Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ a167575e

History | View | Annotate | Download (395.9 kB)

1
#
2
#
3

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

    
21

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

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

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

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

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

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

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

    
63

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

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

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

    
76

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

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

90
  Note that all commands require root permissions.
91

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

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

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

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

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

    
133
    # Tasklets
134
    self.tasklets = None
135

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

    
139
    self.CheckArguments()
140

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

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

    
149
  ssh = property(fget=__GetSSH)
150

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

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

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

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

166
    """
167
    pass
168

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

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

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

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

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

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

194
    Examples::
195

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

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

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

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

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

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

233
    """
234

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

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

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

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

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

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

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

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

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

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

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

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

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

292
    """
293
    raise NotImplementedError
294

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
381
    del self.recalculate_locks[locking.LEVEL_NODE]
382

    
383

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

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

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

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

397
    This just raises an error.
398

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

    
402

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

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

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

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

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

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

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

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

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

435
    """
436
    pass
437

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

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

445
    """
446
    raise NotImplementedError
447

    
448

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

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

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

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

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

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

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

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

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

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

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

    
492
    # Return expanded names
493
    return self.wanted
494

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

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

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

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

507
    See L{LogicalUnit.ExpandNames}.
508

509
    """
510
    raise NotImplementedError()
511

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

515
    See L{LogicalUnit.DeclareLocks}.
516

517
    """
518
    raise NotImplementedError()
519

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

523
    @return: Query data object
524

525
    """
526
    raise NotImplementedError()
527

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

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

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

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

    
540

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

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

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

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

    
558

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

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

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

    
578

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

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

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

    
611

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

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

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

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

    
630

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

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

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

    
645

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

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

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

    
660

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

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

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

    
673

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

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

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

    
686

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

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

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

    
704

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

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

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

    
731

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

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

    
739

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

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

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

    
755

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

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

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

    
772

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

    
777

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

    
782

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

788
  This builds the hook environment from individual variables.
789

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

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

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

    
852
  env["INSTANCE_NIC_COUNT"] = nic_count
853

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

    
862
  env["INSTANCE_DISK_COUNT"] = disk_count
863

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

    
868
  return env
869

    
870

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

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

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

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

    
894

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

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

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

    
932

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

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

    
948

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

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

    
959

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

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

    
973

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

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

    
982

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

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

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

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

    
1002

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

    
1006

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

1010
  """
1011

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

    
1014

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

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

    
1022

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

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

    
1030

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

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

    
1040
  return []
1041

    
1042

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

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

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

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

    
1057
  return faulty
1058

    
1059

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

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

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

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

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

    
1091

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

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

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

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

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

1110
    """
1111
    return True
1112

    
1113

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

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

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

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

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

1131
    This checks whether the cluster is empty.
1132

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

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

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

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

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

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

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

    
1166
    return master
1167

    
1168

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

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

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

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

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

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

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

    
1201

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
1412
    return True
1413

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
1566
    for nname, success, bdev_status, idx in diskdata:
1567
      _ErrorIf(instanceconfig.admin_up and not success,
1568
               self.EINSTANCEFAULTYDISK, instance,
1569
               "couldn't retrieve status for disk/%s on %s: %s",
1570
               idx, nname, bdev_status)
1571
      _ErrorIf((instanceconfig.admin_up and success and
1572
                bdev_status.ldisk_status == constants.LDS_FAULTY),
1573
               self.EINSTANCEFAULTYDISK, instance,
1574
               "disk/%s on %s is faulty", idx, nname)
1575

    
1576
  def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):
1577
    """Verify if there are any unknown volumes in the cluster.
1578

1579
    The .os, .swap and backup volumes are ignored. All other volumes are
1580
    reported as unknown.
1581

1582
    @type reserved: L{ganeti.utils.FieldSet}
1583
    @param reserved: a FieldSet of reserved volume names
1584

1585
    """
1586
    for node, n_img in node_image.items():
1587
      if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1588
        # skip non-healthy nodes
1589
        continue
1590
      for volume in n_img.volumes:
1591
        test = ((node not in node_vol_should or
1592
                volume not in node_vol_should[node]) and
1593
                not reserved.Matches(volume))
1594
        self._ErrorIf(test, self.ENODEORPHANLV, node,
1595
                      "volume %s is unknown", volume)
1596

    
1597
  def _VerifyOrphanInstances(self, instancelist, node_image):
1598
    """Verify the list of running instances.
1599

1600
    This checks what instances are running but unknown to the cluster.
1601

1602
    """
1603
    for node, n_img in node_image.items():
1604
      for o_inst in n_img.instances:
1605
        test = o_inst not in instancelist
1606
        self._ErrorIf(test, self.ENODEORPHANINSTANCE, node,
1607
                      "instance %s on node %s should not exist", o_inst, node)
1608

    
1609
  def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
1610
    """Verify N+1 Memory Resilience.
1611

1612
    Check that if one single node dies we can still start all the
1613
    instances it was primary for.
1614

1615
    """
1616
    cluster_info = self.cfg.GetClusterInfo()
1617
    for node, n_img in node_image.items():
1618
      # This code checks that every node which is now listed as
1619
      # secondary has enough memory to host all instances it is
1620
      # supposed to should a single other node in the cluster fail.
1621
      # FIXME: not ready for failover to an arbitrary node
1622
      # FIXME: does not support file-backed instances
1623
      # WARNING: we currently take into account down instances as well
1624
      # as up ones, considering that even if they're down someone
1625
      # might want to start them even in the event of a node failure.
1626
      for prinode, instances in n_img.sbp.items():
1627
        needed_mem = 0
1628
        for instance in instances:
1629
          bep = cluster_info.FillBE(instance_cfg[instance])
1630
          if bep[constants.BE_AUTO_BALANCE]:
1631
            needed_mem += bep[constants.BE_MEMORY]
1632
        test = n_img.mfree < needed_mem
1633
        self._ErrorIf(test, self.ENODEN1, node,
1634
                      "not enough memory to accomodate instance failovers"
1635
                      " should node %s fail", prinode)
1636

    
1637
  def _VerifyNodeFiles(self, ninfo, nresult, file_list, local_cksum,
1638
                       master_files):
1639
    """Verifies and computes the node required file checksums.
1640

1641
    @type ninfo: L{objects.Node}
1642
    @param ninfo: the node to check
1643
    @param nresult: the remote results for the node
1644
    @param file_list: required list of files
1645
    @param local_cksum: dictionary of local files and their checksums
1646
    @param master_files: list of files that only masters should have
1647

1648
    """
1649
    node = ninfo.name
1650
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1651

    
1652
    remote_cksum = nresult.get(constants.NV_FILELIST, None)
1653
    test = not isinstance(remote_cksum, dict)
1654
    _ErrorIf(test, self.ENODEFILECHECK, node,
1655
             "node hasn't returned file checksum data")
1656
    if test:
1657
      return
1658

    
1659
    for file_name in file_list:
1660
      node_is_mc = ninfo.master_candidate
1661
      must_have = (file_name not in master_files) or node_is_mc
1662
      # missing
1663
      test1 = file_name not in remote_cksum
1664
      # invalid checksum
1665
      test2 = not test1 and remote_cksum[file_name] != local_cksum[file_name]
1666
      # existing and good
1667
      test3 = not test1 and remote_cksum[file_name] == local_cksum[file_name]
1668
      _ErrorIf(test1 and must_have, self.ENODEFILECHECK, node,
1669
               "file '%s' missing", file_name)
1670
      _ErrorIf(test2 and must_have, self.ENODEFILECHECK, node,
1671
               "file '%s' has wrong checksum", file_name)
1672
      # not candidate and this is not a must-have file
1673
      _ErrorIf(test2 and not must_have, self.ENODEFILECHECK, node,
1674
               "file '%s' should not exist on non master"
1675
               " candidates (and the file is outdated)", file_name)
1676
      # all good, except non-master/non-must have combination
1677
      _ErrorIf(test3 and not must_have, self.ENODEFILECHECK, node,
1678
               "file '%s' should not exist"
1679
               " on non master candidates", file_name)
1680

    
1681
  def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper,
1682
                      drbd_map):
1683
    """Verifies and the node DRBD status.
1684

1685
    @type ninfo: L{objects.Node}
1686
    @param ninfo: the node to check
1687
    @param nresult: the remote results for the node
1688
    @param instanceinfo: the dict of instances
1689
    @param drbd_helper: the configured DRBD usermode helper
1690
    @param drbd_map: the DRBD map as returned by
1691
        L{ganeti.config.ConfigWriter.ComputeDRBDMap}
1692

1693
    """
1694
    node = ninfo.name
1695
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1696

    
1697
    if drbd_helper:
1698
      helper_result = nresult.get(constants.NV_DRBDHELPER, None)
1699
      test = (helper_result == None)
1700
      _ErrorIf(test, self.ENODEDRBDHELPER, node,
1701
               "no drbd usermode helper returned")
1702
      if helper_result:
1703
        status, payload = helper_result
1704
        test = not status
1705
        _ErrorIf(test, self.ENODEDRBDHELPER, node,
1706
                 "drbd usermode helper check unsuccessful: %s", payload)
1707
        test = status and (payload != drbd_helper)
1708
        _ErrorIf(test, self.ENODEDRBDHELPER, node,
1709
                 "wrong drbd usermode helper: %s", payload)
1710

    
1711
    # compute the DRBD minors
1712
    node_drbd = {}
1713
    for minor, instance in drbd_map[node].items():
1714
      test = instance not in instanceinfo
1715
      _ErrorIf(test, self.ECLUSTERCFG, None,
1716
               "ghost instance '%s' in temporary DRBD map", instance)
1717
        # ghost instance should not be running, but otherwise we
1718
        # don't give double warnings (both ghost instance and
1719
        # unallocated minor in use)
1720
      if test:
1721
        node_drbd[minor] = (instance, False)
1722
      else:
1723
        instance = instanceinfo[instance]
1724
        node_drbd[minor] = (instance.name, instance.admin_up)
1725

    
1726
    # and now check them
1727
    used_minors = nresult.get(constants.NV_DRBDLIST, [])
1728
    test = not isinstance(used_minors, (tuple, list))
1729
    _ErrorIf(test, self.ENODEDRBD, node,
1730
             "cannot parse drbd status file: %s", str(used_minors))
1731
    if test:
1732
      # we cannot check drbd status
1733
      return
1734

    
1735
    for minor, (iname, must_exist) in node_drbd.items():
1736
      test = minor not in used_minors and must_exist
1737
      _ErrorIf(test, self.ENODEDRBD, node,
1738
               "drbd minor %d of instance %s is not active", minor, iname)
1739
    for minor in used_minors:
1740
      test = minor not in node_drbd
1741
      _ErrorIf(test, self.ENODEDRBD, node,
1742
               "unallocated drbd minor %d is in use", minor)
1743

    
1744
  def _UpdateNodeOS(self, ninfo, nresult, nimg):
1745
    """Builds the node OS structures.
1746

1747
    @type ninfo: L{objects.Node}
1748
    @param ninfo: the node to check
1749
    @param nresult: the remote results for the node
1750
    @param nimg: the node image object
1751

1752
    """
1753
    node = ninfo.name
1754
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1755

    
1756
    remote_os = nresult.get(constants.NV_OSLIST, None)
1757
    test = (not isinstance(remote_os, list) or
1758
            not compat.all(isinstance(v, list) and len(v) == 7
1759
                           for v in remote_os))
1760

    
1761
    _ErrorIf(test, self.ENODEOS, node,
1762
             "node hasn't returned valid OS data")
1763

    
1764
    nimg.os_fail = test
1765

    
1766
    if test:
1767
      return
1768

    
1769
    os_dict = {}
1770

    
1771
    for (name, os_path, status, diagnose,
1772
         variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
1773

    
1774
      if name not in os_dict:
1775
        os_dict[name] = []
1776

    
1777
      # parameters is a list of lists instead of list of tuples due to
1778
      # JSON lacking a real tuple type, fix it:
1779
      parameters = [tuple(v) for v in parameters]
1780
      os_dict[name].append((os_path, status, diagnose,
1781
                            set(variants), set(parameters), set(api_ver)))
1782

    
1783
    nimg.oslist = os_dict
1784

    
1785
  def _VerifyNodeOS(self, ninfo, nimg, base):
1786
    """Verifies the node OS list.
1787

1788
    @type ninfo: L{objects.Node}
1789
    @param ninfo: the node to check
1790
    @param nimg: the node image object
1791
    @param base: the 'template' node we match against (e.g. from the master)
1792

1793
    """
1794
    node = ninfo.name
1795
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1796

    
1797
    assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
1798

    
1799
    for os_name, os_data in nimg.oslist.items():
1800
      assert os_data, "Empty OS status for OS %s?!" % os_name
1801
      f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
1802
      _ErrorIf(not f_status, self.ENODEOS, node,
1803
               "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
1804
      _ErrorIf(len(os_data) > 1, self.ENODEOS, node,
1805
               "OS '%s' has multiple entries (first one shadows the rest): %s",
1806
               os_name, utils.CommaJoin([v[0] for v in os_data]))
1807
      # this will catched in backend too
1808
      _ErrorIf(compat.any(v >= constants.OS_API_V15 for v in f_api)
1809
               and not f_var, self.ENODEOS, node,
1810
               "OS %s with API at least %d does not declare any variant",
1811
               os_name, constants.OS_API_V15)
1812
      # comparisons with the 'base' image
1813
      test = os_name not in base.oslist
1814
      _ErrorIf(test, self.ENODEOS, node,
1815
               "Extra OS %s not present on reference node (%s)",
1816
               os_name, base.name)
1817
      if test:
1818
        continue
1819
      assert base.oslist[os_name], "Base node has empty OS status?"
1820
      _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
1821
      if not b_status:
1822
        # base OS is invalid, skipping
1823
        continue
1824
      for kind, a, b in [("API version", f_api, b_api),
1825
                         ("variants list", f_var, b_var),
1826
                         ("parameters", f_param, b_param)]:
1827
        _ErrorIf(a != b, self.ENODEOS, node,
1828
                 "OS %s %s differs from reference node %s: %s vs. %s",
1829
                 kind, os_name, base.name,
1830
                 utils.CommaJoin(a), utils.CommaJoin(b))
1831

    
1832
    # check any missing OSes
1833
    missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
1834
    _ErrorIf(missing, self.ENODEOS, node,
1835
             "OSes present on reference node %s but missing on this node: %s",
1836
             base.name, utils.CommaJoin(missing))
1837

    
1838
  def _VerifyOob(self, ninfo, nresult):
1839
    """Verifies out of band functionality of a node.
1840

1841
    @type ninfo: L{objects.Node}
1842
    @param ninfo: the node to check
1843
    @param nresult: the remote results for the node
1844

1845
    """
1846
    node = ninfo.name
1847
    # We just have to verify the paths on master and/or master candidates
1848
    # as the oob helper is invoked on the master
1849
    if ((ninfo.master_candidate or ninfo.master_capable) and
1850
        constants.NV_OOB_PATHS in nresult):
1851
      for path_result in nresult[constants.NV_OOB_PATHS]:
1852
        self._ErrorIf(path_result, self.ENODEOOBPATH, node, path_result)
1853

    
1854
  def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
1855
    """Verifies and updates the node volume data.
1856

1857
    This function will update a L{NodeImage}'s internal structures
1858
    with data from the remote call.
1859

1860
    @type ninfo: L{objects.Node}
1861
    @param ninfo: the node to check
1862
    @param nresult: the remote results for the node
1863
    @param nimg: the node image object
1864
    @param vg_name: the configured VG name
1865

1866
    """
1867
    node = ninfo.name
1868
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1869

    
1870
    nimg.lvm_fail = True
1871
    lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
1872
    if vg_name is None:
1873
      pass
1874
    elif isinstance(lvdata, basestring):
1875
      _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
1876
               utils.SafeEncode(lvdata))
1877
    elif not isinstance(lvdata, dict):
1878
      _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
1879
    else:
1880
      nimg.volumes = lvdata
1881
      nimg.lvm_fail = False
1882

    
1883
  def _UpdateNodeInstances(self, ninfo, nresult, nimg):
1884
    """Verifies and updates the node instance list.
1885

1886
    If the listing was successful, then updates this node's instance
1887
    list. Otherwise, it marks the RPC call as failed for the instance
1888
    list key.
1889

1890
    @type ninfo: L{objects.Node}
1891
    @param ninfo: the node to check
1892
    @param nresult: the remote results for the node
1893
    @param nimg: the node image object
1894

1895
    """
1896
    idata = nresult.get(constants.NV_INSTANCELIST, None)
1897
    test = not isinstance(idata, list)
1898
    self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed"
1899
                  " (instancelist): %s", utils.SafeEncode(str(idata)))
1900
    if test:
1901
      nimg.hyp_fail = True
1902
    else:
1903
      nimg.instances = idata
1904

    
1905
  def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
1906
    """Verifies and computes a node information map
1907

1908
    @type ninfo: L{objects.Node}
1909
    @param ninfo: the node to check
1910
    @param nresult: the remote results for the node
1911
    @param nimg: the node image object
1912
    @param vg_name: the configured VG name
1913

1914
    """
1915
    node = ninfo.name
1916
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1917

    
1918
    # try to read free memory (from the hypervisor)
1919
    hv_info = nresult.get(constants.NV_HVINFO, None)
1920
    test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
1921
    _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
1922
    if not test:
1923
      try:
1924
        nimg.mfree = int(hv_info["memory_free"])
1925
      except (ValueError, TypeError):
1926
        _ErrorIf(True, self.ENODERPC, node,
1927
                 "node returned invalid nodeinfo, check hypervisor")
1928

    
1929
    # FIXME: devise a free space model for file based instances as well
1930
    if vg_name is not None:
1931
      test = (constants.NV_VGLIST not in nresult or
1932
              vg_name not in nresult[constants.NV_VGLIST])
1933
      _ErrorIf(test, self.ENODELVM, node,
1934
               "node didn't return data for the volume group '%s'"
1935
               " - it is either missing or broken", vg_name)
1936
      if not test:
1937
        try:
1938
          nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
1939
        except (ValueError, TypeError):
1940
          _ErrorIf(True, self.ENODERPC, node,
1941
                   "node returned invalid LVM info, check LVM status")
1942

    
1943
  def _CollectDiskInfo(self, nodelist, node_image, instanceinfo):
1944
    """Gets per-disk status information for all instances.
1945

1946
    @type nodelist: list of strings
1947
    @param nodelist: Node names
1948
    @type node_image: dict of (name, L{objects.Node})
1949
    @param node_image: Node objects
1950
    @type instanceinfo: dict of (name, L{objects.Instance})
1951
    @param instanceinfo: Instance objects
1952
    @rtype: {instance: {node: [(succes, payload)]}}
1953
    @return: a dictionary of per-instance dictionaries with nodes as
1954
        keys and disk information as values; the disk information is a
1955
        list of tuples (success, payload)
1956

1957
    """
1958
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1959

    
1960
    node_disks = {}
1961
    node_disks_devonly = {}
1962
    diskless_instances = set()
1963
    diskless = constants.DT_DISKLESS
1964

    
1965
    for nname in nodelist:
1966
      node_instances = list(itertools.chain(node_image[nname].pinst,
1967
                                            node_image[nname].sinst))
1968
      diskless_instances.update(inst for inst in node_instances
1969
                                if instanceinfo[inst].disk_template == diskless)
1970
      disks = [(inst, disk)
1971
               for inst in node_instances
1972
               for disk in instanceinfo[inst].disks]
1973

    
1974
      if not disks:
1975
        # No need to collect data
1976
        continue
1977

    
1978
      node_disks[nname] = disks
1979

    
1980
      # Creating copies as SetDiskID below will modify the objects and that can
1981
      # lead to incorrect data returned from nodes
1982
      devonly = [dev.Copy() for (_, dev) in disks]
1983

    
1984
      for dev in devonly:
1985
        self.cfg.SetDiskID(dev, nname)
1986

    
1987
      node_disks_devonly[nname] = devonly
1988

    
1989
    assert len(node_disks) == len(node_disks_devonly)
1990

    
1991
    # Collect data from all nodes with disks
1992
    result = self.rpc.call_blockdev_getmirrorstatus_multi(node_disks.keys(),
1993
                                                          node_disks_devonly)
1994

    
1995
    assert len(result) == len(node_disks)
1996

    
1997
    instdisk = {}
1998

    
1999
    for (nname, nres) in result.items():
2000
      disks = node_disks[nname]
2001

    
2002
      if nres.offline:
2003
        # No data from this node
2004
        data = len(disks) * [(False, "node offline")]
2005
      else:
2006
        msg = nres.fail_msg
2007
        _ErrorIf(msg, self.ENODERPC, nname,
2008
                 "while getting disk information: %s", msg)
2009
        if msg:
2010
          # No data from this node
2011
          data = len(disks) * [(False, msg)]
2012
        else:
2013
          data = []
2014
          for idx, i in enumerate(nres.payload):
2015
            if isinstance(i, (tuple, list)) and len(i) == 2:
2016
              data.append(i)
2017
            else:
2018
              logging.warning("Invalid result from node %s, entry %d: %s",
2019
                              nname, idx, i)
2020
              data.append((False, "Invalid result from the remote node"))
2021

    
2022
      for ((inst, _), status) in zip(disks, data):
2023
        instdisk.setdefault(inst, {}).setdefault(nname, []).append(status)
2024

    
2025
    # Add empty entries for diskless instances.
2026
    for inst in diskless_instances:
2027
      assert inst not in instdisk
2028
      instdisk[inst] = {}
2029

    
2030
    assert compat.all(len(statuses) == len(instanceinfo[inst].disks) and
2031
                      len(nnames) <= len(instanceinfo[inst].all_nodes) and
2032
                      compat.all(isinstance(s, (tuple, list)) and
2033
                                 len(s) == 2 for s in statuses)
2034
                      for inst, nnames in instdisk.items()
2035
                      for nname, statuses in nnames.items())
2036
    assert set(instdisk) == set(instanceinfo), "instdisk consistency failure"
2037

    
2038
    return instdisk
2039

    
2040
  def _VerifyHVP(self, hvp_data):
2041
    """Verifies locally the syntax of the hypervisor parameters.
2042

2043
    """
2044
    for item, hv_name, hv_params in hvp_data:
2045
      msg = ("hypervisor %s parameters syntax check (source %s): %%s" %
2046
             (item, hv_name))
2047
      try:
2048
        hv_class = hypervisor.GetHypervisor(hv_name)
2049
        utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2050
        hv_class.CheckParameterSyntax(hv_params)
2051
      except errors.GenericError, err:
2052
        self._ErrorIf(True, self.ECLUSTERCFG, None, msg % str(err))
2053

    
2054

    
2055
  def BuildHooksEnv(self):
2056
    """Build hooks env.
2057

2058
    Cluster-Verify hooks just ran in the post phase and their failure makes
2059
    the output be logged in the verify output and the verification to fail.
2060

2061
    """
2062
    all_nodes = self.cfg.GetNodeList()
2063
    env = {
2064
      "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
2065
      }
2066
    for node in self.cfg.GetAllNodesInfo().values():
2067
      env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
2068

    
2069
    return env, [], all_nodes
2070

    
2071
  def Exec(self, feedback_fn):
2072
    """Verify integrity of cluster, performing various test on nodes.
2073

2074
    """
2075
    # This method has too many local variables. pylint: disable-msg=R0914
2076
    self.bad = False
2077
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
2078
    verbose = self.op.verbose
2079
    self._feedback_fn = feedback_fn
2080
    feedback_fn("* Verifying global settings")
2081
    for msg in self.cfg.VerifyConfig():
2082
      _ErrorIf(True, self.ECLUSTERCFG, None, msg)
2083

    
2084
    # Check the cluster certificates
2085
    for cert_filename in constants.ALL_CERT_FILES:
2086
      (errcode, msg) = _VerifyCertificate(cert_filename)
2087
      _ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
2088

    
2089
    vg_name = self.cfg.GetVGName()
2090
    drbd_helper = self.cfg.GetDRBDHelper()
2091
    hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
2092
    cluster = self.cfg.GetClusterInfo()
2093
    nodelist = utils.NiceSort(self.cfg.GetNodeList())
2094
    nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
2095
    nodeinfo_byname = dict(zip(nodelist, nodeinfo))
2096
    instancelist = utils.NiceSort(self.cfg.GetInstanceList())
2097
    instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
2098
                        for iname in instancelist)
2099
    groupinfo = self.cfg.GetAllNodeGroupsInfo()
2100
    i_non_redundant = [] # Non redundant instances
2101
    i_non_a_balanced = [] # Non auto-balanced instances
2102
    n_offline = 0 # Count of offline nodes
2103
    n_drained = 0 # Count of nodes being drained
2104
    node_vol_should = {}
2105

    
2106
    # FIXME: verify OS list
2107
    # do local checksums
2108
    master_files = [constants.CLUSTER_CONF_FILE]
2109
    master_node = self.master_node = self.cfg.GetMasterNode()
2110
    master_ip = self.cfg.GetMasterIP()
2111

    
2112
    file_names = ssconf.SimpleStore().GetFileList()
2113
    file_names.extend(constants.ALL_CERT_FILES)
2114
    file_names.extend(master_files)
2115
    if cluster.modify_etc_hosts:
2116
      file_names.append(constants.ETC_HOSTS)
2117

    
2118
    local_checksums = utils.FingerprintFiles(file_names)
2119

    
2120
    # Compute the set of hypervisor parameters
2121
    hvp_data = []
2122
    for hv_name in hypervisors:
2123
      hvp_data.append(("cluster", hv_name, cluster.GetHVDefaults(hv_name)))
2124
    for os_name, os_hvp in cluster.os_hvp.items():
2125
      for hv_name, hv_params in os_hvp.items():
2126
        if not hv_params:
2127
          continue
2128
        full_params = cluster.GetHVDefaults(hv_name, os_name=os_name)
2129
        hvp_data.append(("os %s" % os_name, hv_name, full_params))
2130
    # TODO: collapse identical parameter values in a single one
2131
    for instance in instanceinfo.values():
2132
      if not instance.hvparams:
2133
        continue
2134
      hvp_data.append(("instance %s" % instance.name, instance.hypervisor,
2135
                       cluster.FillHV(instance)))
2136
    # and verify them locally
2137
    self._VerifyHVP(hvp_data)
2138

    
2139
    feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
2140
    node_verify_param = {
2141
      constants.NV_FILELIST: file_names,
2142
      constants.NV_NODELIST: [node.name for node in nodeinfo
2143
                              if not node.offline],
2144
      constants.NV_HYPERVISOR: hypervisors,
2145
      constants.NV_HVPARAMS: hvp_data,
2146
      constants.NV_NODENETTEST: [(node.name, node.primary_ip,
2147
                                  node.secondary_ip) for node in nodeinfo
2148
                                 if not node.offline],
2149
      constants.NV_INSTANCELIST: hypervisors,
2150
      constants.NV_VERSION: None,
2151
      constants.NV_HVINFO: self.cfg.GetHypervisorType(),
2152
      constants.NV_NODESETUP: None,
2153
      constants.NV_TIME: None,
2154
      constants.NV_MASTERIP: (master_node, master_ip),
2155
      constants.NV_OSLIST: None,
2156
      constants.NV_VMNODES: self.cfg.GetNonVmCapableNodeList(),
2157
      }
2158

    
2159
    if vg_name is not None:
2160
      node_verify_param[constants.NV_VGLIST] = None
2161
      node_verify_param[constants.NV_LVLIST] = vg_name
2162
      node_verify_param[constants.NV_PVLIST] = [vg_name]
2163
      node_verify_param[constants.NV_DRBDLIST] = None
2164

    
2165
    if drbd_helper:
2166
      node_verify_param[constants.NV_DRBDHELPER] = drbd_helper
2167

    
2168
    # Build our expected cluster state
2169
    node_image = dict((node.name, self.NodeImage(offline=node.offline,
2170
                                                 name=node.name,
2171
                                                 vm_capable=node.vm_capable))
2172
                      for node in nodeinfo)
2173

    
2174
    # Gather OOB paths
2175
    oob_paths = []
2176
    for node in nodeinfo:
2177
      path = _SupportsOob(self.cfg, node)
2178
      if path and path not in oob_paths:
2179
        oob_paths.append(path)
2180

    
2181
    if oob_paths:
2182
      node_verify_param[constants.NV_OOB_PATHS] = oob_paths
2183

    
2184
    for instance in instancelist:
2185
      inst_config = instanceinfo[instance]
2186

    
2187
      for nname in inst_config.all_nodes:
2188
        if nname not in node_image:
2189
          # ghost node
2190
          gnode = self.NodeImage(name=nname)
2191
          gnode.ghost = True
2192
          node_image[nname] = gnode
2193

    
2194
      inst_config.MapLVsByNode(node_vol_should)
2195

    
2196
      pnode = inst_config.primary_node
2197
      node_image[pnode].pinst.append(instance)
2198

    
2199
      for snode in inst_config.secondary_nodes:
2200
        nimg = node_image[snode]
2201
        nimg.sinst.append(instance)
2202
        if pnode not in nimg.sbp:
2203
          nimg.sbp[pnode] = []
2204
        nimg.sbp[pnode].append(instance)
2205

    
2206
    # At this point, we have the in-memory data structures complete,
2207
    # except for the runtime information, which we'll gather next
2208

    
2209
    # Due to the way our RPC system works, exact response times cannot be
2210
    # guaranteed (e.g. a broken node could run into a timeout). By keeping the
2211
    # time before and after executing the request, we can at least have a time
2212
    # window.
2213
    nvinfo_starttime = time.time()
2214
    all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
2215
                                           self.cfg.GetClusterName())
2216
    nvinfo_endtime = time.time()
2217

    
2218
    all_drbd_map = self.cfg.ComputeDRBDMap()
2219

    
2220
    feedback_fn("* Gathering disk information (%s nodes)" % len(nodelist))
2221
    instdisk = self._CollectDiskInfo(nodelist, node_image, instanceinfo)
2222

    
2223
    feedback_fn("* Verifying node status")
2224

    
2225
    refos_img = None
2226

    
2227
    for node_i in nodeinfo:
2228
      node = node_i.name
2229
      nimg = node_image[node]
2230

    
2231
      if node_i.offline:
2232
        if verbose:
2233
          feedback_fn("* Skipping offline node %s" % (node,))
2234
        n_offline += 1
2235
        continue
2236

    
2237
      if node == master_node:
2238
        ntype = "master"
2239
      elif node_i.master_candidate:
2240
        ntype = "master candidate"
2241
      elif node_i.drained:
2242
        ntype = "drained"
2243
        n_drained += 1
2244
      else:
2245
        ntype = "regular"
2246
      if verbose:
2247
        feedback_fn("* Verifying node %s (%s)" % (node, ntype))
2248

    
2249
      msg = all_nvinfo[node].fail_msg
2250
      _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
2251
      if msg:
2252
        nimg.rpc_fail = True
2253
        continue
2254

    
2255
      nresult = all_nvinfo[node].payload
2256

    
2257
      nimg.call_ok = self._VerifyNode(node_i, nresult)
2258
      self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
2259
      self._VerifyNodeNetwork(node_i, nresult)
2260
      self._VerifyNodeFiles(node_i, nresult, file_names, local_checksums,
2261
                            master_files)
2262

    
2263
      self._VerifyOob(node_i, nresult)
2264

    
2265
      if nimg.vm_capable:
2266
        self._VerifyNodeLVM(node_i, nresult, vg_name)
2267
        self._VerifyNodeDrbd(node_i, nresult, instanceinfo, drbd_helper,
2268
                             all_drbd_map)
2269

    
2270
        self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
2271
        self._UpdateNodeInstances(node_i, nresult, nimg)
2272
        self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
2273
        self._UpdateNodeOS(node_i, nresult, nimg)
2274
        if not nimg.os_fail:
2275
          if refos_img is None:
2276
            refos_img = nimg
2277
          self._VerifyNodeOS(node_i, nimg, refos_img)
2278

    
2279
    feedback_fn("* Verifying instance status")
2280
    for instance in instancelist:
2281
      if verbose:
2282
        feedback_fn("* Verifying instance %s" % instance)
2283
      inst_config = instanceinfo[instance]
2284
      self._VerifyInstance(instance, inst_config, node_image,
2285
                           instdisk[instance])
2286
      inst_nodes_offline = []
2287

    
2288
      pnode = inst_config.primary_node
2289
      pnode_img = node_image[pnode]
2290
      _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
2291
               self.ENODERPC, pnode, "instance %s, connection to"
2292
               " primary node failed", instance)
2293

    
2294
      if pnode_img.offline:
2295
        inst_nodes_offline.append(pnode)
2296

    
2297
      # If the instance is non-redundant we cannot survive losing its primary
2298
      # node, so we are not N+1 compliant. On the other hand we have no disk
2299
      # templates with more than one secondary so that situation is not well
2300
      # supported either.
2301
      # FIXME: does not support file-backed instances
2302
      if not inst_config.secondary_nodes:
2303
        i_non_redundant.append(instance)
2304

    
2305
      _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT,
2306
               instance, "instance has multiple secondary nodes: %s",
2307
               utils.CommaJoin(inst_config.secondary_nodes),
2308
               code=self.ETYPE_WARNING)
2309

    
2310
      if inst_config.disk_template in constants.DTS_NET_MIRROR:
2311
        pnode = inst_config.primary_node
2312
        instance_nodes = utils.NiceSort(inst_config.all_nodes)
2313
        instance_groups = {}
2314

    
2315
        for node in instance_nodes:
2316
          instance_groups.setdefault(nodeinfo_byname[node].group,
2317
                                     []).append(node)
2318

    
2319
        pretty_list = [
2320
          "%s (group %s)" % (utils.CommaJoin(nodes), groupinfo[group].name)
2321
          # Sort so that we always list the primary node first.
2322
          for group, nodes in sorted(instance_groups.items(),
2323
                                     key=lambda (_, nodes): pnode in nodes,
2324
                                     reverse=True)]
2325

    
2326
        self._ErrorIf(len(instance_groups) > 1, self.EINSTANCESPLITGROUPS,
2327
                      instance, "instance has primary and secondary nodes in"
2328
                      " different groups: %s", utils.CommaJoin(pretty_list),
2329
                      code=self.ETYPE_WARNING)
2330

    
2331
      if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
2332
        i_non_a_balanced.append(instance)
2333

    
2334
      for snode in inst_config.secondary_nodes:
2335
        s_img = node_image[snode]
2336
        _ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode,
2337
                 "instance %s, connection to secondary node failed", instance)
2338

    
2339
        if s_img.offline:
2340
          inst_nodes_offline.append(snode)
2341

    
2342
      # warn that the instance lives on offline nodes
2343
      _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
2344
               "instance lives on offline node(s) %s",
2345
               utils.CommaJoin(inst_nodes_offline))
2346
      # ... or ghost/non-vm_capable nodes
2347
      for node in inst_config.all_nodes:
2348
        _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
2349
                 "instance lives on ghost node %s", node)
2350
        _ErrorIf(not node_image[node].vm_capable, self.EINSTANCEBADNODE,
2351
                 instance, "instance lives on non-vm_capable node %s", node)
2352

    
2353
    feedback_fn("* Verifying orphan volumes")
2354
    reserved = utils.FieldSet(*cluster.reserved_lvs)
2355
    self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
2356

    
2357
    feedback_fn("* Verifying orphan instances")
2358
    self._VerifyOrphanInstances(instancelist, node_image)
2359

    
2360
    if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
2361
      feedback_fn("* Verifying N+1 Memory redundancy")
2362
      self._VerifyNPlusOneMemory(node_image, instanceinfo)
2363

    
2364
    feedback_fn("* Other Notes")
2365
    if i_non_redundant:
2366
      feedback_fn("  - NOTICE: %d non-redundant instance(s) found."
2367
                  % len(i_non_redundant))
2368

    
2369
    if i_non_a_balanced:
2370
      feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
2371
                  % len(i_non_a_balanced))
2372

    
2373
    if n_offline:
2374
      feedback_fn("  - NOTICE: %d offline node(s) found." % n_offline)
2375

    
2376
    if n_drained:
2377
      feedback_fn("  - NOTICE: %d drained node(s) found." % n_drained)
2378

    
2379
    return not self.bad
2380

    
2381
  def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
2382
    """Analyze the post-hooks' result
2383

2384
    This method analyses the hook result, handles it, and sends some
2385
    nicely-formatted feedback back to the user.
2386

2387
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
2388
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
2389
    @param hooks_results: the results of the multi-node hooks rpc call
2390
    @param feedback_fn: function used send feedback back to the caller
2391
    @param lu_result: previous Exec result
2392
    @return: the new Exec result, based on the previous result
2393
        and hook results
2394

2395
    """
2396
    # We only really run POST phase hooks, and are only interested in
2397
    # their results
2398
    if phase == constants.HOOKS_PHASE_POST:
2399
      # Used to change hooks' output to proper indentation
2400
      feedback_fn("* Hooks Results")
2401
      assert hooks_results, "invalid result from hooks"
2402

    
2403
      for node_name in hooks_results:
2404
        res = hooks_results[node_name]
2405
        msg = res.fail_msg
2406
        test = msg and not res.offline
2407
        self._ErrorIf(test, self.ENODEHOOKS, node_name,
2408
                      "Communication failure in hooks execution: %s", msg)
2409
        if res.offline or msg:
2410
          # No need to investigate payload if node is offline or gave an error.
2411
          # override manually lu_result here as _ErrorIf only
2412
          # overrides self.bad
2413
          lu_result = 1
2414
          continue
2415
        for script, hkr, output in res.payload:
2416
          test = hkr == constants.HKR_FAIL
2417
          self._ErrorIf(test, self.ENODEHOOKS, node_name,
2418
                        "Script %s failed, output:", script)
2419
          if test:
2420
            output = self._HOOKS_INDENT_RE.sub('      ', output)
2421
            feedback_fn("%s" % output)
2422
            lu_result = 0
2423

    
2424
      return lu_result
2425

    
2426

    
2427
class LUClusterVerifyDisks(NoHooksLU):
2428
  """Verifies the cluster disks status.
2429

2430
  """
2431
  REQ_BGL = False
2432

    
2433
  def ExpandNames(self):
2434
    self.needed_locks = {
2435
      locking.LEVEL_NODE: locking.ALL_SET,
2436
      locking.LEVEL_INSTANCE: locking.ALL_SET,
2437
    }
2438
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
2439

    
2440
  def Exec(self, feedback_fn):
2441
    """Verify integrity of cluster disks.
2442

2443
    @rtype: tuple of three items
2444
    @return: a tuple of (dict of node-to-node_error, list of instances
2445
        which need activate-disks, dict of instance: (node, volume) for
2446
        missing volumes
2447

2448
    """
2449
    result = res_nodes, res_instances, res_missing = {}, [], {}
2450

    
2451
    nodes = utils.NiceSort(self.cfg.GetVmCapableNodeList())
2452
    instances = self.cfg.GetAllInstancesInfo().values()
2453

    
2454
    nv_dict = {}
2455
    for inst in instances:
2456
      inst_lvs = {}
2457
      if not inst.admin_up:
2458
        continue
2459
      inst.MapLVsByNode(inst_lvs)
2460
      # transform { iname: {node: [vol,],},} to {(node, vol): iname}
2461
      for node, vol_list in inst_lvs.iteritems():
2462
        for vol in vol_list:
2463
          nv_dict[(node, vol)] = inst
2464

    
2465
    if not nv_dict:
2466
      return result
2467

    
2468
    node_lvs = self.rpc.call_lv_list(nodes, [])
2469
    for node, node_res in node_lvs.items():
2470
      if node_res.offline:
2471
        continue
2472
      msg = node_res.fail_msg
2473
      if msg:
2474
        logging.warning("Error enumerating LVs on node %s: %s", node, msg)
2475
        res_nodes[node] = msg
2476
        continue
2477

    
2478
      lvs = node_res.payload
2479
      for lv_name, (_, _, lv_online) in lvs.items():
2480
        inst = nv_dict.pop((node, lv_name), None)
2481
        if (not lv_online and inst is not None
2482
            and inst.name not in res_instances):
2483
          res_instances.append(inst.name)
2484

    
2485
    # any leftover items in nv_dict are missing LVs, let's arrange the
2486
    # data better
2487
    for key, inst in nv_dict.iteritems():
2488
      if inst.name not in res_missing:
2489
        res_missing[inst.name] = []
2490
      res_missing[inst.name].append(key)
2491

    
2492
    return result
2493

    
2494

    
2495
class LUClusterRepairDiskSizes(NoHooksLU):
2496
  """Verifies the cluster disks sizes.
2497

2498
  """
2499
  REQ_BGL = False
2500

    
2501
  def ExpandNames(self):
2502
    if self.op.instances:
2503
      self.wanted_names = []
2504
      for name in self.op.instances:
2505
        full_name = _ExpandInstanceName(self.cfg, name)
2506
        self.wanted_names.append(full_name)
2507
      self.needed_locks = {
2508
        locking.LEVEL_NODE: [],
2509
        locking.LEVEL_INSTANCE: self.wanted_names,
2510
        }
2511
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2512
    else:
2513
      self.wanted_names = None
2514
      self.needed_locks = {
2515
        locking.LEVEL_NODE: locking.ALL_SET,
2516
        locking.LEVEL_INSTANCE: locking.ALL_SET,
2517
        }
2518
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
2519

    
2520
  def DeclareLocks(self, level):
2521
    if level == locking.LEVEL_NODE and self.wanted_names is not None:
2522
      self._LockInstancesNodes(primary_only=True)
2523

    
2524
  def CheckPrereq(self):
2525
    """Check prerequisites.
2526

2527
    This only checks the optional instance list against the existing names.
2528

2529
    """
2530
    if self.wanted_names is None:
2531
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
2532

    
2533
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
2534
                             in self.wanted_names]
2535

    
2536
  def _EnsureChildSizes(self, disk):
2537
    """Ensure children of the disk have the needed disk size.
2538

2539
    This is valid mainly for DRBD8 and fixes an issue where the
2540
    children have smaller disk size.
2541

2542
    @param disk: an L{ganeti.objects.Disk} object
2543

2544
    """
2545
    if disk.dev_type == constants.LD_DRBD8:
2546
      assert disk.children, "Empty children for DRBD8?"
2547
      fchild = disk.children[0]
2548
      mismatch = fchild.size < disk.size
2549
      if mismatch:
2550
        self.LogInfo("Child disk has size %d, parent %d, fixing",
2551
                     fchild.size, disk.size)
2552
        fchild.size = disk.size
2553

    
2554
      # and we recurse on this child only, not on the metadev
2555
      return self._EnsureChildSizes(fchild) or mismatch
2556
    else:
2557
      return False
2558

    
2559
  def Exec(self, feedback_fn):
2560
    """Verify the size of cluster disks.
2561

2562
    """
2563
    # TODO: check child disks too
2564
    # TODO: check differences in size between primary/secondary nodes
2565
    per_node_disks = {}
2566
    for instance in self.wanted_instances:
2567
      pnode = instance.primary_node
2568
      if pnode not in per_node_disks:
2569
        per_node_disks[pnode] = []
2570
      for idx, disk in enumerate(instance.disks):
2571
        per_node_disks[pnode].append((instance, idx, disk))
2572

    
2573
    changed = []
2574
    for node, dskl in per_node_disks.items():
2575
      newl = [v[2].Copy() for v in dskl]
2576
      for dsk in newl:
2577
        self.cfg.SetDiskID(dsk, node)
2578
      result = self.rpc.call_blockdev_getsizes(node, newl)
2579
      if result.fail_msg:
2580
        self.LogWarning("Failure in blockdev_getsizes call to node"
2581
                        " %s, ignoring", node)
2582
        continue
2583
      if len(result.data) != len(dskl):
2584
        self.LogWarning("Invalid result from node %s, ignoring node results",
2585
                        node)
2586
        continue
2587
      for ((instance, idx, disk), size) in zip(dskl, result.data):
2588
        if size is None:
2589
          self.LogWarning("Disk %d of instance %s did not return size"
2590
                          " information, ignoring", idx, instance.name)
2591
          continue
2592
        if not isinstance(size, (int, long)):
2593
          self.LogWarning("Disk %d of instance %s did not return valid"
2594
                          " size information, ignoring", idx, instance.name)
2595
          continue
2596
        size = size >> 20
2597
        if size != disk.size:
2598
          self.LogInfo("Disk %d of instance %s has mismatched size,"
2599
                       " correcting: recorded %d, actual %d", idx,
2600
                       instance.name, disk.size, size)
2601
          disk.size = size
2602
          self.cfg.Update(instance, feedback_fn)
2603
          changed.append((instance.name, idx, size))
2604
        if self._EnsureChildSizes(disk):
2605
          self.cfg.Update(instance, feedback_fn)
2606
          changed.append((instance.name, idx, disk.size))
2607
    return changed
2608

    
2609

    
2610
class LUClusterRename(LogicalUnit):
2611
  """Rename the cluster.
2612

2613
  """
2614
  HPATH = "cluster-rename"
2615
  HTYPE = constants.HTYPE_CLUSTER
2616

    
2617
  def BuildHooksEnv(self):
2618
    """Build hooks env.
2619

2620
    """
2621
    env = {
2622
      "OP_TARGET": self.cfg.GetClusterName(),
2623
      "NEW_NAME": self.op.name,
2624
      }
2625
    mn = self.cfg.GetMasterNode()
2626
    all_nodes = self.cfg.GetNodeList()
2627
    return env, [mn], all_nodes
2628

    
2629
  def CheckPrereq(self):
2630
    """Verify that the passed name is a valid one.
2631

2632
    """
2633
    hostname = netutils.GetHostname(name=self.op.name,
2634
                                    family=self.cfg.GetPrimaryIPFamily())
2635

    
2636
    new_name = hostname.name
2637
    self.ip = new_ip = hostname.ip
2638
    old_name = self.cfg.GetClusterName()
2639
    old_ip = self.cfg.GetMasterIP()
2640
    if new_name == old_name and new_ip == old_ip:
2641
      raise errors.OpPrereqError("Neither the name nor the IP address of the"
2642
                                 " cluster has changed",
2643
                                 errors.ECODE_INVAL)
2644
    if new_ip != old_ip:
2645
      if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
2646
        raise errors.OpPrereqError("The given cluster IP address (%s) is"
2647
                                   " reachable on the network" %
2648
                                   new_ip, errors.ECODE_NOTUNIQUE)
2649

    
2650
    self.op.name = new_name
2651

    
2652
  def Exec(self, feedback_fn):
2653
    """Rename the cluster.
2654

2655
    """
2656
    clustername = self.op.name
2657
    ip = self.ip
2658

    
2659
    # shutdown the master IP
2660
    master = self.cfg.GetMasterNode()
2661
    result = self.rpc.call_node_stop_master(master, False)
2662
    result.Raise("Could not disable the master role")
2663

    
2664
    try:
2665
      cluster = self.cfg.GetClusterInfo()
2666
      cluster.cluster_name = clustername
2667
      cluster.master_ip = ip
2668
      self.cfg.Update(cluster, feedback_fn)
2669

    
2670
      # update the known hosts file
2671
      ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
2672
      node_list = self.cfg.GetOnlineNodeList()
2673
      try:
2674
        node_list.remove(master)
2675
      except ValueError:
2676
        pass
2677
      _UploadHelper(self, node_list, constants.SSH_KNOWN_HOSTS_FILE)
2678
    finally:
2679
      result = self.rpc.call_node_start_master(master, False, False)
2680
      msg = result.fail_msg
2681
      if msg:
2682
        self.LogWarning("Could not re-enable the master role on"
2683
                        " the master, please restart manually: %s", msg)
2684

    
2685
    return clustername
2686

    
2687

    
2688
class LUClusterSetParams(LogicalUnit):
2689
  """Change the parameters of the cluster.
2690

2691
  """
2692
  HPATH = "cluster-modify"
2693
  HTYPE = constants.HTYPE_CLUSTER
2694
  REQ_BGL = False
2695

    
2696
  def CheckArguments(self):
2697
    """Check parameters
2698

2699
    """
2700
    if self.op.uid_pool:
2701
      uidpool.CheckUidPool(self.op.uid_pool)
2702

    
2703
    if self.op.add_uids:
2704
      uidpool.CheckUidPool(self.op.add_uids)
2705

    
2706
    if self.op.remove_uids:
2707
      uidpool.CheckUidPool(self.op.remove_uids)
2708

    
2709
  def ExpandNames(self):
2710
    # FIXME: in the future maybe other cluster params won't require checking on
2711
    # all nodes to be modified.
2712
    self.needed_locks = {
2713
      locking.LEVEL_NODE: locking.ALL_SET,
2714
    }
2715
    self.share_locks[locking.LEVEL_NODE] = 1
2716

    
2717
  def BuildHooksEnv(self):
2718
    """Build hooks env.
2719

2720
    """
2721
    env = {
2722
      "OP_TARGET": self.cfg.GetClusterName(),
2723
      "NEW_VG_NAME": self.op.vg_name,
2724
      }
2725
    mn = self.cfg.GetMasterNode()
2726
    return env, [mn], [mn]
2727

    
2728
  def CheckPrereq(self):
2729
    """Check prerequisites.
2730

2731
    This checks whether the given params don't conflict and
2732
    if the given volume group is valid.
2733

2734
    """
2735
    if self.op.vg_name is not None and not self.op.vg_name:
2736
      if self.cfg.HasAnyDiskOfType(constants.LD_LV):
2737
        raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based"
2738
                                   " instances exist", errors.ECODE_INVAL)
2739

    
2740
    if self.op.drbd_helper is not None and not self.op.drbd_helper:
2741
      if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8):
2742
        raise errors.OpPrereqError("Cannot disable drbd helper while"
2743
                                   " drbd-based instances exist",
2744
                                   errors.ECODE_INVAL)
2745

    
2746
    node_list = self.acquired_locks[locking.LEVEL_NODE]
2747

    
2748
    # if vg_name not None, checks given volume group on all nodes
2749
    if self.op.vg_name:
2750
      vglist = self.rpc.call_vg_list(node_list)
2751
      for node in node_list:
2752
        msg = vglist[node].fail_msg
2753
        if msg:
2754
          # ignoring down node
2755
          self.LogWarning("Error while gathering data on node %s"
2756
                          " (ignoring node): %s", node, msg)
2757
          continue
2758
        vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
2759
                                              self.op.vg_name,
2760
                                              constants.MIN_VG_SIZE)
2761
        if vgstatus:
2762
          raise errors.OpPrereqError("Error on node '%s': %s" %
2763
                                     (node, vgstatus), errors.ECODE_ENVIRON)
2764

    
2765
    if self.op.drbd_helper:
2766
      # checks given drbd helper on all nodes
2767
      helpers = self.rpc.call_drbd_helper(node_list)
2768
      for node in node_list:
2769
        ninfo = self.cfg.GetNodeInfo(node)
2770
        if ninfo.offline:
2771
          self.LogInfo("Not checking drbd helper on offline node %s", node)
2772
          continue
2773
        msg = helpers[node].fail_msg
2774
        if msg:
2775
          raise errors.OpPrereqError("Error checking drbd helper on node"
2776
                                     " '%s': %s" % (node, msg),
2777
                                     errors.ECODE_ENVIRON)
2778
        node_helper = helpers[node].payload
2779
        if node_helper != self.op.drbd_helper:
2780
          raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" %
2781
                                     (node, node_helper), errors.ECODE_ENVIRON)
2782

    
2783
    self.cluster = cluster = self.cfg.GetClusterInfo()
2784
    # validate params changes
2785
    if self.op.beparams:
2786
      utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
2787
      self.new_beparams = cluster.SimpleFillBE(self.op.beparams)
2788

    
2789
    if self.op.ndparams:
2790
      utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
2791
      self.new_ndparams = cluster.SimpleFillND(self.op.ndparams)
2792

    
2793
    if self.op.nicparams:
2794
      utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
2795
      self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams)
2796
      objects.NIC.CheckParameterSyntax(self.new_nicparams)
2797
      nic_errors = []
2798

    
2799
      # check all instances for consistency
2800
      for instance in self.cfg.GetAllInstancesInfo().values():
2801
        for nic_idx, nic in enumerate(instance.nics):
2802
          params_copy = copy.deepcopy(nic.nicparams)
2803
          params_filled = objects.FillDict(self.new_nicparams, params_copy)
2804

    
2805
          # check parameter syntax
2806
          try:
2807
            objects.NIC.CheckParameterSyntax(params_filled)
2808
          except errors.ConfigurationError, err:
2809
            nic_errors.append("Instance %s, nic/%d: %s" %
2810
                              (instance.name, nic_idx, err))
2811

    
2812
          # if we're moving instances to routed, check that they have an ip
2813
          target_mode = params_filled[constants.NIC_MODE]
2814
          if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
2815
            nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
2816
                              (instance.name, nic_idx))
2817
      if nic_errors:
2818
        raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
2819
                                   "\n".join(nic_errors))
2820

    
2821
    # hypervisor list/parameters
2822
    self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
2823
    if self.op.hvparams:
2824
      for hv_name, hv_dict in self.op.hvparams.items():
2825
        if hv_name not in self.new_hvparams:
2826
          self.new_hvparams[hv_name] = hv_dict
2827
        else:
2828
          self.new_hvparams[hv_name].update(hv_dict)
2829

    
2830
    # os hypervisor parameters
2831
    self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
2832
    if self.op.os_hvp:
2833
      for os_name, hvs in self.op.os_hvp.items():
2834
        if os_name not in self.new_os_hvp:
2835
          self.new_os_hvp[os_name] = hvs
2836
        else:
2837
          for hv_name, hv_dict in hvs.items():
2838
            if hv_name not in self.new_os_hvp[os_name]:
2839
              self.new_os_hvp[os_name][hv_name] = hv_dict
2840
            else:
2841
              self.new_os_hvp[os_name][hv_name].update(hv_dict)
2842

    
2843
    # os parameters
2844
    self.new_osp = objects.FillDict(cluster.osparams, {})
2845
    if self.op.osparams:
2846
      for os_name, osp in self.op.osparams.items():
2847
        if os_name not in self.new_osp:
2848
          self.new_osp[os_name] = {}
2849

    
2850
        self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp,
2851
                                                  use_none=True)
2852

    
2853
        if not self.new_osp[os_name]:
2854
          # we removed all parameters
2855
          del self.new_osp[os_name]
2856
        else:
2857
          # check the parameter validity (remote check)
2858
          _CheckOSParams(self, False, [self.cfg.GetMasterNode()],
2859
                         os_name, self.new_osp[os_name])
2860

    
2861
    # changes to the hypervisor list
2862
    if self.op.enabled_hypervisors is not None:
2863
      self.hv_list = self.op.enabled_hypervisors
2864
      for hv in self.hv_list:
2865
        # if the hypervisor doesn't already exist in the cluster
2866
        # hvparams, we initialize it to empty, and then (in both
2867
        # cases) we make sure to fill the defaults, as we might not
2868
        # have a complete defaults list if the hypervisor wasn't
2869
        # enabled before
2870
        if hv not in new_hvp:
2871
          new_hvp[hv] = {}
2872
        new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
2873
        utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
2874
    else:
2875
      self.hv_list = cluster.enabled_hypervisors
2876

    
2877
    if self.op.hvparams or self.op.enabled_hypervisors is not None:
2878
      # either the enabled list has changed, or the parameters have, validate
2879
      for hv_name, hv_params in self.new_hvparams.items():
2880
        if ((self.op.hvparams and hv_name in self.op.hvparams) or
2881
            (self.op.enabled_hypervisors and
2882
             hv_name in self.op.enabled_hypervisors)):
2883
          # either this is a new hypervisor, or its parameters have changed
2884
          hv_class = hypervisor.GetHypervisor(hv_name)
2885
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2886
          hv_class.CheckParameterSyntax(hv_params)
2887
          _CheckHVParams(self, node_list, hv_name, hv_params)
2888

    
2889
    if self.op.os_hvp:
2890
      # no need to check any newly-enabled hypervisors, since the
2891
      # defaults have already been checked in the above code-block
2892
      for os_name, os_hvp in self.new_os_hvp.items():
2893
        for hv_name, hv_params in os_hvp.items():
2894
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2895
          # we need to fill in the new os_hvp on top of the actual hv_p
2896
          cluster_defaults = self.new_hvparams.get(hv_name, {})
2897
          new_osp = objects.FillDict(cluster_defaults, hv_params)
2898
          hv_class = hypervisor.GetHypervisor(hv_name)
2899
          hv_class.CheckParameterSyntax(new_osp)
2900
          _CheckHVParams(self, node_list, hv_name, new_osp)
2901

    
2902
    if self.op.default_iallocator:
2903
      alloc_script = utils.FindFile(self.op.default_iallocator,
2904
                                    constants.IALLOCATOR_SEARCH_PATH,
2905
                                    os.path.isfile)
2906
      if alloc_script is None:
2907
        raise errors.OpPrereqError("Invalid default iallocator script '%s'"
2908
                                   " specified" % self.op.default_iallocator,
2909
                                   errors.ECODE_INVAL)
2910

    
2911
  def Exec(self, feedback_fn):
2912
    """Change the parameters of the cluster.
2913

2914
    """
2915
    if self.op.vg_name is not None:
2916
      new_volume = self.op.vg_name
2917
      if not new_volume:
2918
        new_volume = None
2919
      if new_volume != self.cfg.GetVGName():
2920
        self.cfg.SetVGName(new_volume)
2921
      else:
2922
        feedback_fn("Cluster LVM configuration already in desired"
2923
                    " state, not changing")
2924
    if self.op.drbd_helper is not None:
2925
      new_helper = self.op.drbd_helper
2926
      if not new_helper:
2927
        new_helper = None
2928
      if new_helper != self.cfg.GetDRBDHelper():
2929
        self.cfg.SetDRBDHelper(new_helper)
2930
      else:
2931
        feedback_fn("Cluster DRBD helper already in desired state,"
2932
                    " not changing")
2933
    if self.op.hvparams:
2934
      self.cluster.hvparams = self.new_hvparams
2935
    if self.op.os_hvp:
2936
      self.cluster.os_hvp = self.new_os_hvp
2937
    if self.op.enabled_hypervisors is not None:
2938
      self.cluster.hvparams = self.new_hvparams
2939
      self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
2940
    if self.op.beparams:
2941
      self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
2942
    if self.op.nicparams:
2943
      self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
2944
    if self.op.osparams:
2945
      self.cluster.osparams = self.new_osp
2946
    if self.op.ndparams:
2947
      self.cluster.ndparams = self.new_ndparams
2948

    
2949
    if self.op.candidate_pool_size is not None:
2950
      self.cluster.candidate_pool_size = self.op.candidate_pool_size
2951
      # we need to update the pool size here, otherwise the save will fail
2952
      _AdjustCandidatePool(self, [])
2953

    
2954
    if self.op.maintain_node_health is not None:
2955
      self.cluster.maintain_node_health = self.op.maintain_node_health
2956

    
2957
    if self.op.prealloc_wipe_disks is not None:
2958
      self.cluster.prealloc_wipe_disks = self.op.prealloc_wipe_disks
2959

    
2960
    if self.op.add_uids is not None:
2961
      uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
2962

    
2963
    if self.op.remove_uids is not None:
2964
      uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
2965

    
2966
    if self.op.uid_pool is not None:
2967
      self.cluster.uid_pool = self.op.uid_pool
2968

    
2969
    if self.op.default_iallocator is not None:
2970
      self.cluster.default_iallocator = self.op.default_iallocator
2971

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

    
2975
    def helper_os(aname, mods, desc):
2976
      desc += " OS list"
2977
      lst = getattr(self.cluster, aname)
2978
      for key, val in mods:
2979
        if key == constants.DDM_ADD:
2980
          if val in lst:
2981
            feedback_fn("OS %s already in %s, ignoring" % (val, desc))
2982
          else:
2983
            lst.append(val)
2984
        elif key == constants.DDM_REMOVE:
2985
          if val in lst:
2986
            lst.remove(val)
2987
          else:
2988
            feedback_fn("OS %s not found in %s, ignoring" % (val, desc))
2989
        else:
2990
          raise errors.ProgrammerError("Invalid modification '%s'" % key)
2991

    
2992
    if self.op.hidden_os:
2993
      helper_os("hidden_os", self.op.hidden_os, "hidden")
2994

    
2995
    if self.op.blacklisted_os:
2996
      helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted")
2997

    
2998
    if self.op.master_netdev:
2999
      master = self.cfg.GetMasterNode()
3000
      feedback_fn("Shutting down master ip on the current netdev (%s)" %
3001
                  self.cluster.master_netdev)
3002
      result = self.rpc.call_node_stop_master(master, False)
3003
      result.Raise("Could not disable the master ip")
3004
      feedback_fn("Changing master_netdev from %s to %s" %
3005
                  (self.cluster.master_netdev, self.op.master_netdev))
3006
      self.cluster.master_netdev = self.op.master_netdev
3007

    
3008
    self.cfg.Update(self.cluster, feedback_fn)
3009

    
3010
    if self.op.master_netdev:
3011
      feedback_fn("Starting the master ip on the new master netdev (%s)" %
3012
                  self.op.master_netdev)
3013
      result = self.rpc.call_node_start_master(master, False, False)
3014
      if result.fail_msg:
3015
        self.LogWarning("Could not re-enable the master ip on"
3016
                        " the master, please restart manually: %s",
3017
                        result.fail_msg)
3018

    
3019

    
3020
def _UploadHelper(lu, nodes, fname):
3021
  """Helper for uploading a file and showing warnings.
3022

3023
  """
3024
  if os.path.exists(fname):
3025
    result = lu.rpc.call_upload_file(nodes, fname)
3026
    for to_node, to_result in result.items():
3027
      msg = to_result.fail_msg
3028
      if msg:
3029
        msg = ("Copy of file %s to node %s failed: %s" %
3030
               (fname, to_node, msg))
3031
        lu.proc.LogWarning(msg)
3032

    
3033

    
3034
def _RedistributeAncillaryFiles(lu, additional_nodes=None, additional_vm=True):
3035
  """Distribute additional files which are part of the cluster configuration.
3036

3037
  ConfigWriter takes care of distributing the config and ssconf files, but
3038
  there are more files which should be distributed to all nodes. This function
3039
  makes sure those are copied.
3040

3041
  @param lu: calling logical unit
3042
  @param additional_nodes: list of nodes not in the config to distribute to
3043
  @type additional_vm: boolean
3044
  @param additional_vm: whether the additional nodes are vm-capable or not
3045

3046
  """
3047
  # 1. Gather target nodes
3048
  myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
3049
  dist_nodes = lu.cfg.GetOnlineNodeList()
3050
  nvm_nodes = lu.cfg.GetNonVmCapableNodeList()
3051
  vm_nodes = [name for name in dist_nodes if name not in nvm_nodes]
3052
  if additional_nodes is not None:
3053
    dist_nodes.extend(additional_nodes)
3054
    if additional_vm:
3055
      vm_nodes.extend(additional_nodes)
3056
  if myself.name in dist_nodes:
3057
    dist_nodes.remove(myself.name)
3058
  if myself.name in vm_nodes:
3059
    vm_nodes.remove(myself.name)
3060

    
3061
  # 2. Gather files to distribute
3062
  dist_files = set([constants.ETC_HOSTS,
3063
                    constants.SSH_KNOWN_HOSTS_FILE,
3064
                    constants.RAPI_CERT_FILE,
3065
                    constants.RAPI_USERS_FILE,
3066
                    constants.CONFD_HMAC_KEY,
3067
                    constants.CLUSTER_DOMAIN_SECRET_FILE,
3068
                   ])
3069

    
3070
  vm_files = set()
3071
  enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
3072
  for hv_name in enabled_hypervisors:
3073
    hv_class = hypervisor.GetHypervisor(hv_name)
3074
    vm_files.update(hv_class.GetAncillaryFiles())
3075

    
3076
  # 3. Perform the files upload
3077
  for fname in dist_files:
3078
    _UploadHelper(lu, dist_nodes, fname)
3079
  for fname in vm_files:
3080
    _UploadHelper(lu, vm_nodes, fname)
3081

    
3082

    
3083
class LUClusterRedistConf(NoHooksLU):
3084
  """Force the redistribution of cluster configuration.
3085

3086
  This is a very simple LU.
3087

3088
  """
3089
  REQ_BGL = False
3090

    
3091
  def ExpandNames(self):
3092
    self.needed_locks = {
3093
      locking.LEVEL_NODE: locking.ALL_SET,
3094
    }
3095
    self.share_locks[locking.LEVEL_NODE] = 1
3096

    
3097
  def Exec(self, feedback_fn):
3098
    """Redistribute the configuration.
3099

3100
    """
3101
    self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
3102
    _RedistributeAncillaryFiles(self)
3103

    
3104

    
3105
def _WaitForSync(lu, instance, disks=None, oneshot=False):
3106
  """Sleep and poll for an instance's disk to sync.
3107

3108
  """
3109
  if not instance.disks or disks is not None and not disks:
3110
    return True
3111

    
3112
  disks = _ExpandCheckDisks(instance, disks)
3113

    
3114
  if not oneshot:
3115
    lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
3116

    
3117
  node = instance.primary_node
3118

    
3119
  for dev in disks:
3120
    lu.cfg.SetDiskID(dev, node)
3121

    
3122
  # TODO: Convert to utils.Retry
3123

    
3124
  retries = 0
3125
  degr_retries = 10 # in seconds, as we sleep 1 second each time
3126
  while True:
3127
    max_time = 0
3128
    done = True
3129
    cumul_degraded = False
3130
    rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
3131
    msg = rstats.fail_msg
3132
    if msg:
3133
      lu.LogWarning("Can't get any data from node %s: %s", node, msg)
3134
      retries += 1
3135
      if retries >= 10:
3136
        raise errors.RemoteError("Can't contact node %s for mirror data,"
3137
                                 " aborting." % node)
3138
      time.sleep(6)
3139
      continue
3140
    rstats = rstats.payload
3141
    retries = 0
3142
    for i, mstat in enumerate(rstats):
3143
      if mstat is None:
3144
        lu.LogWarning("Can't compute data for node %s/%s",
3145
                           node, disks[i].iv_name)
3146
        continue
3147

    
3148
      cumul_degraded = (cumul_degraded or
3149
                        (mstat.is_degraded and mstat.sync_percent is None))
3150
      if mstat.sync_percent is not None:
3151
        done = False
3152
        if mstat.estimated_time is not None:
3153
          rem_time = ("%s remaining (estimated)" %
3154
                      utils.FormatSeconds(mstat.estimated_time))
3155
          max_time = mstat.estimated_time
3156
        else:
3157
          rem_time = "no time estimate"
3158
        lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
3159
                        (disks[i].iv_name, mstat.sync_percent, rem_time))
3160

    
3161
    # if we're done but degraded, let's do a few small retries, to
3162
    # make sure we see a stable and not transient situation; therefore
3163
    # we force restart of the loop
3164
    if (done or oneshot) and cumul_degraded and degr_retries > 0:
3165
      logging.info("Degraded disks found, %d retries left", degr_retries)
3166
      degr_retries -= 1
3167
      time.sleep(1)
3168
      continue
3169

    
3170
    if done or oneshot:
3171
      break
3172

    
3173
    time.sleep(min(60, max_time))
3174

    
3175
  if done:
3176
    lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
3177
  return not cumul_degraded
3178

    
3179

    
3180
def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
3181
  """Check that mirrors are not degraded.
3182

3183
  The ldisk parameter, if True, will change the test from the
3184
  is_degraded attribute (which represents overall non-ok status for
3185
  the device(s)) to the ldisk (representing the local storage status).
3186

3187
  """
3188
  lu.cfg.SetDiskID(dev, node)
3189

    
3190
  result = True
3191

    
3192
  if on_primary or dev.AssembleOnSecondary():
3193
    rstats = lu.rpc.call_blockdev_find(node, dev)
3194
    msg = rstats.fail_msg
3195
    if msg:
3196
      lu.LogWarning("Can't find disk on node %s: %s", node, msg)
3197
      result = False
3198
    elif not rstats.payload:
3199
      lu.LogWarning("Can't find disk on node %s", node)
3200
      result = False
3201
    else:
3202
      if ldisk:
3203
        result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
3204
      else:
3205
        result = result and not rstats.payload.is_degraded
3206

    
3207
  if dev.children:
3208
    for child in dev.children:
3209
      result = result and _CheckDiskConsistency(lu, child, node, on_primary)
3210

    
3211
  return result
3212

    
3213

    
3214
class LUOobCommand(NoHooksLU):
3215
  """Logical unit for OOB handling.
3216

3217
  """
3218
  REG_BGL = False
3219

    
3220
  def CheckPrereq(self):
3221
    """Check prerequisites.
3222

3223
    This checks:
3224
     - the node exists in the configuration
3225
     - OOB is supported
3226

3227
    Any errors are signaled by raising errors.OpPrereqError.
3228

3229
    """
3230
    self.nodes = []
3231
    master_node = self.cfg.GetMasterNode()
3232
    for node_name in self.op.node_names:
3233
      node = self.cfg.GetNodeInfo(node_name)
3234

    
3235
      if node is None:
3236
        raise errors.OpPrereqError("Node %s not found" % node_name,
3237
                                   errors.ECODE_NOENT)
3238
      else:
3239
        self.nodes.append(node)
3240

    
3241
      if (not self.op.ignore_status and
3242
          (self.op.command == constants.OOB_POWER_OFF and not node.offline)):
3243
        raise errors.OpPrereqError(("Cannot power off node %s because it is"
3244
                                    " not marked offline") % node_name,
3245
                                   errors.ECODE_STATE)
3246

    
3247
    if self.op.command in (constants.OOB_POWER_OFF, constants.OOB_POWER_CYCLE):
3248
      # This does two things, it checks if master is in the list and if so and
3249
      # force_master is set it puts it to the end so the master is done last
3250
      try:
3251
        self.op.node_names.remove(master_node)
3252
      except ValueError:
3253
        pass
3254
      else:
3255
        if self.op.force_master:
3256
          self.op.node_names.append(master_node)
3257
        else:
3258
          self.LogWarning("Master %s was skipped, use the force master"
3259
                          " option to operate on the master too",
3260
                          master_node)
3261
          if not self.op.node_names:
3262
            raise errors.OpPrereqError("No nodes left to operate on, aborting",
3263
                                       errors.ECODE_INVAL)
3264

    
3265
      assert (master_node not in self.op.node_names or
3266
              self.op.node_names[-1] == master_node)
3267

    
3268
  def ExpandNames(self):
3269
    """Gather locks we need.
3270

3271
    """
3272
    if self.op.node_names:
3273
      self.op.node_names = [_ExpandNodeName(self.cfg, name)
3274
                            for name in self.op.node_names]
3275
    else:
3276
      self.op.node_names = self.cfg.GetNodeList()
3277

    
3278
    self.needed_locks = {
3279
      locking.LEVEL_NODE: self.op.node_names,
3280
      }
3281

    
3282
  def Exec(self, feedback_fn):
3283
    """Execute OOB and return result if we expect any.
3284

3285
    """
3286
    master_node = self.cfg.GetMasterNode()
3287
    ret = []
3288

    
3289
    for node in self.nodes:
3290
      node_entry = [(constants.RS_NORMAL, node.name)]
3291
      ret.append(node_entry)
3292

    
3293
      oob_program = _SupportsOob(self.cfg, node)
3294

    
3295
      if not oob_program:
3296
        node_entry.append((constants.RS_UNAVAIL, None))
3297
        continue
3298

    
3299
      logging.info("Executing out-of-band command '%s' using '%s' on %s",
3300
                   self.op.command, oob_program, node.name)
3301
      result = self.rpc.call_run_oob(master_node, oob_program,
3302
                                     self.op.command, node.name,
3303
                                     self.op.timeout)
3304

    
3305
      if result.fail_msg:
3306
        self.LogWarning("On node '%s' out-of-band RPC failed with: %s",
3307
                        node.name, result.fail_msg)
3308
        node_entry.append((constants.RS_NODATA, None))
3309
      else:
3310
        try:
3311
          self._CheckPayload(result)
3312
        except errors.OpExecError, err:
3313
          self.LogWarning("The payload returned by '%s' is not valid: %s",
3314
                          node.name, err)
3315
          node_entry.append((constants.RS_NODATA, None))
3316
        else:
3317
          if self.op.command == constants.OOB_HEALTH:
3318
            # For health we should log important events
3319
            for item, status in result.payload:
3320
              if status in [constants.OOB_STATUS_WARNING,
3321
                            constants.OOB_STATUS_CRITICAL]:
3322
                self.LogWarning("On node '%s' item '%s' has status '%s'",
3323
                                node.name, item, status)
3324

    
3325
          if self.op.command == constants.OOB_POWER_ON:
3326
            node.powered = True
3327
          elif self.op.command == constants.OOB_POWER_OFF:
3328
            node.powered = False
3329
          elif self.op.command == constants.OOB_POWER_STATUS:
3330
            powered = result.payload[constants.OOB_POWER_STATUS_POWERED]
3331
            if powered != node.powered:
3332
              logging.warning(("Recorded power state (%s) of node '%s' does not"
3333
                               " match actual power state (%s)"), node.powered,
3334
                              node.name, powered)
3335

    
3336
          # For configuration changing commands we should update the node
3337
          if self.op.command in (constants.OOB_POWER_ON,
3338
                                 constants.OOB_POWER_OFF):
3339
            self.cfg.Update(node, feedback_fn)
3340

    
3341
          node_entry.append((constants.RS_NORMAL, result.payload))
3342

    
3343
    return ret
3344

    
3345
  def _CheckPayload(self, result):
3346
    """Checks if the payload is valid.
3347

3348
    @param result: RPC result
3349
    @raises errors.OpExecError: If payload is not valid
3350

3351
    """
3352
    errs = []
3353
    if self.op.command == constants.OOB_HEALTH:
3354
      if not isinstance(result.payload, list):
3355
        errs.append("command 'health' is expected to return a list but got %s" %
3356
                    type(result.payload))
3357
      else:
3358
        for item, status in result.payload:
3359
          if status not in constants.OOB_STATUSES:
3360
            errs.append("health item '%s' has invalid status '%s'" %
3361
                        (item, status))
3362

    
3363
    if self.op.command == constants.OOB_POWER_STATUS:
3364
      if not isinstance(result.payload, dict):
3365
        errs.append("power-status is expected to return a dict but got %s" %
3366
                    type(result.payload))
3367

    
3368
    if self.op.command in [
3369
        constants.OOB_POWER_ON,
3370
        constants.OOB_POWER_OFF,
3371
        constants.OOB_POWER_CYCLE,
3372
        ]:
3373
      if result.payload is not None:
3374
        errs.append("%s is expected to not return payload but got '%s'" %
3375
                    (self.op.command, result.payload))
3376

    
3377
    if errs:
3378
      raise errors.OpExecError("Check of out-of-band payload failed due to %s" %
3379
                               utils.CommaJoin(errs))
3380

    
3381

    
3382

    
3383
class LUOsDiagnose(NoHooksLU):
3384
  """Logical unit for OS diagnose/query.
3385

3386
  """
3387
  REQ_BGL = False
3388
  _HID = "hidden"
3389
  _BLK = "blacklisted"
3390
  _VLD = "valid"
3391
  _FIELDS_STATIC = utils.FieldSet()
3392
  _FIELDS_DYNAMIC = utils.FieldSet("name", _VLD, "node_status", "variants",
3393
                                   "parameters", "api_versions", _HID, _BLK)
3394

    
3395
  def CheckArguments(self):
3396
    if self.op.names:
3397
      raise errors.OpPrereqError("Selective OS query not supported",
3398
                                 errors.ECODE_INVAL)
3399

    
3400
    _CheckOutputFields(static=self._FIELDS_STATIC,
3401
                       dynamic=self._FIELDS_DYNAMIC,
3402
                       selected=self.op.output_fields)
3403

    
3404
  def ExpandNames(self):
3405
    # Lock all nodes, in shared mode
3406
    # Temporary removal of locks, should be reverted later
3407
    # TODO: reintroduce locks when they are lighter-weight
3408
    self.needed_locks = {}
3409
    #self.share_locks[locking.LEVEL_NODE] = 1
3410
    #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3411

    
3412
  @staticmethod
3413
  def _DiagnoseByOS(rlist):
3414
    """Remaps a per-node return list into an a per-os per-node dictionary
3415

3416
    @param rlist: a map with node names as keys and OS objects as values
3417

3418
    @rtype: dict
3419
    @return: a dictionary with osnames as keys and as value another
3420
        map, with nodes as keys and tuples of (path, status, diagnose,
3421
        variants, parameters, api_versions) as values, eg::
3422

3423
          {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
3424
                                     (/srv/..., False, "invalid api")],
3425
                           "node2": [(/srv/..., True, "", [], [])]}
3426
          }
3427

3428
    """
3429
    all_os = {}
3430
    # we build here the list of nodes that didn't fail the RPC (at RPC
3431
    # level), so that nodes with a non-responding node daemon don't
3432
    # make all OSes invalid
3433
    good_nodes = [node_name for node_name in rlist
3434
                  if not rlist[node_name].fail_msg]
3435
    for node_name, nr in rlist.items():
3436
      if nr.fail_msg or not nr.payload:
3437
        continue
3438
      for (name, path, status, diagnose, variants,
3439
           params, api_versions) in nr.payload:
3440
        if name not in all_os:
3441
          # build a list of nodes for this os containing empty lists
3442
          # for each node in node_list
3443
          all_os[name] = {}
3444
          for nname in good_nodes:
3445
            all_os[name][nname] = []
3446
        # convert params from [name, help] to (name, help)
3447
        params = [tuple(v) for v in params]
3448
        all_os[name][node_name].append((path, status, diagnose,
3449
                                        variants, params, api_versions))
3450
    return all_os
3451

    
3452
  def Exec(self, feedback_fn):
3453
    """Compute the list of OSes.
3454

3455
    """
3456
    valid_nodes = [node.name
3457
                   for node in self.cfg.GetAllNodesInfo().values()
3458
                   if not node.offline and node.vm_capable]
3459
    node_data = self.rpc.call_os_diagnose(valid_nodes)
3460
    pol = self._DiagnoseByOS(node_data)
3461
    output = []
3462
    cluster = self.cfg.GetClusterInfo()
3463

    
3464
    for os_name in utils.NiceSort(pol.keys()):
3465
      os_data = pol[os_name]
3466
      row = []
3467
      valid = True
3468
      (variants, params, api_versions) = null_state = (set(), set(), set())
3469
      for idx, osl in enumerate(os_data.values()):
3470
        valid = bool(valid and osl and osl[0][1])
3471
        if not valid:
3472
          (variants, params, api_versions) = null_state
3473
          break
3474
        node_variants, node_params, node_api = osl[0][3:6]
3475
        if idx == 0: # first entry
3476
          variants = set(node_variants)
3477
          params = set(node_params)
3478
          api_versions = set(node_api)
3479
        else: # keep consistency
3480
          variants.intersection_update(node_variants)
3481
          params.intersection_update(node_params)
3482
          api_versions.intersection_update(node_api)
3483

    
3484
      is_hid = os_name in cluster.hidden_os
3485
      is_blk = os_name in cluster.blacklisted_os
3486
      if ((self._HID not in self.op.output_fields and is_hid) or
3487
          (self._BLK not in self.op.output_fields and is_blk) or
3488
          (self._VLD not in self.op.output_fields and not valid)):
3489
        continue
3490

    
3491
      for field in self.op.output_fields:
3492
        if field == "name":
3493
          val = os_name
3494
        elif field == self._VLD:
3495
          val = valid
3496
        elif field == "node_status":
3497
          # this is just a copy of the dict
3498
          val = {}
3499
          for node_name, nos_list in os_data.items():
3500
            val[node_name] = nos_list
3501
        elif field == "variants":
3502
          val = utils.NiceSort(list(variants))
3503
        elif field == "parameters":
3504
          val = list(params)
3505
        elif field == "api_versions":
3506
          val = list(api_versions)
3507
        elif field == self._HID:
3508
          val = is_hid
3509
        elif field == self._BLK:
3510
          val = is_blk
3511
        else:
3512
          raise errors.ParameterError(field)
3513
        row.append(val)
3514
      output.append(row)
3515

    
3516
    return output
3517

    
3518

    
3519
class LUNodeRemove(LogicalUnit):
3520
  """Logical unit for removing a node.
3521

3522
  """
3523
  HPATH = "node-remove"
3524
  HTYPE = constants.HTYPE_NODE
3525

    
3526
  def BuildHooksEnv(self):
3527
    """Build hooks env.
3528

3529
    This doesn't run on the target node in the pre phase as a failed
3530
    node would then be impossible to remove.
3531

3532
    """
3533
    env = {
3534
      "OP_TARGET": self.op.node_name,
3535
      "NODE_NAME": self.op.node_name,
3536
      }
3537
    all_nodes = self.cfg.GetNodeList()
3538
    try:
3539
      all_nodes.remove(self.op.node_name)
3540
    except ValueError:
3541
      logging.warning("Node %s which is about to be removed not found"
3542
                      " in the all nodes list", self.op.node_name)
3543
    return env, all_nodes, all_nodes
3544

    
3545
  def CheckPrereq(self):
3546
    """Check prerequisites.
3547

3548
    This checks:
3549
     - the node exists in the configuration
3550
     - it does not have primary or secondary instances
3551
     - it's not the master
3552

3553
    Any errors are signaled by raising errors.OpPrereqError.
3554

3555
    """
3556
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3557
    node = self.cfg.GetNodeInfo(self.op.node_name)
3558
    assert node is not None
3559

    
3560
    instance_list = self.cfg.GetInstanceList()
3561

    
3562
    masternode = self.cfg.GetMasterNode()
3563
    if node.name == masternode:
3564
      raise errors.OpPrereqError("Node is the master node,"
3565
                                 " you need to failover first.",
3566
                                 errors.ECODE_INVAL)
3567

    
3568
    for instance_name in instance_list:
3569
      instance = self.cfg.GetInstanceInfo(instance_name)
3570
      if node.name in instance.all_nodes:
3571
        raise errors.OpPrereqError("Instance %s is still running on the node,"
3572
                                   " please remove first." % instance_name,
3573
                                   errors.ECODE_INVAL)
3574
    self.op.node_name = node.name
3575
    self.node = node
3576

    
3577
  def Exec(self, feedback_fn):
3578
    """Removes the node from the cluster.
3579

3580
    """
3581
    node = self.node
3582
    logging.info("Stopping the node daemon and removing configs from node %s",
3583
                 node.name)
3584

    
3585
    modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
3586

    
3587
    # Promote nodes to master candidate as needed
3588
    _AdjustCandidatePool(self, exceptions=[node.name])
3589
    self.context.RemoveNode(node.name)
3590

    
3591
    # Run post hooks on the node before it's removed
3592
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
3593
    try:
3594
      hm.RunPhase(constants.HOOKS_PHASE_POST, [node.name])
3595
    except:
3596
      # pylint: disable-msg=W0702
3597
      self.LogWarning("Errors occurred running hooks on %s" % node.name)
3598

    
3599
    result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
3600
    msg = result.fail_msg
3601
    if msg:
3602
      self.LogWarning("Errors encountered on the remote node while leaving"
3603
                      " the cluster: %s", msg)
3604

    
3605
    # Remove node from our /etc/hosts
3606
    if self.cfg.GetClusterInfo().modify_etc_hosts:
3607
      master_node = self.cfg.GetMasterNode()
3608
      result = self.rpc.call_etc_hosts_modify(master_node,
3609
                                              constants.ETC_HOSTS_REMOVE,
3610
                                              node.name, None)
3611
      result.Raise("Can't update hosts file with new host data")
3612
      _RedistributeAncillaryFiles(self)
3613

    
3614

    
3615
class _NodeQuery(_QueryBase):
3616
  FIELDS = query.NODE_FIELDS
3617

    
3618
  def ExpandNames(self, lu):
3619
    lu.needed_locks = {}
3620
    lu.share_locks[locking.LEVEL_NODE] = 1
3621

    
3622
    if self.names:
3623
      self.wanted = _GetWantedNodes(lu, self.names)
3624
    else:
3625
      self.wanted = locking.ALL_SET
3626

    
3627
    self.do_locking = (self.use_locking and
3628
                       query.NQ_LIVE in self.requested_data)
3629

    
3630
    if self.do_locking:
3631
      # if we don't request only static fields, we need to lock the nodes
3632
      lu.needed_locks[locking.LEVEL_NODE] = self.wanted
3633

    
3634
  def DeclareLocks(self, lu, level):
3635
    pass
3636

    
3637
  def _GetQueryData(self, lu):
3638
    """Computes the list of nodes and their attributes.
3639

3640
    """
3641
    all_info = lu.cfg.GetAllNodesInfo()
3642

    
3643
    nodenames = self._GetNames(lu, all_info.keys(), locking.LEVEL_NODE)
3644

    
3645
    # Gather data as requested
3646
    if query.NQ_LIVE in self.requested_data:
3647
      node_data = lu.rpc.call_node_info(nodenames, lu.cfg.GetVGName(),
3648
                                        lu.cfg.GetHypervisorType())
3649
      live_data = dict((name, nresult.payload)
3650
                       for (name, nresult) in node_data.items()
3651
                       if not nresult.fail_msg and nresult.payload)
3652
    else:
3653
      live_data = None
3654

    
3655
    if query.NQ_INST in self.requested_data:
3656
      node_to_primary = dict([(name, set()) for name in nodenames])
3657
      node_to_secondary = dict([(name, set()) for name in nodenames])
3658

    
3659
      inst_data = lu.cfg.GetAllInstancesInfo()
3660

    
3661
      for inst in inst_data.values():
3662
        if inst.primary_node in node_to_primary:
3663
          node_to_primary[inst.primary_node].add(inst.name)
3664
        for secnode in inst.secondary_nodes:
3665
          if secnode in node_to_secondary:
3666
            node_to_secondary[secnode].add(inst.name)
3667
    else:
3668
      node_to_primary = None
3669
      node_to_secondary = None
3670

    
3671
    if query.NQ_OOB in self.requested_data:
3672
      oob_support = dict((name, bool(_SupportsOob(lu.cfg, node)))
3673
                         for name, node in all_info.iteritems())
3674
    else:
3675
      oob_support = None
3676

    
3677
    if query.NQ_GROUP in self.requested_data:
3678
      groups = lu.cfg.GetAllNodeGroupsInfo()
3679
    else:
3680
      groups = {}
3681

    
3682
    return query.NodeQueryData([all_info[name] for name in nodenames],
3683
                               live_data, lu.cfg.GetMasterNode(),
3684
                               node_to_primary, node_to_secondary, groups,
3685
                               oob_support, lu.cfg.GetClusterInfo())
3686

    
3687

    
3688
class LUNodeQuery(NoHooksLU):
3689
  """Logical unit for querying nodes.
3690

3691
  """
3692
  # pylint: disable-msg=W0142
3693
  REQ_BGL = False
3694

    
3695
  def CheckArguments(self):
3696
    self.nq = _NodeQuery(self.op.names, self.op.output_fields,
3697
                         self.op.use_locking)
3698

    
3699
  def ExpandNames(self):
3700
    self.nq.ExpandNames(self)
3701

    
3702
  def Exec(self, feedback_fn):
3703
    return self.nq.OldStyleQuery(self)
3704

    
3705

    
3706
class LUNodeQueryvols(NoHooksLU):
3707
  """Logical unit for getting volumes on node(s).
3708

3709
  """
3710
  REQ_BGL = False
3711
  _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
3712
  _FIELDS_STATIC = utils.FieldSet("node")
3713

    
3714
  def CheckArguments(self):
3715
    _CheckOutputFields(static=self._FIELDS_STATIC,
3716
                       dynamic=self._FIELDS_DYNAMIC,
3717
                       selected=self.op.output_fields)
3718

    
3719
  def ExpandNames(self):
3720
    self.needed_locks = {}
3721
    self.share_locks[locking.LEVEL_NODE] = 1
3722
    if not self.op.nodes:
3723
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3724
    else:
3725
      self.needed_locks[locking.LEVEL_NODE] = \
3726
        _GetWantedNodes(self, self.op.nodes)
3727

    
3728
  def Exec(self, feedback_fn):
3729
    """Computes the list of nodes and their attributes.
3730

3731
    """
3732
    nodenames = self.acquired_locks[locking.LEVEL_NODE]
3733
    volumes = self.rpc.call_node_volumes(nodenames)
3734

    
3735
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
3736
             in self.cfg.GetInstanceList()]
3737

    
3738
    lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
3739

    
3740
    output = []
3741
    for node in nodenames:
3742
      nresult = volumes[node]
3743
      if nresult.offline:
3744
        continue
3745
      msg = nresult.fail_msg
3746
      if msg:
3747
        self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
3748
        continue
3749

    
3750
      node_vols = nresult.payload[:]
3751
      node_vols.sort(key=lambda vol: vol['dev'])
3752

    
3753
      for vol in node_vols:
3754
        node_output = []
3755
        for field in self.op.output_fields:
3756
          if field == "node":
3757
            val = node
3758
          elif field == "phys":
3759
            val = vol['dev']
3760
          elif field == "vg":
3761
            val = vol['vg']
3762
          elif field == "name":
3763
            val = vol['name']
3764
          elif field == "size":
3765
            val = int(float(vol['size']))
3766
          elif field == "instance":
3767
            for inst in ilist:
3768
              if node not in lv_by_node[inst]:
3769
                continue
3770
              if vol['name'] in lv_by_node[inst][node]:
3771
                val = inst.name
3772
                break
3773
            else:
3774
              val = '-'
3775
          else:
3776
            raise errors.ParameterError(field)
3777
          node_output.append(str(val))
3778

    
3779
        output.append(node_output)
3780

    
3781
    return output
3782

    
3783

    
3784
class LUNodeQueryStorage(NoHooksLU):
3785
  """Logical unit for getting information on storage units on node(s).
3786

3787
  """
3788
  _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
3789
  REQ_BGL = False
3790

    
3791
  def CheckArguments(self):
3792
    _CheckOutputFields(static=self._FIELDS_STATIC,
3793
                       dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
3794
                       selected=self.op.output_fields)
3795

    
3796
  def ExpandNames(self):
3797
    self.needed_locks = {}
3798
    self.share_locks[locking.LEVEL_NODE] = 1
3799

    
3800
    if self.op.nodes:
3801
      self.needed_locks[locking.LEVEL_NODE] = \
3802
        _GetWantedNodes(self, self.op.nodes)
3803
    else:
3804
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3805

    
3806
  def Exec(self, feedback_fn):
3807
    """Computes the list of nodes and their attributes.
3808

3809
    """
3810
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
3811

    
3812
    # Always get name to sort by
3813
    if constants.SF_NAME in self.op.output_fields:
3814
      fields = self.op.output_fields[:]
3815
    else:
3816
      fields = [constants.SF_NAME] + self.op.output_fields
3817

    
3818
    # Never ask for node or type as it's only known to the LU
3819
    for extra in [constants.SF_NODE, constants.SF_TYPE]:
3820
      while extra in fields:
3821
        fields.remove(extra)
3822

    
3823
    field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
3824
    name_idx = field_idx[constants.SF_NAME]
3825

    
3826
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3827
    data = self.rpc.call_storage_list(self.nodes,
3828
                                      self.op.storage_type, st_args,
3829
                                      self.op.name, fields)
3830

    
3831
    result = []
3832

    
3833
    for node in utils.NiceSort(self.nodes):
3834
      nresult = data[node]
3835
      if nresult.offline:
3836
        continue
3837

    
3838
      msg = nresult.fail_msg
3839
      if msg:
3840
        self.LogWarning("Can't get storage data from node %s: %s", node, msg)
3841
        continue
3842

    
3843
      rows = dict([(row[name_idx], row) for row in nresult.payload])
3844

    
3845
      for name in utils.NiceSort(rows.keys()):
3846
        row = rows[name]
3847

    
3848
        out = []
3849

    
3850
        for field in self.op.output_fields:
3851
          if field == constants.SF_NODE:
3852
            val = node
3853
          elif field == constants.SF_TYPE:
3854
            val = self.op.storage_type
3855
          elif field in field_idx:
3856
            val = row[field_idx[field]]
3857
          else:
3858
            raise errors.ParameterError(field)
3859

    
3860
          out.append(val)
3861

    
3862
        result.append(out)
3863

    
3864
    return result
3865

    
3866

    
3867
class _InstanceQuery(_QueryBase):
3868
  FIELDS = query.INSTANCE_FIELDS
3869

    
3870
  def ExpandNames(self, lu):
3871
    lu.needed_locks = {}
3872
    lu.share_locks[locking.LEVEL_INSTANCE] = 1
3873
    lu.share_locks[locking.LEVEL_NODE] = 1
3874

    
3875
    if self.names:
3876
      self.wanted = _GetWantedInstances(lu, self.names)
3877
    else:
3878
      self.wanted = locking.ALL_SET
3879

    
3880
    self.do_locking = (self.use_locking and
3881
                       query.IQ_LIVE in self.requested_data)
3882
    if self.do_locking:
3883
      lu.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
3884
      lu.needed_locks[locking.LEVEL_NODE] = []
3885
      lu.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3886

    
3887
  def DeclareLocks(self, lu, level):
3888
    if level == locking.LEVEL_NODE and self.do_locking:
3889
      lu._LockInstancesNodes() # pylint: disable-msg=W0212
3890

    
3891
  def _GetQueryData(self, lu):
3892
    """Computes the list of instances and their attributes.
3893

3894
    """
3895
    all_info = lu.cfg.GetAllInstancesInfo()
3896

    
3897
    instance_names = self._GetNames(lu, all_info.keys(), locking.LEVEL_INSTANCE)
3898

    
3899
    instance_list = [all_info[name] for name in instance_names]
3900
    nodes = frozenset([inst.primary_node for inst in instance_list])
3901
    hv_list = list(set([inst.hypervisor for inst in instance_list]))
3902
    bad_nodes = []
3903
    offline_nodes = []
3904

    
3905
    # Gather data as requested
3906
    if query.IQ_LIVE in self.requested_data:
3907
      live_data = {}
3908
      node_data = lu.rpc.call_all_instances_info(nodes, hv_list)
3909
      for name in nodes:
3910
        result = node_data[name]
3911
        if result.offline:
3912
          # offline nodes will be in both lists
3913
          assert result.fail_msg
3914
          offline_nodes.append(name)
3915
        if result.fail_msg:
3916
          bad_nodes.append(name)
3917
        elif result.payload:
3918
          live_data.update(result.payload)
3919
        # else no instance is alive
3920
    else:
3921
      live_data = {}
3922

    
3923
    if query.IQ_DISKUSAGE in self.requested_data:
3924
      disk_usage = dict((inst.name,
3925
                         _ComputeDiskSize(inst.disk_template,
3926
                                          [{"size": disk.size}
3927
                                           for disk in inst.disks]))
3928
                        for inst in instance_list)
3929
    else:
3930
      disk_usage = None
3931

    
3932
    return query.InstanceQueryData(instance_list, lu.cfg.GetClusterInfo(),
3933
                                   disk_usage, offline_nodes, bad_nodes,
3934
                                   live_data)
3935

    
3936

    
3937
class LUQuery(NoHooksLU):
3938
  """Query for resources/items of a certain kind.
3939

3940
  """
3941
  # pylint: disable-msg=W0142
3942
  REQ_BGL = False
3943

    
3944
  def CheckArguments(self):
3945
    qcls = _GetQueryImplementation(self.op.what)
3946
    names = qlang.ReadSimpleFilter("name", self.op.filter)
3947

    
3948
    self.impl = qcls(names, self.op.fields, False)
3949

    
3950
  def ExpandNames(self):
3951
    self.impl.ExpandNames(self)
3952

    
3953
  def DeclareLocks(self, level):
3954
    self.impl.DeclareLocks(self, level)
3955

    
3956
  def Exec(self, feedback_fn):
3957
    return self.impl.NewStyleQuery(self)
3958

    
3959

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

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

    
3967
  def CheckArguments(self):
3968
    self.qcls = _GetQueryImplementation(self.op.what)
3969

    
3970
  def ExpandNames(self):
3971
    self.needed_locks = {}
3972

    
3973
  def Exec(self, feedback_fn):
3974
    return self.qcls.FieldsQuery(self.op.fields)
3975

    
3976

    
3977
class LUNodeModifyStorage(NoHooksLU):
3978
  """Logical unit for modifying a storage volume on a node.
3979

3980
  """
3981
  REQ_BGL = False
3982

    
3983
  def CheckArguments(self):
3984
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3985

    
3986
    storage_type = self.op.storage_type
3987

    
3988
    try:
3989
      modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
3990
    except KeyError:
3991
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
3992
                                 " modified" % storage_type,
3993
                                 errors.ECODE_INVAL)
3994

    
3995
    diff = set(self.op.changes.keys()) - modifiable
3996
    if diff:
3997
      raise errors.OpPrereqError("The following fields can not be modified for"
3998
                                 " storage units of type '%s': %r" %
3999
                                 (storage_type, list(diff)),
4000
                                 errors.ECODE_INVAL)
4001

    
4002
  def ExpandNames(self):
4003
    self.needed_locks = {
4004
      locking.LEVEL_NODE: self.op.node_name,
4005
      }
4006

    
4007
  def Exec(self, feedback_fn):
4008
    """Computes the list of nodes and their attributes.
4009

4010
    """
4011
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
4012
    result = self.rpc.call_storage_modify(self.op.node_name,
4013
                                          self.op.storage_type, st_args,
4014
                                          self.op.name, self.op.changes)
4015
    result.Raise("Failed to modify storage unit '%s' on %s" %
4016
                 (self.op.name, self.op.node_name))
4017

    
4018

    
4019
class LUNodeAdd(LogicalUnit):
4020
  """Logical unit for adding node to the cluster.
4021

4022
  """
4023
  HPATH = "node-add"
4024
  HTYPE = constants.HTYPE_NODE
4025
  _NFLAGS = ["master_capable", "vm_capable"]
4026

    
4027
  def CheckArguments(self):
4028
    self.primary_ip_family = self.cfg.GetPrimaryIPFamily()
4029
    # validate/normalize the node name
4030
    self.hostname = netutils.GetHostname(name=self.op.node_name,
4031
                                         family=self.primary_ip_family)
4032
    self.op.node_name = self.hostname.name
4033
    if self.op.readd and self.op.group:
4034
      raise errors.OpPrereqError("Cannot pass a node group when a node is"
4035
                                 " being readded", errors.ECODE_INVAL)
4036

    
4037
  def BuildHooksEnv(self):
4038
    """Build hooks env.
4039

4040
    This will run on all nodes before, and on all nodes + the new node after.
4041

4042
    """
4043
    env = {
4044
      "OP_TARGET": self.op.node_name,
4045
      "NODE_NAME": self.op.node_name,
4046
      "NODE_PIP": self.op.primary_ip,
4047
      "NODE_SIP": self.op.secondary_ip,
4048
      "MASTER_CAPABLE": str(self.op.master_capable),
4049
      "VM_CAPABLE": str(self.op.vm_capable),
4050
      }
4051
    nodes_0 = self.cfg.GetNodeList()
4052
    nodes_1 = nodes_0 + [self.op.node_name, ]
4053
    return env, nodes_0, nodes_1
4054

    
4055
  def CheckPrereq(self):
4056
    """Check prerequisites.
4057

4058
    This checks:
4059
     - the new node is not already in the config
4060
     - it is resolvable
4061
     - its parameters (single/dual homed) matches the cluster
4062

4063
    Any errors are signaled by raising errors.OpPrereqError.
4064

4065
    """
4066
    cfg = self.cfg
4067
    hostname = self.hostname
4068
    node = hostname.name
4069
    primary_ip = self.op.primary_ip = hostname.ip
4070
    if self.op.secondary_ip is None:
4071
      if self.primary_ip_family == netutils.IP6Address.family:
4072
        raise errors.OpPrereqError("When using a IPv6 primary address, a valid"
4073
                                   " IPv4 address must be given as secondary",
4074