Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ 72740756

History | View | Annotate | Download (401.7 kB)

<
1
#
2
#
3

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

    
21

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

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

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

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

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

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

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

    
63

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

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

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

    
76

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

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

90
  Note that all commands require root permissions.
91

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

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

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

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

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

    
133
    # Tasklets
134
    self.tasklets = None
135

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

    
139
    self.CheckArguments()
140

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

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

    
149
  ssh = property(fget=__GetSSH)
150

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

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

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

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

166
    """
167
    pass
168

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

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

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

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

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

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

194
    Examples::
195

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

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

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

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

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

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

233
    """
234

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

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

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

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

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

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

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

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

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

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

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

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

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

292
    """
293
    raise NotImplementedError
294

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
381
    del self.recalculate_locks[locking.LEVEL_NODE]
382

    
383

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

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

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

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

397
    This just raises an error.
398

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

    
402

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

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

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

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

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

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

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

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

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

435
    """
436
    pass
437

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

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

445
    """
446
    raise NotImplementedError
447

    
448

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

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

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

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

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

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

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

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

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

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

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

    
492
    # Return expanded names
493
    return self.wanted
494

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

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

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

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

507
    See L{LogicalUnit.ExpandNames}.
508

509
    """
510
    raise NotImplementedError()
511

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

515
    See L{LogicalUnit.DeclareLocks}.
516

517
    """
518
    raise NotImplementedError()
519

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

523
    @return: Query data object
524

525
    """
526
    raise NotImplementedError()
527

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

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

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

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

    
540

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

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

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

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

    
558

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

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

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

    
578

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

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

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

    
611

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

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

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

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

    
630

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

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

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

    
645

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

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

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

    
660

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

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

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

    
673

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

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

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

    
686

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

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

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

    
704

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

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

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

    
731

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

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

    
739

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

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

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

    
755

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

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

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

    
772

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

    
777

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

    
782

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

788
  This builds the hook environment from individual variables.
789

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

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

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

    
852
  env["INSTANCE_NIC_COUNT"] = nic_count
853

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

    
862
  env["INSTANCE_DISK_COUNT"] = disk_count
863

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

    
868
  return env
869

    
870

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

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

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

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

    
894

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

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

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

    
932

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

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

    
948

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

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

    
959

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

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

    
973

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

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

    
982

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

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

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

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

    
1002

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

    
1006

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

1010
  """
1011

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

    
1014

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

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

    
1022

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

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

    
1030

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

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

    
1040
  return []
1041

    
1042

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

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

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

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

    
1057
  return faulty
1058

    
1059

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

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

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

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

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

    
1091

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

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

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

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

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

1110
    """
1111
    return True
1112

    
1113

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

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

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

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

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

1131
    This checks whether the cluster is empty.
1132

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

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

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

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

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

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

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

    
1166
    return master
1167

    
1168

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

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

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

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

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

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

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

    
1201

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
1412
    return True
1413

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

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

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

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

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

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

    
1446
  def _VerifyNodeLVM(self, ninfo, nresult, vg_name):
1447
    """Check the node LVM results.
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 _VerifyNodeBridges(self, ninfo, nresult, bridges):
1484
    """Check the node bridges.
1485

1486
    @type ninfo: L{objects.Node}
1487
    @param ninfo: the node to check
1488
    @param nresult: the remote results for the node
1489
    @param bridges: the expected list of bridges
1490

1491
    """
1492
    if not bridges:
1493
      return
1494

    
1495
    node = ninfo.name
1496
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1497

    
1498
    missing = nresult.get(constants.NV_BRIDGES, None)
1499
    test = not isinstance(missing, list)
1500
    _ErrorIf(test, self.ENODENET, node,
1501
             "did not return valid bridge information")
1502
    if not test:
1503
      _ErrorIf(bool(missing), self.ENODENET, node, "missing bridges: %s" %
1504
               utils.CommaJoin(sorted(missing)))
1505

    
1506
  def _VerifyNodeNetwork(self, ninfo, nresult):
1507
    """Check the node network connectivity results.
1508

1509
    @type ninfo: L{objects.Node}
1510
    @param ninfo: the node to check
1511
    @param nresult: the remote results for the node
1512

1513
    """
1514
    node = ninfo.name
1515
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1516

    
1517
    test = constants.NV_NODELIST not in nresult
1518
    _ErrorIf(test, self.ENODESSH, node,
1519
             "node hasn't returned node ssh connectivity data")
1520
    if not test:
1521
      if nresult[constants.NV_NODELIST]:
1522
        for a_node, a_msg in nresult[constants.NV_NODELIST].items():
1523
          _ErrorIf(True, self.ENODESSH, node,
1524
                   "ssh communication with node '%s': %s", a_node, a_msg)
1525

    
1526
    test = constants.NV_NODENETTEST not in nresult
1527
    _ErrorIf(test, self.ENODENET, node,
1528
             "node hasn't returned node tcp connectivity data")
1529
    if not test:
1530
      if nresult[constants.NV_NODENETTEST]:
1531
        nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys())
1532
        for anode in nlist:
1533
          _ErrorIf(True, self.ENODENET, node,
1534
                   "tcp communication with node '%s': %s",
1535
                   anode, nresult[constants.NV_NODENETTEST][anode])
1536

    
1537
    test = constants.NV_MASTERIP not in nresult
1538
    _ErrorIf(test, self.ENODENET, node,
1539
             "node hasn't returned node master IP reachability data")
1540
    if not test:
1541
      if not nresult[constants.NV_MASTERIP]:
1542
        if node == self.master_node:
1543
          msg = "the master node cannot reach the master IP (not configured?)"
1544
        else:
1545
          msg = "cannot reach the master IP"
1546
        _ErrorIf(True, self.ENODENET, node, msg)
1547

    
1548
  def _VerifyInstance(self, instance, instanceconfig, node_image,
1549
                      diskstatus):
1550
    """Verify an instance.
1551

1552
    This function checks to see if the required block devices are
1553
    available on the instance's node.
1554

1555
    """
1556
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1557
    node_current = instanceconfig.primary_node
1558

    
1559
    node_vol_should = {}
1560
    instanceconfig.MapLVsByNode(node_vol_should)
1561

    
1562
    for node in node_vol_should:
1563
      n_img = node_image[node]
1564
      if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1565
        # ignore missing volumes on offline or broken nodes
1566
        continue
1567
      for volume in node_vol_should[node]:
1568
        test = volume not in n_img.volumes
1569
        _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
1570
                 "volume %s missing on node %s", volume, node)
1571

    
1572
    if instanceconfig.admin_up:
1573
      pri_img = node_image[node_current]
1574
      test = instance not in pri_img.instances and not pri_img.offline
1575
      _ErrorIf(test, self.EINSTANCEDOWN, instance,
1576
               "instance not running on its primary node %s",
1577
               node_current)
1578

    
1579
    for node, n_img in node_image.items():
1580
      if node != node_current:
1581
        test = instance in n_img.instances
1582
        _ErrorIf(test, self.EINSTANCEWRONGNODE, instance,
1583
                 "instance should not run on node %s", node)
1584

    
1585
    diskdata = [(nname, success, status, idx)
1586
                for (nname, disks) in diskstatus.items()
1587
                for idx, (success, status) in enumerate(disks)]
1588

    
1589
    for nname, success, bdev_status, idx in diskdata:
1590
      # the 'ghost node' construction in Exec() ensures that we have a
1591
      # node here
1592
      snode = node_image[nname]
1593
      bad_snode = snode.ghost or snode.offline
1594
      _ErrorIf(instanceconfig.admin_up and not success and not bad_snode,
1595
               self.EINSTANCEFAULTYDISK, instance,
1596
               "couldn't retrieve status for disk/%s on %s: %s",
1597
               idx, nname, bdev_status)
1598
      _ErrorIf((instanceconfig.admin_up and success and
1599
                bdev_status.ldisk_status == constants.LDS_FAULTY),
1600
               self.EINSTANCEFAULTYDISK, instance,
1601
               "disk/%s on %s is faulty", idx, nname)
1602

    
1603
  def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):
1604
    """Verify if there are any unknown volumes in the cluster.
1605

1606
    The .os, .swap and backup volumes are ignored. All other volumes are
1607
    reported as unknown.
1608

1609
    @type reserved: L{ganeti.utils.FieldSet}
1610
    @param reserved: a FieldSet of reserved volume names
1611

1612
    """
1613
    for node, n_img in node_image.items():
1614
      if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1615
        # skip non-healthy nodes
1616
        continue
1617
      for volume in n_img.volumes:
1618
        test = ((node not in node_vol_should or
1619
                volume not in node_vol_should[node]) and
1620
                not reserved.Matches(volume))
1621
        self._ErrorIf(test, self.ENODEORPHANLV, node,
1622
                      "volume %s is unknown", volume)
1623

    
1624
  def _VerifyOrphanInstances(self, instancelist, node_image):
1625
    """Verify the list of running instances.
1626

1627
    This checks what instances are running but unknown to the cluster.
1628

1629
    """
1630
    for node, n_img in node_image.items():
1631
      for o_inst in n_img.instances:
1632
        test = o_inst not in instancelist
1633
        self._ErrorIf(test, self.ENODEORPHANINSTANCE, node,
1634
                      "instance %s on node %s should not exist", o_inst, node)
1635

    
1636
  def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
1637
    """Verify N+1 Memory Resilience.
1638

1639
    Check that if one single node dies we can still start all the
1640
    instances it was primary for.
1641

1642
    """
1643
    for node, n_img in node_image.items():
1644
      # This code checks that every node which is now listed as
1645
      # secondary has enough memory to host all instances it is
1646
      # supposed to should a single other node in the cluster fail.
1647
      # FIXME: not ready for failover to an arbitrary node
1648
      # FIXME: does not support file-backed instances
1649
      # WARNING: we currently take into account down instances as well
1650
      # as up ones, considering that even if they're down someone
1651
      # might want to start them even in the event of a node failure.
1652
      if n_img.offline:
1653
        # we're skipping offline nodes from the N+1 warning, since
1654
        # most likely we don't have good memory infromation from them;
1655
        # we already list instances living on such nodes, and that's
1656
        # enough warning
1657
        continue
1658
      for prinode, instances in n_img.sbp.items():
1659
        needed_mem = 0
1660
        for instance in instances:
1661
          bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
1662
          if bep[constants.BE_AUTO_BALANCE]:
1663
            needed_mem += bep[constants.BE_MEMORY]
1664
        test = n_img.mfree < needed_mem
1665
        self._ErrorIf(test, self.ENODEN1, node,
1666
                      "not enough memory to accomodate instance failovers"
1667
                      " should node %s fail (%dMiB needed, %dMiB available)",
1668
                      prinode, needed_mem, n_img.mfree)
1669

    
1670
  def _VerifyNodeFiles(self, ninfo, nresult, file_list, local_cksum,
1671
                       master_files):
1672
    """Verifies and computes the node required file checksums.
1673

1674
    @type ninfo: L{objects.Node}
1675
    @param ninfo: the node to check
1676
    @param nresult: the remote results for the node
1677
    @param file_list: required list of files
1678
    @param local_cksum: dictionary of local files and their checksums
1679
    @param master_files: list of files that only masters should have
1680

1681
    """
1682
    node = ninfo.name
1683
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1684

    
1685
    remote_cksum = nresult.get(constants.NV_FILELIST, None)
1686
    test = not isinstance(remote_cksum, dict)
1687
    _ErrorIf(test, self.ENODEFILECHECK, node,
1688
             "node hasn't returned file checksum data")
1689
    if test:
1690
      return
1691

    
1692
    for file_name in file_list:
1693
      node_is_mc = ninfo.master_candidate
1694
      must_have = (file_name not in master_files) or node_is_mc
1695
      # missing
1696
      test1 = file_name not in remote_cksum
1697
      # invalid checksum
1698
      test2 = not test1 and remote_cksum[file_name] != local_cksum[file_name]
1699
      # existing and good
1700
      test3 = not test1 and remote_cksum[file_name] == local_cksum[file_name]
1701
      _ErrorIf(test1 and must_have, self.ENODEFILECHECK, node,
1702
               "file '%s' missing", file_name)
1703
      _ErrorIf(test2 and must_have, self.ENODEFILECHECK, node,
1704
               "file '%s' has wrong checksum", file_name)
1705
      # not candidate and this is not a must-have file
1706
      _ErrorIf(test2 and not must_have, self.ENODEFILECHECK, node,
1707
               "file '%s' should not exist on non master"
1708
               " candidates (and the file is outdated)", file_name)
1709
      # all good, except non-master/non-must have combination
1710
      _ErrorIf(test3 and not must_have, self.ENODEFILECHECK, node,
1711
               "file '%s' should not exist"
1712
               " on non master candidates", file_name)
1713

    
1714
  def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper,
1715
                      drbd_map):
1716
    """Verifies and the node DRBD status.
1717

1718
    @type ninfo: L{objects.Node}
1719
    @param ninfo: the node to check
1720
    @param nresult: the remote results for the node
1721
    @param instanceinfo: the dict of instances
1722
    @param drbd_helper: the configured DRBD usermode helper
1723
    @param drbd_map: the DRBD map as returned by
1724
        L{ganeti.config.ConfigWriter.ComputeDRBDMap}
1725

1726
    """
1727
    node = ninfo.name
1728
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1729

    
1730
    if drbd_helper:
1731
      helper_result = nresult.get(constants.NV_DRBDHELPER, None)
1732
      test = (helper_result == None)
1733
      _ErrorIf(test, self.ENODEDRBDHELPER, node,
1734
               "no drbd usermode helper returned")
1735
      if helper_result:
1736
        status, payload = helper_result
1737
        test = not status
1738
        _ErrorIf(test, self.ENODEDRBDHELPER, node,
1739
                 "drbd usermode helper check unsuccessful: %s", payload)
1740
        test = status and (payload != drbd_helper)
1741
        _ErrorIf(test, self.ENODEDRBDHELPER, node,
1742
                 "wrong drbd usermode helper: %s", payload)
1743

    
1744
    # compute the DRBD minors
1745
    node_drbd = {}
1746
    for minor, instance in drbd_map[node].items():
1747
      test = instance not in instanceinfo
1748
      _ErrorIf(test, self.ECLUSTERCFG, None,
1749
               "ghost instance '%s' in temporary DRBD map", instance)
1750
        # ghost instance should not be running, but otherwise we
1751
        # don't give double warnings (both ghost instance and
1752
        # unallocated minor in use)
1753
      if test:
1754
        node_drbd[minor] = (instance, False)
1755
      else:
1756
        instance = instanceinfo[instance]
1757
        node_drbd[minor] = (instance.name, instance.admin_up)
1758

    
1759
    # and now check them
1760
    used_minors = nresult.get(constants.NV_DRBDLIST, [])
1761
    test = not isinstance(used_minors, (tuple, list))
1762
    _ErrorIf(test, self.ENODEDRBD, node,
1763
             "cannot parse drbd status file: %s", str(used_minors))
1764
    if test:
1765
      # we cannot check drbd status
1766
      return
1767

    
1768
    for minor, (iname, must_exist) in node_drbd.items():
1769
      test = minor not in used_minors and must_exist
1770
      _ErrorIf(test, self.ENODEDRBD, node,
1771
               "drbd minor %d of instance %s is not active", minor, iname)
1772
    for minor in used_minors:
1773
      test = minor not in node_drbd
1774
      _ErrorIf(test, self.ENODEDRBD, node,
1775
               "unallocated drbd minor %d is in use", minor)
1776

    
1777
  def _UpdateNodeOS(self, ninfo, nresult, nimg):
1778
    """Builds the node OS structures.
1779

1780
    @type ninfo: L{objects.Node}
1781
    @param ninfo: the node to check
1782
    @param nresult: the remote results for the node
1783
    @param nimg: the node image object
1784

1785
    """
1786
    node = ninfo.name
1787
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1788

    
1789
    remote_os = nresult.get(constants.NV_OSLIST, None)
1790
    test = (not isinstance(remote_os, list) or
1791
            not compat.all(isinstance(v, list) and len(v) == 7
1792
                           for v in remote_os))
1793

    
1794
    _ErrorIf(test, self.ENODEOS, node,
1795
             "node hasn't returned valid OS data")
1796

    
1797
    nimg.os_fail = test
1798

    
1799
    if test:
1800
      return
1801

    
1802
    os_dict = {}
1803

    
1804
    for (name, os_path, status, diagnose,
1805
         variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
1806

    
1807
      if name not in os_dict:
1808
        os_dict[name] = []
1809

    
1810
      # parameters is a list of lists instead of list of tuples due to
1811
      # JSON lacking a real tuple type, fix it:
1812
      parameters = [tuple(v) for v in parameters]
1813
      os_dict[name].append((os_path, status, diagnose,
1814
                            set(variants), set(parameters), set(api_ver)))
1815

    
1816
    nimg.oslist = os_dict
1817

    
1818
  def _VerifyNodeOS(self, ninfo, nimg, base):
1819
    """Verifies the node OS list.
1820

1821
    @type ninfo: L{objects.Node}
1822
    @param ninfo: the node to check
1823
    @param nimg: the node image object
1824
    @param base: the 'template' node we match against (e.g. from the master)
1825

1826
    """
1827
    node = ninfo.name
1828
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1829

    
1830
    assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
1831

    
1832
    beautify_params = lambda l: ["%s: %s" % (k, v) for (k, v) in l]
1833
    for os_name, os_data in nimg.oslist.items():
1834
      assert os_data, "Empty OS status for OS %s?!" % os_name
1835
      f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
1836
      _ErrorIf(not f_status, self.ENODEOS, node,
1837
               "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
1838
      _ErrorIf(len(os_data) > 1, self.ENODEOS, node,
1839
               "OS '%s' has multiple entries (first one shadows the rest): %s",
1840
               os_name, utils.CommaJoin([v[0] for v in os_data]))
1841
      # this will catched in backend too
1842
      _ErrorIf(compat.any(v >= constants.OS_API_V15 for v in f_api)
1843
               and not f_var, self.ENODEOS, node,
1844
               "OS %s with API at least %d does not declare any variant",
1845
               os_name, constants.OS_API_V15)
1846
      # comparisons with the 'base' image
1847
      test = os_name not in base.oslist
1848
      _ErrorIf(test, self.ENODEOS, node,
1849
               "Extra OS %s not present on reference node (%s)",
1850
               os_name, base.name)
1851
      if test:
1852
        continue
1853
      assert base.oslist[os_name], "Base node has empty OS status?"
1854
      _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
1855
      if not b_status:
1856
        # base OS is invalid, skipping
1857
        continue
1858
      for kind, a, b in [("API version", f_api, b_api),
1859
                         ("variants list", f_var, b_var),
1860
                         ("parameters", beautify_params(f_param),
1861
                          beautify_params(b_param))]:
1862
        _ErrorIf(a != b, self.ENODEOS, node,
1863
                 "OS %s for %s differs from reference node %s: [%s] vs. [%s]",
1864
                 kind, os_name, base.name,
1865
                 utils.CommaJoin(sorted(a)), utils.CommaJoin(sorted(b)))
1866

    
1867
    # check any missing OSes
1868
    missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
1869
    _ErrorIf(missing, self.ENODEOS, node,
1870
             "OSes present on reference node %s but missing on this node: %s",
1871
             base.name, utils.CommaJoin(missing))
1872

    
1873
  def _VerifyOob(self, ninfo, nresult):
1874
    """Verifies out of band functionality of a node.
1875

1876
    @type ninfo: L{objects.Node}
1877
    @param ninfo: the node to check
1878
    @param nresult: the remote results for the node
1879

1880
    """
1881
    node = ninfo.name
1882
    # We just have to verify the paths on master and/or master candidates
1883
    # as the oob helper is invoked on the master
1884
    if ((ninfo.master_candidate or ninfo.master_capable) and
1885
        constants.NV_OOB_PATHS in nresult):
1886
      for path_result in nresult[constants.NV_OOB_PATHS]:
1887
        self._ErrorIf(path_result, self.ENODEOOBPATH, node, path_result)
1888

    
1889
  def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
1890
    """Verifies and updates the node volume data.
1891

1892
    This function will update a L{NodeImage}'s internal structures
1893
    with data from the remote call.
1894

1895
    @type ninfo: L{objects.Node}
1896
    @param ninfo: the node to check
1897
    @param nresult: the remote results for the node
1898
    @param nimg: the node image object
1899
    @param vg_name: the configured VG name
1900

1901
    """
1902
    node = ninfo.name
1903
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1904

    
1905
    nimg.lvm_fail = True
1906
    lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
1907
    if vg_name is None:
1908
      pass
1909
    elif isinstance(lvdata, basestring):
1910
      _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
1911
               utils.SafeEncode(lvdata))
1912
    elif not isinstance(lvdata, dict):
1913
      _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
1914
    else:
1915
      nimg.volumes = lvdata
1916
      nimg.lvm_fail = False
1917

    
1918
  def _UpdateNodeInstances(self, ninfo, nresult, nimg):
1919
    """Verifies and updates the node instance list.
1920

1921
    If the listing was successful, then updates this node's instance
1922
    list. Otherwise, it marks the RPC call as failed for the instance
1923
    list key.
1924

1925
    @type ninfo: L{objects.Node}
1926
    @param ninfo: the node to check
1927
    @param nresult: the remote results for the node
1928
    @param nimg: the node image object
1929

1930
    """
1931
    idata = nresult.get(constants.NV_INSTANCELIST, None)
1932
    test = not isinstance(idata, list)
1933
    self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed"
1934
                  " (instancelist): %s", utils.SafeEncode(str(idata)))
1935
    if test:
1936
      nimg.hyp_fail = True
1937
    else:
1938
      nimg.instances = idata
1939

    
1940
  def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
1941
    """Verifies and computes a node information map
1942

1943
    @type ninfo: L{objects.Node}
1944
    @param ninfo: the node to check
1945
    @param nresult: the remote results for the node
1946
    @param nimg: the node image object
1947
    @param vg_name: the configured VG name
1948

1949
    """
1950
    node = ninfo.name
1951
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1952

    
1953
    # try to read free memory (from the hypervisor)
1954
    hv_info = nresult.get(constants.NV_HVINFO, None)
1955
    test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
1956
    _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
1957
    if not test:
1958
      try:
1959
        nimg.mfree = int(hv_info["memory_free"])
1960
      except (ValueError, TypeError):
1961
        _ErrorIf(True, self.ENODERPC, node,
1962
                 "node returned invalid nodeinfo, check hypervisor")
1963

    
1964
    # FIXME: devise a free space model for file based instances as well
1965
    if vg_name is not None:
1966
      test = (constants.NV_VGLIST not in nresult or
1967
              vg_name not in nresult[constants.NV_VGLIST])
1968
      _ErrorIf(test, self.ENODELVM, node,
1969
               "node didn't return data for the volume group '%s'"
1970
               " - it is either missing or broken", vg_name)
1971
      if not test:
1972
        try:
1973
          nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
1974
        except (ValueError, TypeError):
1975
          _ErrorIf(True, self.ENODERPC, node,
1976
                   "node returned invalid LVM info, check LVM status")
1977

    
1978
  def _CollectDiskInfo(self, nodelist, node_image, instanceinfo):
1979
    """Gets per-disk status information for all instances.
1980

1981
    @type nodelist: list of strings
1982
    @param nodelist: Node names
1983
    @type node_image: dict of (name, L{objects.Node})
1984
    @param node_image: Node objects
1985
    @type instanceinfo: dict of (name, L{objects.Instance})
1986
    @param instanceinfo: Instance objects
1987
    @rtype: {instance: {node: [(succes, payload)]}}
1988
    @return: a dictionary of per-instance dictionaries with nodes as
1989
        keys and disk information as values; the disk information is a
1990
        list of tuples (success, payload)
1991

1992
    """
1993
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1994

    
1995
    node_disks = {}
1996
    node_disks_devonly = {}
1997
    diskless_instances = set()
1998
    diskless = constants.DT_DISKLESS
1999

    
2000
    for nname in nodelist:
2001
      node_instances = list(itertools.chain(node_image[nname].pinst,
2002
                                            node_image[nname].sinst))
2003
      diskless_instances.update(inst for inst in node_instances
2004
                                if instanceinfo[inst].disk_template == diskless)
2005
      disks = [(inst, disk)
2006
               for inst in node_instances
2007
               for disk in instanceinfo[inst].disks]
2008

    
2009
      if not disks:
2010
        # No need to collect data
2011
        continue
2012

    
2013
      node_disks[nname] = disks
2014

    
2015
      # Creating copies as SetDiskID below will modify the objects and that can
2016
      # lead to incorrect data returned from nodes
2017
      devonly = [dev.Copy() for (_, dev) in disks]
2018

    
2019
      for dev in devonly:
2020
        self.cfg.SetDiskID(dev, nname)
2021

    
2022
      node_disks_devonly[nname] = devonly
2023

    
2024
    assert len(node_disks) == len(node_disks_devonly)
2025

    
2026
    # Collect data from all nodes with disks
2027
    result = self.rpc.call_blockdev_getmirrorstatus_multi(node_disks.keys(),
2028
                                                          node_disks_devonly)
2029

    
2030
    assert len(result) == len(node_disks)
2031

    
2032
    instdisk = {}
2033

    
2034
    for (nname, nres) in result.items():
2035
      disks = node_disks[nname]
2036

    
2037
      if nres.offline:
2038
        # No data from this node
2039
        data = len(disks) * [(False, "node offline")]
2040
      else:
2041
        msg = nres.fail_msg
2042
        _ErrorIf(msg, self.ENODERPC, nname,
2043
                 "while getting disk information: %s", msg)
2044
        if msg:
2045
          # No data from this node
2046
          data = len(disks) * [(False, msg)]
2047
        else:
2048
          data = []
2049
          for idx, i in enumerate(nres.payload):
2050
            if isinstance(i, (tuple, list)) and len(i) == 2:
2051
              data.append(i)
2052
            else:
2053
              logging.warning("Invalid result from node %s, entry %d: %s",
2054
                              nname, idx, i)
2055
              data.append((False, "Invalid result from the remote node"))
2056

    
2057
      for ((inst, _), status) in zip(disks, data):
2058
        instdisk.setdefault(inst, {}).setdefault(nname, []).append(status)
2059

    
2060
    # Add empty entries for diskless instances.
2061
    for inst in diskless_instances:
2062
      assert inst not in instdisk
2063
      instdisk[inst] = {}
2064

    
2065
    assert compat.all(len(statuses) == len(instanceinfo[inst].disks) and
2066
                      len(nnames) <= len(instanceinfo[inst].all_nodes) and
2067
                      compat.all(isinstance(s, (tuple, list)) and
2068
                                 len(s) == 2 for s in statuses)
2069
                      for inst, nnames in instdisk.items()
2070
                      for nname, statuses in nnames.items())
2071
    assert set(instdisk) == set(instanceinfo), "instdisk consistency failure"
2072

    
2073
    return instdisk
2074

    
2075
  def _VerifyHVP(self, hvp_data):
2076
    """Verifies locally the syntax of the hypervisor parameters.
2077

2078
    """
2079
    for item, hv_name, hv_params in hvp_data:
2080
      msg = ("hypervisor %s parameters syntax check (source %s): %%s" %
2081
             (item, hv_name))
2082
      try:
2083
        hv_class = hypervisor.GetHypervisor(hv_name)
2084
        utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2085
        hv_class.CheckParameterSyntax(hv_params)
2086
      except errors.GenericError, err:
2087
        self._ErrorIf(True, self.ECLUSTERCFG, None, msg % str(err))
2088

    
2089

    
2090
  def BuildHooksEnv(self):
2091
    """Build hooks env.
2092

2093
    Cluster-Verify hooks just ran in the post phase and their failure makes
2094
    the output be logged in the verify output and the verification to fail.
2095

2096
    """
2097
    all_nodes = self.cfg.GetNodeList()
2098
    env = {
2099
      "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
2100
      }
2101
    for node in self.cfg.GetAllNodesInfo().values():
2102
      env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
2103

    
2104
    return env, [], all_nodes
2105

    
2106
  def Exec(self, feedback_fn):
2107
    """Verify integrity of cluster, performing various test on nodes.
2108

2109
    """
2110
    # This method has too many local variables. pylint: disable-msg=R0914
2111
    self.bad = False
2112
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
2113
    verbose = self.op.verbose
2114
    self._feedback_fn = feedback_fn
2115
    feedback_fn("* Verifying global settings")
2116
    for msg in self.cfg.VerifyConfig():
2117
      _ErrorIf(True, self.ECLUSTERCFG, None, msg)
2118

    
2119
    # Check the cluster certificates
2120
    for cert_filename in constants.ALL_CERT_FILES:
2121
      (errcode, msg) = _VerifyCertificate(cert_filename)
2122
      _ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
2123

    
2124
    vg_name = self.cfg.GetVGName()
2125
    drbd_helper = self.cfg.GetDRBDHelper()
2126
    hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
2127
    cluster = self.cfg.GetClusterInfo()
2128
    nodelist = utils.NiceSort(self.cfg.GetNodeList())
2129
    nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
2130
    nodeinfo_byname = dict(zip(nodelist, nodeinfo))
2131
    instancelist = utils.NiceSort(self.cfg.GetInstanceList())
2132
    instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
2133
                        for iname in instancelist)
2134
    groupinfo = self.cfg.GetAllNodeGroupsInfo()
2135
    i_non_redundant = [] # Non redundant instances
2136
    i_non_a_balanced = [] # Non auto-balanced instances
2137
    n_offline = 0 # Count of offline nodes
2138
    n_drained = 0 # Count of nodes being drained
2139
    node_vol_should = {}
2140

    
2141
    # FIXME: verify OS list
2142
    # do local checksums
2143
    master_files = [constants.CLUSTER_CONF_FILE]
2144
    master_node = self.master_node = self.cfg.GetMasterNode()
2145
    master_ip = self.cfg.GetMasterIP()
2146

    
2147
    file_names = ssconf.SimpleStore().GetFileList()
2148
    file_names.extend(constants.ALL_CERT_FILES)
2149
    file_names.extend(master_files)
2150
    if cluster.modify_etc_hosts:
2151
      file_names.append(constants.ETC_HOSTS)
2152

    
2153
    local_checksums = utils.FingerprintFiles(file_names)
2154

    
2155
    # Compute the set of hypervisor parameters
2156
    hvp_data = []
2157
    for hv_name in hypervisors:
2158
      hvp_data.append(("cluster", hv_name, cluster.GetHVDefaults(hv_name)))
2159
    for os_name, os_hvp in cluster.os_hvp.items():
2160
      for hv_name, hv_params in os_hvp.items():
2161
        if not hv_params:
2162
          continue
2163
        full_params = cluster.GetHVDefaults(hv_name, os_name=os_name)
2164
        hvp_data.append(("os %s" % os_name, hv_name, full_params))
2165
    # TODO: collapse identical parameter values in a single one
2166
    for instance in instanceinfo.values():
2167
      if not instance.hvparams:
2168
        continue
2169
      hvp_data.append(("instance %s" % instance.name, instance.hypervisor,
2170
                       cluster.FillHV(instance)))
2171
    # and verify them locally
2172
    self._VerifyHVP(hvp_data)
2173

    
2174
    feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
2175
    node_verify_param = {
2176
      constants.NV_FILELIST: file_names,
2177
      constants.NV_NODELIST: [node.name for node in nodeinfo
2178
                              if not node.offline],
2179
      constants.NV_HYPERVISOR: hypervisors,
2180
      constants.NV_HVPARAMS: hvp_data,
2181
      constants.NV_NODENETTEST: [(node.name, node.primary_ip,
2182
                                  node.secondary_ip) for node in nodeinfo
2183
                                 if not node.offline],
2184
      constants.NV_INSTANCELIST: hypervisors,
2185
      constants.NV_VERSION: None,
2186
      constants.NV_HVINFO: self.cfg.GetHypervisorType(),
2187
      constants.NV_NODESETUP: None,
2188
      constants.NV_TIME: None,
2189
      constants.NV_MASTERIP: (master_node, master_ip),
2190
      constants.NV_OSLIST: None,
2191
      constants.NV_VMNODES: self.cfg.GetNonVmCapableNodeList(),
2192
      }
2193

    
2194
    if vg_name is not None:
2195
      node_verify_param[constants.NV_VGLIST] = None
2196
      node_verify_param[constants.NV_LVLIST] = vg_name
2197
      node_verify_param[constants.NV_PVLIST] = [vg_name]
2198
      node_verify_param[constants.NV_DRBDLIST] = None
2199

    
2200
    if drbd_helper:
2201
      node_verify_param[constants.NV_DRBDHELPER] = drbd_helper
2202

    
2203
    # bridge checks
2204
    # FIXME: this needs to be changed per node-group, not cluster-wide
2205
    bridges = set()
2206
    default_nicpp = cluster.nicparams[constants.PP_DEFAULT]
2207
    if default_nicpp[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2208
      bridges.add(default_nicpp[constants.NIC_LINK])
2209
    for instance in instanceinfo.values():
2210
      for nic in instance.nics:
2211
        full_nic = cluster.SimpleFillNIC(nic.nicparams)
2212
        if full_nic[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2213
          bridges.add(full_nic[constants.NIC_LINK])
2214

    
2215
    if bridges:
2216
      node_verify_param[constants.NV_BRIDGES] = list(bridges)
2217

    
2218
    # Build our expected cluster state
2219
    node_image = dict((node.name, self.NodeImage(offline=node.offline,
2220
                                                 name=node.name,
2221
                                                 vm_capable=node.vm_capable))
2222
                      for node in nodeinfo)
2223

    
2224
    # Gather OOB paths
2225
    oob_paths = []
2226
    for node in nodeinfo:
2227
      path = _SupportsOob(self.cfg, node)
2228
      if path and path not in oob_paths:
2229
        oob_paths.append(path)
2230

    
2231
    if oob_paths:
2232
      node_verify_param[constants.NV_OOB_PATHS] = oob_paths
2233

    
2234
    for instance in instancelist:
2235
      inst_config = instanceinfo[instance]
2236

    
2237
      for nname in inst_config.all_nodes:
2238
        if nname not in node_image:
2239
          # ghost node
2240
          gnode = self.NodeImage(name=nname)
2241
          gnode.ghost = True
2242
          node_image[nname] = gnode
2243

    
2244
      inst_config.MapLVsByNode(node_vol_should)
2245

    
2246
      pnode = inst_config.primary_node
2247
      node_image[pnode].pinst.append(instance)
2248

    
2249
      for snode in inst_config.secondary_nodes:
2250
        nimg = node_image[snode]
2251
        nimg.sinst.append(instance)
2252
        if pnode not in nimg.sbp:
2253
          nimg.sbp[pnode] = []
2254
        nimg.sbp[pnode].append(instance)
2255

    
2256
    # At this point, we have the in-memory data structures complete,
2257
    # except for the runtime information, which we'll gather next
2258

    
2259
    # Due to the way our RPC system works, exact response times cannot be
2260
    # guaranteed (e.g. a broken node could run into a timeout). By keeping the
2261
    # time before and after executing the request, we can at least have a time
2262
    # window.
2263
    nvinfo_starttime = time.time()
2264
    all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
2265
                                           self.cfg.GetClusterName())
2266
    nvinfo_endtime = time.time()
2267

    
2268
    all_drbd_map = self.cfg.ComputeDRBDMap()
2269

    
2270
    feedback_fn("* Gathering disk information (%s nodes)" % len(nodelist))
2271
    instdisk = self._CollectDiskInfo(nodelist, node_image, instanceinfo)
2272

    
2273
    feedback_fn("* Verifying node status")
2274

    
2275
    refos_img = None
2276

    
2277
    for node_i in nodeinfo:
2278
      node = node_i.name
2279
      nimg = node_image[node]
2280

    
2281
      if node_i.offline:
2282
        if verbose:
2283
          feedback_fn("* Skipping offline node %s" % (node,))
2284
        n_offline += 1
2285
        continue
2286

    
2287
      if node == master_node:
2288
        ntype = "master"
2289
      elif node_i.master_candidate:
2290
        ntype = "master candidate"
2291
      elif node_i.drained:
2292
        ntype = "drained"
2293
        n_drained += 1
2294
      else:
2295
        ntype = "regular"
2296
      if verbose:
2297
        feedback_fn("* Verifying node %s (%s)" % (node, ntype))
2298

    
2299
      msg = all_nvinfo[node].fail_msg
2300
      _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
2301
      if msg:
2302
        nimg.rpc_fail = True
2303
        continue
2304

    
2305
      nresult = all_nvinfo[node].payload
2306

    
2307
      nimg.call_ok = self._VerifyNode(node_i, nresult)
2308
      self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
2309
      self._VerifyNodeNetwork(node_i, nresult)
2310
      self._VerifyNodeFiles(node_i, nresult, file_names, local_checksums,
2311
                            master_files)
2312

    
2313
      self._VerifyOob(node_i, nresult)
2314

    
2315
      if nimg.vm_capable:
2316
        self._VerifyNodeLVM(node_i, nresult, vg_name)
2317
        self._VerifyNodeDrbd(node_i, nresult, instanceinfo, drbd_helper,
2318
                             all_drbd_map)
2319

    
2320
        self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
2321
        self._UpdateNodeInstances(node_i, nresult, nimg)
2322
        self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
2323
        self._UpdateNodeOS(node_i, nresult, nimg)
2324
        if not nimg.os_fail:
2325
          if refos_img is None:
2326
            refos_img = nimg
2327
          self._VerifyNodeOS(node_i, nimg, refos_img)
2328
        self._VerifyNodeBridges(node_i, nresult, bridges)
2329

    
2330
    feedback_fn("* Verifying instance status")
2331
    for instance in instancelist:
2332
      if verbose:
2333
        feedback_fn("* Verifying instance %s" % instance)
2334
      inst_config = instanceinfo[instance]
2335
      self._VerifyInstance(instance, inst_config, node_image,
2336
                           instdisk[instance])
2337
      inst_nodes_offline = []
2338

    
2339
      pnode = inst_config.primary_node
2340
      pnode_img = node_image[pnode]
2341
      _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
2342
               self.ENODERPC, pnode, "instance %s, connection to"
2343
               " primary node failed", instance)
2344

    
2345
      _ErrorIf(pnode_img.offline, self.EINSTANCEBADNODE, instance,
2346
               "instance lives on offline node %s", inst_config.primary_node)
2347

    
2348
      # If the instance is non-redundant we cannot survive losing its primary
2349
      # node, so we are not N+1 compliant. On the other hand we have no disk
2350
      # templates with more than one secondary so that situation is not well
2351
      # supported either.
2352
      # FIXME: does not support file-backed instances
2353
      if not inst_config.secondary_nodes:
2354
        i_non_redundant.append(instance)
2355

    
2356
      _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT,
2357
               instance, "instance has multiple secondary nodes: %s",
2358
               utils.CommaJoin(inst_config.secondary_nodes),
2359
               code=self.ETYPE_WARNING)
2360

    
2361
      if inst_config.disk_template in constants.DTS_NET_MIRROR:
2362
        pnode = inst_config.primary_node
2363
        instance_nodes = utils.NiceSort(inst_config.all_nodes)
2364
        instance_groups = {}
2365

    
2366
        for node in instance_nodes:
2367
          instance_groups.setdefault(nodeinfo_byname[node].group,
2368
                                     []).append(node)
2369

    
2370
        pretty_list = [
2371
          "%s (group %s)" % (utils.CommaJoin(nodes), groupinfo[group].name)
2372
          # Sort so that we always list the primary node first.
2373
          for group, nodes in sorted(instance_groups.items(),
2374
                                     key=lambda (_, nodes): pnode in nodes,
2375
                                     reverse=True)]
2376

    
2377
        self._ErrorIf(len(instance_groups) > 1, self.EINSTANCESPLITGROUPS,
2378
                      instance, "instance has primary and secondary nodes in"
2379
                      " different groups: %s", utils.CommaJoin(pretty_list),
2380
                      code=self.ETYPE_WARNING)
2381

    
2382
      if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
2383
        i_non_a_balanced.append(instance)
2384

    
2385
      for snode in inst_config.secondary_nodes:
2386
        s_img = node_image[snode]
2387
        _ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode,
2388
                 "instance %s, connection to secondary node failed", instance)
2389

    
2390
        if s_img.offline:
2391
          inst_nodes_offline.append(snode)
2392

    
2393
      # warn that the instance lives on offline nodes
2394
      _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
2395
               "instance has offline secondary node(s) %s",
2396
               utils.CommaJoin(inst_nodes_offline))
2397
      # ... or ghost/non-vm_capable nodes
2398
      for node in inst_config.all_nodes:
2399
        _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
2400
                 "instance lives on ghost node %s", node)
2401
        _ErrorIf(not node_image[node].vm_capable, self.EINSTANCEBADNODE,
2402
                 instance, "instance lives on non-vm_capable node %s", node)
2403

    
2404
    feedback_fn("* Verifying orphan volumes")
2405
    reserved = utils.FieldSet(*cluster.reserved_lvs)
2406
    self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
2407

    
2408
    feedback_fn("* Verifying orphan instances")
2409
    self._VerifyOrphanInstances(instancelist, node_image)
2410

    
2411
    if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
2412
      feedback_fn("* Verifying N+1 Memory redundancy")
2413
      self._VerifyNPlusOneMemory(node_image, instanceinfo)
2414

    
2415
    feedback_fn("* Other Notes")
2416
    if i_non_redundant:
2417
      feedback_fn("  - NOTICE: %d non-redundant instance(s) found."
2418
                  % len(i_non_redundant))
2419

    
2420
    if i_non_a_balanced:
2421
      feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
2422
                  % len(i_non_a_balanced))
2423

    
2424
    if n_offline:
2425
      feedback_fn("  - NOTICE: %d offline node(s) found." % n_offline)
2426

    
2427
    if n_drained:
2428
      feedback_fn("  - NOTICE: %d drained node(s) found." % n_drained)
2429

    
2430
    return not self.bad
2431

    
2432
  def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
2433
    """Analyze the post-hooks' result
2434

2435
    This method analyses the hook result, handles it, and sends some
2436
    nicely-formatted feedback back to the user.
2437

2438
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
2439
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
2440
    @param hooks_results: the results of the multi-node hooks rpc call
2441
    @param feedback_fn: function used send feedback back to the caller
2442
    @param lu_result: previous Exec result
2443
    @return: the new Exec result, based on the previous result
2444
        and hook results
2445

2446
    """
2447
    # We only really run POST phase hooks, and are only interested in
2448
    # their results
2449
    if phase == constants.HOOKS_PHASE_POST:
2450
      # Used to change hooks' output to proper indentation
2451
      feedback_fn("* Hooks Results")
2452
      assert hooks_results, "invalid result from hooks"
2453

    
2454
      for node_name in hooks_results:
2455
        res = hooks_results[node_name]
2456
        msg = res.fail_msg
2457
        test = msg and not res.offline
2458
        self._ErrorIf(test, self.ENODEHOOKS, node_name,
2459
                      "Communication failure in hooks execution: %s", msg)
2460
        if res.offline or msg:
2461
          # No need to investigate payload if node is offline or gave an error.
2462
          # override manually lu_result here as _ErrorIf only
2463
          # overrides self.bad
2464
          lu_result = 1
2465
          continue
2466
        for script, hkr, output in res.payload:
2467
          test = hkr == constants.HKR_FAIL
2468
          self._ErrorIf(test, self.ENODEHOOKS, node_name,
2469
                        "Script %s failed, output:", script)
2470
          if test:
2471
            output = self._HOOKS_INDENT_RE.sub('      ', output)
2472
            feedback_fn("%s" % output)
2473
            lu_result = 0
2474

    
2475
      return lu_result
2476

    
2477

    
2478
class LUClusterVerifyDisks(NoHooksLU):
2479
  """Verifies the cluster disks status.
2480

2481
  """
2482
  REQ_BGL = False
2483

    
2484
  def ExpandNames(self):
2485
    self.needed_locks = {
2486
      locking.LEVEL_NODE: locking.ALL_SET,
2487
      locking.LEVEL_INSTANCE: locking.ALL_SET,
2488
    }
2489
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
2490

    
2491
  def Exec(self, feedback_fn):
2492
    """Verify integrity of cluster disks.
2493

2494
    @rtype: tuple of three items
2495
    @return: a tuple of (dict of node-to-node_error, list of instances
2496
        which need activate-disks, dict of instance: (node, volume) for
2497
        missing volumes
2498

2499
    """
2500
    result = res_nodes, res_instances, res_missing = {}, [], {}
2501

    
2502
    nodes = utils.NiceSort(self.cfg.GetVmCapableNodeList())
2503
    instances = self.cfg.GetAllInstancesInfo().values()
2504

    
2505
    nv_dict = {}
2506
    for inst in instances:
2507
      inst_lvs = {}
2508
      if not inst.admin_up:
2509
        continue
2510
      inst.MapLVsByNode(inst_lvs)
2511
      # transform { iname: {node: [vol,],},} to {(node, vol): iname}
2512
      for node, vol_list in inst_lvs.iteritems():
2513
        for vol in vol_list:
2514
          nv_dict[(node, vol)] = inst
2515

    
2516
    if not nv_dict:
2517
      return result
2518

    
2519
    node_lvs = self.rpc.call_lv_list(nodes, [])
2520
    for node, node_res in node_lvs.items():
2521
      if node_res.offline:
2522
        continue
2523
      msg = node_res.fail_msg
2524
      if msg:
2525
        logging.warning("Error enumerating LVs on node %s: %s", node, msg)
2526
        res_nodes[node] = msg
2527
        continue
2528

    
2529
      lvs = node_res.payload
2530
      for lv_name, (_, _, lv_online) in lvs.items():
2531
        inst = nv_dict.pop((node, lv_name), None)
2532
        if (not lv_online and inst is not None
2533
            and inst.name not in res_instances):
2534
          res_instances.append(inst.name)
2535

    
2536
    # any leftover items in nv_dict are missing LVs, let's arrange the
2537
    # data better
2538
    for key, inst in nv_dict.iteritems():
2539
      if inst.name not in res_missing:
2540
        res_missing[inst.name] = []
2541
      res_missing[inst.name].append(key)
2542

    
2543
    return result
2544

    
2545

    
2546
class LUClusterRepairDiskSizes(NoHooksLU):
2547
  """Verifies the cluster disks sizes.
2548

2549
  """
2550
  REQ_BGL = False
2551

    
2552
  def ExpandNames(self):
2553
    if self.op.instances:
2554
      self.wanted_names = []
2555
      for name in self.op.instances:
2556
        full_name = _ExpandInstanceName(self.cfg, name)
2557
        self.wanted_names.append(full_name)
2558
      self.needed_locks = {
2559
        locking.LEVEL_NODE: [],
2560
        locking.LEVEL_INSTANCE: self.wanted_names,
2561
        }
2562
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2563
    else:
2564
      self.wanted_names = None
2565
      self.needed_locks = {
2566
        locking.LEVEL_NODE: locking.ALL_SET,
2567
        locking.LEVEL_INSTANCE: locking.ALL_SET,
2568
        }
2569
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
2570

    
2571
  def DeclareLocks(self, level):
2572
    if level == locking.LEVEL_NODE and self.wanted_names is not None:
2573
      self._LockInstancesNodes(primary_only=True)
2574

    
2575
  def CheckPrereq(self):
2576
    """Check prerequisites.
2577

2578
    This only checks the optional instance list against the existing names.
2579

2580
    """
2581
    if self.wanted_names is None:
2582
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
2583

    
2584
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
2585
                             in self.wanted_names]
2586

    
2587
  def _EnsureChildSizes(self, disk):
2588
    """Ensure children of the disk have the needed disk size.
2589

2590
    This is valid mainly for DRBD8 and fixes an issue where the
2591
    children have smaller disk size.
2592

2593
    @param disk: an L{ganeti.objects.Disk} object
2594

2595
    """
2596
    if disk.dev_type == constants.LD_DRBD8:
2597
      assert disk.children, "Empty children for DRBD8?"
2598
      fchild = disk.children[0]
2599
      mismatch = fchild.size < disk.size
2600
      if mismatch:
2601
        self.LogInfo("Child disk has size %d, parent %d, fixing",
2602
                     fchild.size, disk.size)
2603
        fchild.size = disk.size
2604

    
2605
      # and we recurse on this child only, not on the metadev
2606
      return self._EnsureChildSizes(fchild) or mismatch
2607
    else:
2608
      return False
2609

    
2610
  def Exec(self, feedback_fn):
2611
    """Verify the size of cluster disks.
2612

2613
    """
2614
    # TODO: check child disks too
2615
    # TODO: check differences in size between primary/secondary nodes
2616
    per_node_disks = {}
2617
    for instance in self.wanted_instances:
2618
      pnode = instance.primary_node
2619
      if pnode not in per_node_disks:
2620
        per_node_disks[pnode] = []
2621
      for idx, disk in enumerate(instance.disks):
2622
        per_node_disks[pnode].append((instance, idx, disk))
2623

    
2624
    changed = []
2625
    for node, dskl in per_node_disks.items():
2626
      newl = [v[2].Copy() for v in dskl]
2627
      for dsk in newl:
2628
        self.cfg.SetDiskID(dsk, node)
2629
      result = self.rpc.call_blockdev_getsize(node, newl)
2630
      if result.fail_msg:
2631
        self.LogWarning("Failure in blockdev_getsize call to node"
2632
                        " %s, ignoring", node)
2633
        continue
2634
      if len(result.payload) != len(dskl):
2635
        logging.warning("Invalid result from node %s: len(dksl)=%d,"
2636
                        " result.payload=%s", node, len(dskl), result.payload)
2637
        self.LogWarning("Invalid result from node %s, ignoring node results",
2638
                        node)
2639
        continue
2640
      for ((instance, idx, disk), size) in zip(dskl, result.payload):
2641
        if size is None:
2642
          self.LogWarning("Disk %d of instance %s did not return size"
2643
                          " information, ignoring", idx, instance.name)
2644
          continue
2645
        if not isinstance(size, (int, long)):
2646
          self.LogWarning("Disk %d of instance %s did not return valid"
2647
                          " size information, ignoring", idx, instance.name)
2648
          continue
2649
        size = size >> 20
2650
        if size != disk.size:
2651
          self.LogInfo("Disk %d of instance %s has mismatched size,"
2652
                       " correcting: recorded %d, actual %d", idx,
2653
                       instance.name, disk.size, size)
2654
          disk.size = size
2655
          self.cfg.Update(instance, feedback_fn)
2656
          changed.append((instance.name, idx, size))
2657
        if self._EnsureChildSizes(disk):
2658
          self.cfg.Update(instance, feedback_fn)
2659
          changed.append((instance.name, idx, disk.size))
2660
    return changed
2661

    
2662

    
2663
class LUClusterRename(LogicalUnit):
2664
  """Rename the cluster.
2665

2666
  """
2667
  HPATH = "cluster-rename"
2668
  HTYPE = constants.HTYPE_CLUSTER
2669

    
2670
  def BuildHooksEnv(self):
2671
    """Build hooks env.
2672

2673
    """
2674
    env = {
2675
      "OP_TARGET": self.cfg.GetClusterName(),
2676
      "NEW_NAME": self.op.name,
2677
      }
2678
    mn = self.cfg.GetMasterNode()
2679
    all_nodes = self.cfg.GetNodeList()
2680
    return env, [mn], all_nodes
2681

    
2682
  def CheckPrereq(self):
2683
    """Verify that the passed name is a valid one.
2684

2685
    """
2686
    hostname = netutils.GetHostname(name=self.op.name,
2687
                                    family=self.cfg.GetPrimaryIPFamily())
2688

    
2689
    new_name = hostname.name
2690
    self.ip = new_ip = hostname.ip
2691
    old_name = self.cfg.GetClusterName()
2692
    old_ip = self.cfg.GetMasterIP()
2693
    if new_name == old_name and new_ip == old_ip:
2694
      raise errors.OpPrereqError("Neither the name nor the IP address of the"
2695
                                 " cluster has changed",
2696
                                 errors.ECODE_INVAL)
2697
    if new_ip != old_ip:
2698
      if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
2699
        raise errors.OpPrereqError("The given cluster IP address (%s) is"
2700
                                   " reachable on the network" %
2701
                                   new_ip, errors.ECODE_NOTUNIQUE)
2702

    
2703
    self.op.name = new_name
2704

    
2705
  def Exec(self, feedback_fn):
2706
    """Rename the cluster.
2707

2708
    """
2709
    clustername = self.op.name
2710
    ip = self.ip
2711

    
2712
    # shutdown the master IP
2713
    master = self.cfg.GetMasterNode()
2714
    result = self.rpc.call_node_stop_master(master, False)
2715
    result.Raise("Could not disable the master role")
2716

    
2717
    try:
2718
      cluster = self.cfg.GetClusterInfo()
2719
      cluster.cluster_name = clustername
2720
      cluster.master_ip = ip
2721
      self.cfg.Update(cluster, feedback_fn)
2722

    
2723
      # update the known hosts file
2724
      ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
2725
      node_list = self.cfg.GetOnlineNodeList()
2726
      try:
2727
        node_list.remove(master)
2728
      except ValueError:
2729
        pass
2730
      _UploadHelper(self, node_list, constants.SSH_KNOWN_HOSTS_FILE)
2731
    finally:
2732
      result = self.rpc.call_node_start_master(master, False, False)
2733
      msg = result.fail_msg
2734
      if msg:
2735
        self.LogWarning("Could not re-enable the master role on"
2736
                        " the master, please restart manually: %s", msg)
2737

    
2738
    return clustername
2739

    
2740

    
2741
class LUClusterSetParams(LogicalUnit):
2742
  """Change the parameters of the cluster.
2743

2744
  """
2745
  HPATH = "cluster-modify"
2746
  HTYPE = constants.HTYPE_CLUSTER
2747
  REQ_BGL = False
2748

    
2749
  def CheckArguments(self):
2750
    """Check parameters
2751

2752
    """
2753
    if self.op.uid_pool:
2754
      uidpool.CheckUidPool(self.op.uid_pool)
2755

    
2756
    if self.op.add_uids:
2757
      uidpool.CheckUidPool(self.op.add_uids)
2758

    
2759
    if self.op.remove_uids:
2760
      uidpool.CheckUidPool(self.op.remove_uids)
2761

    
2762
  def ExpandNames(self):
2763
    # FIXME: in the future maybe other cluster params won't require checking on
2764
    # all nodes to be modified.
2765
    self.needed_locks = {
2766
      locking.LEVEL_NODE: locking.ALL_SET,
2767
    }
2768
    self.share_locks[locking.LEVEL_NODE] = 1
2769

    
2770
  def BuildHooksEnv(self):
2771
    """Build hooks env.
2772

2773
    """
2774
    env = {
2775
      "OP_TARGET": self.cfg.GetClusterName(),
2776
      "NEW_VG_NAME": self.op.vg_name,
2777
      }
2778
    mn = self.cfg.GetMasterNode()
2779
    return env, [mn], [mn]
2780

    
2781
  def CheckPrereq(self):
2782
    """Check prerequisites.
2783

2784
    This checks whether the given params don't conflict and
2785
    if the given volume group is valid.
2786

2787
    """
2788
    if self.op.vg_name is not None and not self.op.vg_name:
2789
      if self.cfg.HasAnyDiskOfType(constants.LD_LV):
2790
        raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based"
2791
                                   " instances exist", errors.ECODE_INVAL)
2792

    
2793
    if self.op.drbd_helper is not None and not self.op.drbd_helper:
2794
      if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8):
2795
        raise errors.OpPrereqError("Cannot disable drbd helper while"
2796
                                   " drbd-based instances exist",
2797
                                   errors.ECODE_INVAL)
2798

    
2799
    node_list = self.acquired_locks[locking.LEVEL_NODE]
2800

    
2801
    # if vg_name not None, checks given volume group on all nodes
2802
    if self.op.vg_name:
2803
      vglist = self.rpc.call_vg_list(node_list)
2804
      for node in node_list:
2805
        msg = vglist[node].fail_msg
2806
        if msg:
2807
          # ignoring down node
2808
          self.LogWarning("Error while gathering data on node %s"
2809
                          " (ignoring node): %s", node, msg)
2810
          continue
2811
        vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
2812
                                              self.op.vg_name,
2813
                                              constants.MIN_VG_SIZE)
2814
        if vgstatus:
2815
          raise errors.OpPrereqError("Error on node '%s': %s" %
2816
                                     (node, vgstatus), errors.ECODE_ENVIRON)
2817

    
2818
    if self.op.drbd_helper:
2819
      # checks given drbd helper on all nodes
2820
      helpers = self.rpc.call_drbd_helper(node_list)
2821
      for node in node_list:
2822
        ninfo = self.cfg.GetNodeInfo(node)
2823
        if ninfo.offline:
2824
          self.LogInfo("Not checking drbd helper on offline node %s", node)
2825
          continue
2826
        msg = helpers[node].fail_msg
2827
        if msg:
2828
          raise errors.OpPrereqError("Error checking drbd helper on node"
2829
                                     " '%s': %s" % (node, msg),
2830
                                     errors.ECODE_ENVIRON)
2831
        node_helper = helpers[node].payload
2832
        if node_helper != self.op.drbd_helper:
2833
          raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" %
2834
                                     (node, node_helper), errors.ECODE_ENVIRON)
2835

    
2836
    self.cluster = cluster = self.cfg.GetClusterInfo()
2837
    # validate params changes
2838
    if self.op.beparams:
2839
      utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
2840
      self.new_beparams = cluster.SimpleFillBE(self.op.beparams)
2841

    
2842
    if self.op.ndparams:
2843
      utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
2844
      self.new_ndparams = cluster.SimpleFillND(self.op.ndparams)
2845

    
2846
      # TODO: we need a more general way to handle resetting
2847
      # cluster-level parameters to default values
2848
      if self.new_ndparams["oob_program"] == "":
2849
        self.new_ndparams["oob_program"] = \
2850
            constants.NDC_DEFAULTS[constants.ND_OOB_PROGRAM]
2851

    
2852
    if self.op.nicparams:
2853
      utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
2854
      self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams)
2855
      objects.NIC.CheckParameterSyntax(self.new_nicparams)
2856
      nic_errors = []
2857

    
2858
      # check all instances for consistency
2859
      for instance in self.cfg.GetAllInstancesInfo().values():
2860
        for nic_idx, nic in enumerate(instance.nics):
2861
          params_copy = copy.deepcopy(nic.nicparams)
2862
          params_filled = objects.FillDict(self.new_nicparams, params_copy)
2863

    
2864
          # check parameter syntax
2865
          try:
2866
            objects.NIC.CheckParameterSyntax(params_filled)
2867
          except errors.ConfigurationError, err:
2868
            nic_errors.append("Instance %s, nic/%d: %s" %
2869
                              (instance.name, nic_idx, err))
2870

    
2871
          # if we're moving instances to routed, check that they have an ip
2872
          target_mode = params_filled[constants.NIC_MODE]
2873
          if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
2874
            nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
2875
                              (instance.name, nic_idx))
2876
      if nic_errors:
2877
        raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
2878
                                   "\n".join(nic_errors))
2879

    
2880
    # hypervisor list/parameters
2881
    self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
2882
    if self.op.hvparams:
2883
      for hv_name, hv_dict in self.op.hvparams.items():
2884
        if hv_name not in self.new_hvparams:
2885
          self.new_hvparams[hv_name] = hv_dict
2886
        else:
2887
          self.new_hvparams[hv_name].update(hv_dict)
2888

    
2889
    # os hypervisor parameters
2890
    self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
2891
    if self.op.os_hvp:
2892
      for os_name, hvs in self.op.os_hvp.items():
2893
        if os_name not in self.new_os_hvp:
2894
          self.new_os_hvp[os_name] = hvs
2895
        else:
2896
          for hv_name, hv_dict in hvs.items():
2897
            if hv_name not in self.new_os_hvp[os_name]:
2898
              self.new_os_hvp[os_name][hv_name] = hv_dict
2899
            else:
2900
              self.new_os_hvp[os_name][hv_name].update(hv_dict)
2901

    
2902
    # os parameters
2903
    self.new_osp = objects.FillDict(cluster.osparams, {})
2904
    if self.op.osparams:
2905
      for os_name, osp in self.op.osparams.items():
2906
        if os_name not in self.new_osp:
2907
          self.new_osp[os_name] = {}
2908

    
2909
        self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp,
2910
                                                  use_none=True)
2911

    
2912
        if not self.new_osp[os_name]:
2913
          # we removed all parameters
2914
          del self.new_osp[os_name]
2915
        else:
2916
          # check the parameter validity (remote check)
2917
          _CheckOSParams(self, False, [self.cfg.GetMasterNode()],
2918
                         os_name, self.new_osp[os_name])
2919

    
2920
    # changes to the hypervisor list
2921
    if self.op.enabled_hypervisors is not None:
2922
      self.hv_list = self.op.enabled_hypervisors
2923
      for hv in self.hv_list:
2924
        # if the hypervisor doesn't already exist in the cluster
2925
        # hvparams, we initialize it to empty, and then (in both
2926
        # cases) we make sure to fill the defaults, as we might not
2927
        # have a complete defaults list if the hypervisor wasn't
2928
        # enabled before
2929
        if hv not in new_hvp:
2930
          new_hvp[hv] = {}
2931
        new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
2932
        utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
2933
    else:
2934
      self.hv_list = cluster.enabled_hypervisors
2935

    
2936
    if self.op.hvparams or self.op.enabled_hypervisors is not None:
2937
      # either the enabled list has changed, or the parameters have, validate
2938
      for hv_name, hv_params in self.new_hvparams.items():
2939
        if ((self.op.hvparams and hv_name in self.op.hvparams) or
2940
            (self.op.enabled_hypervisors and
2941
             hv_name in self.op.enabled_hypervisors)):
2942
          # either this is a new hypervisor, or its parameters have changed
2943
          hv_class = hypervisor.GetHypervisor(hv_name)
2944
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2945
          hv_class.CheckParameterSyntax(hv_params)
2946
          _CheckHVParams(self, node_list, hv_name, hv_params)
2947

    
2948
    if self.op.os_hvp:
2949
      # no need to check any newly-enabled hypervisors, since the
2950
      # defaults have already been checked in the above code-block
2951
      for os_name, os_hvp in self.new_os_hvp.items():
2952
        for hv_name, hv_params in os_hvp.items():
2953
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2954
          # we need to fill in the new os_hvp on top of the actual hv_p
2955
          cluster_defaults = self.new_hvparams.get(hv_name, {})
2956
          new_osp = objects.FillDict(cluster_defaults, hv_params)
2957
          hv_class = hypervisor.GetHypervisor(hv_name)
2958
          hv_class.CheckParameterSyntax(new_osp)
2959
          _CheckHVParams(self, node_list, hv_name, new_osp)
2960

    
2961
    if self.op.default_iallocator:
2962
      alloc_script = utils.FindFile(self.op.default_iallocator,
2963
                                    constants.IALLOCATOR_SEARCH_PATH,
2964
                                    os.path.isfile)
2965
      if alloc_script is None:
2966
        raise errors.OpPrereqError("Invalid default iallocator script '%s'"
2967
                                   " specified" % self.op.default_iallocator,
2968
                                   errors.ECODE_INVAL)
2969

    
2970
  def Exec(self, feedback_fn):
2971
    """Change the parameters of the cluster.
2972

2973
    """
2974
    if self.op.vg_name is not None:
2975
      new_volume = self.op.vg_name
2976
      if not new_volume:
2977
        new_volume = None
2978
      if new_volume != self.cfg.GetVGName():
2979
        self.cfg.SetVGName(new_volume)
2980
      else:
2981
        feedback_fn("Cluster LVM configuration already in desired"
2982
                    " state, not changing")
2983
    if self.op.drbd_helper is not None:
2984
      new_helper = self.op.drbd_helper
2985
      if not new_helper:
2986
        new_helper = None
2987
      if new_helper != self.cfg.GetDRBDHelper():
2988
        self.cfg.SetDRBDHelper(new_helper)
2989
      else:
2990
        feedback_fn("Cluster DRBD helper already in desired state,"
2991
                    " not changing")
2992
    if self.op.hvparams:
2993
      self.cluster.hvparams = self.new_hvparams
2994
    if self.op.os_hvp:
2995
      self.cluster.os_hvp = self.new_os_hvp
2996
    if self.op.enabled_hypervisors is not None:
2997
      self.cluster.hvparams = self.new_hvparams
2998
      self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
2999
    if self.op.beparams:
3000
      self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
3001
    if self.op.nicparams:
3002
      self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
3003
    if self.op.osparams:
3004
      self.cluster.osparams = self.new_osp
3005
    if self.op.ndparams:
3006
      self.cluster.ndparams = self.new_ndparams
3007

    
3008
    if self.op.candidate_pool_size is not None:
3009
      self.cluster.candidate_pool_size = self.op.candidate_pool_size
3010
      # we need to update the pool size here, otherwise the save will fail
3011
      _AdjustCandidatePool(self, [])
3012

    
3013
    if self.op.maintain_node_health is not None:
3014
      self.cluster.maintain_node_health = self.op.maintain_node_health
3015

    
3016
    if self.op.prealloc_wipe_disks is not None:
3017
      self.cluster.prealloc_wipe_disks = self.op.prealloc_wipe_disks
3018

    
3019
    if self.op.add_uids is not None:
3020
      uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
3021

    
3022
    if self.op.remove_uids is not None:
3023
      uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
3024

    
3025
    if self.op.uid_pool is not None:
3026
      self.cluster.uid_pool = self.op.uid_pool
3027

    
3028
    if self.op.default_iallocator is not None:
3029
      self.cluster.default_iallocator = self.op.default_iallocator
3030

    
3031
    if self.op.reserved_lvs is not None:
3032
      self.cluster.reserved_lvs = self.op.reserved_lvs
3033

    
3034
    def helper_os(aname, mods, desc):
3035
      desc += " OS list"
3036
      lst = getattr(self.cluster, aname)
3037
      for key, val in mods:
3038
        if key == constants.DDM_ADD:
3039
          if val in lst:
3040
            feedback_fn("OS %s already in %s, ignoring" % (val, desc))
3041
          else:
3042
            lst.append(val)
3043
        elif key == constants.DDM_REMOVE:
3044
          if val in lst:
3045
            lst.remove(val)
3046
          else:
3047
            feedback_fn("OS %s not found in %s, ignoring" % (val, desc))
3048
        else:
3049
          raise errors.ProgrammerError("Invalid modification '%s'" % key)
3050

    
3051
    if self.op.hidden_os:
3052
      helper_os("hidden_os", self.op.hidden_os, "hidden")
3053

    
3054
    if self.op.blacklisted_os:
3055
      helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted")
3056

    
3057
    if self.op.master_netdev:
3058
      master = self.cfg.GetMasterNode()
3059
      feedback_fn("Shutting down master ip on the current netdev (%s)" %
3060
                  self.cluster.master_netdev)
3061
      result = self.rpc.call_node_stop_master(master, False)
3062
      result.Raise("Could not disable the master ip")
3063
      feedback_fn("Changing master_netdev from %s to %s" %
3064
                  (self.cluster.master_netdev, self.op.master_netdev))
3065
      self.cluster.master_netdev = self.op.master_netdev
3066

    
3067
    self.cfg.Update(self.cluster, feedback_fn)
3068

    
3069
    if self.op.master_netdev:
3070
      feedback_fn("Starting the master ip on the new master netdev (%s)" %
3071
                  self.op.master_netdev)
3072
      result = self.rpc.call_node_start_master(master, False, False)
3073
      if result.fail_msg:
3074
        self.LogWarning("Could not re-enable the master ip on"
3075
                        " the master, please restart manually: %s",
3076
                        result.fail_msg)
3077

    
3078

    
3079
def _UploadHelper(lu, nodes, fname):
3080
  """Helper for uploading a file and showing warnings.
3081

3082
  """
3083
  if os.path.exists(fname):
3084
    result = lu.rpc.call_upload_file(nodes, fname)
3085
    for to_node, to_result in result.items():
3086
      msg = to_result.fail_msg
3087
      if msg:
3088
        msg = ("Copy of file %s to node %s failed: %s" %
3089
               (fname, to_node, msg))
3090
        lu.proc.LogWarning(msg)
3091

    
3092

    
3093
def _RedistributeAncillaryFiles(lu, additional_nodes=None, additional_vm=True):
3094
  """Distribute additional files which are part of the cluster configuration.
3095

3096
  ConfigWriter takes care of distributing the config and ssconf files, but
3097
  there are more files which should be distributed to all nodes. This function
3098
  makes sure those are copied.
3099

3100
  @param lu: calling logical unit
3101
  @param additional_nodes: list of nodes not in the config to distribute to
3102
  @type additional_vm: boolean
3103
  @param additional_vm: whether the additional nodes are vm-capable or not
3104

3105
  """
3106
  # 1. Gather target nodes
3107
  myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
3108
  dist_nodes = lu.cfg.GetOnlineNodeList()
3109
  nvm_nodes = lu.cfg.GetNonVmCapableNodeList()
3110
  vm_nodes = [name for name in dist_nodes if name not in nvm_nodes]
3111
  if additional_nodes is not None:
3112
    dist_nodes.extend(additional_nodes)
3113
    if additional_vm:
3114
      vm_nodes.extend(additional_nodes)
3115
  if myself.name in dist_nodes:
3116
    dist_nodes.remove(myself.name)
3117
  if myself.name in vm_nodes:
3118
    vm_nodes.remove(myself.name)
3119

    
3120
  # 2. Gather files to distribute
3121
  dist_files = set([constants.ETC_HOSTS,
3122
                    constants.SSH_KNOWN_HOSTS_FILE,
3123
                    constants.RAPI_CERT_FILE,
3124
                    constants.RAPI_USERS_FILE,
3125
                    constants.CONFD_HMAC_KEY,
3126
                    constants.CLUSTER_DOMAIN_SECRET_FILE,
3127
                   ])
3128

    
3129
  vm_files = set()
3130
  enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
3131
  for hv_name in enabled_hypervisors:
3132
    hv_class = hypervisor.GetHypervisor(hv_name)
3133
    vm_files.update(hv_class.GetAncillaryFiles())
3134

    
3135
  # 3. Perform the files upload
3136
  for fname in dist_files:
3137
    _UploadHelper(lu, dist_nodes, fname)
3138
  for fname in vm_files:
3139
    _UploadHelper(lu, vm_nodes, fname)
3140

    
3141

    
3142
class LUClusterRedistConf(NoHooksLU):
3143
  """Force the redistribution of cluster configuration.
3144

3145
  This is a very simple LU.
3146

3147
  """
3148
  REQ_BGL = False
3149

    
3150
  def ExpandNames(self):
3151
    self.needed_locks = {
3152
      locking.LEVEL_NODE: locking.ALL_SET,
3153
    }
3154
    self.share_locks[locking.LEVEL_NODE] = 1
3155

    
3156
  def Exec(self, feedback_fn):
3157
    """Redistribute the configuration.
3158

3159
    """
3160
    self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
3161
    _RedistributeAncillaryFiles(self)
3162

    
3163

    
3164
def _WaitForSync(lu, instance, disks=None, oneshot=False):
3165
  """Sleep and poll for an instance's disk to sync.
3166

3167
  """
3168
  if not instance.disks or disks is not None and not disks:
3169
    return True
3170

    
3171
  disks = _ExpandCheckDisks(instance, disks)
3172

    
3173
  if not oneshot:
3174
    lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
3175

    
3176
  node = instance.primary_node
3177

    
3178
  for dev in disks:
3179
    lu.cfg.SetDiskID(dev, node)
3180

    
3181
  # TODO: Convert to utils.Retry
3182

    
3183
  retries = 0
3184
  degr_retries = 10 # in seconds, as we sleep 1 second each time
3185
  while True:
3186
    max_time = 0
3187
    done = True
3188
    cumul_degraded = False
3189
    rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
3190
    msg = rstats.fail_msg
3191
    if msg:
3192
      lu.LogWarning("Can't get any data from node %s: %s", node, msg)
3193
      retries += 1
3194
      if retries >= 10:
3195
        raise errors.RemoteError("Can't contact node %s for mirror data,"
3196
                                 " aborting." % node)
3197
      time.sleep(6)
3198
      continue
3199
    rstats = rstats.payload
3200
    retries = 0
3201
    for i, mstat in enumerate(rstats):
3202
      if mstat is None:
3203
        lu.LogWarning("Can't compute data for node %s/%s",
3204
                           node, disks[i].iv_name)
3205
        continue
3206

    
3207
      cumul_degraded = (cumul_degraded or
3208
                        (mstat.is_degraded and mstat.sync_percent is None))
3209
      if mstat.sync_percent is not None:
3210
        done = False
3211
        if mstat.estimated_time is not None:
3212
          rem_time = ("%s remaining (estimated)" %
3213
                      utils.FormatSeconds(mstat.estimated_time))
3214
          max_time = mstat.estimated_time
3215
        else:
3216
          rem_time = "no time estimate"
3217
        lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
3218
                        (disks[i].iv_name, mstat.sync_percent, rem_time))
3219

    
3220
    # if we're done but degraded, let's do a few small retries, to
3221
    # make sure we see a stable and not transient situation; therefore
3222
    # we force restart of the loop
3223
    if (done or oneshot) and cumul_degraded and degr_retries > 0:
3224
      logging.info("Degraded disks found, %d retries left", degr_retries)
3225
      degr_retries -= 1
3226
      time.sleep(1)
3227
      continue
3228

    
3229
    if done or oneshot:
3230
      break
3231

    
3232
    time.sleep(min(60, max_time))
3233

    
3234
  if done:
3235
    lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
3236
  return not cumul_degraded
3237

    
3238

    
3239
def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
3240
  """Check that mirrors are not degraded.
3241

3242
  The ldisk parameter, if True, will change the test from the
3243
  is_degraded attribute (which represents overall non-ok status for
3244
  the device(s)) to the ldisk (representing the local storage status).
3245

3246
  """
3247
  lu.cfg.SetDiskID(dev, node)
3248

    
3249
  result = True
3250

    
3251
  if on_primary or dev.AssembleOnSecondary():
3252
    rstats = lu.rpc.call_blockdev_find(node, dev)
3253
    msg = rstats.fail_msg
3254
    if msg:
3255
      lu.LogWarning("Can't find disk on node %s: %s", node, msg)
3256
      result = False
3257
    elif not rstats.payload:
3258
      lu.LogWarning("Can't find disk on node %s", node)
3259
      result = False
3260
    else:
3261
      if ldisk:
3262
        result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
3263
      else:
3264
        result = result and not rstats.payload.is_degraded
3265

    
3266
  if dev.children:
3267
    for child in dev.children:
3268
      result = result and _CheckDiskConsistency(lu, child, node, on_primary)
3269

    
3270
  return result
3271

    
3272

    
3273
class LUOobCommand(NoHooksLU):
3274
  """Logical unit for OOB handling.
3275

3276
  """
3277
  REG_BGL = False
3278

    
3279
  def CheckPrereq(self):
3280
    """Check prerequisites.
3281

3282
    This checks:
3283
     - the node exists in the configuration
3284
     - OOB is supported
3285

3286
    Any errors are signaled by raising errors.OpPrereqError.
3287

3288
    """
3289
    self.nodes = []
3290
    for node_name in self.op.node_names:
3291
      node = self.cfg.GetNodeInfo(node_name)
3292

    
3293
      if node is None:
3294
        raise errors.OpPrereqError("Node %s not found" % node_name,
3295
                                   errors.ECODE_NOENT)
3296
      else:
3297
        self.nodes.append(node)
3298

    
3299
      if (self.op.command == constants.OOB_POWER_OFF and not node.offline):
3300
        raise errors.OpPrereqError(("Cannot power off node %s because it is"
3301
                                    " not marked offline") % node_name,
3302
                                   errors.ECODE_STATE)
3303

    
3304
  def ExpandNames(self):
3305
    """Gather locks we need.
3306

3307
    """
3308
    if self.op.node_names:
3309
      self.op.node_names = [_ExpandNodeName(self.cfg, name)
3310
                            for name in self.op.node_names]
3311
    else:
3312
      self.op.node_names = self.cfg.GetNodeList()
3313

    
3314
    self.needed_locks = {
3315
      locking.LEVEL_NODE: self.op.node_names,
3316
      }
3317

    
3318
  def Exec(self, feedback_fn):
3319
    """Execute OOB and return result if we expect any.
3320

3321
    """
3322
    master_node = self.cfg.GetMasterNode()
3323
    ret = []
3324

    
3325
    for node in self.nodes:
3326
      node_entry = [(constants.RS_NORMAL, node.name)]
3327
      ret.append(node_entry)
3328

    
3329
      oob_program = _SupportsOob(self.cfg, node)
3330

    
3331
      if not oob_program:
3332
        node_entry.append((constants.RS_UNAVAIL, None))
3333
        continue
3334

    
3335
      logging.info("Executing out-of-band command '%s' using '%s' on %s",
3336
                   self.op.command, oob_program, node.name)
3337
      result = self.rpc.call_run_oob(master_node, oob_program,
3338
                                     self.op.command, node.name,
3339
                                     self.op.timeout)
3340

    
3341
      if result.fail_msg:
3342
        self.LogWarning("On node '%s' out-of-band RPC failed with: %s",
3343
                        node.name, result.fail_msg)
3344
        node_entry.append((constants.RS_NODATA, None))
3345
      else:
3346
        try:
3347
          self._CheckPayload(result)
3348
        except errors.OpExecError, err:
3349
          self.LogWarning("The payload returned by '%s' is not valid: %s",
3350
                          node.name, err)
3351
          node_entry.append((constants.RS_NODATA, None))
3352
        else:
3353
          if self.op.command == constants.OOB_HEALTH:
3354
            # For health we should log important events
3355
            for item, status in result.payload:
3356
              if status in [constants.OOB_STATUS_WARNING,
3357
                            constants.OOB_STATUS_CRITICAL]:
3358
                self.LogWarning("On node '%s' item '%s' has status '%s'",
3359
                                node.name, item, status)
3360

    
3361
          if self.op.command == constants.OOB_POWER_ON:
3362
            node.powered = True
3363
          elif self.op.command == constants.OOB_POWER_OFF:
3364
            node.powered = False
3365
          elif self.op.command == constants.OOB_POWER_STATUS:
3366
            powered = result.payload[constants.OOB_POWER_STATUS_POWERED]
3367
            if powered != node.powered:
3368
              logging.warning(("Recorded power state (%s) of node '%s' does not"
3369
                               " match actual power state (%s)"), node.powered,
3370
                              node.name, powered)
3371

    
3372
          # For configuration changing commands we should update the node
3373
          if self.op.command in (constants.OOB_POWER_ON,
3374
                                 constants.OOB_POWER_OFF):
3375
            self.cfg.Update(node, feedback_fn)
3376

    
3377
          node_entry.append((constants.RS_NORMAL, result.payload))
3378

    
3379
    return ret
3380

    
3381
  def _CheckPayload(self, result):
3382
    """Checks if the payload is valid.
3383

3384
    @param result: RPC result
3385
    @raises errors.OpExecError: If payload is not valid
3386

3387
    """
3388
    errs = []
3389
    if self.op.command == constants.OOB_HEALTH:
3390
      if not isinstance(result.payload, list):
3391
        errs.append("command 'health' is expected to return a list but got %s" %
3392
                    type(result.payload))
3393
      else:
3394
        for item, status in result.payload:
3395
          if status not in constants.OOB_STATUSES:
3396
            errs.append("health item '%s' has invalid status '%s'" %
3397
                        (item, status))
3398

    
3399
    if self.op.command == constants.OOB_POWER_STATUS:
3400
      if not isinstance(result.payload, dict):
3401
        errs.append("power-status is expected to return a dict but got %s" %
3402
                    type(result.payload))
3403

    
3404
    if self.op.command in [
3405
        constants.OOB_POWER_ON,
3406
        constants.OOB_POWER_OFF,
3407
        constants.OOB_POWER_CYCLE,
3408
        ]:
3409
      if result.payload is not None:
3410
        errs.append("%s is expected to not return payload but got '%s'" %
3411
                    (self.op.command, result.payload))
3412

    
3413
    if errs:
3414
      raise errors.OpExecError("Check of out-of-band payload failed due to %s" %
3415
                               utils.CommaJoin(errs))
3416

    
3417

    
3418

    
3419
class LUOsDiagnose(NoHooksLU):
3420
  """Logical unit for OS diagnose/query.
3421

3422
  """
3423
  REQ_BGL = False
3424
  _HID = "hidden"
3425
  _BLK = "blacklisted"
3426
  _VLD = "valid"
3427
  _FIELDS_STATIC = utils.FieldSet()
3428
  _FIELDS_DYNAMIC = utils.FieldSet("name", _VLD, "node_status", "variants",
3429
                                   "parameters", "api_versions", _HID, _BLK)
3430

    
3431
  def CheckArguments(self):
3432
    if self.op.names:
3433
      raise errors.OpPrereqError("Selective OS query not supported",
3434
                                 errors.ECODE_INVAL)
3435

    
3436
    _CheckOutputFields(static=self._FIELDS_STATIC,
3437
                       dynamic=self._FIELDS_DYNAMIC,
3438
                       selected=self.op.output_fields)
3439

    
3440
  def ExpandNames(self):
3441
    # Lock all nodes, in shared mode
3442
    # Temporary removal of locks, should be reverted later
3443
    # TODO: reintroduce locks when they are lighter-weight
3444
    self.needed_locks = {}
3445
    #self.share_locks[locking.LEVEL_NODE] = 1
3446
    #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3447

    
3448
  @staticmethod
3449
  def _DiagnoseByOS(rlist):
3450
    """Remaps a per-node return list into an a per-os per-node dictionary
3451

3452
    @param rlist: a map with node names as keys and OS objects as values
3453

3454
    @rtype: dict
3455
    @return: a dictionary with osnames as keys and as value another
3456
        map, with nodes as keys and tuples of (path, status, diagnose,
3457
        variants, parameters, api_versions) as values, eg::
3458

3459
          {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
3460
                                     (/srv/..., False, "invalid api")],
3461
                           "node2": [(/srv/..., True, "", [], [])]}
3462
          }
3463

3464
    """
3465
    all_os = {}
3466
    # we build here the list of nodes that didn't fail the RPC (at RPC
3467
    # level), so that nodes with a non-responding node daemon don't
3468
    # make all OSes invalid
3469
    good_nodes = [node_name for node_name in rlist
3470
                  if not rlist[node_name].fail_msg]
3471
    for node_name, nr in rlist.items():
3472
      if nr.fail_msg or not nr.payload:
3473
        continue
3474
      for (name, path, status, diagnose, variants,
3475
           params, api_versions) in nr.payload:
3476
        if name not in all_os:
3477
          # build a list of nodes for this os containing empty lists
3478
          # for each node in node_list
3479
          all_os[name] = {}
3480
          for nname in good_nodes:
3481
            all_os[name][nname] = []
3482
        # convert params from [name, help] to (name, help)
3483
        params = [tuple(v) for v in params]
3484
        all_os[name][node_name].append((path, status, diagnose,
3485
                                        variants, params, api_versions))
3486
    return all_os
3487

    
3488
  def Exec(self, feedback_fn):
3489
    """Compute the list of OSes.
3490

3491
    """
3492
    valid_nodes = [node.name
3493
                   for node in self.cfg.GetAllNodesInfo().values()
3494
                   if not node.offline and node.vm_capable]
3495
    node_data = self.rpc.call_os_diagnose(valid_nodes)
3496
    pol = self._DiagnoseByOS(node_data)
3497
    output = []
3498
    cluster = self.cfg.GetClusterInfo()
3499

    
3500
    for os_name in utils.NiceSort(pol.keys()):
3501
      os_data = pol[os_name]
3502
      row = []
3503
      valid = True
3504
      (variants, params, api_versions) = null_state = (set(), set(), set())
3505
      for idx, osl in enumerate(os_data.values()):
3506
        valid = bool(valid and osl and osl[0][1])
3507
        if not valid:
3508
          (variants, params, api_versions) = null_state
3509
          break
3510
        node_variants, node_params, node_api = osl[0][3:6]
3511
        if idx == 0: # first entry
3512
          variants = set(node_variants)
3513
          params = set(node_params)
3514
          api_versions = set(node_api)
3515
        else: # keep consistency
3516
          variants.intersection_update(node_variants)
3517
          params.intersection_update(node_params)
3518
          api_versions.intersection_update(node_api)
3519

    
3520
      is_hid = os_name in cluster.hidden_os
3521
      is_blk = os_name in cluster.blacklisted_os
3522
      if ((self._HID not in self.op.output_fields and is_hid) or
3523
          (self._BLK not in self.op.output_fields and is_blk) or
3524
          (self._VLD not in self.op.output_fields and not valid)):
3525
        continue
3526

    
3527
      for field in self.op.output_fields:
3528
        if field == "name":
3529
          val = os_name
3530
        elif field == self._VLD:
3531
          val = valid
3532
        elif field == "node_status":
3533
          # this is just a copy of the dict
3534
          val = {}
3535
          for node_name, nos_list in os_data.items():
3536
            val[node_name] = nos_list
3537
        elif field == "variants":
3538
          val = utils.NiceSort(list(variants))
3539
        elif field == "parameters":
3540
          val = list(params)
3541
        elif field == "api_versions":
3542
          val = list(api_versions)
3543
        elif field == self._HID:
3544
          val = is_hid
3545
        elif field == self._BLK:
3546
          val = is_blk
3547
        else:
3548
          raise errors.ParameterError(field)
3549
        row.append(val)
3550
      output.append(row)
3551

    
3552
    return output
3553

    
3554

    
3555
class LUNodeRemove(LogicalUnit):
3556
  """Logical unit for removing a node.
3557

3558
  """
3559
  HPATH = "node-remove"
3560
  HTYPE = constants.HTYPE_NODE
3561

    
3562
  def BuildHooksEnv(self):
3563
    """Build hooks env.
3564

3565
    This doesn't run on the target node in the pre phase as a failed
3566
    node would then be impossible to remove.
3567

3568
    """
3569
    env = {
3570
      "OP_TARGET": self.op.node_name,
3571
      "NODE_NAME": self.op.node_name,
3572
      }
3573
    all_nodes = self.cfg.GetNodeList()
3574
    try:
3575
      all_nodes.remove(self.op.node_name)
3576
    except ValueError:
3577
      logging.warning("Node %s which is about to be removed not found"
3578
                      " in the all nodes list", self.op.node_name)
3579
    return env, all_nodes, all_nodes
3580

    
3581
  def CheckPrereq(self):
3582
    """Check prerequisites.
3583

3584
    This checks:
3585
     - the node exists in the configuration
3586
     - it does not have primary or secondary instances
3587
     - it's not the master
3588

3589
    Any errors are signaled by raising errors.OpPrereqError.
3590

3591
    """
3592
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3593
    node = self.cfg.GetNodeInfo(self.op.node_name)
3594
    assert node is not None
3595

    
3596
    instance_list = self.cfg.GetInstanceList()
3597

    
3598
    masternode = self.cfg.GetMasterNode()
3599
    if node.name == masternode:
3600
      raise errors.OpPrereqError("Node is the master node,"
3601
                                 " you need to failover first.",
3602
                                 errors.ECODE_INVAL)
3603

    
3604
    for instance_name in instance_list:
3605
      instance = self.cfg.GetInstanceInfo(instance_name)
3606
      if node.name in instance.all_nodes:
3607
        raise errors.OpPrereqError("Instance %s is still running on the node,"
3608
                                   " please remove first." % instance_name,
3609
                                   errors.ECODE_INVAL)
3610
    self.op.node_name = node.name
3611
    self.node = node
3612

    
3613
  def Exec(self, feedback_fn):
3614
    """Removes the node from the cluster.
3615

3616
    """
3617
    node = self.node
3618
    logging.info("Stopping the node daemon and removing configs from node %s",
3619
                 node.name)
3620

    
3621
    modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
3622

    
3623
    # Promote nodes to master candidate as needed
3624
    _AdjustCandidatePool(self, exceptions=[node.name])
3625
    self.context.RemoveNode(node.name)
3626

    
3627
    # Run post hooks on the node before it's removed
3628
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
3629
    try:
3630
      hm.RunPhase(constants.HOOKS_PHASE_POST, [node.name])
3631
    except:
3632
      # pylint: disable-msg=W0702
3633
      self.LogWarning("Errors occurred running hooks on %s" % node.name)
3634

    
3635
    result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
3636
    msg = result.fail_msg
3637
    if msg:
3638
      self.LogWarning("Errors encountered on the remote node while leaving"
3639
                      " the cluster: %s", msg)
3640

    
3641
    # Remove node from our /etc/hosts
3642
    if self.cfg.GetClusterInfo().modify_etc_hosts:
3643
      master_node = self.cfg.GetMasterNode()
3644
      result = self.rpc.call_etc_hosts_modify(master_node,
3645
                                              constants.ETC_HOSTS_REMOVE,
3646
                                              node.name, None)
3647
      result.Raise("Can't update hosts file with new host data")
3648
      _RedistributeAncillaryFiles(self)
3649

    
3650

    
3651
class _NodeQuery(_QueryBase):
3652
  FIELDS = query.NODE_FIELDS
3653

    
3654
  def ExpandNames(self, lu):
3655
    lu.needed_locks = {}
3656
    lu.share_locks[locking.LEVEL_NODE] = 1
3657

    
3658
    if self.names:
3659
      self.wanted = _GetWantedNodes(lu, self.names)
3660
    else:
3661
      self.wanted = locking.ALL_SET
3662

    
3663
    self.do_locking = (self.use_locking and
3664
                       query.NQ_LIVE in self.requested_data)
3665

    
3666
    if self.do_locking:
3667
      # if we don't request only static fields, we need to lock the nodes
3668
      lu.needed_locks[locking.LEVEL_NODE] = self.wanted
3669

    
3670
  def DeclareLocks(self, lu, level):
3671
    pass
3672

    
3673
  def _GetQueryData(self, lu):
3674
    """Computes the list of nodes and their attributes.
3675

3676
    """
3677
    all_info = lu.cfg.GetAllNodesInfo()
3678

    
3679
    nodenames = self._GetNames(lu, all_info.keys(), locking.LEVEL_NODE)
3680

    
3681
    # Gather data as requested
3682
    if query.NQ_LIVE in self.requested_data:
3683
      # filter out non-vm_capable nodes
3684
      toquery_nodes = [name for name in nodenames if all_info[name].vm_capable]
3685

    
3686
      node_data = lu.rpc.call_node_info(toquery_nodes, lu.cfg.GetVGName(),
3687
                                        lu.cfg.GetHypervisorType())
3688
      live_data = dict((name, nresult.payload)
3689
                       for (name, nresult) in node_data.items()
3690
                       if not nresult.fail_msg and nresult.payload)
3691
    else:
3692
      live_data = None
3693

    
3694
    if query.NQ_INST in self.requested_data:
3695
      node_to_primary = dict([(name, set()) for name in nodenames])
3696
      node_to_secondary = dict([(name, set()) for name in nodenames])
3697

    
3698
      inst_data = lu.cfg.GetAllInstancesInfo()
3699

    
3700
      for inst in inst_data.values():
3701
        if inst.primary_node in node_to_primary:
3702
          node_to_primary[inst.primary_node].add(inst.name)
3703
        for secnode in inst.secondary_nodes:
3704
          if secnode in node_to_secondary:
3705
            node_to_secondary[secnode].add(inst.name)
3706
    else:
3707
      node_to_primary = None
3708
      node_to_secondary = None
3709

    
3710
    if query.NQ_OOB in self.requested_data:
3711
      oob_support = dict((name, bool(_SupportsOob(lu.cfg, node)))
3712
                         for name, node in all_info.iteritems())
3713
    else:
3714
      oob_support = None
3715

    
3716
    if query.NQ_GROUP in self.requested_data:
3717
      groups = lu.cfg.GetAllNodeGroupsInfo()
3718
    else:
3719
      groups = {}
3720

    
3721
    return query.NodeQueryData([all_info[name] for name in nodenames],
3722
                               live_data, lu.cfg.GetMasterNode(),
3723
                               node_to_primary, node_to_secondary, groups,
3724
                               oob_support, lu.cfg.GetClusterInfo())
3725

    
3726

    
3727
class LUNodeQuery(NoHooksLU):
3728
  """Logical unit for querying nodes.
3729

3730
  """
3731
  # pylint: disable-msg=W0142
3732
  REQ_BGL = False
3733

    
3734
  def CheckArguments(self):
3735
    self.nq = _NodeQuery(self.op.names, self.op.output_fields,
3736
                         self.op.use_locking)
3737

    
3738
  def ExpandNames(self):
3739
    self.nq.ExpandNames(self)
3740

    
3741
  def Exec(self, feedback_fn):
3742
    return self.nq.OldStyleQuery(self)
3743

    
3744

    
3745
class LUNodeQueryvols(NoHooksLU):
3746
  """Logical unit for getting volumes on node(s).
3747

3748
  """
3749
  REQ_BGL = False
3750
  _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
3751
  _FIELDS_STATIC = utils.FieldSet("node")
3752

    
3753
  def CheckArguments(self):
3754
    _CheckOutputFields(static=self._FIELDS_STATIC,
3755
                       dynamic=self._FIELDS_DYNAMIC,
3756
                       selected=self.op.output_fields)
3757

    
3758
  def ExpandNames(self):
3759
    self.needed_locks = {}
3760
    self.share_locks[locking.LEVEL_NODE] = 1
3761
    if not self.op.nodes:
3762
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3763
    else:
3764
      self.needed_locks[locking.LEVEL_NODE] = \
3765
        _GetWantedNodes(self, self.op.nodes)
3766

    
3767
  def Exec(self, feedback_fn):
3768
    """Computes the list of nodes and their attributes.
3769

3770
    """
3771
    nodenames = self.acquired_locks[locking.LEVEL_NODE]
3772
    volumes = self.rpc.call_node_volumes(nodenames)
3773

    
3774
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
3775
             in self.cfg.GetInstanceList()]
3776

    
3777
    lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
3778

    
3779
    output = []
3780
    for node in nodenames:
3781
      nresult = volumes[node]
3782
      if nresult.offline:
3783
        continue
3784
      msg = nresult.fail_msg
3785
      if msg:
3786
        self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
3787
        continue
3788

    
3789
      node_vols = nresult.payload[:]
3790
      node_vols.sort(key=lambda vol: vol['dev'])
3791

    
3792
      for vol in node_vols:
3793
        node_output = []
3794
        for field in self.op.output_fields:
3795
          if field == "node":
3796
            val = node
3797
          elif field == "phys":
3798
            val = vol['dev']
3799
          elif field == "vg":
3800
            val = vol['vg']
3801
          elif field == "name":
3802
            val = vol['name']
3803
          elif field == "size":
3804
            val = int(float(vol['size']))
3805
          elif field == "instance":
3806
            for inst in ilist:
3807
              if node not in lv_by_node[inst]:
3808
                continue
3809
              if vol['name'] in lv_by_node[inst][node]:
3810
                val = inst.name
3811
                break
3812
            else:
3813
              val = '-'
3814
          else:
3815
            raise errors.ParameterError(field)
3816
          node_output.append(str(val))
3817

    
3818
        output.append(node_output)
3819

    
3820
    return output
3821

    
3822

    
3823
class LUNodeQueryStorage(NoHooksLU):
3824
  """Logical unit for getting information on storage units on node(s).
3825

3826
  """
3827
  _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
3828
  REQ_BGL = False
3829

    
3830
  def CheckArguments(self):
3831
    _CheckOutputFields(static=self._FIELDS_STATIC,
3832
                       dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
3833
                       selected=self.op.output_fields)
3834

    
3835
  def ExpandNames(self):
3836
    self.needed_locks = {}
3837
    self.share_locks[locking.LEVEL_NODE] = 1
3838

    
3839
    if self.op.nodes:
3840
      self.needed_locks[locking.LEVEL_NODE] = \
3841
        _GetWantedNodes(self, self.op.nodes)
3842
    else:
3843
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3844

    
3845
  def Exec(self, feedback_fn):
3846
    """Computes the list of nodes and their attributes.
3847

3848
    """
3849
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
3850

    
3851
    # Always get name to sort by
3852
    if constants.SF_NAME in self.op.output_fields:
3853
      fields = self.op.output_fields[:]
3854
    else:
3855
      fields = [constants.SF_NAME] + self.op.output_fields
3856

    
3857
    # Never ask for node or type as it's only known to the LU
3858
    for extra in [constants.SF_NODE, constants.SF_TYPE]:
3859
      while extra in fields:
3860
        fields.remove(extra)
3861

    
3862
    field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
3863
    name_idx = field_idx[constants.SF_NAME]
3864

    
3865
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3866
    data = self.rpc.call_storage_list(self.nodes,
3867
                                      self.op.storage_type, st_args,
3868
                                      self.op.name, fields)
3869

    
3870
    result = []
3871

    
3872
    for node in utils.NiceSort(self.nodes):
3873
      nresult = data[node]
3874
      if nresult.offline:
3875
        continue
3876

    
3877
      msg = nresult.fail_msg
3878
      if msg:
3879
        self.LogWarning("Can't get storage data from node %s: %s", node, msg)
3880
        continue
3881

    
3882
      rows = dict([(row[name_idx], row) for row in nresult.payload])
3883

    
3884
      for name in utils.NiceSort(rows.keys()):
3885
        row = rows[name]
3886

    
3887
        out = []
3888

    
3889
        for field in self.op.output_fields:
3890
          if field == constants.SF_NODE:
3891
            val = node
3892
          elif field == constants.SF_TYPE:
3893
            val = self.op.storage_type
3894
          elif field in field_idx:
3895
            val = row[field_idx[field]]
3896
          else:
3897
            raise errors.ParameterError(field)
3898

    
3899
          out.append(val)
3900

    
3901
        result.append(out)
3902

    
3903
    return result
3904

    
3905

    
3906
class _InstanceQuery(_QueryBase):
3907
  FIELDS = query.INSTANCE_FIELDS
3908

    
3909
  def ExpandNames(self, lu):
3910
    lu.needed_locks = {}
3911
    lu.share_locks[locking.LEVEL_INSTANCE] = 1
3912
    lu.share_locks[locking.LEVEL_NODE] = 1
3913

    
3914
    if self.names:
3915
      self.wanted = _GetWantedInstances(lu, self.names)
3916
    else:
3917
      self.wanted = locking.ALL_SET
3918

    
3919
    self.do_locking = (self.use_locking and
3920
                       query.IQ_LIVE in self.requested_data)
3921
    if self.do_locking:
3922
      lu.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
3923
      lu.needed_locks[locking.LEVEL_NODE] = []
3924
      lu.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3925

    
3926
  def DeclareLocks(self, lu, level):
3927
    if level == locking.LEVEL_NODE and self.do_locking:
3928
      lu._LockInstancesNodes() # pylint: disable-msg=W0212
3929

    
3930
  def _GetQueryData(self, lu):
3931
    """Computes the list of instances and their attributes.
3932

3933
    """
3934
    cluster = lu.cfg.GetClusterInfo()
3935
    all_info = lu.cfg.GetAllInstancesInfo()
3936

    
3937
    instance_names = self._GetNames(lu, all_info.keys(), locking.LEVEL_INSTANCE)
3938

    
3939
    instance_list = [all_info[name] for name in instance_names]
3940
    nodes = frozenset(itertools.chain(*(inst.all_nodes
3941
                                        for inst in instance_list)))
3942
    hv_list = list(set([inst.hypervisor for inst in instance_list]))
3943
    bad_nodes = []
3944
    offline_nodes = []
3945
    wrongnode_inst = set()
3946

    
3947
    # Gather data as requested
3948
    if self.requested_data & set([query.IQ_LIVE, query.IQ_CONSOLE]):
3949
      live_data = {}
3950
      node_data = lu.rpc.call_all_instances_info(nodes, hv_list)
3951
      for name in nodes:
3952
        result = node_data[name]
3953
        if result.offline:
3954
          # offline nodes will be in both lists
3955
          assert result.fail_msg
3956
          offline_nodes.append(name)
3957
        if result.fail_msg:
3958
          bad_nodes.append(name)
3959
        elif result.payload:
3960
          for inst in result.payload:
3961
            if inst in all_info:
3962
              if all_info[inst].primary_node == name:
3963
                live_data.update(result.payload)
3964
              else:
3965
                wrongnode_inst.add(inst)
3966
            else:
3967
              # orphan instance; we don't list it here as we don't
3968
              # handle this case yet in the output of instance listing
3969
              logging.warning("Orphan instance '%s' found on node %s",
3970
                              inst, name)
3971
        # else no instance is alive
3972
    else:
3973
      live_data = {}
3974

    
3975
    if query.IQ_DISKUSAGE in self.requested_data:
3976
      disk_usage = dict((inst.name,
3977
                         _ComputeDiskSize(inst.disk_template,
3978
                                          [{"size": disk.size}
3979
                                           for disk in inst.disks]))
3980
                        for inst in instance_list)
3981
    else:
3982
      disk_usage = None
3983

    
3984
    if query.IQ_CONSOLE in self.requested_data:
3985
      consinfo = {}
3986
      for inst in instance_list:
3987
        if inst.name in live_data:
3988
          # Instance is running
3989
          consinfo[inst.name] = _GetInstanceConsole(cluster, inst)
3990
        else:
3991
          consinfo[inst.name] = None
3992
      assert set(consinfo.keys()) == set(instance_names)
3993
    else:
3994
      consinfo = None
3995

    
3996
    return query.InstanceQueryData(instance_list, lu.cfg.GetClusterInfo(),
3997
                                   disk_usage, offline_nodes, bad_nodes,
3998
                                   live_data, wrongnode_inst, consinfo)
3999

    
4000

    
4001
class LUQuery(NoHooksLU):
4002
  """Query for resources/items of a certain kind.
4003

4004
  """
4005
  # pylint: disable-msg=W0142
4006
  REQ_BGL = False
4007

    
4008
  def CheckArguments(self):
4009
    qcls = _GetQueryImplementation(self.op.what)
4010
    names = qlang.ReadSimpleFilter("name", self.op.filter)
4011

    
4012
    self.impl = qcls(names, self.op.fields, False)
4013

    
4014
  def ExpandNames(self):
4015
    self.impl.ExpandNames(self)
4016

    
4017
  def DeclareLocks(self, level):
4018
    self.impl.DeclareLocks(self, level)
4019

    
4020
  def Exec(self, feedback_fn):
4021
    return self.impl.NewStyleQuery(self)
4022

    
4023

    
4024
class LUQueryFields(NoHooksLU):
4025
  """Query for resources/items of a certain kind.
4026

4027
  """
4028
  # pylint: disable-msg=W0142
4029
  REQ_BGL = False
4030

    
4031
  def CheckArguments(self):
4032
    self.qcls = _GetQueryImplementation(self.op.what)
4033

    
4034
  def ExpandNames(self):
4035
    self.needed_locks = {}
4036

    
4037
  def Exec(self, feedback_fn):
4038
    return self.qcls.FieldsQuery(self.op.fields)
4039

    
4040

    
4041
class LUNodeModifyStorage(NoHooksLU):
4042
  """Logical unit for modifying a storage volume on a node.
4043

4044
  """
4045
  REQ_BGL = False
4046

    
4047
  def CheckArguments(self):
4048
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4049

    
4050
    storage_type = self.op.storage_type
4051

    
4052
    try:
4053
      modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
4054
    except KeyError:
4055
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
4056
                                 " modified" % storage_type,
4057
                                 errors.ECODE_INVAL)
4058

    
4059
    diff = set(self.op.changes.keys()) - modifiable
4060
    if diff:
4061
      raise errors.OpPrereqError("The following fields can not be modified for"
4062
                                 " storage units of type '%s': %r" %
4063
                                 (storage_type, list(diff)),
4064
                                 errors.ECODE_INVAL)
4065

    
4066
  def ExpandNames(self):
4067
    self.needed_locks = {
4068
      locking.LEVEL_NODE: self.op.node_name,
4069
      }
4070

    
4071
  def Exec(self, feedback_fn):
4072
    """Computes the list of nodes and their attributes.
4073

4074
    """
4075
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
4076
    result = self.rpc.call_storage_modify(self.op.node_name,
4077
                                          self.op.storage_type, st_args,
4078
                                          self.op.name, self.op.changes)
4079
    result.Raise("Failed to modify storage unit '%s' on %s" %
4080
                 (self.op.name, self.op.node_name))
4081

    
4082

    
4083
class LUNodeAdd(LogicalUnit):
4084
  """Logical unit for adding node to the cluster.
4085

4086
  """
4087
  HPATH = "node-add"
4088
  HTYPE = constants.HTYPE_NODE
4089
  _NFLAGS = ["master_capable", "vm_capable"]
4090

    
4091
  def CheckArguments(self):
4092
    self.primary_ip_family = self.cfg.GetPrimaryIPFamily()
4093
    # validate/normalize the node name
4094
    self.hostname = netutils.GetHostname(name=self.op.node_name,
4095
                                         family=self.primary_ip_family)
4096
    self.op.node_name = self.hostname.name
4097

    
4098
    if self.op.readd and self.op.node_name == self.cfg.GetMasterNode():
4099
      raise errors.OpPrereqError("Cannot readd the master node",
4100
                                 errors.ECODE_STATE)
4101

    
4102
    if self.op.readd and self.op.group: