Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ 5d28cb6f

History | View | Annotate | Download (395.7 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 time.
1448

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
1636
  def _VerifyNodeFiles(self, ninfo, nresult, file_list, local_cksum,
1637
                       master_files):
1638
    """Verifies and computes the node required file checksums.
1639

1640
    @type ninfo: L{objects.Node}
1641
    @param ninfo: the node to check
1642
    @param nresult: the remote results for the node
1643
    @param file_list: required list of files
1644
    @param local_cksum: dictionary of local files and their checksums
1645
    @param master_files: list of files that only masters should have
1646

1647
    """
1648
    node = ninfo.name
1649
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1650

    
1651
    remote_cksum = nresult.get(constants.NV_FILELIST, None)
1652
    test = not isinstance(remote_cksum, dict)
1653
    _ErrorIf(test, self.ENODEFILECHECK, node,
1654
             "node hasn't returned file checksum data")
1655
    if test:
1656
      return
1657

    
1658
    for file_name in file_list:
1659
      node_is_mc = ninfo.master_candidate
1660
      must_have = (file_name not in master_files) or node_is_mc
1661
      # missing
1662
      test1 = file_name not in remote_cksum
1663
      # invalid checksum
1664
      test2 = not test1 and remote_cksum[file_name] != local_cksum[file_name]
1665
      # existing and good
1666
      test3 = not test1 and remote_cksum[file_name] == local_cksum[file_name]
1667
      _ErrorIf(test1 and must_have, self.ENODEFILECHECK, node,
1668
               "file '%s' missing", file_name)
1669
      _ErrorIf(test2 and must_have, self.ENODEFILECHECK, node,
1670
               "file '%s' has wrong checksum", file_name)
1671
      # not candidate and this is not a must-have file
1672
      _ErrorIf(test2 and not must_have, self.ENODEFILECHECK, node,
1673
               "file '%s' should not exist on non master"
1674
               " candidates (and the file is outdated)", file_name)
1675
      # all good, except non-master/non-must have combination
1676
      _ErrorIf(test3 and not must_have, self.ENODEFILECHECK, node,
1677
               "file '%s' should not exist"
1678
               " on non master candidates", file_name)
1679

    
1680
  def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper,
1681
                      drbd_map):
1682
    """Verifies and the node DRBD status.
1683

1684
    @type ninfo: L{objects.Node}
1685
    @param ninfo: the node to check
1686
    @param nresult: the remote results for the node
1687
    @param instanceinfo: the dict of instances
1688
    @param drbd_helper: the configured DRBD usermode helper
1689
    @param drbd_map: the DRBD map as returned by
1690
        L{ganeti.config.ConfigWriter.ComputeDRBDMap}
1691

1692
    """
1693
    node = ninfo.name
1694
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1695

    
1696
    if drbd_helper:
1697
      helper_result = nresult.get(constants.NV_DRBDHELPER, None)
1698
      test = (helper_result == None)
1699
      _ErrorIf(test, self.ENODEDRBDHELPER, node,
1700
               "no drbd usermode helper returned")
1701
      if helper_result:
1702
        status, payload = helper_result
1703
        test = not status
1704
        _ErrorIf(test, self.ENODEDRBDHELPER, node,
1705
                 "drbd usermode helper check unsuccessful: %s", payload)
1706
        test = status and (payload != drbd_helper)
1707
        _ErrorIf(test, self.ENODEDRBDHELPER, node,
1708
                 "wrong drbd usermode helper: %s", payload)
1709

    
1710
    # compute the DRBD minors
1711
    node_drbd = {}
1712
    for minor, instance in drbd_map[node].items():
1713
      test = instance not in instanceinfo
1714
      _ErrorIf(test, self.ECLUSTERCFG, None,
1715
               "ghost instance '%s' in temporary DRBD map", instance)
1716
        # ghost instance should not be running, but otherwise we
1717
        # don't give double warnings (both ghost instance and
1718
        # unallocated minor in use)
1719
      if test:
1720
        node_drbd[minor] = (instance, False)
1721
      else:
1722
        instance = instanceinfo[instance]
1723
        node_drbd[minor] = (instance.name, instance.admin_up)
1724

    
1725
    # and now check them
1726
    used_minors = nresult.get(constants.NV_DRBDLIST, [])
1727
    test = not isinstance(used_minors, (tuple, list))
1728
    _ErrorIf(test, self.ENODEDRBD, node,
1729
             "cannot parse drbd status file: %s", str(used_minors))
1730
    if test:
1731
      # we cannot check drbd status
1732
      return
1733

    
1734
    for minor, (iname, must_exist) in node_drbd.items():
1735
      test = minor not in used_minors and must_exist
1736
      _ErrorIf(test, self.ENODEDRBD, node,
1737
               "drbd minor %d of instance %s is not active", minor, iname)
1738
    for minor in used_minors:
1739
      test = minor not in node_drbd
1740
      _ErrorIf(test, self.ENODEDRBD, node,
1741
               "unallocated drbd minor %d is in use", minor)
1742

    
1743
  def _UpdateNodeOS(self, ninfo, nresult, nimg):
1744
    """Builds the node OS structures.
1745

1746
    @type ninfo: L{objects.Node}
1747
    @param ninfo: the node to check
1748
    @param nresult: the remote results for the node
1749
    @param nimg: the node image object
1750

1751
    """
1752
    node = ninfo.name
1753
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1754

    
1755
    remote_os = nresult.get(constants.NV_OSLIST, None)
1756
    test = (not isinstance(remote_os, list) or
1757
            not compat.all(isinstance(v, list) and len(v) == 7
1758
                           for v in remote_os))
1759

    
1760
    _ErrorIf(test, self.ENODEOS, node,
1761
             "node hasn't returned valid OS data")
1762

    
1763
    nimg.os_fail = test
1764

    
1765
    if test:
1766
      return
1767

    
1768
    os_dict = {}
1769

    
1770
    for (name, os_path, status, diagnose,
1771
         variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
1772

    
1773
      if name not in os_dict:
1774
        os_dict[name] = []
1775

    
1776
      # parameters is a list of lists instead of list of tuples due to
1777
      # JSON lacking a real tuple type, fix it:
1778
      parameters = [tuple(v) for v in parameters]
1779
      os_dict[name].append((os_path, status, diagnose,
1780
                            set(variants), set(parameters), set(api_ver)))
1781

    
1782
    nimg.oslist = os_dict
1783

    
1784
  def _VerifyNodeOS(self, ninfo, nimg, base):
1785
    """Verifies the node OS list.
1786

1787
    @type ninfo: L{objects.Node}
1788
    @param ninfo: the node to check
1789
    @param nimg: the node image object
1790
    @param base: the 'template' node we match against (e.g. from the master)
1791

1792
    """
1793
    node = ninfo.name
1794
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1795

    
1796
    assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
1797

    
1798
    for os_name, os_data in nimg.oslist.items():
1799
      assert os_data, "Empty OS status for OS %s?!" % os_name
1800
      f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
1801
      _ErrorIf(not f_status, self.ENODEOS, node,
1802
               "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
1803
      _ErrorIf(len(os_data) > 1, self.ENODEOS, node,
1804
               "OS '%s' has multiple entries (first one shadows the rest): %s",
1805
               os_name, utils.CommaJoin([v[0] for v in os_data]))
1806
      # this will catched in backend too
1807
      _ErrorIf(compat.any(v >= constants.OS_API_V15 for v in f_api)
1808
               and not f_var, self.ENODEOS, node,
1809
               "OS %s with API at least %d does not declare any variant",
1810
               os_name, constants.OS_API_V15)
1811
      # comparisons with the 'base' image
1812
      test = os_name not in base.oslist
1813
      _ErrorIf(test, self.ENODEOS, node,
1814
               "Extra OS %s not present on reference node (%s)",
1815
               os_name, base.name)
1816
      if test:
1817
        continue
1818
      assert base.oslist[os_name], "Base node has empty OS status?"
1819
      _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
1820
      if not b_status:
1821
        # base OS is invalid, skipping
1822
        continue
1823
      for kind, a, b in [("API version", f_api, b_api),
1824
                         ("variants list", f_var, b_var),
1825
                         ("parameters", f_param, b_param)]:
1826
        _ErrorIf(a != b, self.ENODEOS, node,
1827
                 "OS %s %s differs from reference node %s: %s vs. %s",
1828
                 kind, os_name, base.name,
1829
                 utils.CommaJoin(a), utils.CommaJoin(b))
1830

    
1831
    # check any missing OSes
1832
    missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
1833
    _ErrorIf(missing, self.ENODEOS, node,
1834
             "OSes present on reference node %s but missing on this node: %s",
1835
             base.name, utils.CommaJoin(missing))
1836

    
1837
  def _VerifyOob(self, ninfo, nresult):
1838
    """Verifies out of band functionality of a node.
1839

1840
    @type ninfo: L{objects.Node}
1841
    @param ninfo: the node to check
1842
    @param nresult: the remote results for the node
1843

1844
    """
1845
    node = ninfo.name
1846
    # We just have to verify the paths on master and/or master candidates
1847
    # as the oob helper is invoked on the master
1848
    if ((ninfo.master_candidate or ninfo.master_capable) and
1849
        constants.NV_OOB_PATHS in nresult):
1850
      for path_result in nresult[constants.NV_OOB_PATHS]:
1851
        self._ErrorIf(path_result, self.ENODEOOBPATH, node, path_result)
1852

    
1853
  def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
1854
    """Verifies and updates the node volume data.
1855

1856
    This function will update a L{NodeImage}'s internal structures
1857
    with data from the remote call.
1858

1859
    @type ninfo: L{objects.Node}
1860
    @param ninfo: the node to check
1861
    @param nresult: the remote results for the node
1862
    @param nimg: the node image object
1863
    @param vg_name: the configured VG name
1864

1865
    """
1866
    node = ninfo.name
1867
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1868

    
1869
    nimg.lvm_fail = True
1870
    lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
1871
    if vg_name is None:
1872
      pass
1873
    elif isinstance(lvdata, basestring):
1874
      _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
1875
               utils.SafeEncode(lvdata))
1876
    elif not isinstance(lvdata, dict):
1877
      _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
1878
    else:
1879
      nimg.volumes = lvdata
1880
      nimg.lvm_fail = False
1881

    
1882
  def _UpdateNodeInstances(self, ninfo, nresult, nimg):
1883
    """Verifies and updates the node instance list.
1884

1885
    If the listing was successful, then updates this node's instance
1886
    list. Otherwise, it marks the RPC call as failed for the instance
1887
    list key.
1888

1889
    @type ninfo: L{objects.Node}
1890
    @param ninfo: the node to check
1891
    @param nresult: the remote results for the node
1892
    @param nimg: the node image object
1893

1894
    """
1895
    idata = nresult.get(constants.NV_INSTANCELIST, None)
1896
    test = not isinstance(idata, list)
1897
    self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed"
1898
                  " (instancelist): %s", utils.SafeEncode(str(idata)))
1899
    if test:
1900
      nimg.hyp_fail = True
1901
    else:
1902
      nimg.instances = idata
1903

    
1904
  def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
1905
    """Verifies and computes a node information map
1906

1907
    @type ninfo: L{objects.Node}
1908
    @param ninfo: the node to check
1909
    @param nresult: the remote results for the node
1910
    @param nimg: the node image object
1911
    @param vg_name: the configured VG name
1912

1913
    """
1914
    node = ninfo.name
1915
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1916

    
1917
    # try to read free memory (from the hypervisor)
1918
    hv_info = nresult.get(constants.NV_HVINFO, None)
1919
    test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
1920
    _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
1921
    if not test:
1922
      try:
1923
        nimg.mfree = int(hv_info["memory_free"])
1924
      except (ValueError, TypeError):
1925
        _ErrorIf(True, self.ENODERPC, node,
1926
                 "node returned invalid nodeinfo, check hypervisor")
1927

    
1928
    # FIXME: devise a free space model for file based instances as well
1929
    if vg_name is not None:
1930
      test = (constants.NV_VGLIST not in nresult or
1931
              vg_name not in nresult[constants.NV_VGLIST])
1932
      _ErrorIf(test, self.ENODELVM, node,
1933
               "node didn't return data for the volume group '%s'"
1934
               " - it is either missing or broken", vg_name)
1935
      if not test:
1936
        try:
1937
          nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
1938
        except (ValueError, TypeError):
1939
          _ErrorIf(True, self.ENODERPC, node,
1940
                   "node returned invalid LVM info, check LVM status")
1941

    
1942
  def _CollectDiskInfo(self, nodelist, node_image, instanceinfo):
1943
    """Gets per-disk status information for all instances.
1944

1945
    @type nodelist: list of strings
1946
    @param nodelist: Node names
1947
    @type node_image: dict of (name, L{objects.Node})
1948
    @param node_image: Node objects
1949
    @type instanceinfo: dict of (name, L{objects.Instance})
1950
    @param instanceinfo: Instance objects
1951
    @rtype: {instance: {node: [(succes, payload)]}}
1952
    @return: a dictionary of per-instance dictionaries with nodes as
1953
        keys and disk information as values; the disk information is a
1954
        list of tuples (success, payload)
1955

1956
    """
1957
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1958

    
1959
    node_disks = {}
1960
    node_disks_devonly = {}
1961
    diskless_instances = set()
1962
    diskless = constants.DT_DISKLESS
1963

    
1964
    for nname in nodelist:
1965
      node_instances = list(itertools.chain(node_image[nname].pinst,
1966
                                            node_image[nname].sinst))
1967
      diskless_instances.update(inst for inst in node_instances
1968
                                if instanceinfo[inst].disk_template == diskless)
1969
      disks = [(inst, disk)
1970
               for inst in node_instances
1971
               for disk in instanceinfo[inst].disks]
1972

    
1973
      if not disks:
1974
        # No need to collect data
1975
        continue
1976

    
1977
      node_disks[nname] = disks
1978

    
1979
      # Creating copies as SetDiskID below will modify the objects and that can
1980
      # lead to incorrect data returned from nodes
1981
      devonly = [dev.Copy() for (_, dev) in disks]
1982

    
1983
      for dev in devonly:
1984
        self.cfg.SetDiskID(dev, nname)
1985

    
1986
      node_disks_devonly[nname] = devonly
1987

    
1988
    assert len(node_disks) == len(node_disks_devonly)
1989

    
1990
    # Collect data from all nodes with disks
1991
    result = self.rpc.call_blockdev_getmirrorstatus_multi(node_disks.keys(),
1992
                                                          node_disks_devonly)
1993

    
1994
    assert len(result) == len(node_disks)
1995

    
1996
    instdisk = {}
1997

    
1998
    for (nname, nres) in result.items():
1999
      disks = node_disks[nname]
2000

    
2001
      if nres.offline:
2002
        # No data from this node
2003
        data = len(disks) * [(False, "node offline")]
2004
      else:
2005
        msg = nres.fail_msg
2006
        _ErrorIf(msg, self.ENODERPC, nname,
2007
                 "while getting disk information: %s", msg)
2008
        if msg:
2009
          # No data from this node
2010
          data = len(disks) * [(False, msg)]
2011
        else:
2012
          data = []
2013
          for idx, i in enumerate(nres.payload):
2014
            if isinstance(i, (tuple, list)) and len(i) == 2:
2015
              data.append(i)
2016
            else:
2017
              logging.warning("Invalid result from node %s, entry %d: %s",
2018
                              nname, idx, i)
2019
              data.append((False, "Invalid result from the remote node"))
2020

    
2021
      for ((inst, _), status) in zip(disks, data):
2022
        instdisk.setdefault(inst, {}).setdefault(nname, []).append(status)
2023

    
2024
    # Add empty entries for diskless instances.
2025
    for inst in diskless_instances:
2026
      assert inst not in instdisk
2027
      instdisk[inst] = {}
2028

    
2029
    assert compat.all(len(statuses) == len(instanceinfo[inst].disks) and
2030
                      len(nnames) <= len(instanceinfo[inst].all_nodes) and
2031
                      compat.all(isinstance(s, (tuple, list)) and
2032
                                 len(s) == 2 for s in statuses)
2033
                      for inst, nnames in instdisk.items()
2034
                      for nname, statuses in nnames.items())
2035
    assert set(instdisk) == set(instanceinfo), "instdisk consistency failure"
2036

    
2037
    return instdisk
2038

    
2039
  def _VerifyHVP(self, hvp_data):
2040
    """Verifies locally the syntax of the hypervisor parameters.
2041

2042
    """
2043
    for item, hv_name, hv_params in hvp_data:
2044
      msg = ("hypervisor %s parameters syntax check (source %s): %%s" %
2045
             (item, hv_name))
2046
      try:
2047
        hv_class = hypervisor.GetHypervisor(hv_name)
2048
        utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2049
        hv_class.CheckParameterSyntax(hv_params)
2050
      except errors.GenericError, err:
2051
        self._ErrorIf(True, self.ECLUSTERCFG, None, msg % str(err))
2052

    
2053

    
2054
  def BuildHooksEnv(self):
2055
    """Build hooks env.
2056

2057
    Cluster-Verify hooks just ran in the post phase and their failure makes
2058
    the output be logged in the verify output and the verification to fail.
2059

2060
    """
2061
    all_nodes = self.cfg.GetNodeList()
2062
    env = {
2063
      "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
2064
      }
2065
    for node in self.cfg.GetAllNodesInfo().values():
2066
      env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
2067

    
2068
    return env, [], all_nodes
2069

    
2070
  def Exec(self, feedback_fn):
2071
    """Verify integrity of cluster, performing various test on nodes.
2072

2073
    """
2074
    # This method has too many local variables. pylint: disable-msg=R0914
2075
    self.bad = False
2076
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
2077
    verbose = self.op.verbose
2078
    self._feedback_fn = feedback_fn
2079
    feedback_fn("* Verifying global settings")
2080
    for msg in self.cfg.VerifyConfig():
2081
      _ErrorIf(True, self.ECLUSTERCFG, None, msg)
2082

    
2083
    # Check the cluster certificates
2084
    for cert_filename in constants.ALL_CERT_FILES:
2085
      (errcode, msg) = _VerifyCertificate(cert_filename)
2086
      _ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
2087

    
2088
    vg_name = self.cfg.GetVGName()
2089
    drbd_helper = self.cfg.GetDRBDHelper()
2090
    hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
2091
    cluster = self.cfg.GetClusterInfo()
2092
    nodelist = utils.NiceSort(self.cfg.GetNodeList())
2093
    nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
2094
    nodeinfo_byname = dict(zip(nodelist, nodeinfo))
2095
    instancelist = utils.NiceSort(self.cfg.GetInstanceList())
2096
    instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
2097
                        for iname in instancelist)
2098
    groupinfo = self.cfg.GetAllNodeGroupsInfo()
2099
    i_non_redundant = [] # Non redundant instances
2100
    i_non_a_balanced = [] # Non auto-balanced instances
2101
    n_offline = 0 # Count of offline nodes
2102
    n_drained = 0 # Count of nodes being drained
2103
    node_vol_should = {}
2104

    
2105
    # FIXME: verify OS list
2106
    # do local checksums
2107
    master_files = [constants.CLUSTER_CONF_FILE]
2108
    master_node = self.master_node = self.cfg.GetMasterNode()
2109
    master_ip = self.cfg.GetMasterIP()
2110

    
2111
    file_names = ssconf.SimpleStore().GetFileList()
2112
    file_names.extend(constants.ALL_CERT_FILES)
2113
    file_names.extend(master_files)
2114
    if cluster.modify_etc_hosts:
2115
      file_names.append(constants.ETC_HOSTS)
2116

    
2117
    local_checksums = utils.FingerprintFiles(file_names)
2118

    
2119
    # Compute the set of hypervisor parameters
2120
    hvp_data = []
2121
    for hv_name in hypervisors:
2122
      hvp_data.append(("cluster", hv_name, cluster.GetHVDefaults(hv_name)))
2123
    for os_name, os_hvp in cluster.os_hvp.items():
2124
      for hv_name, hv_params in os_hvp.items():
2125
        if not hv_params:
2126
          continue
2127
        full_params = cluster.GetHVDefaults(hv_name, os_name=os_name)
2128
        hvp_data.append(("os %s" % os_name, hv_name, full_params))
2129
    # TODO: collapse identical parameter values in a single one
2130
    for instance in instanceinfo.values():
2131
      if not instance.hvparams:
2132
        continue
2133
      hvp_data.append(("instance %s" % instance.name, instance.hypervisor,
2134
                       cluster.FillHV(instance)))
2135
    # and verify them locally
2136
    self._VerifyHVP(hvp_data)
2137

    
2138
    feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
2139
    node_verify_param = {
2140
      constants.NV_FILELIST: file_names,
2141
      constants.NV_NODELIST: [node.name for node in nodeinfo
2142
                              if not node.offline],
2143
      constants.NV_HYPERVISOR: hypervisors,
2144
      constants.NV_HVPARAMS: hvp_data,
2145
      constants.NV_NODENETTEST: [(node.name, node.primary_ip,
2146
                                  node.secondary_ip) for node in nodeinfo
2147
                                 if not node.offline],
2148
      constants.NV_INSTANCELIST: hypervisors,
2149
      constants.NV_VERSION: None,
2150
      constants.NV_HVINFO: self.cfg.GetHypervisorType(),
2151
      constants.NV_NODESETUP: None,
2152
      constants.NV_TIME: None,
2153
      constants.NV_MASTERIP: (master_node, master_ip),
2154
      constants.NV_OSLIST: None,
2155
      constants.NV_VMNODES: self.cfg.GetNonVmCapableNodeList(),
2156
      }
2157

    
2158
    if vg_name is not None:
2159
      node_verify_param[constants.NV_VGLIST] = None
2160
      node_verify_param[constants.NV_LVLIST] = vg_name
2161
      node_verify_param[constants.NV_PVLIST] = [vg_name]
2162
      node_verify_param[constants.NV_DRBDLIST] = None
2163

    
2164
    if drbd_helper:
2165
      node_verify_param[constants.NV_DRBDHELPER] = drbd_helper
2166

    
2167
    # Build our expected cluster state
2168
    node_image = dict((node.name, self.NodeImage(offline=node.offline,
2169
                                                 name=node.name,
2170
                                                 vm_capable=node.vm_capable))
2171
                      for node in nodeinfo)
2172

    
2173
    # Gather OOB paths
2174
    oob_paths = []
2175
    for node in nodeinfo:
2176
      path = _SupportsOob(self.cfg, node)
2177
      if path and path not in oob_paths:
2178
        oob_paths.append(path)
2179

    
2180
    if oob_paths:
2181
      node_verify_param[constants.NV_OOB_PATHS] = oob_paths
2182

    
2183
    for instance in instancelist:
2184
      inst_config = instanceinfo[instance]
2185

    
2186
      for nname in inst_config.all_nodes:
2187
        if nname not in node_image:
2188
          # ghost node
2189
          gnode = self.NodeImage(name=nname)
2190
          gnode.ghost = True
2191
          node_image[nname] = gnode
2192

    
2193
      inst_config.MapLVsByNode(node_vol_should)
2194

    
2195
      pnode = inst_config.primary_node
2196
      node_image[pnode].pinst.append(instance)
2197

    
2198
      for snode in inst_config.secondary_nodes:
2199
        nimg = node_image[snode]
2200
        nimg.sinst.append(instance)
2201
        if pnode not in nimg.sbp:
2202
          nimg.sbp[pnode] = []
2203
        nimg.sbp[pnode].append(instance)
2204

    
2205
    # At this point, we have the in-memory data structures complete,
2206
    # except for the runtime information, which we'll gather next
2207

    
2208
    # Due to the way our RPC system works, exact response times cannot be
2209
    # guaranteed (e.g. a broken node could run into a timeout). By keeping the
2210
    # time before and after executing the request, we can at least have a time
2211
    # window.
2212
    nvinfo_starttime = time.time()
2213
    all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
2214
                                           self.cfg.GetClusterName())
2215
    nvinfo_endtime = time.time()
2216

    
2217
    all_drbd_map = self.cfg.ComputeDRBDMap()
2218

    
2219
    feedback_fn("* Gathering disk information (%s nodes)" % len(nodelist))
2220
    instdisk = self._CollectDiskInfo(nodelist, node_image, instanceinfo)
2221

    
2222
    feedback_fn("* Verifying node status")
2223

    
2224
    refos_img = None
2225

    
2226
    for node_i in nodeinfo:
2227
      node = node_i.name
2228
      nimg = node_image[node]
2229

    
2230
      if node_i.offline:
2231
        if verbose:
2232
          feedback_fn("* Skipping offline node %s" % (node,))
2233
        n_offline += 1
2234
        continue
2235

    
2236
      if node == master_node:
2237
        ntype = "master"
2238
      elif node_i.master_candidate:
2239
        ntype = "master candidate"
2240
      elif node_i.drained:
2241
        ntype = "drained"
2242
        n_drained += 1
2243
      else:
2244
        ntype = "regular"
2245
      if verbose:
2246
        feedback_fn("* Verifying node %s (%s)" % (node, ntype))
2247

    
2248
      msg = all_nvinfo[node].fail_msg
2249
      _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
2250
      if msg:
2251
        nimg.rpc_fail = True
2252
        continue
2253

    
2254
      nresult = all_nvinfo[node].payload
2255

    
2256
      nimg.call_ok = self._VerifyNode(node_i, nresult)
2257
      self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
2258
      self._VerifyNodeNetwork(node_i, nresult)
2259
      self._VerifyNodeFiles(node_i, nresult, file_names, local_checksums,
2260
                            master_files)
2261

    
2262
      self._VerifyOob(node_i, nresult)
2263

    
2264
      if nimg.vm_capable:
2265
        self._VerifyNodeLVM(node_i, nresult, vg_name)
2266
        self._VerifyNodeDrbd(node_i, nresult, instanceinfo, drbd_helper,
2267
                             all_drbd_map)
2268

    
2269
        self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
2270
        self._UpdateNodeInstances(node_i, nresult, nimg)
2271
        self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
2272
        self._UpdateNodeOS(node_i, nresult, nimg)
2273
        if not nimg.os_fail:
2274
          if refos_img is None:
2275
            refos_img = nimg
2276
          self._VerifyNodeOS(node_i, nimg, refos_img)
2277

    
2278
    feedback_fn("* Verifying instance status")
2279
    for instance in instancelist:
2280
      if verbose:
2281
        feedback_fn("* Verifying instance %s" % instance)
2282
      inst_config = instanceinfo[instance]
2283
      self._VerifyInstance(instance, inst_config, node_image,
2284
                           instdisk[instance])
2285
      inst_nodes_offline = []
2286

    
2287
      pnode = inst_config.primary_node
2288
      pnode_img = node_image[pnode]
2289
      _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
2290
               self.ENODERPC, pnode, "instance %s, connection to"
2291
               " primary node failed", instance)
2292

    
2293
      if pnode_img.offline:
2294
        inst_nodes_offline.append(pnode)
2295

    
2296
      # If the instance is non-redundant we cannot survive losing its primary
2297
      # node, so we are not N+1 compliant. On the other hand we have no disk
2298
      # templates with more than one secondary so that situation is not well
2299
      # supported either.
2300
      # FIXME: does not support file-backed instances
2301
      if not inst_config.secondary_nodes:
2302
        i_non_redundant.append(instance)
2303

    
2304
      _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT,
2305
               instance, "instance has multiple secondary nodes: %s",
2306
               utils.CommaJoin(inst_config.secondary_nodes),
2307
               code=self.ETYPE_WARNING)
2308

    
2309
      if inst_config.disk_template in constants.DTS_NET_MIRROR:
2310
        pnode = inst_config.primary_node
2311
        instance_nodes = utils.NiceSort(inst_config.all_nodes)
2312
        instance_groups = {}
2313

    
2314
        for node in instance_nodes:
2315
          instance_groups.setdefault(nodeinfo_byname[node].group,
2316
                                     []).append(node)
2317

    
2318
        pretty_list = [
2319
          "%s (group %s)" % (utils.CommaJoin(nodes), groupinfo[group].name)
2320
          # Sort so that we always list the primary node first.
2321
          for group, nodes in sorted(instance_groups.items(),
2322
                                     key=lambda (_, nodes): pnode in nodes,
2323
                                     reverse=True)]
2324

    
2325
        self._ErrorIf(len(instance_groups) > 1, self.EINSTANCESPLITGROUPS,
2326
                      instance, "instance has primary and secondary nodes in"
2327
                      " different groups: %s", utils.CommaJoin(pretty_list),
2328
                      code=self.ETYPE_WARNING)
2329

    
2330
      if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
2331
        i_non_a_balanced.append(instance)
2332

    
2333
      for snode in inst_config.secondary_nodes:
2334
        s_img = node_image[snode]
2335
        _ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode,
2336
                 "instance %s, connection to secondary node failed", instance)
2337

    
2338
        if s_img.offline:
2339
          inst_nodes_offline.append(snode)
2340

    
2341
      # warn that the instance lives on offline nodes
2342
      _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
2343
               "instance lives on offline node(s) %s",
2344
               utils.CommaJoin(inst_nodes_offline))
2345
      # ... or ghost/non-vm_capable nodes
2346
      for node in inst_config.all_nodes:
2347
        _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
2348
                 "instance lives on ghost node %s", node)
2349
        _ErrorIf(not node_image[node].vm_capable, self.EINSTANCEBADNODE,
2350
                 instance, "instance lives on non-vm_capable node %s", node)
2351

    
2352
    feedback_fn("* Verifying orphan volumes")
2353
    reserved = utils.FieldSet(*cluster.reserved_lvs)
2354
    self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
2355

    
2356
    feedback_fn("* Verifying orphan instances")
2357
    self._VerifyOrphanInstances(instancelist, node_image)
2358

    
2359
    if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
2360
      feedback_fn("* Verifying N+1 Memory redundancy")
2361
      self._VerifyNPlusOneMemory(node_image, instanceinfo)
2362

    
2363
    feedback_fn("* Other Notes")
2364
    if i_non_redundant:
2365
      feedback_fn("  - NOTICE: %d non-redundant instance(s) found."
2366
                  % len(i_non_redundant))
2367

    
2368
    if i_non_a_balanced:
2369
      feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
2370
                  % len(i_non_a_balanced))
2371

    
2372
    if n_offline:
2373
      feedback_fn("  - NOTICE: %d offline node(s) found." % n_offline)
2374

    
2375
    if n_drained:
2376
      feedback_fn("  - NOTICE: %d drained node(s) found." % n_drained)
2377

    
2378
    return not self.bad
2379

    
2380
  def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
2381
    """Analyze the post-hooks' result
2382

2383
    This method analyses the hook result, handles it, and sends some
2384
    nicely-formatted feedback back to the user.
2385

2386
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
2387
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
2388
    @param hooks_results: the results of the multi-node hooks rpc call
2389
    @param feedback_fn: function used send feedback back to the caller
2390
    @param lu_result: previous Exec result
2391
    @return: the new Exec result, based on the previous result
2392
        and hook results
2393

2394
    """
2395
    # We only really run POST phase hooks, and are only interested in
2396
    # their results
2397
    if phase == constants.HOOKS_PHASE_POST:
2398
      # Used to change hooks' output to proper indentation
2399
      feedback_fn("* Hooks Results")
2400
      assert hooks_results, "invalid result from hooks"
2401

    
2402
      for node_name in hooks_results:
2403
        res = hooks_results[node_name]
2404
        msg = res.fail_msg
2405
        test = msg and not res.offline
2406
        self._ErrorIf(test, self.ENODEHOOKS, node_name,
2407
                      "Communication failure in hooks execution: %s", msg)
2408
        if res.offline or msg:
2409
          # No need to investigate payload if node is offline or gave an error.
2410
          # override manually lu_result here as _ErrorIf only
2411
          # overrides self.bad
2412
          lu_result = 1
2413
          continue
2414
        for script, hkr, output in res.payload:
2415
          test = hkr == constants.HKR_FAIL
2416
          self._ErrorIf(test, self.ENODEHOOKS, node_name,
2417
                        "Script %s failed, output:", script)
2418
          if test:
2419
            output = self._HOOKS_INDENT_RE.sub('      ', output)
2420
            feedback_fn("%s" % output)
2421
            lu_result = 0
2422

    
2423
      return lu_result
2424

    
2425

    
2426
class LUClusterVerifyDisks(NoHooksLU):
2427
  """Verifies the cluster disks status.
2428

2429
  """
2430
  REQ_BGL = False
2431

    
2432
  def ExpandNames(self):
2433
    self.needed_locks = {
2434
      locking.LEVEL_NODE: locking.ALL_SET,
2435
      locking.LEVEL_INSTANCE: locking.ALL_SET,
2436
    }
2437
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
2438

    
2439
  def Exec(self, feedback_fn):
2440
    """Verify integrity of cluster disks.
2441

2442
    @rtype: tuple of three items
2443
    @return: a tuple of (dict of node-to-node_error, list of instances
2444
        which need activate-disks, dict of instance: (node, volume) for
2445
        missing volumes
2446

2447
    """
2448
    result = res_nodes, res_instances, res_missing = {}, [], {}
2449

    
2450
    nodes = utils.NiceSort(self.cfg.GetVmCapableNodeList())
2451
    instances = self.cfg.GetAllInstancesInfo().values()
2452

    
2453
    nv_dict = {}
2454
    for inst in instances:
2455
      inst_lvs = {}
2456
      if not inst.admin_up:
2457
        continue
2458
      inst.MapLVsByNode(inst_lvs)
2459
      # transform { iname: {node: [vol,],},} to {(node, vol): iname}
2460
      for node, vol_list in inst_lvs.iteritems():
2461
        for vol in vol_list:
2462
          nv_dict[(node, vol)] = inst
2463

    
2464
    if not nv_dict:
2465
      return result
2466

    
2467
    node_lvs = self.rpc.call_lv_list(nodes, [])
2468
    for node, node_res in node_lvs.items():
2469
      if node_res.offline:
2470
        continue
2471
      msg = node_res.fail_msg
2472
      if msg:
2473
        logging.warning("Error enumerating LVs on node %s: %s", node, msg)
2474
        res_nodes[node] = msg
2475
        continue
2476

    
2477
      lvs = node_res.payload
2478
      for lv_name, (_, _, lv_online) in lvs.items():
2479
        inst = nv_dict.pop((node, lv_name), None)
2480
        if (not lv_online and inst is not None
2481
            and inst.name not in res_instances):
2482
          res_instances.append(inst.name)
2483

    
2484
    # any leftover items in nv_dict are missing LVs, let's arrange the
2485
    # data better
2486
    for key, inst in nv_dict.iteritems():
2487
      if inst.name not in res_missing:
2488
        res_missing[inst.name] = []
2489
      res_missing[inst.name].append(key)
2490

    
2491
    return result
2492

    
2493

    
2494
class LUClusterRepairDiskSizes(NoHooksLU):
2495
  """Verifies the cluster disks sizes.
2496

2497
  """
2498
  REQ_BGL = False
2499

    
2500
  def ExpandNames(self):
2501
    if self.op.instances:
2502
      self.wanted_names = []
2503
      for name in self.op.instances:
2504
        full_name = _ExpandInstanceName(self.cfg, name)
2505
        self.wanted_names.append(full_name)
2506
      self.needed_locks = {
2507
        locking.LEVEL_NODE: [],
2508
        locking.LEVEL_INSTANCE: self.wanted_names,
2509
        }
2510
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2511
    else:
2512
      self.wanted_names = None
2513
      self.needed_locks = {
2514
        locking.LEVEL_NODE: locking.ALL_SET,
2515
        locking.LEVEL_INSTANCE: locking.ALL_SET,
2516
        }
2517
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
2518

    
2519
  def DeclareLocks(self, level):
2520
    if level == locking.LEVEL_NODE and self.wanted_names is not None:
2521
      self._LockInstancesNodes(primary_only=True)
2522

    
2523
  def CheckPrereq(self):
2524
    """Check prerequisites.
2525

2526
    This only checks the optional instance list against the existing names.
2527

2528
    """
2529
    if self.wanted_names is None:
2530
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
2531

    
2532
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
2533
                             in self.wanted_names]
2534

    
2535
  def _EnsureChildSizes(self, disk):
2536
    """Ensure children of the disk have the needed disk size.
2537

2538
    This is valid mainly for DRBD8 and fixes an issue where the
2539
    children have smaller disk size.
2540

2541
    @param disk: an L{ganeti.objects.Disk} object
2542

2543
    """
2544
    if disk.dev_type == constants.LD_DRBD8:
2545
      assert disk.children, "Empty children for DRBD8?"
2546
      fchild = disk.children[0]
2547
      mismatch = fchild.size < disk.size
2548
      if mismatch:
2549
        self.LogInfo("Child disk has size %d, parent %d, fixing",
2550
                     fchild.size, disk.size)
2551
        fchild.size = disk.size
2552

    
2553
      # and we recurse on this child only, not on the metadev
2554
      return self._EnsureChildSizes(fchild) or mismatch
2555
    else:
2556
      return False
2557

    
2558
  def Exec(self, feedback_fn):
2559
    """Verify the size of cluster disks.
2560

2561
    """
2562
    # TODO: check child disks too
2563
    # TODO: check differences in size between primary/secondary nodes
2564
    per_node_disks = {}
2565
    for instance in self.wanted_instances:
2566
      pnode = instance.primary_node
2567
      if pnode not in per_node_disks:
2568
        per_node_disks[pnode] = []
2569
      for idx, disk in enumerate(instance.disks):
2570
        per_node_disks[pnode].append((instance, idx, disk))
2571

    
2572
    changed = []
2573
    for node, dskl in per_node_disks.items():
2574
      newl = [v[2].Copy() for v in dskl]
2575
      for dsk in newl:
2576
        self.cfg.SetDiskID(dsk, node)
2577
      result = self.rpc.call_blockdev_getsizes(node, newl)
2578
      if result.fail_msg:
2579
        self.LogWarning("Failure in blockdev_getsizes call to node"
2580
                        " %s, ignoring", node)
2581
        continue
2582
      if len(result.data) != len(dskl):
2583
        self.LogWarning("Invalid result from node %s, ignoring node results",
2584
                        node)
2585
        continue
2586
      for ((instance, idx, disk), size) in zip(dskl, result.data):
2587
        if size is None:
2588
          self.LogWarning("Disk %d of instance %s did not return size"
2589
                          " information, ignoring", idx, instance.name)
2590
          continue
2591
        if not isinstance(size, (int, long)):
2592
          self.LogWarning("Disk %d of instance %s did not return valid"
2593
                          " size information, ignoring", idx, instance.name)
2594
          continue
2595
        size = size >> 20
2596
        if size != disk.size:
2597
          self.LogInfo("Disk %d of instance %s has mismatched size,"
2598
                       " correcting: recorded %d, actual %d", idx,
2599
                       instance.name, disk.size, size)
2600
          disk.size = size
2601
          self.cfg.Update(instance, feedback_fn)
2602
          changed.append((instance.name, idx, size))
2603
        if self._EnsureChildSizes(disk):
2604
          self.cfg.Update(instance, feedback_fn)
2605
          changed.append((instance.name, idx, disk.size))
2606
    return changed
2607

    
2608

    
2609
class LUClusterRename(LogicalUnit):
2610
  """Rename the cluster.
2611

2612
  """
2613
  HPATH = "cluster-rename"
2614
  HTYPE = constants.HTYPE_CLUSTER
2615

    
2616
  def BuildHooksEnv(self):
2617
    """Build hooks env.
2618

2619
    """
2620
    env = {
2621
      "OP_TARGET": self.cfg.GetClusterName(),
2622
      "NEW_NAME": self.op.name,
2623
      }
2624
    mn = self.cfg.GetMasterNode()
2625
    all_nodes = self.cfg.GetNodeList()
2626
    return env, [mn], all_nodes
2627

    
2628
  def CheckPrereq(self):
2629
    """Verify that the passed name is a valid one.
2630

2631
    """
2632
    hostname = netutils.GetHostname(name=self.op.name,
2633
                                    family=self.cfg.GetPrimaryIPFamily())
2634

    
2635
    new_name = hostname.name
2636
    self.ip = new_ip = hostname.ip
2637
    old_name = self.cfg.GetClusterName()
2638
    old_ip = self.cfg.GetMasterIP()
2639
    if new_name == old_name and new_ip == old_ip:
2640
      raise errors.OpPrereqError("Neither the name nor the IP address of the"
2641
                                 " cluster has changed",
2642
                                 errors.ECODE_INVAL)
2643
    if new_ip != old_ip:
2644
      if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
2645
        raise errors.OpPrereqError("The given cluster IP address (%s) is"
2646
                                   " reachable on the network" %
2647
                                   new_ip, errors.ECODE_NOTUNIQUE)
2648

    
2649
    self.op.name = new_name
2650

    
2651
  def Exec(self, feedback_fn):
2652
    """Rename the cluster.
2653

2654
    """
2655
    clustername = self.op.name
2656
    ip = self.ip
2657

    
2658
    # shutdown the master IP
2659
    master = self.cfg.GetMasterNode()
2660
    result = self.rpc.call_node_stop_master(master, False)
2661
    result.Raise("Could not disable the master role")
2662

    
2663
    try:
2664
      cluster = self.cfg.GetClusterInfo()
2665
      cluster.cluster_name = clustername
2666
      cluster.master_ip = ip
2667
      self.cfg.Update(cluster, feedback_fn)
2668

    
2669
      # update the known hosts file
2670
      ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
2671
      node_list = self.cfg.GetOnlineNodeList()
2672
      try:
2673
        node_list.remove(master)
2674
      except ValueError:
2675
        pass
2676
      _UploadHelper(self, node_list, constants.SSH_KNOWN_HOSTS_FILE)
2677
    finally:
2678
      result = self.rpc.call_node_start_master(master, False, False)
2679
      msg = result.fail_msg
2680
      if msg:
2681
        self.LogWarning("Could not re-enable the master role on"
2682
                        " the master, please restart manually: %s", msg)
2683

    
2684
    return clustername
2685

    
2686

    
2687
class LUClusterSetParams(LogicalUnit):
2688
  """Change the parameters of the cluster.
2689

2690
  """
2691
  HPATH = "cluster-modify"
2692
  HTYPE = constants.HTYPE_CLUSTER
2693
  REQ_BGL = False
2694

    
2695
  def CheckArguments(self):
2696
    """Check parameters
2697

2698
    """
2699
    if self.op.uid_pool:
2700
      uidpool.CheckUidPool(self.op.uid_pool)
2701

    
2702
    if self.op.add_uids:
2703
      uidpool.CheckUidPool(self.op.add_uids)
2704

    
2705
    if self.op.remove_uids:
2706
      uidpool.CheckUidPool(self.op.remove_uids)
2707

    
2708
  def ExpandNames(self):
2709
    # FIXME: in the future maybe other cluster params won't require checking on
2710
    # all nodes to be modified.
2711
    self.needed_locks = {
2712
      locking.LEVEL_NODE: locking.ALL_SET,
2713
    }
2714
    self.share_locks[locking.LEVEL_NODE] = 1
2715

    
2716
  def BuildHooksEnv(self):
2717
    """Build hooks env.
2718

2719
    """
2720
    env = {
2721
      "OP_TARGET": self.cfg.GetClusterName(),
2722
      "NEW_VG_NAME": self.op.vg_name,
2723
      }
2724
    mn = self.cfg.GetMasterNode()
2725
    return env, [mn], [mn]
2726

    
2727
  def CheckPrereq(self):
2728
    """Check prerequisites.
2729

2730
    This checks whether the given params don't conflict and
2731
    if the given volume group is valid.
2732

2733
    """
2734
    if self.op.vg_name is not None and not self.op.vg_name:
2735
      if self.cfg.HasAnyDiskOfType(constants.LD_LV):
2736
        raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based"
2737
                                   " instances exist", errors.ECODE_INVAL)
2738

    
2739
    if self.op.drbd_helper is not None and not self.op.drbd_helper:
2740
      if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8):
2741
        raise errors.OpPrereqError("Cannot disable drbd helper while"
2742
                                   " drbd-based instances exist",
2743
                                   errors.ECODE_INVAL)
2744

    
2745
    node_list = self.acquired_locks[locking.LEVEL_NODE]
2746

    
2747
    # if vg_name not None, checks given volume group on all nodes
2748
    if self.op.vg_name:
2749
      vglist = self.rpc.call_vg_list(node_list)
2750
      for node in node_list:
2751
        msg = vglist[node].fail_msg
2752
        if msg:
2753
          # ignoring down node
2754
          self.LogWarning("Error while gathering data on node %s"
2755
                          " (ignoring node): %s", node, msg)
2756
          continue
2757
        vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
2758
                                              self.op.vg_name,
2759
                                              constants.MIN_VG_SIZE)
2760
        if vgstatus:
2761
          raise errors.OpPrereqError("Error on node '%s': %s" %
2762
                                     (node, vgstatus), errors.ECODE_ENVIRON)
2763

    
2764
    if self.op.drbd_helper:
2765
      # checks given drbd helper on all nodes
2766
      helpers = self.rpc.call_drbd_helper(node_list)
2767
      for node in node_list:
2768
        ninfo = self.cfg.GetNodeInfo(node)
2769
        if ninfo.offline:
2770
          self.LogInfo("Not checking drbd helper on offline node %s", node)
2771
          continue
2772
        msg = helpers[node].fail_msg
2773
        if msg:
2774
          raise errors.OpPrereqError("Error checking drbd helper on node"
2775
                                     " '%s': %s" % (node, msg),
2776
                                     errors.ECODE_ENVIRON)
2777
        node_helper = helpers[node].payload
2778
        if node_helper != self.op.drbd_helper:
2779
          raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" %
2780
                                     (node, node_helper), errors.ECODE_ENVIRON)
2781

    
2782
    self.cluster = cluster = self.cfg.GetClusterInfo()
2783
    # validate params changes
2784
    if self.op.beparams:
2785
      utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
2786
      self.new_beparams = cluster.SimpleFillBE(self.op.beparams)
2787

    
2788
    if self.op.ndparams:
2789
      utils.ForceDictType(self.op.ndparams, constants.NDS_PARAMETER_TYPES)
2790
      self.new_ndparams = cluster.SimpleFillND(self.op.ndparams)
2791

    
2792
    if self.op.nicparams:
2793
      utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
2794
      self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams)
2795
      objects.NIC.CheckParameterSyntax(self.new_nicparams)
2796
      nic_errors = []
2797

    
2798
      # check all instances for consistency
2799
      for instance in self.cfg.GetAllInstancesInfo().values():
2800
        for nic_idx, nic in enumerate(instance.nics):
2801
          params_copy = copy.deepcopy(nic.nicparams)
2802
          params_filled = objects.FillDict(self.new_nicparams, params_copy)
2803

    
2804
          # check parameter syntax
2805
          try:
2806
            objects.NIC.CheckParameterSyntax(params_filled)
2807
          except errors.ConfigurationError, err:
2808
            nic_errors.append("Instance %s, nic/%d: %s" %
2809
                              (instance.name, nic_idx, err))
2810

    
2811
          # if we're moving instances to routed, check that they have an ip
2812
          target_mode = params_filled[constants.NIC_MODE]
2813
          if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
2814
            nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
2815
                              (instance.name, nic_idx))
2816
      if nic_errors:
2817
        raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
2818
                                   "\n".join(nic_errors))
2819

    
2820
    # hypervisor list/parameters
2821
    self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
2822
    if self.op.hvparams:
2823
      for hv_name, hv_dict in self.op.hvparams.items():
2824
        if hv_name not in self.new_hvparams:
2825
          self.new_hvparams[hv_name] = hv_dict
2826
        else:
2827
          self.new_hvparams[hv_name].update(hv_dict)
2828

    
2829
    # os hypervisor parameters
2830
    self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
2831
    if self.op.os_hvp:
2832
      for os_name, hvs in self.op.os_hvp.items():
2833
        if os_name not in self.new_os_hvp:
2834
          self.new_os_hvp[os_name] = hvs
2835
        else:
2836
          for hv_name, hv_dict in hvs.items():
2837
            if hv_name not in self.new_os_hvp[os_name]:
2838
              self.new_os_hvp[os_name][hv_name] = hv_dict
2839
            else:
2840
              self.new_os_hvp[os_name][hv_name].update(hv_dict)
2841

    
2842
    # os parameters
2843
    self.new_osp = objects.FillDict(cluster.osparams, {})
2844
    if self.op.osparams:
2845
      for os_name, osp in self.op.osparams.items():
2846
        if os_name not in self.new_osp:
2847
          self.new_osp[os_name] = {}
2848

    
2849
        self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp,
2850
                                                  use_none=True)
2851

    
2852
        if not self.new_osp[os_name]:
2853
          # we removed all parameters
2854
          del self.new_osp[os_name]
2855
        else:
2856
          # check the parameter validity (remote check)
2857
          _CheckOSParams(self, False, [self.cfg.GetMasterNode()],
2858
                         os_name, self.new_osp[os_name])
2859

    
2860
    # changes to the hypervisor list
2861
    if self.op.enabled_hypervisors is not None:
2862
      self.hv_list = self.op.enabled_hypervisors
2863
      for hv in self.hv_list:
2864
        # if the hypervisor doesn't already exist in the cluster
2865
        # hvparams, we initialize it to empty, and then (in both
2866
        # cases) we make sure to fill the defaults, as we might not
2867
        # have a complete defaults list if the hypervisor wasn't
2868
        # enabled before
2869
        if hv not in new_hvp:
2870
          new_hvp[hv] = {}
2871
        new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
2872
        utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
2873
    else:
2874
      self.hv_list = cluster.enabled_hypervisors
2875

    
2876
    if self.op.hvparams or self.op.enabled_hypervisors is not None:
2877
      # either the enabled list has changed, or the parameters have, validate
2878
      for hv_name, hv_params in self.new_hvparams.items():
2879
        if ((self.op.hvparams and hv_name in self.op.hvparams) or
2880
            (self.op.enabled_hypervisors and
2881
             hv_name in self.op.enabled_hypervisors)):
2882
          # either this is a new hypervisor, or its parameters have changed
2883
          hv_class = hypervisor.GetHypervisor(hv_name)
2884
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2885
          hv_class.CheckParameterSyntax(hv_params)
2886
          _CheckHVParams(self, node_list, hv_name, hv_params)
2887

    
2888
    if self.op.os_hvp:
2889
      # no need to check any newly-enabled hypervisors, since the
2890
      # defaults have already been checked in the above code-block
2891
      for os_name, os_hvp in self.new_os_hvp.items():
2892
        for hv_name, hv_params in os_hvp.items():
2893
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2894
          # we need to fill in the new os_hvp on top of the actual hv_p
2895
          cluster_defaults = self.new_hvparams.get(hv_name, {})
2896
          new_osp = objects.FillDict(cluster_defaults, hv_params)
2897
          hv_class = hypervisor.GetHypervisor(hv_name)
2898
          hv_class.CheckParameterSyntax(new_osp)
2899
          _CheckHVParams(self, node_list, hv_name, new_osp)
2900

    
2901
    if self.op.default_iallocator:
2902
      alloc_script = utils.FindFile(self.op.default_iallocator,
2903
                                    constants.IALLOCATOR_SEARCH_PATH,
2904
                                    os.path.isfile)
2905
      if alloc_script is None:
2906
        raise errors.OpPrereqError("Invalid default iallocator script '%s'"
2907
                                   " specified" % self.op.default_iallocator,
2908
                                   errors.ECODE_INVAL)
2909

    
2910
  def Exec(self, feedback_fn):
2911
    """Change the parameters of the cluster.
2912

2913
    """
2914
    if self.op.vg_name is not None:
2915
      new_volume = self.op.vg_name
2916
      if not new_volume:
2917
        new_volume = None
2918
      if new_volume != self.cfg.GetVGName():
2919
        self.cfg.SetVGName(new_volume)
2920
      else:
2921
        feedback_fn("Cluster LVM configuration already in desired"
2922
                    " state, not changing")
2923
    if self.op.drbd_helper is not None:
2924
      new_helper = self.op.drbd_helper
2925
      if not new_helper:
2926
        new_helper = None
2927
      if new_helper != self.cfg.GetDRBDHelper():
2928
        self.cfg.SetDRBDHelper(new_helper)
2929
      else:
2930
        feedback_fn("Cluster DRBD helper already in desired state,"
2931
                    " not changing")
2932
    if self.op.hvparams:
2933
      self.cluster.hvparams = self.new_hvparams
2934
    if self.op.os_hvp:
2935
      self.cluster.os_hvp = self.new_os_hvp
2936
    if self.op.enabled_hypervisors is not None:
2937
      self.cluster.hvparams = self.new_hvparams
2938
      self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
2939
    if self.op.beparams:
2940
      self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
2941
    if self.op.nicparams:
2942
      self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
2943
    if self.op.osparams:
2944
      self.cluster.osparams = self.new_osp
2945
    if self.op.ndparams:
2946
      self.cluster.ndparams = self.new_ndparams
2947

    
2948
    if self.op.candidate_pool_size is not None:
2949
      self.cluster.candidate_pool_size = self.op.candidate_pool_size
2950
      # we need to update the pool size here, otherwise the save will fail
2951
      _AdjustCandidatePool(self, [])
2952

    
2953
    if self.op.maintain_node_health is not None:
2954
      self.cluster.maintain_node_health = self.op.maintain_node_health
2955

    
2956
    if self.op.prealloc_wipe_disks is not None:
2957
      self.cluster.prealloc_wipe_disks = self.op.prealloc_wipe_disks
2958

    
2959
    if self.op.add_uids is not None:
2960
      uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
2961

    
2962
    if self.op.remove_uids is not None:
2963
      uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
2964

    
2965
    if self.op.uid_pool is not None:
2966
      self.cluster.uid_pool = self.op.uid_pool
2967

    
2968
    if self.op.default_iallocator is not None:
2969
      self.cluster.default_iallocator = self.op.default_iallocator
2970

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

    
2974
    def helper_os(aname, mods, desc):
2975
      desc += " OS list"
2976
      lst = getattr(self.cluster, aname)
2977
      for key, val in mods:
2978
        if key == constants.DDM_ADD:
2979
          if val in lst:
2980
            feedback_fn("OS %s already in %s, ignoring" % (val, desc))
2981
          else:
2982
            lst.append(val)
2983
        elif key == constants.DDM_REMOVE:
2984
          if val in lst:
2985
            lst.remove(val)
2986
          else:
2987
            feedback_fn("OS %s not found in %s, ignoring" % (val, desc))
2988
        else:
2989
          raise errors.ProgrammerError("Invalid modification '%s'" % key)
2990

    
2991
    if self.op.hidden_os:
2992
      helper_os("hidden_os", self.op.hidden_os, "hidden")
2993

    
2994
    if self.op.blacklisted_os:
2995
      helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted")
2996

    
2997
    if self.op.master_netdev:
2998
      master = self.cfg.GetMasterNode()
2999
      feedback_fn("Shutting down master ip on the current netdev (%s)" %
3000
                  self.cluster.master_netdev)
3001
      result = self.rpc.call_node_stop_master(master, False)
3002
      result.Raise("Could not disable the master ip")
3003
      feedback_fn("Changing master_netdev from %s to %s" %
3004
                  (self.cluster.master_netdev, self.op.master_netdev))
3005
      self.cluster.master_netdev = self.op.master_netdev
3006

    
3007
    self.cfg.Update(self.cluster, feedback_fn)
3008

    
3009
    if self.op.master_netdev:
3010
      feedback_fn("Starting the master ip on the new master netdev (%s)" %
3011
                  self.op.master_netdev)
3012
      result = self.rpc.call_node_start_master(master, False, False)
3013
      if result.fail_msg:
3014
        self.LogWarning("Could not re-enable the master ip on"
3015
                        " the master, please restart manually: %s",
3016
                        result.fail_msg)
3017

    
3018

    
3019
def _UploadHelper(lu, nodes, fname):
3020
  """Helper for uploading a file and showing warnings.
3021

3022
  """
3023
  if os.path.exists(fname):
3024
    result = lu.rpc.call_upload_file(nodes, fname)
3025
    for to_node, to_result in result.items():
3026
      msg = to_result.fail_msg
3027
      if msg:
3028
        msg = ("Copy of file %s to node %s failed: %s" %
3029
               (fname, to_node, msg))
3030
        lu.proc.LogWarning(msg)
3031

    
3032

    
3033
def _RedistributeAncillaryFiles(lu, additional_nodes=None, additional_vm=True):
3034
  """Distribute additional files which are part of the cluster configuration.
3035

3036
  ConfigWriter takes care of distributing the config and ssconf files, but
3037
  there are more files which should be distributed to all nodes. This function
3038
  makes sure those are copied.
3039

3040
  @param lu: calling logical unit
3041
  @param additional_nodes: list of nodes not in the config to distribute to
3042
  @type additional_vm: boolean
3043
  @param additional_vm: whether the additional nodes are vm-capable or not
3044

3045
  """
3046
  # 1. Gather target nodes
3047
  myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
3048
  dist_nodes = lu.cfg.GetOnlineNodeList()
3049
  nvm_nodes = lu.cfg.GetNonVmCapableNodeList()
3050
  vm_nodes = [name for name in dist_nodes if name not in nvm_nodes]
3051
  if additional_nodes is not None:
3052
    dist_nodes.extend(additional_nodes)
3053
    if additional_vm:
3054
      vm_nodes.extend(additional_nodes)
3055
  if myself.name in dist_nodes:
3056
    dist_nodes.remove(myself.name)
3057
  if myself.name in vm_nodes:
3058
    vm_nodes.remove(myself.name)
3059

    
3060
  # 2. Gather files to distribute
3061
  dist_files = set([constants.ETC_HOSTS,
3062
                    constants.SSH_KNOWN_HOSTS_FILE,
3063
                    constants.RAPI_CERT_FILE,
3064
                    constants.RAPI_USERS_FILE,
3065
                    constants.CONFD_HMAC_KEY,
3066
                    constants.CLUSTER_DOMAIN_SECRET_FILE,
3067
                   ])
3068

    
3069
  vm_files = set()
3070
  enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
3071
  for hv_name in enabled_hypervisors:
3072
    hv_class = hypervisor.GetHypervisor(hv_name)
3073
    vm_files.update(hv_class.GetAncillaryFiles())
3074

    
3075
  # 3. Perform the files upload
3076
  for fname in dist_files:
3077
    _UploadHelper(lu, dist_nodes, fname)
3078
  for fname in vm_files:
3079
    _UploadHelper(lu, vm_nodes, fname)
3080

    
3081

    
3082
class LUClusterRedistConf(NoHooksLU):
3083
  """Force the redistribution of cluster configuration.
3084

3085
  This is a very simple LU.
3086

3087
  """
3088
  REQ_BGL = False
3089

    
3090
  def ExpandNames(self):
3091
    self.needed_locks = {
3092
      locking.LEVEL_NODE: locking.ALL_SET,
3093
    }
3094
    self.share_locks[locking.LEVEL_NODE] = 1
3095

    
3096
  def Exec(self, feedback_fn):
3097
    """Redistribute the configuration.
3098

3099
    """
3100
    self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
3101
    _RedistributeAncillaryFiles(self)
3102

    
3103

    
3104
def _WaitForSync(lu, instance, disks=None, oneshot=False):
3105
  """Sleep and poll for an instance's disk to sync.
3106

3107
  """
3108
  if not instance.disks or disks is not None and not disks:
3109
    return True
3110

    
3111
  disks = _ExpandCheckDisks(instance, disks)
3112

    
3113
  if not oneshot:
3114
    lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
3115

    
3116
  node = instance.primary_node
3117

    
3118
  for dev in disks:
3119
    lu.cfg.SetDiskID(dev, node)
3120

    
3121
  # TODO: Convert to utils.Retry
3122

    
3123
  retries = 0
3124
  degr_retries = 10 # in seconds, as we sleep 1 second each time
3125
  while True:
3126
    max_time = 0
3127
    done = True
3128
    cumul_degraded = False
3129
    rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
3130
    msg = rstats.fail_msg
3131
    if msg:
3132
      lu.LogWarning("Can't get any data from node %s: %s", node, msg)
3133
      retries += 1
3134
      if retries >= 10:
3135
        raise errors.RemoteError("Can't contact node %s for mirror data,"
3136
                                 " aborting." % node)
3137
      time.sleep(6)
3138
      continue
3139
    rstats = rstats.payload
3140
    retries = 0
3141
    for i, mstat in enumerate(rstats):
3142
      if mstat is None:
3143
        lu.LogWarning("Can't compute data for node %s/%s",
3144
                           node, disks[i].iv_name)
3145
        continue
3146

    
3147
      cumul_degraded = (cumul_degraded or
3148
                        (mstat.is_degraded and mstat.sync_percent is None))
3149
      if mstat.sync_percent is not None:
3150
        done = False
3151
        if mstat.estimated_time is not None:
3152
          rem_time = ("%s remaining (estimated)" %
3153
                      utils.FormatSeconds(mstat.estimated_time))
3154
          max_time = mstat.estimated_time
3155
        else:
3156
          rem_time = "no time estimate"
3157
        lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
3158
                        (disks[i].iv_name, mstat.sync_percent, rem_time))
3159

    
3160
    # if we're done but degraded, let's do a few small retries, to
3161
    # make sure we see a stable and not transient situation; therefore
3162
    # we force restart of the loop
3163
    if (done or oneshot) and cumul_degraded and degr_retries > 0:
3164
      logging.info("Degraded disks found, %d retries left", degr_retries)
3165
      degr_retries -= 1
3166
      time.sleep(1)
3167
      continue
3168

    
3169
    if done or oneshot:
3170
      break
3171

    
3172
    time.sleep(min(60, max_time))
3173

    
3174
  if done:
3175
    lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
3176
  return not cumul_degraded
3177

    
3178

    
3179
def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
3180
  """Check that mirrors are not degraded.
3181

3182
  The ldisk parameter, if True, will change the test from the
3183
  is_degraded attribute (which represents overall non-ok status for
3184
  the device(s)) to the ldisk (representing the local storage status).
3185

3186
  """
3187
  lu.cfg.SetDiskID(dev, node)
3188

    
3189
  result = True
3190

    
3191
  if on_primary or dev.AssembleOnSecondary():
3192
    rstats = lu.rpc.call_blockdev_find(node, dev)
3193
    msg = rstats.fail_msg
3194
    if msg:
3195
      lu.LogWarning("Can't find disk on node %s: %s", node, msg)
3196
      result = False
3197
    elif not rstats.payload:
3198
      lu.LogWarning("Can't find disk on node %s", node)
3199
      result = False
3200
    else:
3201
      if ldisk:
3202
        result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
3203
      else:
3204
        result = result and not rstats.payload.is_degraded
3205

    
3206
  if dev.children:
3207
    for child in dev.children:
3208
      result = result and _CheckDiskConsistency(lu, child, node, on_primary)
3209

    
3210
  return result
3211

    
3212

    
3213
class LUOobCommand(NoHooksLU):
3214
  """Logical unit for OOB handling.
3215

3216
  """
3217
  REG_BGL = False
3218

    
3219
  def CheckPrereq(self):
3220
    """Check prerequisites.
3221

3222
    This checks:
3223
     - the node exists in the configuration
3224
     - OOB is supported
3225

3226
    Any errors are signaled by raising errors.OpPrereqError.
3227

3228
    """
3229
    self.nodes = []
3230
    for node_name in self.op.node_names:
3231
      node = self.cfg.GetNodeInfo(node_name)
3232

    
3233
      if node is None:
3234
        raise errors.OpPrereqError("Node %s not found" % node_name,
3235
                                   errors.ECODE_NOENT)
3236
      else:
3237
        self.nodes.append(node)
3238

    
3239
      if (self.op.command == constants.OOB_POWER_OFF and not node.offline):
3240
        raise errors.OpPrereqError(("Cannot power off node %s because it is"
3241
                                    " not marked offline") % node_name,
3242
                                   errors.ECODE_STATE)
3243

    
3244
  def ExpandNames(self):
3245
    """Gather locks we need.
3246

3247
    """
3248
    if self.op.node_names:
3249
      self.op.node_names = [_ExpandNodeName(self.cfg, name)
3250
                            for name in self.op.node_names]
3251
    else:
3252
      self.op.node_names = self.cfg.GetNodeList()
3253

    
3254
    self.needed_locks = {
3255
      locking.LEVEL_NODE: self.op.node_names,
3256
      }
3257

    
3258
  def Exec(self, feedback_fn):
3259
    """Execute OOB and return result if we expect any.
3260

3261
    """
3262
    master_node = self.cfg.GetMasterNode()
3263
    ret = []
3264

    
3265
    for node in self.nodes:
3266
      node_entry = [(constants.RS_NORMAL, node.name)]
3267
      ret.append(node_entry)
3268

    
3269
      oob_program = _SupportsOob(self.cfg, node)
3270

    
3271
      if not oob_program:
3272
        node_entry.append((constants.RS_UNAVAIL, None))
3273
        continue
3274

    
3275
      logging.info("Executing out-of-band command '%s' using '%s' on %s",
3276
                   self.op.command, oob_program, node.name)
3277
      result = self.rpc.call_run_oob(master_node, oob_program,
3278
                                     self.op.command, node.name,
3279
                                     self.op.timeout)
3280

    
3281
      if result.fail_msg:
3282
        self.LogWarning("On node '%s' out-of-band RPC failed with: %s",
3283
                        node.name, result.fail_msg)
3284
        node_entry.append((constants.RS_NODATA, None))
3285
      else:
3286
        try:
3287
          self._CheckPayload(result)
3288
        except errors.OpExecError, err:
3289
          self.LogWarning("The payload returned by '%s' is not valid: %s",
3290
                          node.name, err)
3291
          node_entry.append((constants.RS_NODATA, None))
3292
        else:
3293
          if self.op.command == constants.OOB_HEALTH:
3294
            # For health we should log important events
3295
            for item, status in result.payload:
3296
              if status in [constants.OOB_STATUS_WARNING,
3297
                            constants.OOB_STATUS_CRITICAL]:
3298
                self.LogWarning("On node '%s' item '%s' has status '%s'",
3299
                                node.name, item, status)
3300

    
3301
          if self.op.command == constants.OOB_POWER_ON:
3302
            node.powered = True
3303
          elif self.op.command == constants.OOB_POWER_OFF:
3304
            node.powered = False
3305
          elif self.op.command == constants.OOB_POWER_STATUS:
3306
            powered = result.payload[constants.OOB_POWER_STATUS_POWERED]
3307
            if powered != node.powered:
3308
              logging.warning(("Recorded power state (%s) of node '%s' does not"
3309
                               " match actual power state (%s)"), node.powered,
3310
                              node.name, powered)
3311

    
3312
          # For configuration changing commands we should update the node
3313
          if self.op.command in (constants.OOB_POWER_ON,
3314
                                 constants.OOB_POWER_OFF):
3315
            self.cfg.Update(node, feedback_fn)
3316

    
3317
          node_entry.append((constants.RS_NORMAL, result.payload))
3318

    
3319
    return ret
3320

    
3321
  def _CheckPayload(self, result):
3322
    """Checks if the payload is valid.
3323

3324
    @param result: RPC result
3325
    @raises errors.OpExecError: If payload is not valid
3326

3327
    """
3328
    errs = []
3329
    if self.op.command == constants.OOB_HEALTH:
3330
      if not isinstance(result.payload, list):
3331
        errs.append("command 'health' is expected to return a list but got %s" %
3332
                    type(result.payload))
3333
      else:
3334
        for item, status in result.payload:
3335
          if status not in constants.OOB_STATUSES:
3336
            errs.append("health item '%s' has invalid status '%s'" %
3337
                        (item, status))
3338

    
3339
    if self.op.command == constants.OOB_POWER_STATUS:
3340
      if not isinstance(result.payload, dict):
3341
        errs.append("power-status is expected to return a dict but got %s" %
3342
                    type(result.payload))
3343

    
3344
    if self.op.command in [
3345
        constants.OOB_POWER_ON,
3346
        constants.OOB_POWER_OFF,
3347
        constants.OOB_POWER_CYCLE,
3348
        ]:
3349
      if result.payload is not None:
3350
        errs.append("%s is expected to not return payload but got '%s'" %
3351
                    (self.op.command, result.payload))
3352

    
3353
    if errs:
3354
      raise errors.OpExecError("Check of out-of-band payload failed due to %s" %
3355
                               utils.CommaJoin(errs))
3356

    
3357

    
3358

    
3359
class LUOsDiagnose(NoHooksLU):
3360
  """Logical unit for OS diagnose/query.
3361

3362
  """
3363
  REQ_BGL = False
3364
  _HID = "hidden"
3365
  _BLK = "blacklisted"
3366
  _VLD = "valid"
3367
  _FIELDS_STATIC = utils.FieldSet()
3368
  _FIELDS_DYNAMIC = utils.FieldSet("name", _VLD, "node_status", "variants",
3369
                                   "parameters", "api_versions", _HID, _BLK)
3370

    
3371
  def CheckArguments(self):
3372
    if self.op.names:
3373
      raise errors.OpPrereqError("Selective OS query not supported",
3374
                                 errors.ECODE_INVAL)
3375

    
3376
    _CheckOutputFields(static=self._FIELDS_STATIC,
3377
                       dynamic=self._FIELDS_DYNAMIC,
3378
                       selected=self.op.output_fields)
3379

    
3380
  def ExpandNames(self):
3381
    # Lock all nodes, in shared mode
3382
    # Temporary removal of locks, should be reverted later
3383
    # TODO: reintroduce locks when they are lighter-weight
3384
    self.needed_locks = {}
3385
    #self.share_locks[locking.LEVEL_NODE] = 1
3386
    #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3387

    
3388
  @staticmethod
3389
  def _DiagnoseByOS(rlist):
3390
    """Remaps a per-node return list into an a per-os per-node dictionary
3391

3392
    @param rlist: a map with node names as keys and OS objects as values
3393

3394
    @rtype: dict
3395
    @return: a dictionary with osnames as keys and as value another
3396
        map, with nodes as keys and tuples of (path, status, diagnose,
3397
        variants, parameters, api_versions) as values, eg::
3398

3399
          {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
3400
                                     (/srv/..., False, "invalid api")],
3401
                           "node2": [(/srv/..., True, "", [], [])]}
3402
          }
3403

3404
    """
3405
    all_os = {}
3406
    # we build here the list of nodes that didn't fail the RPC (at RPC
3407
    # level), so that nodes with a non-responding node daemon don't
3408
    # make all OSes invalid
3409
    good_nodes = [node_name for node_name in rlist
3410
                  if not rlist[node_name].fail_msg]
3411
    for node_name, nr in rlist.items():
3412
      if nr.fail_msg or not nr.payload:
3413
        continue
3414
      for (name, path, status, diagnose, variants,
3415
           params, api_versions) in nr.payload:
3416
        if name not in all_os:
3417
          # build a list of nodes for this os containing empty lists
3418
          # for each node in node_list
3419
          all_os[name] = {}
3420
          for nname in good_nodes:
3421
            all_os[name][nname] = []
3422
        # convert params from [name, help] to (name, help)
3423
        params = [tuple(v) for v in params]
3424
        all_os[name][node_name].append((path, status, diagnose,
3425
                                        variants, params, api_versions))
3426
    return all_os
3427

    
3428
  def Exec(self, feedback_fn):
3429
    """Compute the list of OSes.
3430

3431
    """
3432
    valid_nodes = [node.name
3433
                   for node in self.cfg.GetAllNodesInfo().values()
3434
                   if not node.offline and node.vm_capable]
3435
    node_data = self.rpc.call_os_diagnose(valid_nodes)
3436
    pol = self._DiagnoseByOS(node_data)
3437
    output = []
3438
    cluster = self.cfg.GetClusterInfo()
3439

    
3440
    for os_name in utils.NiceSort(pol.keys()):
3441
      os_data = pol[os_name]
3442
      row = []
3443
      valid = True
3444
      (variants, params, api_versions) = null_state = (set(), set(), set())
3445
      for idx, osl in enumerate(os_data.values()):
3446
        valid = bool(valid and osl and osl[0][1])
3447
        if not valid:
3448
          (variants, params, api_versions) = null_state
3449
          break
3450
        node_variants, node_params, node_api = osl[0][3:6]
3451
        if idx == 0: # first entry
3452
          variants = set(node_variants)
3453
          params = set(node_params)
3454
          api_versions = set(node_api)
3455
        else: # keep consistency
3456
          variants.intersection_update(node_variants)
3457
          params.intersection_update(node_params)
3458
          api_versions.intersection_update(node_api)
3459

    
3460
      is_hid = os_name in cluster.hidden_os
3461
      is_blk = os_name in cluster.blacklisted_os
3462
      if ((self._HID not in self.op.output_fields and is_hid) or
3463
          (self._BLK not in self.op.output_fields and is_blk) or
3464
          (self._VLD not in self.op.output_fields and not valid)):
3465
        continue
3466

    
3467
      for field in self.op.output_fields:
3468
        if field == "name":
3469
          val = os_name
3470
        elif field == self._VLD:
3471
          val = valid
3472
        elif field == "node_status":
3473
          # this is just a copy of the dict
3474
          val = {}
3475
          for node_name, nos_list in os_data.items():
3476
            val[node_name] = nos_list
3477
        elif field == "variants":
3478
          val = utils.NiceSort(list(variants))
3479
        elif field == "parameters":
3480
          val = list(params)
3481
        elif field == "api_versions":
3482
          val = list(api_versions)
3483
        elif field == self._HID:
3484
          val = is_hid
3485
        elif field == self._BLK:
3486
          val = is_blk
3487
        else:
3488
          raise errors.ParameterError(field)
3489
        row.append(val)
3490
      output.append(row)
3491

    
3492
    return output
3493

    
3494

    
3495
class LUNodeRemove(LogicalUnit):
3496
  """Logical unit for removing a node.
3497

3498
  """
3499
  HPATH = "node-remove"
3500
  HTYPE = constants.HTYPE_NODE
3501

    
3502
  def BuildHooksEnv(self):
3503
    """Build hooks env.
3504

3505
    This doesn't run on the target node in the pre phase as a failed
3506
    node would then be impossible to remove.
3507

3508
    """
3509
    env = {
3510
      "OP_TARGET": self.op.node_name,
3511
      "NODE_NAME": self.op.node_name,
3512
      }
3513
    all_nodes = self.cfg.GetNodeList()
3514
    try:
3515
      all_nodes.remove(self.op.node_name)
3516
    except ValueError:
3517
      logging.warning("Node %s which is about to be removed not found"
3518
                      " in the all nodes list", self.op.node_name)
3519
    return env, all_nodes, all_nodes
3520

    
3521
  def CheckPrereq(self):
3522
    """Check prerequisites.
3523

3524
    This checks:
3525
     - the node exists in the configuration
3526
     - it does not have primary or secondary instances
3527
     - it's not the master
3528

3529
    Any errors are signaled by raising errors.OpPrereqError.
3530

3531
    """
3532
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3533
    node = self.cfg.GetNodeInfo(self.op.node_name)
3534
    assert node is not None
3535

    
3536
    instance_list = self.cfg.GetInstanceList()
3537

    
3538
    masternode = self.cfg.GetMasterNode()
3539
    if node.name == masternode:
3540
      raise errors.OpPrereqError("Node is the master node,"
3541
                                 " you need to failover first.",
3542
                                 errors.ECODE_INVAL)
3543

    
3544
    for instance_name in instance_list:
3545
      instance = self.cfg.GetInstanceInfo(instance_name)
3546
      if node.name in instance.all_nodes:
3547
        raise errors.OpPrereqError("Instance %s is still running on the node,"
3548
                                   " please remove first." % instance_name,
3549
                                   errors.ECODE_INVAL)
3550
    self.op.node_name = node.name
3551
    self.node = node
3552

    
3553
  def Exec(self, feedback_fn):
3554
    """Removes the node from the cluster.
3555

3556
    """
3557
    node = self.node
3558
    logging.info("Stopping the node daemon and removing configs from node %s",
3559
                 node.name)
3560

    
3561
    modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
3562

    
3563
    # Promote nodes to master candidate as needed
3564
    _AdjustCandidatePool(self, exceptions=[node.name])
3565
    self.context.RemoveNode(node.name)
3566

    
3567
    # Run post hooks on the node before it's removed
3568
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
3569
    try:
3570
      hm.RunPhase(constants.HOOKS_PHASE_POST, [node.name])
3571
    except:
3572
      # pylint: disable-msg=W0702
3573
      self.LogWarning("Errors occurred running hooks on %s" % node.name)
3574

    
3575
    result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
3576
    msg = result.fail_msg
3577
    if msg:
3578
      self.LogWarning("Errors encountered on the remote node while leaving"
3579
                      " the cluster: %s", msg)
3580

    
3581
    # Remove node from our /etc/hosts
3582
    if self.cfg.GetClusterInfo().modify_etc_hosts:
3583
      master_node = self.cfg.GetMasterNode()
3584
      result = self.rpc.call_etc_hosts_modify(master_node,
3585
                                              constants.ETC_HOSTS_REMOVE,
3586
                                              node.name, None)
3587
      result.Raise("Can't update hosts file with new host data")
3588
      _RedistributeAncillaryFiles(self)
3589

    
3590

    
3591
class _NodeQuery(_QueryBase):
3592
  FIELDS = query.NODE_FIELDS
3593

    
3594
  def ExpandNames(self, lu):
3595
    lu.needed_locks = {}
3596
    lu.share_locks[locking.LEVEL_NODE] = 1
3597

    
3598
    if self.names:
3599
      self.wanted = _GetWantedNodes(lu, self.names)
3600
    else:
3601
      self.wanted = locking.ALL_SET
3602

    
3603
    self.do_locking = (self.use_locking and
3604
                       query.NQ_LIVE in self.requested_data)
3605

    
3606
    if self.do_locking:
3607
      # if we don't request only static fields, we need to lock the nodes
3608
      lu.needed_locks[locking.LEVEL_NODE] = self.wanted
3609

    
3610
  def DeclareLocks(self, lu, level):
3611
    pass
3612

    
3613
  def _GetQueryData(self, lu):
3614
    """Computes the list of nodes and their attributes.
3615

3616
    """
3617
    all_info = lu.cfg.GetAllNodesInfo()
3618

    
3619
    nodenames = self._GetNames(lu, all_info.keys(), locking.LEVEL_NODE)
3620

    
3621
    # Gather data as requested
3622
    if query.NQ_LIVE in self.requested_data:
3623
      node_data = lu.rpc.call_node_info(nodenames, lu.cfg.GetVGName(),
3624
                                        lu.cfg.GetHypervisorType())
3625
      live_data = dict((name, nresult.payload)
3626
                       for (name, nresult) in node_data.items()
3627
                       if not nresult.fail_msg and nresult.payload)
3628
    else:
3629
      live_data = None
3630

    
3631
    if query.NQ_INST in self.requested_data:
3632
      node_to_primary = dict([(name, set()) for name in nodenames])
3633
      node_to_secondary = dict([(name, set()) for name in nodenames])
3634

    
3635
      inst_data = lu.cfg.GetAllInstancesInfo()
3636

    
3637
      for inst in inst_data.values():
3638
        if inst.primary_node in node_to_primary:
3639
          node_to_primary[inst.primary_node].add(inst.name)
3640
        for secnode in inst.secondary_nodes:
3641
          if secnode in node_to_secondary:
3642
            node_to_secondary[secnode].add(inst.name)
3643
    else:
3644
      node_to_primary = None
3645
      node_to_secondary = None
3646

    
3647
    if query.NQ_OOB in self.requested_data:
3648
      oob_support = dict((name, bool(_SupportsOob(lu.cfg, node)))
3649
                         for name, node in all_info.iteritems())
3650
    else:
3651
      oob_support = None
3652

    
3653
    if query.NQ_GROUP in self.requested_data:
3654
      groups = lu.cfg.GetAllNodeGroupsInfo()
3655
    else:
3656
      groups = {}
3657

    
3658
    return query.NodeQueryData([all_info[name] for name in nodenames],
3659
                               live_data, lu.cfg.GetMasterNode(),
3660
                               node_to_primary, node_to_secondary, groups,
3661
                               oob_support, lu.cfg.GetClusterInfo())
3662

    
3663

    
3664
class LUNodeQuery(NoHooksLU):
3665
  """Logical unit for querying nodes.
3666

3667
  """
3668
  # pylint: disable-msg=W0142
3669
  REQ_BGL = False
3670

    
3671
  def CheckArguments(self):
3672
    self.nq = _NodeQuery(self.op.names, self.op.output_fields,
3673
                         self.op.use_locking)
3674

    
3675
  def ExpandNames(self):
3676
    self.nq.ExpandNames(self)
3677

    
3678
  def Exec(self, feedback_fn):
3679
    return self.nq.OldStyleQuery(self)
3680

    
3681

    
3682
class LUNodeQueryvols(NoHooksLU):
3683
  """Logical unit for getting volumes on node(s).
3684

3685
  """
3686
  REQ_BGL = False
3687
  _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
3688
  _FIELDS_STATIC = utils.FieldSet("node")
3689

    
3690
  def CheckArguments(self):
3691
    _CheckOutputFields(static=self._FIELDS_STATIC,
3692
                       dynamic=self._FIELDS_DYNAMIC,
3693
                       selected=self.op.output_fields)
3694

    
3695
  def ExpandNames(self):
3696
    self.needed_locks = {}
3697
    self.share_locks[locking.LEVEL_NODE] = 1
3698
    if not self.op.nodes:
3699
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3700
    else:
3701
      self.needed_locks[locking.LEVEL_NODE] = \
3702
        _GetWantedNodes(self, self.op.nodes)
3703

    
3704
  def Exec(self, feedback_fn):
3705
    """Computes the list of nodes and their attributes.
3706

3707
    """
3708
    nodenames = self.acquired_locks[locking.LEVEL_NODE]
3709
    volumes = self.rpc.call_node_volumes(nodenames)
3710

    
3711
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
3712
             in self.cfg.GetInstanceList()]
3713

    
3714
    lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
3715

    
3716
    output = []
3717
    for node in nodenames:
3718
      nresult = volumes[node]
3719
      if nresult.offline:
3720
        continue
3721
      msg = nresult.fail_msg
3722
      if msg:
3723
        self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
3724
        continue
3725

    
3726
      node_vols = nresult.payload[:]
3727
      node_vols.sort(key=lambda vol: vol['dev'])
3728

    
3729
      for vol in node_vols:
3730
        node_output = []
3731
        for field in self.op.output_fields:
3732
          if field == "node":
3733
            val = node
3734
          elif field == "phys":
3735
            val = vol['dev']
3736
          elif field == "vg":
3737
            val = vol['vg']
3738
          elif field == "name":
3739
            val = vol['name']
3740
          elif field == "size":
3741
            val = int(float(vol['size']))
3742
          elif field == "instance":
3743
            for inst in ilist:
3744
              if node not in lv_by_node[inst]:
3745
                continue
3746
              if vol['name'] in lv_by_node[inst][node]:
3747
                val = inst.name
3748
                break
3749
            else:
3750
              val = '-'
3751
          else:
3752
            raise errors.ParameterError(field)
3753
          node_output.append(str(val))
3754

    
3755
        output.append(node_output)
3756

    
3757
    return output
3758

    
3759

    
3760
class LUNodeQueryStorage(NoHooksLU):
3761
  """Logical unit for getting information on storage units on node(s).
3762

3763
  """
3764
  _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
3765
  REQ_BGL = False
3766

    
3767
  def CheckArguments(self):
3768
    _CheckOutputFields(static=self._FIELDS_STATIC,
3769
                       dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
3770
                       selected=self.op.output_fields)
3771

    
3772
  def ExpandNames(self):
3773
    self.needed_locks = {}
3774
    self.share_locks[locking.LEVEL_NODE] = 1
3775

    
3776
    if self.op.nodes:
3777
      self.needed_locks[locking.LEVEL_NODE] = \
3778
        _GetWantedNodes(self, self.op.nodes)
3779
    else:
3780
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3781

    
3782
  def Exec(self, feedback_fn):
3783
    """Computes the list of nodes and their attributes.
3784

3785
    """
3786
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
3787

    
3788
    # Always get name to sort by
3789
    if constants.SF_NAME in self.op.output_fields:
3790
      fields = self.op.output_fields[:]
3791
    else:
3792
      fields = [constants.SF_NAME] + self.op.output_fields
3793

    
3794
    # Never ask for node or type as it's only known to the LU
3795
    for extra in [constants.SF_NODE, constants.SF_TYPE]:
3796
      while extra in fields:
3797
        fields.remove(extra)
3798

    
3799
    field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
3800
    name_idx = field_idx[constants.SF_NAME]
3801

    
3802
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3803
    data = self.rpc.call_storage_list(self.nodes,
3804
                                      self.op.storage_type, st_args,
3805
                                      self.op.name, fields)
3806

    
3807
    result = []
3808

    
3809
    for node in utils.NiceSort(self.nodes):
3810
      nresult = data[node]
3811
      if nresult.offline:
3812
        continue
3813

    
3814
      msg = nresult.fail_msg
3815
      if msg:
3816
        self.LogWarning("Can't get storage data from node %s: %s", node, msg)
3817
        continue
3818

    
3819
      rows = dict([(row[name_idx], row) for row in nresult.payload])
3820

    
3821
      for name in utils.NiceSort(rows.keys()):
3822
        row = rows[name]
3823

    
3824
        out = []
3825

    
3826
        for field in self.op.output_fields:
3827
          if field == constants.SF_NODE:
3828
            val = node
3829
          elif field == constants.SF_TYPE:
3830
            val = self.op.storage_type
3831
          elif field in field_idx:
3832
            val = row[field_idx[field]]
3833
          else:
3834
            raise errors.ParameterError(field)
3835

    
3836
          out.append(val)
3837

    
3838
        result.append(out)
3839

    
3840
    return result
3841

    
3842

    
3843
class _InstanceQuery(_QueryBase):
3844
  FIELDS = query.INSTANCE_FIELDS
3845

    
3846
  def ExpandNames(self, lu):
3847
    lu.needed_locks = {}
3848
    lu.share_locks[locking.LEVEL_INSTANCE] = 1
3849
    lu.share_locks[locking.LEVEL_NODE] = 1
3850

    
3851
    if self.names:
3852
      self.wanted = _GetWantedInstances(lu, self.names)
3853
    else:
3854
      self.wanted = locking.ALL_SET
3855

    
3856
    self.do_locking = (self.use_locking and
3857
                       query.IQ_LIVE in self.requested_data)
3858
    if self.do_locking:
3859
      lu.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
3860
      lu.needed_locks[locking.LEVEL_NODE] = []
3861
      lu.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3862

    
3863
  def DeclareLocks(self, lu, level):
3864
    if level == locking.LEVEL_NODE and self.do_locking:
3865
      lu._LockInstancesNodes() # pylint: disable-msg=W0212
3866

    
3867
  def _GetQueryData(self, lu):
3868
    """Computes the list of instances and their attributes.
3869

3870
    """
3871
    cluster = lu.cfg.GetClusterInfo()
3872
    all_info = lu.cfg.GetAllInstancesInfo()
3873

    
3874
    instance_names = self._GetNames(lu, all_info.keys(), locking.LEVEL_INSTANCE)
3875

    
3876
    instance_list = [all_info[name] for name in instance_names]
3877
    nodes = frozenset(itertools.chain(*(inst.all_nodes
3878
                                        for inst in instance_list)))
3879
    hv_list = list(set([inst.hypervisor for inst in instance_list]))
3880
    bad_nodes = []
3881
    offline_nodes = []
3882
    wrongnode_inst = set()
3883

    
3884
    # Gather data as requested
3885
    if self.requested_data & set([query.IQ_LIVE, query.IQ_CONSOLE]):
3886
      live_data = {}
3887
      node_data = lu.rpc.call_all_instances_info(nodes, hv_list)
3888
      for name in nodes:
3889
        result = node_data[name]
3890
        if result.offline:
3891
          # offline nodes will be in both lists
3892
          assert result.fail_msg
3893
          offline_nodes.append(name)
3894
        if result.fail_msg:
3895
          bad_nodes.append(name)
3896
        elif result.payload:
3897
          for inst in result.payload:
3898
            if all_info[inst].primary_node == name:
3899
              live_data.update(result.payload)
3900
            else:
3901
              wrongnode_inst.add(inst)
3902
        # else no instance is alive
3903
    else:
3904
      live_data = {}
3905

    
3906
    if query.IQ_DISKUSAGE in self.requested_data:
3907
      disk_usage = dict((inst.name,
3908
                         _ComputeDiskSize(inst.disk_template,
3909
                                          [{"size": disk.size}
3910
                                           for disk in inst.disks]))
3911
                        for inst in instance_list)
3912
    else:
3913
      disk_usage = None
3914

    
3915
    if query.IQ_CONSOLE in self.requested_data:
3916
      consinfo = {}
3917
      for inst in instance_list:
3918
        if inst.name in live_data:
3919
          # Instance is running
3920
          consinfo[inst.name] = _GetInstanceConsole(cluster, inst)
3921
        else:
3922
          consinfo[inst.name] = None
3923
      assert set(consinfo.keys()) == set(instance_names)
3924
    else:
3925
      consinfo = None
3926

    
3927
    return query.InstanceQueryData(instance_list, lu.cfg.GetClusterInfo(),
3928
                                   disk_usage, offline_nodes, bad_nodes,
3929
                                   live_data, wrongnode_inst, consinfo)
3930

    
3931

    
3932
class LUQuery(NoHooksLU):
3933
  """Query for resources/items of a certain kind.
3934

3935
  """
3936
  # pylint: disable-msg=W0142
3937
  REQ_BGL = False
3938

    
3939
  def CheckArguments(self):
3940
    qcls = _GetQueryImplementation(self.op.what)
3941
    names = qlang.ReadSimpleFilter("name", self.op.filter)
3942

    
3943
    self.impl = qcls(names, self.op.fields, False)
3944

    
3945
  def ExpandNames(self):
3946
    self.impl.ExpandNames(self)
3947

    
3948
  def DeclareLocks(self, level):
3949
    self.impl.DeclareLocks(self, level)
3950

    
3951
  def Exec(self, feedback_fn):
3952
    return self.impl.NewStyleQuery(self)
3953

    
3954

    
3955
class LUQueryFields(NoHooksLU):
3956
  """Query for resources/items of a certain kind.
3957

3958
  """
3959
  # pylint: disable-msg=W0142
3960
  REQ_BGL = False
3961

    
3962
  def CheckArguments(self):
3963
    self.qcls = _GetQueryImplementation(self.op.what)
3964

    
3965
  def ExpandNames(self):
3966
    self.needed_locks = {}
3967

    
3968
  def Exec(self, feedback_fn):
3969
    return self.qcls.FieldsQuery(self.op.fields)
3970

    
3971

    
3972
class LUNodeModifyStorage(NoHooksLU):
3973
  """Logical unit for modifying a storage volume on a node.
3974

3975
  """
3976
  REQ_BGL = False
3977

    
3978
  def CheckArguments(self):
3979
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3980

    
3981
    storage_type = self.op.storage_type
3982

    
3983
    try:
3984
      modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
3985
    except KeyError:
3986
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
3987
                                 " modified" % storage_type,
3988
                                 errors.ECODE_INVAL)
3989

    
3990
    diff = set(self.op.changes.keys()) - modifiable
3991
    if diff:
3992
      raise errors.OpPrereqError("The following fields can not be modified for"
3993
                                 " storage units of type '%s': %r" %
3994
                                 (storage_type, list(diff)),
3995
                                 errors.ECODE_INVAL)
3996

    
3997
  def ExpandNames(self):
3998
    self.needed_locks = {
3999
      locking.LEVEL_NODE: self.op.node_name,
4000
      }
4001

    
4002
  def Exec(self, feedback_fn):
4003
    """Computes the list of nodes and their attributes.
4004

4005
    """
4006
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
4007
    result = self.rpc.call_storage_modify(self.op.node_name,
4008
                                          self.op.storage_type, st_args,
4009
                                          self.op.name, self.op.changes)
4010
    result.Raise("Failed to modify storage unit '%s' on %s" %
4011
                 (self.op.name, self.op.node_name))
4012

    
4013

    
4014
class LUNodeAdd(LogicalUnit):
4015
  """Logical unit for adding node to the cluster.
4016

4017
  """
4018
  HPATH = "node-add"
4019
  HTYPE = constants.HTYPE_NODE
4020
  _NFLAGS = ["master_capable", "vm_capable"]
4021

    
4022
  def CheckArguments(self):
4023
    self.primary_ip_family = self.cfg.GetPrimaryIPFamily()
4024
    # validate/normalize the node name
4025
    self.hostname = netutils.GetHostname(name=self.op.node_name,
4026
                                         family=self.primary_ip_family)
4027
    self.op.node_name = self.hostname.name
4028
    if self.op.readd and self.op.group:
4029
      raise errors.OpPrereqError("Cannot pass a node group when a node is"
4030
                                 " being readded", errors.ECODE_INVAL)
4031

    
4032
  def BuildHooksEnv(self):
4033
    """Build hooks env.
4034

4035
    This will run on all nodes before, and on all nodes + the new node after.
4036

4037
    """
4038
    env = {
4039
      "OP_TARGET": self.op.node_name,
4040
      "NODE_NAME": self.op.node_name,
4041
      "NODE_PIP": self.op.primary_ip,
4042
      "NODE_SIP": self.op.secondary_ip,
4043
      "MASTER_CAPABLE": str(self.op.master_capable),
4044
      "VM_CAPABLE": str(self.op.vm_capable),
4045
      }
4046
    nodes_0 = self.cfg.GetNodeList()
4047
    nodes_1 = nodes_0 + [self.op.node_name, ]
4048
    return env, nodes_0, nodes_1
4049

    
4050
  def CheckPrereq(self):
4051
    """Check prerequisites.
4052

4053
    This checks:
4054
     - the new node is not already in the config
4055
     - it is resolvable
4056
     - its parameters (single/dual homed) matches the cluster
4057

4058
    Any errors are signaled by raising errors.OpPrereqError.
4059

4060
    """
4061
    cfg = self.cfg
4062
    hostname = self.hostname
4063
    node = hostname.name
4064
    primary_ip = self.op.primary_ip = hostname.ip
4065
    if self.op.secondary_ip is None:
4066
      if self.primary_ip_family == netutils.IP6Address.family:
4067
        raise errors.OpPrereqError("When using a IPv6 primary address, a valid"
4068
                                   " IPv4 address must be given as secondary",
4069
                                   errors.ECODE_INVAL)
4070
      self.op.secondary_ip = primary_ip
4071

    
4072
    secondary_ip = self.op.secondary_ip
4073
    if not netutils.IP4Address.IsValid(secondary_ip):
4074
      raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
4075
                                 " address" % secondary_ip, errors.ECODE_INVAL)
4076