Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ 9d5d703e

History | View | Annotate | Download (413.4 kB)

1
#
2
#
3

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

    
21

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

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

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

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

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

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

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

    
63

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

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

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

    
76

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

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

90
  Note that all commands require root permissions.
91

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

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

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

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

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

    
133
    # Tasklets
134
    self.tasklets = None
135

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

    
139
    self.CheckArguments()
140

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

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

    
149
  ssh = property(fget=__GetSSH)
150

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

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

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

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

166
    """
167
    pass
168

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

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

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

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

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

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

194
    Examples::
195

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

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

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

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

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

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

233
    """
234

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

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

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

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

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

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

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

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

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

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

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

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

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

292
    """
293
    raise NotImplementedError
294

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
381
    del self.recalculate_locks[locking.LEVEL_NODE]
382

    
383

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

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

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

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

397
    This just raises an error.
398

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

    
402

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

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

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

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

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

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

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

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

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

435
    """
436
    pass
437

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

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

445
    """
446
    raise NotImplementedError
447

    
448

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

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

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

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

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

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

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

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

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

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

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

    
492
    # Return expanded names
493
    return self.wanted
494

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

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

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

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

507
    See L{LogicalUnit.ExpandNames}.
508

509
    """
510
    raise NotImplementedError()
511

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

515
    See L{LogicalUnit.DeclareLocks}.
516

517
    """
518
    raise NotImplementedError()
519

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

523
    @return: Query data object
524

525
    """
526
    raise NotImplementedError()
527

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

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

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

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

    
540

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

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

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

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

    
558

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

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

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

    
578

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

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

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

    
611

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

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

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

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

    
630

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

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

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

    
645

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

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

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

    
660

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

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

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

    
673

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

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

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

    
686

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

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

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

    
704

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

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

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

    
731

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

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

    
739

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

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

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

    
755

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

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

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

    
772

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

    
777

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

    
782

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

788
  This builds the hook environment from individual variables.
789

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

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

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

    
852
  env["INSTANCE_NIC_COUNT"] = nic_count
853

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

    
862
  env["INSTANCE_DISK_COUNT"] = disk_count
863

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

    
868
  return env
869

    
870

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

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

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

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

    
894

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

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

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

    
932

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

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

    
948

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

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

    
959

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

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

    
973

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

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

    
982

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

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

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

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

    
1002

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

    
1006

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

1010
  """
1011

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

    
1014

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

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

    
1022

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

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

    
1030

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

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

    
1040
  return []
1041

    
1042

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

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

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

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

    
1057
  return faulty
1058

    
1059

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

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

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

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

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

    
1091

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

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

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

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

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

1110
    """
1111
    return True
1112

    
1113

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

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

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

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

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

1131
    This checks whether the cluster is empty.
1132

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

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

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

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

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

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

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

    
1166
    return master
1167

    
1168

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

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

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

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

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

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

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

    
1201

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
1412
    return True
1413

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
1646
  def _VerifyNodeFiles(self, ninfo, nresult, file_list, local_cksum,
1647
                       master_files):
1648
    """Verifies and computes the node required file checksums.
1649

1650
    @type ninfo: L{objects.Node}
1651
    @param ninfo: the node to check
1652
    @param nresult: the remote results for the node
1653
    @param file_list: required list of files
1654
    @param local_cksum: dictionary of local files and their checksums
1655
    @param master_files: list of files that only masters should have
1656

1657
    """
1658
    node = ninfo.name
1659
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1660

    
1661
    remote_cksum = nresult.get(constants.NV_FILELIST, None)
1662
    test = not isinstance(remote_cksum, dict)
1663
    _ErrorIf(test, self.ENODEFILECHECK, node,
1664
             "node hasn't returned file checksum data")
1665
    if test:
1666
      return
1667

    
1668
    for file_name in file_list:
1669
      node_is_mc = ninfo.master_candidate
1670
      must_have = (file_name not in master_files) or node_is_mc
1671
      # missing
1672
      test1 = file_name not in remote_cksum
1673
      # invalid checksum
1674
      test2 = not test1 and remote_cksum[file_name] != local_cksum[file_name]
1675
      # existing and good
1676
      test3 = not test1 and remote_cksum[file_name] == local_cksum[file_name]
1677
      _ErrorIf(test1 and must_have, self.ENODEFILECHECK, node,
1678
               "file '%s' missing", file_name)
1679
      _ErrorIf(test2 and must_have, self.ENODEFILECHECK, node,
1680
               "file '%s' has wrong checksum", file_name)
1681
      # not candidate and this is not a must-have file
1682
      _ErrorIf(test2 and not must_have, self.ENODEFILECHECK, node,
1683
               "file '%s' should not exist on non master"
1684
               " candidates (and the file is outdated)", file_name)
1685
      # all good, except non-master/non-must have combination
1686
      _ErrorIf(test3 and not must_have, self.ENODEFILECHECK, node,
1687
               "file '%s' should not exist"
1688
               " on non master candidates", file_name)
1689

    
1690
  def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper,
1691
                      drbd_map):
1692
    """Verifies and the node DRBD status.
1693

1694
    @type ninfo: L{objects.Node}
1695
    @param ninfo: the node to check
1696
    @param nresult: the remote results for the node
1697
    @param instanceinfo: the dict of instances
1698
    @param drbd_helper: the configured DRBD usermode helper
1699
    @param drbd_map: the DRBD map as returned by
1700
        L{ganeti.config.ConfigWriter.ComputeDRBDMap}
1701

1702
    """
1703
    node = ninfo.name
1704
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1705

    
1706
    if drbd_helper:
1707
      helper_result = nresult.get(constants.NV_DRBDHELPER, None)
1708
      test = (helper_result == None)
1709
      _ErrorIf(test, self.ENODEDRBDHELPER, node,
1710
               "no drbd usermode helper returned")
1711
      if helper_result:
1712
        status, payload = helper_result
1713
        test = not status
1714
        _ErrorIf(test, self.ENODEDRBDHELPER, node,
1715
                 "drbd usermode helper check unsuccessful: %s", payload)
1716
        test = status and (payload != drbd_helper)
1717
        _ErrorIf(test, self.ENODEDRBDHELPER, node,
1718
                 "wrong drbd usermode helper: %s", payload)
1719

    
1720
    # compute the DRBD minors
1721
    node_drbd = {}
1722
    for minor, instance in drbd_map[node].items():
1723
      test = instance not in instanceinfo
1724
      _ErrorIf(test, self.ECLUSTERCFG, None,
1725
               "ghost instance '%s' in temporary DRBD map", instance)
1726
        # ghost instance should not be running, but otherwise we
1727
        # don't give double warnings (both ghost instance and
1728
        # unallocated minor in use)
1729
      if test:
1730
        node_drbd[minor] = (instance, False)
1731
      else:
1732
        instance = instanceinfo[instance]
1733
        node_drbd[minor] = (instance.name, instance.admin_up)
1734

    
1735
    # and now check them
1736
    used_minors = nresult.get(constants.NV_DRBDLIST, [])
1737
    test = not isinstance(used_minors, (tuple, list))
1738
    _ErrorIf(test, self.ENODEDRBD, node,
1739
             "cannot parse drbd status file: %s", str(used_minors))
1740
    if test:
1741
      # we cannot check drbd status
1742
      return
1743

    
1744
    for minor, (iname, must_exist) in node_drbd.items():
1745
      test = minor not in used_minors and must_exist
1746
      _ErrorIf(test, self.ENODEDRBD, node,
1747
               "drbd minor %d of instance %s is not active", minor, iname)
1748
    for minor in used_minors:
1749
      test = minor not in node_drbd
1750
      _ErrorIf(test, self.ENODEDRBD, node,
1751
               "unallocated drbd minor %d is in use", minor)
1752

    
1753
  def _UpdateNodeOS(self, ninfo, nresult, nimg):
1754
    """Builds the node OS structures.
1755

1756
    @type ninfo: L{objects.Node}
1757
    @param ninfo: the node to check
1758
    @param nresult: the remote results for the node
1759
    @param nimg: the node image object
1760

1761
    """
1762
    node = ninfo.name
1763
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1764

    
1765
    remote_os = nresult.get(constants.NV_OSLIST, None)
1766
    test = (not isinstance(remote_os, list) or
1767
            not compat.all(isinstance(v, list) and len(v) == 7
1768
                           for v in remote_os))
1769

    
1770
    _ErrorIf(test, self.ENODEOS, node,
1771
             "node hasn't returned valid OS data")
1772

    
1773
    nimg.os_fail = test
1774

    
1775
    if test:
1776
      return
1777

    
1778
    os_dict = {}
1779

    
1780
    for (name, os_path, status, diagnose,
1781
         variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
1782

    
1783
      if name not in os_dict:
1784
        os_dict[name] = []
1785

    
1786
      # parameters is a list of lists instead of list of tuples due to
1787
      # JSON lacking a real tuple type, fix it:
1788
      parameters = [tuple(v) for v in parameters]
1789
      os_dict[name].append((os_path, status, diagnose,
1790
                            set(variants), set(parameters), set(api_ver)))
1791

    
1792
    nimg.oslist = os_dict
1793

    
1794
  def _VerifyNodeOS(self, ninfo, nimg, base):
1795
    """Verifies the node OS list.
1796

1797
    @type ninfo: L{objects.Node}
1798
    @param ninfo: the node to check
1799
    @param nimg: the node image object
1800
    @param base: the 'template' node we match against (e.g. from the master)
1801

1802
    """
1803
    node = ninfo.name
1804
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1805

    
1806
    assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
1807

    
1808
    for os_name, os_data in nimg.oslist.items():
1809
      assert os_data, "Empty OS status for OS %s?!" % os_name
1810
      f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
1811
      _ErrorIf(not f_status, self.ENODEOS, node,
1812
               "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
1813
      _ErrorIf(len(os_data) > 1, self.ENODEOS, node,
1814
               "OS '%s' has multiple entries (first one shadows the rest): %s",
1815
               os_name, utils.CommaJoin([v[0] for v in os_data]))
1816
      # this will catched in backend too
1817
      _ErrorIf(compat.any(v >= constants.OS_API_V15 for v in f_api)
1818
               and not f_var, self.ENODEOS, node,
1819
               "OS %s with API at least %d does not declare any variant",
1820
               os_name, constants.OS_API_V15)
1821
      # comparisons with the 'base' image
1822
      test = os_name not in base.oslist
1823
      _ErrorIf(test, self.ENODEOS, node,
1824
               "Extra OS %s not present on reference node (%s)",
1825
               os_name, base.name)
1826
      if test:
1827
        continue
1828
      assert base.oslist[os_name], "Base node has empty OS status?"
1829
      _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
1830
      if not b_status:
1831
        # base OS is invalid, skipping
1832
        continue
1833
      for kind, a, b in [("API version", f_api, b_api),
1834
                         ("variants list", f_var, b_var),
1835
                         ("parameters", f_param, b_param)]:
1836
        _ErrorIf(a != b, self.ENODEOS, node,
1837
                 "OS %s %s differs from reference node %s: %s vs. %s",
1838
                 kind, os_name, base.name,
1839
                 utils.CommaJoin(a), utils.CommaJoin(b))
1840

    
1841
    # check any missing OSes
1842
    missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
1843
    _ErrorIf(missing, self.ENODEOS, node,
1844
             "OSes present on reference node %s but missing on this node: %s",
1845
             base.name, utils.CommaJoin(missing))
1846

    
1847
  def _VerifyOob(self, ninfo, nresult):
1848
    """Verifies out of band functionality of a node.
1849

1850
    @type ninfo: L{objects.Node}
1851
    @param ninfo: the node to check
1852
    @param nresult: the remote results for the node
1853

1854
    """
1855
    node = ninfo.name
1856
    # We just have to verify the paths on master and/or master candidates
1857
    # as the oob helper is invoked on the master
1858
    if ((ninfo.master_candidate or ninfo.master_capable) and
1859
        constants.NV_OOB_PATHS in nresult):
1860
      for path_result in nresult[constants.NV_OOB_PATHS]:
1861
        self._ErrorIf(path_result, self.ENODEOOBPATH, node, path_result)
1862

    
1863
  def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
1864
    """Verifies and updates the node volume data.
1865

1866
    This function will update a L{NodeImage}'s internal structures
1867
    with data from the remote call.
1868

1869
    @type ninfo: L{objects.Node}
1870
    @param ninfo: the node to check
1871
    @param nresult: the remote results for the node
1872
    @param nimg: the node image object
1873
    @param vg_name: the configured VG name
1874

1875
    """
1876
    node = ninfo.name
1877
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1878

    
1879
    nimg.lvm_fail = True
1880
    lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
1881
    if vg_name is None:
1882
      pass
1883
    elif isinstance(lvdata, basestring):
1884
      _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
1885
               utils.SafeEncode(lvdata))
1886
    elif not isinstance(lvdata, dict):
1887
      _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
1888
    else:
1889
      nimg.volumes = lvdata
1890
      nimg.lvm_fail = False
1891

    
1892
  def _UpdateNodeInstances(self, ninfo, nresult, nimg):
1893
    """Verifies and updates the node instance list.
1894

1895
    If the listing was successful, then updates this node's instance
1896
    list. Otherwise, it marks the RPC call as failed for the instance
1897
    list key.
1898

1899
    @type ninfo: L{objects.Node}
1900
    @param ninfo: the node to check
1901
    @param nresult: the remote results for the node
1902
    @param nimg: the node image object
1903

1904
    """
1905
    idata = nresult.get(constants.NV_INSTANCELIST, None)
1906
    test = not isinstance(idata, list)
1907
    self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed"
1908
                  " (instancelist): %s", utils.SafeEncode(str(idata)))
1909
    if test:
1910
      nimg.hyp_fail = True
1911
    else:
1912
      nimg.instances = idata
1913

    
1914
  def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
1915
    """Verifies and computes a node information map
1916

1917
    @type ninfo: L{objects.Node}
1918
    @param ninfo: the node to check
1919
    @param nresult: the remote results for the node
1920
    @param nimg: the node image object
1921
    @param vg_name: the configured VG name
1922

1923
    """
1924
    node = ninfo.name
1925
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1926

    
1927
    # try to read free memory (from the hypervisor)
1928
    hv_info = nresult.get(constants.NV_HVINFO, None)
1929
    test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
1930
    _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
1931
    if not test:
1932
      try:
1933
        nimg.mfree = int(hv_info["memory_free"])
1934
      except (ValueError, TypeError):
1935
        _ErrorIf(True, self.ENODERPC, node,
1936
                 "node returned invalid nodeinfo, check hypervisor")
1937

    
1938
    # FIXME: devise a free space model for file based instances as well
1939
    if vg_name is not None:
1940
      test = (constants.NV_VGLIST not in nresult or
1941
              vg_name not in nresult[constants.NV_VGLIST])
1942
      _ErrorIf(test, self.ENODELVM, node,
1943
               "node didn't return data for the volume group '%s'"
1944
               " - it is either missing or broken", vg_name)
1945
      if not test:
1946
        try:
1947
          nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
1948
        except (ValueError, TypeError):
1949
          _ErrorIf(True, self.ENODERPC, node,
1950
                   "node returned invalid LVM info, check LVM status")
1951

    
1952
  def _CollectDiskInfo(self, nodelist, node_image, instanceinfo):
1953
    """Gets per-disk status information for all instances.
1954

1955
    @type nodelist: list of strings
1956
    @param nodelist: Node names
1957
    @type node_image: dict of (name, L{objects.Node})
1958
    @param node_image: Node objects
1959
    @type instanceinfo: dict of (name, L{objects.Instance})
1960
    @param instanceinfo: Instance objects
1961
    @rtype: {instance: {node: [(succes, payload)]}}
1962
    @return: a dictionary of per-instance dictionaries with nodes as
1963
        keys and disk information as values; the disk information is a
1964
        list of tuples (success, payload)
1965

1966
    """
1967
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1968

    
1969
    node_disks = {}
1970
    node_disks_devonly = {}
1971
    diskless_instances = set()
1972
    diskless = constants.DT_DISKLESS
1973

    
1974
    for nname in nodelist:
1975
      node_instances = list(itertools.chain(node_image[nname].pinst,
1976
                                            node_image[nname].sinst))
1977
      diskless_instances.update(inst for inst in node_instances
1978
                                if instanceinfo[inst].disk_template == diskless)
1979
      disks = [(inst, disk)
1980
               for inst in node_instances
1981
               for disk in instanceinfo[inst].disks]
1982

    
1983
      if not disks:
1984
        # No need to collect data
1985
        continue
1986

    
1987
      node_disks[nname] = disks
1988

    
1989
      # Creating copies as SetDiskID below will modify the objects and that can
1990
      # lead to incorrect data returned from nodes
1991
      devonly = [dev.Copy() for (_, dev) in disks]
1992

    
1993
      for dev in devonly:
1994
        self.cfg.SetDiskID(dev, nname)
1995

    
1996
      node_disks_devonly[nname] = devonly
1997

    
1998
    assert len(node_disks) == len(node_disks_devonly)
1999

    
2000
    # Collect data from all nodes with disks
2001
    result = self.rpc.call_blockdev_getmirrorstatus_multi(node_disks.keys(),
2002
                                                          node_disks_devonly)
2003

    
2004
    assert len(result) == len(node_disks)
2005

    
2006
    instdisk = {}
2007

    
2008
    for (nname, nres) in result.items():
2009
      disks = node_disks[nname]
2010

    
2011
      if nres.offline:
2012
        # No data from this node
2013
        data = len(disks) * [(False, "node offline")]
2014
      else:
2015
        msg = nres.fail_msg
2016
        _ErrorIf(msg, self.ENODERPC, nname,
2017
                 "while getting disk information: %s", msg)
2018
        if msg:
2019
          # No data from this node
2020
          data = len(disks) * [(False, msg)]
2021
        else:
2022
          data = []
2023
          for idx, i in enumerate(nres.payload):
2024
            if isinstance(i, (tuple, list)) and len(i) == 2:
2025
              data.append(i)
2026
            else:
2027
              logging.warning("Invalid result from node %s, entry %d: %s",
2028
                              nname, idx, i)
2029
              data.append((False, "Invalid result from the remote node"))
2030

    
2031
      for ((inst, _), status) in zip(disks, data):
2032
        instdisk.setdefault(inst, {}).setdefault(nname, []).append(status)
2033

    
2034
    # Add empty entries for diskless instances.
2035
    for inst in diskless_instances:
2036
      assert inst not in instdisk
2037
      instdisk[inst] = {}
2038

    
2039
    assert compat.all(len(statuses) == len(instanceinfo[inst].disks) and
2040
                      len(nnames) <= len(instanceinfo[inst].all_nodes) and
2041
                      compat.all(isinstance(s, (tuple, list)) and
2042
                                 len(s) == 2 for s in statuses)
2043
                      for inst, nnames in instdisk.items()
2044
                      for nname, statuses in nnames.items())
2045
    assert set(instdisk) == set(instanceinfo), "instdisk consistency failure"
2046

    
2047
    return instdisk
2048

    
2049
  def _VerifyHVP(self, hvp_data):
2050
    """Verifies locally the syntax of the hypervisor parameters.
2051

2052
    """
2053
    for item, hv_name, hv_params in hvp_data:
2054
      msg = ("hypervisor %s parameters syntax check (source %s): %%s" %
2055
             (item, hv_name))
2056
      try:
2057
        hv_class = hypervisor.GetHypervisor(hv_name)
2058
        utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2059
        hv_class.CheckParameterSyntax(hv_params)
2060
      except errors.GenericError, err:
2061
        self._ErrorIf(True, self.ECLUSTERCFG, None, msg % str(err))
2062

    
2063

    
2064
  def BuildHooksEnv(self):
2065
    """Build hooks env.
2066

2067
    Cluster-Verify hooks just ran in the post phase and their failure makes
2068
    the output be logged in the verify output and the verification to fail.
2069

2070
    """
2071
    all_nodes = self.cfg.GetNodeList()
2072
    env = {
2073
      "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
2074
      }
2075
    for node in self.cfg.GetAllNodesInfo().values():
2076
      env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
2077

    
2078
    return env, [], all_nodes
2079

    
2080
  def Exec(self, feedback_fn):
2081
    """Verify integrity of cluster, performing various test on nodes.
2082

2083
    """
2084
    # This method has too many local variables. pylint: disable-msg=R0914
2085
    self.bad = False
2086
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
2087
    verbose = self.op.verbose
2088
    self._feedback_fn = feedback_fn
2089
    feedback_fn("* Verifying global settings")
2090
    for msg in self.cfg.VerifyConfig():
2091
      _ErrorIf(True, self.ECLUSTERCFG, None, msg)
2092

    
2093
    # Check the cluster certificates
2094
    for cert_filename in constants.ALL_CERT_FILES:
2095
      (errcode, msg) = _VerifyCertificate(cert_filename)
2096
      _ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
2097

    
2098
    vg_name = self.cfg.GetVGName()
2099
    drbd_helper = self.cfg.GetDRBDHelper()
2100
    hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
2101
    cluster = self.cfg.GetClusterInfo()
2102
    nodelist = utils.NiceSort(self.cfg.GetNodeList())
2103
    nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
2104
    nodeinfo_byname = dict(zip(nodelist, nodeinfo))
2105
    instancelist = utils.NiceSort(self.cfg.GetInstanceList())
2106
    instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
2107
                        for iname in instancelist)
2108
    groupinfo = self.cfg.GetAllNodeGroupsInfo()
2109
    i_non_redundant = [] # Non redundant instances
2110
    i_non_a_balanced = [] # Non auto-balanced instances
2111
    n_offline = 0 # Count of offline nodes
2112
    n_drained = 0 # Count of nodes being drained
2113
    node_vol_should = {}
2114

    
2115
    # FIXME: verify OS list
2116
    # do local checksums
2117
    master_files = [constants.CLUSTER_CONF_FILE]
2118
    master_node = self.master_node = self.cfg.GetMasterNode()
2119
    master_ip = self.cfg.GetMasterIP()
2120

    
2121
    file_names = ssconf.SimpleStore().GetFileList()
2122
    file_names.extend(constants.ALL_CERT_FILES)
2123
    file_names.extend(master_files)
2124
    if cluster.modify_etc_hosts:
2125
      file_names.append(constants.ETC_HOSTS)
2126

    
2127
    local_checksums = utils.FingerprintFiles(file_names)
2128

    
2129
    # Compute the set of hypervisor parameters
2130
    hvp_data = []
2131
    for hv_name in hypervisors:
2132
      hvp_data.append(("cluster", hv_name, cluster.GetHVDefaults(hv_name)))
2133
    for os_name, os_hvp in cluster.os_hvp.items():
2134
      for hv_name, hv_params in os_hvp.items():
2135
        if not hv_params:
2136
          continue
2137
        full_params = cluster.GetHVDefaults(hv_name, os_name=os_name)
2138
        hvp_data.append(("os %s" % os_name, hv_name, full_params))
2139
    # TODO: collapse identical parameter values in a single one
2140
    for instance in instanceinfo.values():
2141
      if not instance.hvparams:
2142
        continue
2143
      hvp_data.append(("instance %s" % instance.name, instance.hypervisor,
2144
                       cluster.FillHV(instance)))
2145
    # and verify them locally
2146
    self._VerifyHVP(hvp_data)
2147

    
2148
    feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
2149
    node_verify_param = {
2150
      constants.NV_FILELIST: file_names,
2151
      constants.NV_NODELIST: [node.name for node in nodeinfo
2152
                              if not node.offline],
2153
      constants.NV_HYPERVISOR: hypervisors,
2154
      constants.NV_HVPARAMS: hvp_data,
2155
      constants.NV_NODENETTEST: [(node.name, node.primary_ip,
2156
                                  node.secondary_ip) for node in nodeinfo
2157
                                 if not node.offline],
2158
      constants.NV_INSTANCELIST: hypervisors,
2159
      constants.NV_VERSION: None,
2160
      constants.NV_HVINFO: self.cfg.GetHypervisorType(),
2161
      constants.NV_NODESETUP: None,
2162
      constants.NV_TIME: None,
2163
      constants.NV_MASTERIP: (master_node, master_ip),
2164
      constants.NV_OSLIST: None,
2165
      constants.NV_VMNODES: self.cfg.GetNonVmCapableNodeList(),
2166
      }
2167

    
2168
    if vg_name is not None:
2169
      node_verify_param[constants.NV_VGLIST] = None
2170
      node_verify_param[constants.NV_LVLIST] = vg_name
2171
      node_verify_param[constants.NV_PVLIST] = [vg_name]
2172
      node_verify_param[constants.NV_DRBDLIST] = None
2173

    
2174
    if drbd_helper:
2175
      node_verify_param[constants.NV_DRBDHELPER] = drbd_helper
2176

    
2177
    # Build our expected cluster state
2178
    node_image = dict((node.name, self.NodeImage(offline=node.offline,
2179
                                                 name=node.name,
2180
                                                 vm_capable=node.vm_capable))
2181
                      for node in nodeinfo)
2182

    
2183
    # Gather OOB paths
2184
    oob_paths = []
2185
    for node in nodeinfo:
2186
      path = _SupportsOob(self.cfg, node)
2187
      if path and path not in oob_paths:
2188
        oob_paths.append(path)
2189

    
2190
    if oob_paths:
2191
      node_verify_param[constants.NV_OOB_PATHS] = oob_paths
2192

    
2193
    for instance in instancelist:
2194
      inst_config = instanceinfo[instance]
2195

    
2196
      for nname in inst_config.all_nodes:
2197
        if nname not in node_image:
2198
          # ghost node
2199
          gnode = self.NodeImage(name=nname)
2200
          gnode.ghost = True
2201
          node_image[nname] = gnode
2202

    
2203
      inst_config.MapLVsByNode(node_vol_should)
2204

    
2205
      pnode = inst_config.primary_node
2206
      node_image[pnode].pinst.append(instance)
2207

    
2208
      for snode in inst_config.secondary_nodes:
2209
        nimg = node_image[snode]
2210
        nimg.sinst.append(instance)
2211
        if pnode not in nimg.sbp:
2212
          nimg.sbp[pnode] = []
2213
        nimg.sbp[pnode].append(instance)
2214

    
2215
    # At this point, we have the in-memory data structures complete,
2216
    # except for the runtime information, which we'll gather next
2217

    
2218
    # Due to the way our RPC system works, exact response times cannot be
2219
    # guaranteed (e.g. a broken node could run into a timeout). By keeping the
2220
    # time before and after executing the request, we can at least have a time
2221
    # window.
2222
    nvinfo_starttime = time.time()
2223
    all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
2224
                                           self.cfg.GetClusterName())
2225
    nvinfo_endtime = time.time()
2226

    
2227
    all_drbd_map = self.cfg.ComputeDRBDMap()
2228

    
2229
    feedback_fn("* Gathering disk information (%s nodes)" % len(nodelist))
2230
    instdisk = self._CollectDiskInfo(nodelist, node_image, instanceinfo)
2231

    
2232
    feedback_fn("* Verifying node status")
2233

    
2234
    refos_img = None
2235

    
2236
    for node_i in nodeinfo:
2237
      node = node_i.name
2238
      nimg = node_image[node]
2239

    
2240
      if node_i.offline:
2241
        if verbose:
2242
          feedback_fn("* Skipping offline node %s" % (node,))
2243
        n_offline += 1
2244
        continue
2245

    
2246
      if node == master_node:
2247
        ntype = "master"
2248
      elif node_i.master_candidate:
2249
        ntype = "master candidate"
2250
      elif node_i.drained:
2251
        ntype = "drained"
2252
        n_drained += 1
2253
      else:
2254
        ntype = "regular"
2255
      if verbose:
2256
        feedback_fn("* Verifying node %s (%s)" % (node, ntype))
2257

    
2258
      msg = all_nvinfo[node].fail_msg
2259
      _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
2260
      if msg:
2261
        nimg.rpc_fail = True
2262
        continue
2263

    
2264
      nresult = all_nvinfo[node].payload
2265

    
2266
      nimg.call_ok = self._VerifyNode(node_i, nresult)
2267
      self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
2268
      self._VerifyNodeNetwork(node_i, nresult)
2269
      self._VerifyNodeFiles(node_i, nresult, file_names, local_checksums,
2270
                            master_files)
2271

    
2272
      self._VerifyOob(node_i, nresult)
2273

    
2274
      if nimg.vm_capable:
2275
        self._VerifyNodeLVM(node_i, nresult, vg_name)
2276
        self._VerifyNodeDrbd(node_i, nresult, instanceinfo, drbd_helper,
2277
                             all_drbd_map)
2278

    
2279
        self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
2280
        self._UpdateNodeInstances(node_i, nresult, nimg)
2281
        self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
2282
        self._UpdateNodeOS(node_i, nresult, nimg)
2283
        if not nimg.os_fail:
2284
          if refos_img is None:
2285
            refos_img = nimg
2286
          self._VerifyNodeOS(node_i, nimg, refos_img)
2287

    
2288
    feedback_fn("* Verifying instance status")
2289
    for instance in instancelist:
2290
      if verbose:
2291
        feedback_fn("* Verifying instance %s" % instance)
2292
      inst_config = instanceinfo[instance]
2293
      self._VerifyInstance(instance, inst_config, node_image,
2294
                           instdisk[instance])
2295
      inst_nodes_offline = []
2296

    
2297
      pnode = inst_config.primary_node
2298
      pnode_img = node_image[pnode]
2299
      _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
2300
               self.ENODERPC, pnode, "instance %s, connection to"
2301
               " primary node failed", instance)
2302

    
2303
      _ErrorIf(pnode_img.offline, self.EINSTANCEBADNODE, instance,
2304
               "instance lives on offline node %s", inst_config.primary_node)
2305

    
2306
      # If the instance is non-redundant we cannot survive losing its primary
2307
      # node, so we are not N+1 compliant. On the other hand we have no disk
2308
      # templates with more than one secondary so that situation is not well
2309
      # supported either.
2310
      # FIXME: does not support file-backed instances
2311
      if not inst_config.secondary_nodes:
2312
        i_non_redundant.append(instance)
2313

    
2314
      _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT,
2315
               instance, "instance has multiple secondary nodes: %s",
2316
               utils.CommaJoin(inst_config.secondary_nodes),
2317
               code=self.ETYPE_WARNING)
2318

    
2319
      if inst_config.disk_template in constants.DTS_INT_MIRROR:
2320
        pnode = inst_config.primary_node
2321
        instance_nodes = utils.NiceSort(inst_config.all_nodes)
2322
        instance_groups = {}
2323

    
2324
        for node in instance_nodes:
2325
          instance_groups.setdefault(nodeinfo_byname[node].group,
2326
                                     []).append(node)
2327

    
2328
        pretty_list = [
2329
          "%s (group %s)" % (utils.CommaJoin(nodes), groupinfo[group].name)
2330
          # Sort so that we always list the primary node first.
2331
          for group, nodes in sorted(instance_groups.items(),
2332
                                     key=lambda (_, nodes): pnode in nodes,
2333
                                     reverse=True)]
2334

    
2335
        self._ErrorIf(len(instance_groups) > 1, self.EINSTANCESPLITGROUPS,
2336
                      instance, "instance has primary and secondary nodes in"
2337
                      " different groups: %s", utils.CommaJoin(pretty_list),
2338
                      code=self.ETYPE_WARNING)
2339

    
2340
      if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
2341
        i_non_a_balanced.append(instance)
2342

    
2343
      for snode in inst_config.secondary_nodes:
2344
        s_img = node_image[snode]
2345
        _ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode,
2346
                 "instance %s, connection to secondary node failed", instance)
2347

    
2348
        if s_img.offline:
2349
          inst_nodes_offline.append(snode)
2350

    
2351
      # warn that the instance lives on offline nodes
2352
      _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
2353
               "instance has offline secondary node(s) %s",
2354
               utils.CommaJoin(inst_nodes_offline))
2355
      # ... or ghost/non-vm_capable nodes
2356
      for node in inst_config.all_nodes:
2357
        _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
2358
                 "instance lives on ghost node %s", node)
2359
        _ErrorIf(not node_image[node].vm_capable, self.EINSTANCEBADNODE,
2360
                 instance, "instance lives on non-vm_capable node %s", node)
2361

    
2362
    feedback_fn("* Verifying orphan volumes")
2363
    reserved = utils.FieldSet(*cluster.reserved_lvs)
2364
    self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
2365

    
2366
    feedback_fn("* Verifying orphan instances")
2367
    self._VerifyOrphanInstances(instancelist, node_image)
2368

    
2369
    if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
2370
      feedback_fn("* Verifying N+1 Memory redundancy")
2371
      self._VerifyNPlusOneMemory(node_image, instanceinfo)
2372

    
2373
    feedback_fn("* Other Notes")
2374
    if i_non_redundant:
2375
      feedback_fn("  - NOTICE: %d non-redundant instance(s) found."
2376
                  % len(i_non_redundant))
2377

    
2378
    if i_non_a_balanced:
2379
      feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
2380
                  % len(i_non_a_balanced))
2381

    
2382
    if n_offline:
2383
      feedback_fn("  - NOTICE: %d offline node(s) found." % n_offline)
2384

    
2385
    if n_drained:
2386
      feedback_fn("  - NOTICE: %d drained node(s) found." % n_drained)
2387

    
2388
    return not self.bad
2389

    
2390
  def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
2391
    """Analyze the post-hooks' result
2392

2393
    This method analyses the hook result, handles it, and sends some
2394
    nicely-formatted feedback back to the user.
2395

2396
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
2397
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
2398
    @param hooks_results: the results of the multi-node hooks rpc call
2399
    @param feedback_fn: function used send feedback back to the caller
2400
    @param lu_result: previous Exec result
2401
    @return: the new Exec result, based on the previous result
2402
        and hook results
2403

2404
    """
2405
    # We only really run POST phase hooks, and are only interested in
2406
    # their results
2407
    if phase == constants.HOOKS_PHASE_POST:
2408
      # Used to change hooks' output to proper indentation
2409
      feedback_fn("* Hooks Results")
2410
      assert hooks_results, "invalid result from hooks"
2411

    
2412
      for node_name in hooks_results:
2413
        res = hooks_results[node_name]
2414
        msg = res.fail_msg
2415
        test = msg and not res.offline
2416
        self._ErrorIf(test, self.ENODEHOOKS, node_name,
2417
                      "Communication failure in hooks execution: %s", msg)
2418
        if res.offline or msg:
2419
          # No need to investigate payload if node is offline or gave an error.
2420
          # override manually lu_result here as _ErrorIf only
2421
          # overrides self.bad
2422
          lu_result = 1
2423
          continue
2424
        for script, hkr, output in res.payload:
2425
          test = hkr == constants.HKR_FAIL
2426
          self._ErrorIf(test, self.ENODEHOOKS, node_name,
2427
                        "Script %s failed, output:", script)
2428
          if test:
2429
            output = self._HOOKS_INDENT_RE.sub('      ', output)
2430
            feedback_fn("%s" % output)
2431
            lu_result = 0
2432

    
2433
      return lu_result
2434

    
2435

    
2436
class LUClusterVerifyDisks(NoHooksLU):
2437
  """Verifies the cluster disks status.
2438

2439
  """
2440
  REQ_BGL = False
2441

    
2442
  def ExpandNames(self):
2443
    self.needed_locks = {
2444
      locking.LEVEL_NODE: locking.ALL_SET,
2445
      locking.LEVEL_INSTANCE: locking.ALL_SET,
2446
    }
2447
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
2448

    
2449
  def Exec(self, feedback_fn):
2450
    """Verify integrity of cluster disks.
2451

2452
    @rtype: tuple of three items
2453
    @return: a tuple of (dict of node-to-node_error, list of instances
2454
        which need activate-disks, dict of instance: (node, volume) for
2455
        missing volumes
2456

2457
    """
2458
    result = res_nodes, res_instances, res_missing = {}, [], {}
2459

    
2460
    nodes = utils.NiceSort(self.cfg.GetVmCapableNodeList())
2461
    instances = self.cfg.GetAllInstancesInfo().values()
2462

    
2463
    nv_dict = {}
2464
    for inst in instances:
2465
      inst_lvs = {}
2466
      if not inst.admin_up:
2467
        continue
2468
      inst.MapLVsByNode(inst_lvs)
2469
      # transform { iname: {node: [vol,],},} to {(node, vol): iname}
2470
      for node, vol_list in inst_lvs.iteritems():
2471
        for vol in vol_list:
2472
          nv_dict[(node, vol)] = inst
2473

    
2474
    if not nv_dict:
2475
      return result
2476

    
2477
    node_lvs = self.rpc.call_lv_list(nodes, [])
2478
    for node, node_res in node_lvs.items():
2479
      if node_res.offline:
2480
        continue
2481
      msg = node_res.fail_msg
2482
      if msg:
2483
        logging.warning("Error enumerating LVs on node %s: %s", node, msg)
2484
        res_nodes[node] = msg
2485
        continue
2486

    
2487
      lvs = node_res.payload
2488
      for lv_name, (_, _, lv_online) in lvs.items():
2489
        inst = nv_dict.pop((node, lv_name), None)
2490
        if (not lv_online and inst is not None
2491
            and inst.name not in res_instances):
2492
          res_instances.append(inst.name)
2493

    
2494
    # any leftover items in nv_dict are missing LVs, let's arrange the
2495
    # data better
2496
    for key, inst in nv_dict.iteritems():
2497
      if inst.name not in res_missing:
2498
        res_missing[inst.name] = []
2499
      res_missing[inst.name].append(key)
2500

    
2501
    return result
2502

    
2503

    
2504
class LUClusterRepairDiskSizes(NoHooksLU):
2505
  """Verifies the cluster disks sizes.
2506

2507
  """
2508
  REQ_BGL = False
2509

    
2510
  def ExpandNames(self):
2511
    if self.op.instances:
2512
      self.wanted_names = []
2513
      for name in self.op.instances:
2514
        full_name = _ExpandInstanceName(self.cfg, name)
2515
        self.wanted_names.append(full_name)
2516
      self.needed_locks = {
2517
        locking.LEVEL_NODE: [],
2518
        locking.LEVEL_INSTANCE: self.wanted_names,
2519
        }
2520
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2521
    else:
2522
      self.wanted_names = None
2523
      self.needed_locks = {
2524
        locking.LEVEL_NODE: locking.ALL_SET,
2525
        locking.LEVEL_INSTANCE: locking.ALL_SET,
2526
        }
2527
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
2528

    
2529
  def DeclareLocks(self, level):
2530
    if level == locking.LEVEL_NODE and self.wanted_names is not None:
2531
      self._LockInstancesNodes(primary_only=True)
2532

    
2533
  def CheckPrereq(self):
2534
    """Check prerequisites.
2535

2536
    This only checks the optional instance list against the existing names.
2537

2538
    """
2539
    if self.wanted_names is None:
2540
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
2541

    
2542
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
2543
                             in self.wanted_names]
2544

    
2545
  def _EnsureChildSizes(self, disk):
2546
    """Ensure children of the disk have the needed disk size.
2547

2548
    This is valid mainly for DRBD8 and fixes an issue where the
2549
    children have smaller disk size.
2550

2551
    @param disk: an L{ganeti.objects.Disk} object
2552

2553
    """
2554
    if disk.dev_type == constants.LD_DRBD8:
2555
      assert disk.children, "Empty children for DRBD8?"
2556
      fchild = disk.children[0]
2557
      mismatch = fchild.size < disk.size
2558
      if mismatch:
2559
        self.LogInfo("Child disk has size %d, parent %d, fixing",
2560
                     fchild.size, disk.size)
2561
        fchild.size = disk.size
2562

    
2563
      # and we recurse on this child only, not on the metadev
2564
      return self._EnsureChildSizes(fchild) or mismatch
2565
    else:
2566
      return False
2567

    
2568
  def Exec(self, feedback_fn):
2569
    """Verify the size of cluster disks.
2570

2571
    """
2572
    # TODO: check child disks too
2573
    # TODO: check differences in size between primary/secondary nodes
2574
    per_node_disks = {}
2575
    for instance in self.wanted_instances:
2576
      pnode = instance.primary_node
2577
      if pnode not in per_node_disks:
2578
        per_node_disks[pnode] = []
2579
      for idx, disk in enumerate(instance.disks):
2580
        per_node_disks[pnode].append((instance, idx, disk))
2581

    
2582
    changed = []
2583
    for node, dskl in per_node_disks.items():
2584
      newl = [v[2].Copy() for v in dskl]
2585
      for dsk in newl:
2586
        self.cfg.SetDiskID(dsk, node)
2587
      result = self.rpc.call_blockdev_getsize(node, newl)
2588
      if result.fail_msg:
2589
        self.LogWarning("Failure in blockdev_getsize call to node"
2590
                        " %s, ignoring", node)
2591
        continue
2592
      if len(result.payload) != len(dskl):
2593
        logging.warning("Invalid result from node %s: len(dksl)=%d,"
2594
                        " result.payload=%s", node, len(dskl), result.payload)
2595
        self.LogWarning("Invalid result from node %s, ignoring node results",
2596
                        node)
2597
        continue
2598
      for ((instance, idx, disk), size) in zip(dskl, result.payload):
2599
        if size is None:
2600
          self.LogWarning("Disk %d of instance %s did not return size"
2601
                          " information, ignoring", idx, instance.name)
2602
          continue
2603
        if not isinstance(size, (int, long)):
2604
          self.LogWarning("Disk %d of instance %s did not return valid"
2605
                          " size information, ignoring", idx, instance.name)
2606
          continue
2607
        size = size >> 20
2608
        if size != disk.size:
2609
          self.LogInfo("Disk %d of instance %s has mismatched size,"
2610
                       " correcting: recorded %d, actual %d", idx,
2611
                       instance.name, disk.size, size)
2612
          disk.size = size
2613
          self.cfg.Update(instance, feedback_fn)
2614
          changed.append((instance.name, idx, size))
2615
        if self._EnsureChildSizes(disk):
2616
          self.cfg.Update(instance, feedback_fn)
2617
          changed.append((instance.name, idx, disk.size))
2618
    return changed
2619

    
2620

    
2621
class LUClusterRename(LogicalUnit):
2622
  """Rename the cluster.
2623

2624
  """
2625
  HPATH = "cluster-rename"
2626
  HTYPE = constants.HTYPE_CLUSTER
2627

    
2628
  def BuildHooksEnv(self):
2629
    """Build hooks env.
2630

2631
    """
2632
    env = {
2633
      "OP_TARGET": self.cfg.GetClusterName(),
2634
      "NEW_NAME": self.op.name,
2635
      }
2636
    mn = self.cfg.GetMasterNode()
2637
    all_nodes = self.cfg.GetNodeList()
2638
    return env, [mn], all_nodes
2639

    
2640
  def CheckPrereq(self):
2641
    """Verify that the passed name is a valid one.
2642

2643
    """
2644
    hostname = netutils.GetHostname(name=self.op.name,
2645
                                    family=self.cfg.GetPrimaryIPFamily())
2646

    
2647
    new_name = hostname.name
2648
    self.ip = new_ip = hostname.ip
2649
    old_name = self.cfg.GetClusterName()
2650
    old_ip = self.cfg.GetMasterIP()
2651
    if new_name == old_name and new_ip == old_ip:
2652
      raise errors.OpPrereqError("Neither the name nor the IP address of the"
2653
                                 " cluster has changed",
2654
                                 errors.ECODE_INVAL)
2655
    if new_ip != old_ip:
2656
      if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
2657
        raise errors.OpPrereqError("The given cluster IP address (%s) is"
2658
                                   " reachable on the network" %
2659
                                   new_ip, errors.ECODE_NOTUNIQUE)
2660

    
2661
    self.op.name = new_name
2662

    
2663
  def Exec(self, feedback_fn):
2664
    """Rename the cluster.
2665

2666
    """
2667
    clustername = self.op.name
2668
    ip = self.ip
2669

    
2670
    # shutdown the master IP
2671
    master = self.cfg.GetMasterNode()
2672
    result = self.rpc.call_node_stop_master(master, False)
2673
    result.Raise("Could not disable the master role")
2674

    
2675
    try:
2676
      cluster = self.cfg.GetClusterInfo()
2677
      cluster.cluster_name = clustername
2678
      cluster.master_ip = ip
2679
      self.cfg.Update(cluster, feedback_fn)
2680

    
2681
      # update the known hosts file
2682
      ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
2683
      node_list = self.cfg.GetOnlineNodeList()
2684
      try:
2685
        node_list.remove(master)
2686
      except ValueError:
2687
        pass
2688
      _UploadHelper(self, node_list, constants.SSH_KNOWN_HOSTS_FILE)
2689
    finally:
2690
      result = self.rpc.call_node_start_master(master, False, False)
2691
      msg = result.fail_msg
2692
      if msg:
2693
        self.LogWarning("Could not re-enable the master role on"
2694
                        " the master, please restart manually: %s", msg)
2695

    
2696
    return clustername
2697

    
2698

    
2699
class LUClusterSetParams(LogicalUnit):
2700
  """Change the parameters of the cluster.
2701

2702
  """
2703
  HPATH = "cluster-modify"
2704
  HTYPE = constants.HTYPE_CLUSTER
2705
  REQ_BGL = False
2706

    
2707
  def CheckArguments(self):
2708
    """Check parameters
2709

2710
    """
2711
    if self.op.uid_pool:
2712
      uidpool.CheckUidPool(self.op.uid_pool)
2713

    
2714
    if self.op.add_uids:
2715
      uidpool.CheckUidPool(self.op.add_uids)
2716

    
2717
    if self.op.remove_uids:
2718
      uidpool.CheckUidPool(self.op.remove_uids)
2719

    
2720
  def ExpandNames(self):
2721
    # FIXME: in the future maybe other cluster params won't require checking on
2722
    # all nodes to be modified.
2723
    self.needed_locks = {
2724
      locking.LEVEL_NODE: locking.ALL_SET,
2725
    }
2726
    self.share_locks[locking.LEVEL_NODE] = 1
2727

    
2728
  def BuildHooksEnv(self):
2729
    """Build hooks env.
2730

2731
    """
2732
    env = {
2733
      "OP_TARGET": self.cfg.GetClusterName(),
2734
      "NEW_VG_NAME": self.op.vg_name,
2735
      }
2736
    mn = self.cfg.GetMasterNode()
2737
    return env, [mn], [mn]
2738

    
2739
  def CheckPrereq(self):
2740
    """Check prerequisites.
2741

2742
    This checks whether the given params don't conflict and
2743
    if the given volume group is valid.
2744

2745
    """
2746
    if self.op.vg_name is not None and not self.op.vg_name:
2747
      if self.cfg.HasAnyDiskOfType(constants.LD_LV):
2748
        raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based"
2749
                                   " instances exist", errors.ECODE_INVAL)
2750

    
2751
    if self.op.drbd_helper is not None and not self.op.drbd_helper:
2752
      if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8):
2753
        raise errors.OpPrereqError("Cannot disable drbd helper while"
2754
                                   " drbd-based instances exist",
2755
                                   errors.ECODE_INVAL)
2756

    
2757
    node_list = self.acquired_locks[locking.LEVEL_NODE]
2758

    
2759
    # if vg_name not None, checks given volume group on all nodes
2760
    if self.op.vg_name:
2761
      vglist = self.rpc.call_vg_list(node_list)
2762
      for node in node_list:
2763
        msg = vglist[node].fail_msg
2764
        if msg:
2765
          # ignoring down node
2766
          self.LogWarning("Error while gathering data on node %s"
2767
                          " (ignoring node): %s", node, msg)
2768
          continue
2769
        vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
2770
                                              self.op.vg_name,
2771
                                              constants.MIN_VG_SIZE)
2772
        if vgstatus:
2773
          raise errors.OpPrereqError("Error on node '%s': %s" %
2774
                                     (node, vgstatus), errors.ECODE_ENVIRON)
2775

    
2776
    if self.op.drbd_helper:
2777
      # checks given drbd helper on all nodes
2778
      helpers = self.rpc.call_drbd_helper(node_list)
2779
      for node in node_list:
2780
        ninfo = self.cfg.GetNodeInfo(node)
2781
        if ninfo.offline:
2782
          self.LogInfo("Not checking drbd helper on offline node %s", node)
2783
          continue
2784
        msg = helpers[node].fail_msg
2785
        if msg:
2786
          raise errors.OpPrereqError("Error checking drbd helper on node"
2787
                                     " '%s': %s" % (node, msg),
2788
                                     errors.ECODE_ENVIRON)
2789
        node_helper = helpers[node].payload
2790
        if node_helper != self.op.drbd_helper:
2791
          raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" %
2792
                                     (node, node_helper), errors.ECODE_ENVIRON)
2793

    
2794
    self.cluster = cluster = self.cfg.GetClusterInfo()
2795
    # validate params changes
2796
    if self.op.beparams:
2797
      utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
2798
      self.new_beparams = cluster.SimpleFillBE(self.op.beparams)
2799

    
2800
    if self.op.ndparams:
2801
      utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
2802
      self.new_ndparams = cluster.SimpleFillND(self.op.ndparams)
2803

    
2804
      # TODO: we need a more general way to handle resetting
2805
      # cluster-level parameters to default values
2806
      if self.new_ndparams["oob_program"] == "":
2807
        self.new_ndparams["oob_program"] = \
2808
            constants.NDC_DEFAULTS[constants.ND_OOB_PROGRAM]
2809

    
2810
    if self.op.nicparams:
2811
      utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
2812
      self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams)
2813
      objects.NIC.CheckParameterSyntax(self.new_nicparams)
2814
      nic_errors = []
2815

    
2816
      # check all instances for consistency
2817
      for instance in self.cfg.GetAllInstancesInfo().values():
2818
        for nic_idx, nic in enumerate(instance.nics):
2819
          params_copy = copy.deepcopy(nic.nicparams)
2820
          params_filled = objects.FillDict(self.new_nicparams, params_copy)
2821

    
2822
          # check parameter syntax
2823
          try:
2824
            objects.NIC.CheckParameterSyntax(params_filled)
2825
          except errors.ConfigurationError, err:
2826
            nic_errors.append("Instance %s, nic/%d: %s" %
2827
                              (instance.name, nic_idx, err))
2828

    
2829
          # if we're moving instances to routed, check that they have an ip
2830
          target_mode = params_filled[constants.NIC_MODE]
2831
          if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
2832
            nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
2833
                              (instance.name, nic_idx))
2834
      if nic_errors:
2835
        raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
2836
                                   "\n".join(nic_errors))
2837

    
2838
    # hypervisor list/parameters
2839
    self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
2840
    if self.op.hvparams:
2841
      for hv_name, hv_dict in self.op.hvparams.items():
2842
        if hv_name not in self.new_hvparams:
2843
          self.new_hvparams[hv_name] = hv_dict
2844
        else:
2845
          self.new_hvparams[hv_name].update(hv_dict)
2846

    
2847
    # os hypervisor parameters
2848
    self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
2849
    if self.op.os_hvp:
2850
      for os_name, hvs in self.op.os_hvp.items():
2851
        if os_name not in self.new_os_hvp:
2852
          self.new_os_hvp[os_name] = hvs
2853
        else:
2854
          for hv_name, hv_dict in hvs.items():
2855
            if hv_name not in self.new_os_hvp[os_name]:
2856
              self.new_os_hvp[os_name][hv_name] = hv_dict
2857
            else:
2858
              self.new_os_hvp[os_name][hv_name].update(hv_dict)
2859

    
2860
    # os parameters
2861
    self.new_osp = objects.FillDict(cluster.osparams, {})
2862
    if self.op.osparams:
2863
      for os_name, osp in self.op.osparams.items():
2864
        if os_name not in self.new_osp:
2865
          self.new_osp[os_name] = {}
2866

    
2867
        self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp,
2868
                                                  use_none=True)
2869

    
2870
        if not self.new_osp[os_name]:
2871
          # we removed all parameters
2872
          del self.new_osp[os_name]
2873
        else:
2874
          # check the parameter validity (remote check)
2875
          _CheckOSParams(self, False, [self.cfg.GetMasterNode()],
2876
                         os_name, self.new_osp[os_name])
2877

    
2878
    # changes to the hypervisor list
2879
    if self.op.enabled_hypervisors is not None:
2880
      self.hv_list = self.op.enabled_hypervisors
2881
      for hv in self.hv_list:
2882
        # if the hypervisor doesn't already exist in the cluster
2883
        # hvparams, we initialize it to empty, and then (in both
2884
        # cases) we make sure to fill the defaults, as we might not
2885
        # have a complete defaults list if the hypervisor wasn't
2886
        # enabled before
2887
        if hv not in new_hvp:
2888
          new_hvp[hv] = {}
2889
        new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
2890
        utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
2891
    else:
2892
      self.hv_list = cluster.enabled_hypervisors
2893

    
2894
    if self.op.hvparams or self.op.enabled_hypervisors is not None:
2895
      # either the enabled list has changed, or the parameters have, validate
2896
      for hv_name, hv_params in self.new_hvparams.items():
2897
        if ((self.op.hvparams and hv_name in self.op.hvparams) or
2898
            (self.op.enabled_hypervisors and
2899
             hv_name in self.op.enabled_hypervisors)):
2900
          # either this is a new hypervisor, or its parameters have changed
2901
          hv_class = hypervisor.GetHypervisor(hv_name)
2902
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2903
          hv_class.CheckParameterSyntax(hv_params)
2904
          _CheckHVParams(self, node_list, hv_name, hv_params)
2905

    
2906
    if self.op.os_hvp:
2907
      # no need to check any newly-enabled hypervisors, since the
2908
      # defaults have already been checked in the above code-block
2909
      for os_name, os_hvp in self.new_os_hvp.items():
2910
        for hv_name, hv_params in os_hvp.items():
2911
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2912
          # we need to fill in the new os_hvp on top of the actual hv_p
2913
          cluster_defaults = self.new_hvparams.get(hv_name, {})
2914
          new_osp = objects.FillDict(cluster_defaults, hv_params)
2915
          hv_class = hypervisor.GetHypervisor(hv_name)
2916
          hv_class.CheckParameterSyntax(new_osp)
2917
          _CheckHVParams(self, node_list, hv_name, new_osp)
2918

    
2919
    if self.op.default_iallocator:
2920
      alloc_script = utils.FindFile(self.op.default_iallocator,
2921
                                    constants.IALLOCATOR_SEARCH_PATH,
2922
                                    os.path.isfile)
2923
      if alloc_script is None:
2924
        raise errors.OpPrereqError("Invalid default iallocator script '%s'"
2925
                                   " specified" % self.op.default_iallocator,
2926
                                   errors.ECODE_INVAL)
2927

    
2928
  def Exec(self, feedback_fn):
2929
    """Change the parameters of the cluster.
2930

2931
    """
2932
    if self.op.vg_name is not None:
2933
      new_volume = self.op.vg_name
2934
      if not new_volume:
2935
        new_volume = None
2936
      if new_volume != self.cfg.GetVGName():
2937
        self.cfg.SetVGName(new_volume)
2938
      else:
2939
        feedback_fn("Cluster LVM configuration already in desired"
2940
                    " state, not changing")
2941
    if self.op.drbd_helper is not None:
2942
      new_helper = self.op.drbd_helper
2943
      if not new_helper:
2944
        new_helper = None
2945
      if new_helper != self.cfg.GetDRBDHelper():
2946
        self.cfg.SetDRBDHelper(new_helper)
2947
      else:
2948
        feedback_fn("Cluster DRBD helper already in desired state,"
2949
                    " not changing")
2950
    if self.op.hvparams:
2951
      self.cluster.hvparams = self.new_hvparams
2952
    if self.op.os_hvp:
2953
      self.cluster.os_hvp = self.new_os_hvp
2954
    if self.op.enabled_hypervisors is not None:
2955
      self.cluster.hvparams = self.new_hvparams
2956
      self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
2957
    if self.op.beparams:
2958
      self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
2959
    if self.op.nicparams:
2960
      self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
2961
    if self.op.osparams:
2962
      self.cluster.osparams = self.new_osp
2963
    if self.op.ndparams:
2964
      self.cluster.ndparams = self.new_ndparams
2965

    
2966
    if self.op.candidate_pool_size is not None:
2967
      self.cluster.candidate_pool_size = self.op.candidate_pool_size
2968
      # we need to update the pool size here, otherwise the save will fail
2969
      _AdjustCandidatePool(self, [])
2970

    
2971
    if self.op.maintain_node_health is not None:
2972
      self.cluster.maintain_node_health = self.op.maintain_node_health
2973

    
2974
    if self.op.prealloc_wipe_disks is not None:
2975
      self.cluster.prealloc_wipe_disks = self.op.prealloc_wipe_disks
2976

    
2977
    if self.op.add_uids is not None:
2978
      uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
2979

    
2980
    if self.op.remove_uids is not None:
2981
      uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
2982

    
2983
    if self.op.uid_pool is not None:
2984
      self.cluster.uid_pool = self.op.uid_pool
2985

    
2986
    if self.op.default_iallocator is not None:
2987
      self.cluster.default_iallocator = self.op.default_iallocator
2988

    
2989
    if self.op.reserved_lvs is not None:
2990
      self.cluster.reserved_lvs = self.op.reserved_lvs
2991

    
2992
    def helper_os(aname, mods, desc):
2993
      desc += " OS list"
2994
      lst = getattr(self.cluster, aname)
2995
      for key, val in mods:
2996
        if key == constants.DDM_ADD:
2997
          if val in lst:
2998
            feedback_fn("OS %s already in %s, ignoring" % (val, desc))
2999
          else:
3000
            lst.append(val)
3001
        elif key == constants.DDM_REMOVE:
3002
          if val in lst:
3003
            lst.remove(val)
3004
          else:
3005
            feedback_fn("OS %s not found in %s, ignoring" % (val, desc))
3006
        else:
3007
          raise errors.ProgrammerError("Invalid modification '%s'" % key)
3008

    
3009
    if self.op.hidden_os:
3010
      helper_os("hidden_os", self.op.hidden_os, "hidden")
3011

    
3012
    if self.op.blacklisted_os:
3013
      helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted")
3014

    
3015
    if self.op.master_netdev:
3016
      master = self.cfg.GetMasterNode()
3017
      feedback_fn("Shutting down master ip on the current netdev (%s)" %
3018
                  self.cluster.master_netdev)
3019
      result = self.rpc.call_node_stop_master(master, False)
3020
      result.Raise("Could not disable the master ip")
3021
      feedback_fn("Changing master_netdev from %s to %s" %
3022
                  (self.cluster.master_netdev, self.op.master_netdev))
3023
      self.cluster.master_netdev = self.op.master_netdev
3024

    
3025
    self.cfg.Update(self.cluster, feedback_fn)
3026

    
3027
    if self.op.master_netdev:
3028
      feedback_fn("Starting the master ip on the new master netdev (%s)" %
3029
                  self.op.master_netdev)
3030
      result = self.rpc.call_node_start_master(master, False, False)
3031
      if result.fail_msg:
3032
        self.LogWarning("Could not re-enable the master ip on"
3033
                        " the master, please restart manually: %s",
3034
                        result.fail_msg)
3035

    
3036

    
3037
def _UploadHelper(lu, nodes, fname):
3038
  """Helper for uploading a file and showing warnings.
3039

3040
  """
3041
  if os.path.exists(fname):
3042
    result = lu.rpc.call_upload_file(nodes, fname)
3043
    for to_node, to_result in result.items():
3044
      msg = to_result.fail_msg
3045
      if msg:
3046
        msg = ("Copy of file %s to node %s failed: %s" %
3047
               (fname, to_node, msg))
3048
        lu.proc.LogWarning(msg)
3049

    
3050

    
3051
def _RedistributeAncillaryFiles(lu, additional_nodes=None, additional_vm=True):
3052
  """Distribute additional files which are part of the cluster configuration.
3053

3054
  ConfigWriter takes care of distributing the config and ssconf files, but
3055
  there are more files which should be distributed to all nodes. This function
3056
  makes sure those are copied.
3057

3058
  @param lu: calling logical unit
3059
  @param additional_nodes: list of nodes not in the config to distribute to
3060
  @type additional_vm: boolean
3061
  @param additional_vm: whether the additional nodes are vm-capable or not
3062

3063
  """
3064
  # 1. Gather target nodes
3065
  myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
3066
  dist_nodes = lu.cfg.GetOnlineNodeList()
3067
  nvm_nodes = lu.cfg.GetNonVmCapableNodeList()
3068
  vm_nodes = [name for name in dist_nodes if name not in nvm_nodes]
3069
  if additional_nodes is not None:
3070
    dist_nodes.extend(additional_nodes)
3071
    if additional_vm:
3072
      vm_nodes.extend(additional_nodes)
3073
  if myself.name in dist_nodes:
3074
    dist_nodes.remove(myself.name)
3075
  if myself.name in vm_nodes:
3076
    vm_nodes.remove(myself.name)
3077

    
3078
  # 2. Gather files to distribute
3079
  dist_files = set([constants.ETC_HOSTS,
3080
                    constants.SSH_KNOWN_HOSTS_FILE,
3081
                    constants.RAPI_CERT_FILE,
3082
                    constants.RAPI_USERS_FILE,
3083
                    constants.CONFD_HMAC_KEY,
3084
                    constants.CLUSTER_DOMAIN_SECRET_FILE,
3085
                   ])
3086

    
3087
  vm_files = set()
3088
  enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
3089
  for hv_name in enabled_hypervisors:
3090
    hv_class = hypervisor.GetHypervisor(hv_name)
3091
    vm_files.update(hv_class.GetAncillaryFiles())
3092

    
3093
  # 3. Perform the files upload
3094
  for fname in dist_files:
3095
    _UploadHelper(lu, dist_nodes, fname)
3096
  for fname in vm_files:
3097
    _UploadHelper(lu, vm_nodes, fname)
3098

    
3099

    
3100
class LUClusterRedistConf(NoHooksLU):
3101
  """Force the redistribution of cluster configuration.
3102

3103
  This is a very simple LU.
3104

3105
  """
3106
  REQ_BGL = False
3107

    
3108
  def ExpandNames(self):
3109
    self.needed_locks = {
3110
      locking.LEVEL_NODE: locking.ALL_SET,
3111
    }
3112
    self.share_locks[locking.LEVEL_NODE] = 1
3113

    
3114
  def Exec(self, feedback_fn):
3115
    """Redistribute the configuration.
3116

3117
    """
3118
    self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
3119
    _RedistributeAncillaryFiles(self)
3120

    
3121

    
3122
def _WaitForSync(lu, instance, disks=None, oneshot=False):
3123
  """Sleep and poll for an instance's disk to sync.
3124

3125
  """
3126
  if not instance.disks or disks is not None and not disks:
3127
    return True
3128

    
3129
  disks = _ExpandCheckDisks(instance, disks)
3130

    
3131
  if not oneshot:
3132
    lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
3133

    
3134
  node = instance.primary_node
3135

    
3136
  for dev in disks:
3137
    lu.cfg.SetDiskID(dev, node)
3138

    
3139
  # TODO: Convert to utils.Retry
3140

    
3141
  retries = 0
3142
  degr_retries = 10 # in seconds, as we sleep 1 second each time
3143
  while True:
3144
    max_time = 0
3145
    done = True
3146
    cumul_degraded = False
3147
    rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
3148
    msg = rstats.fail_msg
3149
    if msg:
3150
      lu.LogWarning("Can't get any data from node %s: %s", node, msg)
3151
      retries += 1
3152
      if retries >= 10:
3153
        raise errors.RemoteError("Can't contact node %s for mirror data,"
3154
                                 " aborting." % node)
3155
      time.sleep(6)
3156
      continue
3157
    rstats = rstats.payload
3158
    retries = 0
3159
    for i, mstat in enumerate(rstats):
3160
      if mstat is None:
3161
        lu.LogWarning("Can't compute data for node %s/%s",
3162
                           node, disks[i].iv_name)
3163
        continue
3164

    
3165
      cumul_degraded = (cumul_degraded or
3166
                        (mstat.is_degraded and mstat.sync_percent is None))
3167
      if mstat.sync_percent is not None:
3168
        done = False
3169
        if mstat.estimated_time is not None:
3170
          rem_time = ("%s remaining (estimated)" %
3171
                      utils.FormatSeconds(mstat.estimated_time))
3172
          max_time = mstat.estimated_time
3173
        else:
3174
          rem_time = "no time estimate"
3175
        lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
3176
                        (disks[i].iv_name, mstat.sync_percent, rem_time))
3177

    
3178
    # if we're done but degraded, let's do a few small retries, to
3179
    # make sure we see a stable and not transient situation; therefore
3180
    # we force restart of the loop
3181
    if (done or oneshot) and cumul_degraded and degr_retries > 0:
3182
      logging.info("Degraded disks found, %d retries left", degr_retries)
3183
      degr_retries -= 1
3184
      time.sleep(1)
3185
      continue
3186

    
3187
    if done or oneshot:
3188
      break
3189

    
3190
    time.sleep(min(60, max_time))
3191

    
3192
  if done:
3193
    lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
3194
  return not cumul_degraded
3195

    
3196

    
3197
def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
3198
  """Check that mirrors are not degraded.
3199

3200
  The ldisk parameter, if True, will change the test from the
3201
  is_degraded attribute (which represents overall non-ok status for
3202
  the device(s)) to the ldisk (representing the local storage status).
3203

3204
  """
3205
  lu.cfg.SetDiskID(dev, node)
3206

    
3207
  result = True
3208

    
3209
  if on_primary or dev.AssembleOnSecondary():
3210
    rstats = lu.rpc.call_blockdev_find(node, dev)
3211
    msg = rstats.fail_msg
3212
    if msg:
3213
      lu.LogWarning("Can't find disk on node %s: %s", node, msg)
3214
      result = False
3215
    elif not rstats.payload:
3216
      lu.LogWarning("Can't find disk on node %s", node)
3217
      result = False
3218
    else:
3219
      if ldisk:
3220
        result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
3221
      else:
3222
        result = result and not rstats.payload.is_degraded
3223

    
3224
  if dev.children:
3225
    for child in dev.children:
3226
      result = result and _CheckDiskConsistency(lu, child, node, on_primary)
3227

    
3228
  return result
3229

    
3230

    
3231
class LUOobCommand(NoHooksLU):
3232
  """Logical unit for OOB handling.
3233

3234
  """
3235
  REG_BGL = False
3236

    
3237
  def CheckPrereq(self):
3238
    """Check prerequisites.
3239

3240
    This checks:
3241
     - the node exists in the configuration
3242
     - OOB is supported
3243

3244
    Any errors are signaled by raising errors.OpPrereqError.
3245

3246
    """
3247
    self.nodes = []
3248
    for node_name in self.op.node_names:
3249
      node = self.cfg.GetNodeInfo(node_name)
3250

    
3251
      if node is None:
3252
        raise errors.OpPrereqError("Node %s not found" % node_name,
3253
                                   errors.ECODE_NOENT)
3254
      else:
3255
        self.nodes.append(node)
3256

    
3257
      if (self.op.command == constants.OOB_POWER_OFF and not node.offline):
3258
        raise errors.OpPrereqError(("Cannot power off node %s because it is"
3259
                                    " not marked offline") % node_name,
3260
                                   errors.ECODE_STATE)
3261

    
3262
  def ExpandNames(self):
3263
    """Gather locks we need.
3264

3265
    """
3266
    if self.op.node_names:
3267
      self.op.node_names = [_ExpandNodeName(self.cfg, name)
3268
                            for name in self.op.node_names]
3269
    else:
3270
      self.op.node_names = self.cfg.GetNodeList()
3271

    
3272
    self.needed_locks = {
3273
      locking.LEVEL_NODE: self.op.node_names,
3274
      }
3275

    
3276
  def Exec(self, feedback_fn):
3277
    """Execute OOB and return result if we expect any.
3278

3279
    """
3280
    master_node = self.cfg.GetMasterNode()
3281
    ret = []
3282

    
3283
    for node in self.nodes:
3284
      node_entry = [(constants.RS_NORMAL, node.name)]
3285
      ret.append(node_entry)
3286

    
3287
      oob_program = _SupportsOob(self.cfg, node)
3288

    
3289
      if not oob_program:
3290
        node_entry.append((constants.RS_UNAVAIL, None))
3291
        continue
3292

    
3293
      logging.info("Executing out-of-band command '%s' using '%s' on %s",
3294
                   self.op.command, oob_program, node.name)
3295
      result = self.rpc.call_run_oob(master_node, oob_program,
3296
                                     self.op.command, node.name,
3297
                                     self.op.timeout)
3298

    
3299
      if result.fail_msg:
3300
        self.LogWarning("On node '%s' out-of-band RPC failed with: %s",
3301
                        node.name, result.fail_msg)
3302
        node_entry.append((constants.RS_NODATA, None))
3303
      else:
3304
        try:
3305
          self._CheckPayload(result)
3306
        except errors.OpExecError, err:
3307
          self.LogWarning("The payload returned by '%s' is not valid: %s",
3308
                          node.name, err)
3309
          node_entry.append((constants.RS_NODATA, None))
3310
        else:
3311
          if self.op.command == constants.OOB_HEALTH:
3312
            # For health we should log important events
3313
            for item, status in result.payload:
3314
              if status in [constants.OOB_STATUS_WARNING,
3315
                            constants.OOB_STATUS_CRITICAL]:
3316
                self.LogWarning("On node '%s' item '%s' has status '%s'",
3317
                                node.name, item, status)
3318

    
3319
          if self.op.command == constants.OOB_POWER_ON:
3320
            node.powered = True
3321
          elif self.op.command == constants.OOB_POWER_OFF:
3322
            node.powered = False
3323
          elif self.op.command == constants.OOB_POWER_STATUS:
3324
            powered = result.payload[constants.OOB_POWER_STATUS_POWERED]
3325
            if powered != node.powered:
3326
              logging.warning(("Recorded power state (%s) of node '%s' does not"
3327
                               " match actual power state (%s)"), node.powered,
3328
                              node.name, powered)
3329

    
3330
          # For configuration changing commands we should update the node
3331
          if self.op.command in (constants.OOB_POWER_ON,
3332
                                 constants.OOB_POWER_OFF):
3333
            self.cfg.Update(node, feedback_fn)
3334

    
3335
          node_entry.append((constants.RS_NORMAL, result.payload))
3336

    
3337
    return ret
3338

    
3339
  def _CheckPayload(self, result):
3340
    """Checks if the payload is valid.
3341

3342
    @param result: RPC result
3343
    @raises errors.OpExecError: If payload is not valid
3344

3345
    """
3346
    errs = []
3347
    if self.op.command == constants.OOB_HEALTH:
3348
      if not isinstance(result.payload, list):
3349
        errs.append("command 'health' is expected to return a list but got %s" %
3350
                    type(result.payload))
3351
      else:
3352
        for item, status in result.payload:
3353
          if status not in constants.OOB_STATUSES:
3354
            errs.append("health item '%s' has invalid status '%s'" %
3355
                        (item, status))
3356

    
3357
    if self.op.command == constants.OOB_POWER_STATUS:
3358
      if not isinstance(result.payload, dict):
3359
        errs.append("power-status is expected to return a dict but got %s" %
3360
                    type(result.payload))
3361

    
3362
    if self.op.command in [
3363
        constants.OOB_POWER_ON,
3364
        constants.OOB_POWER_OFF,
3365
        constants.OOB_POWER_CYCLE,
3366
        ]:
3367
      if result.payload is not None:
3368
        errs.append("%s is expected to not return payload but got '%s'" %
3369
                    (self.op.command, result.payload))
3370

    
3371
    if errs:
3372
      raise errors.OpExecError("Check of out-of-band payload failed due to %s" %
3373
                               utils.CommaJoin(errs))
3374

    
3375

    
3376

    
3377
class LUOsDiagnose(NoHooksLU):
3378
  """Logical unit for OS diagnose/query.
3379

3380
  """
3381
  REQ_BGL = False
3382
  _HID = "hidden"
3383
  _BLK = "blacklisted"
3384
  _VLD = "valid"
3385
  _FIELDS_STATIC = utils.FieldSet()
3386
  _FIELDS_DYNAMIC = utils.FieldSet("name", _VLD, "node_status", "variants",
3387
                                   "parameters", "api_versions", _HID, _BLK)
3388

    
3389
  def CheckArguments(self):
3390
    if self.op.names:
3391
      raise errors.OpPrereqError("Selective OS query not supported",
3392
                                 errors.ECODE_INVAL)
3393

    
3394
    _CheckOutputFields(static=self._FIELDS_STATIC,
3395
                       dynamic=self._FIELDS_DYNAMIC,
3396
                       selected=self.op.output_fields)
3397

    
3398
  def ExpandNames(self):
3399
    # Lock all nodes, in shared mode
3400
    # Temporary removal of locks, should be reverted later
3401
    # TODO: reintroduce locks when they are lighter-weight
3402
    self.needed_locks = {}
3403
    #self.share_locks[locking.LEVEL_NODE] = 1
3404
    #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3405

    
3406
  @staticmethod
3407
  def _DiagnoseByOS(rlist):
3408
    """Remaps a per-node return list into an a per-os per-node dictionary
3409

3410
    @param rlist: a map with node names as keys and OS objects as values
3411

3412
    @rtype: dict
3413
    @return: a dictionary with osnames as keys and as value another
3414
        map, with nodes as keys and tuples of (path, status, diagnose,
3415
        variants, parameters, api_versions) as values, eg::
3416

3417
          {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
3418
                                     (/srv/..., False, "invalid api")],
3419
                           "node2": [(/srv/..., True, "", [], [])]}
3420
          }
3421

3422
    """
3423
    all_os = {}
3424
    # we build here the list of nodes that didn't fail the RPC (at RPC
3425
    # level), so that nodes with a non-responding node daemon don't
3426
    # make all OSes invalid
3427
    good_nodes = [node_name for node_name in rlist
3428
                  if not rlist[node_name].fail_msg]
3429
    for node_name, nr in rlist.items():
3430
      if nr.fail_msg or not nr.payload:
3431
        continue
3432
      for (name, path, status, diagnose, variants,
3433
           params, api_versions) in nr.payload:
3434
        if name not in all_os:
3435
          # build a list of nodes for this os containing empty lists
3436
          # for each node in node_list
3437
          all_os[name] = {}
3438
          for nname in good_nodes:
3439
            all_os[name][nname] = []
3440
        # convert params from [name, help] to (name, help)
3441
        params = [tuple(v) for v in params]
3442
        all_os[name][node_name].append((path, status, diagnose,
3443
                                        variants, params, api_versions))
3444
    return all_os
3445

    
3446
  def Exec(self, feedback_fn):
3447
    """Compute the list of OSes.
3448

3449
    """
3450
    valid_nodes = [node.name
3451
                   for node in self.cfg.GetAllNodesInfo().values()
3452
                   if not node.offline and node.vm_capable]
3453
    node_data = self.rpc.call_os_diagnose(valid_nodes)
3454
    pol = self._DiagnoseByOS(node_data)
3455
    output = []
3456
    cluster = self.cfg.GetClusterInfo()
3457

    
3458
    for os_name in utils.NiceSort(pol.keys()):
3459
      os_data = pol[os_name]
3460
      row = []
3461
      valid = True
3462
      (variants, params, api_versions) = null_state = (set(), set(), set())
3463
      for idx, osl in enumerate(os_data.values()):
3464
        valid = bool(valid and osl and osl[0][1])
3465
        if not valid:
3466
          (variants, params, api_versions) = null_state
3467
          break
3468
        node_variants, node_params, node_api = osl[0][3:6]
3469
        if idx == 0: # first entry
3470
          variants = set(node_variants)
3471
          params = set(node_params)
3472
          api_versions = set(node_api)
3473
        else: # keep consistency
3474
          variants.intersection_update(node_variants)
3475
          params.intersection_update(node_params)
3476
          api_versions.intersection_update(node_api)
3477

    
3478
      is_hid = os_name in cluster.hidden_os
3479
      is_blk = os_name in cluster.blacklisted_os
3480
      if ((self._HID not in self.op.output_fields and is_hid) or
3481
          (self._BLK not in self.op.output_fields and is_blk) or
3482
          (self._VLD not in self.op.output_fields and not valid)):
3483
        continue
3484

    
3485
      for field in self.op.output_fields:
3486
        if field == "name":
3487
          val = os_name
3488
        elif field == self._VLD:
3489
          val = valid
3490
        elif field == "node_status":
3491
          # this is just a copy of the dict
3492
          val = {}
3493
          for node_name, nos_list in os_data.items():
3494
            val[node_name] = nos_list
3495
        elif field == "variants":
3496
          val = utils.NiceSort(list(variants))
3497
        elif field == "parameters":
3498
          val = list(params)
3499
        elif field == "api_versions":
3500
          val = list(api_versions)
3501
        elif field == self._HID:
3502
          val = is_hid
3503
        elif field == self._BLK:
3504
          val = is_blk
3505
        else:
3506
          raise errors.ParameterError(field)
3507
        row.append(val)
3508
      output.append(row)
3509

    
3510
    return output
3511

    
3512

    
3513
class LUNodeRemove(LogicalUnit):
3514
  """Logical unit for removing a node.
3515

3516
  """
3517
  HPATH = "node-remove"
3518
  HTYPE = constants.HTYPE_NODE
3519

    
3520
  def BuildHooksEnv(self):
3521
    """Build hooks env.
3522

3523
    This doesn't run on the target node in the pre phase as a failed
3524
    node would then be impossible to remove.
3525

3526
    """
3527
    env = {
3528
      "OP_TARGET": self.op.node_name,
3529
      "NODE_NAME": self.op.node_name,
3530
      }
3531
    all_nodes = self.cfg.GetNodeList()
3532
    try:
3533
      all_nodes.remove(self.op.node_name)
3534
    except ValueError:
3535
      logging.warning("Node %s which is about to be removed not found"
3536
                      " in the all nodes list", self.op.node_name)
3537
    return env, all_nodes, all_nodes
3538

    
3539
  def CheckPrereq(self):
3540
    """Check prerequisites.
3541

3542
    This checks:
3543
     - the node exists in the configuration
3544
     - it does not have primary or secondary instances
3545
     - it's not the master
3546

3547
    Any errors are signaled by raising errors.OpPrereqError.
3548

3549
    """
3550
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3551
    node = self.cfg.GetNodeInfo(self.op.node_name)
3552
    assert node is not None
3553

    
3554
    instance_list = self.cfg.GetInstanceList()
3555

    
3556
    masternode = self.cfg.GetMasterNode()
3557
    if node.name == masternode:
3558
      raise errors.OpPrereqError("Node is the master node,"
3559
                                 " you need to failover first.",
3560
                                 errors.ECODE_INVAL)
3561

    
3562
    for instance_name in instance_list:
3563
      instance = self.cfg.GetInstanceInfo(instance_name)
3564
      if node.name in instance.all_nodes:
3565
        raise errors.OpPrereqError("Instance %s is still running on the node,"
3566
                                   " please remove first." % instance_name,
3567
                                   errors.ECODE_INVAL)
3568
    self.op.node_name = node.name
3569
    self.node = node
3570

    
3571
  def Exec(self, feedback_fn):
3572
    """Removes the node from the cluster.
3573

3574
    """
3575
    node = self.node
3576
    logging.info("Stopping the node daemon and removing configs from node %s",
3577
                 node.name)
3578

    
3579
    modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
3580

    
3581
    # Promote nodes to master candidate as needed
3582
    _AdjustCandidatePool(self, exceptions=[node.name])
3583
    self.context.RemoveNode(node.name)
3584

    
3585
    # Run post hooks on the node before it's removed
3586
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
3587
    try:
3588
      hm.RunPhase(constants.HOOKS_PHASE_POST, [node.name])
3589
    except:
3590
      # pylint: disable-msg=W0702
3591
      self.LogWarning("Errors occurred running hooks on %s" % node.name)
3592

    
3593
    result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
3594
    msg = result.fail_msg
3595
    if msg:
3596
      self.LogWarning("Errors encountered on the remote node while leaving"
3597
                      " the cluster: %s", msg)
3598

    
3599
    # Remove node from our /etc/hosts
3600
    if self.cfg.GetClusterInfo().modify_etc_hosts:
3601
      master_node = self.cfg.GetMasterNode()
3602
      result = self.rpc.call_etc_hosts_modify(master_node,
3603
                                              constants.ETC_HOSTS_REMOVE,
3604
                                              node.name, None)
3605
      result.Raise("Can't update hosts file with new host data")
3606
      _RedistributeAncillaryFiles(self)
3607

    
3608

    
3609
class _NodeQuery(_QueryBase):
3610
  FIELDS = query.NODE_FIELDS
3611

    
3612
  def ExpandNames(self, lu):
3613
    lu.needed_locks = {}
3614
    lu.share_locks[locking.LEVEL_NODE] = 1
3615

    
3616
    if self.names:
3617
      self.wanted = _GetWantedNodes(lu, self.names)
3618
    else:
3619
      self.wanted = locking.ALL_SET
3620

    
3621
    self.do_locking = (self.use_locking and
3622
                       query.NQ_LIVE in self.requested_data)
3623

    
3624
    if self.do_locking:
3625
      # if we don't request only static fields, we need to lock the nodes
3626
      lu.needed_locks[locking.LEVEL_NODE] = self.wanted
3627

    
3628
  def DeclareLocks(self, lu, level):
3629
    pass
3630

    
3631
  def _GetQueryData(self, lu):
3632
    """Computes the list of nodes and their attributes.
3633

3634
    """
3635
    all_info = lu.cfg.GetAllNodesInfo()
3636

    
3637
    nodenames = self._GetNames(lu, all_info.keys(), locking.LEVEL_NODE)
3638

    
3639
    # Gather data as requested
3640
    if query.NQ_LIVE in self.requested_data:
3641
      # filter out non-vm_capable nodes
3642
      toquery_nodes = [name for name in nodenames if all_info[name].vm_capable]
3643

    
3644
      node_data = lu.rpc.call_node_info(toquery_nodes, lu.cfg.GetVGName(),
3645
                                        lu.cfg.GetHypervisorType())
3646
      live_data = dict((name, nresult.payload)
3647
                       for (name, nresult) in node_data.items()
3648
                       if not nresult.fail_msg and nresult.payload)
3649
    else:
3650
      live_data = None
3651

    
3652
    if query.NQ_INST in self.requested_data:
3653
      node_to_primary = dict([(name, set()) for name in nodenames])
3654
      node_to_secondary = dict([(name, set()) for name in nodenames])
3655

    
3656
      inst_data = lu.cfg.GetAllInstancesInfo()
3657

    
3658
      for inst in inst_data.values():
3659
        if inst.primary_node in node_to_primary:
3660
          node_to_primary[inst.primary_node].add(inst.name)
3661
        for secnode in inst.secondary_nodes:
3662
          if secnode in node_to_secondary:
3663
            node_to_secondary[secnode].add(inst.name)
3664
    else:
3665
      node_to_primary = None
3666
      node_to_secondary = None
3667

    
3668
    if query.NQ_OOB in self.requested_data:
3669
      oob_support = dict((name, bool(_SupportsOob(lu.cfg, node)))
3670
                         for name, node in all_info.iteritems())
3671
    else:
3672
      oob_support = None
3673

    
3674
    if query.NQ_GROUP in self.requested_data:
3675
      groups = lu.cfg.GetAllNodeGroupsInfo()
3676
    else:
3677
      groups = {}
3678

    
3679
    return query.NodeQueryData([all_info[name] for name in nodenames],
3680
                               live_data, lu.cfg.GetMasterNode(),
3681
                               node_to_primary, node_to_secondary, groups,
3682
                               oob_support, lu.cfg.GetClusterInfo())
3683

    
3684

    
3685
class LUNodeQuery(NoHooksLU):
3686
  """Logical unit for querying nodes.
3687

3688
  """
3689
  # pylint: disable-msg=W0142
3690
  REQ_BGL = False
3691

    
3692
  def CheckArguments(self):
3693
    self.nq = _NodeQuery(self.op.names, self.op.output_fields,
3694
                         self.op.use_locking)
3695

    
3696
  def ExpandNames(self):
3697
    self.nq.ExpandNames(self)
3698

    
3699
  def Exec(self, feedback_fn):
3700
    return self.nq.OldStyleQuery(self)
3701

    
3702

    
3703
class LUNodeQueryvols(NoHooksLU):
3704
  """Logical unit for getting volumes on node(s).
3705

3706
  """
3707
  REQ_BGL = False
3708
  _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
3709
  _FIELDS_STATIC = utils.FieldSet("node")
3710

    
3711
  def CheckArguments(self):
3712
    _CheckOutputFields(static=self._FIELDS_STATIC,
3713
                       dynamic=self._FIELDS_DYNAMIC,
3714
                       selected=self.op.output_fields)
3715

    
3716
  def ExpandNames(self):
3717
    self.needed_locks = {}
3718
    self.share_locks[locking.LEVEL_NODE] = 1
3719
    if not self.op.nodes:
3720
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3721
    else:
3722
      self.needed_locks[locking.LEVEL_NODE] = \
3723
        _GetWantedNodes(self, self.op.nodes)
3724

    
3725
  def Exec(self, feedback_fn):
3726
    """Computes the list of nodes and their attributes.
3727

3728
    """
3729
    nodenames = self.acquired_locks[locking.LEVEL_NODE]
3730
    volumes = self.rpc.call_node_volumes(nodenames)
3731

    
3732
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
3733
             in self.cfg.GetInstanceList()]
3734

    
3735
    lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
3736

    
3737
    output = []
3738
    for node in nodenames:
3739
      nresult = volumes[node]
3740
      if nresult.offline:
3741
        continue
3742
      msg = nresult.fail_msg
3743
      if msg:
3744
        self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
3745
        continue
3746

    
3747
      node_vols = nresult.payload[:]
3748
      node_vols.sort(key=lambda vol: vol['dev'])
3749

    
3750
      for vol in node_vols:
3751
        node_output = []
3752
        for field in self.op.output_fields:
3753
          if field == "node":
3754
            val = node
3755
          elif field == "phys":
3756
            val = vol['dev']
3757
          elif field == "vg":
3758
            val = vol['vg']
3759
          elif field == "name":
3760
            val = vol['name']
3761
          elif field == "size":
3762
            val = int(float(vol['size']))
3763
          elif field == "instance":
3764
            for inst in ilist:
3765
              if node not in lv_by_node[inst]:
3766
                continue
3767
              if vol['name'] in lv_by_node[inst][node]:
3768
                val = inst.name
3769
                break
3770
            else:
3771
              val = '-'
3772
          else:
3773
            raise errors.ParameterError(field)
3774
          node_output.append(str(val))
3775

    
3776
        output.append(node_output)
3777

    
3778
    return output
3779

    
3780

    
3781
class LUNodeQueryStorage(NoHooksLU):
3782
  """Logical unit for getting information on storage units on node(s).
3783

3784
  """
3785
  _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
3786
  REQ_BGL = False
3787

    
3788
  def CheckArguments(self):
3789
    _CheckOutputFields(static=self._FIELDS_STATIC,
3790
                       dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
3791
                       selected=self.op.output_fields)
3792

    
3793
  def ExpandNames(self):
3794
    self.needed_locks = {}
3795
    self.share_locks[locking.LEVEL_NODE] = 1
3796

    
3797
    if self.op.nodes:
3798
      self.needed_locks[locking.LEVEL_NODE] = \
3799
        _GetWantedNodes(self, self.op.nodes)
3800
    else:
3801
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3802

    
3803
  def Exec(self, feedback_fn):
3804
    """Computes the list of nodes and their attributes.
3805

3806
    """
3807
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
3808

    
3809
    # Always get name to sort by
3810
    if constants.SF_NAME in self.op.output_fields:
3811
      fields = self.op.output_fields[:]
3812
    else:
3813
      fields = [constants.SF_NAME] + self.op.output_fields
3814

    
3815
    # Never ask for node or type as it's only known to the LU
3816
    for extra in [constants.SF_NODE, constants.SF_TYPE]:
3817
      while extra in fields:
3818
        fields.remove(extra)
3819

    
3820
    field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
3821
    name_idx = field_idx[constants.SF_NAME]
3822

    
3823
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3824
    data = self.rpc.call_storage_list(self.nodes,
3825
                                      self.op.storage_type, st_args,
3826
                                      self.op.name, fields)
3827

    
3828
    result = []
3829

    
3830
    for node in utils.NiceSort(self.nodes):
3831
      nresult = data[node]
3832
      if nresult.offline:
3833
        continue
3834

    
3835
      msg = nresult.fail_msg
3836
      if msg:
3837
        self.LogWarning("Can't get storage data from node %s: %s", node, msg)
3838
        continue
3839

    
3840
      rows = dict([(row[name_idx], row) for row in nresult.payload])
3841

    
3842
      for name in utils.NiceSort(rows.keys()):
3843
        row = rows[name]
3844

    
3845
        out = []
3846

    
3847
        for field in self.op.output_fields:
3848
          if field == constants.SF_NODE:
3849
            val = node
3850
          elif field == constants.SF_TYPE:
3851
            val = self.op.storage_type
3852
          elif field in field_idx:
3853
            val = row[field_idx[field]]
3854
          else:
3855
            raise errors.ParameterError(field)
3856

    
3857
          out.append(val)
3858

    
3859
        result.append(out)
3860

    
3861
    return result
3862

    
3863

    
3864
class _InstanceQuery(_QueryBase):
3865
  FIELDS = query.INSTANCE_FIELDS
3866

    
3867
  def ExpandNames(self, lu):
3868
    lu.needed_locks = {}
3869
    lu.share_locks[locking.LEVEL_INSTANCE] = 1
3870
    lu.share_locks[locking.LEVEL_NODE] = 1
3871

    
3872
    if self.names:
3873
      self.wanted = _GetWantedInstances(lu, self.names)
3874
    else:
3875
      self.wanted = locking.ALL_SET
3876

    
3877
    self.do_locking = (self.use_locking and
3878
                       query.IQ_LIVE in self.requested_data)
3879
    if self.do_locking:
3880
      lu.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
3881
      lu.needed_locks[locking.LEVEL_NODE] = []
3882
      lu.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3883

    
3884
  def DeclareLocks(self, lu, level):
3885
    if level == locking.LEVEL_NODE and self.do_locking:
3886
      lu._LockInstancesNodes() # pylint: disable-msg=W0212
3887

    
3888
  def _GetQueryData(self, lu):
3889
    """Computes the list of instances and their attributes.
3890

3891
    """
3892
    cluster = lu.cfg.GetClusterInfo()
3893
    all_info = lu.cfg.GetAllInstancesInfo()
3894

    
3895
    instance_names = self._GetNames(lu, all_info.keys(), locking.LEVEL_INSTANCE)
3896

    
3897
    instance_list = [all_info[name] for name in instance_names]
3898
    nodes = frozenset(itertools.chain(*(inst.all_nodes
3899
                                        for inst in instance_list)))
3900
    hv_list = list(set([inst.hypervisor for inst in instance_list]))
3901
    bad_nodes = []
3902
    offline_nodes = []
3903
    wrongnode_inst = set()
3904

    
3905
    # Gather data as requested
3906
    if self.requested_data & set([query.IQ_LIVE, query.IQ_CONSOLE]):
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
          for inst in result.payload:
3919
            if inst in all_info:
3920
              if all_info[inst].primary_node == name:
3921
                live_data.update(result.payload)
3922
              else:
3923
                wrongnode_inst.add(inst)
3924
            else:
3925
              # orphan instance; we don't list it here as we don't
3926
              # handle this case yet in the output of instance listing
3927
              logging.warning("Orphan instance '%s' found on node %s",
3928
                              inst, name)
3929
        # else no instance is alive
3930
    else:
3931
      live_data = {}
3932

    
3933
    if query.IQ_DISKUSAGE in self.requested_data:
3934
      disk_usage = dict((inst.name,
3935
                         _ComputeDiskSize(inst.disk_template,
3936
                                          [{"size": disk.size}
3937
                                           for disk in inst.disks]))
3938
                        for inst in instance_list)
3939
    else:
3940
      disk_usage = None
3941

    
3942
    if query.IQ_CONSOLE in self.requested_data:
3943
      consinfo = {}
3944
      for inst in instance_list:
3945
        if inst.name in live_data:
3946
          # Instance is running
3947
          consinfo[inst.name] = _GetInstanceConsole(cluster, inst)
3948
        else:
3949
          consinfo[inst.name] = None
3950
      assert set(consinfo.keys()) == set(instance_names)
3951
    else:
3952
      consinfo = None
3953

    
3954
    return query.InstanceQueryData(instance_list, lu.cfg.GetClusterInfo(),
3955
                                   disk_usage, offline_nodes, bad_nodes,
3956
                                   live_data, wrongnode_inst, consinfo)
3957

    
3958

    
3959
class LUQuery(NoHooksLU):
3960
  """Query for resources/items of a certain kind.
3961

3962
  """
3963
  # pylint: disable-msg=W0142
3964
  REQ_BGL = False
3965

    
3966
  def CheckArguments(self):
3967
    qcls = _GetQueryImplementation(self.op.what)
3968
    names = qlang.ReadSimpleFilter("name", self.op.filter)
3969

    
3970
    self.impl = qcls(names, self.op.fields, False)
3971

    
3972
  def ExpandNames(self):
3973
    self.impl.ExpandNames(self)
3974

    
3975
  def DeclareLocks(self, level):
3976
    self.impl.DeclareLocks(self, level)
3977

    
3978
  def Exec(self, feedback_fn):
3979
    return self.impl.NewStyleQuery(self)
3980

    
3981

    
3982
class LUQueryFields(NoHooksLU):
3983
  """Query for resources/items of a certain kind.
3984

3985
  """
3986
  # pylint: disable-msg=W0142
3987
  REQ_BGL = False
3988

    
3989
  def CheckArguments(self):
3990
    self.qcls = _GetQueryImplementation(self.op.what)
3991

    
3992
  def ExpandNames(self):
3993
    self.needed_locks = {}
3994

    
3995
  def Exec(self, feedback_fn):
3996
    return self.qcls.FieldsQuery(self.op.fields)
3997

    
3998

    
3999
class LUNodeModifyStorage(NoHooksLU):
4000
  """Logical unit for modifying a storage volume on a node.
4001

4002
  """
4003
  REQ_BGL = False
4004

    
4005
  def CheckArguments(self):
4006
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4007

    
4008
    storage_type = self.op.storage_type
4009

    
4010
    try:
4011
      modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
4012
    except KeyError:
4013
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
4014
                                 " modified" % storage_type,
4015
                                 errors.ECODE_INVAL)
4016

    
4017
    diff = set(self.op.changes.keys()) - modifiable
4018
    if diff:
4019
      raise errors.OpPrereqError("The following fields can not be modified for"
4020
                                 " storage units of type '%s': %r" %
4021
                                 (storage_type, list(diff)),
4022
                                 errors.ECODE_INVAL)
4023

    
4024
  def ExpandNames(self):
4025
    self.needed_locks = {
4026
      locking.LEVEL_NODE: self.op.node_name,
4027
      }
4028

    
4029
  def Exec(self, feedback_fn):
4030
    """Computes the list of nodes and their attributes.
4031

4032
    """
4033
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
4034
    result = self.rpc.call_storage_modify(self.op.node_name,
4035
                                          self.op.storage_type, st_args,
4036
                                          self.op.name, self.op.changes)
4037
    result.Raise("Failed to modify storage unit '%s' on %s" %
4038
                 (self.op.name, self.op.node_name))
4039

    
4040

    
4041
class LUNodeAdd(LogicalUnit):
4042
  """Logical unit for adding node to the cluster.
4043

4044
  """
4045
  HPATH = "node-add"
4046
  HTYPE = constants.HTYPE_NODE
4047
  _NFLAGS = ["master_capable", "vm_capable"]
4048

    
4049
  def CheckArguments(self):
4050
    self.primary_ip_family = self.cfg.GetPrimaryIPFamily()
4051
    # validate/normalize the node name
4052
    self.hostname = netutils.GetHostname(name=self.op.node_name,
4053
                                         family=self.primary_ip_family)
4054
    self.op.node_name = self.hostname.name
4055
    if self.op.readd and self.op.group:
4056
      raise errors.OpPrereqError("Cannot pass a node group when a node is"
4057
                                 " being readded", errors.ECODE_INVAL)
4058

    
4059
  def BuildHooksEnv(self):
4060
    """Build hooks env.
4061

4062
    This will run on all nodes before, and on all nodes + the new node after.
4063

4064
    """
4065
    env = {
4066
      "OP_TARGET": self.op.node_name,
4067
      "NODE_NAME": self.op.node_name,
4068
      "NODE_PIP": self.op.primary_ip,
4069
      "NODE_SIP": self.op.secondary_ip,
4070
      "MASTER_CAPABLE": str(self.op.master_capable),
4071
      "VM_CAPABLE": str(self.op.vm_capable),
4072
      }
4073
    nodes_0 = self.cfg.GetNodeList()
4074
    nodes_1 = nodes_0 + [self.op.node_name, ]
4075
    return env, nodes_0, nodes_1
4076

    
4077
  def CheckPrereq(self):
<