2067
                 "instance %s, connection to secondary node failed", instance)
2068

    
2069
        if s_img.offline:
2070
          inst_nodes_offline.append(snode)
2071

    
2072
      # warn that the instance lives on offline nodes
2073
      _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
2074
               "instance lives on offline node(s) %s",
2075
               utils.CommaJoin(inst_nodes_offline))
2076
      # ... or ghost nodes
2077
      for node in inst_config.all_nodes:
2078
        _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
2079
                 "instance lives on ghost node %s", node)
2080

    
2081
    feedback_fn("* Verifying orphan volumes")
2082
    reserved = utils.FieldSet(*cluster.reserved_lvs)
2083
    self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
2084

    
2085
    feedback_fn("* Verifying orphan instances")
2086
    self._VerifyOrphanInstances(instancelist, node_image)
2087

    
2088
    if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
2089
      feedback_fn("* Verifying N+1 Memory redundancy")
2090
      self._VerifyNPlusOneMemory(node_image, instanceinfo)
2091

    
2092
    feedback_fn("* Other Notes")
2093
    if i_non_redundant:
2094
      feedback_fn("  - NOTICE: %d non-redundant instance(s) found."
2095
                  % len(i_non_redundant))
2096

    
2097
    if i_non_a_balanced:
2098
      feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
2099
                  % len(i_non_a_balanced))
2100

    
2101
    if n_offline:
2102
      feedback_fn("  - NOTICE: %d offline node(s) found." % n_offline)
2103

    
2104
    if n_drained:
2105
      feedback_fn("  - NOTICE: %d drained node(s) found." % n_drained)
2106

    
2107
    return not self.bad
2108

    
2109
  def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
2110
    """Analyze the post-hooks' result
2111

2112
    This method analyses the hook result, handles it, and sends some
2113
    nicely-formatted feedback back to the user.
2114

2115
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
2116
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
2117
    @param hooks_results: the results of the multi-node hooks rpc call
2118
    @param feedback_fn: function used send feedback back to the caller
2119
    @param lu_result: previous Exec result
2120
    @return: the new Exec result, based on the previous result
2121
        and hook results
2122

2123
    """
2124
    # We only really run POST phase hooks, and are only interested in
2125
    # their results
2126
    if phase == constants.HOOKS_PHASE_POST:
2127
      # Used to change hooks' output to proper indentation
2128
      indent_re = re.compile('^', re.M)
2129
      feedback_fn("* Hooks Results")
2130
      assert hooks_results, "invalid result from hooks"
2131

    
2132
      for node_name in hooks_results:
2133
        res = hooks_results[node_name]
2134
        msg = res.fail_msg
2135
        test = msg and not res.offline
2136
        self._ErrorIf(test, self.ENODEHOOKS, node_name,
2137
                      "Communication failure in hooks execution: %s", msg)
2138
        if res.offline or msg:
2139
          # No need to investigate payload if node is offline or gave an error.
2140
          # override manually lu_result here as _ErrorIf only
2141
          # overrides self.bad
2142
          lu_result = 1
2143
          continue
2144
        for script, hkr, output in res.payload:
2145
          test = hkr == constants.HKR_FAIL
2146
          self._ErrorIf(test, self.ENODEHOOKS, node_name,
2147
                        "Script %s failed, output:", script)
2148
          if test:
2149
            output = indent_re.sub('      ', output)
2150
            feedback_fn("%s" % output)
2151
            lu_result = 0
2152

    
2153
      return lu_result
2154

    
2155

    
2156
class LUVerifyDisks(NoHooksLU):
2157
  """Verifies the cluster disks status.
2158

2159
  """
2160
  REQ_BGL = False
2161

    
2162
  def ExpandNames(self):
2163
    self.needed_locks = {
2164
      locking.LEVEL_NODE: locking.ALL_SET,
2165
      locking.LEVEL_INSTANCE: locking.ALL_SET,
2166
    }
2167
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
2168

    
2169
  def Exec(self, feedback_fn):
2170
    """Verify integrity of cluster disks.
2171

2172
    @rtype: tuple of three items
2173
    @return: a tuple of (dict of node-to-node_error, list of instances
2174
        which need activate-disks, dict of instance: (node, volume) for
2175
        missing volumes
2176

2177
    """
2178
    result = res_nodes, res_instances, res_missing = {}, [], {}
2179

    
2180
    vg_name = self.cfg.GetVGName()
2181
    nodes = utils.NiceSort(self.cfg.GetNodeList())
2182
    instances = [self.cfg.GetInstanceInfo(name)
2183
                 for name in self.cfg.GetInstanceList()]
2184

    
2185
    nv_dict = {}
2186
    for inst in instances:
2187
      inst_lvs = {}
2188
      if (not inst.admin_up or
2189
          inst.disk_template not in constants.DTS_NET_MIRROR):
2190
        continue
2191
      inst.MapLVsByNode(inst_lvs)
2192
      # transform { iname: {node: [vol,],},} to {(node, vol): iname}
2193
      for node, vol_list in inst_lvs.iteritems():
2194
        for vol in vol_list:
2195
          nv_dict[(node, vol)] = inst
2196

    
2197
    if not nv_dict:
2198
      return result
2199

    
2200
    node_lvs = self.rpc.call_lv_list(nodes, vg_name)
2201

    
2202
    for node in nodes:
2203
      # node_volume
2204
      node_res = node_lvs[node]
2205
      if node_res.offline:
2206
        continue
2207
      msg = node_res.fail_msg
2208
      if msg:
2209
        logging.warning("Error enumerating LVs on node %s: %s", node, msg)
2210
        res_nodes[node] = msg
2211
        continue
2212

    
2213
      lvs = node_res.payload
2214
      for lv_name, (_, _, lv_online) in lvs.items():
2215
        inst = nv_dict.pop((node, lv_name), None)
2216
        if (not lv_online and inst is not None
2217
            and inst.name not in res_instances):
2218
          res_instances.append(inst.name)
2219

    
2220
    # any leftover items in nv_dict are missing LVs, let's arrange the
2221
    # data better
2222
    for key, inst in nv_dict.iteritems():
2223
      if inst.name not in res_missing:
2224
        res_missing[inst.name] = []
2225
      res_missing[inst.name].append(key)
2226

    
2227
    return result
2228

    
2229

    
2230
class LURepairDiskSizes(NoHooksLU):
2231
  """Verifies the cluster disks sizes.
2232

2233
  """
2234
  _OP_PARAMS = [("instances", ht.EmptyList, ht.TListOf(ht.TNonEmptyString))]
2235
  REQ_BGL = False
2236

    
2237
  def ExpandNames(self):
2238
    if self.op.instances:
2239
      self.wanted_names = []
2240
      for name in self.op.instances:
2241
        full_name = _ExpandInstanceName(self.cfg, name)
2242
        self.wanted_names.append(full_name)
2243
      self.needed_locks = {
2244
        locking.LEVEL_NODE: [],
2245
        locking.LEVEL_INSTANCE: self.wanted_names,
2246
        }
2247
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2248
    else:
2249
      self.wanted_names = None
2250
      self.needed_locks = {
2251
        locking.LEVEL_NODE: locking.ALL_SET,
2252
        locking.LEVEL_INSTANCE: locking.ALL_SET,
2253
        }
2254
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
2255

    
2256
  def DeclareLocks(self, level):
2257
    if level == locking.LEVEL_NODE and self.wanted_names is not None:
2258
      self._LockInstancesNodes(primary_only=True)
2259

    
2260
  def CheckPrereq(self):
2261
    """Check prerequisites.
2262

2263
    This only checks the optional instance list against the existing names.
2264

2265
    """
2266
    if self.wanted_names is None:
2267
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
2268

    
2269
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
2270
                             in self.wanted_names]
2271

    
2272
  def _EnsureChildSizes(self, disk):
2273
    """Ensure children of the disk have the needed disk size.
2274

2275
    This is valid mainly for DRBD8 and fixes an issue where the
2276
    children have smaller disk size.
2277

2278
    @param disk: an L{ganeti.objects.Disk} object
2279

2280
    """
2281
    if disk.dev_type == constants.LD_DRBD8:
2282
      assert disk.children, "Empty children for DRBD8?"
2283
      fchild = disk.children[0]
2284
      mismatch = fchild.size < disk.size
2285
      if mismatch:
2286
        self.LogInfo("Child disk has size %d, parent %d, fixing",
2287
                     fchild.size, disk.size)
2288
        fchild.size = disk.size
2289

    
2290
      # and we recurse on this child only, not on the metadev
2291
      return self._EnsureChildSizes(fchild) or mismatch
2292
    else:
2293
      return False
2294

    
2295
  def Exec(self, feedback_fn):
2296
    """Verify the size of cluster disks.
2297

2298
    """
2299
    # TODO: check child disks too
2300
    # TODO: check differences in size between primary/secondary nodes
2301
    per_node_disks = {}
2302
    for instance in self.wanted_instances:
2303
      pnode = instance.primary_node
2304
      if pnode not in per_node_disks:
2305
        per_node_disks[pnode] = []
2306
      for idx, disk in enumerate(instance.disks):
2307
        per_node_disks[pnode].append((instance, idx, disk))
2308

    
2309
    changed = []
2310
    for node, dskl in per_node_disks.items():
2311
      newl = [v[2].Copy() for v in dskl]
2312
      for dsk in newl:
2313
        self.cfg.SetDiskID(dsk, node)
2314
      result = self.rpc.call_blockdev_getsizes(node, newl)
2315
      if result.fail_msg:
2316
        self.LogWarning("Failure in blockdev_getsizes call to node"
2317
                        " %s, ignoring", node)
2318
        continue
2319
      if len(result.data) != len(dskl):
2320
        self.LogWarning("Invalid result from node %s, ignoring node results",
2321
                        node)
2322
        continue
2323
      for ((instance, idx, disk), size) in zip(dskl, result.data):
2324
        if size is None:
2325
          self.LogWarning("Disk %d of instance %s did not return size"
2326
                          " information, ignoring", idx, instance.name)
2327
          continue
2328
        if not isinstance(size, (int, long)):
2329
          self.LogWarning("Disk %d of instance %s did not return valid"
2330
                          " size information, ignoring", idx, instance.name)
2331
          continue
2332
        size = size >> 20
2333
        if size != disk.size:
2334
          self.LogInfo("Disk %d of instance %s has mismatched size,"
2335
                       " correcting: recorded %d, actual %d", idx,
2336
                       instance.name, disk.size, size)
2337
          disk.size = size
2338
          self.cfg.Update(instance, feedback_fn)
2339
          changed.append((instance.name, idx, size))
2340
        if self._EnsureChildSizes(disk):
2341
          self.cfg.Update(instance, feedback_fn)
2342
          changed.append((instance.name, idx, disk.size))
2343
    return changed
2344

    
2345

    
2346
class LURenameCluster(LogicalUnit):
2347
  """Rename the cluster.
2348

2349
  """
2350
  HPATH = "cluster-rename"
2351
  HTYPE = constants.HTYPE_CLUSTER
2352
  _OP_PARAMS = [("name", ht.NoDefault, ht.TNonEmptyString)]
2353

    
2354
  def BuildHooksEnv(self):
2355
    """Build hooks env.
2356

2357
    """
2358
    env = {
2359
      "OP_TARGET": self.cfg.GetClusterName(),
2360
      "NEW_NAME": self.op.name,
2361
      }
2362
    mn = self.cfg.GetMasterNode()
2363
    all_nodes = self.cfg.GetNodeList()
2364
    return env, [mn], all_nodes
2365

    
2366
  def CheckPrereq(self):
2367
    """Verify that the passed name is a valid one.
2368

2369
    """
2370
    hostname = netutils.GetHostname(name=self.op.name,
2371
                                    family=self.cfg.GetPrimaryIPFamily())
2372

    
2373
    new_name = hostname.name
2374
    self.ip = new_ip = hostname.ip
2375
    old_name = self.cfg.GetClusterName()
2376
    old_ip = self.cfg.GetMasterIP()
2377
    if new_name == old_name and new_ip == old_ip:
2378
      raise errors.OpPrereqError("Neither the name nor the IP address of the"
2379
                                 " cluster has changed",
2380
                                 errors.ECODE_INVAL)
2381
    if new_ip != old_ip:
2382
      if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
2383
        raise errors.OpPrereqError("The given cluster IP address (%s) is"
2384
                                   " reachable on the network" %
2385
                                   new_ip, errors.ECODE_NOTUNIQUE)
2386

    
2387
    self.op.name = new_name
2388

    
2389
  def Exec(self, feedback_fn):
2390
    """Rename the cluster.
2391

2392
    """
2393
    clustername = self.op.name
2394
    ip = self.ip
2395

    
2396
    # shutdown the master IP
2397
    master = self.cfg.GetMasterNode()
2398
    result = self.rpc.call_node_stop_master(master, False)
2399
    result.Raise("Could not disable the master role")
2400

    
2401
    try:
2402
      cluster = self.cfg.GetClusterInfo()
2403
      cluster.cluster_name = clustername
2404
      cluster.master_ip = ip
2405
      self.cfg.Update(cluster, feedback_fn)
2406

    
2407
      # update the known hosts file
2408
      ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
2409
      node_list = self.cfg.GetNodeList()
2410
      try:
2411
        node_list.remove(master)
2412
      except ValueError:
2413
        pass
2414
      result = self.rpc.call_upload_file(node_list,
2415
                                         constants.SSH_KNOWN_HOSTS_FILE)
2416
      for to_node, to_result in result.iteritems():
2417
        msg = to_result.fail_msg
2418
        if msg:
2419
          msg = ("Copy of file %s to node %s failed: %s" %
2420
                 (constants.SSH_KNOWN_HOSTS_FILE, to_node, msg))
2421
          self.proc.LogWarning(msg)
2422

    
2423
    finally:
2424
      result = self.rpc.call_node_start_master(master, False, False)
2425
      msg = result.fail_msg
2426
      if msg:
2427
        self.LogWarning("Could not re-enable the master role on"
2428
                        " the master, please restart manually: %s", msg)
2429

    
2430
    return clustername
2431

    
2432

    
2433
class LUSetClusterParams(LogicalUnit):
2434
  """Change the parameters of the cluster.
2435

2436
  """
2437
  HPATH = "cluster-modify"
2438
  HTYPE = constants.HTYPE_CLUSTER
2439
  _OP_PARAMS = [
2440
    ("vg_name", None, ht.TMaybeString),
2441
    ("enabled_hypervisors", None,
2442
     ht.TOr(ht.TAnd(ht.TListOf(ht.TElemOf(constants.HYPER_TYPES)), ht.TTrue),
2443
            ht.TNone)),
2444
    ("hvparams", None, ht.TOr(ht.TDictOf(ht.TNonEmptyString, ht.TDict),
2445
                              ht.TNone)),
2446
    ("beparams", None, ht.TOr(ht.TDictOf(ht.TNonEmptyString, ht.TDict),
2447
                              ht.TNone)),
2448
    ("os_hvp", None, ht.TOr(ht.TDictOf(ht.TNonEmptyString, ht.TDict),
2449
                            ht.TNone)),
2450
    ("osparams", None, ht.TOr(ht.TDictOf(ht.TNonEmptyString, ht.TDict),
2451
                              ht.TNone)),
2452
    ("candidate_pool_size", None, ht.TOr(ht.TStrictPositiveInt, ht.TNone)),
2453
    ("uid_pool", None, ht.NoType),
2454
    ("add_uids", None, ht.NoType),
2455
    ("remove_uids", None, ht.NoType),
2456
    ("maintain_node_health", None, ht.TMaybeBool),
2457
    ("prealloc_wipe_disks", None, ht.TMaybeBool),
2458
    ("nicparams", None, ht.TOr(ht.TDict, ht.TNone)),
2459
    ("drbd_helper", None, ht.TOr(ht.TString, ht.TNone)),
2460
    ("default_iallocator", None, ht.TOr(ht.TString, ht.TNone)),
2461
    ("reserved_lvs", None, ht.TOr(ht.TListOf(ht.TNonEmptyString), ht.TNone)),
2462
    ("hidden_os", None, ht.TOr(ht.TListOf(\
2463
          ht.TAnd(ht.TList,
2464
                ht.TIsLength(2),
2465
                ht.TMap(lambda v: v[0], ht.TElemOf(constants.DDMS_VALUES)))),
2466
          ht.TNone)),
2467
    ("blacklisted_os", None, ht.TOr(ht.TListOf(\
2468
          ht.TAnd(ht.TList,
2469
                ht.TIsLength(2),
2470
                ht.TMap(lambda v: v[0], ht.TElemOf(constants.DDMS_VALUES)))),
2471
          ht.TNone)),
2472
    ]
2473
  REQ_BGL = False
2474

    
2475
  def CheckArguments(self):
2476
    """Check parameters
2477

2478
    """
2479
    if self.op.uid_pool:
2480
      uidpool.CheckUidPool(self.op.uid_pool)
2481

    
2482
    if self.op.add_uids:
2483
      uidpool.CheckUidPool(self.op.add_uids)
2484

    
2485
    if self.op.remove_uids:
2486
      uidpool.CheckUidPool(self.op.remove_uids)
2487

    
2488
  def ExpandNames(self):
2489
    # FIXME: in the future maybe other cluster params won't require checking on
2490
    # all nodes to be modified.
2491
    self.needed_locks = {
2492
      locking.LEVEL_NODE: locking.ALL_SET,
2493
    }
2494
    self.share_locks[locking.LEVEL_NODE] = 1
2495

    
2496
  def BuildHooksEnv(self):
2497
    """Build hooks env.
2498

2499
    """
2500
    env = {
2501
      "OP_TARGET": self.cfg.GetClusterName(),
2502
      "NEW_VG_NAME": self.op.vg_name,
2503
      }
2504
    mn = self.cfg.GetMasterNode()
2505
    return env, [mn], [mn]
2506

    
2507
  def CheckPrereq(self):
2508
    """Check prerequisites.
2509

2510
    This checks whether the given params don't conflict and
2511
    if the given volume group is valid.
2512

2513
    """
2514
    if self.op.vg_name is not None and not self.op.vg_name:
2515
      if self.cfg.HasAnyDiskOfType(constants.LD_LV):
2516
        raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based"
2517
                                   " instances exist", errors.ECODE_INVAL)
2518

    
2519
    if self.op.drbd_helper is not None and not self.op.drbd_helper:
2520
      if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8):
2521
        raise errors.OpPrereqError("Cannot disable drbd helper while"
2522
                                   " drbd-based instances exist",
2523
                                   errors.ECODE_INVAL)
2524

    
2525
    node_list = self.acquired_locks[locking.LEVEL_NODE]
2526

    
2527
    # if vg_name not None, checks given volume group on all nodes
2528
    if self.op.vg_name:
2529
      vglist = self.rpc.call_vg_list(node_list)
2530
      for node in node_list:
2531
        msg = vglist[node].fail_msg
2532
        if msg:
2533
          # ignoring down node
2534
          self.LogWarning("Error while gathering data on node %s"
2535
                          " (ignoring node): %s", node, msg)
2536
          continue
2537
        vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
2538
                                              self.op.vg_name,
2539
                                              constants.MIN_VG_SIZE)
2540
        if vgstatus:
2541
          raise errors.OpPrereqError("Error on node '%s': %s" %
2542
                                     (node, vgstatus), errors.ECODE_ENVIRON)
2543

    
2544
    if self.op.drbd_helper:
2545
      # checks given drbd helper on all nodes
2546
      helpers = self.rpc.call_drbd_helper(node_list)
2547
      for node in node_list:
2548
        ninfo = self.cfg.GetNodeInfo(node)
2549
        if ninfo.offline:
2550
          self.LogInfo("Not checking drbd helper on offline node %s", node)
2551
          continue
2552
        msg = helpers[node].fail_msg
2553
        if msg:
2554
          raise errors.OpPrereqError("Error checking drbd helper on node"
2555
                                     " '%s': %s" % (node, msg),
2556
                                     errors.ECODE_ENVIRON)
2557
        node_helper = helpers[node].payload
2558
        if node_helper != self.op.drbd_helper:
2559
          raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" %
2560
                                     (node, node_helper), errors.ECODE_ENVIRON)
2561

    
2562
    self.cluster = cluster = self.cfg.GetClusterInfo()
2563
    # validate params changes
2564
    if self.op.beparams:
2565
      utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
2566
      self.new_beparams = cluster.SimpleFillBE(self.op.beparams)
2567

    
2568
    if self.op.nicparams:
2569
      utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
2570
      self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams)
2571
      objects.NIC.CheckParameterSyntax(self.new_nicparams)
2572
      nic_errors = []
2573

    
2574
      # check all instances for consistency
2575
      for instance in self.cfg.GetAllInstancesInfo().values():
2576
        for nic_idx, nic in enumerate(instance.nics):
2577
          params_copy = copy.deepcopy(nic.nicparams)
2578
          params_filled = objects.FillDict(self.new_nicparams, params_copy)
2579

    
2580
          # check parameter syntax
2581
          try:
2582
            objects.NIC.CheckParameterSyntax(params_filled)
2583
          except errors.ConfigurationError, err:
2584
            nic_errors.append("Instance %s, nic/%d: %s" %
2585
                              (instance.name, nic_idx, err))
2586

    
2587
          # if we're moving instances to routed, check that they have an ip
2588
          target_mode = params_filled[constants.NIC_MODE]
2589
          if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
2590
            nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
2591
                              (instance.name, nic_idx))
2592
      if nic_errors:
2593
        raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
2594
                                   "\n".join(nic_errors))
2595

    
2596
    # hypervisor list/parameters
2597
    self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
2598
    if self.op.hvparams:
2599
      for hv_name, hv_dict in self.op.hvparams.items():
2600
        if hv_name not in self.new_hvparams:
2601
          self.new_hvparams[hv_name] = hv_dict
2602
        else:
2603
          self.new_hvparams[hv_name].update(hv_dict)
2604

    
2605
    # os hypervisor parameters
2606
    self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
2607
    if self.op.os_hvp:
2608
      for os_name, hvs in self.op.os_hvp.items():
2609
        if os_name not in self.new_os_hvp:
2610
          self.new_os_hvp[os_name] = hvs
2611
        else:
2612
          for hv_name, hv_dict in hvs.items():
2613
            if hv_name not in self.new_os_hvp[os_name]:
2614
              self.new_os_hvp[os_name][hv_name] = hv_dict
2615
            else:
2616
              self.new_os_hvp[os_name][hv_name].update(hv_dict)
2617

    
2618
    # os parameters
2619
    self.new_osp = objects.FillDict(cluster.osparams, {})
2620
    if self.op.osparams:
2621
      for os_name, osp in self.op.osparams.items():
2622
        if os_name not in self.new_osp:
2623
          self.new_osp[os_name] = {}
2624

    
2625
        self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp,
2626
                                                  use_none=True)
2627

    
2628
        if not self.new_osp[os_name]:
2629
          # we removed all parameters
2630
          del self.new_osp[os_name]
2631
        else:
2632
          # check the parameter validity (remote check)
2633
          _CheckOSParams(self, False, [self.cfg.GetMasterNode()],
2634
                         os_name, self.new_osp[os_name])
2635

    
2636
    # changes to the hypervisor list
2637
    if self.op.enabled_hypervisors is not None:
2638
      self.hv_list = self.op.enabled_hypervisors
2639
      for hv in self.hv_list:
2640
        # if the hypervisor doesn't already exist in the cluster
2641
        # hvparams, we initialize it to empty, and then (in both
2642
        # cases) we make sure to fill the defaults, as we might not
2643
        # have a complete defaults list if the hypervisor wasn't
2644
        # enabled before
2645
        if hv not in new_hvp:
2646
          new_hvp[hv] = {}
2647
        new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
2648
        utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
2649
    else:
2650
      self.hv_list = cluster.enabled_hypervisors
2651

    
2652
    if self.op.hvparams or self.op.enabled_hypervisors is not None:
2653
      # either the enabled list has changed, or the parameters have, validate
2654
      for hv_name, hv_params in self.new_hvparams.items():
2655
        if ((self.op.hvparams and hv_name in self.op.hvparams) or
2656
            (self.op.enabled_hypervisors and
2657
             hv_name in self.op.enabled_hypervisors)):
2658
          # either this is a new hypervisor, or its parameters have changed
2659
          hv_class = hypervisor.GetHypervisor(hv_name)
2660
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2661
          hv_class.CheckParameterSyntax(hv_params)
2662
          _CheckHVParams(self, node_list, hv_name, hv_params)
2663

    
2664
    if self.op.os_hvp:
2665
      # no need to check any newly-enabled hypervisors, since the
2666
      # defaults have already been checked in the above code-block
2667
      for os_name, os_hvp in self.new_os_hvp.items():
2668
        for hv_name, hv_params in os_hvp.items():
2669
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2670
          # we need to fill in the new os_hvp on top of the actual hv_p
2671
          cluster_defaults = self.new_hvparams.get(hv_name, {})
2672
          new_osp = objects.FillDict(cluster_defaults, hv_params)
2673
          hv_class = hypervisor.GetHypervisor(hv_name)
2674
          hv_class.CheckParameterSyntax(new_osp)
2675
          _CheckHVParams(self, node_list, hv_name, new_osp)
2676

    
2677
    if self.op.default_iallocator:
2678
      alloc_script = utils.FindFile(self.op.default_iallocator,
2679
                                    constants.IALLOCATOR_SEARCH_PATH,
2680
                                    os.path.isfile)
2681
      if alloc_script is None:
2682
        raise errors.OpPrereqError("Invalid default iallocator script '%s'"
2683
                                   " specified" % self.op.default_iallocator,
2684
                                   errors.ECODE_INVAL)
2685

    
2686
  def Exec(self, feedback_fn):
2687
    """Change the parameters of the cluster.
2688

2689
    """
2690
    if self.op.vg_name is not None:
2691
      new_volume = self.op.vg_name
2692
      if not new_volume:
2693
        new_volume = None
2694
      if new_volume != self.cfg.GetVGName():
2695
        self.cfg.SetVGName(new_volume)
2696
      else:
2697
        feedback_fn("Cluster LVM configuration already in desired"
2698
                    " state, not changing")
2699
    if self.op.drbd_helper is not None:
2700
      new_helper = self.op.drbd_helper
2701
      if not new_helper:
2702
        new_helper = None
2703
      if new_helper != self.cfg.GetDRBDHelper():
2704
        self.cfg.SetDRBDHelper(new_helper)
2705
      else:
2706
        feedback_fn("Cluster DRBD helper already in desired state,"
2707
                    " not changing")
2708
    if self.op.hvparams:
2709
      self.cluster.hvparams = self.new_hvparams
2710
    if self.op.os_hvp:
2711
      self.cluster.os_hvp = self.new_os_hvp
2712
    if self.op.enabled_hypervisors is not None:
2713
      self.cluster.hvparams = self.new_hvparams
2714
      self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
2715
    if self.op.beparams:
2716
      self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
2717
    if self.op.nicparams:
2718
      self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
2719
    if self.op.osparams:
2720
      self.cluster.osparams = self.new_osp
2721

    
2722
    if self.op.candidate_pool_size is not None:
2723
      self.cluster.candidate_pool_size = self.op.candidate_pool_size
2724
      # we need to update the pool size here, otherwise the save will fail
2725
      _AdjustCandidatePool(self, [])
2726

    
2727
    if self.op.maintain_node_health is not None:
2728
      self.cluster.maintain_node_health = self.op.maintain_node_health
2729

    
2730
    if self.op.prealloc_wipe_disks is not None:
2731
      self.cluster.prealloc_wipe_disks = self.op.prealloc_wipe_disks
2732

    
2733
    if self.op.add_uids is not None:
2734
      uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
2735

    
2736
    if self.op.remove_uids is not None:
2737
      uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
2738

    
2739
    if self.op.uid_pool is not None:
2740
      self.cluster.uid_pool = self.op.uid_pool
2741

    
2742
    if self.op.default_iallocator is not None:
2743
      self.cluster.default_iallocator = self.op.default_iallocator
2744

    
2745
    if self.op.reserved_lvs is not None:
2746
      self.cluster.reserved_lvs = self.op.reserved_lvs
2747

    
2748
    def helper_os(aname, mods, desc):
2749
      desc += " OS list"
2750
      lst = getattr(self.cluster, aname)
2751
      for key, val in mods:
2752
        if key == constants.DDM_ADD:
2753
          if val in lst:
2754
            feedback_fn("OS %s already in %s, ignoring", val, desc)
2755
          else:
2756
            lst.append(val)
2757
        elif key == constants.DDM_REMOVE:
2758
          if val in lst:
2759
            lst.remove(val)
2760
          else:
2761
            feedback_fn("OS %s not found in %s, ignoring", val, desc)
2762
        else:
2763
          raise errors.ProgrammerError("Invalid modification '%s'" % key)
2764

    
2765
    if self.op.hidden_os:
2766
      helper_os("hidden_os", self.op.hidden_os, "hidden")
2767

    
2768
    if self.op.blacklisted_os:
2769
      helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted")
2770

    
2771
    self.cfg.Update(self.cluster, feedback_fn)
2772

    
2773

    
2774
def _RedistributeAncillaryFiles(lu, additional_nodes=None):
2775
  """Distribute additional files which are part of the cluster configuration.
2776

2777
  ConfigWriter takes care of distributing the config and ssconf files, but
2778
  there are more files which should be distributed to all nodes. This function
2779
  makes sure those are copied.
2780

2781
  @param lu: calling logical unit
2782
  @param additional_nodes: list of nodes not in the config to distribute to
2783

2784
  """
2785
  # 1. Gather target nodes
2786
  myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
2787
  dist_nodes = lu.cfg.GetOnlineNodeList()
2788
  if additional_nodes is not None:
2789
    dist_nodes.extend(additional_nodes)
2790
  if myself.name in dist_nodes:
2791
    dist_nodes.remove(myself.name)
2792

    
2793
  # 2. Gather files to distribute
2794
  dist_files = set([constants.ETC_HOSTS,
2795
                    constants.SSH_KNOWN_HOSTS_FILE,
2796
                    constants.RAPI_CERT_FILE,
2797
                    constants.RAPI_USERS_FILE,
2798
                    constants.CONFD_HMAC_KEY,
2799
                    constants.CLUSTER_DOMAIN_SECRET_FILE,
2800
                   ])
2801

    
2802
  enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
2803
  for hv_name in enabled_hypervisors:
2804
    hv_class = hypervisor.GetHypervisor(hv_name)
2805
    dist_files.update(hv_class.GetAncillaryFiles())
2806

    
2807
  # 3. Perform the files upload
2808
  for fname in dist_files:
2809
    if os.path.exists(fname):
2810
      result = lu.rpc.call_upload_file(dist_nodes, fname)
2811
      for to_node, to_result in result.items():
2812
        msg = to_result.fail_msg
2813
        if msg:
2814
          msg = ("Copy of file %s to node %s failed: %s" %
2815
                 (fname, to_node, msg))
2816
          lu.proc.LogWarning(msg)
2817

    
2818

    
2819
class LURedistributeConfig(NoHooksLU):
2820
  """Force the redistribution of cluster configuration.
2821

2822
  This is a very simple LU.
2823

2824
  """
2825
  REQ_BGL = False
2826

    
2827
  def ExpandNames(self):
2828
    self.needed_locks = {
2829
      locking.LEVEL_NODE: locking.ALL_SET,
2830
    }
2831
    self.share_locks[locking.LEVEL_NODE] = 1
2832

    
2833
  def Exec(self, feedback_fn):
2834
    """Redistribute the configuration.
2835

2836
    """
2837
    self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
2838
    _RedistributeAncillaryFiles(self)
2839

    
2840

    
2841
def _WaitForSync(lu, instance, disks=None, oneshot=False):
2842
  """Sleep and poll for an instance's disk to sync.
2843

2844
  """
2845
  if not instance.disks or disks is not None and not disks:
2846
    return True
2847

    
2848
  disks = _ExpandCheckDisks(instance, disks)
2849

    
2850
  if not oneshot:
2851
    lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
2852

    
2853
  node = instance.primary_node
2854

    
2855
  for dev in disks:
2856
    lu.cfg.SetDiskID(dev, node)
2857

    
2858
  # TODO: Convert to utils.Retry
2859

    
2860
  retries = 0
2861
  degr_retries = 10 # in seconds, as we sleep 1 second each time
2862
  while True:
2863
    max_time = 0
2864
    done = True
2865
    cumul_degraded = False
2866
    rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
2867
    msg = rstats.fail_msg
2868
    if msg:
2869
      lu.LogWarning("Can't get any data from node %s: %s", node, msg)
2870
      retries += 1
2871
      if retries >= 10:
2872
        raise errors.RemoteError("Can't contact node %s for mirror data,"
2873
                                 " aborting." % node)
2874
      time.sleep(6)
2875
      continue
2876
    rstats = rstats.payload
2877
    retries = 0
2878
    for i, mstat in enumerate(rstats):
2879
      if mstat is None:
2880
        lu.LogWarning("Can't compute data for node %s/%s",
2881
                           node, disks[i].iv_name)
2882
        continue
2883

    
2884
      cumul_degraded = (cumul_degraded or
2885
                        (mstat.is_degraded and mstat.sync_percent is None))
2886
      if mstat.sync_percent is not None:
2887
        done = False
2888
        if mstat.estimated_time is not None:
2889
          rem_time = ("%s remaining (estimated)" %
2890
                      utils.FormatSeconds(mstat.estimated_time))
2891
          max_time = mstat.estimated_time
2892
        else:
2893
          rem_time = "no time estimate"
2894
        lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
2895
                        (disks[i].iv_name, mstat.sync_percent, rem_time))
2896

    
2897
    # if we're done but degraded, let's do a few small retries, to
2898
    # make sure we see a stable and not transient situation; therefore
2899
    # we force restart of the loop
2900
    if (done or oneshot) and cumul_degraded and degr_retries > 0:
2901
      logging.info("Degraded disks found, %d retries left", degr_retries)
2902
      degr_retries -= 1
2903
      time.sleep(1)
2904
      continue
2905

    
2906
    if done or oneshot:
2907
      break
2908

    
2909
    time.sleep(min(60, max_time))
2910

    
2911
  if done:
2912
    lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
2913
  return not cumul_degraded
2914

    
2915

    
2916
def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
2917
  """Check that mirrors are not degraded.
2918

2919
  The ldisk parameter, if True, will change the test from the
2920
  is_degraded attribute (which represents overall non-ok status for
2921
  the device(s)) to the ldisk (representing the local storage status).
2922

2923
  """
2924
  lu.cfg.SetDiskID(dev, node)
2925

    
2926
  result = True
2927

    
2928
  if on_primary or dev.AssembleOnSecondary():
2929
    rstats = lu.rpc.call_blockdev_find(node, dev)
2930
    msg = rstats.fail_msg
2931
    if msg:
2932
      lu.LogWarning("Can't find disk on node %s: %s", node, msg)
2933
      result = False
2934
    elif not rstats.payload:
2935
      lu.LogWarning("Can't find disk on node %s", node)
2936
      result = False
2937
    else:
2938
      if ldisk:
2939
        result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
2940
      else:
2941
        result = result and not rstats.payload.is_degraded
2942

    
2943
  if dev.children:
2944
    for child in dev.children:
2945
      result = result and _CheckDiskConsistency(lu, child, node, on_primary)
2946

    
2947
  return result
2948

    
2949

    
2950
class LUDiagnoseOS(NoHooksLU):
2951
  """Logical unit for OS diagnose/query.
2952

2953
  """
2954
  _OP_PARAMS = [
2955
    _POutputFields,
2956
    ("names", ht.EmptyList, ht.TListOf(ht.TNonEmptyString)),
2957
    ]
2958
  REQ_BGL = False
2959
  _HID = "hidden"
2960
  _BLK = "blacklisted"
2961
  _VLD = "valid"
2962
  _FIELDS_STATIC = utils.FieldSet()
2963
  _FIELDS_DYNAMIC = utils.FieldSet("name", _VLD, "node_status", "variants",
2964
                                   "parameters", "api_versions", _HID, _BLK)
2965

    
2966
  def CheckArguments(self):
2967
    if self.op.names:
2968
      raise errors.OpPrereqError("Selective OS query not supported",
2969
                                 errors.ECODE_INVAL)
2970

    
2971
    _CheckOutputFields(static=self._FIELDS_STATIC,
2972
                       dynamic=self._FIELDS_DYNAMIC,
2973
                       selected=self.op.output_fields)
2974

    
2975
  def ExpandNames(self):
2976
    # Lock all nodes, in shared mode
2977
    # Temporary removal of locks, should be reverted later
2978
    # TODO: reintroduce locks when they are lighter-weight
2979
    self.needed_locks = {}
2980
    #self.share_locks[locking.LEVEL_NODE] = 1
2981
    #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
2982

    
2983
  @staticmethod
2984
  def _DiagnoseByOS(rlist):
2985
    """Remaps a per-node return list into an a per-os per-node dictionary
2986

2987
    @param rlist: a map with node names as keys and OS objects as values
2988

2989
    @rtype: dict
2990
    @return: a dictionary with osnames as keys and as value another
2991
        map, with nodes as keys and tuples of (path, status, diagnose,
2992
        variants, parameters, api_versions) as values, eg::
2993

2994
          {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
2995
                                     (/srv/..., False, "invalid api")],
2996
                           "node2": [(/srv/..., True, "", [], [])]}
2997
          }
2998

2999
    """
3000
    all_os = {}
3001
    # we build here the list of nodes that didn't fail the RPC (at RPC
3002
    # level), so that nodes with a non-responding node daemon don't
3003
    # make all OSes invalid
3004
    good_nodes = [node_name for node_name in rlist
3005
                  if not rlist[node_name].fail_msg]
3006
    for node_name, nr in rlist.items():
3007
      if nr.fail_msg or not nr.payload:
3008
        continue
3009
      for (name, path, status, diagnose, variants,
3010
           params, api_versions) in nr.payload:
3011
        if name not in all_os:
3012
          # build a list of nodes for this os containing empty lists
3013
          # for each node in node_list
3014
          all_os[name] = {}
3015
          for nname in good_nodes:
3016
            all_os[name][nname] = []
3017
        # convert params from [name, help] to (name, help)
3018
        params = [tuple(v) for v in params]
3019
        all_os[name][node_name].append((path, status, diagnose,
3020
                                        variants, params, api_versions))
3021
    return all_os
3022

    
3023
  def Exec(self, feedback_fn):
3024
    """Compute the list of OSes.
3025

3026
    """
3027
    valid_nodes = [node for node in self.cfg.GetOnlineNodeList()]
3028
    node_data = self.rpc.call_os_diagnose(valid_nodes)
3029
    pol = self._DiagnoseByOS(node_data)
3030
    output = []
3031
    cluster = self.cfg.GetClusterInfo()
3032

    
3033
    for os_name in utils.NiceSort(pol.keys()):
3034
      os_data = pol[os_name]
3035
      row = []
3036
      valid = True
3037
      (variants, params, api_versions) = null_state = (set(), set(), set())
3038
      for idx, osl in enumerate(os_data.values()):
3039
        valid = bool(valid and osl and osl[0][1])
3040
        if not valid:
3041
          (variants, params, api_versions) = null_state
3042
          break
3043
        node_variants, node_params, node_api = osl[0][3:6]
3044
        if idx == 0: # first entry
3045
          variants = set(node_variants)
3046
          params = set(node_params)
3047
          api_versions = set(node_api)
3048
        else: # keep consistency
3049
          variants.intersection_update(node_variants)
3050
          params.intersection_update(node_params)
3051
          api_versions.intersection_update(node_api)
3052

    
3053
      is_hid = os_name in cluster.hidden_os
3054
      is_blk = os_name in cluster.blacklisted_os
3055
      if ((self._HID not in self.op.output_fields and is_hid) or
3056
          (self._BLK not in self.op.output_fields and is_blk) or
3057
          (self._VLD not in self.op.output_fields and not valid)):
3058
        continue
3059

    
3060
      for field in self.op.output_fields:
3061
        if field == "name":
3062
          val = os_name
3063
        elif field == self._VLD:
3064
          val = valid
3065
        elif field == "node_status":
3066
          # this is just a copy of the dict
3067
          val = {}
3068
          for node_name, nos_list in os_data.items():
3069
            val[node_name] = nos_list
3070
        elif field == "variants":
3071
          val = utils.NiceSort(list(variants))
3072
        elif field == "parameters":
3073
          val = list(params)
3074
        elif field == "api_versions":
3075
          val = list(api_versions)
3076
        elif field == self._HID:
3077
          val = is_hid
3078
        elif field == self._BLK:
3079
          val = is_blk
3080
        else:
3081
          raise errors.ParameterError(field)
3082
        row.append(val)
3083
      output.append(row)
3084

    
3085
    return output
3086

    
3087

    
3088
class LURemoveNode(LogicalUnit):
3089
  """Logical unit for removing a node.
3090

3091
  """
3092
  HPATH = "node-remove"
3093
  HTYPE = constants.HTYPE_NODE
3094
  _OP_PARAMS = [
3095
    _PNodeName,
3096
    ]
3097

    
3098
  def BuildHooksEnv(self):
3099
    """Build hooks env.
3100

3101
    This doesn't run on the target node in the pre phase as a failed
3102
    node would then be impossible to remove.
3103

3104
    """
3105
    env = {
3106
      "OP_TARGET": self.op.node_name,
3107
      "NODE_NAME": self.op.node_name,
3108
      }
3109
    all_nodes = self.cfg.GetNodeList()
3110
    try:
3111
      all_nodes.remove(self.op.node_name)
3112
    except ValueError:
3113
      logging.warning("Node %s which is about to be removed not found"
3114
                      " in the all nodes list", self.op.node_name)
3115
    return env, all_nodes, all_nodes
3116

    
3117
  def CheckPrereq(self):
3118
    """Check prerequisites.
3119

3120
    This checks:
3121
     - the node exists in the configuration
3122
     - it does not have primary or secondary instances
3123
     - it's not the master
3124

3125
    Any errors are signaled by raising errors.OpPrereqError.
3126

3127
    """
3128
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3129
    node = self.cfg.GetNodeInfo(self.op.node_name)
3130
    assert node is not None
3131

    
3132
    instance_list = self.cfg.GetInstanceList()
3133

    
3134
    masternode = self.cfg.GetMasterNode()
3135
    if node.name == masternode:
3136
      raise errors.OpPrereqError("Node is the master node,"
3137
                                 " you need to failover first.",
3138
                                 errors.ECODE_INVAL)
3139

    
3140
    for instance_name in instance_list:
3141
      instance = self.cfg.GetInstanceInfo(instance_name)
3142
      if node.name in instance.all_nodes:
3143
        raise errors.OpPrereqError("Instance %s is still running on the node,"
3144
                                   " please remove first." % instance_name,
3145
                                   errors.ECODE_INVAL)
3146
    self.op.node_name = node.name
3147
    self.node = node
3148

    
3149
  def Exec(self, feedback_fn):
3150
    """Removes the node from the cluster.
3151

3152
    """
3153
    node = self.node
3154
    logging.info("Stopping the node daemon and removing configs from node %s",
3155
                 node.name)
3156

    
3157
    modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
3158

    
3159
    # Promote nodes to master candidate as needed
3160
    _AdjustCandidatePool(self, exceptions=[node.name])
3161
    self.context.RemoveNode(node.name)
3162

    
3163
    # Run post hooks on the node before it's removed
3164
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
3165
    try:
3166
      hm.RunPhase(constants.HOOKS_PHASE_POST, [node.name])
3167
    except:
3168
      # pylint: disable-msg=W0702
3169
      self.LogWarning("Errors occurred running hooks on %s" % node.name)
3170

    
3171
    result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
3172
    msg = result.fail_msg
3173
    if msg:
3174
      self.LogWarning("Errors encountered on the remote node while leaving"
3175
                      " the cluster: %s", msg)
3176

    
3177
    # Remove node from our /etc/hosts
3178
    if self.cfg.GetClusterInfo().modify_etc_hosts:
3179
      master_node = self.cfg.GetMasterNode()
3180
      result = self.rpc.call_etc_hosts_modify(master_node,
3181
                                              constants.ETC_HOSTS_REMOVE,
3182
                                              node.name, None)
3183
      result.Raise("Can't update hosts file with new host data")
3184
      _RedistributeAncillaryFiles(self)
3185

    
3186

    
3187
class LUQueryNodes(NoHooksLU):
3188
  """Logical unit for querying nodes.
3189

3190
  """
3191
  # pylint: disable-msg=W0142
3192
  _OP_PARAMS = [
3193
    _POutputFields,
3194
    ("names", ht.EmptyList, ht.TListOf(ht.TNonEmptyString)),
3195
    ("use_locking", False, ht.TBool),
3196
    ]
3197
  REQ_BGL = False
3198

    
3199
  _SIMPLE_FIELDS = ["name", "serial_no", "ctime", "mtime", "uuid",
3200
                    "master_candidate", "offline", "drained",
3201
                    "master_capable", "vm_capable"]
3202

    
3203
  _FIELDS_DYNAMIC = utils.FieldSet(
3204
    "dtotal", "dfree",
3205
    "mtotal", "mnode", "mfree",
3206
    "bootid",
3207
    "ctotal", "cnodes", "csockets",
3208
    )
3209

    
3210
  _FIELDS_STATIC = utils.FieldSet(*[
3211
    "pinst_cnt", "sinst_cnt",
3212
    "pinst_list", "sinst_list",
3213
    "pip", "sip", "tags",
3214
    "master",
3215
    "role"] + _SIMPLE_FIELDS
3216
    )
3217

    
3218
  def CheckArguments(self):
3219
    _CheckOutputFields(static=self._FIELDS_STATIC,
3220
                       dynamic=self._FIELDS_DYNAMIC,
3221
                       selected=self.op.output_fields)
3222

    
3223
  def ExpandNames(self):
3224
    self.needed_locks = {}
3225
    self.share_locks[locking.LEVEL_NODE] = 1
3226

    
3227
    if self.op.names:
3228
      self.wanted = _GetWantedNodes(self, self.op.names)
3229
    else:
3230
      self.wanted = locking.ALL_SET
3231

    
3232
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
3233
    self.do_locking = self.do_node_query and self.op.use_locking
3234
    if self.do_locking:
3235
      # if we don't request only static fields, we need to lock the nodes
3236
      self.needed_locks[locking.LEVEL_NODE] = self.wanted
3237

    
3238
  def Exec(self, feedback_fn):
3239
    """Computes the list of nodes and their attributes.
3240

3241
    """
3242
    all_info = self.cfg.GetAllNodesInfo()
3243
    if self.do_locking:
3244
      nodenames = self.acquired_locks[locking.LEVEL_NODE]
3245
    elif self.wanted != locking.ALL_SET:
3246
      nodenames = self.wanted
3247
      missing = set(nodenames).difference(all_info.keys())
3248
      if missing:
3249
        raise errors.OpExecError(
3250
          "Some nodes were removed before retrieving their data: %s" % missing)
3251
    else:
3252
      nodenames = all_info.keys()
3253

    
3254
    nodenames = utils.NiceSort(nodenames)
3255
    nodelist = [all_info[name] for name in nodenames]
3256

    
3257
    # begin data gathering
3258

    
3259
    if self.do_node_query:
3260
      live_data = {}
3261
      node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
3262
                                          self.cfg.GetHypervisorType())
3263
      for name in nodenames:
3264
        nodeinfo = node_data[name]
3265
        if not nodeinfo.fail_msg and nodeinfo.payload:
3266
          nodeinfo = nodeinfo.payload
3267
          fn = utils.TryConvert
3268
          live_data[name] = {
3269
            "mtotal": fn(int, nodeinfo.get('memory_total', None)),
3270
            "mnode": fn(int, nodeinfo.get('memory_dom0', None)),
3271
            "mfree": fn(int, nodeinfo.get('memory_free', None)),
3272
            "dtotal": fn(int, nodeinfo.get('vg_size', None)),
3273
            "dfree": fn(int, nodeinfo.get('vg_free', None)),
3274
            "ctotal": fn(int, nodeinfo.get('cpu_total', None)),
3275
            "bootid": nodeinfo.get('bootid', None),
3276
            "cnodes": fn(int, nodeinfo.get('cpu_nodes', None)),
3277
            "csockets": fn(int, nodeinfo.get('cpu_sockets', None)),
3278
            }
3279
        else:
3280
          live_data[name] = {}
3281
    else:
3282
      live_data = dict.fromkeys(nodenames, {})
3283

    
3284
    node_to_primary = dict([(name, set()) for name in nodenames])
3285
    node_to_secondary = dict([(name, set()) for name in nodenames])
3286

    
3287
    inst_fields = frozenset(("pinst_cnt", "pinst_list",
3288
                             "sinst_cnt", "sinst_list"))
3289
    if inst_fields & frozenset(self.op.output_fields):
3290
      inst_data = self.cfg.GetAllInstancesInfo()
3291

    
3292
      for inst in inst_data.values():
3293
        if inst.primary_node in node_to_primary:
3294
          node_to_primary[inst.primary_node].add(inst.name)
3295
        for secnode in inst.secondary_nodes:
3296
          if secnode in node_to_secondary:
3297
            node_to_secondary[secnode].add(inst.name)
3298

    
3299
    master_node = self.cfg.GetMasterNode()
3300

    
3301
    # end data gathering
3302

    
3303
    output = []
3304
    for node in nodelist:
3305
      node_output = []
3306
      for field in self.op.output_fields:
3307
        if field in self._SIMPLE_FIELDS:
3308
          val = getattr(node, field)
3309
        elif field == "pinst_list":
3310
          val = list(node_to_primary[node.name])
3311
        elif field == "sinst_list":
3312
          val = list(node_to_secondary[node.name])
3313
        elif field == "pinst_cnt":
3314
          val = len(node_to_primary[node.name])
3315
        elif field == "sinst_cnt":
3316
          val = len(node_to_secondary[node.name])
3317
        elif field == "pip":
3318
          val = node.primary_ip
3319
        elif field == "sip":
3320
          val = node.secondary_ip
3321
        elif field == "tags":
3322
          val = list(node.GetTags())
3323
        elif field == "master":
3324
          val = node.name == master_node
3325
        elif self._FIELDS_DYNAMIC.Matches(field):
3326
          val = live_data[node.name].get(field, None)
3327
        elif field == "role":
3328
          if node.name == master_node:
3329
            val = "M"
3330
          elif node.master_candidate:
3331
            val = "C"
3332
          elif node.drained:
3333
            val = "D"
3334
          elif node.offline:
3335
            val = "O"
3336
          else:
3337
            val = "R"
3338
        else:
3339
          raise errors.ParameterError(field)
3340
        node_output.append(val)
3341
      output.append(node_output)
3342

    
3343
    return output
3344

    
3345

    
3346
class LUQueryNodeVolumes(NoHooksLU):
3347
  """Logical unit for getting volumes on node(s).
3348

3349
  """
3350
  _OP_PARAMS = [
3351
    ("nodes", ht.EmptyList, ht.TListOf(ht.TNonEmptyString)),
3352
    ("output_fields", ht.NoDefault, ht.TListOf(ht.TNonEmptyString)),
3353
    ]
3354
  REQ_BGL = False
3355
  _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
3356
  _FIELDS_STATIC = utils.FieldSet("node")
3357

    
3358
  def CheckArguments(self):
3359
    _CheckOutputFields(static=self._FIELDS_STATIC,
3360
                       dynamic=self._FIELDS_DYNAMIC,
3361
                       selected=self.op.output_fields)
3362

    
3363
  def ExpandNames(self):
3364
    self.needed_locks = {}
3365
    self.share_locks[locking.LEVEL_NODE] = 1
3366
    if not self.op.nodes:
3367
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3368
    else:
3369
      self.needed_locks[locking.LEVEL_NODE] = \
3370
        _GetWantedNodes(self, self.op.nodes)
3371

    
3372
  def Exec(self, feedback_fn):
3373
    """Computes the list of nodes and their attributes.
3374

3375
    """
3376
    nodenames = self.acquired_locks[locking.LEVEL_NODE]
3377
    volumes = self.rpc.call_node_volumes(nodenames)
3378

    
3379
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
3380
             in self.cfg.GetInstanceList()]
3381

    
3382
    lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
3383

    
3384
    output = []
3385
    for node in nodenames:
3386
      nresult = volumes[node]
3387
      if nresult.offline:
3388
        continue
3389
      msg = nresult.fail_msg
3390
      if msg:
3391
        self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
3392
        continue
3393

    
3394
      node_vols = nresult.payload[:]
3395
      node_vols.sort(key=lambda vol: vol['dev'])
3396

    
3397
      for vol in node_vols:
3398
        node_output = []
3399
        for field in self.op.output_fields:
3400
          if field == "node":
3401
            val = node
3402
          elif field == "phys":
3403
            val = vol['dev']
3404
          elif field == "vg":
3405
            val = vol['vg']
3406
          elif field == "name":
3407
            val = vol['name']
3408
          elif field == "size":
3409
            val = int(float(vol['size']))
3410
          elif field == "instance":
3411
            for inst in ilist:
3412
              if node not in lv_by_node[inst]:
3413
                continue
3414
              if vol['name'] in lv_by_node[inst][node]:
3415
                val = inst.name
3416
                break
3417
            else:
3418
              val = '-'
3419
          else:
3420
            raise errors.ParameterError(field)
3421
          node_output.append(str(val))
3422

    
3423
        output.append(node_output)
3424

    
3425
    return output
3426

    
3427

    
3428
class LUQueryNodeStorage(NoHooksLU):
3429
  """Logical unit for getting information on storage units on node(s).
3430

3431
  """
3432
  _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
3433
  _OP_PARAMS = [
3434
    ("nodes", ht.EmptyList, ht.TListOf(ht.TNonEmptyString)),
3435
    ("storage_type", ht.NoDefault, _CheckStorageType),
3436
    ("output_fields", ht.NoDefault, ht.TListOf(ht.TNonEmptyString)),
3437
    ("name", None, ht.TMaybeString),
3438
    ]
3439
  REQ_BGL = False
3440

    
3441
  def CheckArguments(self):
3442
    _CheckOutputFields(static=self._FIELDS_STATIC,
3443
                       dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
3444
                       selected=self.op.output_fields)
3445

    
3446
  def ExpandNames(self):
3447
    self.needed_locks = {}
3448
    self.share_locks[locking.LEVEL_NODE] = 1
3449

    
3450
    if self.op.nodes:
3451
      self.needed_locks[locking.LEVEL_NODE] = \
3452
        _GetWantedNodes(self, self.op.nodes)
3453
    else:
3454
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3455

    
3456
  def Exec(self, feedback_fn):
3457
    """Computes the list of nodes and their attributes.
3458

3459
    """
3460
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
3461

    
3462
    # Always get name to sort by
3463
    if constants.SF_NAME in self.op.output_fields:
3464
      fields = self.op.output_fields[:]
3465
    else:
3466
      fields = [constants.SF_NAME] + self.op.output_fields
3467

    
3468
    # Never ask for node or type as it's only known to the LU
3469
    for extra in [constants.SF_NODE, constants.SF_TYPE]:
3470
      while extra in fields:
3471
        fields.remove(extra)
3472

    
3473
    field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
3474
    name_idx = field_idx[constants.SF_NAME]
3475

    
3476
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3477
    data = self.rpc.call_storage_list(self.nodes,
3478
                                      self.op.storage_type, st_args,
3479
                                      self.op.name, fields)
3480

    
3481
    result = []
3482

    
3483
    for node in utils.NiceSort(self.nodes):
3484
      nresult = data[node]
3485
      if nresult.offline:
3486
        continue
3487

    
3488
      msg = nresult.fail_msg
3489
      if msg:
3490
        self.LogWarning("Can't get storage data from node %s: %s", node, msg)
3491
        continue
3492

    
3493
      rows = dict([(row[name_idx], row) for row in nresult.payload])
3494

    
3495
      for name in utils.NiceSort(rows.keys()):
3496
        row = rows[name]
3497

    
3498
        out = []
3499

    
3500
        for field in self.op.output_fields:
3501
          if field == constants.SF_NODE:
3502
            val = node
3503
          elif field == constants.SF_TYPE:
3504
            val = self.op.storage_type
3505
          elif field in field_idx:
3506
            val = row[field_idx[field]]
3507
          else:
3508
            raise errors.ParameterError(field)
3509

    
3510
          out.append(val)
3511

    
3512
        result.append(out)
3513

    
3514
    return result
3515

    
3516

    
3517
class LUModifyNodeStorage(NoHooksLU):
3518
  """Logical unit for modifying a storage volume on a node.
3519

3520
  """
3521
  _OP_PARAMS = [
3522
    _PNodeName,
3523
    ("storage_type", ht.NoDefault, _CheckStorageType),
3524
    ("name", ht.NoDefault, ht.TNonEmptyString),
3525
    ("changes", ht.NoDefault, ht.TDict),
3526
    ]
3527
  REQ_BGL = False
3528

    
3529
  def CheckArguments(self):
3530
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3531

    
3532
    storage_type = self.op.storage_type
3533

    
3534
    try:
3535
      modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
3536
    except KeyError:
3537
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
3538
                                 " modified" % storage_type,
3539
                                 errors.ECODE_INVAL)
3540

    
3541
    diff = set(self.op.changes.keys()) - modifiable
3542
    if diff:
3543
      raise errors.OpPrereqError("The following fields can not be modified for"
3544
                                 " storage units of type '%s': %r" %
3545
                                 (storage_type, list(diff)),
3546
                                 errors.ECODE_INVAL)
3547

    
3548
  def ExpandNames(self):
3549
    self.needed_locks = {
3550
      locking.LEVEL_NODE: self.op.node_name,
3551
      }
3552

    
3553
  def Exec(self, feedback_fn):
3554
    """Computes the list of nodes and their attributes.
3555

3556
    """
3557
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3558
    result = self.rpc.call_storage_modify(self.op.node_name,
3559
                                          self.op.storage_type, st_args,
3560
                                          self.op.name, self.op.changes)
3561
    result.Raise("Failed to modify storage unit '%s' on %s" %
3562
                 (self.op.name, self.op.node_name))
3563

    
3564

    
3565
class LUAddNode(LogicalUnit):
3566
  """Logical unit for adding node to the cluster.
3567

3568
  """
3569
  HPATH = "node-add"
3570
  HTYPE = constants.HTYPE_NODE
3571
  _OP_PARAMS = [
3572
    _PNodeName,
3573
    ("primary_ip", None, ht.NoType),
3574
    ("secondary_ip", None, ht.TMaybeString),
3575
    ("readd", False, ht.TBool),
3576
    ("group", None, ht.TMaybeString)
3577
    ]
3578

    
3579
  def CheckArguments(self):
3580
    self.primary_ip_family = self.cfg.GetPrimaryIPFamily()
3581
    # validate/normalize the node name
3582
    self.hostname = netutils.GetHostname(name=self.op.node_name,
3583
                                         family=self.primary_ip_family)
3584
    self.op.node_name = self.hostname.name
3585
    if self.op.readd and self.op.group:
3586
      raise errors.OpPrereqError("Cannot pass a node group when a node is"
3587
                                 " being readded", errors.ECODE_INVAL)
3588

    
3589
  def BuildHooksEnv(self):
3590
    """Build hooks env.
3591

3592
    This will run on all nodes before, and on all nodes + the new node after.
3593

3594
    """
3595
    env = {
3596
      "OP_TARGET": self.op.node_name,
3597
      "NODE_NAME": self.op.node_name,
3598
      "NODE_PIP": self.op.primary_ip,
3599
      "NODE_SIP": self.op.secondary_ip,
3600
      }
3601
    nodes_0 = self.cfg.GetNodeList()
3602
    nodes_1 = nodes_0 + [self.op.node_name, ]
3603
    return env, nodes_0, nodes_1
3604

    
3605
  def CheckPrereq(self):
3606
    """Check prerequisites.
3607

3608
    This checks:
3609
     - the new node is not already in the config
3610
     - it is resolvable
3611
     - its parameters (single/dual homed) matches the cluster
3612

3613
    Any errors are signaled by raising errors.OpPrereqError.
3614

3615
    """
3616
    cfg = self.cfg
3617
    hostname = self.hostname
3618
    node = hostname.name
3619
    primary_ip = self.op.primary_ip = hostname.ip
3620
    if self.op.secondary_ip is None:
3621
      if self.primary_ip_family == netutils.IP6Address.family:
3622
        raise errors.OpPrereqError("When using a IPv6 primary address, a valid"
3623
                                   " IPv4 address must be given as secondary",
3624
                                   errors.ECODE_INVAL)
3625
      self.op.secondary_ip = primary_ip
3626

    
3627
    secondary_ip = self.op.secondary_ip
3628
    if not netutils.IP4Address.IsValid(secondary_ip):
3629
      raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
3630
                                 " address" % secondary_ip, errors.ECODE_INVAL)
3631

    
3632
    node_list = cfg.GetNodeList()
3633
    if not self.op.readd and node in node_list:
3634
      raise errors.OpPrereqError("Node %s is already in the configuration" %
3635
                                 node, errors.ECODE_EXISTS)
3636
    elif self.op.readd and node not in node_list:
3637
      raise errors.OpPrereqError("Node %s is not in the configuration" % node,
3638
                                 errors.ECODE_NOENT)
3639

    
3640
    self.changed_primary_ip = False
3641

    
3642
    for existing_node_name in node_list:
3643
      existing_node = cfg.GetNodeInfo(existing_node_name)
3644

    
3645
      if self.op.readd and node == existing_node_name:
3646
        if existing_node.secondary_ip != secondary_ip:
3647
          raise errors.OpPrereqError("Readded node doesn't have the same IP"
3648
                                     " address configuration as before",
3649
                                     errors.ECODE_INVAL)
3650
        if existing_node.primary_ip != primary_ip:
3651
          self.changed_primary_ip = True
3652

    
3653
        continue
3654

    
3655
      if (existing_node.primary_ip == primary_ip or
3656
          existing_node.secondary_ip == primary_ip or
3657
          existing_node.primary_ip == secondary_ip or
3658
          existing_node.secondary_ip == secondary_ip):
3659
        raise errors.OpPrereqError("New node ip address(es) conflict with"
3660
                                   " existing node %s" % existing_node.name,
3661
                                   errors.ECODE_NOTUNIQUE)
3662

    
3663
    # check that the type of the node (single versus dual homed) is the
3664
    # same as for the master
3665
    myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
3666
    master_singlehomed = myself.secondary_ip == myself.primary_ip
3667
    newbie_singlehomed = secondary_ip == primary_ip
3668
    if master_singlehomed != newbie_singlehomed:
3669
      if master_singlehomed:
3670
        raise errors.OpPrereqError("The master has no private ip but the"
3671
                                   " new node has one",
3672
                                   errors.ECODE_INVAL)
3673
      else:
3674
        raise errors.OpPrereqError("The master has a private ip but the"
3675
                                   " new node doesn't have one",
3676
                                   errors.ECODE_INVAL)
3677

    
3678
    # checks reachability
3679
    if not netutils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
3680
      raise errors.OpPrereqError("Node not reachable by ping",
3681
                                 errors.ECODE_ENVIRON)
3682

    
3683
    if not newbie_singlehomed:
3684
      # check reachability from my secondary ip to newbie's secondary ip
3685
      if not netutils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
3686
                           source=myself.secondary_ip):
3687
        raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
3688
                                   " based ping to noded port",
3689
                                   errors.ECODE_ENVIRON)
3690

    
3691
    if self.op.readd:
3692
      exceptions = [node]
3693
    else:
3694
      exceptions = []
3695

    
3696
    self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
3697

    
3698
    if self.op.readd:
3699
      self.new_node = self.cfg.GetNodeInfo(node)
3700
      assert self.new_node is not None, "Can't retrieve locked node %s" % node
3701
    else:
3702
      node_group = cfg.LookupNodeGroup(self.op.group)
3703
      self.new_node = objects.Node(name=node,
3704
                                   primary_ip=primary_ip,
3705
                                   secondary_ip=secondary_ip,
3706
                                   master_candidate=self.master_candidate,
3707
                                   master_capable=True,
3708
                                   vm_capable=True,
3709
                                   offline=False, drained=False,
3710
                                   group=node_group)
3711

    
3712
  def Exec(self, feedback_fn):
3713
    """Adds the new node to the cluster.
3714

3715
    """
3716
    new_node = self.new_node
3717
    node = new_node.name
3718

    
3719
    # for re-adds, reset the offline/drained/master-candidate flags;
3720
    # we need to reset here, otherwise offline would prevent RPC calls
3721
    # later in the procedure; this also means that if the re-add
3722
    # fails, we are left with a non-offlined, broken node
3723
    if self.op.readd:
3724
      new_node.drained = new_node.offline = False # pylint: disable-msg=W0201
3725
      self.LogInfo("Readding a node, the offline/drained flags were reset")
3726
      # if we demote the node, we do cleanup later in the procedure
3727
      new_node.master_candidate = self.master_candidate
3728
      if self.changed_primary_ip:
3729
        new_node.primary_ip = self.op.primary_ip
3730

    
3731
    # notify the user about any possible mc promotion
3732
    if new_node.master_candidate:
3733
      self.LogInfo("Node will be a master candidate")
3734

    
3735
    # check connectivity
3736
    result = self.rpc.call_version([node])[node]
3737
    result.Raise("Can't get version information from node %s" % node)
3738
    if constants.PROTOCOL_VERSION == result.payload:
3739
      logging.info("Communication to node %s fine, sw version %s match",
3740
                   node, result.payload)
3741
    else:
3742
      raise errors.OpExecError("Version mismatch master version %s,"
3743
                               " node version %s" %
3744
                               (constants.PROTOCOL_VERSION, result.payload))
3745

    
3746
    # Add node to our /etc/hosts, and add key to known_hosts
3747
    if self.cfg.GetClusterInfo().modify_etc_hosts:
3748
      master_node = self.cfg.GetMasterNode()
3749
      result = self.rpc.call_etc_hosts_modify(master_node,
3750
                                              constants.ETC_HOSTS_ADD,
3751
                                              self.hostname.name,
3752
                                              self.hostname.ip)
3753
      result.Raise("Can't update hosts file with new host data")
3754

    
3755
    if new_node.secondary_ip != new_node.primary_ip:
3756
      result = self.rpc.call_node_has_ip_address(new_node.name,
3757
                                                 new_node.secondary_ip)
3758
      result.Raise("Failure checking secondary ip on node %s" % new_node.name,
3759
                   prereq=True, ecode=errors.ECODE_ENVIRON)
3760
      if not result.payload:
3761
        raise errors.OpExecError("Node claims it doesn't have the secondary ip"
3762
                                 " you gave (%s). Please fix and re-run this"
3763
                                 " command." % new_node.secondary_ip)
3764

    
3765
    node_verify_list = [self.cfg.GetMasterNode()]
3766
    node_verify_param = {
3767
      constants.NV_NODELIST: [node],
3768
      # TODO: do a node-net-test as well?
3769
    }
3770

    
3771
    result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
3772
                                       self.cfg.GetClusterName())
3773
    for verifier in node_verify_list:
3774
      result[verifier].Raise("Cannot communicate with node %s" % verifier)
3775
      nl_payload = result[verifier].payload[constants.NV_NODELIST]
3776
      if nl_payload:
3777
        for failed in nl_payload:
3778
          feedback_fn("ssh/hostname verification failed"
3779
                      " (checking from %s): %s" %
3780
                      (verifier, nl_payload[failed]))
3781
        raise errors.OpExecError("ssh/hostname verification failed.")
3782

    
3783
    if self.op.readd:
3784
      _RedistributeAncillaryFiles(self)
3785
      self.context.ReaddNode(new_node)
3786
      # make sure we redistribute the config
3787
      self.cfg.Update(new_node, feedback_fn)
3788
      # and make sure the new node will not have old files around
3789
      if not new_node.master_candidate:
3790
        result = self.rpc.call_node_demote_from_mc(new_node.name)
3791
        msg = result.fail_msg
3792
        if msg:
3793
          self.LogWarning("Node failed to demote itself from master"
3794
                          " candidate status: %s" % msg)
3795
    else:
3796
      _RedistributeAncillaryFiles(self, additional_nodes=[node])
3797
      self.context.AddNode(new_node, self.proc.GetECId())
3798

    
3799

    
3800
class LUSetNodeParams(LogicalUnit):
3801
  """Modifies the parameters of a node.
3802

3803
  @cvar _F2R: a dictionary from tuples of flags (mc, drained, offline)
3804
      to the node role (as _ROLE_*)
3805
  @cvar _R2F: a dictionary from node role to tuples of flags
3806
  @cvar _FLAGS: a list of attribute names corresponding to the flags
3807

3808
  """
3809
  HPATH = "node-modify"
3810
  HTYPE = constants.HTYPE_NODE
3811
  _OP_PARAMS = [
3812
    _PNodeName,
3813
    ("master_candidate", None, ht.TMaybeBool),
3814
    ("offline", None, ht.TMaybeBool),
3815
    ("drained", None, ht.TMaybeBool),
3816
    ("auto_promote", False, ht.TBool),
3817
    _PForce,
3818
    ]
3819
  REQ_BGL = False
3820
  (_ROLE_CANDIDATE, _ROLE_DRAINED, _ROLE_OFFLINE, _ROLE_REGULAR) = range(4)
3821
  _F2R = {
3822
    (True, False, False): _ROLE_CANDIDATE,
3823
    (False, True, False): _ROLE_DRAINED,
3824
    (False, False, True): _ROLE_OFFLINE,
3825
    (False, False, False): _ROLE_REGULAR,
3826
    }
3827
  _R2F = dict((v, k) for k, v in _F2R.items())
3828
  _FLAGS = ["master_candidate", "drained", "offline"]
3829

    
3830
  def CheckArguments(self):
3831
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3832
    all_mods = [self.op.offline, self.op.master_candidate, self.op.drained]
3833
    if all_mods.count(None) == len(all_mods):
3834
      raise errors.OpPrereqError("Please pass at least one modification",
3835
                                 errors.ECODE_INVAL)
3836
    if all_mods.count(True) > 1:
3837
      raise errors.OpPrereqError("Can't set the node into more than one"
3838
                                 " state at the same time",
3839
                                 errors.ECODE_INVAL)
3840

    
3841
    # Boolean value that tells us whether we might be demoting from MC
3842
    self.might_demote = (self.op.master_candidate == False or
3843
                         self.op.offline == True or
3844
                         self.op.drained == True)
3845

    
3846
    self.lock_all = self.op.auto_promote and self.might_demote
3847

    
3848
  def ExpandNames(self):
3849
    if self.lock_all:
3850
      self.needed_locks = {locking.LEVEL_NODE: locking.ALL_SET}
3851
    else:
3852
      self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
3853

    
3854
  def BuildHooksEnv(self):
3855
    """Build hooks env.
3856

3857
    This runs on the master node.
3858

3859
    """
3860
    env = {
3861
      "OP_TARGET": self.op.node_name,
3862
      "MASTER_CANDIDATE": str(self.op.master_candidate),
3863
      "OFFLINE": str(self.op.offline),
3864
      "DRAINED": str(self.op.drained),
3865
      }
3866
    nl = [self.cfg.GetMasterNode(),
3867
          self.op.node_name]
3868
    return env, nl, nl
3869

    
3870
  def CheckPrereq(self):
3871
    """Check prerequisites.
3872

3873
    This only checks the instance list against the existing names.
3874

3875
    """
3876
    node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
3877

    
3878
    if (self.op.master_candidate is not None or
3879
        self.op.drained is not None or
3880
        self.op.offline is not None):
3881
      # we can't change the master's node flags
3882
      if self.op.node_name == self.cfg.GetMasterNode():
3883
        raise errors.OpPrereqError("The master role can be changed"
3884
                                   " only via master-failover",
3885
                                   errors.ECODE_INVAL)
3886

    
3887

    
3888
    if node.master_candidate and self.might_demote and not self.lock_all:
3889
      assert not self.op.auto_promote, "auto-promote set but lock_all not"
3890
      # check if after removing the current node, we're missing master
3891
      # candidates
3892
      (mc_remaining, mc_should, _) = \
3893
          self.cfg.GetMasterCandidateStats(exceptions=[node.name])
3894
      if mc_remaining < mc_should:
3895
        raise errors.OpPrereqError("Not enough master candidates, please"
3896
                                   " pass auto_promote to allow promotion",
3897
                                   errors.ECODE_STATE)
3898

    
3899
    self.old_flags = old_flags = (node.master_candidate,
3900
                                  node.drained, node.offline)
3901
    assert old_flags in self._F2R, "Un-handled old flags  %s" % str(old_flags)
3902
    self.old_role = self._F2R[old_flags]
3903

    
3904
    # Check for ineffective changes
3905
    for attr in self._FLAGS:
3906
      if (getattr(self.op, attr) == False and getattr(node, attr) == False):
3907
        self.LogInfo("Ignoring request to unset flag %s, already unset", attr)
3908
        setattr(self.op, attr, None)
3909

    
3910
    # Past this point, any flag change to False means a transition
3911
    # away from the respective state, as only real changes are kept
3912

    
3913
    # If we're being deofflined/drained, we'll MC ourself if needed
3914
    if self.op.drained == False or self.op.offline == False:
3915
      if _DecideSelfPromotion(self):
3916
        self.op.master_candidate = True
3917
        self.LogInfo("Auto-promoting node to master candidate")
3918

    
3919
  def Exec(self, feedback_fn):
3920
    """Modifies a node.
3921

3922
    """
3923
    node = self.node
3924
    old_role = self.old_role
3925

    
3926
    assert [getattr(self.op, attr) for attr in self._FLAGS].count(True) <= 1
3927

    
3928
    # compute new flags
3929
    if self.op.master_candidate:
3930
      new_role = self._ROLE_CANDIDATE
3931
    elif self.op.drained:
3932
      new_role = self._ROLE_DRAINED
3933
    elif self.op.offline:
3934
      new_role = self._ROLE_OFFLINE
3935
    elif False in [self.op.master_candidate, self.op.drained, self.op.offline]:
3936
      # False is still in new flags, which means we're un-setting (the
3937
      # only) True flag
3938
      new_role = self._ROLE_REGULAR
3939
    else: # no new flags, nothing, keep old role
3940
      new_role = old_role
3941

    
3942
    result = []
3943
    changed_mc = [old_role, new_role].count(self._ROLE_CANDIDATE) == 1
3944

    
3945
    # Tell the node to demote itself, if no longer MC and not offline
3946
    if (old_role == self._ROLE_CANDIDATE and
3947
        new_role != self._ROLE_OFFLINE and new_role != old_role):
3948
      msg = self.rpc.call_node_demote_from_mc(node.name).fail_msg
3949
      if msg:
3950
        self.LogWarning("Node failed to demote itself: %s", msg)
3951

    
3952
    new_flags = self._R2F[new_role]
3953
    for of, nf, desc in zip(self.old_flags, new_flags, self._FLAGS):
3954
      if of != nf:
3955
        result.append((desc, str(nf)))
3956
    (node.master_candidate, node.drained, node.offline) = new_flags
3957

    
3958
    # we locked all nodes, we adjust the CP before updating this node
3959
    if self.lock_all:
3960
      _AdjustCandidatePool(self, [node.name])
3961

    
3962
    # this will trigger configuration file update, if needed
3963
    self.cfg.Update(node, feedback_fn)
3964

    
3965
    # this will trigger job queue propagation or cleanup
3966
    if changed_mc:
3967
      self.context.ReaddNode(node)
3968

    
3969
    return result
3970

    
3971

    
3972
class LUPowercycleNode(NoHooksLU):
3973
  """Powercycles a node.
3974

3975
  """
3976
  _OP_PARAMS = [
3977
    _PNodeName,
3978
    _PForce,
3979
    ]
3980
  REQ_BGL = False
3981

    
3982
  def CheckArguments(self):
3983
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3984
    if self.op.node_name == self.cfg.GetMasterNode() and not self.op.force:
3985
      raise errors.OpPrereqError("The node is the master and the force"
3986
                                 " parameter was not set",
3987
                                 errors.ECODE_INVAL)
3988

    
3989
  def ExpandNames(self):
3990
    """Locking for PowercycleNode.
3991

3992
    This is a last-resort option and shouldn't block on other
3993
    jobs. Therefore, we grab no locks.
3994

3995
    """
3996
    self.needed_locks = {}
3997

    
3998
  def Exec(self, feedback_fn):
3999
    """Reboots a node.
4000

4001
    """
4002
    result = self.rpc.call_node_powercycle(self.op.node_name,
4003
                                           self.cfg.GetHypervisorType())
4004
    result.Raise("Failed to schedule the reboot")
4005
    return result.payload
4006

    
4007

    
4008
class LUQueryClusterInfo(NoHooksLU):
4009
  """Query cluster configuration.
4010

4011
  """
4012
  REQ_BGL = False
4013

    
4014
  def ExpandNames(self):
4015
    self.needed_locks = {}
4016

    
4017
  def Exec(self, feedback_fn):
4018
    """Return cluster config.
4019

4020
    """
4021
    cluster = self.cfg.GetClusterInfo()
4022
    os_hvp = {}
4023

    
4024
    # Filter just for enabled hypervisors
4025
    for os_name, hv_dict in cluster.os_hvp.items():
4026
      os_hvp[os_name] = {}
4027
      for hv_name, hv_params in hv_dict.items():
4028
        if hv_name in cluster.enabled_hypervisors:
4029
          os_hvp[os_name][hv_name] = hv_params
4030

    
4031
    # Convert ip_family to ip_version
4032
    primary_ip_version = constants.IP4_VERSION
4033
    if cluster.primary_ip_family == netutils.IP6Address.family:
4034
      primary_ip_version = constants.IP6_VERSION
4035

    
4036
    result = {
4037
      "software_version": constants.RELEASE_VERSION,
4038
      "protocol_version": constants.PROTOCOL_VERSION,
4039
      "config_version": constants.CONFIG_VERSION,
4040
      "os_api_version": max(constants.OS_API_VERSIONS),
4041
      "export_version": constants.EXPORT_VERSION,
4042
      "architecture": (platform.architecture()[0], platform.machine()),
4043
      "name": cluster.cluster_name,
4044
      "master": cluster.master_node,
4045
      "default_hypervisor": cluster.enabled_hypervisors[0],
4046
      "enabled_hypervisors": cluster.enabled_hypervisors,
4047
      "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
4048
                        for hypervisor_name in cluster.enabled_hypervisors]),
4049
      "os_hvp": os_hvp,
4050
      "beparams": cluster.beparams,
4051
      "osparams": cluster.osparams,
4052
      "nicparams": cluster.nicparams,
4053
      "candidate_pool_size": cluster.candidate_pool_size,
4054
      "master_netdev": cluster.master_netdev,
4055
      "volume_group_name": cluster.volume_group_name,
4056
      "drbd_usermode_helper": cluster.drbd_usermode_helper,
4057
      "file_storage_dir": cluster.file_storage_dir,
4058
      "maintain_node_health": cluster.maintain_node_health,
4059
      "ctime": cluster.ctime,
4060
      "mtime": cluster.mtime,
4061
      "uuid": cluster.uuid,
4062
      "tags": list(cluster.GetTags()),
4063
      "uid_pool": cluster.uid_pool,
4064
      "default_iallocator": cluster.default_iallocator,
4065
      "reserved_lvs": cluster.reserved_lvs,
4066
      "primary_ip_version": primary_ip_version,
4067
      "prealloc_wipe_disks": cluster.prealloc_wipe_disks,
4068
      }
4069

    
4070
    return result
4071

    
4072

    
4073
class LUQueryConfigValues(NoHooksLU):
4074
  """Return configuration values.
4075

4076
  """
4077
  _OP_PARAMS = [_POutputFields]
4078
  REQ_BGL = False
4079
  _FIELDS_DYNAMIC = utils.FieldSet()
4080
  _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag",
4081
                                  "watcher_pause", "volume_group_name")
4082

    
4083
  def CheckArguments(self):
4084
    _CheckOutputFields(static=self._FIELDS_STATIC,
4085
                       dynamic=self._FIELDS_DYNAMIC,
4086
                       selected=self.op.output_fields)
4087

    
4088
  def ExpandNames(self):
4089
    self.needed_locks = {}
4090

    
4091
  def Exec(self, feedback_fn):
4092
    """Dump a representation of the cluster config to the standard output.
4093

4094
    """
4095
    values = []
4096
    for field in self.op.output_fields:
4097
      if field == "cluster_name":
4098
        entry = self.cfg.GetClusterName()
4099
      elif field == "master_node":
4100
        entry = self.cfg.GetMasterNode()
4101
      elif field == "drain_flag":
4102
        entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
4103
      elif field == "watcher_pause":
4104
        entry = utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
4105
      elif field == "volume_group_name":
4106
        entry = self.cfg.GetVGName()
4107
      else:
4108
        raise errors.ParameterError(field)
4109
      values.append(entry)
4110
    return values
4111

    
4112

    
4113
class LUActivateInstanceDisks(NoHooksLU):
4114
  """Bring up an instance's disks.
4115

4116
  """
4117
  _OP_PARAMS = [
4118
    _PInstanceName,
4119
    ("ignore_size", False, ht.TBool),
4120
    ]
4121
  REQ_BGL = False
4122

    
4123
  def ExpandNames(self):
4124
    self._ExpandAndLockInstance()
4125
    self.needed_locks[locking.LEVEL_NODE] = []
4126
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4127

    
4128
  def DeclareLocks(self, level):
4129
    if level == locking.LEVEL_NODE:
4130
      self._LockInstancesNodes()
4131

    
4132
  def CheckPrereq(self):
4133
    """Check prerequisites.
4134

4135
    This checks that the instance is in the cluster.
4136

4137
    """
4138
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4139
    assert self.instance is not None, \
4140
      "Cannot retrieve locked instance %s" % self.op.instance_name
4141
    _CheckNodeOnline(self, self.instance.primary_node)
4142

    
4143
  def Exec(self, feedback_fn):
4144
    """Activate the disks.
4145

4146
    """
4147
    disks_ok, disks_info = \
4148
              _AssembleInstanceDisks(self, self.instance,
4149
                                     ignore_size=self.op.ignore_size)
4150
    if not disks_ok:
4151
      raise errors.OpExecError("Cannot activate block devices")
4152

    
4153
    return disks_info
4154

    
4155

    
4156
def _AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False,
4157
                           ignore_size=False):
4158
  """Prepare the block devices for an instance.
4159

4160
  This sets up the block devices on all nodes.
4161

4162
  @type lu: L{LogicalUnit}
4163
  @param lu: the logical unit on whose behalf we execute
4164
  @type instance: L{objects.Instance}
4165
  @param instance: the instance for whose disks we assemble
4166
  @type disks: list of L{objects.Disk} or None
4167
  @param disks: which disks to assemble (or all, if None)
4168
  @type ignore_secondaries: boolean
4169
  @param ignore_secondaries: if true, errors on secondary nodes
4170
      won't result in an error return from the function
4171
  @type ignore_size: boolean
4172
  @param ignore_size: if true, the current known size of the disk
4173
      will not be used during the disk activation, useful for cases
4174
      when the size is wrong
4175
  @return: False if the operation failed, otherwise a list of
4176
      (host, instance_visible_name, node_visible_name)
4177
      with the mapping from node devices to instance devices
4178

4179
  """
4180
  device_info = []
4181
  disks_ok = True
4182
  iname = instance.name
4183
  disks = _ExpandCheckDisks(instance, disks)
4184

    
4185
  # With the two passes mechanism we try to reduce the window of
4186
  # opportunity for the race condition of switching DRBD to primary
4187
  # before handshaking occured, but we do not eliminate it
4188

    
4189
  # The proper fix would be to wait (with some limits) until the
4190
  # connection has been made and drbd transitions from WFConnection
4191
  # into any other network-connected state (Connected, SyncTarget,
4192
  # SyncSource, etc.)
4193

    
4194
  # 1st pass, assemble on all nodes in secondary mode
4195
  for inst_disk in disks:
4196
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
4197
      if ignore_size:
4198
        node_disk = node_disk.Copy()
4199
        node_disk.UnsetSize()
4200
      lu.cfg.SetDiskID(node_disk, node)
4201
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
4202
      msg = result.fail_msg
4203
      if msg:
4204
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
4205
                           " (is_primary=False, pass=1): %s",
4206
                           inst_disk.iv_name, node, msg)
4207
        if not ignore_secondaries:
4208
          disks_ok = False
4209

    
4210
  # FIXME: race condition on drbd migration to primary
4211

    
4212
  # 2nd pass, do only the primary node
4213
  for inst_disk in disks:
4214
    dev_path = None
4215

    
4216
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
4217
      if node != instance.primary_node:
4218
        continue
4219
      if ignore_size:
4220
        node_disk = node_disk.Copy()
4221
        node_disk.UnsetSize()
4222
      lu.cfg.SetDiskID(node_disk, node)
4223
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
4224
      msg = result.fail_msg
4225
      if msg:
4226
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
4227
                           " (is_primary=True, pass=2): %s",
4228
                           inst_disk.iv_name, node, msg)
4229
        disks_ok = False
4230
      else:
4231
        dev_path = result.payload
4232

    
4233
    device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
4234

    
4235
  # leave the disks configured for the primary node
4236
  # this is a workaround that would be fixed better by
4237
  # improving the logical/physical id handling
4238
  for disk in disks:
4239
    lu.cfg.SetDiskID(disk, instance.primary_node)
4240

    
4241
  return disks_ok, device_info
4242

    
4243

    
4244
def _StartInstanceDisks(lu, instance, force):
4245
  """Start the disks of an instance.
4246

4247
  """
4248
  disks_ok, _ = _AssembleInstanceDisks(lu, instance,
4249
                                           ignore_secondaries=force)
4250
  if not disks_ok:
4251
    _ShutdownInstanceDisks(lu, instance)
4252
    if force is not None and not force:
4253
      lu.proc.LogWarning("", hint="If the message above refers to a"
4254
                         " secondary node,"
4255
                         " you can retry the operation using '--force'.")
4256
    raise errors.OpExecError("Disk consistency error")
4257

    
4258

    
4259
class LUDeactivateInstanceDisks(NoHooksLU):
4260
  """Shutdown an instance's disks.
4261

4262
  """
4263
  _OP_PARAMS = [
4264
    _PInstanceName,
4265
    ]
4266
  REQ_BGL = False
4267

    
4268
  def ExpandNames(self):
4269
    self._ExpandAndLockInstance()
4270
    self.needed_locks[locking.LEVEL_NODE] = []
4271
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4272

    
4273
  def DeclareLocks(self, level):
4274
    if level == locking.LEVEL_NODE:
4275
      self._LockInstancesNodes()
4276

    
4277
  def CheckPrereq(self):
4278
    """Check prerequisites.
4279

4280
    This checks that the instance is in the cluster.
4281

4282
    """
4283
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4284
    assert self.instance is not None, \
4285
      "Cannot retrieve locked instance %s" % self.op.instance_name
4286

    
4287
  def Exec(self, feedback_fn):
4288
    """Deactivate the disks
4289

4290
    """
4291
    instance = self.instance
4292
    _SafeShutdownInstanceDisks(self, instance)
4293

    
4294

    
4295
def _SafeShutdownInstanceDisks(lu, instance, disks=None):
4296
  """Shutdown block devices of an instance.
4297

4298
  This function checks if an instance is running, before calling
4299
  _ShutdownInstanceDisks.
4300

4301
  """
4302
  _CheckInstanceDown(lu, instance, "cannot shutdown disks")
4303
  _ShutdownInstanceDisks(lu, instance, disks=disks)
4304

    
4305

    
4306
def _ExpandCheckDisks(instance, disks):
4307
  """Return the instance disks selected by the disks list
4308

4309
  @type disks: list of L{objects.Disk} or None
4310
  @param disks: selected disks
4311
  @rtype: list of L{objects.Disk}
4312
  @return: selected instance disks to act on
4313

4314
  """
4315
  if disks is None:
4316
    return instance.disks
4317
  else:
4318
    if not set(disks).issubset(instance.disks):
4319
      raise errors.ProgrammerError("Can only act on disks belonging to the"
4320
                                   " target instance")
4321
    return disks
4322

    
4323

    
4324
def _ShutdownInstanceDisks(lu, instance, disks=None, ignore_primary=False):
4325
  """Shutdown block devices of an instance.
4326

4327
  This does the shutdown on all nodes of the instance.
4328

4329
  If the ignore_primary is false, errors on the primary node are
4330
  ignored.
4331

4332
  """
4333
  all_result = True
4334
  disks = _ExpandCheckDisks(instance, disks)
4335

    
4336
  for disk in disks:
4337
    for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
4338
      lu.cfg.SetDiskID(top_disk, node)
4339
      result = lu.rpc.call_blockdev_shutdown(node, top_disk)
4340
      msg = result.fail_msg
4341
      if msg:
4342
        lu.LogWarning("Could not shutdown block device %s on node %s: %s",
4343
                      disk.iv_name, node, msg)
4344
        if not ignore_primary or node != instance.primary_node:
4345
          all_result = False
4346
  return all_result
4347

    
4348

    
4349
def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
4350
  """Checks if a node has enough free memory.
4351

4352
  This function check if a given node has the needed amount of free
4353
  memory. In case the node has less memory or we cannot get the
4354
  information from the node, this function raise an OpPrereqError
4355
  exception.
4356

4357
  @type lu: C{LogicalUnit}
4358
  @param lu: a logical unit from which we get configuration data
4359
  @type node: C{str}
4360
  @param node: the node to check
4361
  @type reason: C{str}
4362
  @param reason: string to use in the error message
4363
  @type requested: C{int}
4364
  @param requested: the amount of memory in MiB to check for
4365
  @type hypervisor_name: C{str}
4366
  @param hypervisor_name: the hypervisor to ask for memory stats
4367
  @raise errors.OpPrereqError: if the node doesn't have enough memory, or
4368
      we cannot check the node
4369

4370
  """
4371
  nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
4372
  nodeinfo[node].Raise("Can't get data from node %s" % node,
4373
                       prereq=True, ecode=errors.ECODE_ENVIRON)
4374
  free_mem = nodeinfo[node].payload.get('memory_free', None)
4375
  if not isinstance(free_mem, int):
4376
    raise errors.OpPrereqError("Can't compute free memory on node %s, result"
4377
                               " was '%s'" % (node, free_mem),
4378
                               errors.ECODE_ENVIRON)
4379
  if requested > free_mem:
4380
    raise errors.OpPrereqError("Not enough memory on node %s for %s:"
4381
                               " needed %s MiB, available %s MiB" %
4382
                               (node, reason, requested, free_mem),
4383
                               errors.ECODE_NORES)
4384

    
4385

    
4386
def _CheckNodesFreeDisk(lu, nodenames, requested):
4387
  """Checks if nodes have enough free disk space in the default VG.
4388

4389
  This function check if all given nodes have the needed amount of
4390
  free disk. In case any node has less disk or we cannot get the
4391
  information from the node, this function raise an OpPrereqError
4392
  exception.
4393

4394
  @type lu: C{LogicalUnit}
4395
  @param lu: a logical unit from which we get configuration data
4396
  @type nodenames: C{list}
4397
  @param nodenames: the list of node names to check
4398
  @type requested: C{int}
4399
  @param requested: the amount of disk in MiB to check for
4400
  @raise errors.OpPrereqError: if the node doesn't have enough disk, or
4401
      we cannot check the node
4402

4403
  """
4404
  nodeinfo = lu.rpc.call_node_info(nodenames, lu.cfg.GetVGName(),
4405
                                   lu.cfg.GetHypervisorType())
4406
  for node in nodenames:
4407
    info = nodeinfo[node]
4408
    info.Raise("Cannot get current information from node %s" % node,
4409
               prereq=True, ecode=errors.ECODE_ENVIRON)
4410
    vg_free = info.payload.get("vg_free", None)
4411
    if not isinstance(vg_free, int):
4412
      raise errors.OpPrereqError("Can't compute free disk space on node %s,"
4413
                                 " result was '%s'" % (node, vg_free),
4414
                                 errors.ECODE_ENVIRON)
4415
    if requested > vg_free:
4416
      raise errors.OpPrereqError("Not enough disk space on target node %s:"
4417
                                 " required %d MiB, available %d MiB" %
4418
                                 (node, requested, vg_free),
4419
                                 errors.ECODE_NORES)
4420

    
4421

    
4422
class LUStartupInstance(LogicalUnit):
4423
  """Starts an instance.
4424

4425
  """
4426
  HPATH = "instance-start"
4427
  HTYPE = constants.HTYPE_INSTANCE
4428
  _OP_PARAMS = [
4429
    _PInstanceName,
4430
    _PForce,
4431
    _PIgnoreOfflineNodes,
4432
    ("hvparams", ht.EmptyDict, ht.TDict),
4433
    ("beparams", ht.EmptyDict, ht.TDict),
4434
    ]
4435
  REQ_BGL = False
4436

    
4437
  def CheckArguments(self):
4438
    # extra beparams
4439
    if self.op.beparams:
4440
      # fill the beparams dict
4441
      utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
4442

    
4443
  def ExpandNames(self):
4444
    self._ExpandAndLockInstance()
4445

    
4446
  def BuildHooksEnv(self):
4447
    """Build hooks env.
4448

4449
    This runs on master, primary and secondary nodes of the instance.
4450

4451
    """
4452
    env = {
4453
      "FORCE": self.op.force,
4454
      }
4455
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4456
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4457
    return env, nl, nl
4458

    
4459
  def CheckPrereq(self):
4460
    """Check prerequisites.
4461

4462
    This checks that the instance is in the cluster.
4463

4464
    """
4465
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4466
    assert self.instance is not None, \
4467
      "Cannot retrieve locked instance %s" % self.op.instance_name
4468

    
4469
    # extra hvparams
4470
    if self.op.hvparams:
4471
      # check hypervisor parameter syntax (locally)
4472
      cluster = self.cfg.GetClusterInfo()
4473
      utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
4474
      filled_hvp = cluster.FillHV(instance)
4475
      filled_hvp.update(self.op.hvparams)
4476
      hv_type = hypervisor.GetHypervisor(instance.hypervisor)
4477
      hv_type.CheckParameterSyntax(filled_hvp)
4478
      _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
4479

    
4480
    self.primary_offline = self.cfg.GetNodeInfo(instance.primary_node).offline
4481

    
4482
    if self.primary_offline and self.op.ignore_offline_nodes:
4483
      self.proc.LogWarning("Ignoring offline primary node")
4484

    
4485
      if self.op.hvparams or self.op.beparams:
4486
        self.proc.LogWarning("Overridden parameters are ignored")
4487
    else:
4488
      _CheckNodeOnline(self, instance.primary_node)
4489

    
4490
      bep = self.cfg.GetClusterInfo().FillBE(instance)
4491

    
4492
      # check bridges existence
4493
      _CheckInstanceBridgesExist(self, instance)
4494

    
4495
      remote_info = self.rpc.call_instance_info(instance.primary_node,
4496
                                                instance.name,
4497
                                                instance.hypervisor)
4498
      remote_info.Raise("Error checking node %s" % instance.primary_node,
4499
                        prereq=True, ecode=errors.ECODE_ENVIRON)
4500
      if not remote_info.payload: # not running already
4501
        _CheckNodeFreeMemory(self, instance.primary_node,
4502
                             "starting instance %s" % instance.name,
4503
                             bep[constants.BE_MEMORY], instance.hypervisor)
4504

    
4505
  def Exec(self, feedback_fn):
4506
    """Start the instance.
4507

4508
    """
4509
    instance = self.instance
4510
    force = self.op.force
4511

    
4512
    self.cfg.MarkInstanceUp(instance.name)
4513

    
4514
    if self.primary_offline:
4515
      assert self.op.ignore_offline_nodes
4516
      self.proc.LogInfo("Primary node offline, marked instance as started")
4517
    else:
4518
      node_current = instance.primary_node
4519

    
4520
      _StartInstanceDisks(self, instance, force)
4521

    
4522
      result = self.rpc.call_instance_start(node_current, instance,
4523
                                            self.op.hvparams, self.op.beparams)
4524
      msg = result.fail_msg
4525
      if msg:
4526
        _ShutdownInstanceDisks(self, instance)
4527
        raise errors.OpExecError("Could not start instance: %s" % msg)
4528

    
4529

    
4530
class LURebootInstance(LogicalUnit):
4531
  """Reboot an instance.
4532

4533
  """
4534
  HPATH = "instance-reboot"
4535
  HTYPE = constants.HTYPE_INSTANCE
4536
  _OP_PARAMS = [
4537
    _PInstanceName,
4538
    ("ignore_secondaries", False, ht.TBool),
4539
    ("reboot_type", ht.NoDefault, ht.TElemOf(constants.REBOOT_TYPES)),
4540
    _PShutdownTimeout,
4541
    ]
4542
  REQ_BGL = False
4543

    
4544
  def ExpandNames(self):
4545
    self._ExpandAndLockInstance()
4546

    
4547
  def BuildHooksEnv(self):
4548
    """Build hooks env.
4549

4550
    This runs on master, primary and secondary nodes of the instance.
4551

4552
    """
4553
    env = {
4554
      "IGNORE_SECONDARIES": self.op.ignore_secondaries,
4555
      "REBOOT_TYPE": self.op.reboot_type,
4556
      "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
4557
      }
4558
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4559
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4560
    return env, nl, nl
4561

    
4562
  def CheckPrereq(self):
4563
    """Check prerequisites.
4564

4565
    This checks that the instance is in the cluster.
4566

4567
    """
4568
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4569
    assert self.instance is not None, \
4570
      "Cannot retrieve locked instance %s" % self.op.instance_name
4571

    
4572
    _CheckNodeOnline(self, instance.primary_node)
4573

    
4574
    # check bridges existence
4575
    _CheckInstanceBridgesExist(self, instance)
4576

    
4577
  def Exec(self, feedback_fn):
4578
    """Reboot the instance.
4579

4580
    """
4581
    instance = self.instance
4582
    ignore_secondaries = self.op.ignore_secondaries
4583
    reboot_type = self.op.reboot_type
4584

    
4585
    node_current = instance.primary_node
4586

    
4587
    if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
4588
                       constants.INSTANCE_REBOOT_HARD]:
4589
      for disk in instance.disks:
4590
        self.cfg.SetDiskID(disk, node_current)
4591
      result = self.rpc.call_instance_reboot(node_current, instance,
4592
                                             reboot_type,
4593
                                             self.op.shutdown_timeout)
4594
      result.Raise("Could not reboot instance")
4595
    else:
4596
      result = self.rpc.call_instance_shutdown(node_current, instance,
4597
                                               self.op.shutdown_timeout)
4598
      result.Raise("Could not shutdown instance for full reboot")
4599
      _ShutdownInstanceDisks(self, instance)
4600
      _StartInstanceDisks(self, instance, ignore_secondaries)
4601
      result = self.rpc.call_instance_start(node_current, instance, None, None)
4602
      msg = result.fail_msg
4603
      if msg:
4604
        _ShutdownInstanceDisks(self, instance)
4605
        raise errors.OpExecError("Could not start instance for"
4606
                                 " full reboot: %s" % msg)
4607

    
4608
    self.cfg.MarkInstanceUp(instance.name)
4609

    
4610

    
4611
class LUShutdownInstance(LogicalUnit):
4612
  """Shutdown an instance.
4613

4614
  """
4615
  HPATH = "instance-stop"
4616
  HTYPE = constants.HTYPE_INSTANCE
4617
  _OP_PARAMS = [
4618
    _PInstanceName,
4619
    _PIgnoreOfflineNodes,
4620
    ("timeout", constants.DEFAULT_SHUTDOWN_TIMEOUT, ht.TPositiveInt),
4621
    ]
4622
  REQ_BGL = False
4623

    
4624
  def ExpandNames(self):
4625
    self._ExpandAndLockInstance()
4626

    
4627
  def BuildHooksEnv(self):
4628
    """Build hooks env.
4629

4630
    This runs on master, primary and secondary nodes of the instance.
4631

4632
    """
4633
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4634
    env["TIMEOUT"] = self.op.timeout
4635
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4636
    return env, nl, nl
4637

    
4638
  def CheckPrereq(self):
4639
    """Check prerequisites.
4640

4641
    This checks that the instance is in the cluster.
4642

4643
    """
4644
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4645
    assert self.instance is not None, \
4646
      "Cannot retrieve locked instance %s" % self.op.instance_name
4647

    
4648
    self.primary_offline = \
4649
      self.cfg.GetNodeInfo(self.instance.primary_node).offline
4650

    
4651
    if self.primary_offline and self.op.ignore_offline_nodes:
4652
      self.proc.LogWarning("Ignoring offline primary node")
4653
    else:
4654
      _CheckNodeOnline(self, self.instance.primary_node)
4655

    
4656
  def Exec(self, feedback_fn):
4657
    """Shutdown the instance.
4658

4659
    """
4660
    instance = self.instance
4661
    node_current = instance.primary_node
4662
    timeout = self.op.timeout
4663

    
4664
    self.cfg.MarkInstanceDown(instance.name)
4665

    
4666
    if self.primary_offline:
4667
      assert self.op.ignore_offline_nodes
4668
      self.proc.LogInfo("Primary node offline, marked instance as stopped")
4669
    else:
4670
      result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
4671
      msg = result.fail_msg
4672
      if msg:
4673
        self.proc.LogWarning("Could not shutdown instance: %s" % msg)
4674

    
4675
      _ShutdownInstanceDisks(self, instance)
4676

    
4677

    
4678
class LUReinstallInstance(LogicalUnit):
4679
  """Reinstall an instance.
4680

4681
  """
4682
  HPATH = "instance-reinstall"
4683
  HTYPE = constants.HTYPE_INSTANCE
4684
  _OP_PARAMS = [
4685
    _PInstanceName,
4686
    ("os_type", None, ht.TMaybeString),
4687
    ("force_variant", False, ht.TBool),
4688
    ("osparams", None, ht.TOr(ht.TDict, ht.TNone)),
4689
    ]
4690
  REQ_BGL = False
4691

    
4692
  def ExpandNames(self):
4693
    self._ExpandAndLockInstance()
4694

    
4695
  def BuildHooksEnv(self):
4696
    """Build hooks env.
4697

4698
    This runs on master, primary and secondary nodes of the instance.
4699

4700
    """
4701
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4702
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4703
    return env, nl, nl
4704

    
4705
  def CheckPrereq(self):
4706
    """Check prerequisites.
4707

4708
    This checks that the instance is in the cluster and is not running.
4709

4710
    """
4711
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4712
    assert instance is not None, \
4713
      "Cannot retrieve locked instance %s" % self.op.instance_name
4714
    _CheckNodeOnline(self, instance.primary_node)
4715

    
4716
    if instance.disk_template == constants.DT_DISKLESS:
4717
      raise errors.OpPrereqError("Instance '%s' has no disks" %
4718
                                 self.op.instance_name,
4719
                                 errors.ECODE_INVAL)
4720
    _CheckInstanceDown(self, instance, "cannot reinstall")
4721

    
4722
    if self.op.os_type is not None:
4723
      # OS verification
4724
      pnode = _ExpandNodeName(self.cfg, instance.primary_node)
4725
      _CheckNodeHasOS(self, pnode, self.op.os_type, self.op.force_variant)
4726
      instance_os = self.op.os_type
4727
    else:
4728
      instance_os = instance.os
4729

    
4730
    nodelist = list(instance.all_nodes)
4731

    
4732
    if self.op.osparams:
4733
      i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
4734
      _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
4735
      self.os_inst = i_osdict # the new dict (without defaults)
4736
    else:
4737
      self.os_inst = None
4738

    
4739
    self.instance = instance
4740

    
4741
  def Exec(self, feedback_fn):
4742
    """Reinstall the instance.
4743

4744
    """
4745
    inst = self.instance
4746

    
4747
    if self.op.os_type is not None:
4748
      feedback_fn("Changing OS to '%s'..." % self.op.os_type)
4749
      inst.os = self.op.os_type
4750
      # Write to configuration
4751
      self.cfg.Update(inst, feedback_fn)
4752

    
4753
    _StartInstanceDisks(self, inst, None)
4754
    try:
4755
      feedback_fn("Running the instance OS create scripts...")
4756
      # FIXME: pass debug option from opcode to backend
4757
      result = self.rpc.call_instance_os_add(inst.primary_node, inst, True,
4758
                                             self.op.debug_level,
4759
                                             osparams=self.os_inst)
4760
      result.Raise("Could not install OS for instance %s on node %s" %
4761
                   (inst.name, inst.primary_node))
4762
    finally:
4763
      _ShutdownInstanceDisks(self, inst)
4764

    
4765

    
4766
class LURecreateInstanceDisks(LogicalUnit):
4767
  """Recreate an instance's missing disks.
4768

4769
  """
4770
  HPATH = "instance-recreate-disks"
4771
  HTYPE = constants.HTYPE_INSTANCE
4772
  _OP_PARAMS = [
4773
    _PInstanceName,
4774
    ("disks", ht.EmptyList, ht.TListOf(ht.TPositiveInt)),
4775
    ]
4776
  REQ_BGL = False
4777

    
4778
  def ExpandNames(self):
4779
    self._ExpandAndLockInstance()
4780

    
4781
  def BuildHooksEnv(self):
4782
    """Build hooks env.
4783

4784
    This runs on master, primary and secondary nodes of the instance.
4785

4786
    """
4787
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4788
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4789
    return env, nl, nl
4790

    
4791
  def CheckPrereq(self):
4792
    """Check prerequisites.
4793

4794
    This checks that the instance is in the cluster and is not running.
4795

4796
    """
4797
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4798
    assert instance is not None, \
4799
      "Cannot retrieve locked instance %s" % self.op.instance_name
4800
    _CheckNodeOnline(self, instance.primary_node)
4801

    
4802
    if instance.disk_template == constants.DT_DISKLESS:
4803
      raise errors.OpPrereqError("Instance '%s' has no disks" %
4804
                                 self.op.instance_name, errors.ECODE_INVAL)
4805
    _CheckInstanceDown(self, instance, "cannot recreate disks")
4806

    
4807
    if not self.op.disks:
4808
      self.op.disks = range(len(instance.disks))
4809
    else:
4810
      for idx in self.op.disks:
4811
        if idx >= len(instance.disks):
4812
          raise errors.OpPrereqError("Invalid disk index passed '%s'" % idx,
4813
                                     errors.ECODE_INVAL)
4814

    
4815
    self.instance = instance
4816

    
4817
  def Exec(self, feedback_fn):
4818
    """Recreate the disks.
4819

4820
    """
4821
    to_skip = []
4822
    for idx, _ in enumerate(self.instance.disks):
4823
      if idx not in self.op.disks: # disk idx has not been passed in
4824
        to_skip.append(idx)
4825
        continue
4826

    
4827
    _CreateDisks(self, self.instance, to_skip=to_skip)
4828

    
4829

    
4830
class LURenameInstance(LogicalUnit):
4831
  """Rename an instance.
4832

4833
  """
4834
  HPATH = "instance-rename"
4835
  HTYPE = constants.HTYPE_INSTANCE
4836
  _OP_PARAMS = [
4837
    _PInstanceName,
4838
    ("new_name", ht.NoDefault, ht.TNonEmptyString),
4839
    ("ip_check", False, ht.TBool),
4840
    ("name_check", True, ht.TBool),
4841
    ]
4842

    
4843
  def CheckArguments(self):
4844
    """Check arguments.
4845

4846
    """
4847
    if self.op.ip_check and not self.op.name_check:
4848
      # TODO: make the ip check more flexible and not depend on the name check
4849
      raise errors.OpPrereqError("Cannot do ip check without a name check",
4850
                                 errors.ECODE_INVAL)
4851

    
4852
  def BuildHooksEnv(self):
4853
    """Build hooks env.
4854

4855
    This runs on master, primary and secondary nodes of the instance.
4856

4857
    """
4858
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4859
    env["INSTANCE_NEW_NAME"] = self.op.new_name
4860
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4861
    return env, nl, nl
4862

    
4863
  def CheckPrereq(self):
4864
    """Check prerequisites.
4865

4866
    This checks that the instance is in the cluster and is not running.
4867

4868
    """
4869
    self.op.instance_name = _ExpandInstanceName(self.cfg,
4870
                                                self.op.instance_name)
4871
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4872
    assert instance is not None
4873
    _CheckNodeOnline(self, instance.primary_node)
4874
    _CheckInstanceDown(self, instance, "cannot rename")
4875
    self.instance = instance
4876

    
4877
    new_name = self.op.new_name
4878
    if self.op.name_check:
4879
      hostname = netutils.GetHostname(name=new_name)
4880
      new_name = self.op.new_name = hostname.name
4881
      if (self.op.ip_check and
4882
          netutils.TcpPing(hostname.ip, constants.DEFAULT_NODED_PORT)):
4883
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
4884
                                   (hostname.ip, new_name),
4885
                                   errors.ECODE_NOTUNIQUE)
4886

    
4887
    instance_list = self.cfg.GetInstanceList()
4888
    if new_name in instance_list:
4889
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
4890
                                 new_name, errors.ECODE_EXISTS)
4891

    
4892
  def Exec(self, feedback_fn):
4893
    """Reinstall the instance.
4894

4895
    """
4896
    inst = self.instance
4897
    old_name = inst.name
4898

    
4899
    if inst.disk_template == constants.DT_FILE:
4900
      old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
4901

    
4902
    self.cfg.RenameInstance(inst.name, self.op.new_name)
4903
    # Change the instance lock. This is definitely safe while we hold the BGL
4904
    self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
4905
    self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
4906

    
4907
    # re-read the instance from the configuration after rename
4908
    inst = self.cfg.GetInstanceInfo(self.op.new_name)
4909

    
4910
    if inst.disk_template == constants.DT_FILE:
4911
      new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
4912
      result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
4913
                                                     old_file_storage_dir,
4914
                                                     new_file_storage_dir)
4915
      result.Raise("Could not rename on node %s directory '%s' to '%s'"
4916
                   " (but the instance has been renamed in Ganeti)" %
4917
                   (inst.primary_node, old_file_storage_dir,
4918
                    new_file_storage_dir))
4919

    
4920
    _StartInstanceDisks(self, inst, None)
4921
    try:
4922
      result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
4923
                                                 old_name, self.op.debug_level)
4924
      msg = result.fail_msg
4925
      if msg:
4926
        msg = ("Could not run OS rename script for instance %s on node %s"
4927
               " (but the instance has been renamed in Ganeti): %s" %
4928
               (inst.name, inst.primary_node, msg))
4929
        self.proc.LogWarning(msg)
4930
    finally:
4931
      _ShutdownInstanceDisks(self, inst)
4932

    
4933
    return inst.name
4934

    
4935

    
4936
class LURemoveInstance(LogicalUnit):
4937
  """Remove an instance.
4938

4939
  """
4940
  HPATH = "instance-remove"
4941
  HTYPE = constants.HTYPE_INSTANCE
4942
  _OP_PARAMS = [
4943
    _PInstanceName,
4944
    ("ignore_failures", False, ht.TBool),
4945
    _PShutdownTimeout,
4946
    ]
4947
  REQ_BGL = False
4948

    
4949
  def ExpandNames(self):
4950
    self._ExpandAndLockInstance()
4951
    self.needed_locks[locking.LEVEL_NODE] = []
4952
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4953

    
4954
  def DeclareLocks(self, level):
4955
    if level == locking.LEVEL_NODE:
4956
      self._LockInstancesNodes()
4957

    
4958
  def BuildHooksEnv(self):
4959
    """Build hooks env.
4960

4961
    This runs on master, primary and secondary nodes of the instance.
4962

4963
    """
4964
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4965
    env["SHUTDOWN_TIMEOUT"] = self.op.shutdown_timeout
4966
    nl = [self.cfg.GetMasterNode()]
4967
    nl_post = list(self.instance.all_nodes) + nl
4968
    return env, nl, nl_post
4969

    
4970
  def CheckPrereq(self):
4971
    """Check prerequisites.
4972

4973
    This checks that the instance is in the cluster.
4974

4975
    """
4976
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4977
    assert self.instance is not None, \
4978
      "Cannot retrieve locked instance %s" % self.op.instance_name
4979

    
4980
  def Exec(self, feedback_fn):
4981
    """Remove the instance.
4982

4983
    """
4984
    instance = self.instance
4985
    logging.info("Shutting down instance %s on node %s",
4986
                 instance.name, instance.primary_node)
4987

    
4988
    result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
4989
                                             self.op.shutdown_timeout)
4990
    msg = result.fail_msg
4991
    if msg:
4992
      if self.op.ignore_failures:
4993
        feedback_fn("Warning: can't shutdown instance: %s" % msg)
4994
      else:
4995
        raise errors.OpExecError("Could not shutdown instance %s on"
4996
                                 " node %s: %s" %
4997
                                 (instance.name, instance.primary_node, msg))
4998

    
4999
    _RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures)
5000

    
5001

    
5002
def _RemoveInstance(lu, feedback_fn, instance, ignore_failures):
5003
  """Utility function to remove an instance.
5004

5005
  """
5006
  logging.info("Removing block devices for instance %s", instance.name)
5007

    
5008
  if not _RemoveDisks(lu, instance):
5009
    if not ignore_failures:
5010
      raise errors.OpExecError("Can't remove instance's disks")
5011
    feedback_fn("Warning: can't remove instance's disks")
5012

    
5013
  logging.info("Removing instance %s out of cluster config", instance.name)
5014

    
5015
  lu.cfg.RemoveInstance(instance.name)
5016

    
5017
  assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \
5018
    "Instance lock removal conflict"
5019

    
5020
  # Remove lock for the instance
5021
  lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name
5022

    
5023

    
5024
class LUQueryInstances(NoHooksLU):
5025
  """Logical unit for querying instances.
5026

5027
  """
5028
  # pylint: disable-msg=W0142
5029
  _OP_PARAMS = [
5030
    ("output_fields", ht.NoDefault, ht.TListOf(ht.TNonEmptyString)),
5031
    ("names", ht.EmptyList, ht.TListOf(ht.TNonEmptyString)),
5032
    ("use_locking", False, ht.TBool),
5033
    ]
5034
  REQ_BGL = False
5035
  _SIMPLE_FIELDS = ["name", "os", "network_port", "hypervisor",
5036
                    "serial_no", "ctime", "mtime", "uuid"]
5037
  _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
5038
                                    "admin_state",
5039
                                    "disk_template", "ip", "mac", "bridge",
5040
                                    "nic_mode", "nic_link",
5041
                                    "sda_size", "sdb_size", "vcpus", "tags",
5042
                                    "network_port", "beparams",
5043
                                    r"(disk)\.(size)/([0-9]+)",
5044
                                    r"(disk)\.(sizes)", "disk_usage",
5045
                                    r"(nic)\.(mac|ip|mode|link)/([0-9]+)",
5046
                                    r"(nic)\.(bridge)/([0-9]+)",
5047
                                    r"(nic)\.(macs|ips|modes|links|bridges)",
5048
                                    r"(disk|nic)\.(count)",
5049
                                    "hvparams", "custom_hvparams",
5050
                                    "custom_beparams", "custom_nicparams",
5051
                                    ] + _SIMPLE_FIELDS +
5052
                                  ["hv/%s" % name
5053
                                   for name in constants.HVS_PARAMETERS
5054
                                   if name not in constants.HVC_GLOBALS] +
5055
                                  ["be/%s" % name
5056
                                   for name in constants.BES_PARAMETERS])
5057
  _FIELDS_DYNAMIC = utils.FieldSet("oper_state",
5058
                                   "oper_ram",
5059
                                   "oper_vcpus",
5060
                                   "status")
5061

    
5062

    
5063
  def CheckArguments(self):
5064
    _CheckOutputFields(static=self._FIELDS_STATIC,
5065
                       dynamic=self._FIELDS_DYNAMIC,
5066
                       selected=self.op.output_fields)
5067

    
5068
  def ExpandNames(self):
5069
    self.needed_locks = {}
5070
    self.share_locks[locking.LEVEL_INSTANCE] = 1
5071
    self.share_locks[locking.LEVEL_NODE] = 1
5072

    
5073
    if self.op.names:
5074
      self.wanted = _GetWantedInstances(self, self.op.names)
5075
    else:
5076
      self.wanted = locking.ALL_SET
5077

    
5078
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
5079
    self.do_locking = self.do_node_query and self.op.use_locking
5080
    if self.do_locking:
5081
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
5082
      self.needed_locks[locking.LEVEL_NODE] = []
5083
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5084

    
5085
  def DeclareLocks(self, level):
5086
    if level == locking.LEVEL_NODE and self.do_locking:
5087
      self._LockInstancesNodes()
5088

    
5089
  def Exec(self, feedback_fn):
5090
    """Computes the list of nodes and their attributes.
5091

5092
    """
5093
    # pylint: disable-msg=R0912
5094
    # way too many branches here
5095
    all_info = self.cfg.GetAllInstancesInfo()
5096
    if self.wanted == locking.ALL_SET:
5097
      # caller didn't specify instance names, so ordering is not important
5098
      if self.do_locking:
5099
        instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
5100
      else:
5101
        instance_names = all_info.keys()
5102
      instance_names = utils.NiceSort(instance_names)
5103
    else:
5104
      # caller did specify names, so we must keep the ordering
5105
      if self.do_locking:
5106
        tgt_set = self.acquired_locks[locking.LEVEL_INSTANCE]
5107
      else:
5108
        tgt_set = all_info.keys()
5109
      missing = set(self.wanted).difference(tgt_set)
5110
      if missing:
5111
        raise errors.OpExecError("Some instances were removed before"
5112
                                 " retrieving their data: %s" % missing)
5113
      instance_names = self.wanted
5114

    
5115
    instance_list = [all_info[iname] for iname in instance_names]
5116

    
5117
    # begin data gathering
5118

    
5119
    nodes = frozenset([inst.primary_node for inst in instance_list])
5120
    hv_list = list(set([inst.hypervisor for inst in instance_list]))
5121

    
5122
    bad_nodes = []
5123
    off_nodes = []
5124
    if self.do_node_query:
5125
      live_data = {}
5126
      node_data = self.rpc.call_all_instances_info(nodes, hv_list)
5127
      for name in nodes:
5128
        result = node_data[name]
5129
        if result.offline:
5130
          # offline nodes will be in both lists
5131
          off_nodes.append(name)
5132
        if result.fail_msg:
5133
          bad_nodes.append(name)
5134
        else:
5135
          if result.payload:
5136
            live_data.update(result.payload)
5137
          # else no instance is alive
5138
    else:
5139
      live_data = dict([(name, {}) for name in instance_names])
5140

    
5141
    # end data gathering
5142

    
5143
    HVPREFIX = "hv/"
5144
    BEPREFIX = "be/"
5145
    output = []
5146
    cluster = self.cfg.GetClusterInfo()
5147
    for instance in instance_list:
5148
      iout = []
5149
      i_hv = cluster.FillHV(instance, skip_globals=True)
5150
      i_be = cluster.FillBE(instance)
5151
      i_nicp = [cluster.SimpleFillNIC(nic.nicparams) for nic in instance.nics]
5152
      for field in self.op.output_fields:
5153
        st_match = self._FIELDS_STATIC.Matches(field)
5154
        if field in self._SIMPLE_FIELDS:
5155
          val = getattr(instance, field)
5156
        elif field == "pnode":
5157
          val = instance.primary_node
5158
        elif field == "snodes":
5159
          val = list(instance.secondary_nodes)
5160
        elif field == "admin_state":
5161
          val = instance.admin_up
5162
        elif field == "oper_state":
5163
          if instance.primary_node in bad_nodes:
5164
            val = None
5165
          else:
5166
            val = bool(live_data.get(instance.name))
5167
        elif field == "status":
5168
          if instance.primary_node in off_nodes:
5169
            val = "ERROR_nodeoffline"
5170
          elif instance.primary_node in bad_nodes:
5171
            val = "ERROR_nodedown"
5172
          else:
5173
            running = bool(live_data.get(instance.name))
5174
            if running:
5175
              if instance.admin_up:
5176
                val = "running"
5177
              else:
5178
                val = "ERROR_up"
5179
            else:
5180
              if instance.admin_up:
5181
                val = "ERROR_down"
5182
              else:
5183
                val = "ADMIN_down"
5184
        elif field == "oper_ram":
5185
          if instance.primary_node in bad_nodes:
5186
            val = None
5187
          elif instance.name in live_data:
5188
            val = live_data[instance.name].get("memory", "?")
5189
          else:
5190
            val = "-"
5191
        elif field == "oper_vcpus":
5192
          if instance.primary_node in bad_nodes:
5193
            val = None
5194
          elif instance.name in live_data:
5195
            val = live_data[instance.name].get("vcpus", "?")
5196
          else:
5197
            val = "-"
5198
        elif field == "vcpus":
5199
          val = i_be[constants.BE_VCPUS]
5200
        elif field == "disk_template":
5201
          val = instance.disk_template
5202
        elif field == "ip":
5203
          if instance.nics:
5204
            val = instance.nics[0].ip
5205
          else:
5206
            val = None
5207
        elif field == "nic_mode":
5208
          if instance.nics:
5209
            val = i_nicp[0][constants.NIC_MODE]
5210
          else:
5211
            val = None
5212
        elif field == "nic_link":
5213
          if instance.nics:
5214
            val = i_nicp[0][constants.NIC_LINK]
5215
          else:
5216
            val = None
5217
        elif field == "bridge":
5218
          if (instance.nics and
5219
              i_nicp[0][constants.NIC_MODE] == constants.NIC_MODE_BRIDGED):
5220
            val = i_nicp[0][constants.NIC_LINK]
5221
          else:
5222
            val = None
5223
        elif field == "mac":
5224
          if instance.nics:
5225
            val = instance.nics[0].mac
5226
          else:
5227
            val = None
5228
        elif field == "custom_nicparams":
5229
          val = [nic.nicparams for nic in instance.nics]
5230
        elif field == "sda_size" or field == "sdb_size":
5231
          idx = ord(field[2]) - ord('a')
5232
          try:
5233
            val = instance.FindDisk(idx).size
5234
          except errors.OpPrereqError:
5235
            val = None
5236
        elif field == "disk_usage": # total disk usage per node
5237
          disk_sizes = [{'size': disk.size} for disk in instance.disks]
5238
          val = _ComputeDiskSize(instance.disk_template, disk_sizes)
5239
        elif field == "tags":
5240
          val = list(instance.GetTags())
5241
        elif field == "custom_hvparams":
5242
          val = instance.hvparams # not filled!
5243
        elif field == "hvparams":
5244
          val = i_hv
5245
        elif (field.startswith(HVPREFIX) and
5246
              field[len(HVPREFIX):] in constants.HVS_PARAMETERS and
5247
              field[len(HVPREFIX):] not in constants.HVC_GLOBALS):
5248
          val = i_hv.get(field[len(HVPREFIX):], None)
5249
        elif field == "custom_beparams":
5250
          val = instance.beparams
5251
        elif field == "beparams":
5252
          val = i_be
5253
        elif (field.startswith(BEPREFIX) and
5254
              field[len(BEPREFIX):] in constants.BES_PARAMETERS):
5255
          val = i_be.get(field[len(BEPREFIX):], None)
5256
        elif st_match and st_match.groups():
5257
          # matches a variable list
5258
          st_groups = st_match.groups()
5259
          if st_groups and st_groups[0] == "disk":
5260
            if st_groups[1] == "count":
5261
              val = len(instance.disks)
5262
            elif st_groups[1] == "sizes":
5263
              val = [disk.size for disk in instance.disks]
5264
            elif st_groups[1] == "size":
5265
              try:
5266
                val = instance.FindDisk(st_groups[2]).size
5267
              except errors.OpPrereqError:
5268
                val = None
5269
            else:
5270
              assert False, "Unhandled disk parameter"
5271
          elif st_groups[0] == "nic":
5272
            if st_groups[1] == "count":
5273
              val = len(instance.nics)
5274
            elif st_groups[1] == "macs":
5275
              val = [nic.mac for nic in instance.nics]
5276
            elif st_groups[1] == "ips":
5277
              val = [nic.ip for nic in instance.nics]
5278
            elif st_groups[1] == "modes":
5279
              val = [nicp[constants.NIC_MODE] for nicp in i_nicp]
5280
            elif st_groups[1] == "links":
5281
              val = [nicp[constants.NIC_LINK] for nicp in i_nicp]
5282
            elif st_groups[1] == "bridges":
5283
              val = []
5284
              for nicp in i_nicp:
5285
                if nicp[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
5286
                  val.append(nicp[constants.NIC_LINK])
5287
                else:
5288
                  val.append(None)
5289
            else:
5290
              # index-based item
5291
              nic_idx = int(st_groups[2])
5292
              if nic_idx >= len(instance.nics):
5293
                val = None
5294
              else:
5295
                if st_groups[1] == "mac":
5296
                  val = instance.nics[nic_idx].mac
5297
                elif st_groups[1] == "ip":
5298
                  val = instance.nics[nic_idx].ip
5299
                elif st_groups[1] == "mode":
5300
                  val = i_nicp[nic_idx][constants.NIC_MODE]
5301
                elif st_groups[1] == "link":
5302
                  val = i_nicp[nic_idx][constants.NIC_LINK]
5303
                elif st_groups[1] == "bridge":
5304
                  nic_mode = i_nicp[nic_idx][constants.NIC_MODE]
5305
                  if nic_mode == constants.NIC_MODE_BRIDGED:
5306
                    val = i_nicp[nic_idx][constants.NIC_LINK]
5307
                  else:
5308
                    val = None
5309
                else:
5310
                  assert False, "Unhandled NIC parameter"
5311
          else:
5312
            assert False, ("Declared but unhandled variable parameter '%s'" %
5313
                           field)
5314
        else:
5315
          assert False, "Declared but unhandled parameter '%s'" % field
5316
        iout.append(val)
5317
      output.append(iout)
5318

    
5319
    return output
5320

    
5321

    
5322
class LUFailoverInstance(LogicalUnit):
5323
  """Failover an instance.
5324

5325
  """
5326
  HPATH = "instance-failover"
5327
  HTYPE = constants.HTYPE_INSTANCE
5328
  _OP_PARAMS = [
5329
    _PInstanceName,
5330
    ("ignore_consistency", False, ht.TBool),
5331
    _PShutdownTimeout,
5332
    ]
5333
  REQ_BGL = False
5334

    
5335
  def ExpandNames(self):
5336
    self._ExpandAndLockInstance()
5337
    self.needed_locks[locking.LEVEL_NODE] = []
5338
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5339

    
5340
  def DeclareLocks(self, level):
5341
    if level == locking.LEVEL_NODE:
5342
      self._LockInstancesNodes()
5343

    
5344
  def BuildHooksEnv(self):
5345
    """Build hooks env.
5346

5347
    This runs on master, primary and secondary nodes of the instance.
5348

5349
    """
5350
    instance = self.instance
5351
    source_node = instance.primary_node
5352
    target_node = instance.secondary_nodes[0]
5353
    env = {
5354
      "IGNORE_CONSISTENCY": self.op.ignore_consistency,
5355
      "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5356
      "OLD_PRIMARY": source_node,
5357
      "OLD_SECONDARY": target_node,
5358
      "NEW_PRIMARY": target_node,
5359
      "NEW_SECONDARY": source_node,
5360
      }
5361
    env.update(_BuildInstanceHookEnvByObject(self, instance))
5362
    nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5363
    nl_post = list(nl)
5364
    nl_post.append(source_node)
5365
    return env, nl, nl_post
5366

    
5367
  def CheckPrereq(self):
5368
    """Check prerequisites.
5369

5370
    This checks that the instance is in the cluster.
5371

5372
    """
5373
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5374
    assert self.instance is not None, \
5375
      "Cannot retrieve locked instance %s" % self.op.instance_name
5376

    
5377
    bep = self.cfg.GetClusterInfo().FillBE(instance)
5378
    if instance.disk_template not in constants.DTS_NET_MIRROR:
5379
      raise errors.OpPrereqError("Instance's disk layout is not"
5380
                                 " network mirrored, cannot failover.",
5381
                                 errors.ECODE_STATE)
5382

    
5383
    secondary_nodes = instance.secondary_nodes
5384
    if not secondary_nodes:
5385
      raise errors.ProgrammerError("no secondary node but using "
5386
                                   "a mirrored disk template")
5387

    
5388
    target_node = secondary_nodes[0]
5389
    _CheckNodeOnline(self, target_node)
5390
    _CheckNodeNotDrained(self, target_node)
5391
    if instance.admin_up:
5392
      # check memory requirements on the secondary node
5393
      _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5394
                           instance.name, bep[constants.BE_MEMORY],
5395
                           instance.hypervisor)
5396
    else:
5397
      self.LogInfo("Not checking memory on the secondary node as"
5398
                   " instance will not be started")
5399

    
5400
    # check bridge existance
5401
    _CheckInstanceBridgesExist(self, instance, node=target_node)
5402

    
5403
  def Exec(self, feedback_fn):
5404
    """Failover an instance.
5405

5406
    The failover is done by shutting it down on its present node and
5407
    starting it on the secondary.
5408

5409
    """
5410
    instance = self.instance
5411
    primary_node = self.cfg.GetNodeInfo(instance.primary_node)
5412

    
5413
    source_node = instance.primary_node
5414
    target_node = instance.secondary_nodes[0]
5415

    
5416
    if instance.admin_up:
5417
      feedback_fn("* checking disk consistency between source and target")
5418
      for dev in instance.disks:
5419
        # for drbd, these are drbd over lvm
5420
        if not _CheckDiskConsistency(self, dev, target_node, False):
5421
          if not self.op.ignore_consistency:
5422
            raise errors.OpExecError("Disk %s is degraded on target node,"
5423
                                     " aborting failover." % dev.iv_name)
5424
    else:
5425
      feedback_fn("* not checking disk consistency as instance is not running")
5426

    
5427
    feedback_fn("* shutting down instance on source node")
5428
    logging.info("Shutting down instance %s on node %s",
5429
                 instance.name, source_node)
5430

    
5431
    result = self.rpc.call_instance_shutdown(source_node, instance,
5432
                                             self.op.shutdown_timeout)
5433
    msg = result.fail_msg
5434
    if msg:
5435
      if self.op.ignore_consistency or primary_node.offline:
5436
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
5437
                             " Proceeding anyway. Please make sure node"
5438
                             " %s is down. Error details: %s",
5439
                             instance.name, source_node, source_node, msg)
5440
      else:
5441
        raise errors.OpExecError("Could not shutdown instance %s on"
5442
                                 " node %s: %s" %
5443
                                 (instance.name, source_node, msg))
5444

    
5445
    feedback_fn("* deactivating the instance's disks on source node")
5446
    if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
5447
      raise errors.OpExecError("Can't shut down the instance's disks.")
5448

    
5449
    instance.primary_node = target_node
5450
    # distribute new instance config to the other nodes
5451
    self.cfg.Update(instance, feedback_fn)
5452

    
5453
    # Only start the instance if it's marked as up
5454
    if instance.admin_up:
5455
      feedback_fn("* activating the instance's disks on target node")
5456
      logging.info("Starting instance %s on node %s",
5457
                   instance.name, target_node)
5458

    
5459
      disks_ok, _ = _AssembleInstanceDisks(self, instance,
5460
                                           ignore_secondaries=True)
5461
      if not disks_ok:
5462
        _ShutdownInstanceDisks(self, instance)
5463
        raise errors.OpExecError("Can't activate the instance's disks")
5464

    
5465
      feedback_fn("* starting the instance on the target node")
5466
      result = self.rpc.call_instance_start(target_node, instance, None, None)
5467
      msg = result.fail_msg
5468
      if msg:
5469
        _ShutdownInstanceDisks(self, instance)
5470
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
5471
                                 (instance.name, target_node, msg))
5472

    
5473

    
5474
class LUMigrateInstance(LogicalUnit):
5475
  """Migrate an instance.
5476

5477
  This is migration without shutting down, compared to the failover,
5478
  which is done with shutdown.
5479

5480
  """
5481
  HPATH = "instance-migrate"
5482
  HTYPE = constants.HTYPE_INSTANCE
5483
  _OP_PARAMS = [
5484
    _PInstanceName,
5485
    _PMigrationMode,
5486
    _PMigrationLive,
5487
    ("cleanup", False, ht.TBool),
5488
    ]
5489

    
5490
  REQ_BGL = False
5491

    
5492
  def ExpandNames(self):
5493
    self._ExpandAndLockInstance()
5494

    
5495
    self.needed_locks[locking.LEVEL_NODE] = []
5496
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5497

    
5498
    self._migrater = TLMigrateInstance(self, self.op.instance_name,
5499
                                       self.op.cleanup)
5500
    self.tasklets = [self._migrater]
5501

    
5502
  def DeclareLocks(self, level):
5503
    if level == locking.LEVEL_NODE:
5504
      self._LockInstancesNodes()
5505

    
5506
  def BuildHooksEnv(self):
5507
    """Build hooks env.
5508

5509
    This runs on master, primary and secondary nodes of the instance.
5510

5511
    """
5512
    instance = self._migrater.instance
5513
    source_node = instance.primary_node
5514
    target_node = instance.secondary_nodes[0]
5515
    env = _BuildInstanceHookEnvByObject(self, instance)
5516
    env["MIGRATE_LIVE"] = self._migrater.live
5517
    env["MIGRATE_CLEANUP"] = self.op.cleanup
5518
    env.update({
5519
        "OLD_PRIMARY": source_node,
5520
        "OLD_SECONDARY": target_node,
5521
        "NEW_PRIMARY": target_node,
5522
        "NEW_SECONDARY": source_node,
5523
        })
5524
    nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5525
    nl_post = list(nl)
5526
    nl_post.append(source_node)
5527
    return env, nl, nl_post
5528

    
5529

    
5530
class LUMoveInstance(LogicalUnit):
5531
  """Move an instance by data-copying.
5532

5533
  """
5534
  HPATH = "instance-move"
5535
  HTYPE = constants.HTYPE_INSTANCE
5536
  _OP_PARAMS = [
5537
    _PInstanceName,
5538
    ("target_node", ht.NoDefault, ht.TNonEmptyString),
5539
    _PShutdownTimeout,
5540
    ]
5541
  REQ_BGL = False
5542

    
5543
  def ExpandNames(self):
5544
    self._ExpandAndLockInstance()
5545
    target_node = _ExpandNodeName(self.cfg, self.op.target_node)
5546
    self.op.target_node = target_node
5547
    self.needed_locks[locking.LEVEL_NODE] = [target_node]
5548
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5549

    
5550
  def DeclareLocks(self, level):
5551
    if level == locking.LEVEL_NODE:
5552
      self._LockInstancesNodes(primary_only=True)
5553

    
5554
  def BuildHooksEnv(self):
5555
    """Build hooks env.
5556

5557
    This runs on master, primary and secondary nodes of the instance.
5558

5559
    """
5560
    env = {
5561
      "TARGET_NODE": self.op.target_node,
5562
      "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5563
      }
5564
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5565
    nl = [self.cfg.GetMasterNode()] + [self.instance.primary_node,
5566
                                       self.op.target_node]
5567
    return env, nl, nl
5568

    
5569
  def CheckPrereq(self):
5570
    """Check prerequisites.
5571

5572
    This checks that the instance is in the cluster.
5573

5574
    """
5575
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5576
    assert self.instance is not None, \
5577
      "Cannot retrieve locked instance %s" % self.op.instance_name
5578

    
5579
    node = self.cfg.GetNodeInfo(self.op.target_node)
5580
    assert node is not None, \
5581
      "Cannot retrieve locked node %s" % self.op.target_node
5582

    
5583
    self.target_node = target_node = node.name
5584

    
5585
    if target_node == instance.primary_node:
5586
      raise errors.OpPrereqError("Instance %s is already on the node %s" %
5587
                                 (instance.name, target_node),
5588
                                 errors.ECODE_STATE)
5589

    
5590
    bep = self.cfg.GetClusterInfo().FillBE(instance)
5591

    
5592
    for idx, dsk in enumerate(instance.disks):
5593
      if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
5594
        raise errors.OpPrereqError("Instance disk %d has a complex layout,"
5595
                                   " cannot copy" % idx, errors.ECODE_STATE)
5596

    
5597
    _CheckNodeOnline(self, target_node)
5598
    _CheckNodeNotDrained(self, target_node)
5599

    
5600
    if instance.admin_up:
5601
      # check memory requirements on the secondary node
5602
      _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5603
                           instance.name, bep[constants.BE_MEMORY],
5604
                           instance.hypervisor)
5605
    else:
5606
      self.LogInfo("Not checking memory on the secondary node as"
5607
                   " instance will not be started")
5608

    
5609
    # check bridge existance
5610
    _CheckInstanceBridgesExist(self, instance, node=target_node)
5611

    
5612
  def Exec(self, feedback_fn):
5613
    """Move an instance.
5614

5615
    The move is done by shutting it down on its present node, copying
5616
    the data over (slow) and starting it on the new node.
5617

5618
    """
5619
    instance = self.instance
5620

    
5621
    source_node = instance.primary_node
5622
    target_node = self.target_node
5623

    
5624
    self.LogInfo("Shutting down instance %s on source node %s",
5625
                 instance.name, source_node)
5626

    
5627
    result = self.rpc.call_instance_shutdown(source_node, instance,
5628
                                             self.op.shutdown_timeout)
5629
    msg = result.fail_msg
5630
    if msg:
5631
      if self.op.ignore_consistency:
5632
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
5633
                             " Proceeding anyway. Please make sure node"
5634
                             " %s is down. Error details: %s",
5635
                             instance.name, source_node, source_node, msg)
5636
      else:
5637
        raise errors.OpExecError("Could not shutdown instance %s on"
5638
                                 " node %s: %s" %
5639
                                 (instance.name, source_node, msg))
5640

    
5641
    # create the target disks
5642
    try:
5643
      _CreateDisks(self, instance, target_node=target_node)
5644
    except errors.OpExecError:
5645
      self.LogWarning("Device creation failed, reverting...")
5646
      try:
5647
        _RemoveDisks(self, instance, target_node=target_node)
5648
      finally:
5649
        self.cfg.ReleaseDRBDMinors(instance.name)
5650
        raise
5651

    
5652
    cluster_name = self.cfg.GetClusterInfo().cluster_name
5653

    
5654
    errs = []
5655
    # activate, get path, copy the data over
5656
    for idx, disk in enumerate(instance.disks):
5657
      self.LogInfo("Copying data for disk %d", idx)
5658
      result = self.rpc.call_blockdev_assemble(target_node, disk,
5659
                                               instance.name, True)
5660
      if result.fail_msg:
5661
        self.LogWarning("Can't assemble newly created disk %d: %s",
5662
                        idx, result.fail_msg)
5663
        errs.append(result.fail_msg)
5664
        break
5665
      dev_path = result.payload
5666
      result = self.rpc.call_blockdev_export(source_node, disk,
5667
                                             target_node, dev_path,
5668
                                             cluster_name)
5669
      if result.fail_msg:
5670
        self.LogWarning("Can't copy data over for disk %d: %s",
5671
                        idx, result.fail_msg)
5672
        errs.append(result.fail_msg)
5673
        break
5674

    
5675
    if errs:
5676
      self.LogWarning("Some disks failed to copy, aborting")
5677
      try:
5678
        _RemoveDisks(self, instance, target_node=target_node)
5679
      finally:
5680
        self.cfg.ReleaseDRBDMinors(instance.name)
5681
        raise errors.OpExecError("Errors during disk copy: %s" %
5682
                                 (",".join(errs),))
5683

    
5684
    instance.primary_node = target_node
5685
    self.cfg.Update(instance, feedback_fn)
5686

    
5687
    self.LogInfo("Removing the disks on the original node")
5688
    _RemoveDisks(self, instance, target_node=source_node)
5689

    
5690
    # Only start the instance if it's marked as up
5691
    if instance.admin_up:
5692
      self.LogInfo("Starting instance %s on node %s",
5693
                   instance.name, target_node)
5694

    
5695
      disks_ok, _ = _AssembleInstanceDisks(self, instance,
5696
                                           ignore_secondaries=True)
5697
      if not disks_ok:
5698
        _ShutdownInstanceDisks(self, instance)
5699
        raise errors.OpExecError("Can't activate the instance's disks")
5700

    
5701
      result = self.rpc.call_instance_start(target_node, instance, None, None)
5702
      msg = result.fail_msg
5703
      if msg:
5704
        _ShutdownInstanceDisks(self, instance)
5705
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
5706
                                 (instance.name, target_node, msg))
5707

    
5708

    
5709
class LUMigrateNode(LogicalUnit):
5710
  """Migrate all instances from a node.
5711

5712
  """
5713
  HPATH = "node-migrate"
5714
  HTYPE = constants.HTYPE_NODE
5715
  _OP_PARAMS = [
5716
    _PNodeName,
5717
    _PMigrationMode,
5718
    _PMigrationLive,
5719
    ]
5720
  REQ_BGL = False
5721

    
5722
  def ExpandNames(self):
5723
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
5724

    
5725
    self.needed_locks = {
5726
      locking.LEVEL_NODE: [self.op.node_name],
5727
      }
5728

    
5729
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5730

    
5731
    # Create tasklets for migrating instances for all instances on this node
5732
    names = []
5733
    tasklets = []
5734

    
5735
    for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name):
5736
      logging.debug("Migrating instance %s", inst.name)
5737
      names.append(inst.name)
5738

    
5739
      tasklets.append(TLMigrateInstance(self, inst.name, False))
5740

    
5741
    self.tasklets = tasklets
5742

    
5743
    # Declare instance locks
5744
    self.needed_locks[locking.LEVEL_INSTANCE] = names
5745

    
5746
  def DeclareLocks(self, level):
5747
    if level == locking.LEVEL_NODE:
5748
      self._LockInstancesNodes()
5749

    
5750
  def BuildHooksEnv(self):
5751
    """Build hooks env.
5752

5753
    This runs on the master, the primary and all the secondaries.
5754

5755
    """
5756
    env = {
5757
      "NODE_NAME": self.op.node_name,
5758
      }
5759

    
5760
    nl = [self.cfg.GetMasterNode()]
5761

    
5762
    return (env, nl, nl)
5763

    
5764

    
5765
class TLMigrateInstance(Tasklet):
5766
  """Tasklet class for instance migration.
5767

5768
  @type live: boolean
5769
  @ivar live: whether the migration will be done live or non-live;
5770
      this variable is initalized only after CheckPrereq has run
5771

5772
  """
5773
  def __init__(self, lu, instance_name, cleanup):
5774
    """Initializes this class.
5775

5776
    """
5777
    Tasklet.__init__(self, lu)
5778

    
5779
    # Parameters
5780
    self.instance_name = instance_name
5781
    self.cleanup = cleanup
5782
    self.live = False # will be overridden later
5783

    
5784
  def CheckPrereq(self):
5785
    """Check prerequisites.
5786

5787
    This checks that the instance is in the cluster.
5788

5789
    """
5790
    instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name)
5791
    instance = self.cfg.GetInstanceInfo(instance_name)
5792
    assert instance is not None
5793

    
5794
    if instance.disk_template != constants.DT_DRBD8:
5795
      raise errors.OpPrereqError("Instance's disk layout is not"
5796
                                 " drbd8, cannot migrate.", errors.ECODE_STATE)
5797

    
5798
    secondary_nodes = instance.secondary_nodes
5799
    if not secondary_nodes:
5800
      raise errors.ConfigurationError("No secondary node but using"
5801
                                      " drbd8 disk template")
5802

    
5803
    i_be = self.cfg.GetClusterInfo().FillBE(instance)
5804

    
5805
    target_node = secondary_nodes[0]
5806
    # check memory requirements on the secondary node
5807
    _CheckNodeFreeMemory(self.lu, target_node, "migrating instance %s" %
5808
                         instance.name, i_be[constants.BE_MEMORY],
5809
                         instance.hypervisor)
5810

    
5811
    # check bridge existance
5812
    _CheckInstanceBridgesExist(self.lu, instance, node=target_node)
5813

    
5814
    if not self.cleanup:
5815
      _CheckNodeNotDrained(self.lu, target_node)
5816
      result = self.rpc.call_instance_migratable(instance.primary_node,
5817
                                                 instance)
5818
      result.Raise("Can't migrate, please use failover",
5819
                   prereq=True, ecode=errors.ECODE_STATE)
5820

    
5821
    self.instance = instance
5822

    
5823
    if self.lu.op.live is not None and self.lu.op.mode is not None:
5824
      raise errors.OpPrereqError("Only one of the 'live' and 'mode'"
5825
                                 " parameters are accepted",
5826
                                 errors.ECODE_INVAL)
5827
    if self.lu.op.live is not None:
5828
      if self.lu.op.live:
5829
        self.lu.op.mode = constants.HT_MIGRATION_LIVE
5830
      else:
5831
        self.lu.op.mode = constants.HT_MIGRATION_NONLIVE
5832
      # reset the 'live' parameter to None so that repeated
5833
      # invocations of CheckPrereq do not raise an exception
5834
      self.lu.op.live = None
5835
    elif self.lu.op.mode is None:
5836
      # read the default value from the hypervisor
5837
      i_hv = self.cfg.GetClusterInfo().FillHV(instance, skip_globals=False)
5838
      self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE]
5839

    
5840
    self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE
5841

    
5842
  def _WaitUntilSync(self):
5843
    """Poll with custom rpc for disk sync.
5844

5845
    This uses our own step-based rpc call.
5846

5847
    """
5848
    self.feedback_fn("* wait until resync is done")
5849
    all_done = False
5850
    while not all_done:
5851
      all_done = True
5852
      result = self.rpc.call_drbd_wait_sync(self.all_nodes,
5853
                                            self.nodes_ip,
5854
                                            self.instance.disks)
5855
      min_percent = 100
5856
      for node, nres in result.items():
5857
        nres.Raise("Cannot resync disks on node %s" % node)
5858
        node_done, node_percent = nres.payload
5859
        all_done = all_done and node_done
5860
        if node_percent is not None:
5861
          min_percent = min(min_percent, node_percent)
5862
      if not all_done:
5863
        if min_percent < 100:
5864
          self.feedback_fn("   - progress: %.1f%%" % min_percent)
5865
        time.sleep(2)
5866

    
5867
  def _EnsureSecondary(self, node):
5868
    """Demote a node to secondary.
5869

5870
    """
5871
    self.feedback_fn("* switching node %s to secondary mode" % node)
5872

    
5873
    for dev in self.instance.disks:
5874
      self.cfg.SetDiskID(dev, node)
5875

    
5876
    result = self.rpc.call_blockdev_close(node, self.instance.name,
5877
                                          self.instance.disks)
5878
    result.Raise("Cannot change disk to secondary on node %s" % node)
5879

    
5880
  def _GoStandalone(self):
5881
    """Disconnect from the network.
5882

5883
    """
5884
    self.feedback_fn("* changing into standalone mode")
5885
    result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
5886
                                               self.instance.disks)
5887
    for node, nres in result.items():
5888
      nres.Raise("Cannot disconnect disks node %s" % node)
5889

    
5890
  def _GoReconnect(self, multimaster):
5891
    """Reconnect to the network.
5892

5893
    """
5894
    if multimaster:
5895
      msg = "dual-master"
5896
    else:
5897
      msg = "single-master"
5898
    self.feedback_fn("* changing disks into %s mode" % msg)
5899
    result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
5900
                                           self.instance.disks,
5901
                                           self.instance.name, multimaster)
5902
    for node, nres in result.items():
5903
      nres.Raise("Cannot change disks config on node %s" % node)
5904

    
5905
  def _ExecCleanup(self):
5906
    """Try to cleanup after a failed migration.
5907

5908
    The cleanup is done by:
5909
      - check that the instance is running only on one node
5910
        (and update the config if needed)
5911
      - change disks on its secondary node to secondary
5912
      - wait until disks are fully synchronized
5913
      - disconnect from the network
5914
      - change disks into single-master mode
5915
      - wait again until disks are fully synchronized
5916

5917
    """
5918
    instance = self.instance
5919
    target_node = self.target_node
5920
    source_node = self.source_node
5921

    
5922
    # check running on only one node
5923
    self.feedback_fn("* checking where the instance actually runs"
5924
                     " (if this hangs, the hypervisor might be in"
5925
                     " a bad state)")
5926
    ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
5927
    for node, result in ins_l.items():
5928
      result.Raise("Can't contact node %s" % node)
5929

    
5930
    runningon_source = instance.name in ins_l[source_node].payload
5931
    runningon_target = instance.name in ins_l[target_node].payload
5932

    
5933
    if runningon_source and runningon_target:
5934
      raise errors.OpExecError("Instance seems to be running on two nodes,"
5935
                               " or the hypervisor is confused. You will have"
5936
                               " to ensure manually that it runs only on one"
5937
                               " and restart this operation.")
5938

    
5939
    if not (runningon_source or runningon_target):
5940
      raise errors.OpExecError("Instance does not seem to be running at all."
5941
                               " In this case, it's safer to repair by"
5942
                               " running 'gnt-instance stop' to ensure disk"
5943
                               " shutdown, and then restarting it.")
5944

    
5945
    if runningon_target:
5946
      # the migration has actually succeeded, we need to update the config
5947
      self.feedback_fn("* instance running on secondary node (%s),"
5948
                       " updating config" % target_node)
5949
      instance.primary_node = target_node
5950
      self.cfg.Update(instance, self.feedback_fn)
5951
      demoted_node = source_node
5952
    else:
5953
      self.feedback_fn("* instance confirmed to be running on its"
5954
                       " primary node (%s)" % source_node)
5955
      demoted_node = target_node
5956

    
5957
    self._EnsureSecondary(demoted_node)
5958
    try:
5959
      self._WaitUntilSync()
5960
    except errors.OpExecError:
5961
      # we ignore here errors, since if the device is standalone, it
5962
      # won't be able to sync
5963
      pass
5964
    self._GoStandalone()
5965
    self._GoReconnect(False)
5966
    self._WaitUntilSync()
5967

    
5968
    self.feedback_fn("* done")
5969

    
5970
  def _RevertDiskStatus(self):
5971
    """Try to revert the disk status after a failed migration.
5972

5973
    """
5974
    target_node = self.target_node
5975
    try:
5976
      self._EnsureSecondary(target_node)
5977
      self._GoStandalone()
5978
      self._GoReconnect(False)
5979
      self._WaitUntilSync()
5980
    except errors.OpExecError, err:
5981
      self.lu.LogWarning("Migration failed and I can't reconnect the"
5982
                         " drives: error '%s'\n"
5983
                         "Please look and recover the instance status" %
5984
                         str(err))
5985

    
5986
  def _AbortMigration(self):
5987
    """Call the hypervisor code to abort a started migration.
5988

5989
    """
5990
    instance = self.instance
5991
    target_node = self.target_node
5992
    migration_info = self.migration_info
5993

    
5994
    abort_result = self.rpc.call_finalize_migration(target_node,
5995
                                                    instance,
5996
                                                    migration_info,
5997
                                                    False)
5998
    abort_msg = abort_result.fail_msg
5999
    if abort_msg:
6000
      logging.error("Aborting migration failed on target node %s: %s",
6001
                    target_node, abort_msg)
6002
      # Don't raise an exception here, as we stil have to try to revert the
6003
      # disk status, even if this step failed.
6004

    
6005
  def _ExecMigration(self):
6006
    """Migrate an instance.
6007

6008
    The migrate is done by:
6009
      - change the disks into dual-master mode
6010
      - wait until disks are fully synchronized again
6011
      - migrate the instance
6012
      - change disks on the new secondary node (the old primary) to secondary
6013
      - wait until disks are fully synchronized
6014
      - change disks into single-master mode
6015

6016
    """
6017
    instance = self.instance
6018
    target_node = self.target_node
6019
    source_node = self.source_node
6020

    
6021
    self.feedback_fn("* checking disk consistency between source and target")
6022
    for dev in instance.disks:
6023
      if not _CheckDiskConsistency(self.lu, dev, target_node, False):
6024
        raise errors.OpExecError("Disk %s is degraded or not fully"
6025
                                 " synchronized on target node,"
6026
                                 " aborting migrate." % dev.iv_name)
6027

    
6028
    # First get the migration information from the remote node
6029
    result = self.rpc.call_migration_info(source_node, instance)
6030
    msg = result.fail_msg
6031
    if msg:
6032
      log_err = ("Failed fetching source migration information from %s: %s" %
6033
                 (source_node, msg))
6034
      logging.error(log_err)
6035
      raise errors.OpExecError(log_err)
6036

    
6037
    self.migration_info = migration_info = result.payload
6038

    
6039
    # Then switch the disks to master/master mode
6040
    self._EnsureSecondary(target_node)
6041
    self._GoStandalone()
6042
    self._GoReconnect(True)
6043
    self._WaitUntilSync()
6044

    
6045
    self.feedback_fn("* preparing %s to accept the instance" % target_node)
6046
    result = self.rpc.call_accept_instance(target_node,
6047
                                           instance,
6048
                                           migration_info,
6049
                                           self.nodes_ip[target_node])
6050

    
6051
    msg = result.fail_msg
6052
    if msg:
6053
      logging.error("Instance pre-migration failed, trying to revert"
6054
                    " disk status: %s", msg)
6055
      self.feedback_fn("Pre-migration failed, aborting")
6056
      self._AbortMigration()
6057
      self._RevertDiskStatus()
6058
      raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
6059
                               (instance.name, msg))
6060

    
6061
    self.feedback_fn("* migrating instance to %s" % target_node)
6062
    time.sleep(10)
6063
    result = self.rpc.call_instance_migrate(source_node, instance,
6064
                                            self.nodes_ip[target_node],
6065
                                            self.live)
6066
    msg = result.fail_msg
6067
    if msg:
6068
      logging.error("Instance migration failed, trying to revert"
6069
                    " disk status: %s", msg)
6070
      self.feedback_fn("Migration failed, aborting")
6071
      self._AbortMigration()
6072
      self._RevertDiskStatus()
6073
      raise errors.OpExecError("Could not migrate instance %s: %s" %
6074
                               (instance.name, msg))
6075
    time.sleep(10)
6076

    
6077
    instance.primary_node = target_node
6078
    # distribute new instance config to the other nodes
6079
    self.cfg.Update(instance, self.feedback_fn)
6080

    
6081
    result = self.rpc.call_finalize_migration(target_node,
6082
                                              instance,
6083
                                              migration_info,
6084
                                              True)
6085
    msg = result.fail_msg
6086
    if msg:
6087
      logging.error("Instance migration succeeded, but finalization failed:"
6088
                    " %s", msg)
6089
      raise errors.OpExecError("Could not finalize instance migration: %s" %
6090
                               msg)
6091

    
6092
    self._EnsureSecondary(source_node)
6093
    self._WaitUntilSync()
6094
    self._GoStandalone()
6095
    self._GoReconnect(False)
6096
    self._WaitUntilSync()
6097

    
6098
    self.feedback_fn("* done")
6099

    
6100
  def Exec(self, feedback_fn):
6101
    """Perform the migration.
6102

6103
    """
6104
    feedback_fn("Migrating instance %s" % self.instance.name)
6105

    
6106
    self.feedback_fn = feedback_fn
6107

    
6108
    self.source_node = self.instance.primary_node
6109
    self.target_node = self.instance.secondary_nodes[0]
6110
    self.all_nodes = [self.source_node, self.target_node]
6111
    self.nodes_ip = {
6112
      self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
6113
      self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
6114
      }
6115

    
6116
    if self.cleanup:
6117
      return self._ExecCleanup()
6118
    else:
6119
      return self._ExecMigration()
6120

    
6121

    
6122
def _CreateBlockDev(lu, node, instance, device, force_create,
6123
                    info, force_open):
6124
  """Create a tree of block devices on a given node.
6125

6126
  If this device type has to be created on secondaries, create it and
6127
  all its children.
6128

6129
  If not, just recurse to children keeping the same 'force' value.
6130

6131
  @param lu: the lu on whose behalf we execute
6132
  @param node: the node on which to create the device
6133
  @type instance: L{objects.Instance}
6134
  @param instance: the instance which owns the device
6135
  @type device: L{objects.Disk}
6136
  @param device: the device to create
6137
  @type force_create: boolean
6138
  @param force_create: whether to force creation of this device; this
6139
      will be change to True whenever we find a device which has
6140
      CreateOnSecondary() attribute
6141
  @param info: the extra 'metadata' we should attach to the device
6142
      (this will be represented as a LVM tag)
6143
  @type force_open: boolean
6144
  @param force_open: this parameter will be passes to the
6145
      L{backend.BlockdevCreate} function where it specifies
6146
      whether we run on primary or not, and it affects both
6147
      the child assembly and the device own Open() execution
6148

6149
  """
6150
  if device.CreateOnSecondary():
6151
    force_create = True
6152

    
6153
  if device.children:
6154
    for child in device.children:
6155
      _CreateBlockDev(lu, node, instance, child, force_create,
6156
                      info, force_open)
6157

    
6158
  if not force_create:
6159
    return
6160

    
6161
  _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
6162

    
6163

    
6164
def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
6165
  """Create a single block device on a given node.
6166

6167
  This will not recurse over children of the device, so they must be
6168
  created in advance.
6169

6170
  @param lu: the lu on whose behalf we execute
6171
  @param node: the node on which to create the device
6172
  @type instance: L{objects.Instance}
6173
  @param instance: the instance which owns the device
6174
  @type device: L{objects.Disk}
6175
  @param device: the device to create
6176
  @param info: the extra 'metadata' we should attach to the device
6177
      (this will be represented as a LVM tag)
6178
  @type force_open: boolean
6179
  @param force_open: this parameter will be passes to the
6180
      L{backend.BlockdevCreate} function where it specifies
6181
      whether we run on primary or not, and it affects both
6182
      the child assembly and the device own Open() execution
6183

6184
  """
6185
  lu.cfg.SetDiskID(device, node)
6186
  result = lu.rpc.call_blockdev_create(node, device, device.size,
6187
                                       instance.name, force_open, info)
6188
  result.Raise("Can't create block device %s on"
6189
               " node %s for instance %s" % (device, node, instance.name))
6190
  if device.physical_id is None:
6191
    device.physical_id = result.payload
6192

    
6193

    
6194
def _GenerateUniqueNames(lu, exts):
6195
  """Generate a suitable LV name.
6196

6197
  This will generate a logical volume name for the given instance.
6198

6199
  """
6200
  results = []
6201
  for val in exts:
6202
    new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
6203
    results.append("%s%s" % (new_id, val))
6204
  return results
6205

    
6206

    
6207
def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
6208
                         p_minor, s_minor):
6209
  """Generate a drbd8 device complete with its children.
6210

6211
  """
6212
  port = lu.cfg.AllocatePort()
6213
  vgname = lu.cfg.GetVGName()
6214
  shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
6215
  dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
6216
                          logical_id=(vgname, names[0]))
6217
  dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
6218
                          logical_id=(vgname, names[1]))
6219
  drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
6220
                          logical_id=(primary, secondary, port,
6221
                                      p_minor, s_minor,
6222
                                      shared_secret),
6223
                          children=[dev_data, dev_meta],
6224
                          iv_name=iv_name)
6225
  return drbd_dev
6226

    
6227

    
6228
def _GenerateDiskTemplate(lu, template_name,
6229
                          instance_name, primary_node,
6230
                          secondary_nodes, disk_info,
6231
                          file_storage_dir, file_driver,
6232
                          base_index):
6233
  """Generate the entire disk layout for a given template type.
6234

6235
  """
6236
  #TODO: compute space requirements
6237

    
6238
  vgname = lu.cfg.GetVGName()
6239
  disk_count = len(disk_info)
6240
  disks = []
6241
  if template_name == constants.DT_DISKLESS:
6242
    pass
6243
  elif template_name == constants.DT_PLAIN:
6244
    if len(secondary_nodes) != 0:
6245
      raise errors.ProgrammerError("Wrong template configuration")
6246

    
6247
    names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
6248
                                      for i in range(disk_count)])
6249
    for idx, disk in enumerate(disk_info):
6250
      disk_index = idx + base_index
6251
      disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
6252
                              logical_id=(vgname, names[idx]),
6253
                              iv_name="disk/%d" % disk_index,
6254
                              mode=disk["mode"])
6255
      disks.append(disk_dev)
6256
  elif template_name == constants.DT_DRBD8:
6257
    if len(secondary_nodes) != 1:
6258
      raise errors.ProgrammerError("Wrong template configuration")
6259
    remote_node = secondary_nodes[0]
6260
    minors = lu.cfg.AllocateDRBDMinor(
6261
      [primary_node, remote_node] * len(disk_info), instance_name)
6262

    
6263
    names = []
6264
    for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
6265
                                               for i in range(disk_count)]):
6266
      names.append(lv_prefix + "_data")
6267
      names.append(lv_prefix + "_meta")
6268
    for idx, disk in enumerate(disk_info):
6269
      disk_index = idx + base_index
6270
      disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
6271
                                      disk["size"], names[idx*2:idx*2+2],
6272
                                      "disk/%d" % disk_index,
6273
                                      minors[idx*2], minors[idx*2+1])
6274
      disk_dev.mode = disk["mode"]
6275
      disks.append(disk_dev)
6276
  elif template_name == constants.DT_FILE:
6277
    if len(secondary_nodes) != 0:
6278
      raise errors.ProgrammerError("Wrong template configuration")
6279

    
6280
    _RequireFileStorage()
6281

    
6282
    for idx, disk in enumerate(disk_info):
6283
      disk_index = idx + base_index
6284
      disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
6285
                              iv_name="disk/%d" % disk_index,
6286
                              logical_id=(file_driver,
6287
                                          "%s/disk%d" % (file_storage_dir,
6288
                                                         disk_index)),
6289
                              mode=disk["mode"])
6290
      disks.append(disk_dev)
6291
  else:
6292
    raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
6293
  return disks
6294

    
6295

    
6296
def _GetInstanceInfoText(instance):
6297
  """Compute that text that should be added to the disk's metadata.
6298

6299
  """
6300
  return "originstname+%s" % instance.name
6301

    
6302

    
6303
def _CalcEta(time_taken, written, total_size):
6304
  """Calculates the ETA based on size written and total size.
6305

6306
  @param time_taken: The time taken so far
6307
  @param written: amount written so far
6308
  @param total_size: The total size of data to be written
6309
  @return: The remaining time in seconds
6310

6311
  """
6312
  avg_time = time_taken / float(written)
6313
  return (total_size - written) * avg_time
6314

    
6315

    
6316
def _WipeDisks(lu, instance):
6317
  """Wipes instance disks.
6318

6319
  @type lu: L{LogicalUnit}
6320
  @param lu: the logical unit on whose behalf we execute
6321
  @type instance: L{objects.Instance}
6322
  @param instance: the instance whose disks we should create
6323
  @return: the success of the wipe
6324

6325
  """
6326
  node = instance.primary_node
6327
  for idx, device in enumerate(instance.disks):
6328
    lu.LogInfo("* Wiping disk %d", idx)
6329
    logging.info("Wiping disk %d for instance %s", idx, instance.name)
6330

    
6331
    # The wipe size is MIN_WIPE_CHUNK_PERCENT % of the instance disk but
6332
    # MAX_WIPE_CHUNK at max
6333
    wipe_chunk_size = min(constants.MAX_WIPE_CHUNK, device.size / 100.0 *
6334
                          constants.MIN_WIPE_CHUNK_PERCENT)
6335

    
6336
    offset = 0
6337
    size = device.size
6338
    last_output = 0
6339
    start_time = time.time()
6340

    
6341
    while offset < size:
6342
      wipe_size = min(wipe_chunk_size, size - offset)
6343
      result = lu.rpc.call_blockdev_wipe(node, device, offset, wipe_size)
6344
      result.Raise("Could not wipe disk %d at offset %d for size %d" %
6345
                   (idx, offset, wipe_size))
6346
      now = time.time()
6347
      offset += wipe_size
6348
      if now - last_output >= 60:
6349
        eta = _CalcEta(now - start_time, offset, size)
6350
        lu.LogInfo(" - done: %.1f%% ETA: %s" %
6351
                   (offset / float(size) * 100, utils.FormatSeconds(eta)))
6352
        last_output = now
6353

    
6354

    
6355
def _CreateDisks(lu, instance, to_skip=None, target_node=None):
6356
  """Create all disks for an instance.
6357

6358
  This abstracts away some work from AddInstance.
6359

6360
  @type lu: L{LogicalUnit}
6361
  @param lu: the logical unit on whose behalf we execute
6362
  @type instance: L{objects.Instance}
6363
  @param instance: the instance whose disks we should create
6364
  @type to_skip: list
6365
  @param to_skip: list of indices to skip
6366
  @type target_node: string
6367
  @param target_node: if passed, overrides the target node for creation
6368
  @rtype: boolean
6369
  @return: the success of the creation
6370

6371
  """
6372
  info = _GetInstanceInfoText(instance)
6373
  if target_node is None:
6374
    pnode = instance.primary_node
6375
    all_nodes = instance.all_nodes
6376
  else:
6377
    pnode = target_node
6378
    all_nodes = [pnode]
6379

    
6380
  if instance.disk_template == constants.DT_FILE:
6381
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
6382
    result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
6383

    
6384
    result.Raise("Failed to create directory '%s' on"
6385
                 " node %s" % (file_storage_dir, pnode))
6386

    
6387
  # Note: this needs to be kept in sync with adding of disks in
6388
  # LUSetInstanceParams
6389
  for idx, device in enumerate(instance.disks):
6390
    if to_skip and idx in to_skip:
6391
      continue
6392
    logging.info("Creating volume %s for instance %s",
6393
                 device.iv_name, instance.name)
6394
    #HARDCODE
6395
    for node in all_nodes:
6396
      f_create = node == pnode
6397
      _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
6398

    
6399

    
6400
def _RemoveDisks(lu, instance, target_node=None):
6401
  """Remove all disks for an instance.
6402

6403
  This abstracts away some work from `AddInstance()` and
6404
  `RemoveInstance()`. Note that in case some of the devices couldn't
6405
  be removed, the removal will continue with the other ones (compare
6406
  with `_CreateDisks()`).
6407

6408
  @type lu: L{LogicalUnit}
6409
  @param lu: the logical unit on whose behalf we execute
6410
  @type instance: L{objects.Instance}
6411
  @param instance: the instance whose disks we should remove
6412
  @type target_node: string
6413
  @param target_node: used to override the node on which to remove the disks
6414
  @rtype: boolean
6415
  @return: the success of the removal
6416

6417
  """
6418
  logging.info("Removing block devices for instance %s", instance.name)
6419

    
6420
  all_result = True
6421
  for device in instance.disks:
6422
    if target_node:
6423
      edata = [(target_node, device)]
6424
    else:
6425
      edata = device.ComputeNodeTree(instance.primary_node)
6426
    for node, disk in edata:
6427
      lu.cfg.SetDiskID(disk, node)
6428
      msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
6429
      if msg:
6430
        lu.LogWarning("Could not remove block device %s on node %s,"
6431
                      " continuing anyway: %s", device.iv_name, node, msg)
6432
        all_result = False
6433

    
6434
  if instance.disk_template == constants.DT_FILE:
6435
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
6436
    if target_node:
6437
      tgt = target_node
6438
    else:
6439
      tgt = instance.primary_node
6440
    result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
6441
    if result.fail_msg:
6442
      lu.LogWarning("Could not remove directory '%s' on node %s: %s",
6443
                    file_storage_dir, instance.primary_node, result.fail_msg)
6444
      all_result = False
6445

    
6446
  return all_result
6447

    
6448

    
6449
def _ComputeDiskSize(disk_template, disks):
6450
  """Compute disk size requirements in the volume group
6451

6452
  """
6453
  # Required free disk space as a function of disk and swap space
6454
  req_size_dict = {
6455
    constants.DT_DISKLESS: None,
6456
    constants.DT_PLAIN: sum(d["size"] for d in disks),
6457
    # 128 MB are added for drbd metadata for each disk
6458
    constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
6459
    constants.DT_FILE: None,
6460
  }
6461

    
6462
  if disk_template not in req_size_dict:
6463
    raise errors.ProgrammerError("Disk template '%s' size requirement"
6464
                                 " is unknown" %  disk_template)
6465

    
6466
  return req_size_dict[disk_template]
6467

    
6468

    
6469
def _CheckHVParams(lu, nodenames, hvname, hvparams):
6470
  """Hypervisor parameter validation.
6471

6472
  This function abstract the hypervisor parameter validation to be
6473
  used in both instance create and instance modify.
6474

6475
  @type lu: L{LogicalUnit}
6476
  @param lu: the logical unit for which we check
6477
  @type nodenames: list
6478
  @param nodenames: the list of nodes on which we should check
6479
  @type hvname: string
6480
  @param hvname: the name of the hypervisor we should use
6481
  @type hvparams: dict
6482
  @param hvparams: the parameters which we need to check
6483
  @raise errors.OpPrereqError: if the parameters are not valid
6484

6485
  """
6486
  hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
6487
                                                  hvname,
6488
                                                  hvparams)
6489
  for node in nodenames:
6490
    info = hvinfo[node]
6491
    if info.offline:
6492
      continue
6493
    info.Raise("Hypervisor parameter validation failed on node %s" % node)
6494

    
6495

    
6496
def _CheckOSParams(lu, required, nodenames, osname, osparams):
6497
  """OS parameters validation.
6498

6499
  @type lu: L{LogicalUnit}
6500
  @param lu: the logical unit for which we check
6501
  @type required: boolean
6502
  @param required: whether the validation should fail if the OS is not
6503
      found
6504
  @type nodenames: list
6505
  @param nodenames: the list of nodes on which we should check
6506
  @type osname: string
6507
  @param osname: the name of the hypervisor we should use
6508
  @type osparams: dict
6509
  @param osparams: the parameters which we need to check
6510
  @raise errors.OpPrereqError: if the parameters are not valid
6511

6512
  """
6513
  result = lu.rpc.call_os_validate(required, nodenames, osname,
6514
                                   [constants.OS_VALIDATE_PARAMETERS],
6515
                                   osparams)
6516
  for node, nres in result.items():
6517
    # we don't check for offline cases since this should be run only
6518
    # against the master node and/or an instance's nodes
6519
    nres.Raise("OS Parameters validation failed on node %s" % node)
6520
    if not nres.payload:
6521
      lu.LogInfo("OS %s not found on node %s, validation skipped",
6522
                 osname, node)
6523

    
6524

    
6525
class LUCreateInstance(LogicalUnit):
6526
  """Create an instance.
6527

6528
  """
6529
  HPATH = "instance-add"
6530
  HTYPE = constants.HTYPE_INSTANCE
6531
  _OP_PARAMS = [
6532
    _PInstanceName,
6533
    ("mode", ht.NoDefault, ht.TElemOf(constants.INSTANCE_CREATE_MODES)),
6534
    ("start", True, ht.TBool),
6535
    ("wait_for_sync", True, ht.TBool),
6536
    ("ip_check", True, ht.TBool),
6537
    ("name_check", True, ht.TBool),
6538
    ("disks", ht.NoDefault, ht.TListOf(ht.TDict)),
6539
    ("nics", ht.NoDefault, ht.TListOf(ht.TDict)),
6540
    ("hvparams", ht.EmptyDict, ht.TDict),
6541
    ("beparams", ht.EmptyDict, ht.TDict),
6542
    ("osparams", ht.EmptyDict, ht.TDict),
6543
    ("no_install", None, ht.TMaybeBool),
6544
    ("os_type", None, ht.TMaybeString),
6545
    ("force_variant", False, ht.TBool),
6546
    ("source_handshake", None, ht.TOr(ht.TList, ht.TNone)),
6547
    ("source_x509_ca", None, ht.TMaybeString),
6548
    ("source_instance_name", None, ht.TMaybeString),
6549
    ("src_node", None, ht.TMaybeString),
6550
    ("src_path", None, ht.TMaybeString),
6551
    ("pnode", None, ht.TMaybeString),
6552
    ("snode", None, ht.TMaybeString),
6553
    ("iallocator", None, ht.TMaybeString),
6554
    ("hypervisor", None, ht.TMaybeString),
6555
    ("disk_template", ht.NoDefault, _CheckDiskTemplate),
6556
    ("identify_defaults", False, ht.TBool),
6557
    ("file_driver", None, ht.TOr(ht.TNone, ht.TElemOf(constants.FILE_DRIVER))),
6558
    ("file_storage_dir", None, ht.TMaybeString),
6559
    ]
6560
  REQ_BGL = False
6561

    
6562
  def CheckArguments(self):
6563
    """Check arguments.
6564

6565
    """
6566
    # do not require name_check to ease forward/backward compatibility
6567
    # for tools
6568
    if self.op.no_install and self.op.start:
6569
      self.LogInfo("No-installation mode selected, disabling startup")
6570
      self.op.start = False
6571
    # validate/normalize the instance name
6572
    self.op.instance_name = \
6573
      netutils.Hostname.GetNormalizedName(self.op.instance_name)
6574

    
6575
    if self.op.ip_check and not self.op.name_check:
6576
      # TODO: make the ip check more flexible and not depend on the name check
6577
      raise errors.OpPrereqError("Cannot do ip check without a name check",
6578
                                 errors.ECODE_INVAL)
6579

    
6580
    # check nics' parameter names
6581
    for nic in self.op.nics:
6582
      utils.ForceDictType(nic, constants.INIC_PARAMS_TYPES)
6583

    
6584
    # check disks. parameter names and consistent adopt/no-adopt strategy
6585
    has_adopt = has_no_adopt = False
6586
    for disk in self.op.disks:
6587
      utils.ForceDictType(disk, constants.IDISK_PARAMS_TYPES)
6588
      if "adopt" in disk:
6589
        has_adopt = True
6590
      else:
6591
        has_no_adopt = True
6592
    if has_adopt and has_no_adopt:
6593
      raise errors.OpPrereqError("Either all disks are adopted or none is",
6594
                                 errors.ECODE_INVAL)
6595
    if has_adopt:
6596
      if self.op.disk_template not in constants.DTS_MAY_ADOPT:
6597
        raise errors.OpPrereqError("Disk adoption is not supported for the"
6598
                                   " '%s' disk template" %
6599
                                   self.op.disk_template,
6600
                                   errors.ECODE_INVAL)
6601
      if self.op.iallocator is not None:
6602
        raise errors.OpPrereqError("Disk adoption not allowed with an"
6603
                                   " iallocator script", errors.ECODE_INVAL)
6604
      if self.op.mode == constants.INSTANCE_IMPORT:
6605
        raise errors.OpPrereqError("Disk adoption not allowed for"
6606
                                   " instance import", errors.ECODE_INVAL)
6607

    
6608
    self.adopt_disks = has_adopt
6609

    
6610
    # instance name verification
6611
    if self.op.name_check:
6612
      self.hostname1 = netutils.GetHostname(name=self.op.instance_name)
6613
      self.op.instance_name = self.hostname1.name
6614
      # used in CheckPrereq for ip ping check
6615
      self.check_ip = self.hostname1.ip
6616
    else:
6617
      self.check_ip = None
6618

    
6619
    # file storage checks
6620
    if (self.op.file_driver and
6621
        not self.op.file_driver in constants.FILE_DRIVER):
6622
      raise errors.OpPrereqError("Invalid file driver name '%s'" %
6623
                                 self.op.file_driver, errors.ECODE_INVAL)
6624

    
6625
    if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
6626
      raise errors.OpPrereqError("File storage directory path not absolute",
6627
                                 errors.ECODE_INVAL)
6628

    
6629
    ### Node/iallocator related checks
6630
    _CheckIAllocatorOrNode(self, "iallocator", "pnode")
6631

    
6632
    if self.op.pnode is not None:
6633
      if self.op.disk_template in constants.DTS_NET_MIRROR:
6634
        if self.op.snode is None:
6635
          raise errors.OpPrereqError("The networked disk templates need"
6636
                                     " a mirror node", errors.ECODE_INVAL)
6637
      elif self.op.snode:
6638
        self.LogWarning("Secondary node will be ignored on non-mirrored disk"
6639
                        " template")
6640
        self.op.snode = None
6641

    
6642
    self._cds = _GetClusterDomainSecret()
6643

    
6644
    if self.op.mode == constants.INSTANCE_IMPORT:
6645
      # On import force_variant must be True, because if we forced it at
6646
      # initial install, our only chance when importing it back is that it
6647
      # works again!
6648
      self.op.force_variant = True
6649

    
6650
      if self.op.no_install:
6651
        self.LogInfo("No-installation mode has no effect during import")
6652

    
6653
    elif self.op.mode == constants.INSTANCE_CREATE:
6654
      if self.op.os_type is None:
6655
        raise errors.OpPrereqError("No guest OS specified",
6656
                                   errors.ECODE_INVAL)
6657
      if self.op.os_type in self.cfg.GetClusterInfo().blacklisted_os:
6658
        raise errors.OpPrereqError("Guest OS '%s' is not allowed for"
6659
                                   " installation" % self.op.os_type,
6660
                                   errors.ECODE_STATE)
6661
      if self.op.disk_template is None:
6662
        raise errors.OpPrereqError("No disk template specified",
6663
                                   errors.ECODE_INVAL)
6664

    
6665
    elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
6666
      # Check handshake to ensure both clusters have the same domain secret
6667
      src_handshake = self.op.source_handshake
6668
      if not src_handshake:
6669
        raise errors.OpPrereqError("Missing source handshake",
6670
                                   errors.ECODE_INVAL)
6671

    
6672
      errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds,
6673
                                                           src_handshake)
6674
      if errmsg:
6675
        raise errors.OpPrereqError("Invalid handshake: %s" % errmsg,
6676
                                   errors.ECODE_INVAL)
6677

    
6678
      # Load and check source CA
6679
      self.source_x509_ca_pem = self.op.source_x509_ca
6680
      if not self.source_x509_ca_pem:
6681
        raise errors.OpPrereqError("Missing source X509 CA",
6682
                                   errors.ECODE_INVAL)
6683

    
6684
      try:
6685
        (cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem,
6686
                                                    self._cds)
6687
      except OpenSSL.crypto.Error, err:
6688
        raise errors.OpPrereqError("Unable to load source X509 CA (%s)" %
6689
                                   (err, ), errors.ECODE_INVAL)
6690

    
6691
      (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
6692
      if errcode is not None:
6693
        raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ),
6694
                                   errors.ECODE_INVAL)
6695

    
6696
      self.source_x509_ca = cert
6697

    
6698
      src_instance_name = self.op.source_instance_name
6699
      if not src_instance_name:
6700
        raise errors.OpPrereqError("Missing source instance name",
6701
                                   errors.ECODE_INVAL)
6702

    
6703
      self.source_instance_name = \
6704
          netutils.GetHostname(name=src_instance_name).name
6705

    
6706
    else:
6707
      raise errors.OpPrereqError("Invalid instance creation mode %r" %
6708
                                 self.op.mode, errors.ECODE_INVAL)
6709

    
6710
  def ExpandNames(self):
6711
    """ExpandNames for CreateInstance.
6712

6713
    Figure out the right locks for instance creation.
6714

6715
    """
6716
    self.needed_locks = {}
6717

    
6718
    instance_name = self.op.instance_name
6719
    # this is just a preventive check, but someone might still add this
6720
    # instance in the meantime, and creation will fail at lock-add time
6721
    if instance_name in self.cfg.GetInstanceList():
6722
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
6723
                                 instance_name, errors.ECODE_EXISTS)
6724

    
6725
    self.add_locks[locking.LEVEL_INSTANCE] = instance_name
6726

    
6727
    if self.op.iallocator:
6728
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6729
    else:
6730
      self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode)
6731
      nodelist = [self.op.pnode]
6732
      if self.op.snode is not None:
6733
        self.op.snode = _ExpandNodeName(self.cfg, self.op.snode)
6734
        nodelist.append(self.op.snode)
6735
      self.needed_locks[locking.LEVEL_NODE] = nodelist
6736

    
6737
    # in case of import lock the source node too
6738
    if self.op.mode == constants.INSTANCE_IMPORT:
6739
      src_node = self.op.src_node
6740
      src_path = self.op.src_path
6741

    
6742
      if src_path is None:
6743
        self.op.src_path = src_path = self.op.instance_name
6744

    
6745
      if src_node is None:
6746
        self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6747
        self.op.src_node = None
6748
        if os.path.isabs(src_path):
6749
          raise errors.OpPrereqError("Importing an instance from an absolute"
6750
                                     " path requires a source node option.",
6751
                                     errors.ECODE_INVAL)
6752
      else:
6753
        self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node)
6754
        if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
6755
          self.needed_locks[locking.LEVEL_NODE].append(src_node)
6756
        if not os.path.isabs(src_path):
6757
          self.op.src_path = src_path = \
6758
            utils.PathJoin(constants.EXPORT_DIR, src_path)
6759

    
6760
  def _RunAllocator(self):
6761
    """Run the allocator based on input opcode.
6762

6763
    """
6764
    nics = [n.ToDict() for n in self.nics]
6765
    ial = IAllocator(self.cfg, self.rpc,
6766
                     mode=constants.IALLOCATOR_MODE_ALLOC,
6767
                     name=self.op.instance_name,
6768
                     disk_template=self.op.disk_template,
6769
                     tags=[],
6770
                     os=self.op.os_type,
6771
                     vcpus=self.be_full[constants.BE_VCPUS],
6772
                     mem_size=self.be_full[constants.BE_MEMORY],
6773
                     disks=self.disks,
6774
                     nics=nics,
6775
                     hypervisor=self.op.hypervisor,
6776
                     )
6777

    
6778
    ial.Run(self.op.iallocator)
6779

    
6780
    if not ial.success:
6781
      raise errors.OpPrereqError("Can't compute nodes using"
6782
                                 " iallocator '%s': %s" %
6783
                                 (self.op.iallocator, ial.info),
6784
                                 errors.ECODE_NORES)
6785
    if len(ial.result) != ial.required_nodes:
6786
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
6787
                                 " of nodes (%s), required %s" %
6788
                                 (self.op.iallocator, len(ial.result),
6789
                                  ial.required_nodes), errors.ECODE_FAULT)
6790
    self.op.pnode = ial.result[0]
6791
    self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
6792
                 self.op.instance_name, self.op.iallocator,
6793
                 utils.CommaJoin(ial.result))
6794
    if ial.required_nodes == 2:
6795
      self.op.snode = ial.result[1]
6796

    
6797
  def BuildHooksEnv(self):
6798
    """Build hooks env.
6799

6800
    This runs on master, primary and secondary nodes of the instance.
6801

6802
    """
6803
    env = {
6804
      "ADD_MODE": self.op.mode,
6805
      }
6806
    if self.op.mode == constants.INSTANCE_IMPORT:
6807
      env["SRC_NODE"] = self.op.src_node
6808
      env["SRC_PATH"] = self.op.src_path
6809
      env["SRC_IMAGES"] = self.src_images
6810

    
6811
    env.update(_BuildInstanceHookEnv(
6812
      name=self.op.instance_name,
6813
      primary_node=self.op.pnode,
6814
      secondary_nodes=self.secondaries,
6815
      status=self.op.start,
6816
      os_type=self.op.os_type,
6817
      memory=self.be_full[constants.BE_MEMORY],
6818
      vcpus=self.be_full[constants.BE_VCPUS],
6819
      nics=_NICListToTuple(self, self.nics),
6820
      disk_template=self.op.disk_template,
6821
      disks=[(d["size"], d["mode"]) for d in self.disks],
6822
      bep=self.be_full,
6823
      hvp=self.hv_full,
6824
      hypervisor_name=self.op.hypervisor,
6825
    ))
6826

    
6827
    nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
6828
          self.secondaries)
6829
    return env, nl, nl
6830

    
6831
  def _ReadExportInfo(self):
6832
    """Reads the export information from disk.
6833

6834
    It will override the opcode source node and path with the actual
6835
    information, if these two were not specified before.
6836

6837
    @return: the export information
6838

6839
    """
6840
    assert self.op.mode == constants.INSTANCE_IMPORT
6841

    
6842
    src_node = self.op.src_node
6843
    src_path = self.op.src_path
6844

    
6845
    if src_node is None:
6846
      locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
6847
      exp_list = self.rpc.call_export_list(locked_nodes)
6848
      found = False
6849
      for node in exp_list:
6850
        if exp_list[node].fail_msg:
6851
          continue
6852
        if src_path in exp_list[node].payload:
6853
          found = True
6854
          self.op.src_node = src_node = node
6855
          self.op.src_path = src_path = utils.PathJoin(constants.EXPORT_DIR,
6856
                                                       src_path)
6857
          break
6858
      if not found:
6859
        raise errors.OpPrereqError("No export found for relative path %s" %
6860
                                    src_path, errors.ECODE_INVAL)
6861

    
6862
    _CheckNodeOnline(self, src_node)
6863
    result = self.rpc.call_export_info(src_node, src_path)
6864
    result.Raise("No export or invalid export found in dir %s" % src_path)
6865

    
6866
    export_info = objects.SerializableConfigParser.Loads(str(result.payload))
6867
    if not export_info.has_section(constants.INISECT_EXP):
6868
      raise errors.ProgrammerError("Corrupted export config",
6869
                                   errors.ECODE_ENVIRON)
6870

    
6871
    ei_version = export_info.get(constants.INISECT_EXP, "version")
6872
    if (int(ei_version) != constants.EXPORT_VERSION):
6873
      raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
6874
                                 (ei_version, constants.EXPORT_VERSION),
6875
                                 errors.ECODE_ENVIRON)
6876
    return export_info
6877

    
6878
  def _ReadExportParams(self, einfo):
6879
    """Use export parameters as defaults.
6880

6881
    In case the opcode doesn't specify (as in override) some instance
6882
    parameters, then try to use them from the export information, if
6883
    that declares them.
6884

6885
    """
6886
    self.op.os_type = einfo.get(constants.INISECT_EXP, "os")
6887

    
6888
    if self.op.disk_template is None:
6889
      if einfo.has_option(constants.INISECT_INS, "disk_template"):
6890
        self.op.disk_template = einfo.get(constants.INISECT_INS,
6891
                                          "disk_template")
6892
      else:
6893
        raise errors.OpPrereqError("No disk template specified and the export"
6894
                                   " is missing the disk_template information",
6895
                                   errors.ECODE_INVAL)
6896

    
6897
    if not self.op.disks:
6898
      if einfo.has_option(constants.INISECT_INS, "disk_count"):
6899
        disks = []
6900
        # TODO: import the disk iv_name too
6901
        for idx in range(einfo.getint(constants.INISECT_INS, "disk_count")):
6902
          disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
6903
          disks.append({"size": disk_sz})
6904
        self.op.disks = disks
6905
      else:
6906
        raise errors.OpPrereqError("No disk info specified and the export"
6907
                                   " is missing the disk information",
6908
                                   errors.ECODE_INVAL)
6909

    
6910
    if (not self.op.nics and
6911
        einfo.has_option(constants.INISECT_INS, "nic_count")):
6912
      nics = []
6913
      for idx in range(einfo.getint(constants.INISECT_INS, "nic_count")):
6914
        ndict = {}
6915
        for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]:
6916
          v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
6917
          ndict[name] = v
6918
        nics.append(ndict)
6919
      self.op.nics = nics
6920

    
6921
    if (self.op.hypervisor is None and
6922
        einfo.has_option(constants.INISECT_INS, "hypervisor")):
6923
      self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor")
6924
    if einfo.has_section(constants.INISECT_HYP):
6925
      # use the export parameters but do not override the ones
6926
      # specified by the user
6927
      for name, value in einfo.items(constants.INISECT_HYP):
6928
        if name not in self.op.hvparams:
6929
          self.op.hvparams[name] = value
6930

    
6931
    if einfo.has_section(constants.INISECT_BEP):
6932
      # use the parameters, without overriding
6933
      for name, value in einfo.items(constants.INISECT_BEP):
6934
        if name not in self.op.beparams:
6935
          self.op.beparams[name] = value
6936
    else:
6937
      # try to read the parameters old style, from the main section
6938
      for name in constants.BES_PARAMETERS:
6939
        if (name not in self.op.beparams and
6940
            einfo.has_option(constants.INISECT_INS, name)):
6941
          self.op.beparams[name] = einfo.get(constants.INISECT_INS, name)
6942

    
6943
    if einfo.has_section(constants.INISECT_OSP):
6944
      # use the parameters, without overriding
6945
      for name, value in einfo.items(constants.INISECT_OSP):
6946
        if name not in self.op.osparams:
6947
          self.op.osparams[name] = value
6948

    
6949
  def _RevertToDefaults(self, cluster):
6950
    """Revert the instance parameters to the default values.
6951

6952
    """
6953
    # hvparams
6954
    hv_defs = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, {})
6955
    for name in self.op.hvparams.keys():
6956
      if name in hv_defs and hv_defs[name] == self.op.hvparams[name]:
6957
        del self.op.hvparams[name]
6958
    # beparams
6959
    be_defs = cluster.SimpleFillBE({})
6960
    for name in self.op.beparams.keys():
6961
      if name in be_defs and be_defs[name] == self.op.beparams[name]:
6962
        del self.op.beparams[name]
6963
    # nic params
6964
    nic_defs = cluster.SimpleFillNIC({})
6965
    for nic in self.op.nics:
6966
      for name in constants.NICS_PARAMETERS:
6967
        if name in nic and name in nic_defs and nic[name] == nic_defs[name]:
6968
          del nic[name]
6969
    # osparams
6970
    os_defs = cluster.SimpleFillOS(self.op.os_type, {})
6971
    for name in self.op.osparams.keys():
6972
      if name in os_defs and os_defs[name] == self.op.osparams[name]:
6973
        del self.op.osparams[name]
6974

    
6975
  def CheckPrereq(self):
6976
    """Check prerequisites.
6977

6978
    """
6979
    if self.op.mode == constants.INSTANCE_IMPORT:
6980
      export_info = self._ReadExportInfo()
6981
      self._ReadExportParams(export_info)
6982

    
6983
    _CheckDiskTemplate(self.op.disk_template)
6984

    
6985
    if (not self.cfg.GetVGName() and
6986
        self.op.disk_template not in constants.DTS_NOT_LVM):
6987
      raise errors.OpPrereqError("Cluster does not support lvm-based"
6988
                                 " instances", errors.ECODE_STATE)
6989

    
6990
    if self.op.hypervisor is None:
6991
      self.op.hypervisor = self.cfg.GetHypervisorType()
6992

    
6993
    cluster = self.cfg.GetClusterInfo()
6994
    enabled_hvs = cluster.enabled_hypervisors
6995
    if self.op.hypervisor not in enabled_hvs:
6996
      raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
6997
                                 " cluster (%s)" % (self.op.hypervisor,
6998
                                  ",".join(enabled_hvs)),
6999
                                 errors.ECODE_STATE)
7000

    
7001
    # check hypervisor parameter syntax (locally)
7002
    utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
7003
    filled_hvp = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type,
7004
                                      self.op.hvparams)
7005
    hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
7006
    hv_type.CheckParameterSyntax(filled_hvp)
7007
    self.hv_full = filled_hvp
7008
    # check that we don't specify global parameters on an instance
7009
    _CheckGlobalHvParams(self.op.hvparams)
7010

    
7011
    # fill and remember the beparams dict
7012
    utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
7013
    self.be_full = cluster.SimpleFillBE(self.op.beparams)
7014

    
7015
    # build os parameters
7016
    self.os_full = cluster.SimpleFillOS(self.op.os_type, self.op.osparams)
7017

    
7018
    # now that hvp/bep are in final format, let's reset to defaults,
7019
    # if told to do so
7020
    if self.op.identify_defaults:
7021
      self._RevertToDefaults(cluster)
7022

    
7023
    # NIC buildup
7024
    self.nics = []
7025
    for idx, nic in enumerate(self.op.nics):
7026
      nic_mode_req = nic.get("mode", None)
7027
      nic_mode = nic_mode_req
7028
      if nic_mode is None:
7029
        nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
7030

    
7031
      # in routed mode, for the first nic, the default ip is 'auto'
7032
      if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
7033
        default_ip_mode = constants.VALUE_AUTO
7034
      else:
7035
        default_ip_mode = constants.VALUE_NONE
7036

    
7037
      # ip validity checks
7038
      ip = nic.get("ip", default_ip_mode)
7039
      if ip is None or ip.lower() == constants.VALUE_NONE:
7040
        nic_ip = None
7041
      elif ip.lower() == constants.VALUE_AUTO:
7042
        if not self.op.name_check:
7043
          raise errors.OpPrereqError("IP address set to auto but name checks"
7044
                                     " have been skipped",
7045
                                     errors.ECODE_INVAL)
7046
        nic_ip = self.hostname1.ip
7047
      else:
7048
        if not netutils.IPAddress.IsValid(ip):
7049
          raise errors.OpPrereqError("Invalid IP address '%s'" % ip,
7050
                                     errors.ECODE_INVAL)
7051
        nic_ip = ip
7052

    
7053
      # TODO: check the ip address for uniqueness
7054
      if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
7055
        raise errors.OpPrereqError("Routed nic mode requires an ip address",
7056
                                   errors.ECODE_INVAL)
7057

    
7058
      # MAC address verification
7059
      mac = nic.get("mac", constants.VALUE_AUTO)
7060
      if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7061
        mac = utils.NormalizeAndValidateMac(mac)
7062

    
7063
        try:
7064
          self.cfg.ReserveMAC(mac, self.proc.GetECId())
7065
        except errors.ReservationError:
7066
          raise errors.OpPrereqError("MAC address %s already in use"
7067
                                     " in cluster" % mac,
7068
                                     errors.ECODE_NOTUNIQUE)
7069

    
7070
      # bridge verification
7071
      bridge = nic.get("bridge", None)
7072
      link = nic.get("link", None)
7073
      if bridge and link:
7074
        raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
7075
                                   " at the same time", errors.ECODE_INVAL)
7076
      elif bridge and nic_mode == constants.NIC_MODE_ROUTED:
7077
        raise errors.OpPrereqError("Cannot pass 'bridge' on a routed nic",
7078
                                   errors.ECODE_INVAL)
7079
      elif bridge:
7080
        link = bridge
7081

    
7082
      nicparams = {}
7083
      if nic_mode_req:
7084
        nicparams[constants.NIC_MODE] = nic_mode_req
7085
      if link:
7086
        nicparams[constants.NIC_LINK] = link
7087

    
7088
      check_params = cluster.SimpleFillNIC(nicparams)
7089
      objects.NIC.CheckParameterSyntax(check_params)
7090
      self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
7091

    
7092
    # disk checks/pre-build
7093
    self.disks = []
7094
    for disk in self.op.disks:
7095
      mode = disk.get("mode", constants.DISK_RDWR)
7096
      if mode not in constants.DISK_ACCESS_SET:
7097
        raise errors.OpPrereqError("Invalid disk access mode '%s'" %
7098
                                   mode, errors.ECODE_INVAL)
7099
      size = disk.get("size", None)
7100
      if size is None:
7101
        raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
7102
      try:
7103
        size = int(size)
7104
      except (TypeError, ValueError):
7105
        raise errors.OpPrereqError("Invalid disk size '%s'" % size,
7106
                                   errors.ECODE_INVAL)
7107
      new_disk = {"size": size, "mode": mode}
7108
      if "adopt" in disk:
7109
        new_disk["adopt"] = disk["adopt"]
7110
      self.disks.append(new_disk)
7111

    
7112
    if self.op.mode == constants.INSTANCE_IMPORT:
7113

    
7114
      # Check that the new instance doesn't have less disks than the export
7115
      instance_disks = len(self.disks)
7116
      export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
7117
      if instance_disks < export_disks:
7118
        raise errors.OpPrereqError("Not enough disks to import."
7119
                                   " (instance: %d, export: %d)" %
7120
                                   (instance_disks, export_disks),
7121
                                   errors.ECODE_INVAL)
7122

    
7123
      disk_images = []
7124
      for idx in range(export_disks):
7125
        option = 'disk%d_dump' % idx
7126
        if export_info.has_option(constants.INISECT_INS, option):
7127
          # FIXME: are the old os-es, disk sizes, etc. useful?
7128
          export_name = export_info.get(constants.INISECT_INS, option)
7129
          image = utils.PathJoin(self.op.src_path, export_name)
7130
          disk_images.append(image)
7131
        else:
7132
          disk_images.append(False)
7133

    
7134
      self.src_images = disk_images
7135

    
7136
      old_name = export_info.get(constants.INISECT_INS, 'name')
7137
      try:
7138
        exp_nic_count = export_info.getint(constants.INISECT_INS, 'nic_count')
7139
      except (TypeError, ValueError), err:
7140
        raise errors.OpPrereqError("Invalid export file, nic_count is not"
7141
                                   " an integer: %s" % str(err),
7142
                                   errors.ECODE_STATE)
7143
      if self.op.instance_name == old_name:
7144
        for idx, nic in enumerate(self.nics):
7145
          if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
7146
            nic_mac_ini = 'nic%d_mac' % idx
7147
            nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
7148

    
7149
    # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
7150

    
7151
    # ip ping checks (we use the same ip that was resolved in ExpandNames)
7152
    if self.op.ip_check:
7153
      if netutils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
7154
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
7155
                                   (self.check_ip, self.op.instance_name),
7156
                                   errors.ECODE_NOTUNIQUE)
7157

    
7158
    #### mac address generation
7159
    # By generating here the mac address both the allocator and the hooks get
7160
    # the real final mac address rather than the 'auto' or 'generate' value.
7161
    # There is a race condition between the generation and the instance object
7162
    # creation, which means that we know the mac is valid now, but we're not
7163
    # sure it will be when we actually add the instance. If things go bad
7164
    # adding the instance will abort because of a duplicate mac, and the
7165
    # creation job will fail.
7166
    for nic in self.nics:
7167
      if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7168
        nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
7169

    
7170
    #### allocator run
7171

    
7172
    if self.op.iallocator is not None:
7173
      self._RunAllocator()
7174

    
7175
    #### node related checks
7176

    
7177
    # check primary node
7178
    self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
7179
    assert self.pnode is not None, \
7180
      "Cannot retrieve locked node %s" % self.op.pnode
7181
    if pnode.offline:
7182
      raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
7183
                                 pnode.name, errors.ECODE_STATE)
7184
    if pnode.drained:
7185
      raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
7186
                                 pnode.name, errors.ECODE_STATE)
7187

    
7188
    self.secondaries = []
7189

    
7190
    # mirror node verification
7191
    if self.op.disk_template in constants.DTS_NET_MIRROR:
7192
      if self.op.snode == pnode.name:
7193
        raise errors.OpPrereqError("The secondary node cannot be the"
7194
                                   " primary node.", errors.ECODE_INVAL)
7195
      _CheckNodeOnline(self, self.op.snode)
7196
      _CheckNodeNotDrained(self, self.op.snode)
7197
      self.secondaries.append(self.op.snode)
7198

    
7199
    nodenames = [pnode.name] + self.secondaries
7200

    
7201
    req_size = _ComputeDiskSize(self.op.disk_template,
7202
                                self.disks)
7203

    
7204
    # Check lv size requirements, if not adopting
7205
    if req_size is not None and not self.adopt_disks:
7206
      _CheckNodesFreeDisk(self, nodenames, req_size)
7207

    
7208
    if self.adopt_disks: # instead, we must check the adoption data
7209
      all_lvs = set([i["adopt"] for i in self.disks])
7210
      if len(all_lvs) != len(self.disks):
7211
        raise errors.OpPrereqError("Duplicate volume names given for adoption",
7212
                                   errors.ECODE_INVAL)
7213
      for lv_name in all_lvs:
7214
        try:
7215
          self.cfg.ReserveLV(lv_name, self.proc.GetECId())
7216
        except errors.ReservationError:
7217
          raise errors.OpPrereqError("LV named %s used by another instance" %
7218
                                     lv_name, errors.ECODE_NOTUNIQUE)
7219

    
7220
      node_lvs = self.rpc.call_lv_list([pnode.name],
7221
                                       self.cfg.GetVGName())[pnode.name]
7222
      node_lvs.Raise("Cannot get LV information from node %s" % pnode.name)
7223
      node_lvs = node_lvs.payload
7224
      delta = all_lvs.difference(node_lvs.keys())
7225
      if delta:
7226
        raise errors.OpPrereqError("Missing logical volume(s): %s" %
7227
                                   utils.CommaJoin(delta),
7228
                                   errors.ECODE_INVAL)
7229
      online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]]
7230
      if online_lvs:
7231
        raise errors.OpPrereqError("Online logical volumes found, cannot"
7232
                                   " adopt: %s" % utils.CommaJoin(online_lvs),
7233
                                   errors.ECODE_STATE)
7234
      # update the size of disk based on what is found
7235
      for dsk in self.disks:
7236
        dsk["size"] = int(float(node_lvs[dsk["adopt"]][0]))
7237

    
7238
    _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
7239

    
7240
    _CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant)
7241
    # check OS parameters (remotely)
7242
    _CheckOSParams(self, True, nodenames, self.op.os_type, self.os_full)
7243

    
7244
    _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
7245

    
7246
    # memory check on primary node
7247
    if self.op.start:
7248
      _CheckNodeFreeMemory(self, self.pnode.name,
7249
                           "creating instance %s" % self.op.instance_name,
7250
                           self.be_full[constants.BE_MEMORY],
7251
                           self.op.hypervisor)
7252

    
7253
    self.dry_run_result = list(nodenames)
7254

    
7255
  def Exec(self, feedback_fn):
7256
    """Create and add the instance to the cluster.
7257

7258
    """
7259
    instance = self.op.instance_name
7260
    pnode_name = self.pnode.name
7261

    
7262
    ht_kind = self.op.hypervisor
7263
    if ht_kind in constants.HTS_REQ_PORT:
7264
      network_port = self.cfg.AllocatePort()
7265
    else:
7266
      network_port = None
7267

    
7268
    if constants.ENABLE_FILE_STORAGE:
7269
      # this is needed because os.path.join does not accept None arguments
7270
      if self.op.file_storage_dir is None:
7271
        string_file_storage_dir = ""
7272
      else:
7273
        string_file_storage_dir = self.op.file_storage_dir
7274

    
7275
      # build the full file storage dir path
7276
      file_storage_dir = utils.PathJoin(self.cfg.GetFileStorageDir(),
7277
                                        string_file_storage_dir, instance)
7278
    else:
7279
      file_storage_dir = ""
7280

    
7281
    disks = _GenerateDiskTemplate(self,
7282
                                  self.op.disk_template,
7283
                                  instance, pnode_name,
7284
                                  self.secondaries,
7285
                                  self.disks,
7286
                                  file_storage_dir,
7287
                                  self.op.file_driver,
7288
                                  0)
7289

    
7290
    iobj = objects.Instance(name=instance, os=self.op.os_type,
7291
                            primary_node=pnode_name,
7292
                            nics=self.nics, disks=disks,
7293
                            disk_template=self.op.disk_template,
7294
                            admin_up=False,
7295
                            network_port=network_port,
7296
                            beparams=self.op.beparams,
7297
                            hvparams=self.op.hvparams,
7298
                            hypervisor=self.op.hypervisor,
7299
                            osparams=self.op.osparams,
7300
                            )
7301

    
7302
    if self.adopt_disks:
7303
      # rename LVs to the newly-generated names; we need to construct
7304
      # 'fake' LV disks with the old data, plus the new unique_id
7305
      tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks]
7306
      rename_to = []
7307
      for t_dsk, a_dsk in zip (tmp_disks, self.disks):
7308
        rename_to.append(t_dsk.logical_id)
7309
        t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk["adopt"])
7310
        self.cfg.SetDiskID(t_dsk, pnode_name)
7311
      result = self.rpc.call_blockdev_rename(pnode_name,
7312
                                             zip(tmp_disks, rename_to))
7313
      result.Raise("Failed to rename adoped LVs")
7314
    else:
7315
      feedback_fn("* creating instance disks...")
7316
      try:
7317
        _CreateDisks(self, iobj)
7318
      except errors.OpExecError:
7319
        self.LogWarning("Device creation failed, reverting...")
7320
        try:
7321
          _RemoveDisks(self, iobj)
7322
        finally:
7323
          self.cfg.ReleaseDRBDMinors(instance)
7324
          raise
7325

    
7326
      if self.cfg.GetClusterInfo().prealloc_wipe_disks:
7327
        feedback_fn("* wiping instance disks...")
7328
        try:
7329
          _WipeDisks(self, iobj)
7330
        except errors.OpExecError:
7331
          self.LogWarning("Device wiping failed, reverting...")
7332
          try:
7333
            _RemoveDisks(self, iobj)
7334
          finally:
7335
            self.cfg.ReleaseDRBDMinors(instance)
7336
            raise
7337

    
7338
    feedback_fn("adding instance %s to cluster config" % instance)
7339

    
7340
    self.cfg.AddInstance(iobj, self.proc.GetECId())
7341

    
7342
    # Declare that we don't want to remove the instance lock anymore, as we've
7343
    # added the instance to the config
7344
    del self.remove_locks[locking.LEVEL_INSTANCE]
7345
    # Unlock all the nodes
7346
    if self.op.mode == constants.INSTANCE_IMPORT:
7347
      nodes_keep = [self.op.src_node]
7348
      nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
7349
                       if node != self.op.src_node]
7350
      self.context.glm.release(locking.LEVEL_NODE, nodes_release)
7351
      self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
7352
    else:
7353
      self.context.glm.release(locking.LEVEL_NODE)
7354
      del self.acquired_locks[locking.LEVEL_NODE]
7355

    
7356
    if self.op.wait_for_sync:
7357
      disk_abort = not _WaitForSync(self, iobj)
7358
    elif iobj.disk_template in constants.DTS_NET_MIRROR:
7359
      # make sure the disks are not degraded (still sync-ing is ok)
7360
      time.sleep(15)
7361
      feedback_fn("* checking mirrors status")
7362
      disk_abort = not _WaitForSync(self, iobj, oneshot=True)
7363
    else:
7364
      disk_abort = False
7365

    
7366
    if disk_abort:
7367
      _RemoveDisks(self, iobj)
7368
      self.cfg.RemoveInstance(iobj.name)
7369
      # Make sure the instance lock gets removed
7370
      self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
7371
      raise errors.OpExecError("There are some degraded disks for"
7372
                               " this instance")
7373

    
7374
    if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks:
7375
      if self.op.mode == constants.INSTANCE_CREATE:
7376
        if not self.op.no_install:
7377
          feedback_fn("* running the instance OS create scripts...")
7378
          # FIXME: pass debug option from opcode to backend
7379
          result = self.rpc.call_instance_os_add(pnode_name, iobj, False,
7380
                                                 self.op.debug_level)
7381
          result.Raise("Could not add os for instance %s"
7382
                       " on node %s" % (instance, pnode_name))
7383

    
7384
      elif self.op.mode == constants.INSTANCE_IMPORT:
7385
        feedback_fn("* running the instance OS import scripts...")
7386

    
7387
        transfers = []
7388

    
7389
        for idx, image in enumerate(self.src_images):
7390
          if not image:
7391
            continue
7392

    
7393
          # FIXME: pass debug option from opcode to backend
7394
          dt = masterd.instance.DiskTransfer("disk/%s" % idx,
7395
                                             constants.IEIO_FILE, (image, ),
7396
                                             constants.IEIO_SCRIPT,
7397
                                             (iobj.disks[idx], idx),
7398
                                             None)
7399
          transfers.append(dt)
7400

    
7401
        import_result = \
7402
          masterd.instance.TransferInstanceData(self, feedback_fn,
7403
                                                self.op.src_node, pnode_name,
7404
                                                self.pnode.secondary_ip,
7405
                                                iobj, transfers)
7406
        if not compat.all(import_result):
7407
          self.LogWarning("Some disks for instance %s on node %s were not"
7408
                          " imported successfully" % (instance, pnode_name))
7409

    
7410
      elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
7411
        feedback_fn("* preparing remote import...")
7412
        connect_timeout = constants.RIE_CONNECT_TIMEOUT
7413
        timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
7414

    
7415
        disk_results = masterd.instance.RemoteImport(self, feedback_fn, iobj,
7416
                                                     self.source_x509_ca,
7417
                                                     self._cds, timeouts)
7418
        if not compat.all(disk_results):
7419
          # TODO: Should the instance still be started, even if some disks
7420
          # failed to import (valid for local imports, too)?
7421
          self.LogWarning("Some disks for instance %s on node %s were not"
7422
                          " imported successfully" % (instance, pnode_name))
7423

    
7424
        # Run rename script on newly imported instance
7425
        assert iobj.name == instance
7426
        feedback_fn("Running rename script for %s" % instance)
7427
        result = self.rpc.call_instance_run_rename(pnode_name, iobj,
7428
                                                   self.source_instance_name,
7429
                                                   self.op.debug_level)
7430
        if result.fail_msg:
7431
          self.LogWarning("Failed to run rename script for %s on node"
7432
                          " %s: %s" % (instance, pnode_name, result.fail_msg))
7433

    
7434
      else:
7435
        # also checked in the prereq part
7436
        raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
7437
                                     % self.op.mode)
7438

    
7439
    if self.op.start:
7440
      iobj.admin_up = True
7441
      self.cfg.Update(iobj, feedback_fn)
7442
      logging.info("Starting instance %s on node %s", instance, pnode_name)
7443
      feedback_fn("* starting instance...")
7444
      result = self.rpc.call_instance_start(pnode_name, iobj, None, None)
7445
      result.Raise("Could not start instance")
7446

    
7447
    return list(iobj.all_nodes)
7448

    
7449

    
7450
class LUConnectConsole(NoHooksLU):
7451
  """Connect to an instance's console.
7452

7453
  This is somewhat special in that it returns the command line that
7454
  you need to run on the master node in order to connect to the
7455
  console.
7456

7457
  """
7458
  _OP_PARAMS = [
7459
    _PInstanceName
7460
    ]
7461
  REQ_BGL = False
7462

    
7463
  def ExpandNames(self):
7464
    self._ExpandAndLockInstance()
7465

    
7466
  def CheckPrereq(self):
7467
    """Check prerequisites.
7468

7469
    This checks that the instance is in the cluster.
7470

7471
    """
7472
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7473
    assert self.instance is not None, \
7474
      "Cannot retrieve locked instance %s" % self.op.instance_name
7475
    _CheckNodeOnline(self, self.instance.primary_node)
7476

    
7477
  def Exec(self, feedback_fn):
7478
    """Connect to the console of an instance
7479

7480
    """
7481
    instance = self.instance
7482
    node = instance.primary_node
7483

    
7484
    node_insts = self.rpc.call_instance_list([node],
7485
                                             [instance.hypervisor])[node]
7486
    node_insts.Raise("Can't get node information from %s" % node)
7487

    
7488
    if instance.name not in node_insts.payload:
7489
      if instance.admin_up:
7490
        state = "ERROR_down"
7491
      else:
7492
        state = "ADMIN_down"
7493
      raise errors.OpExecError("Instance %s is not running (state %s)" %
7494
                               (instance.name, state))
7495

    
7496
    logging.debug("Connecting to console of %s on %s", instance.name, node)
7497

    
7498
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
7499
    cluster = self.cfg.GetClusterInfo()
7500
    # beparams and hvparams are passed separately, to avoid editing the
7501
    # instance and then saving the defaults in the instance itself.
7502
    hvparams = cluster.FillHV(instance)
7503
    beparams = cluster.FillBE(instance)
7504
    console_cmd = hyper.GetShellCommandForConsole(instance, hvparams, beparams)
7505

    
7506
    # build ssh cmdline
7507
    return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
7508

    
7509

    
7510
class LUReplaceDisks(LogicalUnit):
7511
  """Replace the disks of an instance.
7512

7513
  """
7514
  HPATH = "mirrors-replace"
7515
  HTYPE = constants.HTYPE_INSTANCE
7516
  _OP_PARAMS = [
7517
    _PInstanceName,
7518
    ("mode", ht.NoDefault, ht.TElemOf(constants.REPLACE_MODES)),
7519
    ("disks", ht.EmptyList, ht.TListOf(ht.TPositiveInt)),
7520
    ("remote_node", None, ht.TMaybeString),
7521
    ("iallocator", None, ht.TMaybeString),
7522
    ("early_release", False, ht.TBool),
7523
    ]
7524
  REQ_BGL = False
7525

    
7526
  def CheckArguments(self):
7527
    TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
7528
                                  self.op.iallocator)
7529

    
7530
  def ExpandNames(self):
7531
    self._ExpandAndLockInstance()
7532

    
7533
    if self.op.iallocator is not None:
7534
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7535

    
7536
    elif self.op.remote_node is not None:
7537
      remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
7538
      self.op.remote_node = remote_node
7539

    
7540
      # Warning: do not remove the locking of the new secondary here
7541
      # unless DRBD8.AddChildren is changed to work in parallel;
7542
      # currently it doesn't since parallel invocations of
7543
      # FindUnusedMinor will conflict
7544
      self.needed_locks[locking.LEVEL_NODE] = [remote_node]
7545
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
7546

    
7547
    else:
7548
      self.needed_locks[locking.LEVEL_NODE] = []
7549
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7550

    
7551
    self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
7552
                                   self.op.iallocator, self.op.remote_node,
7553
                                   self.op.disks, False, self.op.early_release)
7554

    
7555
    self.tasklets = [self.replacer]
7556

    
7557
  def DeclareLocks(self, level):
7558
    # If we're not already locking all nodes in the set we have to declare the
7559
    # instance's primary/secondary nodes.
7560
    if (level == locking.LEVEL_NODE and
7561
        self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
7562
      self._LockInstancesNodes()
7563

    
7564
  def BuildHooksEnv(self):
7565
    """Build hooks env.
7566

7567
    This runs on the master, the primary and all the secondaries.
7568

7569
    """
7570
    instance = self.replacer.instance
7571
    env = {
7572
      "MODE": self.op.mode,
7573
      "NEW_SECONDARY": self.op.remote_node,
7574
      "OLD_SECONDARY": instance.secondary_nodes[0],
7575
      }
7576
    env.update(_BuildInstanceHookEnvByObject(self, instance))
7577
    nl = [
7578
      self.cfg.GetMasterNode(),
7579
      instance.primary_node,
7580
      ]
7581
    if self.op.remote_node is not None:
7582
      nl.append(self.op.remote_node)
7583
    return env, nl, nl
7584

    
7585

    
7586
class TLReplaceDisks(Tasklet):
7587
  """Replaces disks for an instance.
7588

7589
  Note: Locking is not within the scope of this class.
7590

7591
  """
7592
  def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
7593
               disks, delay_iallocator, early_release):
7594
    """Initializes this class.
7595

7596
    """
7597
    Tasklet.__init__(self, lu)
7598

    
7599
    # Parameters
7600
    self.instance_name = instance_name
7601
    self.mode = mode
7602
    self.iallocator_name = iallocator_name
7603
    self.remote_node = remote_node
7604
    self.disks = disks
7605
    self.delay_iallocator = delay_iallocator
7606
    self.early_release = early_release
7607

    
7608
    # Runtime data
7609
    self.instance = None
7610
    self.new_node = None
7611
    self.target_node = None
7612
    self.other_node = None
7613
    self.remote_node_info = None
7614
    self.node_secondary_ip = None
7615

    
7616
  @staticmethod
7617
  def CheckArguments(mode, remote_node, iallocator):
7618
    """Helper function for users of this class.
7619

7620
    """
7621
    # check for valid parameter combination
7622
    if mode == constants.REPLACE_DISK_CHG:
7623
      if remote_node is None and iallocator is None:
7624
        raise errors.OpPrereqError("When changing the secondary either an"
7625
                                   " iallocator script must be used or the"
7626
                                   " new node given", errors.ECODE_INVAL)
7627

    
7628
      if remote_node is not None and iallocator is not None:
7629
        raise errors.OpPrereqError("Give either the iallocator or the new"
7630
                                   " secondary, not both", errors.ECODE_INVAL)
7631

    
7632
    elif remote_node is not None or iallocator is not None:
7633
      # Not replacing the secondary
7634
      raise errors.OpPrereqError("The iallocator and new node options can"
7635
                                 " only be used when changing the"
7636
                                 " secondary node", errors.ECODE_INVAL)
7637

    
7638
  @staticmethod
7639
  def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
7640
    """Compute a new secondary node using an IAllocator.
7641

7642
    """
7643
    ial = IAllocator(lu.cfg, lu.rpc,
7644
                     mode=constants.IALLOCATOR_MODE_RELOC,
7645
                     name=instance_name,
7646
                     relocate_from=relocate_from)
7647

    
7648
    ial.Run(iallocator_name)
7649

    
7650
    if not ial.success:
7651
      raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
7652
                                 " %s" % (iallocator_name, ial.info),
7653
                                 errors.ECODE_NORES)
7654

    
7655
    if len(ial.result) != ial.required_nodes:
7656
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
7657
                                 " of nodes (%s), required %s" %
7658
                                 (iallocator_name,
7659
                                  len(ial.result), ial.required_nodes),
7660
                                 errors.ECODE_FAULT)
7661

    
7662
    remote_node_name = ial.result[0]
7663

    
7664
    lu.LogInfo("Selected new secondary for instance '%s': %s",
7665
               instance_name, remote_node_name)
7666

    
7667
    return remote_node_name
7668

    
7669
  def _FindFaultyDisks(self, node_name):
7670
    return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
7671
                                    node_name, True)
7672

    
7673
  def CheckPrereq(self):
7674
    """Check prerequisites.
7675

7676
    This checks that the instance is in the cluster.
7677

7678
    """
7679
    self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
7680
    assert instance is not None, \
7681
      "Cannot retrieve locked instance %s" % self.instance_name
7682

    
7683
    if instance.disk_template != constants.DT_DRBD8:
7684
      raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
7685
                                 " instances", errors.ECODE_INVAL)
7686

    
7687
    if len(instance.secondary_nodes) != 1:
7688
      raise errors.OpPrereqError("The instance has a strange layout,"
7689
                                 " expected one secondary but found %d" %
7690
                                 len(instance.secondary_nodes),
7691
                                 errors.ECODE_FAULT)
7692

    
7693
    if not self.delay_iallocator:
7694
      self._CheckPrereq2()
7695

    
7696
  def _CheckPrereq2(self):
7697
    """Check prerequisites, second part.
7698

7699
    This function should always be part of CheckPrereq. It was separated and is
7700
    now called from Exec because during node evacuation iallocator was only
7701
    called with an unmodified cluster model, not taking planned changes into
7702
    account.
7703

7704
    """
7705
    instance = self.instance
7706
    secondary_node = instance.secondary_nodes[0]
7707

    
7708
    if self.iallocator_name is None:
7709
      remote_node = self.remote_node
7710
    else:
7711
      remote_node = self._RunAllocator(self.lu, self.iallocator_name,
7712
                                       instance.name, instance.secondary_nodes)
7713

    
7714
    if remote_node is not None:
7715
      self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
7716
      assert self.remote_node_info is not None, \
7717
        "Cannot retrieve locked node %s" % remote_node
7718
    else:
7719
      self.remote_node_info = None
7720

    
7721
    if remote_node == self.instance.primary_node:
7722
      raise errors.OpPrereqError("The specified node is the primary node of"
7723
                                 " the instance.", errors.ECODE_INVAL)
7724

    
7725
    if remote_node == secondary_node:
7726
      raise errors.OpPrereqError("The specified node is already the"
7727
                                 " secondary node of the instance.",
7728
                                 errors.ECODE_INVAL)
7729

    
7730
    if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
7731
                                    constants.REPLACE_DISK_CHG):
7732
      raise errors.OpPrereqError("Cannot specify disks to be replaced",
7733
                                 errors.ECODE_INVAL)
7734

    
7735
    if self.mode == constants.REPLACE_DISK_AUTO:
7736
      faulty_primary = self._FindFaultyDisks(instance.primary_node)
7737
      faulty_secondary = self._FindFaultyDisks(secondary_node)
7738

    
7739
      if faulty_primary and faulty_secondary:
7740
        raise errors.OpPrereqError("Instance %s has faulty disks on more than"
7741
                                   " one node and can not be repaired"
7742
                                   " automatically" % self.instance_name,
7743
                                   errors.ECODE_STATE)
7744

    
7745
      if faulty_primary:
7746
        self.disks = faulty_primary
7747
        self.target_node = instance.primary_node
7748
        self.other_node = secondary_node
7749
        check_nodes = [self.target_node, self.other_node]
7750
      elif faulty_secondary:
7751
        self.disks = faulty_secondary
7752
        self.target_node = secondary_node
7753
        self.other_node = instance.primary_node
7754
        check_nodes = [self.target_node, self.other_node]
7755
      else:
7756
        self.disks = []
7757
        check_nodes = []
7758

    
7759
    else:
7760
      # Non-automatic modes
7761
      if self.mode == constants.REPLACE_DISK_PRI:
7762
        self.target_node = instance.primary_node
7763
        self.other_node = secondary_node
7764
        check_nodes = [self.target_node, self.other_node]
7765

    
7766
      elif self.mode == constants.REPLACE_DISK_SEC:
7767
        self.target_node = secondary_node
7768
        self.other_node = instance.primary_node
7769
        check_nodes = [self.target_node, self.other_node]
7770

    
7771
      elif self.mode == constants.REPLACE_DISK_CHG:
7772
        self.new_node = remote_node
7773
        self.other_node = instance.primary_node
7774
        self.target_node = secondary_node
7775
        check_nodes = [self.new_node, self.other_node]
7776

    
7777
        _CheckNodeNotDrained(self.lu, remote_node)
7778

    
7779
        old_node_info = self.cfg.GetNodeInfo(secondary_node)
7780
        assert old_node_info is not None
7781
        if old_node_info.offline and not self.early_release:
7782
          # doesn't make sense to delay the release
7783
          self.early_release = True
7784
          self.lu.LogInfo("Old secondary %s is offline, automatically enabling"
7785
                          " early-release mode", secondary_node)
7786

    
7787
      else:
7788
        raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
7789
                                     self.mode)
7790

    
7791
      # If not specified all disks should be replaced
7792
      if not self.disks:
7793
        self.disks = range(len(self.instance.disks))
7794

    
7795
    for node in check_nodes:
7796
      _CheckNodeOnline(self.lu, node)
7797

    
7798
    # Check whether disks are valid
7799
    for disk_idx in self.disks:
7800
      instance.FindDisk(disk_idx)
7801

    
7802
    # Get secondary node IP addresses
7803
    node_2nd_ip = {}
7804

    
7805
    for node_name in [self.target_node, self.other_node, self.new_node]:
7806
      if node_name is not None:
7807
        node_2nd_ip[node_name] = self.cfg.GetNodeInfo(node_name).secondary_ip
7808

    
7809
    self.node_secondary_ip = node_2nd_ip
7810

    
7811
  def Exec(self, feedback_fn):
7812
    """Execute disk replacement.
7813

7814
    This dispatches the disk replacement to the appropriate handler.
7815

7816
    """
7817
    if self.delay_iallocator:
7818
      self._CheckPrereq2()
7819

    
7820
    if not self.disks:
7821
      feedback_fn("No disks need replacement")
7822
      return
7823

    
7824
    feedback_fn("Replacing disk(s) %s for %s" %
7825
                (utils.CommaJoin(self.disks), self.instance.name))
7826

    
7827
    activate_disks = (not self.instance.admin_up)
7828

    
7829
    # Activate the instance disks if we're replacing them on a down instance
7830
    if activate_disks:
7831
      _StartInstanceDisks(self.lu, self.instance, True)
7832

    
7833
    try:
7834
      # Should we replace the secondary node?
7835
      if self.new_node is not None:
7836
        fn = self._ExecDrbd8Secondary
7837
      else:
7838
        fn = self._ExecDrbd8DiskOnly
7839

    
7840
      return fn(feedback_fn)
7841

    
7842
    finally:
7843
      # Deactivate the instance disks if we're replacing them on a
7844
      # down instance
7845
      if activate_disks:
7846
        _SafeShutdownInstanceDisks(self.lu, self.instance)
7847

    
7848
  def _CheckVolumeGroup(self, nodes):
7849
    self.lu.LogInfo("Checking volume groups")
7850

    
7851
    vgname = self.cfg.GetVGName()
7852

    
7853
    # Make sure volume group exists on all involved nodes
7854
    results = self.rpc.call_vg_list(nodes)
7855
    if not results:
7856
      raise errors.OpExecError("Can't list volume groups on the nodes")
7857

    
7858
    for node in nodes:
7859
      res = results[node]
7860
      res.Raise("Error checking node %s" % node)
7861
      if vgname not in res.payload:
7862
        raise errors.OpExecError("Volume group '%s' not found on node %s" %
7863
                                 (vgname, node))
7864

    
7865
  def _CheckDisksExistence(self, nodes):
7866
    # Check disk existence
7867
    for idx, dev in enumerate(self.instance.disks):
7868
      if idx not in self.disks:
7869
        continue
7870

    
7871
      for node in nodes:
7872
        self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
7873
        self.cfg.SetDiskID(dev, node)
7874

    
7875
        result = self.rpc.call_blockdev_find(node, dev)
7876

    
7877
        msg = result.fail_msg
7878
        if msg or not result.payload:
7879
          if not msg:
7880
            msg = "disk not found"
7881
          raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
7882
                                   (idx, node, msg))
7883

    
7884
  def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
7885
    for idx, dev in enumerate(self.instance.disks):
7886
      if idx not in self.disks:
7887
        continue
7888

    
7889
      self.lu.LogInfo("Checking disk/%d consistency on node %s" %
7890
                      (idx, node_name))
7891

    
7892
      if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
7893
                                   ldisk=ldisk):
7894
        raise errors.OpExecError("Node %s has degraded storage, unsafe to"
7895
                                 " replace disks for instance %s" %
7896
                                 (node_name, self.instance.name))
7897

    
7898
  def _CreateNewStorage(self, node_name):
7899
    vgname = self.cfg.GetVGName()
7900
    iv_names = {}
7901

    
7902
    for idx, dev in enumerate(self.instance.disks):
7903
      if idx not in self.disks:
7904
        continue
7905

    
7906
      self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
7907

    
7908
      self.cfg.SetDiskID(dev, node_name)
7909

    
7910
      lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
7911
      names = _GenerateUniqueNames(self.lu, lv_names)
7912

    
7913
      lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
7914
                             logical_id=(vgname, names[0]))
7915
      lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
7916
                             logical_id=(vgname, names[1]))
7917

    
7918
      new_lvs = [lv_data, lv_meta]
7919
      old_lvs = dev.children
7920
      iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
7921

    
7922
      # we pass force_create=True to force the LVM creation
7923
      for new_lv in new_lvs:
7924
        _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
7925
                        _GetInstanceInfoText(self.instance), False)
7926

    
7927
    return iv_names
7928

    
7929
  def _CheckDevices(self, node_name, iv_names):
7930
    for name, (dev, _, _) in iv_names.iteritems():
7931
      self.cfg.SetDiskID(dev, node_name)
7932

    
7933
      result = self.rpc.call_blockdev_find(node_name, dev)
7934

    
7935
      msg = result.fail_msg
7936
      if msg or not result.payload:
7937
        if not msg:
7938
          msg = "disk not found"
7939
        raise errors.OpExecError("Can't find DRBD device %s: %s" %
7940
                                 (name, msg))
7941

    
7942
      if result.payload.is_degraded:
7943
        raise errors.OpExecError("DRBD device %s is degraded!" % name)
7944

    
7945
  def _RemoveOldStorage(self, node_name, iv_names):
7946
    for name, (_, old_lvs, _) in iv_names.iteritems():
7947
      self.lu.LogInfo("Remove logical volumes for %s" % name)
7948

    
7949
      for lv in old_lvs:
7950
        self.cfg.SetDiskID(lv, node_name)
7951

    
7952
        msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
7953
        if msg:
7954
          self.lu.LogWarning("Can't remove old LV: %s" % msg,
7955
                             hint="remove unused LVs manually")
7956

    
7957
  def _ReleaseNodeLock(self, node_name):
7958
    """Releases the lock for a given node."""
7959
    self.lu.context.glm.release(locking.LEVEL_NODE, node_name)
7960

    
7961
  def _ExecDrbd8DiskOnly(self, feedback_fn):
7962
    """Replace a disk on the primary or secondary for DRBD 8.
7963

7964
    The algorithm for replace is quite complicated:
7965

7966
      1. for each disk to be replaced:
7967

7968
        1. create new LVs on the target node with unique names
7969
        1. detach old LVs from the drbd device
7970
        1. rename old LVs to name_replaced.<time_t>
7971
        1. rename new LVs to old LVs
7972
        1. attach the new LVs (with the old names now) to the drbd device
7973

7974
      1. wait for sync across all devices
7975

7976
      1. for each modified disk:
7977

7978
        1. remove old LVs (which have the name name_replaces.<time_t>)
7979

7980
    Failures are not very well handled.
7981

7982
    """
7983
    steps_total = 6
7984

    
7985
    # Step: check device activation
7986
    self.lu.LogStep(1, steps_total, "Check device existence")
7987
    self._CheckDisksExistence([self.other_node, self.target_node])
7988
    self._CheckVolumeGroup([self.target_node, self.other_node])
7989

    
7990
    # Step: check other node consistency
7991
    self.lu.LogStep(2, steps_total, "Check peer consistency")
7992
    self._CheckDisksConsistency(self.other_node,
7993
                                self.other_node == self.instance.primary_node,
7994
                                False)
7995

    
7996
    # Step: create new storage
7997
    self.lu.LogStep(3, steps_total, "Allocate new storage")
7998
    iv_names = self._CreateNewStorage(self.target_node)
7999

    
8000
    # Step: for each lv, detach+rename*2+attach
8001
    self.lu.LogStep(4, steps_total, "Changing drbd configuration")
8002
    for dev, old_lvs, new_lvs in iv_names.itervalues():
8003
      self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
8004

    
8005
      result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
8006
                                                     old_lvs)
8007
      result.Raise("Can't detach drbd from local storage on node"
8008
                   " %s for device %s" % (self.target_node, dev.iv_name))
8009
      #dev.children = []
8010
      #cfg.Update(instance)
8011

    
8012
      # ok, we created the new LVs, so now we know we have the needed
8013
      # storage; as such, we proceed on the target node to rename
8014
      # old_lv to _old, and new_lv to old_lv; note that we rename LVs
8015
      # using the assumption that logical_id == physical_id (which in
8016
      # turn is the unique_id on that node)
8017

    
8018
      # FIXME(iustin): use a better name for the replaced LVs
8019
      temp_suffix = int(time.time())
8020
      ren_fn = lambda d, suff: (d.physical_id[0],
8021
                                d.physical_id[1] + "_replaced-%s" % suff)
8022

    
8023
      # Build the rename list based on what LVs exist on the node
8024
      rename_old_to_new = []
8025
      for to_ren in old_lvs:
8026
        result = self.rpc.call_blockdev_find(self.target_node, to_ren)
8027
        if not result.fail_msg and result.payload:
8028
          # device exists
8029
          rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
8030

    
8031
      self.lu.LogInfo("Renaming the old LVs on the target node")
8032
      result = self.rpc.call_blockdev_rename(self.target_node,
8033
                                             rename_old_to_new)
8034
      result.Raise("Can't rename old LVs on node %s" % self.target_node)
8035

    
8036
      # Now we rename the new LVs to the old LVs
8037
      self.lu.LogInfo("Renaming the new LVs on the target node")
8038
      rename_new_to_old = [(new, old.physical_id)
8039
                           for old, new in zip(old_lvs, new_lvs)]
8040
      result = self.rpc.call_blockdev_rename(self.target_node,
8041
                                             rename_new_to_old)
8042
      result.Raise("Can't rename new LVs on node %s" % self.target_node)
8043

    
8044
      for old, new in zip(old_lvs, new_lvs):
8045
        new.logical_id = old.logical_id
8046
        self.cfg.SetDiskID(new, self.target_node)
8047

    
8048
      for disk in old_lvs:
8049
        disk.logical_id = ren_fn(disk, temp_suffix)
8050
        self.cfg.SetDiskID(disk, self.target_node)
8051

    
8052
      # Now that the new lvs have the old name, we can add them to the device
8053
      self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
8054
      result = self.rpc.call_blockdev_addchildren(self.target_node, dev,
8055
                                                  new_lvs)
8056
      msg = result.fail_msg
8057
      if msg:
8058
        for new_lv in new_lvs:
8059
          msg2 = self.rpc.call_blockdev_remove(self.target_node,
8060
                                               new_lv).fail_msg
8061
          if msg2:
8062
            self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
8063
                               hint=("cleanup manually the unused logical"
8064
                                     "volumes"))
8065
        raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
8066

    
8067
      dev.children = new_lvs
8068

    
8069
      self.cfg.Update(self.instance, feedback_fn)
8070

    
8071
    cstep = 5
8072
    if self.early_release:
8073
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
8074
      cstep += 1
8075
      self._RemoveOldStorage(self.target_node, iv_names)
8076
      # WARNING: we release both node locks here, do not do other RPCs
8077
      # than WaitForSync to the primary node
8078
      self._ReleaseNodeLock([self.target_node, self.other_node])
8079

    
8080
    # Wait for sync
8081
    # This can fail as the old devices are degraded and _WaitForSync
8082
    # does a combined result over all disks, so we don't check its return value
8083
    self.lu.LogStep(cstep, steps_total, "Sync devices")
8084
    cstep += 1
8085
    _WaitForSync(self.lu, self.instance)
8086

    
8087
    # Check all devices manually
8088
    self._CheckDevices(self.instance.primary_node, iv_names)
8089

    
8090
    # Step: remove old storage
8091
    if not self.early_release:
8092
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
8093
      cstep += 1
8094
      self._RemoveOldStorage(self.target_node, iv_names)
8095

    
8096
  def _ExecDrbd8Secondary(self, feedback_fn):
8097
    """Replace the secondary node for DRBD 8.
8098

8099
    The algorithm for replace is quite complicated:
8100
      - for all disks of the instance:
8101
        - create new LVs on the new node with same names
8102
        - shutdown the drbd device on the old secondary
8103
        - disconnect the drbd network on the primary
8104
        - create the drbd device on the new secondary
8105
        - network attach the drbd on the primary, using an artifice:
8106
          the drbd code for Attach() will connect to the network if it
8107
          finds a device which is connected to the good local disks but
8108
          not network enabled
8109
      - wait for sync across all devices
8110
      - remove all disks from the old secondary
8111

8112
    Failures are not very well handled.
8113

8114
    """
8115
    steps_total = 6
8116

    
8117
    # Step: check device activation
8118
    self.lu.LogStep(1, steps_total, "Check device existence")
8119
    self._CheckDisksExistence([self.instance.primary_node])
8120
    self._CheckVolumeGroup([self.instance.primary_node])
8121

    
8122
    # Step: check other node consistency
8123
    self.lu.LogStep(2, steps_total, "Check peer consistency")
8124
    self._CheckDisksConsistency(self.instance.primary_node, True, True)
8125

    
8126
    # Step: create new storage
8127
    self.lu.LogStep(3, steps_total, "Allocate new storage")
8128
    for idx, dev in enumerate(self.instance.disks):
8129
      self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
8130
                      (self.new_node, idx))
8131
      # we pass force_create=True to force LVM creation
8132
      for new_lv in dev.children:
8133
        _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
8134
                        _GetInstanceInfoText(self.instance), False)
8135

    
8136
    # Step 4: dbrd minors and drbd setups changes
8137
    # after this, we must manually remove the drbd minors on both the
8138
    # error and the success paths
8139
    self.lu.LogStep(4, steps_total, "Changing drbd configuration")
8140
    minors = self.cfg.AllocateDRBDMinor([self.new_node
8141
                                         for dev in self.instance.disks],
8142
                                        self.instance.name)
8143
    logging.debug("Allocated minors %r", minors)
8144

    
8145
    iv_names = {}
8146
    for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
8147
      self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
8148
                      (self.new_node, idx))
8149
      # create new devices on new_node; note that we create two IDs:
8150
      # one without port, so the drbd will be activated without
8151
      # networking information on the new node at this stage, and one
8152
      # with network, for the latter activation in step 4
8153
      (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
8154
      if self.instance.primary_node == o_node1:
8155
        p_minor = o_minor1
8156
      else:
8157
        assert self.instance.primary_node == o_node2, "Three-node instance?"
8158
        p_minor = o_minor2
8159

    
8160
      new_alone_id = (self.instance.primary_node, self.new_node, None,
8161
                      p_minor, new_minor, o_secret)
8162
      new_net_id = (self.instance.primary_node, self.new_node, o_port,
8163
                    p_minor, new_minor, o_secret)
8164

    
8165
      iv_names[idx] = (dev, dev.children, new_net_id)
8166
      logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
8167
                    new_net_id)
8168
      new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
8169
                              logical_id=new_alone_id,
8170
                              children=dev.children,
8171
                              size=dev.size)
8172
      try:
8173
        _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
8174
                              _GetInstanceInfoText(self.instance), False)
8175
      except errors.GenericError:
8176
        self.cfg.ReleaseDRBDMinors(self.instance.name)
8177
        raise
8178

    
8179
    # We have new devices, shutdown the drbd on the old secondary
8180
    for idx, dev in enumerate(self.instance.disks):
8181
      self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
8182
      self.cfg.SetDiskID(dev, self.target_node)
8183
      msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
8184
      if msg:
8185
        self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
8186
                           "node: %s" % (idx, msg),
8187
                           hint=("Please cleanup this device manually as"
8188
                                 " soon as possible"))
8189

    
8190
    self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
8191
    result = self.rpc.call_drbd_disconnect_net([self.instance.primary_node],
8192
                                               self.node_secondary_ip,
8193
                                               self.instance.disks)\
8194
                                              [self.instance.primary_node]
8195

    
8196
    msg = result.fail_msg
8197
    if msg:
8198
      # detaches didn't succeed (unlikely)
8199
      self.cfg.ReleaseDRBDMinors(self.instance.name)
8200
      raise errors.OpExecError("Can't detach the disks from the network on"
8201
                               " old node: %s" % (msg,))
8202

    
8203
    # if we managed to detach at least one, we update all the disks of
8204
    # the instance to point to the new secondary
8205
    self.lu.LogInfo("Updating instance configuration")
8206
    for dev, _, new_logical_id in iv_names.itervalues():
8207
      dev.logical_id = new_logical_id
8208
      self.cfg.SetDiskID(dev, self.instance.primary_node)
8209

    
8210
    self.cfg.Update(self.instance, feedback_fn)
8211

    
8212
    # and now perform the drbd attach
8213
    self.lu.LogInfo("Attaching primary drbds to new secondary"
8214
                    " (standalone => connected)")
8215
    result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
8216
                                            self.new_node],
8217
                                           self.node_secondary_ip,
8218
                                           self.instance.disks,
8219
                                           self.instance.name,
8220
                                           False)
8221
    for to_node, to_result in result.items():
8222
      msg = to_result.fail_msg
8223
      if msg:
8224
        self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
8225
                           to_node, msg,
8226
                           hint=("please do a gnt-instance info to see the"
8227
                                 " status of disks"))
8228
    cstep = 5
8229
    if self.early_release:
8230
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
8231
      cstep += 1
8232
      self._RemoveOldStorage(self.target_node, iv_names)
8233
      # WARNING: we release all node locks here, do not do other RPCs
8234
      # than WaitForSync to the primary node
8235
      self._ReleaseNodeLock([self.instance.primary_node,
8236
                             self.target_node,
8237
                             self.new_node])
8238

    
8239
    # Wait for sync
8240
    # This can fail as the old devices are degraded and _WaitForSync
8241
    # does a combined result over all disks, so we don't check its return value
8242
    self.lu.LogStep(cstep, steps_total, "Sync devices")
8243
    cstep += 1
8244
    _WaitForSync(self.lu, self.instance)
8245

    
8246
    # Check all devices manually
8247
    self._CheckDevices(self.instance.primary_node, iv_names)
8248

    
8249
    # Step: remove old storage
8250
    if not self.early_release:
8251
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
8252
      self._RemoveOldStorage(self.target_node, iv_names)
8253

    
8254

    
8255
class LURepairNodeStorage(NoHooksLU):
8256
  """Repairs the volume group on a node.
8257

8258
  """
8259
  _OP_PARAMS = [
8260
    _PNodeName,
8261
    ("storage_type", ht.NoDefault, _CheckStorageType),
8262
    ("name", ht.NoDefault, ht.TNonEmptyString),
8263
    ("ignore_consistency", False, ht.TBool),
8264
    ]
8265
  REQ_BGL = False
8266

    
8267
  def CheckArguments(self):
8268
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
8269

    
8270
    storage_type = self.op.storage_type
8271

    
8272
    if (constants.SO_FIX_CONSISTENCY not in
8273
        constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
8274
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
8275
                                 " repaired" % storage_type,
8276
                                 errors.ECODE_INVAL)
8277

    
8278
  def ExpandNames(self):
8279
    self.needed_locks = {
8280
      locking.LEVEL_NODE: [self.op.node_name],
8281
      }
8282

    
8283
  def _CheckFaultyDisks(self, instance, node_name):
8284
    """Ensure faulty disks abort the opcode or at least warn."""
8285
    try:
8286
      if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
8287
                                  node_name, True):
8288
        raise errors.OpPrereqError("Instance '%s' has faulty disks on"
8289
                                   " node '%s'" % (instance.name, node_name),
8290
                                   errors.ECODE_STATE)
8291
    except errors.OpPrereqError, err:
8292
      if self.op.ignore_consistency:
8293
        self.proc.LogWarning(str(err.args[0]))
8294
      else:
8295
        raise
8296

    
8297
  def CheckPrereq(self):
8298
    """Check prerequisites.
8299

8300
    """
8301
    # Check whether any instance on this node has faulty disks
8302
    for inst in _GetNodeInstances(self.cfg, self.op.node_name):
8303
      if not inst.admin_up:
8304
        continue
8305
      check_nodes = set(inst.all_nodes)
8306
      check_nodes.discard(self.op.node_name)
8307
      for inst_node_name in check_nodes:
8308
        self._CheckFaultyDisks(inst, inst_node_name)
8309

    
8310
  def Exec(self, feedback_fn):
8311
    feedback_fn("Repairing storage unit '%s' on %s ..." %
8312
                (self.op.name, self.op.node_name))
8313

    
8314
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
8315
    result = self.rpc.call_storage_execute(self.op.node_name,
8316
                                           self.op.storage_type, st_args,
8317
                                           self.op.name,
8318
                                           constants.SO_FIX_CONSISTENCY)
8319
    result.Raise("Failed to repair storage unit '%s' on %s" %
8320
                 (self.op.name, self.op.node_name))
8321

    
8322

    
8323
class LUNodeEvacuationStrategy(NoHooksLU):
8324
  """Computes the node evacuation strategy.
8325

8326
  """
8327
  _OP_PARAMS = [
8328
    ("nodes", ht.NoDefault, ht.TListOf(ht.TNonEmptyString)),
8329
    ("remote_node", None, ht.TMaybeString),
8330
    ("iallocator", None, ht.TMaybeString),
8331
    ]
8332
  REQ_BGL = False
8333

    
8334
  def CheckArguments(self):
8335
    _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
8336

    
8337
  def ExpandNames(self):
8338
    self.op.nodes = _GetWantedNodes(self, self.op.nodes)
8339
    self.needed_locks = locks = {}
8340
    if self.op.remote_node is None:
8341
      locks[locking.LEVEL_NODE] = locking.ALL_SET
8342
    else:
8343
      self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8344
      locks[locking.LEVEL_NODE] = self.op.nodes + [self.op.remote_node]
8345

    
8346
  def Exec(self, feedback_fn):
8347
    if self.op.remote_node is not None:
8348
      instances = []
8349
      for node in self.op.nodes:
8350
        instances.extend(_GetNodeSecondaryInstances(self.cfg, node))
8351
      result = []
8352
      for i in instances:
8353
        if i.primary_node == self.op.remote_node:
8354
          raise errors.OpPrereqError("Node %s is the primary node of"
8355
                                     " instance %s, cannot use it as"
8356
                                     " secondary" %
8357
                                     (self.op.remote_node, i.name),
8358
                                     errors.ECODE_INVAL)
8359
        result.append([i.name, self.op.remote_node])
8360
    else:
8361
      ial = IAllocator(self.cfg, self.rpc,
8362
                       mode=constants.IALLOCATOR_MODE_MEVAC,
8363
                       evac_nodes=self.op.nodes)
8364
      ial.Run(self.op.iallocator, validate=True)
8365
      if not ial.success:
8366
        raise errors.OpExecError("No valid evacuation solution: %s" % ial.info,
8367
                                 errors.ECODE_NORES)
8368
      result = ial.result
8369
    return result
8370

    
8371

    
8372
class LUGrowDisk(LogicalUnit):
8373
  """Grow a disk of an instance.
8374

8375
  """
8376
  HPATH = "disk-grow"
8377
  HTYPE = constants.HTYPE_INSTANCE
8378
  _OP_PARAMS = [
8379
    _PInstanceName,
8380
    ("disk", ht.NoDefault, ht.TInt),
8381
    ("amount", ht.NoDefault, ht.TInt),
8382
    ("wait_for_sync", True, ht.TBool),
8383
    ]
8384
  REQ_BGL = False
8385

    
8386
  def ExpandNames(self):
8387
    self._ExpandAndLockInstance()
8388
    self.needed_locks[locking.LEVEL_NODE] = []
8389
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8390

    
8391
  def DeclareLocks(self, level):
8392
    if level == locking.LEVEL_NODE:
8393
      self._LockInstancesNodes()
8394

    
8395
  def BuildHooksEnv(self):
8396
    """Build hooks env.
8397

8398
    This runs on the master, the primary and all the secondaries.
8399

8400
    """
8401
    env = {
8402
      "DISK": self.op.disk,
8403
      "AMOUNT": self.op.amount,
8404
      }
8405
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
8406
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
8407
    return env, nl, nl
8408

    
8409
  def CheckPrereq(self):
8410
    """Check prerequisites.
8411

8412
    This checks that the instance is in the cluster.
8413

8414
    """
8415
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
8416
    assert instance is not None, \
8417
      "Cannot retrieve locked instance %s" % self.op.instance_name
8418
    nodenames = list(instance.all_nodes)
8419
    for node in nodenames:
8420
      _CheckNodeOnline(self, node)
8421

    
8422
    self.instance = instance
8423

    
8424
    if instance.disk_template not in constants.DTS_GROWABLE:
8425
      raise errors.OpPrereqError("Instance's disk layout does not support"
8426
                                 " growing.", errors.ECODE_INVAL)
8427

    
8428
    self.disk = instance.FindDisk(self.op.disk)
8429

    
8430
    if instance.disk_template != constants.DT_FILE:
8431
      # TODO: check the free disk space for file, when that feature will be
8432
      # supported
8433
      _CheckNodesFreeDisk(self, nodenames, self.op.amount)
8434

    
8435
  def Exec(self, feedback_fn):
8436
    """Execute disk grow.
8437

8438
    """
8439
    instance = self.instance
8440
    disk = self.disk
8441

    
8442
    disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk])
8443
    if not disks_ok:
8444
      raise errors.OpExecError("Cannot activate block device to grow")
8445

    
8446
    for node in instance.all_nodes:
8447
      self.cfg.SetDiskID(disk, node)
8448
      result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
8449
      result.Raise("Grow request failed to node %s" % node)
8450

    
8451
      # TODO: Rewrite code to work properly
8452
      # DRBD goes into sync mode for a short amount of time after executing the
8453
      # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
8454
      # calling "resize" in sync mode fails. Sleeping for a short amount of
8455
      # time is a work-around.
8456
      time.sleep(5)
8457

    
8458
    disk.RecordGrow(self.op.amount)
8459
    self.cfg.Update(instance, feedback_fn)
8460
    if self.op.wait_for_sync:
8461
      disk_abort = not _WaitForSync(self, instance, disks=[disk])
8462
      if disk_abort:
8463
        self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
8464
                             " status.\nPlease check the instance.")
8465
      if not instance.admin_up:
8466
        _SafeShutdownInstanceDisks(self, instance, disks=[disk])
8467
    elif not instance.admin_up:
8468
      self.proc.LogWarning("Not shutting down the disk even if the instance is"
8469
                           " not supposed to be running because no wait for"
8470
                           " sync mode was requested.")
8471

    
8472

    
8473
class LUQueryInstanceData(NoHooksLU):
8474
  """Query runtime instance data.
8475

8476
  """
8477
  _OP_PARAMS = [
8478
    ("instances", ht.EmptyList, ht.TListOf(ht.TNonEmptyString)),
8479
    ("static", False, ht.TBool),
8480
    ]
8481
  REQ_BGL = False
8482

    
8483
  def ExpandNames(self):
8484
    self.needed_locks = {}
8485
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
8486

    
8487
    if self.op.instances:
8488
      self.wanted_names = []
8489
      for name in self.op.instances:
8490
        full_name = _ExpandInstanceName(self.cfg, name)
8491
        self.wanted_names.append(full_name)
8492
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
8493
    else:
8494
      self.wanted_names = None
8495
      self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
8496

    
8497
    self.needed_locks[locking.LEVEL_NODE] = []
8498
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8499

    
8500
  def DeclareLocks(self, level):
8501
    if level == locking.LEVEL_NODE:
8502
      self._LockInstancesNodes()
8503

    
8504
  def CheckPrereq(self):
8505
    """Check prerequisites.
8506

8507
    This only checks the optional instance list against the existing names.
8508

8509
    """
8510
    if self.wanted_names is None:
8511
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
8512

    
8513
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
8514
                             in self.wanted_names]
8515

    
8516
  def _ComputeBlockdevStatus(self, node, instance_name, dev):
8517
    """Returns the status of a block device
8518

8519
    """
8520
    if self.op.static or not node:
8521
      return None
8522

    
8523
    self.cfg.SetDiskID(dev, node)
8524

    
8525
    result = self.rpc.call_blockdev_find(node, dev)
8526
    if result.offline:
8527
      return None
8528

    
8529
    result.Raise("Can't compute disk status for %s" % instance_name)
8530

    
8531
    status = result.payload
8532
    if status is None:
8533
      return None
8534

    
8535
    return (status.dev_path, status.major, status.minor,
8536
            status.sync_percent, status.estimated_time,
8537
            status.is_degraded, status.ldisk_status)
8538

    
8539
  def _ComputeDiskStatus(self, instance, snode, dev):
8540
    """Compute block device status.
8541

8542
    """
8543
    if dev.dev_type in constants.LDS_DRBD:
8544
      # we change the snode then (otherwise we use the one passed in)
8545
      if dev.logical_id[0] == instance.primary_node:
8546
        snode = dev.logical_id[1]
8547
      else:
8548
        snode = dev.logical_id[0]
8549

    
8550
    dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
8551
                                              instance.name, dev)
8552
    dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
8553

    
8554
    if dev.children:
8555
      dev_children = [self._ComputeDiskStatus(instance, snode, child)
8556
                      for child in dev.children]
8557
    else:
8558
      dev_children = []
8559

    
8560
    data = {
8561
      "iv_name": dev.iv_name,
8562
      "dev_type": dev.dev_type,
8563
      "logical_id": dev.logical_id,
8564
      "physical_id": dev.physical_id,
8565
      "pstatus": dev_pstatus,
8566
      "sstatus": dev_sstatus,
8567
      "children": dev_children,
8568
      "mode": dev.mode,
8569
      "size": dev.size,
8570
      }
8571

    
8572
    return data
8573

    
8574
  def Exec(self, feedback_fn):
8575
    """Gather and return data"""
8576
    result = {}
8577

    
8578
    cluster = self.cfg.GetClusterInfo()
8579

    
8580
    for instance in self.wanted_instances:
8581
      if not self.op.static:
8582
        remote_info = self.rpc.call_instance_info(instance.primary_node,
8583
                                                  instance.name,
8584
                                                  instance.hypervisor)
8585
        remote_info.Raise("Error checking node %s" % instance.primary_node)
8586
        remote_info = remote_info.payload
8587
        if remote_info and "state" in remote_info:
8588
          remote_state = "up"
8589
        else:
8590
          remote_state = "down"
8591
      else:
8592
        remote_state = None
8593
      if instance.admin_up:
8594
        config_state = "up"
8595
      else:
8596
        config_state = "down"
8597

    
8598
      disks = [self._ComputeDiskStatus(instance, None, device)
8599
               for device in instance.disks]
8600

    
8601
      idict = {
8602
        "name": instance.name,
8603
        "config_state": config_state,
8604
        "run_state": remote_state,
8605
        "pnode": instance.primary_node,
8606
        "snodes": instance.secondary_nodes,
8607
        "os": instance.os,
8608
        # this happens to be the same format used for hooks
8609
        "nics": _NICListToTuple(self, instance.nics),
8610
        "disk_template": instance.disk_template,
8611
        "disks": disks,
8612
        "hypervisor": instance.hypervisor,
8613
        "network_port": instance.network_port,
8614
        "hv_instance": instance.hvparams,
8615
        "hv_actual": cluster.FillHV(instance, skip_globals=True),
8616
        "be_instance": instance.beparams,
8617
        "be_actual": cluster.FillBE(instance),
8618
        "os_instance": instance.osparams,
8619
        "os_actual": cluster.SimpleFillOS(instance.os, instance.osparams),
8620
        "serial_no": instance.serial_no,
8621
        "mtime": instance.mtime,
8622
        "ctime": instance.ctime,
8623
        "uuid": instance.uuid,
8624
        }
8625

    
8626
      result[instance.name] = idict
8627

    
8628
    return result
8629

    
8630

    
8631
class LUSetInstanceParams(LogicalUnit):
8632
  """Modifies an instances's parameters.
8633

8634
  """
8635
  HPATH = "instance-modify"
8636
  HTYPE = constants.HTYPE_INSTANCE
8637
  _OP_PARAMS = [
8638
    _PInstanceName,
8639
    ("nics", ht.EmptyList, ht.TList),
8640
    ("disks", ht.EmptyList, ht.TList),
8641
    ("beparams", ht.EmptyDict, ht.TDict),
8642
    ("hvparams", ht.EmptyDict, ht.TDict),
8643
    ("disk_template", None, ht.TMaybeString),
8644
    ("remote_node", None, ht.TMaybeString),
8645
    ("os_name", None, ht.TMaybeString),
8646
    ("force_variant", False, ht.TBool),
8647
    ("osparams", None, ht.TOr(ht.TDict, ht.TNone)),
8648
    _PForce,
8649
    ]
8650
  REQ_BGL = False
8651

    
8652
  def CheckArguments(self):
8653
    if not (self.op.nics or self.op.disks or self.op.disk_template or
8654
            self.op.hvparams or self.op.beparams or self.op.os_name):
8655
      raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
8656

    
8657
    if self.op.hvparams:
8658
      _CheckGlobalHvParams(self.op.hvparams)
8659

    
8660
    # Disk validation
8661
    disk_addremove = 0
8662
    for disk_op, disk_dict in self.op.disks:
8663
      utils.ForceDictType(disk_dict, constants.IDISK_PARAMS_TYPES)
8664
      if disk_op == constants.DDM_REMOVE:
8665
        disk_addremove += 1
8666
        continue
8667
      elif disk_op == constants.DDM_ADD:
8668
        disk_addremove += 1
8669
      else:
8670
        if not isinstance(disk_op, int):
8671
          raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL)
8672
        if not isinstance(disk_dict, dict):
8673
          msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
8674
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
8675

    
8676
      if disk_op == constants.DDM_ADD:
8677
        mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
8678
        if mode not in constants.DISK_ACCESS_SET:
8679
          raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
8680
                                     errors.ECODE_INVAL)
8681
        size = disk_dict.get('size', None)
8682
        if size is None:
8683
          raise errors.OpPrereqError("Required disk parameter size missing",
8684
                                     errors.ECODE_INVAL)
8685
        try:
8686
          size = int(size)
8687
        except (TypeError, ValueError), err:
8688
          raise errors.OpPrereqError("Invalid disk size parameter: %s" %
8689
                                     str(err), errors.ECODE_INVAL)
8690
        disk_dict['size'] = size
8691
      else:
8692
        # modification of disk
8693
        if 'size' in disk_dict:
8694
          raise errors.OpPrereqError("Disk size change not possible, use"
8695
                                     " grow-disk", errors.ECODE_INVAL)
8696

    
8697
    if disk_addremove > 1:
8698
      raise errors.OpPrereqError("Only one disk add or remove operation"
8699
                                 " supported at a time", errors.ECODE_INVAL)
8700

    
8701
    if self.op.disks and self.op.disk_template is not None:
8702
      raise errors.OpPrereqError("Disk template conversion and other disk"
8703
                                 " changes not supported at the same time",
8704
                                 errors.ECODE_INVAL)
8705

    
8706
    if self.op.disk_template:
8707
      _CheckDiskTemplate(self.op.disk_template)
8708
      if (self.op.disk_template in constants.DTS_NET_MIRROR and
8709
          self.op.remote_node is None):
8710
        raise errors.OpPrereqError("Changing the disk template to a mirrored"
8711
                                   " one requires specifying a secondary node",
8712
                                   errors.ECODE_INVAL)
8713

    
8714
    # NIC validation
8715
    nic_addremove = 0
8716
    for nic_op, nic_dict in self.op.nics:
8717
      utils.ForceDictType(nic_dict, constants.INIC_PARAMS_TYPES)
8718
      if nic_op == constants.DDM_REMOVE:
8719
        nic_addremove += 1
8720
        continue
8721
      elif nic_op == constants.DDM_ADD:
8722
        nic_addremove += 1
8723
      else:
8724
        if not isinstance(nic_op, int):
8725
          raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL)
8726
        if not isinstance(nic_dict, dict):
8727
          msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
8728
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
8729

    
8730
      # nic_dict should be a dict
8731
      nic_ip = nic_dict.get('ip', None)
8732
      if nic_ip is not None:
8733
        if nic_ip.lower() == constants.VALUE_NONE:
8734
          nic_dict['ip'] = None
8735
        else:
8736
          if not netutils.IPAddress.IsValid(nic_ip):
8737
            raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip,
8738
                                       errors.ECODE_INVAL)
8739

    
8740
      nic_bridge = nic_dict.get('bridge', None)
8741
      nic_link = nic_dict.get('link', None)
8742
      if nic_bridge and nic_link:
8743
        raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
8744
                                   " at the same time", errors.ECODE_INVAL)
8745
      elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
8746
        nic_dict['bridge'] = None
8747
      elif nic_link and nic_link.lower() == constants.VALUE_NONE:
8748
        nic_dict['link'] = None
8749

    
8750
      if nic_op == constants.DDM_ADD:
8751
        nic_mac = nic_dict.get('mac', None)
8752
        if nic_mac is None:
8753
          nic_dict['mac'] = constants.VALUE_AUTO
8754

    
8755
      if 'mac' in nic_dict:
8756
        nic_mac = nic_dict['mac']
8757
        if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
8758
          nic_mac = utils.NormalizeAndValidateMac(nic_mac)
8759

    
8760
        if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
8761
          raise errors.OpPrereqError("'auto' is not a valid MAC address when"
8762
                                     " modifying an existing nic",
8763
                                     errors.ECODE_INVAL)
8764

    
8765
    if nic_addremove > 1:
8766
      raise errors.OpPrereqError("Only one NIC add or remove operation"
8767
                                 " supported at a time", errors.ECODE_INVAL)
8768

    
8769
  def ExpandNames(self):
8770
    self._ExpandAndLockInstance()
8771
    self.needed_locks[locking.LEVEL_NODE] = []
8772
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8773

    
8774
  def DeclareLocks(self, level):
8775
    if level == locking.LEVEL_NODE:
8776
      self._LockInstancesNodes()
8777
      if self.op.disk_template and self.op.remote_node:
8778
        self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8779
        self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)
8780

    
8781
  def BuildHooksEnv(self):
8782
    """Build hooks env.
8783

8784
    This runs on the master, primary and secondaries.
8785

8786
    """
8787
    args = dict()
8788
    if constants.BE_MEMORY in self.be_new:
8789
      args['memory'] = self.be_new[constants.BE_MEMORY]
8790
    if constants.BE_VCPUS in self.be_new:
8791
      args['vcpus'] = self.be_new[constants.BE_VCPUS]
8792
    # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
8793
    # information at all.
8794
    if self.op.nics:
8795
      args['nics'] = []
8796
      nic_override = dict(self.op.nics)
8797
      for idx, nic in enumerate(self.instance.nics):
8798
        if idx in nic_override:
8799
          this_nic_override = nic_override[idx]
8800
        else:
8801
          this_nic_override = {}
8802
        if 'ip' in this_nic_override:
8803
          ip = this_nic_override['ip']
8804
        else:
8805
          ip = nic.ip
8806
        if 'mac' in this_nic_override:
8807
          mac = this_nic_override['mac']
8808
        else:
8809
          mac = nic.mac
8810
        if idx in self.nic_pnew:
8811
          nicparams = self.nic_pnew[idx]
8812
        else:
8813
          nicparams = self.cluster.SimpleFillNIC(nic.nicparams)
8814
        mode = nicparams[constants.NIC_MODE]
8815
        link = nicparams[constants.NIC_LINK]
8816
        args['nics'].append((ip, mac, mode, link))
8817
      if constants.DDM_ADD in nic_override:
8818
        ip = nic_override[constants.DDM_ADD].get('ip', None)
8819
        mac = nic_override[constants.DDM_ADD]['mac']
8820
        nicparams = self.nic_pnew[constants.DDM_ADD]
8821
        mode = nicparams[constants.NIC_MODE]
8822
        link = nicparams[constants.NIC_LINK]
8823
        args['nics'].append((ip, mac, mode, link))
8824
      elif constants.DDM_REMOVE in nic_override:
8825
        del args['nics'][-1]
8826

    
8827
    env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
8828
    if self.op.disk_template:
8829
      env["NEW_DISK_TEMPLATE"] = self.op.disk_template
8830
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
8831
    return env, nl, nl
8832

    
8833
  def CheckPrereq(self):
8834
    """Check prerequisites.
8835

8836
    This only checks the instance list against the existing names.
8837

8838
    """
8839
    # checking the new params on the primary/secondary nodes
8840

    
8841
    instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
8842
    cluster = self.cluster = self.cfg.GetClusterInfo()
8843
    assert self.instance is not None, \
8844
      "Cannot retrieve locked instance %s" % self.op.instance_name
8845
    pnode = instance.primary_node
8846
    nodelist = list(instance.all_nodes)
8847

    
8848
    # OS change
8849
    if self.op.os_name and not self.op.force:
8850
      _CheckNodeHasOS(self, instance.primary_node, self.op.os_name,
8851
                      self.op.force_variant)
8852
      instance_os = self.op.os_name
8853
    else:
8854
      instance_os = instance.os
8855

    
8856
    if self.op.disk_template:
8857
      if instance.disk_template == self.op.disk_template:
8858
        raise errors.OpPrereqError("Instance already has disk template %s" %
8859
                                   instance.disk_template, errors.ECODE_INVAL)
8860

    
8861
      if (instance.disk_template,
8862
          self.op.disk_template) not in self._DISK_CONVERSIONS:
8863
        raise errors.OpPrereqError("Unsupported disk template conversion from"
8864
                                   " %s to %s" % (instance.disk_template,
8865
                                                  self.op.disk_template),
8866
                                   errors.ECODE_INVAL)
8867
      _CheckInstanceDown(self, instance, "cannot change disk template")
8868
      if self.op.disk_template in constants.DTS_NET_MIRROR:
8869
        if self.op.remote_node == pnode:
8870
          raise errors.OpPrereqError("Given new secondary node %s is the same"
8871
                                     " as the primary node of the instance" %
8872
                                     self.op.remote_node, errors.ECODE_STATE)
8873
        _CheckNodeOnline(self, self.op.remote_node)
8874
        _CheckNodeNotDrained(self, self.op.remote_node)
8875
        disks = [{"size": d.size} for d in instance.disks]
8876
        required = _ComputeDiskSize(self.op.disk_template, disks)
8877
        _CheckNodesFreeDisk(self, [self.op.remote_node], required)
8878

    
8879
    # hvparams processing
8880
    if self.op.hvparams:
8881
      hv_type = instance.hypervisor
8882
      i_hvdict = _GetUpdatedParams(instance.hvparams, self.op.hvparams)
8883
      utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES)
8884
      hv_new = cluster.SimpleFillHV(hv_type, instance.os, i_hvdict)
8885

    
8886
      # local check
8887
      hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new)
8888
      _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
8889
      self.hv_new = hv_new # the new actual values
8890
      self.hv_inst = i_hvdict # the new dict (without defaults)
8891
    else:
8892
      self.hv_new = self.hv_inst = {}
8893

    
8894
    # beparams processing
8895
    if self.op.beparams:
8896
      i_bedict = _GetUpdatedParams(instance.beparams, self.op.beparams,
8897
                                   use_none=True)
8898
      utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
8899
      be_new = cluster.SimpleFillBE(i_bedict)
8900
      self.be_new = be_new # the new actual values
8901
      self.be_inst = i_bedict # the new dict (without defaults)
8902
    else:
8903
      self.be_new = self.be_inst = {}
8904

    
8905
    # osparams processing
8906
    if self.op.osparams:
8907
      i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
8908
      _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
8909
      self.os_inst = i_osdict # the new dict (without defaults)
8910
    else:
8911
      self.os_inst = {}
8912

    
8913
    self.warn = []
8914

    
8915
    if constants.BE_MEMORY in self.op.beparams and not self.op.force:
8916
      mem_check_list = [pnode]
8917
      if be_new[constants.BE_AUTO_BALANCE]:
8918
        # either we changed auto_balance to yes or it was from before
8919
        mem_check_list.extend(instance.secondary_nodes)
8920
      instance_info = self.rpc.call_instance_info(pnode, instance.name,
8921
                                                  instance.hypervisor)
8922
      nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
8923
                                         instance.hypervisor)
8924
      pninfo = nodeinfo[pnode]
8925
      msg = pninfo.fail_msg
8926
      if msg:
8927
        # Assume the primary node is unreachable and go ahead
8928
        self.warn.append("Can't get info from primary node %s: %s" %
8929
                         (pnode,  msg))
8930
      elif not isinstance(pninfo.payload.get('memory_free', None), int):
8931
        self.warn.append("Node data from primary node %s doesn't contain"
8932
                         " free memory information" % pnode)
8933
      elif instance_info.fail_msg:
8934
        self.warn.append("Can't get instance runtime information: %s" %
8935
                        instance_info.fail_msg)
8936
      else:
8937
        if instance_info.payload:
8938
          current_mem = int(instance_info.payload['memory'])
8939
        else:
8940
          # Assume instance not running
8941
          # (there is a slight race condition here, but it's not very probable,
8942
          # and we have no other way to check)
8943
          current_mem = 0
8944
        miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
8945
                    pninfo.payload['memory_free'])
8946
        if miss_mem > 0:
8947
          raise errors.OpPrereqError("This change will prevent the instance"
8948
                                     " from starting, due to %d MB of memory"
8949
                                     " missing on its primary node" % miss_mem,
8950
                                     errors.ECODE_NORES)
8951

    
8952
      if be_new[constants.BE_AUTO_BALANCE]:
8953
        for node, nres in nodeinfo.items():
8954
          if node not in instance.secondary_nodes:
8955
            continue
8956
          msg = nres.fail_msg
8957
          if msg:
8958
            self.warn.append("Can't get info from secondary node %s: %s" %
8959
                             (node, msg))
8960
          elif not isinstance(nres.payload.get('memory_free', None), int):
8961
            self.warn.append("Secondary node %s didn't return free"
8962
                             " memory information" % node)
8963
          elif be_new[constants.BE_MEMORY] > nres.payload['memory_free']:
8964
            self.warn.append("Not enough memory to failover instance to"
8965
                             " secondary node %s" % node)
8966

    
8967
    # NIC processing
8968
    self.nic_pnew = {}
8969
    self.nic_pinst = {}
8970
    for nic_op, nic_dict in self.op.nics:
8971
      if nic_op == constants.DDM_REMOVE:
8972
        if not instance.nics:
8973
          raise errors.OpPrereqError("Instance has no NICs, cannot remove",
8974
                                     errors.ECODE_INVAL)
8975
        continue
8976
      if nic_op != constants.DDM_ADD:
8977
        # an existing nic
8978
        if not instance.nics:
8979
          raise errors.OpPrereqError("Invalid NIC index %s, instance has"
8980
                                     " no NICs" % nic_op,
8981
                                     errors.ECODE_INVAL)
8982
        if nic_op < 0 or nic_op >= len(instance.nics):
8983
          raise errors.OpPrereqError("Invalid NIC index %s, valid values"
8984
                                     " are 0 to %d" %
8985
                                     (nic_op, len(instance.nics) - 1),
8986
                                     errors.ECODE_INVAL)
8987
        old_nic_params = instance.nics[nic_op].nicparams
8988
        old_nic_ip = instance.nics[nic_op].ip
8989
      else:
8990
        old_nic_params = {}
8991
        old_nic_ip = None
8992

    
8993
      update_params_dict = dict([(key, nic_dict[key])
8994
                                 for key in constants.NICS_PARAMETERS
8995
                                 if key in nic_dict])
8996

    
8997
      if 'bridge' in nic_dict:
8998
        update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
8999

    
9000
      new_nic_params = _GetUpdatedParams(old_nic_params,
9001
                                         update_params_dict)
9002
      utils.ForceDictType(new_nic_params, constants.NICS_PARAMETER_TYPES)
9003
      new_filled_nic_params = cluster.SimpleFillNIC(new_nic_params)
9004
      objects.NIC.CheckParameterSyntax(new_filled_nic_params)
9005
      self.nic_pinst[nic_op] = new_nic_params
9006
      self.nic_pnew[nic_op] = new_filled_nic_params
9007
      new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
9008

    
9009
      if new_nic_mode == constants.NIC_MODE_BRIDGED:
9010
        nic_bridge = new_filled_nic_params[constants.NIC_LINK]
9011
        msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
9012
        if msg:
9013
          msg = "Error checking bridges on node %s: %s" % (pnode, msg)
9014
          if self.op.force:
9015
            self.warn.append(msg)
9016
          else:
9017
            raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
9018
      if new_nic_mode == constants.NIC_MODE_ROUTED:
9019
        if 'ip' in nic_dict:
9020
          nic_ip = nic_dict['ip']
9021
        else:
9022
          nic_ip = old_nic_ip
9023
        if nic_ip is None:
9024
          raise errors.OpPrereqError('Cannot set the nic ip to None'
9025
                                     ' on a routed nic', errors.ECODE_INVAL)
9026
      if 'mac' in nic_dict:
9027
        nic_mac = nic_dict['mac']
9028
        if nic_mac is None:
9029
          raise errors.OpPrereqError('Cannot set the nic mac to None',
9030
                                     errors.ECODE_INVAL)
9031
        elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
9032
          # otherwise generate the mac
9033
          nic_dict['mac'] = self.cfg.GenerateMAC(self.proc.GetECId())
9034
        else:
9035
          # or validate/reserve the current one
9036
          try:
9037
            self.cfg.ReserveMAC(nic_mac, self.proc.GetECId())
9038
          except errors.ReservationError:
9039
            raise errors.OpPrereqError("MAC address %s already in use"
9040
                                       " in cluster" % nic_mac,
9041
                                       errors.ECODE_NOTUNIQUE)
9042

    
9043
    # DISK processing
9044
    if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
9045
      raise errors.OpPrereqError("Disk operations not supported for"
9046
                                 " diskless instances",
9047
                                 errors.ECODE_INVAL)
9048
    for disk_op, _ in self.op.disks:
9049
      if disk_op == constants.DDM_REMOVE:
9050
        if len(instance.disks) == 1:
9051
          raise errors.OpPrereqError("Cannot remove the last disk of"
9052
                                     " an instance", errors.ECODE_INVAL)
9053
        _CheckInstanceDown(self, instance, "cannot remove disks")
9054

    
9055
      if (disk_op == constants.DDM_ADD and
9056
          len(instance.nics) >= constants.MAX_DISKS):
9057
        raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
9058
                                   " add more" % constants.MAX_DISKS,
9059
                                   errors.ECODE_STATE)
9060
      if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
9061
        # an existing disk
9062
        if disk_op < 0 or disk_op >= len(instance.disks):
9063
          raise errors.OpPrereqError("Invalid disk index %s, valid values"
9064
                                     " are 0 to %d" %
9065
                                     (disk_op, len(instance.disks)),
9066
                                     errors.ECODE_INVAL)
9067

    
9068
    return
9069

    
9070
  def _ConvertPlainToDrbd(self, feedback_fn):
9071
    """Converts an instance from plain to drbd.
9072

9073
    """
9074
    feedback_fn("Converting template to drbd")
9075
    instance = self.instance
9076
    pnode = instance.primary_node
9077
    snode = self.op.remote_node
9078

    
9079
    # create a fake disk info for _GenerateDiskTemplate
9080
    disk_info = [{"size": d.size, "mode": d.mode} for d in instance.disks]
9081
    new_disks = _GenerateDiskTemplate(self, self.op.disk_template,
9082
                                      instance.name, pnode, [snode],
9083
                                      disk_info, None, None, 0)
9084
    info = _GetInstanceInfoText(instance)
9085
    feedback_fn("Creating aditional volumes...")
9086
    # first, create the missing data and meta devices
9087
    for disk in new_disks:
9088
      # unfortunately this is... not too nice
9089
      _CreateSingleBlockDev(self, pnode, instance, disk.children[1],
9090
                            info, True)
9091
      for child in disk.children:
9092
        _CreateSingleBlockDev(self, snode, instance, child, info, True)
9093
    # at this stage, all new LVs have been created, we can rename the
9094
    # old ones
9095
    feedback_fn("Renaming original volumes...")
9096
    rename_list = [(o, n.children[0].logical_id)
9097
                   for (o, n) in zip(instance.disks, new_disks)]
9098
    result = self.rpc.call_blockdev_rename(pnode, rename_list)
9099
    result.Raise("Failed to rename original LVs")
9100

    
9101
    feedback_fn("Initializing DRBD devices...")
9102
    # all child devices are in place, we can now create the DRBD devices
9103
    for disk in new_disks:
9104
      for node in [pnode, snode]:
9105
        f_create = node == pnode
9106
        _CreateSingleBlockDev(self, node, instance, disk, info, f_create)
9107

    
9108
    # at this point, the instance has been modified
9109
    instance.disk_template = constants.DT_DRBD8
9110
    instance.disks = new_disks
9111
    self.cfg.Update(instance, feedback_fn)
9112

    
9113
    # disks are created, waiting for sync
9114
    disk_abort = not _WaitForSync(self, instance)
9115
    if disk_abort:
9116
      raise errors.OpExecError("There are some degraded disks for"
9117
                               " this instance, please cleanup manually")
9118

    
9119
  def _ConvertDrbdToPlain(self, feedback_fn):
9120
    """Converts an instance from drbd to plain.
9121

9122
    """
9123
    instance = self.instance
9124
    assert len(instance.secondary_nodes) == 1
9125
    pnode = instance.primary_node
9126
    snode = instance.secondary_nodes[0]
9127
    feedback_fn("Converting template to plain")
9128

    
9129
    old_disks = instance.disks
9130
    new_disks = [d.children[0] for d in old_disks]
9131

    
9132
    # copy over size and mode
9133
    for parent, child in zip(old_disks, new_disks):
9134
      child.size = parent.size
9135
      child.mode = parent.mode
9136

    
9137
    # update instance structure
9138
    instance.disks = new_disks
9139
    instance.disk_template = constants.DT_PLAIN
9140
    self.cfg.Update(instance, feedback_fn)
9141

    
9142
    feedback_fn("Removing volumes on the secondary node...")
9143
    for disk in old_disks:
9144
      self.cfg.SetDiskID(disk, snode)
9145
      msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg
9146
      if msg:
9147
        self.LogWarning("Could not remove block device %s on node %s,"
9148
                        " continuing anyway: %s", disk.iv_name, snode, msg)
9149

    
9150
    feedback_fn("Removing unneeded volumes on the primary node...")
9151
    for idx, disk in enumerate(old_disks):
9152
      meta = disk.children[1]
9153
      self.cfg.SetDiskID(meta, pnode)
9154
      msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg
9155
      if msg:
9156
        self.LogWarning("Could not remove metadata for disk %d on node %s,"
9157
                        " continuing anyway: %s", idx, pnode, msg)
9158

    
9159

    
9160
  def Exec(self, feedback_fn):
9161
    """Modifies an instance.
9162

9163
    All parameters take effect only at the next restart of the instance.
9164

9165
    """
9166
    # Process here the warnings from CheckPrereq, as we don't have a
9167
    # feedback_fn there.
9168
    for warn in self.warn:
9169
      feedback_fn("WARNING: %s" % warn)
9170

    
9171
    result = []
9172
    instance = self.instance
9173
    # disk changes
9174
    for disk_op, disk_dict in self.op.disks:
9175
      if disk_op == constants.DDM_REMOVE:
9176
        # remove the last disk
9177
        device = instance.disks.pop()
9178
        device_idx = len(instance.disks)
9179
        for node, disk in device.ComputeNodeTree(instance.primary_node):
9180
          self.cfg.SetDiskID(disk, node)
9181
          msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
9182
          if msg:
9183
            self.LogWarning("Could not remove disk/%d on node %s: %s,"
9184
                            " continuing anyway", device_idx, node, msg)
9185
        result.append(("disk/%d" % device_idx, "remove"))
9186
      elif disk_op == constants.DDM_ADD:
9187
        # add a new disk
9188
        if instance.disk_template == constants.DT_FILE:
9189
          file_driver, file_path = instance.disks[0].logical_id
9190
          file_path = os.path.dirname(file_path)
9191
        else:
9192
          file_driver = file_path = None
9193
        disk_idx_base = len(instance.disks)
9194
        new_disk = _GenerateDiskTemplate(self,
9195
                                         instance.disk_template,
9196
                                         instance.name, instance.primary_node,
9197
                                         instance.secondary_nodes,
9198
                                         [disk_dict],
9199
                                         file_path,
9200
                                         file_driver,
9201
                                         disk_idx_base)[0]
9202
        instance.disks.append(new_disk)
9203
        info = _GetInstanceInfoText(instance)
9204

    
9205
        logging.info("Creating volume %s for instance %s",
9206
                     new_disk.iv_name, instance.name)
9207
        # Note: this needs to be kept in sync with _CreateDisks
9208
        #HARDCODE
9209
        for node in instance.all_nodes:
9210
          f_create = node == instance.primary_node
9211
          try:
9212
            _CreateBlockDev(self, node, instance, new_disk,
9213
                            f_create, info, f_create)
9214
          except errors.OpExecError, err:
9215
            self.LogWarning("Failed to create volume %s (%s) on"
9216
                            " node %s: %s",
9217
                            new_disk.iv_name, new_disk, node, err)
9218
        result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
9219
                       (new_disk.size, new_disk.mode)))
9220
      else:
9221
        # change a given disk
9222
        instance.disks[disk_op].mode = disk_dict['mode']
9223
        result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
9224

    
9225
    if self.op.disk_template:
9226
      r_shut = _ShutdownInstanceDisks(self, instance)
9227
      if not r_shut:
9228
        raise errors.OpExecError("Cannot shutdow instance disks, unable to"
9229
                                 " proceed with disk template conversion")
9230
      mode = (instance.disk_template, self.op.disk_template)
9231
      try:
9232
        self._DISK_CONVERSIONS[mode](self, feedback_fn)
9233
      except:
9234
        self.cfg.ReleaseDRBDMinors(instance.name)
9235
        raise
9236
      result.append(("disk_template", self.op.disk_template))
9237

    
9238
    # NIC changes
9239
    for nic_op, nic_dict in self.op.nics:
9240
      if nic_op == constants.DDM_REMOVE:
9241
        # remove the last nic
9242
        del instance.nics[-1]
9243
        result.append(("nic.%d" % len(instance.nics), "remove"))
9244
      elif nic_op == constants.DDM_ADD:
9245
        # mac and bridge should be set, by now
9246
        mac = nic_dict['mac']
9247
        ip = nic_dict.get('ip', None)
9248
        nicparams = self.nic_pinst[constants.DDM_ADD]
9249
        new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
9250
        instance.nics.append(new_nic)
9251
        result.append(("nic.%d" % (len(instance.nics) - 1),
9252
                       "add:mac=%s,ip=%s,mode=%s,link=%s" %
9253
                       (new_nic.mac, new_nic.ip,
9254
                        self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
9255
                        self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
9256
                       )))
9257
      else:
9258
        for key in 'mac', 'ip':
9259
          if key in nic_dict:
9260
            setattr(instance.nics[nic_op], key, nic_dict[key])
9261
        if nic_op in self.nic_pinst:
9262
          instance.nics[nic_op].nicparams = self.nic_pinst[nic_op]
9263
        for key, val in nic_dict.iteritems():
9264
          result.append(("nic.%s/%d" % (key, nic_op), val))
9265

    
9266
    # hvparams changes
9267
    if self.op.hvparams:
9268
      instance.hvparams = self.hv_inst
9269
      for key, val in self.op.hvparams.iteritems():
9270
        result.append(("hv/%s" % key, val))
9271

    
9272
    # beparams changes
9273
    if self.op.beparams:
9274
      instance.beparams = self.be_inst
9275
      for key, val in self.op.beparams.iteritems():
9276
        result.append(("be/%s" % key, val))
9277

    
9278
    # OS change
9279
    if self.op.os_name:
9280
      instance.os = self.op.os_name
9281

    
9282
    # osparams changes
9283
    if self.op.osparams:
9284
      instance.osparams = self.os_inst
9285
      for key, val in self.op.osparams.iteritems():
9286
        result.append(("os/%s" % key, val))
9287

    
9288
    self.cfg.Update(instance, feedback_fn)
9289

    
9290
    return result
9291

    
9292
  _DISK_CONVERSIONS = {
9293
    (constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd,
9294
    (constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain,
9295
    }
9296

    
9297

    
9298
class LUQueryExports(NoHooksLU):
9299
  """Query the exports list
9300

9301
  """
9302
  _OP_PARAMS = [
9303
    ("nodes", ht.EmptyList, ht.TListOf(ht.TNonEmptyString)),
9304
    ("use_locking", False, ht.TBool),
9305
    ]
9306
  REQ_BGL = False
9307

    
9308
  def ExpandNames(self):
9309
    self.needed_locks = {}
9310
    self.share_locks[locking.LEVEL_NODE] = 1
9311
    if not self.op.nodes:
9312
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9313
    else:
9314
      self.needed_locks[locking.LEVEL_NODE] = \
9315
        _GetWantedNodes(self, self.op.nodes)
9316

    
9317
  def Exec(self, feedback_fn):
9318
    """Compute the list of all the exported system images.
9319

9320
    @rtype: dict
9321
    @return: a dictionary with the structure node->(export-list)
9322
        where export-list is a list of the instances exported on
9323
        that node.
9324

9325
    """
9326
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
9327
    rpcresult = self.rpc.call_export_list(self.nodes)
9328
    result = {}
9329
    for node in rpcresult:
9330
      if rpcresult[node].fail_msg:
9331
        result[node] = False
9332
      else:
9333
        result[node] = rpcresult[node].payload
9334

    
9335
    return result
9336

    
9337

    
9338
class LUPrepareExport(NoHooksLU):
9339
  """Prepares an instance for an export and returns useful information.
9340

9341
  """
9342
  _OP_PARAMS = [
9343
    _PInstanceName,
9344
    ("mode", ht.NoDefault, ht.TElemOf(constants.EXPORT_MODES)),
9345
    ]
9346
  REQ_BGL = False
9347

    
9348
  def ExpandNames(self):
9349
    self._ExpandAndLockInstance()
9350

    
9351
  def CheckPrereq(self):
9352
    """Check prerequisites.
9353

9354
    """
9355
    instance_name = self.op.instance_name
9356

    
9357
    self.instance = self.cfg.GetInstanceInfo(instance_name)
9358
    assert self.instance is not None, \
9359
          "Cannot retrieve locked instance %s" % self.op.instance_name
9360
    _CheckNodeOnline(self, self.instance.primary_node)
9361

    
9362
    self._cds = _GetClusterDomainSecret()
9363

    
9364
  def Exec(self, feedback_fn):
9365
    """Prepares an instance for an export.
9366

9367
    """
9368
    instance = self.instance
9369

    
9370
    if self.op.mode == constants.EXPORT_MODE_REMOTE:
9371
      salt = utils.GenerateSecret(8)
9372

    
9373
      feedback_fn("Generating X509 certificate on %s" % instance.primary_node)
9374
      result = self.rpc.call_x509_cert_create(instance.primary_node,
9375
                                              constants.RIE_CERT_VALIDITY)
9376
      result.Raise("Can't create X509 key and certificate on %s" % result.node)
9377

    
9378
      (name, cert_pem) = result.payload
9379

    
9380
      cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
9381
                                             cert_pem)
9382

    
9383
      return {
9384
        "handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds),
9385
        "x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt),
9386
                          salt),
9387
        "x509_ca": utils.SignX509Certificate(cert, self._cds, salt),
9388
        }
9389

    
9390
    return None
9391

    
9392

    
9393
class LUExportInstance(LogicalUnit):
9394
  """Export an instance to an image in the cluster.
9395

9396
  """
9397
  HPATH = "instance-export"
9398
  HTYPE = constants.HTYPE_INSTANCE
9399
  _OP_PARAMS = [
9400
    _PInstanceName,
9401
    ("target_node", ht.NoDefault, ht.TOr(ht.TNonEmptyString, ht.TList)),
9402
    ("shutdown", True, ht.TBool),
9403
    _PShutdownTimeout,
9404
    ("remove_instance", False, ht.TBool),
9405
    ("ignore_remove_failures", False, ht.TBool),
9406
    ("mode", constants.EXPORT_MODE_LOCAL, ht.TElemOf(constants.EXPORT_MODES)),
9407
    ("x509_key_name", None, ht.TOr(ht.TList, ht.TNone)),
9408
    ("destination_x509_ca", None, ht.TMaybeString),
9409
    ]
9410
  REQ_BGL = False
9411

    
9412
  def CheckArguments(self):
9413
    """Check the arguments.
9414

9415
    """
9416
    self.x509_key_name = self.op.x509_key_name
9417
    self.dest_x509_ca_pem = self.op.destination_x509_ca
9418

    
9419
    if self.op.remove_instance and not self.op.shutdown:
9420
      raise errors.OpPrereqError("Can not remove instance without shutting it"
9421
                                 " down before")
9422

    
9423
    if self.op.mode == constants.EXPORT_MODE_REMOTE:
9424
      if not self.x509_key_name:
9425
        raise errors.OpPrereqError("Missing X509 key name for encryption",
9426
                                   errors.ECODE_INVAL)
9427

    
9428
      if not self.dest_x509_ca_pem:
9429
        raise errors.OpPrereqError("Missing destination X509 CA",
9430
                                   errors.ECODE_INVAL)
9431

    
9432
  def ExpandNames(self):
9433
    self._ExpandAndLockInstance()
9434

    
9435
    # Lock all nodes for local exports
9436
    if self.op.mode == constants.EXPORT_MODE_LOCAL:
9437
      # FIXME: lock only instance primary and destination node
9438
      #
9439
      # Sad but true, for now we have do lock all nodes, as we don't know where
9440
      # the previous export might be, and in this LU we search for it and
9441
      # remove it from its current node. In the future we could fix this by:
9442
      #  - making a tasklet to search (share-lock all), then create the
9443
      #    new one, then one to remove, after
9444
      #  - removing the removal operation altogether
9445
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9446

    
9447
  def DeclareLocks(self, level):
9448
    """Last minute lock declaration."""
9449
    # All nodes are locked anyway, so nothing to do here.
9450

    
9451
  def BuildHooksEnv(self):
9452
    """Build hooks env.
9453

9454
    This will run on the master, primary node and target node.
9455

9456
    """
9457
    env = {
9458
      "EXPORT_MODE": self.op.mode,
9459
      "EXPORT_NODE": self.op.target_node,
9460
      "EXPORT_DO_SHUTDOWN": self.op.shutdown,
9461
      "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
9462
      # TODO: Generic function for boolean env variables
9463
      "REMOVE_INSTANCE": str(bool(self.op.remove_instance)),
9464
      }
9465

    
9466
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
9467

    
9468
    nl = [self.cfg.GetMasterNode(), self.instance.primary_node]
9469

    
9470
    if self.op.mode == constants.EXPORT_MODE_LOCAL:
9471
      nl.append(self.op.target_node)
9472

    
9473
    return env, nl, nl
9474

    
9475
  def CheckPrereq(self):
9476
    """Check prerequisites.
9477

9478
    This checks that the instance and node names are valid.
9479

9480
    """
9481
    instance_name = self.op.instance_name
9482

    
9483
    self.instance = self.cfg.GetInstanceInfo(instance_name)
9484
    assert self.instance is not None, \
9485
          "Cannot retrieve locked instance %s" % self.op.instance_name
9486
    _CheckNodeOnline(self, self.instance.primary_node)
9487

    
9488
    if self.op.mode == constants.EXPORT_MODE_LOCAL:
9489
      self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
9490
      self.dst_node = self.cfg.GetNodeInfo(self.op.target_node)
9491
      assert self.dst_node is not None
9492

    
9493
      _CheckNodeOnline(self, self.dst_node.name)
9494
      _CheckNodeNotDrained(self, self.dst_node.name)
9495

    
9496
      self._cds = None
9497
      self.dest_disk_info = None
9498
      self.dest_x509_ca = None
9499

    
9500
    elif self.op.mode == constants.EXPORT_MODE_REMOTE:
9501
      self.dst_node = None
9502

    
9503
      if len(self.op.target_node) != len(self.instance.disks):
9504
        raise errors.OpPrereqError(("Received destination information for %s"
9505
                                    " disks, but instance %s has %s disks") %
9506
                                   (len(self.op.target_node), instance_name,
9507
                                    len(self.instance.disks)),
9508
                                   errors.ECODE_INVAL)
9509

    
9510
      cds = _GetClusterDomainSecret()
9511

    
9512
      # Check X509 key name
9513
      try:
9514
        (key_name, hmac_digest, hmac_salt) = self.x509_key_name
9515
      except (TypeError, ValueError), err:
9516
        raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err)
9517

    
9518
      if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt):
9519
        raise errors.OpPrereqError("HMAC for X509 key name is wrong",
9520
                                   errors.ECODE_INVAL)
9521

    
9522
      # Load and verify CA
9523
      try:
9524
        (cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds)
9525
      except OpenSSL.crypto.Error, err:
9526
        raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" %
9527
                                   (err, ), errors.ECODE_INVAL)
9528

    
9529
      (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
9530
      if errcode is not None:
9531
        raise errors.OpPrereqError("Invalid destination X509 CA (%s)" %
9532
                                   (msg, ), errors.ECODE_INVAL)
9533

    
9534
      self.dest_x509_ca = cert
9535

    
9536
      # Verify target information
9537
      disk_info = []
9538
      for idx, disk_data in enumerate(self.op.target_node):
9539
        try:
9540
          (host, port, magic) = \
9541
            masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data)
9542
        except errors.GenericError, err:
9543
          raise errors.OpPrereqError("Target info for disk %s: %s" %
9544
                                     (idx, err), errors.ECODE_INVAL)
9545

    
9546
        disk_info.append((host, port, magic))
9547

    
9548
      assert len(disk_info) == len(self.op.target_node)
9549
      self.dest_disk_info = disk_info
9550

    
9551
    else:
9552
      raise errors.ProgrammerError("Unhandled export mode %r" %
9553
                                   self.op.mode)
9554

    
9555
    # instance disk type verification
9556
    # TODO: Implement export support for file-based disks
9557
    for disk in self.instance.disks:
9558
      if disk.dev_type == constants.LD_FILE:
9559
        raise errors.OpPrereqError("Export not supported for instances with"
9560
                                   " file-based disks", errors.ECODE_INVAL)
9561

    
9562
  def _CleanupExports(self, feedback_fn):
9563
    """Removes exports of current instance from all other nodes.
9564

9565
    If an instance in a cluster with nodes A..D was exported to node C, its
9566
    exports will be removed from the nodes A, B and D.
9567

9568
    """
9569
    assert self.op.mode != constants.EXPORT_MODE_REMOTE
9570

    
9571
    nodelist = self.cfg.GetNodeList()
9572
    nodelist.remove(self.dst_node.name)
9573

    
9574
    # on one-node clusters nodelist will be empty after the removal
9575
    # if we proceed the backup would be removed because OpQueryExports
9576
    # substitutes an empty list with the full cluster node list.
9577
    iname = self.instance.name
9578
    if nodelist:
9579
      feedback_fn("Removing old exports for instance %s" % iname)
9580
      exportlist = self.rpc.call_export_list(nodelist)
9581
      for node in exportlist:
9582
        if exportlist[node].fail_msg:
9583
          continue
9584
        if iname in exportlist[node].payload:
9585
          msg = self.rpc.call_export_remove(node, iname).fail_msg
9586
          if msg:
9587
            self.LogWarning("Could not remove older export for instance %s"
9588
                            " on node %s: %s", iname, node, msg)
9589

    
9590
  def Exec(self, feedback_fn):
9591
    """Export an instance to an image in the cluster.
9592

9593
    """
9594
    assert self.op.mode in constants.EXPORT_MODES
9595

    
9596
    instance = self.instance
9597
    src_node = instance.primary_node
9598

    
9599
    if self.op.shutdown:
9600
      # shutdown the instance, but not the disks
9601
      feedback_fn("Shutting down instance %s" % instance.name)
9602
      result = self.rpc.call_instance_shutdown(src_node, instance,
9603
                                               self.op.shutdown_timeout)
9604
      # TODO: Maybe ignore failures if ignore_remove_failures is set
9605
      result.Raise("Could not shutdown instance %s on"
9606
                   " node %s" % (instance.name, src_node))
9607

    
9608
    # set the disks ID correctly since call_instance_start needs the
9609
    # correct drbd minor to create the symlinks
9610
    for disk in instance.disks:
9611
      self.cfg.SetDiskID(disk, src_node)
9612

    
9613
    activate_disks = (not instance.admin_up)
9614

    
9615
    if activate_disks:
9616
      # Activate the instance disks if we'exporting a stopped instance
9617
      feedback_fn("Activating disks for %s" % instance.name)
9618
      _StartInstanceDisks(self, instance, None)
9619

    
9620
    try:
9621
      helper = masterd.instance.ExportInstanceHelper(self, feedback_fn,
9622
                                                     instance)
9623

    
9624
      helper.CreateSnapshots()
9625
      try:
9626
        if (self.op.shutdown and instance.admin_up and
9627
            not self.op.remove_instance):
9628
          assert not activate_disks
9629
          feedback_fn("Starting instance %s" % instance.name)
9630
          result = self.rpc.call_instance_start(src_node, instance, None, None)
9631
          msg = result.fail_msg
9632
          if msg:
9633
            feedback_fn("Failed to start instance: %s" % msg)
9634
            _ShutdownInstanceDisks(self, instance)
9635
            raise errors.OpExecError("Could not start instance: %s" % msg)
9636

    
9637
        if self.op.mode == constants.EXPORT_MODE_LOCAL:
9638
          (fin_resu, dresults) = helper.LocalExport(self.dst_node)
9639
        elif self.op.mode == constants.EXPORT_MODE_REMOTE:
9640
          connect_timeout = constants.RIE_CONNECT_TIMEOUT
9641
          timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
9642

    
9643
          (key_name, _, _) = self.x509_key_name
9644

    
9645
          dest_ca_pem = \
9646
            OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM,
9647
                                            self.dest_x509_ca)
9648

    
9649
          (fin_resu, dresults) = helper.RemoteExport(self.dest_disk_info,
9650
                                                     key_name, dest_ca_pem,
9651
                                                     timeouts)
9652
      finally:
9653
        helper.Cleanup()
9654

    
9655
      # Check for backwards compatibility
9656
      assert len(dresults) == len(instance.disks)
9657
      assert compat.all(isinstance(i, bool) for i in dresults), \
9658
             "Not all results are boolean: %r" % dresults
9659

    
9660
    finally:
9661
      if activate_disks:
9662
        feedback_fn("Deactivating disks for %s" % instance.name)
9663
        _ShutdownInstanceDisks(self, instance)
9664

    
9665
    if not (compat.all(dresults) and fin_resu):
9666
      failures = []
9667
      if not fin_resu:
9668
        failures.append("export finalization")
9669
      if not compat.all(dresults):
9670
        fdsk = utils.CommaJoin(idx for (idx, dsk) in enumerate(dresults)
9671
                               if not dsk)
9672
        failures.append("disk export: disk(s) %s" % fdsk)
9673

    
9674
      raise errors.OpExecError("Export failed, errors in %s" %
9675
                               utils.CommaJoin(failures))
9676

    
9677
    # At this point, the export was successful, we can cleanup/finish
9678

    
9679
    # Remove instance if requested
9680
    if self.op.remove_instance:
9681
      feedback_fn("Removing instance %s" % instance.name)
9682
      _RemoveInstance(self, feedback_fn, instance,
9683
                      self.op.ignore_remove_failures)
9684

    
9685
    if self.op.mode == constants.EXPORT_MODE_LOCAL:
9686
      self._CleanupExports(feedback_fn)
9687

    
9688
    return fin_resu, dresults
9689

    
9690

    
9691
class LURemoveExport(NoHooksLU):
9692
  """Remove exports related to the named instance.
9693

9694
  """
9695
  _OP_PARAMS = [
9696
    _PInstanceName,
9697
    ]
9698
  REQ_BGL = False
9699

    
9700
  def ExpandNames(self):
9701
    self.needed_locks = {}
9702
    # We need all nodes to be locked in order for RemoveExport to work, but we
9703
    # don't need to lock the instance itself, as nothing will happen to it (and
9704
    # we can remove exports also for a removed instance)
9705
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9706

    
9707
  def Exec(self, feedback_fn):
9708
    """Remove any export.
9709

9710
    """
9711
    instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
9712
    # If the instance was not found we'll try with the name that was passed in.
9713
    # This will only work if it was an FQDN, though.
9714
    fqdn_warn = False
9715
    if not instance_name:
9716
      fqdn_warn = True
9717
      instance_name = self.op.instance_name
9718

    
9719
    locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
9720
    exportlist = self.rpc.call_export_list(locked_nodes)
9721
    found = False
9722
    for node in exportlist:
9723
      msg = exportlist[node].fail_msg
9724
      if msg:
9725
        self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
9726
        continue
9727
      if instance_name in exportlist[node].payload:
9728
        found = True
9729
        result = self.rpc.call_export_remove(node, instance_name)
9730
        msg = result.fail_msg
9731
        if msg:
9732
          logging.error("Could not remove export for instance %s"
9733
                        " on node %s: %s", instance_name, node, msg)
9734

    
9735
    if fqdn_warn and not found:
9736
      feedback_fn("Export not found. If trying to remove an export belonging"
9737
                  " to a deleted instance please use its Fully Qualified"
9738
                  " Domain Name.")
9739

    
9740

    
9741
class TagsLU(NoHooksLU): # pylint: disable-msg=W0223
9742
  """Generic tags LU.
9743

9744
  This is an abstract class which is the parent of all the other tags LUs.
9745

9746
  """
9747

    
9748
  def ExpandNames(self):
9749
    self.needed_locks = {}
9750
    if self.op.kind == constants.TAG_NODE:
9751
      self.op.name = _ExpandNodeName(self.cfg, self.op.name)
9752
      self.needed_locks[locking.LEVEL_NODE] = self.op.name
9753
    elif self.op.kind == constants.TAG_INSTANCE:
9754
      self.op.name = _ExpandInstanceName(self.cfg, self.op.name)
9755
      self.needed_locks[locking.LEVEL_INSTANCE] = self.op.name
9756

    
9757
    # FIXME: Acquire BGL for cluster tag operations (as of this writing it's
9758
    # not possible to acquire the BGL based on opcode parameters)
9759

    
9760
  def CheckPrereq(self):
9761
    """Check prerequisites.
9762

9763
    """
9764
    if self.op.kind == constants.TAG_CLUSTER:
9765
      self.target = self.cfg.GetClusterInfo()
9766
    elif self.op.kind == constants.TAG_NODE:
9767
      self.target = self.cfg.GetNodeInfo(self.op.name)
9768
    elif self.op.kind == constants.TAG_INSTANCE:
9769
      self.target = self.cfg.GetInstanceInfo(self.op.name)
9770
    else:
9771
      raise errors.OpPrereqError("Wrong tag type requested (%s)" %
9772
                                 str(self.op.kind), errors.ECODE_INVAL)
9773

    
9774

    
9775
class LUGetTags(TagsLU):
9776
  """Returns the tags of a given object.
9777

9778
  """
9779
  _OP_PARAMS = [
9780
    ("kind", ht.NoDefault, ht.TElemOf(constants.VALID_TAG_TYPES)),
9781
    # Name is only meaningful for nodes and instances
9782
    ("name", ht.NoDefault, ht.TMaybeString),
9783
    ]
9784
  REQ_BGL = False
9785

    
9786
  def ExpandNames(self):
9787
    TagsLU.ExpandNames(self)
9788

    
9789
    # Share locks as this is only a read operation
9790
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
9791

    
9792
  def Exec(self, feedback_fn):
9793
    """Returns the tag list.
9794

9795
    """
9796
    return list(self.target.GetTags())
9797

    
9798

    
9799
class LUSearchTags(NoHooksLU):
9800
  """Searches the tags for a given pattern.
9801

9802
  """
9803
  _OP_PARAMS = [
9804
    ("pattern", ht.NoDefault, ht.TNonEmptyString),
9805
    ]
9806
  REQ_BGL = False
9807

    
9808
  def ExpandNames(self):
9809
    self.needed_locks = {}
9810

    
9811
  def CheckPrereq(self):
9812
    """Check prerequisites.
9813

9814
    This checks the pattern passed for validity by compiling it.
9815

9816
    """
9817
    try:
9818
      self.re = re.compile(self.op.pattern)
9819
    except re.error, err:
9820
      raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
9821
                                 (self.op.pattern, err), errors.ECODE_INVAL)
9822

    
9823
  def Exec(self, feedback_fn):
9824
    """Returns the tag list.
9825

9826
    """
9827
    cfg = self.cfg
9828
    tgts = [("/cluster", cfg.GetClusterInfo())]
9829
    ilist = cfg.GetAllInstancesInfo().values()
9830
    tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
9831
    nlist = cfg.GetAllNodesInfo().values()
9832
    tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
9833
    results = []
9834
    for path, target in tgts:
9835
      for tag in target.GetTags():
9836
        if self.re.search(tag):
9837
          results.append((path, tag))
9838
    return results
9839

    
9840

    
9841
class LUAddTags(TagsLU):
9842
  """Sets a tag on a given object.
9843

9844
  """
9845
  _OP_PARAMS = [
9846
    ("kind", ht.NoDefault, ht.TElemOf(constants.VALID_TAG_TYPES)),
9847
    # Name is only meaningful for nodes and instances
9848
    ("name", ht.NoDefault, ht.TMaybeString),
9849
    ("tags", ht.NoDefault, ht.TListOf(ht.TNonEmptyString)),
9850
    ]
9851
  REQ_BGL = False
9852

    
9853
  def CheckPrereq(self):
9854
    """Check prerequisites.
9855

9856
    This checks the type and length of the tag name and value.
9857

9858
    """
9859
    TagsLU.CheckPrereq(self)
9860
    for tag in self.op.tags:
9861
      objects.TaggableObject.ValidateTag(tag)
9862

    
9863
  def Exec(self, feedback_fn):
9864
    """Sets the tag.
9865

9866
    """
9867
    try:
9868
      for tag in self.op.tags:
9869
        self.target.AddTag(tag)
9870
    except errors.TagError, err:
9871
      raise errors.OpExecError("Error while setting tag: %s" % str(err))
9872
    self.cfg.Update(self.target, feedback_fn)
9873

    
9874

    
9875
class LUDelTags(TagsLU):
9876
  """Delete a list of tags from a given object.
9877

9878
  """
9879
  _OP_PARAMS = [
9880
    ("kind", ht.NoDefault, ht.TElemOf(constants.VALID_TAG_TYPES)),
9881
    # Name is only meaningful for nodes and instances
9882
    ("name", ht.NoDefault, ht.TMaybeString),
9883
    ("tags", ht.NoDefault, ht.TListOf(ht.TNonEmptyString)),
9884
    ]
9885
  REQ_BGL = False
9886

    
9887
  def CheckPrereq(self):
9888
    """Check prerequisites.
9889

9890
    This checks that we have the given tag.
9891

9892
    """
9893
    TagsLU.CheckPrereq(self)
9894
    for tag in self.op.tags:
9895
      objects.TaggableObject.ValidateTag(tag)
9896
    del_tags = frozenset(self.op.tags)
9897
    cur_tags = self.target.GetTags()
9898

    
9899
    diff_tags = del_tags - cur_tags
9900
    if diff_tags:
9901
      diff_names = ("'%s'" % i for i in sorted(diff_tags))
9902
      raise errors.OpPrereqError("Tag(s) %s not found" %
9903
                                 (utils.CommaJoin(diff_names), ),
9904
                                 errors.ECODE_NOENT)
9905

    
9906
  def Exec(self, feedback_fn):
9907
    """Remove the tag from the object.
9908

9909
    """
9910
    for tag in self.op.tags:
9911
      self.target.RemoveTag(tag)
9912
    self.cfg.Update(self.target, feedback_fn)
9913

    
9914

    
9915
class LUTestDelay(NoHooksLU):
9916
  """Sleep for a specified amount of time.
9917

9918
  This LU sleeps on the master and/or nodes for a specified amount of
9919
  time.
9920

9921
  """
9922
  _OP_PARAMS = [
9923
    ("duration", ht.NoDefault, ht.TFloat),
9924
    ("on_master", True, ht.TBool),
9925
    ("on_nodes", ht.EmptyList, ht.TListOf(ht.TNonEmptyString)),
9926
    ("repeat", 0, ht.TPositiveInt)
9927
    ]
9928
  REQ_BGL = False
9929

    
9930
  def ExpandNames(self):
9931
    """Expand names and set required locks.
9932

9933
    This expands the node list, if any.
9934

9935
    """
9936
    self.needed_locks = {}
9937
    if self.op.on_nodes:
9938
      # _GetWantedNodes can be used here, but is not always appropriate to use
9939
      # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
9940
      # more information.
9941
      self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
9942
      self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
9943

    
9944
  def _TestDelay(self):
9945
    """Do the actual sleep.
9946

9947
    """
9948
    if self.op.on_master:
9949
      if not utils.TestDelay(self.op.duration):
9950
        raise errors.OpExecError("Error during master delay test")
9951
    if self.op.on_nodes:
9952
      result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
9953
      for node, node_result in result.items():
9954
        node_result.Raise("Failure during rpc call to node %s" % node)
9955

    
9956
  def Exec(self, feedback_fn):
9957
    """Execute the test delay opcode, with the wanted repetitions.
9958

9959
    """
9960
    if self.op.repeat == 0:
9961
      self._TestDelay()
9962
    else:
9963
      top_value = self.op.repeat - 1
9964
      for i in range(self.op.repeat):
9965
        self.LogInfo("Test delay iteration %d/%d" % (i, top_value))
9966
        self._TestDelay()
9967

    
9968

    
9969
class LUTestJobqueue(NoHooksLU):
9970
  """Utility LU to test some aspects of the job queue.
9971

9972
  """
9973
  _OP_PARAMS = [
9974
    ("notify_waitlock", False, ht.TBool),
9975
    ("notify_exec", False, ht.TBool),
9976
    ("log_messages", ht.EmptyList, ht.TListOf(ht.TString)),
9977
    ("fail", False, ht.TBool),
9978
    ]
9979
  REQ_BGL = False
9980

    
9981
  # Must be lower than default timeout for WaitForJobChange to see whether it
9982
  # notices changed jobs
9983
  _CLIENT_CONNECT_TIMEOUT = 20.0
9984
  _CLIENT_CONFIRM_TIMEOUT = 60.0
9985

    
9986
  @classmethod
9987
  def _NotifyUsingSocket(cls, cb, errcls):
9988
    """Opens a Unix socket and waits for another program to connect.
9989

9990
    @type cb: callable
9991
    @param cb: Callback to send socket name to client
9992
    @type errcls: class
9993
    @param errcls: Exception class to use for errors
9994

9995
    """
9996
    # Using a temporary directory as there's no easy way to create temporary
9997
    # sockets without writing a custom loop around tempfile.mktemp and
9998
    # socket.bind
9999
    tmpdir = tempfile.mkdtemp()
10000
    try:
10001
      tmpsock = utils.PathJoin(tmpdir, "sock")
10002

    
10003
      logging.debug("Creating temporary socket at %s", tmpsock)
10004
      sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
10005
      try:
10006
        sock.bind(tmpsock)
10007
        sock.listen(1)
10008

    
10009
        # Send details to client
10010
        cb(tmpsock)
10011

    
10012
        # Wait for client to connect before continuing
10013
        sock.settimeout(cls._CLIENT_CONNECT_TIMEOUT)
10014
        try:
10015
          (conn, _) = sock.accept()
10016
        except socket.error, err:
10017
          raise errcls("Client didn't connect in time (%s)" % err)
10018
      finally:
10019
        sock.close()
10020
    finally:
10021
      # Remove as soon as client is connected
10022
      shutil.rmtree(tmpdir)
10023

    
10024
    # Wait for client to close
10025
    try:
10026
      try:
10027
        # pylint: disable-msg=E1101
10028
        # Instance of '_socketobject' has no ... member
10029
        conn.settimeout(cls._CLIENT_CONFIRM_TIMEOUT)
10030
        conn.recv(1)
10031
      except socket.error, err:
10032
        raise errcls("Client failed to confirm notification (%s)" % err)
10033
    finally:
10034
      conn.close()
10035

    
10036
  def _SendNotification(self, test, arg, sockname):
10037
    """Sends a notification to the client.
10038

10039
    @type test: string
10040
    @param test: Test name
10041
    @param arg: Test argument (depends on test)
10042
    @type sockname: string
10043
    @param sockname: Socket path
10044

10045
    """
10046
    self.Log(constants.ELOG_JQUEUE_TEST, (sockname, test, arg))
10047

    
10048
  def _Notify(self, prereq, test, arg):
10049
    """Notifies the client of a test.
10050

10051
    @type prereq: bool
10052
    @param prereq: Whether this is a prereq-phase test
10053
    @type test: string
10054
    @param test: Test name
10055
    @param arg: Test argument (depends on test)
10056

10057
    """
10058
    if prereq:
10059
      errcls = errors.OpPrereqError
10060
    else:
10061
      errcls = errors.OpExecError
10062

    
10063
    return self._NotifyUsingSocket(compat.partial(self._SendNotification,
10064
                                                  test, arg),
10065
                                   errcls)
10066

    
10067
  def CheckArguments(self):
10068
    self.checkargs_calls = getattr(self, "checkargs_calls", 0) + 1
10069
    self.expandnames_calls = 0
10070

    
10071
  def ExpandNames(self):
10072
    checkargs_calls = getattr(self, "checkargs_calls", 0)
10073
    if checkargs_calls < 1:
10074
      raise errors.ProgrammerError("CheckArguments was not called")
10075

    
10076
    self.expandnames_calls += 1
10077

    
10078
    if self.op.notify_waitlock:
10079
      self._Notify(True, constants.JQT_EXPANDNAMES, None)
10080

    
10081
    self.LogInfo("Expanding names")
10082

    
10083
    # Get lock on master node (just to get a lock, not for a particular reason)
10084
    self.needed_locks = {
10085
      locking.LEVEL_NODE: self.cfg.GetMasterNode(),
10086
      }
10087

    
10088
  def Exec(self, feedback_fn):
10089
    if self.expandnames_calls < 1:
10090
      raise errors.ProgrammerError("ExpandNames was not called")
10091

    
10092
    if self.op.notify_exec:
10093
      self._Notify(False, constants.JQT_EXEC, None)
10094

    
10095
    self.LogInfo("Executing")
10096

    
10097
    if self.op.log_messages:
10098
      self._Notify(False, constants.JQT_STARTMSG, len(self.op.log_messages))
10099
      for idx, msg in enumerate(self.op.log_messages):
10100
        self.LogInfo("Sending log message %s", idx + 1)
10101
        feedback_fn(constants.JQT_MSGPREFIX + msg)
10102
        # Report how many test messages have been sent
10103
        self._Notify(False, constants.JQT_LOGMSG, idx + 1)
10104

    
10105
    if self.op.fail:
10106
      raise errors.OpExecError("Opcode failure was requested")
10107

    
10108
    return True
10109

    
10110

    
10111
class IAllocator(object):
10112
  """IAllocator framework.
10113

10114
  An IAllocator instance has three sets of attributes:
10115
    - cfg that is needed to query the cluster
10116
    - input data (all members of the _KEYS class attribute are required)
10117
    - four buffer attributes (in|out_data|text), that represent the
10118
      input (to the external script) in text and data structure format,
10119
      and the output from it, again in two formats
10120
    - the result variables from the script (success, info, nodes) for
10121
      easy usage
10122

10123
  """
10124
  # pylint: disable-msg=R0902
10125
  # lots of instance attributes
10126
  _ALLO_KEYS = [
10127
    "name", "mem_size", "disks", "disk_template",
10128
    "os", "tags", "nics", "vcpus", "hypervisor",
10129
    ]
10130
  _RELO_KEYS = [
10131
    "name", "relocate_from",
10132
    ]
10133
  _EVAC_KEYS = [
10134
    "evac_nodes",
10135
    ]
10136

    
10137
  def __init__(self, cfg, rpc, mode, **kwargs):
10138
    self.cfg = cfg
10139
    self.rpc = rpc
10140
    # init buffer variables
10141
    self.in_text = self.out_text = self.in_data = self.out_data = None
10142
    # init all input fields so that pylint is happy
10143
    self.mode = mode
10144
    self.mem_size = self.disks = self.disk_template = None
10145
    self.os = self.tags = self.nics = self.vcpus = None
10146
    self.hypervisor = None
10147
    self.relocate_from = None
10148
    self.name = None
10149
    self.evac_nodes = None
10150
    # computed fields
10151
    self.required_nodes = None
10152
    # init result fields
10153
    self.success = self.info = self.result = None
10154
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
10155
      keyset = self._ALLO_KEYS
10156
      fn = self._AddNewInstance
10157
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
10158
      keyset = self._RELO_KEYS
10159
      fn = self._AddRelocateInstance
10160
    elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
10161
      keyset = self._EVAC_KEYS
10162
      fn = self._AddEvacuateNodes
10163
    else:
10164
      raise errors.ProgrammerError("Unknown mode '%s' passed to the"
10165
                                   " IAllocator" % self.mode)
10166
    for key in kwargs:
10167
      if key not in keyset:
10168
        raise errors.ProgrammerError("Invalid input parameter '%s' to"
10169
                                     " IAllocator" % key)
10170
      setattr(self, key, kwargs[key])
10171

    
10172
    for key in keyset:
10173
      if key not in kwargs:
10174
        raise errors.ProgrammerError("Missing input parameter '%s' to"
10175
                                     " IAllocator" % key)
10176
    self._BuildInputData(fn)
10177

    
10178
  def _ComputeClusterData(self):
10179
    """Compute the generic allocator input data.
10180

10181
    This is the data that is independent of the actual operation.
10182

10183
    """
10184
    cfg = self.cfg
10185
    cluster_info = cfg.GetClusterInfo()
10186
    # cluster data
10187
    data = {
10188
      "version": constants.IALLOCATOR_VERSION,
10189
      "cluster_name": cfg.GetClusterName(),
10190
      "cluster_tags": list(cluster_info.GetTags()),
10191
      "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
10192
      # we don't have job IDs
10193
      }
10194
    iinfo = cfg.GetAllInstancesInfo().values()
10195
    i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
10196

    
10197
    # node data
10198
    node_list = cfg.GetNodeList()
10199

    
10200
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
10201
      hypervisor_name = self.hypervisor
10202
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
10203
      hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
10204
    elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
10205
      hypervisor_name = cluster_info.enabled_hypervisors[0]
10206

    
10207
    node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(),
10208
                                        hypervisor_name)
10209
    node_iinfo = \
10210
      self.rpc.call_all_instances_info(node_list,
10211
                                       cluster_info.enabled_hypervisors)
10212

    
10213
    data["nodegroups"] = self._ComputeNodeGroupData(cfg)
10214

    
10215
    data["nodes"] = self._ComputeNodeData(cfg, node_data, node_iinfo, i_list)
10216

    
10217
    data["instances"] = self._ComputeInstanceData(cluster_info, i_list)
10218

    
10219
    self.in_data = data
10220

    
10221
  @staticmethod
10222
  def _ComputeNodeGroupData(cfg):
10223
    """Compute node groups data.
10224

10225
    """
10226
    ng = {}
10227
    for guuid, gdata in cfg.GetAllNodeGroupsInfo().items():
10228
      ng[guuid] = { "name": gdata.name }
10229
    return ng
10230

    
10231
  @staticmethod
10232
  def _ComputeNodeData(cfg, node_data, node_iinfo, i_list):
10233
    """Compute global node data.
10234

10235
    """
10236
    node_results = {}
10237
    for nname, nresult in node_data.items():
10238
      # first fill in static (config-based) values
10239
      ninfo = cfg.GetNodeInfo(nname)
10240
      pnr = {
10241
        "tags": list(ninfo.GetTags()),
10242
        "primary_ip": ninfo.primary_ip,
10243
        "secondary_ip": ninfo.secondary_ip,
10244
        "offline": ninfo.offline,
10245
        "drained": ninfo.drained,
10246
        "master_candidate": ninfo.master_candidate,
10247
        "group": ninfo.group,
10248
        "master_capable": ninfo.master_capable,
10249
        "vm_capable": ninfo.vm_capable,
10250
        }
10251

    
10252
      if not (ninfo.offline or ninfo.drained):
10253
        nresult.Raise("Can't get data for node %s" % nname)
10254
        node_iinfo[nname].Raise("Can't get node instance info from node %s" %
10255
                                nname)
10256
        remote_info = nresult.payload
10257

    
10258
        for attr in ['memory_total', 'memory_free', 'memory_dom0',
10259
                     'vg_size', 'vg_free', 'cpu_total']:
10260
          if attr not in remote_info:
10261
            raise errors.OpExecError("Node '%s' didn't return attribute"
10262
                                     " '%s'" % (nname, attr))
10263
          if not isinstance(remote_info[attr], int):
10264
            raise errors.OpExecError("Node '%s' returned invalid value"
10265
                                     " for '%s': %s" %
10266
                                     (nname, attr, remote_info[attr]))
10267
        # compute memory used by primary instances
10268
        i_p_mem = i_p_up_mem = 0
10269
        for iinfo, beinfo in i_list:
10270
          if iinfo.primary_node == nname:
10271
            i_p_mem += beinfo[constants.BE_MEMORY]
10272
            if iinfo.name not in node_iinfo[nname].payload:
10273
              i_used_mem = 0
10274
            else:
10275
              i_used_mem = int(node_iinfo[nname].payload[iinfo.name]['memory'])
10276
            i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
10277
            remote_info['memory_free'] -= max(0, i_mem_diff)
10278

    
10279
            if iinfo.admin_up:
10280
              i_p_up_mem += beinfo[constants.BE_MEMORY]
10281

    
10282
        # compute memory used by instances
10283
        pnr_dyn = {
10284
          "total_memory": remote_info['memory_total'],
10285
          "reserved_memory": remote_info['memory_dom0'],
10286
          "free_memory": remote_info['memory_free'],
10287
          "total_disk": remote_info['vg_size'],
10288
          "free_disk": remote_info['vg_free'],
10289
          "total_cpus": remote_info['cpu_total'],
10290
          "i_pri_memory": i_p_mem,
10291
          "i_pri_up_memory": i_p_up_mem,
10292
          }
10293
        pnr.update(pnr_dyn)
10294

    
10295
      node_results[nname] = pnr
10296

    
10297
    return node_results
10298

    
10299
  @staticmethod
10300
  def _ComputeInstanceData(cluster_info, i_list):
10301
    """Compute global instance data.
10302

10303
    """
10304
    instance_data = {}
10305
    for iinfo, beinfo in i_list:
10306
      nic_data = []
10307
      for nic in iinfo.nics:
10308
        filled_params = cluster_info.SimpleFillNIC(nic.nicparams)
10309
        nic_dict = {"mac": nic.mac,
10310
                    "ip": nic.ip,
10311
                    "mode": filled_params[constants.NIC_MODE],
10312
                    "link": filled_params[constants.NIC_LINK],
10313
                   }
10314
        if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
10315
          nic_dict["bridge"] = filled_params[constants.NIC_LINK]
10316
        nic_data.append(nic_dict)
10317
      pir = {
10318
        "tags": list(iinfo.GetTags()),
10319
        "admin_up": iinfo.admin_up,
10320
        "vcpus": beinfo[constants.BE_VCPUS],
10321
        "memory": beinfo[constants.BE_MEMORY],
10322
        "os": iinfo.os,
10323
        "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
10324
        "nics": nic_data,
10325
        "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
10326
        "disk_template": iinfo.disk_template,
10327
        "hypervisor": iinfo.hypervisor,
10328
        }
10329
      pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
10330
                                                 pir["disks"])
10331
      instance_data[iinfo.name] = pir
10332

    
10333
    return instance_data
10334

    
10335
  def _AddNewInstance(self):
10336
    """Add new instance data to allocator structure.
10337

10338
    This in combination with _AllocatorGetClusterData will create the
10339
    correct structure needed as input for the allocator.
10340

10341
    The checks for the completeness of the opcode must have already been
10342
    done.
10343

10344
    """
10345
    disk_space = _ComputeDiskSize(self.disk_template, self.disks)
10346

    
10347
    if self.disk_template in constants.DTS_NET_MIRROR:
10348
      self.required_nodes = 2
10349
    else:
10350
      self.required_nodes = 1
10351
    request = {
10352
      "name": self.name,
10353
      "disk_template": self.disk_template,
10354
      "tags": self.tags,
10355
      "os": self.os,
10356
      "vcpus": self.vcpus,
10357
      "memory": self.mem_size,
10358
      "disks": self.disks,
10359
      "disk_space_total": disk_space,
10360
      "nics": self.nics,
10361
      "required_nodes": self.required_nodes,
10362
      }
10363
    return request
10364

    
10365
  def _AddRelocateInstance(self):
10366
    """Add relocate instance data to allocator structure.
10367

10368
    This in combination with _IAllocatorGetClusterData will create the
10369
    correct structure needed as input for the allocator.
10370

10371
    The checks for the completeness of the opcode must have already been
10372
    done.
10373

10374
    """
10375
    instance = self.cfg.GetInstanceInfo(self.name)
10376
    if instance is None:
10377
      raise errors.ProgrammerError("Unknown instance '%s' passed to"
10378
                                   " IAllocator" % self.name)
10379

    
10380
    if instance.disk_template not in constants.DTS_NET_MIRROR:
10381
      raise errors.OpPrereqError("Can't relocate non-mirrored instances",
10382
                                 errors.ECODE_INVAL)
10383

    
10384
    if len(instance.secondary_nodes) != 1:
10385
      raise errors.OpPrereqError("Instance has not exactly one secondary node",
10386
                                 errors.ECODE_STATE)
10387

    
10388
    self.required_nodes = 1
10389
    disk_sizes = [{'size': disk.size} for disk in instance.disks]
10390
    disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
10391

    
10392
    request = {
10393
      "name": self.name,
10394
      "disk_space_total": disk_space,
10395
      "required_nodes": self.required_nodes,
10396
      "relocate_from": self.relocate_from,
10397
      }
10398
    return request
10399

    
10400
  def _AddEvacuateNodes(self):
10401
    """Add evacuate nodes data to allocator structure.
10402

10403
    """
10404
    request = {
10405
      "evac_nodes": self.evac_nodes
10406
      }
10407
    return request
10408

    
10409
  def _BuildInputData(self, fn):
10410
    """Build input data structures.
10411

10412
    """
10413
    self._ComputeClusterData()
10414

    
10415
    request = fn()
10416
    request["type"] = self.mode
10417
    self.in_data["request"] = request
10418

    
10419
    self.in_text = serializer.Dump(self.in_data)
10420

    
10421
  def Run(self, name, validate=True, call_fn=None):
10422
    """Run an instance allocator and return the results.
10423

10424
    """
10425
    if call_fn is None:
10426
      call_fn = self.rpc.call_iallocator_runner
10427

    
10428
    result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
10429
    result.Raise("Failure while running the iallocator script")
10430

    
10431
    self.out_text = result.payload
10432
    if validate:
10433
      self._ValidateResult()
10434

    
10435
  def _ValidateResult(self):
10436
    """Process the allocator results.
10437

10438
    This will process and if successful save the result in
10439
    self.out_data and the other parameters.
10440

10441
    """
10442
    try:
10443
      rdict = serializer.Load(self.out_text)
10444
    except Exception, err:
10445
      raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
10446

    
10447
    if not isinstance(rdict, dict):
10448
      raise errors.OpExecError("Can't parse iallocator results: not a dict")
10449

    
10450
    # TODO: remove backwards compatiblity in later versions
10451
    if "nodes" in rdict and "result" not in rdict:
10452
      rdict["result"] = rdict["nodes"]
10453
      del rdict["nodes"]
10454

    
10455
    for key in "success", "info", "result":
10456
      if key not in rdict:
10457
        raise errors.OpExecError("Can't parse iallocator results:"
10458
                                 " missing key '%s'" % key)
10459
      setattr(self, key, rdict[key])
10460

    
10461
    if not isinstance(rdict["result"], list):
10462
      raise errors.OpExecError("Can't parse iallocator results: 'result' key"
10463
                               " is not a list")
10464
    self.out_data = rdict
10465

    
10466

    
10467
class LUTestAllocator(NoHooksLU):
10468
  """Run allocator tests.
10469

10470
  This LU runs the allocator tests
10471

10472
  """
10473
  _OP_PARAMS = [
10474
    ("direction", ht.NoDefault,
10475
     ht.TElemOf(constants.VALID_IALLOCATOR_DIRECTIONS)),
10476
    ("mode", ht.NoDefault, ht.TElemOf(constants.VALID_IALLOCATOR_MODES)),
10477
    ("name", ht.NoDefault, ht.TNonEmptyString),
10478
    ("nics", ht.NoDefault, ht.TOr(ht.TNone, ht.TListOf(
10479
      ht.TDictOf(ht.TElemOf(["mac", "ip", "bridge"]),
10480
               ht.TOr(ht.TNone, ht.TNonEmptyString))))),
10481
    ("disks", ht.NoDefault, ht.TOr(ht.TNone, ht.TList)),
10482
    ("hypervisor", None, ht.TMaybeString),
10483
    ("allocator", None, ht.TMaybeString),
10484
    ("tags", ht.EmptyList, ht.TListOf(ht.TNonEmptyString)),
10485
    ("mem_size", None, ht.TOr(ht.TNone, ht.TPositiveInt)),
10486
    ("vcpus", None, ht.TOr(ht.TNone, ht.TPositiveInt)),
10487
    ("os", None, ht.TMaybeString),
10488
    ("disk_template", None, ht.TMaybeString),
10489
    ("evac_nodes", None, ht.TOr(ht.TNone, ht.TListOf(ht.TNonEmptyString))),
10490
    ]
10491

    
10492
  def CheckPrereq(self):
10493
    """Check prerequisites.
10494

10495
    This checks the opcode parameters depending on the director and mode test.
10496

10497
    """
10498
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
10499
      for attr in ["mem_size", "disks", "disk_template",
10500
                   "os", "tags", "nics", "vcpus"]:
10501
        if not hasattr(self.op, attr):
10502
          raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
10503
                                     attr, errors.ECODE_INVAL)
10504
      iname = self.cfg.ExpandInstanceName(self.op.name)
10505
      if iname is not None:
10506
        raise errors.OpPrereqError("Instance '%s' already in the cluster" %
10507
                                   iname, errors.ECODE_EXISTS)
10508
      if not isinstance(self.op.nics, list):
10509
        raise errors.OpPrereqError("Invalid parameter 'nics'",
10510
                                   errors.ECODE_INVAL)
10511
      if not isinstance(self.op.disks, list):
10512
        raise errors.OpPrereqError("Invalid parameter 'disks'",
10513
                                   errors.ECODE_INVAL)
10514
      for row in self.op.disks:
10515
        if (not isinstance(row, dict) or
10516
            "size" not in row or
10517
            not isinstance(row["size"], int) or
10518
            "mode" not in row or
10519
            row["mode"] not in ['r', 'w']):
10520
          raise errors.OpPrereqError("Invalid contents of the 'disks'"
10521
                                     " parameter", errors.ECODE_INVAL)
10522
      if self.op.hypervisor is None:
10523
        self.op.hypervisor = self.cfg.GetHypervisorType()
10524
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
10525
      fname = _ExpandInstanceName(self.cfg, self.op.name)
10526
      self.op.name = fname
10527
      self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
10528
    elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
10529
      if not hasattr(self.op, "evac_nodes"):
10530
        raise errors.OpPrereqError("Missing attribute 'evac_nodes' on"
10531
                                   " opcode input", errors.ECODE_INVAL)
10532
    else:
10533
      raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
10534
                                 self.op.mode, errors.ECODE_INVAL)
10535

    
10536
    if self.op.direction == constants.IALLOCATOR_DIR_OUT:
10537
      if self.op.allocator is None:
10538
        raise errors.OpPrereqError("Missing allocator name",
10539
                                   errors.ECODE_INVAL)
10540
    elif self.op.direction != constants.IALLOCATOR_DIR_IN:
10541
      raise errors.OpPrereqError("Wrong allocator test '%s'" %
10542
                                 self.op.direction, errors.ECODE_INVAL)
10543

    
10544
  def Exec(self, feedback_fn):
10545
    """Run the allocator test.
10546

10547
    """
10548
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
10549
      ial = IAllocator(self.cfg, self.rpc,
10550
                       mode=self.op.mode,
10551
                       name=self.op.name,
10552
                       mem_size=self.op.mem_size,
10553
                       disks=self.op.disks,
10554
                       disk_template=self.op.disk_template,
10555
                       os=self.op.os,
10556
                       tags=self.op.tags,
10557
                       nics=self.op.nics,
10558
                       vcpus=self.op.vcpus,
10559
                       hypervisor=self.op.hypervisor,
10560
                       )
10561
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
10562
      ial = IAllocator(self.cfg, self.rpc,
10563
                       mode=self.op.mode,
10564
                       name=self.op.name,
10565
                       relocate_from=list(self.relocate_from),
10566
                       )
10567
    elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
10568
      ial = IAllocator(self.cfg, self.rpc,
10569
                       mode=self.op.mode,
10570
                       evac_nodes=self.op.evac_nodes)
10571
    else:
10572
      raise errors.ProgrammerError("Uncatched mode %s in"
10573
                                   " LUTestAllocator.Exec", self.op.mode)
10574

    
10575
    if self.op.direction == constants.IALLOCATOR_DIR_IN:
10576
      result = ial.in_text
10577
    else:
10578
      ial.Run(self.op.allocator, validate=False)
10579
      result = ial.out_text
10580
    return result