Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ 5fbbd028

History | View | Annotate | Download (366.1 kB)

1
#
2
#
3

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

    
21

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

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

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

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

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

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

    
57
import ganeti.masterd.instance # pylint: disable-msg=W0611
58

    
59

    
60
# Modifiable default values; need to define these here before the
61
# actual LUs
62

    
63
def _EmptyList():
64
  """Returns an empty list.
65

66
  """
67
  return []
68

    
69

    
70
def _EmptyDict():
71
  """Returns an empty dict.
72

73
  """
74
  return {}
75

    
76

    
77
#: The without-default default value
78
_NoDefault = object()
79

    
80

    
81
#: The no-type (value to complex to check it in the type system)
82
_NoType = object()
83

    
84

    
85
# Some basic types
86
def _TNotNone(val):
87
  """Checks if the given value is not None.
88

89
  """
90
  return val is not None
91

    
92

    
93
def _TNone(val):
94
  """Checks if the given value is None.
95

96
  """
97
  return val is None
98

    
99

    
100
def _TBool(val):
101
  """Checks if the given value is a boolean.
102

103
  """
104
  return isinstance(val, bool)
105

    
106

    
107
def _TInt(val):
108
  """Checks if the given value is an integer.
109

110
  """
111
  return isinstance(val, int)
112

    
113

    
114
def _TFloat(val):
115
  """Checks if the given value is a float.
116

117
  """
118
  return isinstance(val, float)
119

    
120

    
121
def _TString(val):
122
  """Checks if the given value is a string.
123

124
  """
125
  return isinstance(val, basestring)
126

    
127

    
128
def _TTrue(val):
129
  """Checks if a given value evaluates to a boolean True value.
130

131
  """
132
  return bool(val)
133

    
134

    
135
def _TElemOf(target_list):
136
  """Builds a function that checks if a given value is a member of a list.
137

138
  """
139
  return lambda val: val in target_list
140

    
141

    
142
# Container types
143
def _TList(val):
144
  """Checks if the given value is a list.
145

146
  """
147
  return isinstance(val, list)
148

    
149

    
150
def _TDict(val):
151
  """Checks if the given value is a dictionary.
152

153
  """
154
  return isinstance(val, dict)
155

    
156

    
157
# Combinator types
158
def _TAnd(*args):
159
  """Combine multiple functions using an AND operation.
160

161
  """
162
  def fn(val):
163
    return compat.all(t(val) for t in args)
164
  return fn
165

    
166

    
167
def _TOr(*args):
168
  """Combine multiple functions using an AND operation.
169

170
  """
171
  def fn(val):
172
    return compat.any(t(val) for t in args)
173
  return fn
174

    
175

    
176
# Type aliases
177

    
178
#: a non-empty string
179
_TNonEmptyString = _TAnd(_TString, _TTrue)
180

    
181

    
182
#: a maybe non-empty string
183
_TMaybeString = _TOr(_TNonEmptyString, _TNone)
184

    
185

    
186
#: a maybe boolean (bool or none)
187
_TMaybeBool = _TOr(_TBool, _TNone)
188

    
189

    
190
#: a positive integer
191
_TPositiveInt = _TAnd(_TInt, lambda v: v >= 0)
192

    
193
#: a strictly positive integer
194
_TStrictPositiveInt = _TAnd(_TInt, lambda v: v > 0)
195

    
196

    
197
def _TListOf(my_type):
198
  """Checks if a given value is a list with all elements of the same type.
199

200
  """
201
  return _TAnd(_TList,
202
               lambda lst: compat.all(my_type(v) for v in lst))
203

    
204

    
205
def _TDictOf(key_type, val_type):
206
  """Checks a dict type for the type of its key/values.
207

208
  """
209
  return _TAnd(_TDict,
210
               lambda my_dict: (compat.all(key_type(v) for v in my_dict.keys())
211
                                and compat.all(val_type(v)
212
                                               for v in my_dict.values())))
213

    
214

    
215
# Common opcode attributes
216

    
217
#: output fields for a query operation
218
_POutputFields = ("output_fields", _NoDefault, _TListOf(_TNonEmptyString))
219

    
220

    
221
#: the shutdown timeout
222
_PShutdownTimeout = ("shutdown_timeout", constants.DEFAULT_SHUTDOWN_TIMEOUT,
223
                     _TPositiveInt)
224

    
225
#: the force parameter
226
_PForce = ("force", False, _TBool)
227

    
228
#: a required instance name (for single-instance LUs)
229
_PInstanceName = ("instance_name", _NoDefault, _TNonEmptyString)
230

    
231

    
232
#: a required node name (for single-node LUs)
233
_PNodeName = ("node_name", _NoDefault, _TNonEmptyString)
234

    
235
#: the migration type (live/non-live)
236
_PMigrationMode = ("mode", None, _TOr(_TNone,
237
                                      _TElemOf(constants.HT_MIGRATION_MODES)))
238

    
239
#: the obsolete 'live' mode (boolean)
240
_PMigrationLive = ("live", None, _TMaybeBool)
241

    
242

    
243
# End types
244
class LogicalUnit(object):
245
  """Logical Unit base class.
246

247
  Subclasses must follow these rules:
248
    - implement ExpandNames
249
    - implement CheckPrereq (except when tasklets are used)
250
    - implement Exec (except when tasklets are used)
251
    - implement BuildHooksEnv
252
    - redefine HPATH and HTYPE
253
    - optionally redefine their run requirements:
254
        REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
255

256
  Note that all commands require root permissions.
257

258
  @ivar dry_run_result: the value (if any) that will be returned to the caller
259
      in dry-run mode (signalled by opcode dry_run parameter)
260
  @cvar _OP_PARAMS: a list of opcode attributes, their defaults values
261
      they should get if not already defined, and types they must match
262

263
  """
264
  HPATH = None
265
  HTYPE = None
266
  _OP_PARAMS = []
267
  REQ_BGL = True
268

    
269
  def __init__(self, processor, op, context, rpc):
270
    """Constructor for LogicalUnit.
271

272
    This needs to be overridden in derived classes in order to check op
273
    validity.
274

275
    """
276
    self.proc = processor
277
    self.op = op
278
    self.cfg = context.cfg
279
    self.context = context
280
    self.rpc = rpc
281
    # Dicts used to declare locking needs to mcpu
282
    self.needed_locks = None
283
    self.acquired_locks = {}
284
    self.share_locks = dict.fromkeys(locking.LEVELS, 0)
285
    self.add_locks = {}
286
    self.remove_locks = {}
287
    # Used to force good behavior when calling helper functions
288
    self.recalculate_locks = {}
289
    self.__ssh = None
290
    # logging
291
    self.Log = processor.Log # pylint: disable-msg=C0103
292
    self.LogWarning = processor.LogWarning # pylint: disable-msg=C0103
293
    self.LogInfo = processor.LogInfo # pylint: disable-msg=C0103
294
    self.LogStep = processor.LogStep # pylint: disable-msg=C0103
295
    # support for dry-run
296
    self.dry_run_result = None
297
    # support for generic debug attribute
298
    if (not hasattr(self.op, "debug_level") or
299
        not isinstance(self.op.debug_level, int)):
300
      self.op.debug_level = 0
301

    
302
    # Tasklets
303
    self.tasklets = None
304

    
305
    # The new kind-of-type-system
306
    op_id = self.op.OP_ID
307
    for attr_name, aval, test in self._OP_PARAMS:
308
      if not hasattr(op, attr_name):
309
        if aval == _NoDefault:
310
          raise errors.OpPrereqError("Required parameter '%s.%s' missing" %
311
                                     (op_id, attr_name), errors.ECODE_INVAL)
312
        else:
313
          if callable(aval):
314
            dval = aval()
315
          else:
316
            dval = aval
317
          setattr(self.op, attr_name, dval)
318
      attr_val = getattr(op, attr_name)
319
      if test == _NoType:
320
        # no tests here
321
        continue
322
      if not callable(test):
323
        raise errors.ProgrammerError("Validation for parameter '%s.%s' failed,"
324
                                     " given type is not a proper type (%s)" %
325
                                     (op_id, attr_name, test))
326
      if not test(attr_val):
327
        logging.error("OpCode %s, parameter %s, has invalid type %s/value %s",
328
                      self.op.OP_ID, attr_name, type(attr_val), attr_val)
329
        raise errors.OpPrereqError("Parameter '%s.%s' fails validation" %
330
                                   (op_id, attr_name), errors.ECODE_INVAL)
331

    
332
    self.CheckArguments()
333

    
334
  def __GetSSH(self):
335
    """Returns the SshRunner object
336

337
    """
338
    if not self.__ssh:
339
      self.__ssh = ssh.SshRunner(self.cfg.GetClusterName())
340
    return self.__ssh
341

    
342
  ssh = property(fget=__GetSSH)
343

    
344
  def CheckArguments(self):
345
    """Check syntactic validity for the opcode arguments.
346

347
    This method is for doing a simple syntactic check and ensure
348
    validity of opcode parameters, without any cluster-related
349
    checks. While the same can be accomplished in ExpandNames and/or
350
    CheckPrereq, doing these separate is better because:
351

352
      - ExpandNames is left as as purely a lock-related function
353
      - CheckPrereq is run after we have acquired locks (and possible
354
        waited for them)
355

356
    The function is allowed to change the self.op attribute so that
357
    later methods can no longer worry about missing parameters.
358

359
    """
360
    pass
361

    
362
  def ExpandNames(self):
363
    """Expand names for this LU.
364

365
    This method is called before starting to execute the opcode, and it should
366
    update all the parameters of the opcode to their canonical form (e.g. a
367
    short node name must be fully expanded after this method has successfully
368
    completed). This way locking, hooks, logging, ecc. can work correctly.
369

370
    LUs which implement this method must also populate the self.needed_locks
371
    member, as a dict with lock levels as keys, and a list of needed lock names
372
    as values. Rules:
373

374
      - use an empty dict if you don't need any lock
375
      - if you don't need any lock at a particular level omit that level
376
      - don't put anything for the BGL level
377
      - if you want all locks at a level use locking.ALL_SET as a value
378

379
    If you need to share locks (rather than acquire them exclusively) at one
380
    level you can modify self.share_locks, setting a true value (usually 1) for
381
    that level. By default locks are not shared.
382

383
    This function can also define a list of tasklets, which then will be
384
    executed in order instead of the usual LU-level CheckPrereq and Exec
385
    functions, if those are not defined by the LU.
386

387
    Examples::
388

389
      # Acquire all nodes and one instance
390
      self.needed_locks = {
391
        locking.LEVEL_NODE: locking.ALL_SET,
392
        locking.LEVEL_INSTANCE: ['instance1.example.com'],
393
      }
394
      # Acquire just two nodes
395
      self.needed_locks = {
396
        locking.LEVEL_NODE: ['node1.example.com', 'node2.example.com'],
397
      }
398
      # Acquire no locks
399
      self.needed_locks = {} # No, you can't leave it to the default value None
400

401
    """
402
    # The implementation of this method is mandatory only if the new LU is
403
    # concurrent, so that old LUs don't need to be changed all at the same
404
    # time.
405
    if self.REQ_BGL:
406
      self.needed_locks = {} # Exclusive LUs don't need locks.
407
    else:
408
      raise NotImplementedError
409

    
410
  def DeclareLocks(self, level):
411
    """Declare LU locking needs for a level
412

413
    While most LUs can just declare their locking needs at ExpandNames time,
414
    sometimes there's the need to calculate some locks after having acquired
415
    the ones before. This function is called just before acquiring locks at a
416
    particular level, but after acquiring the ones at lower levels, and permits
417
    such calculations. It can be used to modify self.needed_locks, and by
418
    default it does nothing.
419

420
    This function is only called if you have something already set in
421
    self.needed_locks for the level.
422

423
    @param level: Locking level which is going to be locked
424
    @type level: member of ganeti.locking.LEVELS
425

426
    """
427

    
428
  def CheckPrereq(self):
429
    """Check prerequisites for this LU.
430

431
    This method should check that the prerequisites for the execution
432
    of this LU are fulfilled. It can do internode communication, but
433
    it should be idempotent - no cluster or system changes are
434
    allowed.
435

436
    The method should raise errors.OpPrereqError in case something is
437
    not fulfilled. Its return value is ignored.
438

439
    This method should also update all the parameters of the opcode to
440
    their canonical form if it hasn't been done by ExpandNames before.
441

442
    """
443
    if self.tasklets is not None:
444
      for (idx, tl) in enumerate(self.tasklets):
445
        logging.debug("Checking prerequisites for tasklet %s/%s",
446
                      idx + 1, len(self.tasklets))
447
        tl.CheckPrereq()
448
    else:
449
      pass
450

    
451
  def Exec(self, feedback_fn):
452
    """Execute the LU.
453

454
    This method should implement the actual work. It should raise
455
    errors.OpExecError for failures that are somewhat dealt with in
456
    code, or expected.
457

458
    """
459
    if self.tasklets is not None:
460
      for (idx, tl) in enumerate(self.tasklets):
461
        logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets))
462
        tl.Exec(feedback_fn)
463
    else:
464
      raise NotImplementedError
465

    
466
  def BuildHooksEnv(self):
467
    """Build hooks environment for this LU.
468

469
    This method should return a three-node tuple consisting of: a dict
470
    containing the environment that will be used for running the
471
    specific hook for this LU, a list of node names on which the hook
472
    should run before the execution, and a list of node names on which
473
    the hook should run after the execution.
474

475
    The keys of the dict must not have 'GANETI_' prefixed as this will
476
    be handled in the hooks runner. Also note additional keys will be
477
    added by the hooks runner. If the LU doesn't define any
478
    environment, an empty dict (and not None) should be returned.
479

480
    No nodes should be returned as an empty list (and not None).
481

482
    Note that if the HPATH for a LU class is None, this function will
483
    not be called.
484

485
    """
486
    raise NotImplementedError
487

    
488
  def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
489
    """Notify the LU about the results of its hooks.
490

491
    This method is called every time a hooks phase is executed, and notifies
492
    the Logical Unit about the hooks' result. The LU can then use it to alter
493
    its result based on the hooks.  By default the method does nothing and the
494
    previous result is passed back unchanged but any LU can define it if it
495
    wants to use the local cluster hook-scripts somehow.
496

497
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
498
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
499
    @param hook_results: the results of the multi-node hooks rpc call
500
    @param feedback_fn: function used send feedback back to the caller
501
    @param lu_result: the previous Exec result this LU had, or None
502
        in the PRE phase
503
    @return: the new Exec result, based on the previous result
504
        and hook results
505

506
    """
507
    # API must be kept, thus we ignore the unused argument and could
508
    # be a function warnings
509
    # pylint: disable-msg=W0613,R0201
510
    return lu_result
511

    
512
  def _ExpandAndLockInstance(self):
513
    """Helper function to expand and lock an instance.
514

515
    Many LUs that work on an instance take its name in self.op.instance_name
516
    and need to expand it and then declare the expanded name for locking. This
517
    function does it, and then updates self.op.instance_name to the expanded
518
    name. It also initializes needed_locks as a dict, if this hasn't been done
519
    before.
520

521
    """
522
    if self.needed_locks is None:
523
      self.needed_locks = {}
524
    else:
525
      assert locking.LEVEL_INSTANCE not in self.needed_locks, \
526
        "_ExpandAndLockInstance called with instance-level locks set"
527
    self.op.instance_name = _ExpandInstanceName(self.cfg,
528
                                                self.op.instance_name)
529
    self.needed_locks[locking.LEVEL_INSTANCE] = self.op.instance_name
530

    
531
  def _LockInstancesNodes(self, primary_only=False):
532
    """Helper function to declare instances' nodes for locking.
533

534
    This function should be called after locking one or more instances to lock
535
    their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
536
    with all primary or secondary nodes for instances already locked and
537
    present in self.needed_locks[locking.LEVEL_INSTANCE].
538

539
    It should be called from DeclareLocks, and for safety only works if
540
    self.recalculate_locks[locking.LEVEL_NODE] is set.
541

542
    In the future it may grow parameters to just lock some instance's nodes, or
543
    to just lock primaries or secondary nodes, if needed.
544

545
    If should be called in DeclareLocks in a way similar to::
546

547
      if level == locking.LEVEL_NODE:
548
        self._LockInstancesNodes()
549

550
    @type primary_only: boolean
551
    @param primary_only: only lock primary nodes of locked instances
552

553
    """
554
    assert locking.LEVEL_NODE in self.recalculate_locks, \
555
      "_LockInstancesNodes helper function called with no nodes to recalculate"
556

    
557
    # TODO: check if we're really been called with the instance locks held
558

    
559
    # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
560
    # future we might want to have different behaviors depending on the value
561
    # of self.recalculate_locks[locking.LEVEL_NODE]
562
    wanted_nodes = []
563
    for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
564
      instance = self.context.cfg.GetInstanceInfo(instance_name)
565
      wanted_nodes.append(instance.primary_node)
566
      if not primary_only:
567
        wanted_nodes.extend(instance.secondary_nodes)
568

    
569
    if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE:
570
      self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
571
    elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND:
572
      self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
573

    
574
    del self.recalculate_locks[locking.LEVEL_NODE]
575

    
576

    
577
class NoHooksLU(LogicalUnit): # pylint: disable-msg=W0223
578
  """Simple LU which runs no hooks.
579

580
  This LU is intended as a parent for other LogicalUnits which will
581
  run no hooks, in order to reduce duplicate code.
582

583
  """
584
  HPATH = None
585
  HTYPE = None
586

    
587
  def BuildHooksEnv(self):
588
    """Empty BuildHooksEnv for NoHooksLu.
589

590
    This just raises an error.
591

592
    """
593
    assert False, "BuildHooksEnv called for NoHooksLUs"
594

    
595

    
596
class Tasklet:
597
  """Tasklet base class.
598

599
  Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
600
  they can mix legacy code with tasklets. Locking needs to be done in the LU,
601
  tasklets know nothing about locks.
602

603
  Subclasses must follow these rules:
604
    - Implement CheckPrereq
605
    - Implement Exec
606

607
  """
608
  def __init__(self, lu):
609
    self.lu = lu
610

    
611
    # Shortcuts
612
    self.cfg = lu.cfg
613
    self.rpc = lu.rpc
614

    
615
  def CheckPrereq(self):
616
    """Check prerequisites for this tasklets.
617

618
    This method should check whether the prerequisites for the execution of
619
    this tasklet are fulfilled. It can do internode communication, but it
620
    should be idempotent - no cluster or system changes are allowed.
621

622
    The method should raise errors.OpPrereqError in case something is not
623
    fulfilled. Its return value is ignored.
624

625
    This method should also update all parameters to their canonical form if it
626
    hasn't been done before.
627

628
    """
629
    pass
630

    
631
  def Exec(self, feedback_fn):
632
    """Execute the tasklet.
633

634
    This method should implement the actual work. It should raise
635
    errors.OpExecError for failures that are somewhat dealt with in code, or
636
    expected.
637

638
    """
639
    raise NotImplementedError
640

    
641

    
642
def _GetWantedNodes(lu, nodes):
643
  """Returns list of checked and expanded node names.
644

645
  @type lu: L{LogicalUnit}
646
  @param lu: the logical unit on whose behalf we execute
647
  @type nodes: list
648
  @param nodes: list of node names or None for all nodes
649
  @rtype: list
650
  @return: the list of nodes, sorted
651
  @raise errors.ProgrammerError: if the nodes parameter is wrong type
652

653
  """
654
  if not nodes:
655
    raise errors.ProgrammerError("_GetWantedNodes should only be called with a"
656
      " non-empty list of nodes whose name is to be expanded.")
657

    
658
  wanted = [_ExpandNodeName(lu.cfg, name) for name in nodes]
659
  return utils.NiceSort(wanted)
660

    
661

    
662
def _GetWantedInstances(lu, instances):
663
  """Returns list of checked and expanded instance names.
664

665
  @type lu: L{LogicalUnit}
666
  @param lu: the logical unit on whose behalf we execute
667
  @type instances: list
668
  @param instances: list of instance names or None for all instances
669
  @rtype: list
670
  @return: the list of instances, sorted
671
  @raise errors.OpPrereqError: if the instances parameter is wrong type
672
  @raise errors.OpPrereqError: if any of the passed instances is not found
673

674
  """
675
  if instances:
676
    wanted = [_ExpandInstanceName(lu.cfg, name) for name in instances]
677
  else:
678
    wanted = utils.NiceSort(lu.cfg.GetInstanceList())
679
  return wanted
680

    
681

    
682
def _GetUpdatedParams(old_params, update_dict,
683
                      use_default=True, use_none=False):
684
  """Return the new version of a parameter dictionary.
685

686
  @type old_params: dict
687
  @param old_params: old parameters
688
  @type update_dict: dict
689
  @param update_dict: dict containing new parameter values, or
690
      constants.VALUE_DEFAULT to reset the parameter to its default
691
      value
692
  @param use_default: boolean
693
  @type use_default: whether to recognise L{constants.VALUE_DEFAULT}
694
      values as 'to be deleted' values
695
  @param use_none: boolean
696
  @type use_none: whether to recognise C{None} values as 'to be
697
      deleted' values
698
  @rtype: dict
699
  @return: the new parameter dictionary
700

701
  """
702
  params_copy = copy.deepcopy(old_params)
703
  for key, val in update_dict.iteritems():
704
    if ((use_default and val == constants.VALUE_DEFAULT) or
705
        (use_none and val is None)):
706
      try:
707
        del params_copy[key]
708
      except KeyError:
709
        pass
710
    else:
711
      params_copy[key] = val
712
  return params_copy
713

    
714

    
715
def _CheckOutputFields(static, dynamic, selected):
716
  """Checks whether all selected fields are valid.
717

718
  @type static: L{utils.FieldSet}
719
  @param static: static fields set
720
  @type dynamic: L{utils.FieldSet}
721
  @param dynamic: dynamic fields set
722

723
  """
724
  f = utils.FieldSet()
725
  f.Extend(static)
726
  f.Extend(dynamic)
727

    
728
  delta = f.NonMatching(selected)
729
  if delta:
730
    raise errors.OpPrereqError("Unknown output fields selected: %s"
731
                               % ",".join(delta), errors.ECODE_INVAL)
732

    
733

    
734
def _CheckGlobalHvParams(params):
735
  """Validates that given hypervisor params are not global ones.
736

737
  This will ensure that instances don't get customised versions of
738
  global params.
739

740
  """
741
  used_globals = constants.HVC_GLOBALS.intersection(params)
742
  if used_globals:
743
    msg = ("The following hypervisor parameters are global and cannot"
744
           " be customized at instance level, please modify them at"
745
           " cluster level: %s" % utils.CommaJoin(used_globals))
746
    raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
747

    
748

    
749
def _CheckNodeOnline(lu, node):
750
  """Ensure that a given node is online.
751

752
  @param lu: the LU on behalf of which we make the check
753
  @param node: the node to check
754
  @raise errors.OpPrereqError: if the node is offline
755

756
  """
757
  if lu.cfg.GetNodeInfo(node).offline:
758
    raise errors.OpPrereqError("Can't use offline node %s" % node,
759
                               errors.ECODE_INVAL)
760

    
761

    
762
def _CheckNodeNotDrained(lu, node):
763
  """Ensure that a given node is not drained.
764

765
  @param lu: the LU on behalf of which we make the check
766
  @param node: the node to check
767
  @raise errors.OpPrereqError: if the node is drained
768

769
  """
770
  if lu.cfg.GetNodeInfo(node).drained:
771
    raise errors.OpPrereqError("Can't use drained node %s" % node,
772
                               errors.ECODE_INVAL)
773

    
774

    
775
def _CheckNodeHasOS(lu, node, os_name, force_variant):
776
  """Ensure that a node supports a given OS.
777

778
  @param lu: the LU on behalf of which we make the check
779
  @param node: the node to check
780
  @param os_name: the OS to query about
781
  @param force_variant: whether to ignore variant errors
782
  @raise errors.OpPrereqError: if the node is not supporting the OS
783

784
  """
785
  result = lu.rpc.call_os_get(node, os_name)
786
  result.Raise("OS '%s' not in supported OS list for node %s" %
787
               (os_name, node),
788
               prereq=True, ecode=errors.ECODE_INVAL)
789
  if not force_variant:
790
    _CheckOSVariant(result.payload, os_name)
791

    
792

    
793
def _RequireFileStorage():
794
  """Checks that file storage is enabled.
795

796
  @raise errors.OpPrereqError: when file storage is disabled
797

798
  """
799
  if not constants.ENABLE_FILE_STORAGE:
800
    raise errors.OpPrereqError("File storage disabled at configure time",
801
                               errors.ECODE_INVAL)
802

    
803

    
804
def _CheckDiskTemplate(template):
805
  """Ensure a given disk template is valid.
806

807
  """
808
  if template not in constants.DISK_TEMPLATES:
809
    msg = ("Invalid disk template name '%s', valid templates are: %s" %
810
           (template, utils.CommaJoin(constants.DISK_TEMPLATES)))
811
    raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
812
  if template == constants.DT_FILE:
813
    _RequireFileStorage()
814
  return True
815

    
816

    
817
def _CheckStorageType(storage_type):
818
  """Ensure a given storage type is valid.
819

820
  """
821
  if storage_type not in constants.VALID_STORAGE_TYPES:
822
    raise errors.OpPrereqError("Unknown storage type: %s" % storage_type,
823
                               errors.ECODE_INVAL)
824
  if storage_type == constants.ST_FILE:
825
    _RequireFileStorage()
826
  return True
827

    
828

    
829
def _GetClusterDomainSecret():
830
  """Reads the cluster domain secret.
831

832
  """
833
  return utils.ReadOneLineFile(constants.CLUSTER_DOMAIN_SECRET_FILE,
834
                               strict=True)
835

    
836

    
837
def _CheckInstanceDown(lu, instance, reason):
838
  """Ensure that an instance is not running."""
839
  if instance.admin_up:
840
    raise errors.OpPrereqError("Instance %s is marked to be up, %s" %
841
                               (instance.name, reason), errors.ECODE_STATE)
842

    
843
  pnode = instance.primary_node
844
  ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
845
  ins_l.Raise("Can't contact node %s for instance information" % pnode,
846
              prereq=True, ecode=errors.ECODE_ENVIRON)
847

    
848
  if instance.name in ins_l.payload:
849
    raise errors.OpPrereqError("Instance %s is running, %s" %
850
                               (instance.name, reason), errors.ECODE_STATE)
851

    
852

    
853
def _ExpandItemName(fn, name, kind):
854
  """Expand an item name.
855

856
  @param fn: the function to use for expansion
857
  @param name: requested item name
858
  @param kind: text description ('Node' or 'Instance')
859
  @return: the resolved (full) name
860
  @raise errors.OpPrereqError: if the item is not found
861

862
  """
863
  full_name = fn(name)
864
  if full_name is None:
865
    raise errors.OpPrereqError("%s '%s' not known" % (kind, name),
866
                               errors.ECODE_NOENT)
867
  return full_name
868

    
869

    
870
def _ExpandNodeName(cfg, name):
871
  """Wrapper over L{_ExpandItemName} for nodes."""
872
  return _ExpandItemName(cfg.ExpandNodeName, name, "Node")
873

    
874

    
875
def _ExpandInstanceName(cfg, name):
876
  """Wrapper over L{_ExpandItemName} for instance."""
877
  return _ExpandItemName(cfg.ExpandInstanceName, name, "Instance")
878

    
879

    
880
def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
881
                          memory, vcpus, nics, disk_template, disks,
882
                          bep, hvp, hypervisor_name):
883
  """Builds instance related env variables for hooks
884

885
  This builds the hook environment from individual variables.
886

887
  @type name: string
888
  @param name: the name of the instance
889
  @type primary_node: string
890
  @param primary_node: the name of the instance's primary node
891
  @type secondary_nodes: list
892
  @param secondary_nodes: list of secondary nodes as strings
893
  @type os_type: string
894
  @param os_type: the name of the instance's OS
895
  @type status: boolean
896
  @param status: the should_run status of the instance
897
  @type memory: string
898
  @param memory: the memory size of the instance
899
  @type vcpus: string
900
  @param vcpus: the count of VCPUs the instance has
901
  @type nics: list
902
  @param nics: list of tuples (ip, mac, mode, link) representing
903
      the NICs the instance has
904
  @type disk_template: string
905
  @param disk_template: the disk template of the instance
906
  @type disks: list
907
  @param disks: the list of (size, mode) pairs
908
  @type bep: dict
909
  @param bep: the backend parameters for the instance
910
  @type hvp: dict
911
  @param hvp: the hypervisor parameters for the instance
912
  @type hypervisor_name: string
913
  @param hypervisor_name: the hypervisor for the instance
914
  @rtype: dict
915
  @return: the hook environment for this instance
916

917
  """
918
  if status:
919
    str_status = "up"
920
  else:
921
    str_status = "down"
922
  env = {
923
    "OP_TARGET": name,
924
    "INSTANCE_NAME": name,
925
    "INSTANCE_PRIMARY": primary_node,
926
    "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
927
    "INSTANCE_OS_TYPE": os_type,
928
    "INSTANCE_STATUS": str_status,
929
    "INSTANCE_MEMORY": memory,
930
    "INSTANCE_VCPUS": vcpus,
931
    "INSTANCE_DISK_TEMPLATE": disk_template,
932
    "INSTANCE_HYPERVISOR": hypervisor_name,
933
  }
934

    
935
  if nics:
936
    nic_count = len(nics)
937
    for idx, (ip, mac, mode, link) in enumerate(nics):
938
      if ip is None:
939
        ip = ""
940
      env["INSTANCE_NIC%d_IP" % idx] = ip
941
      env["INSTANCE_NIC%d_MAC" % idx] = mac
942
      env["INSTANCE_NIC%d_MODE" % idx] = mode
943
      env["INSTANCE_NIC%d_LINK" % idx] = link
944
      if mode == constants.NIC_MODE_BRIDGED:
945
        env["INSTANCE_NIC%d_BRIDGE" % idx] = link
946
  else:
947
    nic_count = 0
948

    
949
  env["INSTANCE_NIC_COUNT"] = nic_count
950

    
951
  if disks:
952
    disk_count = len(disks)
953
    for idx, (size, mode) in enumerate(disks):
954
      env["INSTANCE_DISK%d_SIZE" % idx] = size
955
      env["INSTANCE_DISK%d_MODE" % idx] = mode
956
  else:
957
    disk_count = 0
958

    
959
  env["INSTANCE_DISK_COUNT"] = disk_count
960

    
961
  for source, kind in [(bep, "BE"), (hvp, "HV")]:
962
    for key, value in source.items():
963
      env["INSTANCE_%s_%s" % (kind, key)] = value
964

    
965
  return env
966

    
967

    
968
def _NICListToTuple(lu, nics):
969
  """Build a list of nic information tuples.
970

971
  This list is suitable to be passed to _BuildInstanceHookEnv or as a return
972
  value in LUQueryInstanceData.
973

974
  @type lu:  L{LogicalUnit}
975
  @param lu: the logical unit on whose behalf we execute
976
  @type nics: list of L{objects.NIC}
977
  @param nics: list of nics to convert to hooks tuples
978

979
  """
980
  hooks_nics = []
981
  cluster = lu.cfg.GetClusterInfo()
982
  for nic in nics:
983
    ip = nic.ip
984
    mac = nic.mac
985
    filled_params = cluster.SimpleFillNIC(nic.nicparams)
986
    mode = filled_params[constants.NIC_MODE]
987
    link = filled_params[constants.NIC_LINK]
988
    hooks_nics.append((ip, mac, mode, link))
989
  return hooks_nics
990

    
991

    
992
def _BuildInstanceHookEnvByObject(lu, instance, override=None):
993
  """Builds instance related env variables for hooks from an object.
994

995
  @type lu: L{LogicalUnit}
996
  @param lu: the logical unit on whose behalf we execute
997
  @type instance: L{objects.Instance}
998
  @param instance: the instance for which we should build the
999
      environment
1000
  @type override: dict
1001
  @param override: dictionary with key/values that will override
1002
      our values
1003
  @rtype: dict
1004
  @return: the hook environment dictionary
1005

1006
  """
1007
  cluster = lu.cfg.GetClusterInfo()
1008
  bep = cluster.FillBE(instance)
1009
  hvp = cluster.FillHV(instance)
1010
  args = {
1011
    'name': instance.name,
1012
    'primary_node': instance.primary_node,
1013
    'secondary_nodes': instance.secondary_nodes,
1014
    'os_type': instance.os,
1015
    'status': instance.admin_up,
1016
    'memory': bep[constants.BE_MEMORY],
1017
    'vcpus': bep[constants.BE_VCPUS],
1018
    'nics': _NICListToTuple(lu, instance.nics),
1019
    'disk_template': instance.disk_template,
1020
    'disks': [(disk.size, disk.mode) for disk in instance.disks],
1021
    'bep': bep,
1022
    'hvp': hvp,
1023
    'hypervisor_name': instance.hypervisor,
1024
  }
1025
  if override:
1026
    args.update(override)
1027
  return _BuildInstanceHookEnv(**args) # pylint: disable-msg=W0142
1028

    
1029

    
1030
def _AdjustCandidatePool(lu, exceptions):
1031
  """Adjust the candidate pool after node operations.
1032

1033
  """
1034
  mod_list = lu.cfg.MaintainCandidatePool(exceptions)
1035
  if mod_list:
1036
    lu.LogInfo("Promoted nodes to master candidate role: %s",
1037
               utils.CommaJoin(node.name for node in mod_list))
1038
    for name in mod_list:
1039
      lu.context.ReaddNode(name)
1040
  mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
1041
  if mc_now > mc_max:
1042
    lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
1043
               (mc_now, mc_max))
1044

    
1045

    
1046
def _DecideSelfPromotion(lu, exceptions=None):
1047
  """Decide whether I should promote myself as a master candidate.
1048

1049
  """
1050
  cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
1051
  mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
1052
  # the new node will increase mc_max with one, so:
1053
  mc_should = min(mc_should + 1, cp_size)
1054
  return mc_now < mc_should
1055

    
1056

    
1057
def _CheckNicsBridgesExist(lu, target_nics, target_node):
1058
  """Check that the brigdes needed by a list of nics exist.
1059

1060
  """
1061
  cluster = lu.cfg.GetClusterInfo()
1062
  paramslist = [cluster.SimpleFillNIC(nic.nicparams) for nic in target_nics]
1063
  brlist = [params[constants.NIC_LINK] for params in paramslist
1064
            if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
1065
  if brlist:
1066
    result = lu.rpc.call_bridges_exist(target_node, brlist)
1067
    result.Raise("Error checking bridges on destination node '%s'" %
1068
                 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
1069

    
1070

    
1071
def _CheckInstanceBridgesExist(lu, instance, node=None):
1072
  """Check that the brigdes needed by an instance exist.
1073

1074
  """
1075
  if node is None:
1076
    node = instance.primary_node
1077
  _CheckNicsBridgesExist(lu, instance.nics, node)
1078

    
1079

    
1080
def _CheckOSVariant(os_obj, name):
1081
  """Check whether an OS name conforms to the os variants specification.
1082

1083
  @type os_obj: L{objects.OS}
1084
  @param os_obj: OS object to check
1085
  @type name: string
1086
  @param name: OS name passed by the user, to check for validity
1087

1088
  """
1089
  if not os_obj.supported_variants:
1090
    return
1091
  try:
1092
    variant = name.split("+", 1)[1]
1093
  except IndexError:
1094
    raise errors.OpPrereqError("OS name must include a variant",
1095
                               errors.ECODE_INVAL)
1096

    
1097
  if variant not in os_obj.supported_variants:
1098
    raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
1099

    
1100

    
1101
def _GetNodeInstancesInner(cfg, fn):
1102
  return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
1103

    
1104

    
1105
def _GetNodeInstances(cfg, node_name):
1106
  """Returns a list of all primary and secondary instances on a node.
1107

1108
  """
1109

    
1110
  return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
1111

    
1112

    
1113
def _GetNodePrimaryInstances(cfg, node_name):
1114
  """Returns primary instances on a node.
1115

1116
  """
1117
  return _GetNodeInstancesInner(cfg,
1118
                                lambda inst: node_name == inst.primary_node)
1119

    
1120

    
1121
def _GetNodeSecondaryInstances(cfg, node_name):
1122
  """Returns secondary instances on a node.
1123

1124
  """
1125
  return _GetNodeInstancesInner(cfg,
1126
                                lambda inst: node_name in inst.secondary_nodes)
1127

    
1128

    
1129
def _GetStorageTypeArgs(cfg, storage_type):
1130
  """Returns the arguments for a storage type.
1131

1132
  """
1133
  # Special case for file storage
1134
  if storage_type == constants.ST_FILE:
1135
    # storage.FileStorage wants a list of storage directories
1136
    return [[cfg.GetFileStorageDir()]]
1137

    
1138
  return []
1139

    
1140

    
1141
def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq):
1142
  faulty = []
1143

    
1144
  for dev in instance.disks:
1145
    cfg.SetDiskID(dev, node_name)
1146

    
1147
  result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks)
1148
  result.Raise("Failed to get disk status from node %s" % node_name,
1149
               prereq=prereq, ecode=errors.ECODE_ENVIRON)
1150

    
1151
  for idx, bdev_status in enumerate(result.payload):
1152
    if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
1153
      faulty.append(idx)
1154

    
1155
  return faulty
1156

    
1157

    
1158
def _CheckIAllocatorOrNode(lu, iallocator_slot, node_slot):
1159
  """Check the sanity of iallocator and node arguments and use the
1160
  cluster-wide iallocator if appropriate.
1161

1162
  Check that at most one of (iallocator, node) is specified. If none is
1163
  specified, then the LU's opcode's iallocator slot is filled with the
1164
  cluster-wide default iallocator.
1165

1166
  @type iallocator_slot: string
1167
  @param iallocator_slot: the name of the opcode iallocator slot
1168
  @type node_slot: string
1169
  @param node_slot: the name of the opcode target node slot
1170

1171
  """
1172
  node = getattr(lu.op, node_slot, None)
1173
  iallocator = getattr(lu.op, iallocator_slot, None)
1174

    
1175
  if node is not None and iallocator is not None:
1176
    raise errors.OpPrereqError("Do not specify both, iallocator and node.",
1177
                               errors.ECODE_INVAL)
1178
  elif node is None and iallocator is None:
1179
    default_iallocator = lu.cfg.GetDefaultIAllocator()
1180
    if default_iallocator:
1181
      setattr(lu.op, iallocator_slot, default_iallocator)
1182
    else:
1183
      raise errors.OpPrereqError("No iallocator or node given and no"
1184
                                 " cluster-wide default iallocator found."
1185
                                 " Please specify either an iallocator or a"
1186
                                 " node, or set a cluster-wide default"
1187
                                 " iallocator.")
1188

    
1189

    
1190
class LUPostInitCluster(LogicalUnit):
1191
  """Logical unit for running hooks after cluster initialization.
1192

1193
  """
1194
  HPATH = "cluster-init"
1195
  HTYPE = constants.HTYPE_CLUSTER
1196

    
1197
  def BuildHooksEnv(self):
1198
    """Build hooks env.
1199

1200
    """
1201
    env = {"OP_TARGET": self.cfg.GetClusterName()}
1202
    mn = self.cfg.GetMasterNode()
1203
    return env, [], [mn]
1204

    
1205
  def Exec(self, feedback_fn):
1206
    """Nothing to do.
1207

1208
    """
1209
    return True
1210

    
1211

    
1212
class LUDestroyCluster(LogicalUnit):
1213
  """Logical unit for destroying the cluster.
1214

1215
  """
1216
  HPATH = "cluster-destroy"
1217
  HTYPE = constants.HTYPE_CLUSTER
1218

    
1219
  def BuildHooksEnv(self):
1220
    """Build hooks env.
1221

1222
    """
1223
    env = {"OP_TARGET": self.cfg.GetClusterName()}
1224
    return env, [], []
1225

    
1226
  def CheckPrereq(self):
1227
    """Check prerequisites.
1228

1229
    This checks whether the cluster is empty.
1230

1231
    Any errors are signaled by raising errors.OpPrereqError.
1232

1233
    """
1234
    master = self.cfg.GetMasterNode()
1235

    
1236
    nodelist = self.cfg.GetNodeList()
1237
    if len(nodelist) != 1 or nodelist[0] != master:
1238
      raise errors.OpPrereqError("There are still %d node(s) in"
1239
                                 " this cluster." % (len(nodelist) - 1),
1240
                                 errors.ECODE_INVAL)
1241
    instancelist = self.cfg.GetInstanceList()
1242
    if instancelist:
1243
      raise errors.OpPrereqError("There are still %d instance(s) in"
1244
                                 " this cluster." % len(instancelist),
1245
                                 errors.ECODE_INVAL)
1246

    
1247
  def Exec(self, feedback_fn):
1248
    """Destroys the cluster.
1249

1250
    """
1251
    master = self.cfg.GetMasterNode()
1252

    
1253
    # Run post hooks on master node before it's removed
1254
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
1255
    try:
1256
      hm.RunPhase(constants.HOOKS_PHASE_POST, [master])
1257
    except:
1258
      # pylint: disable-msg=W0702
1259
      self.LogWarning("Errors occurred running hooks on %s" % master)
1260

    
1261
    result = self.rpc.call_node_stop_master(master, False)
1262
    result.Raise("Could not disable the master role")
1263

    
1264
    return master
1265

    
1266

    
1267
def _VerifyCertificate(filename):
1268
  """Verifies a certificate for LUVerifyCluster.
1269

1270
  @type filename: string
1271
  @param filename: Path to PEM file
1272

1273
  """
1274
  try:
1275
    cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
1276
                                           utils.ReadFile(filename))
1277
  except Exception, err: # pylint: disable-msg=W0703
1278
    return (LUVerifyCluster.ETYPE_ERROR,
1279
            "Failed to load X509 certificate %s: %s" % (filename, err))
1280

    
1281
  (errcode, msg) = \
1282
    utils.VerifyX509Certificate(cert, constants.SSL_CERT_EXPIRATION_WARN,
1283
                                constants.SSL_CERT_EXPIRATION_ERROR)
1284

    
1285
  if msg:
1286
    fnamemsg = "While verifying %s: %s" % (filename, msg)
1287
  else:
1288
    fnamemsg = None
1289

    
1290
  if errcode is None:
1291
    return (None, fnamemsg)
1292
  elif errcode == utils.CERT_WARNING:
1293
    return (LUVerifyCluster.ETYPE_WARNING, fnamemsg)
1294
  elif errcode == utils.CERT_ERROR:
1295
    return (LUVerifyCluster.ETYPE_ERROR, fnamemsg)
1296

    
1297
  raise errors.ProgrammerError("Unhandled certificate error code %r" % errcode)
1298

    
1299

    
1300
class LUVerifyCluster(LogicalUnit):
1301
  """Verifies the cluster status.
1302

1303
  """
1304
  HPATH = "cluster-verify"
1305
  HTYPE = constants.HTYPE_CLUSTER
1306
  _OP_PARAMS = [
1307
    ("skip_checks", _EmptyList,
1308
     _TListOf(_TElemOf(constants.VERIFY_OPTIONAL_CHECKS))),
1309
    ("verbose", False, _TBool),
1310
    ("error_codes", False, _TBool),
1311
    ("debug_simulate_errors", False, _TBool),
1312
    ]
1313
  REQ_BGL = False
1314

    
1315
  TCLUSTER = "cluster"
1316
  TNODE = "node"
1317
  TINSTANCE = "instance"
1318

    
1319
  ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
1320
  ECLUSTERCERT = (TCLUSTER, "ECLUSTERCERT")
1321
  EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
1322
  EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
1323
  EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
1324
  EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
1325
  EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
1326
  EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE")
1327
  ENODEDRBD = (TNODE, "ENODEDRBD")
1328
  ENODEDRBDHELPER = (TNODE, "ENODEDRBDHELPER")
1329
  ENODEFILECHECK = (TNODE, "ENODEFILECHECK")
1330
  ENODEHOOKS = (TNODE, "ENODEHOOKS")
1331
  ENODEHV = (TNODE, "ENODEHV")
1332
  ENODELVM = (TNODE, "ENODELVM")
1333
  ENODEN1 = (TNODE, "ENODEN1")
1334
  ENODENET = (TNODE, "ENODENET")
1335
  ENODEOS = (TNODE, "ENODEOS")
1336
  ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE")
1337
  ENODEORPHANLV = (TNODE, "ENODEORPHANLV")
1338
  ENODERPC = (TNODE, "ENODERPC")
1339
  ENODESSH = (TNODE, "ENODESSH")
1340
  ENODEVERSION = (TNODE, "ENODEVERSION")
1341
  ENODESETUP = (TNODE, "ENODESETUP")
1342
  ENODETIME = (TNODE, "ENODETIME")
1343

    
1344
  ETYPE_FIELD = "code"
1345
  ETYPE_ERROR = "ERROR"
1346
  ETYPE_WARNING = "WARNING"
1347

    
1348
  class NodeImage(object):
1349
    """A class representing the logical and physical status of a node.
1350

1351
    @type name: string
1352
    @ivar name: the node name to which this object refers
1353
    @ivar volumes: a structure as returned from
1354
        L{ganeti.backend.GetVolumeList} (runtime)
1355
    @ivar instances: a list of running instances (runtime)
1356
    @ivar pinst: list of configured primary instances (config)
1357
    @ivar sinst: list of configured secondary instances (config)
1358
    @ivar sbp: diction of {secondary-node: list of instances} of all peers
1359
        of this node (config)
1360
    @ivar mfree: free memory, as reported by hypervisor (runtime)
1361
    @ivar dfree: free disk, as reported by the node (runtime)
1362
    @ivar offline: the offline status (config)
1363
    @type rpc_fail: boolean
1364
    @ivar rpc_fail: whether the RPC verify call was successfull (overall,
1365
        not whether the individual keys were correct) (runtime)
1366
    @type lvm_fail: boolean
1367
    @ivar lvm_fail: whether the RPC call didn't return valid LVM data
1368
    @type hyp_fail: boolean
1369
    @ivar hyp_fail: whether the RPC call didn't return the instance list
1370
    @type ghost: boolean
1371
    @ivar ghost: whether this is a known node or not (config)
1372
    @type os_fail: boolean
1373
    @ivar os_fail: whether the RPC call didn't return valid OS data
1374
    @type oslist: list
1375
    @ivar oslist: list of OSes as diagnosed by DiagnoseOS
1376

1377
    """
1378
    def __init__(self, offline=False, name=None):
1379
      self.name = name
1380
      self.volumes = {}
1381
      self.instances = []
1382
      self.pinst = []
1383
      self.sinst = []
1384
      self.sbp = {}
1385
      self.mfree = 0
1386
      self.dfree = 0
1387
      self.offline = offline
1388
      self.rpc_fail = False
1389
      self.lvm_fail = False
1390
      self.hyp_fail = False
1391
      self.ghost = False
1392
      self.os_fail = False
1393
      self.oslist = {}
1394

    
1395
  def ExpandNames(self):
1396
    self.needed_locks = {
1397
      locking.LEVEL_NODE: locking.ALL_SET,
1398
      locking.LEVEL_INSTANCE: locking.ALL_SET,
1399
    }
1400
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
1401

    
1402
  def _Error(self, ecode, item, msg, *args, **kwargs):
1403
    """Format an error message.
1404

1405
    Based on the opcode's error_codes parameter, either format a
1406
    parseable error code, or a simpler error string.
1407

1408
    This must be called only from Exec and functions called from Exec.
1409

1410
    """
1411
    ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
1412
    itype, etxt = ecode
1413
    # first complete the msg
1414
    if args:
1415
      msg = msg % args
1416
    # then format the whole message
1417
    if self.op.error_codes:
1418
      msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
1419
    else:
1420
      if item:
1421
        item = " " + item
1422
      else:
1423
        item = ""
1424
      msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
1425
    # and finally report it via the feedback_fn
1426
    self._feedback_fn("  - %s" % msg)
1427

    
1428
  def _ErrorIf(self, cond, *args, **kwargs):
1429
    """Log an error message if the passed condition is True.
1430

1431
    """
1432
    cond = bool(cond) or self.op.debug_simulate_errors
1433
    if cond:
1434
      self._Error(*args, **kwargs)
1435
    # do not mark the operation as failed for WARN cases only
1436
    if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
1437
      self.bad = self.bad or cond
1438

    
1439
  def _VerifyNode(self, ninfo, nresult):
1440
    """Perform some basic validation on data returned from a node.
1441

1442
      - check the result data structure is well formed and has all the
1443
        mandatory fields
1444
      - check ganeti version
1445

1446
    @type ninfo: L{objects.Node}
1447
    @param ninfo: the node to check
1448
    @param nresult: the results from the node
1449
    @rtype: boolean
1450
    @return: whether overall this call was successful (and we can expect
1451
         reasonable values in the respose)
1452

1453
    """
1454
    node = ninfo.name
1455
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1456

    
1457
    # main result, nresult should be a non-empty dict
1458
    test = not nresult or not isinstance(nresult, dict)
1459
    _ErrorIf(test, self.ENODERPC, node,
1460
                  "unable to verify node: no data returned")
1461
    if test:
1462
      return False
1463

    
1464
    # compares ganeti version
1465
    local_version = constants.PROTOCOL_VERSION
1466
    remote_version = nresult.get("version", None)
1467
    test = not (remote_version and
1468
                isinstance(remote_version, (list, tuple)) and
1469
                len(remote_version) == 2)
1470
    _ErrorIf(test, self.ENODERPC, node,
1471
             "connection to node returned invalid data")
1472
    if test:
1473
      return False
1474

    
1475
    test = local_version != remote_version[0]
1476
    _ErrorIf(test, self.ENODEVERSION, node,
1477
             "incompatible protocol versions: master %s,"
1478
             " node %s", local_version, remote_version[0])
1479
    if test:
1480
      return False
1481

    
1482
    # node seems compatible, we can actually try to look into its results
1483

    
1484
    # full package version
1485
    self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
1486
                  self.ENODEVERSION, node,
1487
                  "software version mismatch: master %s, node %s",
1488
                  constants.RELEASE_VERSION, remote_version[1],
1489
                  code=self.ETYPE_WARNING)
1490

    
1491
    hyp_result = nresult.get(constants.NV_HYPERVISOR, None)
1492
    if isinstance(hyp_result, dict):
1493
      for hv_name, hv_result in hyp_result.iteritems():
1494
        test = hv_result is not None
1495
        _ErrorIf(test, self.ENODEHV, node,
1496
                 "hypervisor %s verify failure: '%s'", hv_name, hv_result)
1497

    
1498

    
1499
    test = nresult.get(constants.NV_NODESETUP,
1500
                           ["Missing NODESETUP results"])
1501
    _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s",
1502
             "; ".join(test))
1503

    
1504
    return True
1505

    
1506
  def _VerifyNodeTime(self, ninfo, nresult,
1507
                      nvinfo_starttime, nvinfo_endtime):
1508
    """Check the node time.
1509

1510
    @type ninfo: L{objects.Node}
1511
    @param ninfo: the node to check
1512
    @param nresult: the remote results for the node
1513
    @param nvinfo_starttime: the start time of the RPC call
1514
    @param nvinfo_endtime: the end time of the RPC call
1515

1516
    """
1517
    node = ninfo.name
1518
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1519

    
1520
    ntime = nresult.get(constants.NV_TIME, None)
1521
    try:
1522
      ntime_merged = utils.MergeTime(ntime)
1523
    except (ValueError, TypeError):
1524
      _ErrorIf(True, self.ENODETIME, node, "Node returned invalid time")
1525
      return
1526

    
1527
    if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
1528
      ntime_diff = "%.01fs" % abs(nvinfo_starttime - ntime_merged)
1529
    elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
1530
      ntime_diff = "%.01fs" % abs(ntime_merged - nvinfo_endtime)
1531
    else:
1532
      ntime_diff = None
1533

    
1534
    _ErrorIf(ntime_diff is not None, self.ENODETIME, node,
1535
             "Node time diverges by at least %s from master node time",
1536
             ntime_diff)
1537

    
1538
  def _VerifyNodeLVM(self, ninfo, nresult, vg_name):
1539
    """Check the node time.
1540

1541
    @type ninfo: L{objects.Node}
1542
    @param ninfo: the node to check
1543
    @param nresult: the remote results for the node
1544
    @param vg_name: the configured VG name
1545

1546
    """
1547
    if vg_name is None:
1548
      return
1549

    
1550
    node = ninfo.name
1551
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1552

    
1553
    # checks vg existence and size > 20G
1554
    vglist = nresult.get(constants.NV_VGLIST, None)
1555
    test = not vglist
1556
    _ErrorIf(test, self.ENODELVM, node, "unable to check volume groups")
1557
    if not test:
1558
      vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
1559
                                            constants.MIN_VG_SIZE)
1560
      _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
1561

    
1562
    # check pv names
1563
    pvlist = nresult.get(constants.NV_PVLIST, None)
1564
    test = pvlist is None
1565
    _ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node")
1566
    if not test:
1567
      # check that ':' is not present in PV names, since it's a
1568
      # special character for lvcreate (denotes the range of PEs to
1569
      # use on the PV)
1570
      for _, pvname, owner_vg in pvlist:
1571
        test = ":" in pvname
1572
        _ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV"
1573
                 " '%s' of VG '%s'", pvname, owner_vg)
1574

    
1575
  def _VerifyNodeNetwork(self, ninfo, nresult):
1576
    """Check the node time.
1577

1578
    @type ninfo: L{objects.Node}
1579
    @param ninfo: the node to check
1580
    @param nresult: the remote results for the node
1581

1582
    """
1583
    node = ninfo.name
1584
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1585

    
1586
    test = constants.NV_NODELIST not in nresult
1587
    _ErrorIf(test, self.ENODESSH, node,
1588
             "node hasn't returned node ssh connectivity data")
1589
    if not test:
1590
      if nresult[constants.NV_NODELIST]:
1591
        for a_node, a_msg in nresult[constants.NV_NODELIST].items():
1592
          _ErrorIf(True, self.ENODESSH, node,
1593
                   "ssh communication with node '%s': %s", a_node, a_msg)
1594

    
1595
    test = constants.NV_NODENETTEST not in nresult
1596
    _ErrorIf(test, self.ENODENET, node,
1597
             "node hasn't returned node tcp connectivity data")
1598
    if not test:
1599
      if nresult[constants.NV_NODENETTEST]:
1600
        nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys())
1601
        for anode in nlist:
1602
          _ErrorIf(True, self.ENODENET, node,
1603
                   "tcp communication with node '%s': %s",
1604
                   anode, nresult[constants.NV_NODENETTEST][anode])
1605

    
1606
    test = constants.NV_MASTERIP not in nresult
1607
    _ErrorIf(test, self.ENODENET, node,
1608
             "node hasn't returned node master IP reachability data")
1609
    if not test:
1610
      if not nresult[constants.NV_MASTERIP]:
1611
        if node == self.master_node:
1612
          msg = "the master node cannot reach the master IP (not configured?)"
1613
        else:
1614
          msg = "cannot reach the master IP"
1615
        _ErrorIf(True, self.ENODENET, node, msg)
1616

    
1617

    
1618
  def _VerifyInstance(self, instance, instanceconfig, node_image):
1619
    """Verify an instance.
1620

1621
    This function checks to see if the required block devices are
1622
    available on the instance's node.
1623

1624
    """
1625
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1626
    node_current = instanceconfig.primary_node
1627

    
1628
    node_vol_should = {}
1629
    instanceconfig.MapLVsByNode(node_vol_should)
1630

    
1631
    for node in node_vol_should:
1632
      n_img = node_image[node]
1633
      if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1634
        # ignore missing volumes on offline or broken nodes
1635
        continue
1636
      for volume in node_vol_should[node]:
1637
        test = volume not in n_img.volumes
1638
        _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
1639
                 "volume %s missing on node %s", volume, node)
1640

    
1641
    if instanceconfig.admin_up:
1642
      pri_img = node_image[node_current]
1643
      test = instance not in pri_img.instances and not pri_img.offline
1644
      _ErrorIf(test, self.EINSTANCEDOWN, instance,
1645
               "instance not running on its primary node %s",
1646
               node_current)
1647

    
1648
    for node, n_img in node_image.items():
1649
      if (not node == node_current):
1650
        test = instance in n_img.instances
1651
        _ErrorIf(test, self.EINSTANCEWRONGNODE, instance,
1652
                 "instance should not run on node %s", node)
1653

    
1654
  def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):
1655
    """Verify if there are any unknown volumes in the cluster.
1656

1657
    The .os, .swap and backup volumes are ignored. All other volumes are
1658
    reported as unknown.
1659

1660
    @type reserved: L{ganeti.utils.FieldSet}
1661
    @param reserved: a FieldSet of reserved volume names
1662

1663
    """
1664
    for node, n_img in node_image.items():
1665
      if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1666
        # skip non-healthy nodes
1667
        continue
1668
      for volume in n_img.volumes:
1669
        test = ((node not in node_vol_should or
1670
                volume not in node_vol_should[node]) and
1671
                not reserved.Matches(volume))
1672
        self._ErrorIf(test, self.ENODEORPHANLV, node,
1673
                      "volume %s is unknown", volume)
1674

    
1675
  def _VerifyOrphanInstances(self, instancelist, node_image):
1676
    """Verify the list of running instances.
1677

1678
    This checks what instances are running but unknown to the cluster.
1679

1680
    """
1681
    for node, n_img in node_image.items():
1682
      for o_inst in n_img.instances:
1683
        test = o_inst not in instancelist
1684
        self._ErrorIf(test, self.ENODEORPHANINSTANCE, node,
1685
                      "instance %s on node %s should not exist", o_inst, node)
1686

    
1687
  def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
1688
    """Verify N+1 Memory Resilience.
1689

1690
    Check that if one single node dies we can still start all the
1691
    instances it was primary for.
1692

1693
    """
1694
    for node, n_img in node_image.items():
1695
      # This code checks that every node which is now listed as
1696
      # secondary has enough memory to host all instances it is
1697
      # supposed to should a single other node in the cluster fail.
1698
      # FIXME: not ready for failover to an arbitrary node
1699
      # FIXME: does not support file-backed instances
1700
      # WARNING: we currently take into account down instances as well
1701
      # as up ones, considering that even if they're down someone
1702
      # might want to start them even in the event of a node failure.
1703
      for prinode, instances in n_img.sbp.items():
1704
        needed_mem = 0
1705
        for instance in instances:
1706
          bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
1707
          if bep[constants.BE_AUTO_BALANCE]:
1708
            needed_mem += bep[constants.BE_MEMORY]
1709
        test = n_img.mfree < needed_mem
1710
        self._ErrorIf(test, self.ENODEN1, node,
1711
                      "not enough memory on to accommodate"
1712
                      " failovers should peer node %s fail", prinode)
1713

    
1714
  def _VerifyNodeFiles(self, ninfo, nresult, file_list, local_cksum,
1715
                       master_files):
1716
    """Verifies and computes the node required file checksums.
1717

1718
    @type ninfo: L{objects.Node}
1719
    @param ninfo: the node to check
1720
    @param nresult: the remote results for the node
1721
    @param file_list: required list of files
1722
    @param local_cksum: dictionary of local files and their checksums
1723
    @param master_files: list of files that only masters should have
1724

1725
    """
1726
    node = ninfo.name
1727
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1728

    
1729
    remote_cksum = nresult.get(constants.NV_FILELIST, None)
1730
    test = not isinstance(remote_cksum, dict)
1731
    _ErrorIf(test, self.ENODEFILECHECK, node,
1732
             "node hasn't returned file checksum data")
1733
    if test:
1734
      return
1735

    
1736
    for file_name in file_list:
1737
      node_is_mc = ninfo.master_candidate
1738
      must_have = (file_name not in master_files) or node_is_mc
1739
      # missing
1740
      test1 = file_name not in remote_cksum
1741
      # invalid checksum
1742
      test2 = not test1 and remote_cksum[file_name] != local_cksum[file_name]
1743
      # existing and good
1744
      test3 = not test1 and remote_cksum[file_name] == local_cksum[file_name]
1745
      _ErrorIf(test1 and must_have, self.ENODEFILECHECK, node,
1746
               "file '%s' missing", file_name)
1747
      _ErrorIf(test2 and must_have, self.ENODEFILECHECK, node,
1748
               "file '%s' has wrong checksum", file_name)
1749
      # not candidate and this is not a must-have file
1750
      _ErrorIf(test2 and not must_have, self.ENODEFILECHECK, node,
1751
               "file '%s' should not exist on non master"
1752
               " candidates (and the file is outdated)", file_name)
1753
      # all good, except non-master/non-must have combination
1754
      _ErrorIf(test3 and not must_have, self.ENODEFILECHECK, node,
1755
               "file '%s' should not exist"
1756
               " on non master candidates", file_name)
1757

    
1758
  def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper,
1759
                      drbd_map):
1760
    """Verifies and the node DRBD status.
1761

1762
    @type ninfo: L{objects.Node}
1763
    @param ninfo: the node to check
1764
    @param nresult: the remote results for the node
1765
    @param instanceinfo: the dict of instances
1766
    @param drbd_helper: the configured DRBD usermode helper
1767
    @param drbd_map: the DRBD map as returned by
1768
        L{ganeti.config.ConfigWriter.ComputeDRBDMap}
1769

1770
    """
1771
    node = ninfo.name
1772
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1773

    
1774
    if drbd_helper:
1775
      helper_result = nresult.get(constants.NV_DRBDHELPER, None)
1776
      test = (helper_result == None)
1777
      _ErrorIf(test, self.ENODEDRBDHELPER, node,
1778
               "no drbd usermode helper returned")
1779
      if helper_result:
1780
        status, payload = helper_result
1781
        test = not status
1782
        _ErrorIf(test, self.ENODEDRBDHELPER, node,
1783
                 "drbd usermode helper check unsuccessful: %s", payload)
1784
        test = status and (payload != drbd_helper)
1785
        _ErrorIf(test, self.ENODEDRBDHELPER, node,
1786
                 "wrong drbd usermode helper: %s", payload)
1787

    
1788
    # compute the DRBD minors
1789
    node_drbd = {}
1790
    for minor, instance in drbd_map[node].items():
1791
      test = instance not in instanceinfo
1792
      _ErrorIf(test, self.ECLUSTERCFG, None,
1793
               "ghost instance '%s' in temporary DRBD map", instance)
1794
        # ghost instance should not be running, but otherwise we
1795
        # don't give double warnings (both ghost instance and
1796
        # unallocated minor in use)
1797
      if test:
1798
        node_drbd[minor] = (instance, False)
1799
      else:
1800
        instance = instanceinfo[instance]
1801
        node_drbd[minor] = (instance.name, instance.admin_up)
1802

    
1803
    # and now check them
1804
    used_minors = nresult.get(constants.NV_DRBDLIST, [])
1805
    test = not isinstance(used_minors, (tuple, list))
1806
    _ErrorIf(test, self.ENODEDRBD, node,
1807
             "cannot parse drbd status file: %s", str(used_minors))
1808
    if test:
1809
      # we cannot check drbd status
1810
      return
1811

    
1812
    for minor, (iname, must_exist) in node_drbd.items():
1813
      test = minor not in used_minors and must_exist
1814
      _ErrorIf(test, self.ENODEDRBD, node,
1815
               "drbd minor %d of instance %s is not active", minor, iname)
1816
    for minor in used_minors:
1817
      test = minor not in node_drbd
1818
      _ErrorIf(test, self.ENODEDRBD, node,
1819
               "unallocated drbd minor %d is in use", minor)
1820

    
1821
  def _UpdateNodeOS(self, ninfo, nresult, nimg):
1822
    """Builds the node OS structures.
1823

1824
    @type ninfo: L{objects.Node}
1825
    @param ninfo: the node to check
1826
    @param nresult: the remote results for the node
1827
    @param nimg: the node image object
1828

1829
    """
1830
    node = ninfo.name
1831
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1832

    
1833
    remote_os = nresult.get(constants.NV_OSLIST, None)
1834
    test = (not isinstance(remote_os, list) or
1835
            not compat.all(isinstance(v, list) and len(v) == 7
1836
                           for v in remote_os))
1837

    
1838
    _ErrorIf(test, self.ENODEOS, node,
1839
             "node hasn't returned valid OS data")
1840

    
1841
    nimg.os_fail = test
1842

    
1843
    if test:
1844
      return
1845

    
1846
    os_dict = {}
1847

    
1848
    for (name, os_path, status, diagnose,
1849
         variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
1850

    
1851
      if name not in os_dict:
1852
        os_dict[name] = []
1853

    
1854
      # parameters is a list of lists instead of list of tuples due to
1855
      # JSON lacking a real tuple type, fix it:
1856
      parameters = [tuple(v) for v in parameters]
1857
      os_dict[name].append((os_path, status, diagnose,
1858
                            set(variants), set(parameters), set(api_ver)))
1859

    
1860
    nimg.oslist = os_dict
1861

    
1862
  def _VerifyNodeOS(self, ninfo, nimg, base):
1863
    """Verifies the node OS list.
1864

1865
    @type ninfo: L{objects.Node}
1866
    @param ninfo: the node to check
1867
    @param nimg: the node image object
1868
    @param base: the 'template' node we match against (e.g. from the master)
1869

1870
    """
1871
    node = ninfo.name
1872
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1873

    
1874
    assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
1875

    
1876
    for os_name, os_data in nimg.oslist.items():
1877
      assert os_data, "Empty OS status for OS %s?!" % os_name
1878
      f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
1879
      _ErrorIf(not f_status, self.ENODEOS, node,
1880
               "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
1881
      _ErrorIf(len(os_data) > 1, self.ENODEOS, node,
1882
               "OS '%s' has multiple entries (first one shadows the rest): %s",
1883
               os_name, utils.CommaJoin([v[0] for v in os_data]))
1884
      # this will catched in backend too
1885
      _ErrorIf(compat.any(v >= constants.OS_API_V15 for v in f_api)
1886
               and not f_var, self.ENODEOS, node,
1887
               "OS %s with API at least %d does not declare any variant",
1888
               os_name, constants.OS_API_V15)
1889
      # comparisons with the 'base' image
1890
      test = os_name not in base.oslist
1891
      _ErrorIf(test, self.ENODEOS, node,
1892
               "Extra OS %s not present on reference node (%s)",
1893
               os_name, base.name)
1894
      if test:
1895
        continue
1896
      assert base.oslist[os_name], "Base node has empty OS status?"
1897
      _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
1898
      if not b_status:
1899
        # base OS is invalid, skipping
1900
        continue
1901
      for kind, a, b in [("API version", f_api, b_api),
1902
                         ("variants list", f_var, b_var),
1903
                         ("parameters", f_param, b_param)]:
1904
        _ErrorIf(a != b, self.ENODEOS, node,
1905
                 "OS %s %s differs from reference node %s: %s vs. %s",
1906
                 kind, os_name, base.name,
1907
                 utils.CommaJoin(a), utils.CommaJoin(b))
1908

    
1909
    # check any missing OSes
1910
    missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
1911
    _ErrorIf(missing, self.ENODEOS, node,
1912
             "OSes present on reference node %s but missing on this node: %s",
1913
             base.name, utils.CommaJoin(missing))
1914

    
1915
  def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
1916
    """Verifies and updates the node volume data.
1917

1918
    This function will update a L{NodeImage}'s internal structures
1919
    with data from the remote call.
1920

1921
    @type ninfo: L{objects.Node}
1922
    @param ninfo: the node to check
1923
    @param nresult: the remote results for the node
1924
    @param nimg: the node image object
1925
    @param vg_name: the configured VG name
1926

1927
    """
1928
    node = ninfo.name
1929
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1930

    
1931
    nimg.lvm_fail = True
1932
    lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
1933
    if vg_name is None:
1934
      pass
1935
    elif isinstance(lvdata, basestring):
1936
      _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
1937
               utils.SafeEncode(lvdata))
1938
    elif not isinstance(lvdata, dict):
1939
      _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
1940
    else:
1941
      nimg.volumes = lvdata
1942
      nimg.lvm_fail = False
1943

    
1944
  def _UpdateNodeInstances(self, ninfo, nresult, nimg):
1945
    """Verifies and updates the node instance list.
1946

1947
    If the listing was successful, then updates this node's instance
1948
    list. Otherwise, it marks the RPC call as failed for the instance
1949
    list key.
1950

1951
    @type ninfo: L{objects.Node}
1952
    @param ninfo: the node to check
1953
    @param nresult: the remote results for the node
1954
    @param nimg: the node image object
1955

1956
    """
1957
    idata = nresult.get(constants.NV_INSTANCELIST, None)
1958
    test = not isinstance(idata, list)
1959
    self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed"
1960
                  " (instancelist): %s", utils.SafeEncode(str(idata)))
1961
    if test:
1962
      nimg.hyp_fail = True
1963
    else:
1964
      nimg.instances = idata
1965

    
1966
  def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
1967
    """Verifies and computes a node information map
1968

1969
    @type ninfo: L{objects.Node}
1970
    @param ninfo: the node to check
1971
    @param nresult: the remote results for the node
1972
    @param nimg: the node image object
1973
    @param vg_name: the configured VG name
1974

1975
    """
1976
    node = ninfo.name
1977
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1978

    
1979
    # try to read free memory (from the hypervisor)
1980
    hv_info = nresult.get(constants.NV_HVINFO, None)
1981
    test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
1982
    _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
1983
    if not test:
1984
      try:
1985
        nimg.mfree = int(hv_info["memory_free"])
1986
      except (ValueError, TypeError):
1987
        _ErrorIf(True, self.ENODERPC, node,
1988
                 "node returned invalid nodeinfo, check hypervisor")
1989

    
1990
    # FIXME: devise a free space model for file based instances as well
1991
    if vg_name is not None:
1992
      test = (constants.NV_VGLIST not in nresult or
1993
              vg_name not in nresult[constants.NV_VGLIST])
1994
      _ErrorIf(test, self.ENODELVM, node,
1995
               "node didn't return data for the volume group '%s'"
1996
               " - it is either missing or broken", vg_name)
1997
      if not test:
1998
        try:
1999
          nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
2000
        except (ValueError, TypeError):
2001
          _ErrorIf(True, self.ENODERPC, node,
2002
                   "node returned invalid LVM info, check LVM status")
2003

    
2004
  def BuildHooksEnv(self):
2005
    """Build hooks env.
2006

2007
    Cluster-Verify hooks just ran in the post phase and their failure makes
2008
    the output be logged in the verify output and the verification to fail.
2009

2010
    """
2011
    all_nodes = self.cfg.GetNodeList()
2012
    env = {
2013
      "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
2014
      }
2015
    for node in self.cfg.GetAllNodesInfo().values():
2016
      env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
2017

    
2018
    return env, [], all_nodes
2019

    
2020
  def Exec(self, feedback_fn):
2021
    """Verify integrity of cluster, performing various test on nodes.
2022

2023
    """
2024
    self.bad = False
2025
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
2026
    verbose = self.op.verbose
2027
    self._feedback_fn = feedback_fn
2028
    feedback_fn("* Verifying global settings")
2029
    for msg in self.cfg.VerifyConfig():
2030
      _ErrorIf(True, self.ECLUSTERCFG, None, msg)
2031

    
2032
    # Check the cluster certificates
2033
    for cert_filename in constants.ALL_CERT_FILES:
2034
      (errcode, msg) = _VerifyCertificate(cert_filename)
2035
      _ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
2036

    
2037
    vg_name = self.cfg.GetVGName()
2038
    drbd_helper = self.cfg.GetDRBDHelper()
2039
    hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
2040
    cluster = self.cfg.GetClusterInfo()
2041
    nodelist = utils.NiceSort(self.cfg.GetNodeList())
2042
    nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
2043
    instancelist = utils.NiceSort(self.cfg.GetInstanceList())
2044
    instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
2045
                        for iname in instancelist)
2046
    i_non_redundant = [] # Non redundant instances
2047
    i_non_a_balanced = [] # Non auto-balanced instances
2048
    n_offline = 0 # Count of offline nodes
2049
    n_drained = 0 # Count of nodes being drained
2050
    node_vol_should = {}
2051

    
2052
    # FIXME: verify OS list
2053
    # do local checksums
2054
    master_files = [constants.CLUSTER_CONF_FILE]
2055
    master_node = self.master_node = self.cfg.GetMasterNode()
2056
    master_ip = self.cfg.GetMasterIP()
2057

    
2058
    file_names = ssconf.SimpleStore().GetFileList()
2059
    file_names.extend(constants.ALL_CERT_FILES)
2060
    file_names.extend(master_files)
2061
    if cluster.modify_etc_hosts:
2062
      file_names.append(constants.ETC_HOSTS)
2063

    
2064
    local_checksums = utils.FingerprintFiles(file_names)
2065

    
2066
    feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
2067
    node_verify_param = {
2068
      constants.NV_FILELIST: file_names,
2069
      constants.NV_NODELIST: [node.name for node in nodeinfo
2070
                              if not node.offline],
2071
      constants.NV_HYPERVISOR: hypervisors,
2072
      constants.NV_NODENETTEST: [(node.name, node.primary_ip,
2073
                                  node.secondary_ip) for node in nodeinfo
2074
                                 if not node.offline],
2075
      constants.NV_INSTANCELIST: hypervisors,
2076
      constants.NV_VERSION: None,
2077
      constants.NV_HVINFO: self.cfg.GetHypervisorType(),
2078
      constants.NV_NODESETUP: None,
2079
      constants.NV_TIME: None,
2080
      constants.NV_MASTERIP: (master_node, master_ip),
2081
      constants.NV_OSLIST: None,
2082
      }
2083

    
2084
    if vg_name is not None:
2085
      node_verify_param[constants.NV_VGLIST] = None
2086
      node_verify_param[constants.NV_LVLIST] = vg_name
2087
      node_verify_param[constants.NV_PVLIST] = [vg_name]
2088
      node_verify_param[constants.NV_DRBDLIST] = None
2089

    
2090
    if drbd_helper:
2091
      node_verify_param[constants.NV_DRBDHELPER] = drbd_helper
2092

    
2093
    # Build our expected cluster state
2094
    node_image = dict((node.name, self.NodeImage(offline=node.offline,
2095
                                                 name=node.name))
2096
                      for node in nodeinfo)
2097

    
2098
    for instance in instancelist:
2099
      inst_config = instanceinfo[instance]
2100

    
2101
      for nname in inst_config.all_nodes:
2102
        if nname not in node_image:
2103
          # ghost node
2104
          gnode = self.NodeImage(name=nname)
2105
          gnode.ghost = True
2106
          node_image[nname] = gnode
2107

    
2108
      inst_config.MapLVsByNode(node_vol_should)
2109

    
2110
      pnode = inst_config.primary_node
2111
      node_image[pnode].pinst.append(instance)
2112

    
2113
      for snode in inst_config.secondary_nodes:
2114
        nimg = node_image[snode]
2115
        nimg.sinst.append(instance)
2116
        if pnode not in nimg.sbp:
2117
          nimg.sbp[pnode] = []
2118
        nimg.sbp[pnode].append(instance)
2119

    
2120
    # At this point, we have the in-memory data structures complete,
2121
    # except for the runtime information, which we'll gather next
2122

    
2123
    # Due to the way our RPC system works, exact response times cannot be
2124
    # guaranteed (e.g. a broken node could run into a timeout). By keeping the
2125
    # time before and after executing the request, we can at least have a time
2126
    # window.
2127
    nvinfo_starttime = time.time()
2128
    all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
2129
                                           self.cfg.GetClusterName())
2130
    nvinfo_endtime = time.time()
2131

    
2132
    all_drbd_map = self.cfg.ComputeDRBDMap()
2133

    
2134
    feedback_fn("* Verifying node status")
2135

    
2136
    refos_img = None
2137

    
2138
    for node_i in nodeinfo:
2139
      node = node_i.name
2140
      nimg = node_image[node]
2141

    
2142
      if node_i.offline:
2143
        if verbose:
2144
          feedback_fn("* Skipping offline node %s" % (node,))
2145
        n_offline += 1
2146
        continue
2147

    
2148
      if node == master_node:
2149
        ntype = "master"
2150
      elif node_i.master_candidate:
2151
        ntype = "master candidate"
2152
      elif node_i.drained:
2153
        ntype = "drained"
2154
        n_drained += 1
2155
      else:
2156
        ntype = "regular"
2157
      if verbose:
2158
        feedback_fn("* Verifying node %s (%s)" % (node, ntype))
2159

    
2160
      msg = all_nvinfo[node].fail_msg
2161
      _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
2162
      if msg:
2163
        nimg.rpc_fail = True
2164
        continue
2165

    
2166
      nresult = all_nvinfo[node].payload
2167

    
2168
      nimg.call_ok = self._VerifyNode(node_i, nresult)
2169
      self._VerifyNodeNetwork(node_i, nresult)
2170
      self._VerifyNodeLVM(node_i, nresult, vg_name)
2171
      self._VerifyNodeFiles(node_i, nresult, file_names, local_checksums,
2172
                            master_files)
2173
      self._VerifyNodeDrbd(node_i, nresult, instanceinfo, drbd_helper,
2174
                           all_drbd_map)
2175
      self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
2176

    
2177
      self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
2178
      self._UpdateNodeInstances(node_i, nresult, nimg)
2179
      self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
2180
      self._UpdateNodeOS(node_i, nresult, nimg)
2181
      if not nimg.os_fail:
2182
        if refos_img is None:
2183
          refos_img = nimg
2184
        self._VerifyNodeOS(node_i, nimg, refos_img)
2185

    
2186
    feedback_fn("* Verifying instance status")
2187
    for instance in instancelist:
2188
      if verbose:
2189
        feedback_fn("* Verifying instance %s" % instance)
2190
      inst_config = instanceinfo[instance]
2191
      self._VerifyInstance(instance, inst_config, node_image)
2192
      inst_nodes_offline = []
2193

    
2194
      pnode = inst_config.primary_node
2195
      pnode_img = node_image[pnode]
2196
      _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
2197
               self.ENODERPC, pnode, "instance %s, connection to"
2198
               " primary node failed", instance)
2199

    
2200
      if pnode_img.offline:
2201
        inst_nodes_offline.append(pnode)
2202

    
2203
      # If the instance is non-redundant we cannot survive losing its primary
2204
      # node, so we are not N+1 compliant. On the other hand we have no disk
2205
      # templates with more than one secondary so that situation is not well
2206
      # supported either.
2207
      # FIXME: does not support file-backed instances
2208
      if not inst_config.secondary_nodes:
2209
        i_non_redundant.append(instance)
2210
      _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT,
2211
               instance, "instance has multiple secondary nodes: %s",
2212
               utils.CommaJoin(inst_config.secondary_nodes),
2213
               code=self.ETYPE_WARNING)
2214

    
2215
      if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
2216
        i_non_a_balanced.append(instance)
2217

    
2218
      for snode in inst_config.secondary_nodes:
2219
        s_img = node_image[snode]
2220
        _ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode,
2221
                 "instance %s, connection to secondary node failed", instance)
2222

    
2223
        if s_img.offline:
2224
          inst_nodes_offline.append(snode)
2225

    
2226
      # warn that the instance lives on offline nodes
2227
      _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
2228
               "instance lives on offline node(s) %s",
2229
               utils.CommaJoin(inst_nodes_offline))
2230
      # ... or ghost nodes
2231
      for node in inst_config.all_nodes:
2232
        _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
2233
                 "instance lives on ghost node %s", node)
2234

    
2235
    feedback_fn("* Verifying orphan volumes")
2236
    reserved = utils.FieldSet(*cluster.reserved_lvs)
2237
    self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
2238

    
2239
    feedback_fn("* Verifying orphan instances")
2240
    self._VerifyOrphanInstances(instancelist, node_image)
2241

    
2242
    if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
2243
      feedback_fn("* Verifying N+1 Memory redundancy")
2244
      self._VerifyNPlusOneMemory(node_image, instanceinfo)
2245

    
2246
    feedback_fn("* Other Notes")
2247
    if i_non_redundant:
2248
      feedback_fn("  - NOTICE: %d non-redundant instance(s) found."
2249
                  % len(i_non_redundant))
2250

    
2251
    if i_non_a_balanced:
2252
      feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
2253
                  % len(i_non_a_balanced))
2254

    
2255
    if n_offline:
2256
      feedback_fn("  - NOTICE: %d offline node(s) found." % n_offline)
2257

    
2258
    if n_drained:
2259
      feedback_fn("  - NOTICE: %d drained node(s) found." % n_drained)
2260

    
2261
    return not self.bad
2262

    
2263
  def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
2264
    """Analyze the post-hooks' result
2265

2266
    This method analyses the hook result, handles it, and sends some
2267
    nicely-formatted feedback back to the user.
2268

2269
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
2270
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
2271
    @param hooks_results: the results of the multi-node hooks rpc call
2272
    @param feedback_fn: function used send feedback back to the caller
2273
    @param lu_result: previous Exec result
2274
    @return: the new Exec result, based on the previous result
2275
        and hook results
2276

2277
    """
2278
    # We only really run POST phase hooks, and are only interested in
2279
    # their results
2280
    if phase == constants.HOOKS_PHASE_POST:
2281
      # Used to change hooks' output to proper indentation
2282
      indent_re = re.compile('^', re.M)
2283
      feedback_fn("* Hooks Results")
2284
      assert hooks_results, "invalid result from hooks"
2285

    
2286
      for node_name in hooks_results:
2287
        res = hooks_results[node_name]
2288
        msg = res.fail_msg
2289
        test = msg and not res.offline
2290
        self._ErrorIf(test, self.ENODEHOOKS, node_name,
2291
                      "Communication failure in hooks execution: %s", msg)
2292
        if res.offline or msg:
2293
          # No need to investigate payload if node is offline or gave an error.
2294
          # override manually lu_result here as _ErrorIf only
2295
          # overrides self.bad
2296
          lu_result = 1
2297
          continue
2298
        for script, hkr, output in res.payload:
2299
          test = hkr == constants.HKR_FAIL
2300
          self._ErrorIf(test, self.ENODEHOOKS, node_name,
2301
                        "Script %s failed, output:", script)
2302
          if test:
2303
            output = indent_re.sub('      ', output)
2304
            feedback_fn("%s" % output)
2305
            lu_result = 0
2306

    
2307
      return lu_result
2308

    
2309

    
2310
class LUVerifyDisks(NoHooksLU):
2311
  """Verifies the cluster disks status.
2312

2313
  """
2314
  REQ_BGL = False
2315

    
2316
  def ExpandNames(self):
2317
    self.needed_locks = {
2318
      locking.LEVEL_NODE: locking.ALL_SET,
2319
      locking.LEVEL_INSTANCE: locking.ALL_SET,
2320
    }
2321
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
2322

    
2323
  def Exec(self, feedback_fn):
2324
    """Verify integrity of cluster disks.
2325

2326
    @rtype: tuple of three items
2327
    @return: a tuple of (dict of node-to-node_error, list of instances
2328
        which need activate-disks, dict of instance: (node, volume) for
2329
        missing volumes
2330

2331
    """
2332
    result = res_nodes, res_instances, res_missing = {}, [], {}
2333

    
2334
    vg_name = self.cfg.GetVGName()
2335
    nodes = utils.NiceSort(self.cfg.GetNodeList())
2336
    instances = [self.cfg.GetInstanceInfo(name)
2337
                 for name in self.cfg.GetInstanceList()]
2338

    
2339
    nv_dict = {}
2340
    for inst in instances:
2341
      inst_lvs = {}
2342
      if (not inst.admin_up or
2343
          inst.disk_template not in constants.DTS_NET_MIRROR):
2344
        continue
2345
      inst.MapLVsByNode(inst_lvs)
2346
      # transform { iname: {node: [vol,],},} to {(node, vol): iname}
2347
      for node, vol_list in inst_lvs.iteritems():
2348
        for vol in vol_list:
2349
          nv_dict[(node, vol)] = inst
2350

    
2351
    if not nv_dict:
2352
      return result
2353

    
2354
    node_lvs = self.rpc.call_lv_list(nodes, vg_name)
2355

    
2356
    for node in nodes:
2357
      # node_volume
2358
      node_res = node_lvs[node]
2359
      if node_res.offline:
2360
        continue
2361
      msg = node_res.fail_msg
2362
      if msg:
2363
        logging.warning("Error enumerating LVs on node %s: %s", node, msg)
2364
        res_nodes[node] = msg
2365
        continue
2366

    
2367
      lvs = node_res.payload
2368
      for lv_name, (_, _, lv_online) in lvs.items():
2369
        inst = nv_dict.pop((node, lv_name), None)
2370
        if (not lv_online and inst is not None
2371
            and inst.name not in res_instances):
2372
          res_instances.append(inst.name)
2373

    
2374
    # any leftover items in nv_dict are missing LVs, let's arrange the
2375
    # data better
2376
    for key, inst in nv_dict.iteritems():
2377
      if inst.name not in res_missing:
2378
        res_missing[inst.name] = []
2379
      res_missing[inst.name].append(key)
2380

    
2381
    return result
2382

    
2383

    
2384
class LURepairDiskSizes(NoHooksLU):
2385
  """Verifies the cluster disks sizes.
2386

2387
  """
2388
  _OP_PARAMS = [("instances", _EmptyList, _TListOf(_TNonEmptyString))]
2389
  REQ_BGL = False
2390

    
2391
  def ExpandNames(self):
2392
    if self.op.instances:
2393
      self.wanted_names = []
2394
      for name in self.op.instances:
2395
        full_name = _ExpandInstanceName(self.cfg, name)
2396
        self.wanted_names.append(full_name)
2397
      self.needed_locks = {
2398
        locking.LEVEL_NODE: [],
2399
        locking.LEVEL_INSTANCE: self.wanted_names,
2400
        }
2401
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2402
    else:
2403
      self.wanted_names = None
2404
      self.needed_locks = {
2405
        locking.LEVEL_NODE: locking.ALL_SET,
2406
        locking.LEVEL_INSTANCE: locking.ALL_SET,
2407
        }
2408
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
2409

    
2410
  def DeclareLocks(self, level):
2411
    if level == locking.LEVEL_NODE and self.wanted_names is not None:
2412
      self._LockInstancesNodes(primary_only=True)
2413

    
2414
  def CheckPrereq(self):
2415
    """Check prerequisites.
2416

2417
    This only checks the optional instance list against the existing names.
2418

2419
    """
2420
    if self.wanted_names is None:
2421
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
2422

    
2423
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
2424
                             in self.wanted_names]
2425

    
2426
  def _EnsureChildSizes(self, disk):
2427
    """Ensure children of the disk have the needed disk size.
2428

2429
    This is valid mainly for DRBD8 and fixes an issue where the
2430
    children have smaller disk size.
2431

2432
    @param disk: an L{ganeti.objects.Disk} object
2433

2434
    """
2435
    if disk.dev_type == constants.LD_DRBD8:
2436
      assert disk.children, "Empty children for DRBD8?"
2437
      fchild = disk.children[0]
2438
      mismatch = fchild.size < disk.size
2439
      if mismatch:
2440
        self.LogInfo("Child disk has size %d, parent %d, fixing",
2441
                     fchild.size, disk.size)
2442
        fchild.size = disk.size
2443

    
2444
      # and we recurse on this child only, not on the metadev
2445
      return self._EnsureChildSizes(fchild) or mismatch
2446
    else:
2447
      return False
2448

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

2452
    """
2453
    # TODO: check child disks too
2454
    # TODO: check differences in size between primary/secondary nodes
2455
    per_node_disks = {}
2456
    for instance in self.wanted_instances:
2457
      pnode = instance.primary_node
2458
      if pnode not in per_node_disks:
2459
        per_node_disks[pnode] = []
2460
      for idx, disk in enumerate(instance.disks):
2461
        per_node_disks[pnode].append((instance, idx, disk))
2462

    
2463
    changed = []
2464
    for node, dskl in per_node_disks.items():
2465
      newl = [v[2].Copy() for v in dskl]
2466
      for dsk in newl:
2467
        self.cfg.SetDiskID(dsk, node)
2468
      result = self.rpc.call_blockdev_getsizes(node, newl)
2469
      if result.fail_msg:
2470
        self.LogWarning("Failure in blockdev_getsizes call to node"
2471
                        " %s, ignoring", node)
2472
        continue
2473
      if len(result.data) != len(dskl):
2474
        self.LogWarning("Invalid result from node %s, ignoring node results",
2475
                        node)
2476
        continue
2477
      for ((instance, idx, disk), size) in zip(dskl, result.data):
2478
        if size is None:
2479
          self.LogWarning("Disk %d of instance %s did not return size"
2480
                          " information, ignoring", idx, instance.name)
2481
          continue
2482
        if not isinstance(size, (int, long)):
2483
          self.LogWarning("Disk %d of instance %s did not return valid"
2484
                          " size information, ignoring", idx, instance.name)
2485
          continue
2486
        size = size >> 20
2487
        if size != disk.size:
2488
          self.LogInfo("Disk %d of instance %s has mismatched size,"
2489
                       " correcting: recorded %d, actual %d", idx,
2490
                       instance.name, disk.size, size)
2491
          disk.size = size
2492
          self.cfg.Update(instance, feedback_fn)
2493
          changed.append((instance.name, idx, size))
2494
        if self._EnsureChildSizes(disk):
2495
          self.cfg.Update(instance, feedback_fn)
2496
          changed.append((instance.name, idx, disk.size))
2497
    return changed
2498

    
2499

    
2500
class LURenameCluster(LogicalUnit):
2501
  """Rename the cluster.
2502

2503
  """
2504
  HPATH = "cluster-rename"
2505
  HTYPE = constants.HTYPE_CLUSTER
2506
  _OP_PARAMS = [("name", _NoDefault, _TNonEmptyString)]
2507

    
2508
  def BuildHooksEnv(self):
2509
    """Build hooks env.
2510

2511
    """
2512
    env = {
2513
      "OP_TARGET": self.cfg.GetClusterName(),
2514
      "NEW_NAME": self.op.name,
2515
      }
2516
    mn = self.cfg.GetMasterNode()
2517
    all_nodes = self.cfg.GetNodeList()
2518
    return env, [mn], all_nodes
2519

    
2520
  def CheckPrereq(self):
2521
    """Verify that the passed name is a valid one.
2522

2523
    """
2524
    hostname = netutils.GetHostname(name=self.op.name,
2525
                                    family=self.cfg.GetPrimaryIPFamily())
2526

    
2527
    new_name = hostname.name
2528
    self.ip = new_ip = hostname.ip
2529
    old_name = self.cfg.GetClusterName()
2530
    old_ip = self.cfg.GetMasterIP()
2531
    if new_name == old_name and new_ip == old_ip:
2532
      raise errors.OpPrereqError("Neither the name nor the IP address of the"
2533
                                 " cluster has changed",
2534
                                 errors.ECODE_INVAL)
2535
    if new_ip != old_ip:
2536
      if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
2537
        raise errors.OpPrereqError("The given cluster IP address (%s) is"
2538
                                   " reachable on the network" %
2539
                                   new_ip, errors.ECODE_NOTUNIQUE)
2540

    
2541
    self.op.name = new_name
2542

    
2543
  def Exec(self, feedback_fn):
2544
    """Rename the cluster.
2545

2546
    """
2547
    clustername = self.op.name
2548
    ip = self.ip
2549

    
2550
    # shutdown the master IP
2551
    master = self.cfg.GetMasterNode()
2552
    result = self.rpc.call_node_stop_master(master, False)
2553
    result.Raise("Could not disable the master role")
2554

    
2555
    try:
2556
      cluster = self.cfg.GetClusterInfo()
2557
      cluster.cluster_name = clustername
2558
      cluster.master_ip = ip
2559
      self.cfg.Update(cluster, feedback_fn)
2560

    
2561
      # update the known hosts file
2562
      ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
2563
      node_list = self.cfg.GetNodeList()
2564
      try:
2565
        node_list.remove(master)
2566
      except ValueError:
2567
        pass
2568
      result = self.rpc.call_upload_file(node_list,
2569
                                         constants.SSH_KNOWN_HOSTS_FILE)
2570
      for to_node, to_result in result.iteritems():
2571
        msg = to_result.fail_msg
2572
        if msg:
2573
          msg = ("Copy of file %s to node %s failed: %s" %
2574
                 (constants.SSH_KNOWN_HOSTS_FILE, to_node, msg))
2575
          self.proc.LogWarning(msg)
2576

    
2577
    finally:
2578
      result = self.rpc.call_node_start_master(master, False, False)
2579
      msg = result.fail_msg
2580
      if msg:
2581
        self.LogWarning("Could not re-enable the master role on"
2582
                        " the master, please restart manually: %s", msg)
2583

    
2584
    return clustername
2585

    
2586

    
2587
class LUSetClusterParams(LogicalUnit):
2588
  """Change the parameters of the cluster.
2589

2590
  """
2591
  HPATH = "cluster-modify"
2592
  HTYPE = constants.HTYPE_CLUSTER
2593
  _OP_PARAMS = [
2594
    ("vg_name", None, _TMaybeString),
2595
    ("enabled_hypervisors", None,
2596
     _TOr(_TAnd(_TListOf(_TElemOf(constants.HYPER_TYPES)), _TTrue), _TNone)),
2597
    ("hvparams", None, _TOr(_TDictOf(_TNonEmptyString, _TDict), _TNone)),
2598
    ("beparams", None, _TOr(_TDictOf(_TNonEmptyString, _TDict), _TNone)),
2599
    ("os_hvp", None, _TOr(_TDictOf(_TNonEmptyString, _TDict), _TNone)),
2600
    ("osparams", None, _TOr(_TDictOf(_TNonEmptyString, _TDict), _TNone)),
2601
    ("candidate_pool_size", None, _TOr(_TStrictPositiveInt, _TNone)),
2602
    ("uid_pool", None, _NoType),
2603
    ("add_uids", None, _NoType),
2604
    ("remove_uids", None, _NoType),
2605
    ("maintain_node_health", None, _TMaybeBool),
2606
    ("nicparams", None, _TOr(_TDict, _TNone)),
2607
    ("drbd_helper", None, _TOr(_TString, _TNone)),
2608
    ("default_iallocator", None, _TMaybeString),
2609
    ("reserved_lvs", None, _TOr(_TListOf(_TNonEmptyString), _TNone)),
2610
    ]
2611
  REQ_BGL = False
2612

    
2613
  def CheckArguments(self):
2614
    """Check parameters
2615

2616
    """
2617
    if self.op.uid_pool:
2618
      uidpool.CheckUidPool(self.op.uid_pool)
2619

    
2620
    if self.op.add_uids:
2621
      uidpool.CheckUidPool(self.op.add_uids)
2622

    
2623
    if self.op.remove_uids:
2624
      uidpool.CheckUidPool(self.op.remove_uids)
2625

    
2626
  def ExpandNames(self):
2627
    # FIXME: in the future maybe other cluster params won't require checking on
2628
    # all nodes to be modified.
2629
    self.needed_locks = {
2630
      locking.LEVEL_NODE: locking.ALL_SET,
2631
    }
2632
    self.share_locks[locking.LEVEL_NODE] = 1
2633

    
2634
  def BuildHooksEnv(self):
2635
    """Build hooks env.
2636

2637
    """
2638
    env = {
2639
      "OP_TARGET": self.cfg.GetClusterName(),
2640
      "NEW_VG_NAME": self.op.vg_name,
2641
      }
2642
    mn = self.cfg.GetMasterNode()
2643
    return env, [mn], [mn]
2644

    
2645
  def CheckPrereq(self):
2646
    """Check prerequisites.
2647

2648
    This checks whether the given params don't conflict and
2649
    if the given volume group is valid.
2650

2651
    """
2652
    if self.op.vg_name is not None and not self.op.vg_name:
2653
      if self.cfg.HasAnyDiskOfType(constants.LD_LV):
2654
        raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based"
2655
                                   " instances exist", errors.ECODE_INVAL)
2656

    
2657
    if self.op.drbd_helper is not None and not self.op.drbd_helper:
2658
      if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8):
2659
        raise errors.OpPrereqError("Cannot disable drbd helper while"
2660
                                   " drbd-based instances exist",
2661
                                   errors.ECODE_INVAL)
2662

    
2663
    node_list = self.acquired_locks[locking.LEVEL_NODE]
2664

    
2665
    # if vg_name not None, checks given volume group on all nodes
2666
    if self.op.vg_name:
2667
      vglist = self.rpc.call_vg_list(node_list)
2668
      for node in node_list:
2669
        msg = vglist[node].fail_msg
2670
        if msg:
2671
          # ignoring down node
2672
          self.LogWarning("Error while gathering data on node %s"
2673
                          " (ignoring node): %s", node, msg)
2674
          continue
2675
        vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
2676
                                              self.op.vg_name,
2677
                                              constants.MIN_VG_SIZE)
2678
        if vgstatus:
2679
          raise errors.OpPrereqError("Error on node '%s': %s" %
2680
                                     (node, vgstatus), errors.ECODE_ENVIRON)
2681

    
2682
    if self.op.drbd_helper:
2683
      # checks given drbd helper on all nodes
2684
      helpers = self.rpc.call_drbd_helper(node_list)
2685
      for node in node_list:
2686
        ninfo = self.cfg.GetNodeInfo(node)
2687
        if ninfo.offline:
2688
          self.LogInfo("Not checking drbd helper on offline node %s", node)
2689
          continue
2690
        msg = helpers[node].fail_msg
2691
        if msg:
2692
          raise errors.OpPrereqError("Error checking drbd helper on node"
2693
                                     " '%s': %s" % (node, msg),
2694
                                     errors.ECODE_ENVIRON)
2695
        node_helper = helpers[node].payload
2696
        if node_helper != self.op.drbd_helper:
2697
          raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" %
2698
                                     (node, node_helper), errors.ECODE_ENVIRON)
2699

    
2700
    self.cluster = cluster = self.cfg.GetClusterInfo()
2701
    # validate params changes
2702
    if self.op.beparams:
2703
      utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
2704
      self.new_beparams = cluster.SimpleFillBE(self.op.beparams)
2705

    
2706
    if self.op.nicparams:
2707
      utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
2708
      self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams)
2709
      objects.NIC.CheckParameterSyntax(self.new_nicparams)
2710
      nic_errors = []
2711

    
2712
      # check all instances for consistency
2713
      for instance in self.cfg.GetAllInstancesInfo().values():
2714
        for nic_idx, nic in enumerate(instance.nics):
2715
          params_copy = copy.deepcopy(nic.nicparams)
2716
          params_filled = objects.FillDict(self.new_nicparams, params_copy)
2717

    
2718
          # check parameter syntax
2719
          try:
2720
            objects.NIC.CheckParameterSyntax(params_filled)
2721
          except errors.ConfigurationError, err:
2722
            nic_errors.append("Instance %s, nic/%d: %s" %
2723
                              (instance.name, nic_idx, err))
2724

    
2725
          # if we're moving instances to routed, check that they have an ip
2726
          target_mode = params_filled[constants.NIC_MODE]
2727
          if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
2728
            nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
2729
                              (instance.name, nic_idx))
2730
      if nic_errors:
2731
        raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
2732
                                   "\n".join(nic_errors))
2733

    
2734
    # hypervisor list/parameters
2735
    self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
2736
    if self.op.hvparams:
2737
      for hv_name, hv_dict in self.op.hvparams.items():
2738
        if hv_name not in self.new_hvparams:
2739
          self.new_hvparams[hv_name] = hv_dict
2740
        else:
2741
          self.new_hvparams[hv_name].update(hv_dict)
2742

    
2743
    # os hypervisor parameters
2744
    self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
2745
    if self.op.os_hvp:
2746
      for os_name, hvs in self.op.os_hvp.items():
2747
        if os_name not in self.new_os_hvp:
2748
          self.new_os_hvp[os_name] = hvs
2749
        else:
2750
          for hv_name, hv_dict in hvs.items():
2751
            if hv_name not in self.new_os_hvp[os_name]:
2752
              self.new_os_hvp[os_name][hv_name] = hv_dict
2753
            else:
2754
              self.new_os_hvp[os_name][hv_name].update(hv_dict)
2755

    
2756
    # os parameters
2757
    self.new_osp = objects.FillDict(cluster.osparams, {})
2758
    if self.op.osparams:
2759
      for os_name, osp in self.op.osparams.items():
2760
        if os_name not in self.new_osp:
2761
          self.new_osp[os_name] = {}
2762

    
2763
        self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp,
2764
                                                  use_none=True)
2765

    
2766
        if not self.new_osp[os_name]:
2767
          # we removed all parameters
2768
          del self.new_osp[os_name]
2769
        else:
2770
          # check the parameter validity (remote check)
2771
          _CheckOSParams(self, False, [self.cfg.GetMasterNode()],
2772
                         os_name, self.new_osp[os_name])
2773

    
2774
    # changes to the hypervisor list
2775
    if self.op.enabled_hypervisors is not None:
2776
      self.hv_list = self.op.enabled_hypervisors
2777
      for hv in self.hv_list:
2778
        # if the hypervisor doesn't already exist in the cluster
2779
        # hvparams, we initialize it to empty, and then (in both
2780
        # cases) we make sure to fill the defaults, as we might not
2781
        # have a complete defaults list if the hypervisor wasn't
2782
        # enabled before
2783
        if hv not in new_hvp:
2784
          new_hvp[hv] = {}
2785
        new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
2786
        utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
2787
    else:
2788
      self.hv_list = cluster.enabled_hypervisors
2789

    
2790
    if self.op.hvparams or self.op.enabled_hypervisors is not None:
2791
      # either the enabled list has changed, or the parameters have, validate
2792
      for hv_name, hv_params in self.new_hvparams.items():
2793
        if ((self.op.hvparams and hv_name in self.op.hvparams) or
2794
            (self.op.enabled_hypervisors and
2795
             hv_name in self.op.enabled_hypervisors)):
2796
          # either this is a new hypervisor, or its parameters have changed
2797
          hv_class = hypervisor.GetHypervisor(hv_name)
2798
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2799
          hv_class.CheckParameterSyntax(hv_params)
2800
          _CheckHVParams(self, node_list, hv_name, hv_params)
2801

    
2802
    if self.op.os_hvp:
2803
      # no need to check any newly-enabled hypervisors, since the
2804
      # defaults have already been checked in the above code-block
2805
      for os_name, os_hvp in self.new_os_hvp.items():
2806
        for hv_name, hv_params in os_hvp.items():
2807
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2808
          # we need to fill in the new os_hvp on top of the actual hv_p
2809
          cluster_defaults = self.new_hvparams.get(hv_name, {})
2810
          new_osp = objects.FillDict(cluster_defaults, hv_params)
2811
          hv_class = hypervisor.GetHypervisor(hv_name)
2812
          hv_class.CheckParameterSyntax(new_osp)
2813
          _CheckHVParams(self, node_list, hv_name, new_osp)
2814

    
2815
    if self.op.default_iallocator:
2816
      alloc_script = utils.FindFile(self.op.default_iallocator,
2817
                                    constants.IALLOCATOR_SEARCH_PATH,
2818
                                    os.path.isfile)
2819
      if alloc_script is None:
2820
        raise errors.OpPrereqError("Invalid default iallocator script '%s'"
2821
                                   " specified" % self.op.default_iallocator,
2822
                                   errors.ECODE_INVAL)
2823

    
2824
  def Exec(self, feedback_fn):
2825
    """Change the parameters of the cluster.
2826

2827
    """
2828
    if self.op.vg_name is not None:
2829
      new_volume = self.op.vg_name
2830
      if not new_volume:
2831
        new_volume = None
2832
      if new_volume != self.cfg.GetVGName():
2833
        self.cfg.SetVGName(new_volume)
2834
      else:
2835
        feedback_fn("Cluster LVM configuration already in desired"
2836
                    " state, not changing")
2837
    if self.op.drbd_helper is not None:
2838
      new_helper = self.op.drbd_helper
2839
      if not new_helper:
2840
        new_helper = None
2841
      if new_helper != self.cfg.GetDRBDHelper():
2842
        self.cfg.SetDRBDHelper(new_helper)
2843
      else:
2844
        feedback_fn("Cluster DRBD helper already in desired state,"
2845
                    " not changing")
2846
    if self.op.hvparams:
2847
      self.cluster.hvparams = self.new_hvparams
2848
    if self.op.os_hvp:
2849
      self.cluster.os_hvp = self.new_os_hvp
2850
    if self.op.enabled_hypervisors is not None:
2851
      self.cluster.hvparams = self.new_hvparams
2852
      self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
2853
    if self.op.beparams:
2854
      self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
2855
    if self.op.nicparams:
2856
      self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
2857
    if self.op.osparams:
2858
      self.cluster.osparams = self.new_osp
2859

    
2860
    if self.op.candidate_pool_size is not None:
2861
      self.cluster.candidate_pool_size = self.op.candidate_pool_size
2862
      # we need to update the pool size here, otherwise the save will fail
2863
      _AdjustCandidatePool(self, [])
2864

    
2865
    if self.op.maintain_node_health is not None:
2866
      self.cluster.maintain_node_health = self.op.maintain_node_health
2867

    
2868
    if self.op.add_uids is not None:
2869
      uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
2870

    
2871
    if self.op.remove_uids is not None:
2872
      uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
2873

    
2874
    if self.op.uid_pool is not None:
2875
      self.cluster.uid_pool = self.op.uid_pool
2876

    
2877
    if self.op.default_iallocator is not None:
2878
      self.cluster.default_iallocator = self.op.default_iallocator
2879

    
2880
    if self.op.reserved_lvs is not None:
2881
      self.cluster.reserved_lvs = self.op.reserved_lvs
2882

    
2883
    self.cfg.Update(self.cluster, feedback_fn)
2884

    
2885

    
2886
def _RedistributeAncillaryFiles(lu, additional_nodes=None):
2887
  """Distribute additional files which are part of the cluster configuration.
2888

2889
  ConfigWriter takes care of distributing the config and ssconf files, but
2890
  there are more files which should be distributed to all nodes. This function
2891
  makes sure those are copied.
2892

2893
  @param lu: calling logical unit
2894
  @param additional_nodes: list of nodes not in the config to distribute to
2895

2896
  """
2897
  # 1. Gather target nodes
2898
  myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
2899
  dist_nodes = lu.cfg.GetOnlineNodeList()
2900
  if additional_nodes is not None:
2901
    dist_nodes.extend(additional_nodes)
2902
  if myself.name in dist_nodes:
2903
    dist_nodes.remove(myself.name)
2904

    
2905
  # 2. Gather files to distribute
2906
  dist_files = set([constants.ETC_HOSTS,
2907
                    constants.SSH_KNOWN_HOSTS_FILE,
2908
                    constants.RAPI_CERT_FILE,
2909
                    constants.RAPI_USERS_FILE,
2910
                    constants.CONFD_HMAC_KEY,
2911
                    constants.CLUSTER_DOMAIN_SECRET_FILE,
2912
                   ])
2913

    
2914
  enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
2915
  for hv_name in enabled_hypervisors:
2916
    hv_class = hypervisor.GetHypervisor(hv_name)
2917
    dist_files.update(hv_class.GetAncillaryFiles())
2918

    
2919
  # 3. Perform the files upload
2920
  for fname in dist_files:
2921
    if os.path.exists(fname):
2922
      result = lu.rpc.call_upload_file(dist_nodes, fname)
2923
      for to_node, to_result in result.items():
2924
        msg = to_result.fail_msg
2925
        if msg:
2926
          msg = ("Copy of file %s to node %s failed: %s" %
2927
                 (fname, to_node, msg))
2928
          lu.proc.LogWarning(msg)
2929

    
2930

    
2931
class LURedistributeConfig(NoHooksLU):
2932
  """Force the redistribution of cluster configuration.
2933

2934
  This is a very simple LU.
2935

2936
  """
2937
  REQ_BGL = False
2938

    
2939
  def ExpandNames(self):
2940
    self.needed_locks = {
2941
      locking.LEVEL_NODE: locking.ALL_SET,
2942
    }
2943
    self.share_locks[locking.LEVEL_NODE] = 1
2944

    
2945
  def Exec(self, feedback_fn):
2946
    """Redistribute the configuration.
2947

2948
    """
2949
    self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
2950
    _RedistributeAncillaryFiles(self)
2951

    
2952

    
2953
def _WaitForSync(lu, instance, disks=None, oneshot=False):
2954
  """Sleep and poll for an instance's disk to sync.
2955

2956
  """
2957
  if not instance.disks or disks is not None and not disks:
2958
    return True
2959

    
2960
  disks = _ExpandCheckDisks(instance, disks)
2961

    
2962
  if not oneshot:
2963
    lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
2964

    
2965
  node = instance.primary_node
2966

    
2967
  for dev in disks:
2968
    lu.cfg.SetDiskID(dev, node)
2969

    
2970
  # TODO: Convert to utils.Retry
2971

    
2972
  retries = 0
2973
  degr_retries = 10 # in seconds, as we sleep 1 second each time
2974
  while True:
2975
    max_time = 0
2976
    done = True
2977
    cumul_degraded = False
2978
    rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
2979
    msg = rstats.fail_msg
2980
    if msg:
2981
      lu.LogWarning("Can't get any data from node %s: %s", node, msg)
2982
      retries += 1
2983
      if retries >= 10:
2984
        raise errors.RemoteError("Can't contact node %s for mirror data,"
2985
                                 " aborting." % node)
2986
      time.sleep(6)
2987
      continue
2988
    rstats = rstats.payload
2989
    retries = 0
2990
    for i, mstat in enumerate(rstats):
2991
      if mstat is None:
2992
        lu.LogWarning("Can't compute data for node %s/%s",
2993
                           node, disks[i].iv_name)
2994
        continue
2995

    
2996
      cumul_degraded = (cumul_degraded or
2997
                        (mstat.is_degraded and mstat.sync_percent is None))
2998
      if mstat.sync_percent is not None:
2999
        done = False
3000
        if mstat.estimated_time is not None:
3001
          rem_time = ("%s remaining (estimated)" %
3002
                      utils.FormatSeconds(mstat.estimated_time))
3003
          max_time = mstat.estimated_time
3004
        else:
3005
          rem_time = "no time estimate"
3006
        lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
3007
                        (disks[i].iv_name, mstat.sync_percent, rem_time))
3008

    
3009
    # if we're done but degraded, let's do a few small retries, to
3010
    # make sure we see a stable and not transient situation; therefore
3011
    # we force restart of the loop
3012
    if (done or oneshot) and cumul_degraded and degr_retries > 0:
3013
      logging.info("Degraded disks found, %d retries left", degr_retries)
3014
      degr_retries -= 1
3015
      time.sleep(1)
3016
      continue
3017

    
3018
    if done or oneshot:
3019
      break
3020

    
3021
    time.sleep(min(60, max_time))
3022

    
3023
  if done:
3024
    lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
3025
  return not cumul_degraded
3026

    
3027

    
3028
def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
3029
  """Check that mirrors are not degraded.
3030

3031
  The ldisk parameter, if True, will change the test from the
3032
  is_degraded attribute (which represents overall non-ok status for
3033
  the device(s)) to the ldisk (representing the local storage status).
3034

3035
  """
3036
  lu.cfg.SetDiskID(dev, node)
3037

    
3038
  result = True
3039

    
3040
  if on_primary or dev.AssembleOnSecondary():
3041
    rstats = lu.rpc.call_blockdev_find(node, dev)
3042
    msg = rstats.fail_msg
3043
    if msg:
3044
      lu.LogWarning("Can't find disk on node %s: %s", node, msg)
3045
      result = False
3046
    elif not rstats.payload:
3047
      lu.LogWarning("Can't find disk on node %s", node)
3048
      result = False
3049
    else:
3050
      if ldisk:
3051
        result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
3052
      else:
3053
        result = result and not rstats.payload.is_degraded
3054

    
3055
  if dev.children:
3056
    for child in dev.children:
3057
      result = result and _CheckDiskConsistency(lu, child, node, on_primary)
3058

    
3059
  return result
3060

    
3061

    
3062
class LUDiagnoseOS(NoHooksLU):
3063
  """Logical unit for OS diagnose/query.
3064

3065
  """
3066
  _OP_PARAMS = [
3067
    _POutputFields,
3068
    ("names", _EmptyList, _TListOf(_TNonEmptyString)),
3069
    ]
3070
  REQ_BGL = False
3071
  _FIELDS_STATIC = utils.FieldSet()
3072
  _FIELDS_DYNAMIC = utils.FieldSet("name", "valid", "node_status", "variants",
3073
                                   "parameters", "api_versions")
3074

    
3075
  def CheckArguments(self):
3076
    if self.op.names:
3077
      raise errors.OpPrereqError("Selective OS query not supported",
3078
                                 errors.ECODE_INVAL)
3079

    
3080
    _CheckOutputFields(static=self._FIELDS_STATIC,
3081
                       dynamic=self._FIELDS_DYNAMIC,
3082
                       selected=self.op.output_fields)
3083

    
3084
  def ExpandNames(self):
3085
    # Lock all nodes, in shared mode
3086
    # Temporary removal of locks, should be reverted later
3087
    # TODO: reintroduce locks when they are lighter-weight
3088
    self.needed_locks = {}
3089
    #self.share_locks[locking.LEVEL_NODE] = 1
3090
    #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3091

    
3092
  @staticmethod
3093
  def _DiagnoseByOS(rlist):
3094
    """Remaps a per-node return list into an a per-os per-node dictionary
3095

3096
    @param rlist: a map with node names as keys and OS objects as values
3097

3098
    @rtype: dict
3099
    @return: a dictionary with osnames as keys and as value another
3100
        map, with nodes as keys and tuples of (path, status, diagnose,
3101
        variants, parameters, api_versions) as values, eg::
3102

3103
          {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
3104
                                     (/srv/..., False, "invalid api")],
3105
                           "node2": [(/srv/..., True, "", [], [])]}
3106
          }
3107

3108
    """
3109
    all_os = {}
3110
    # we build here the list of nodes that didn't fail the RPC (at RPC
3111
    # level), so that nodes with a non-responding node daemon don't
3112
    # make all OSes invalid
3113
    good_nodes = [node_name for node_name in rlist
3114
                  if not rlist[node_name].fail_msg]
3115
    for node_name, nr in rlist.items():
3116
      if nr.fail_msg or not nr.payload:
3117
        continue
3118
      for (name, path, status, diagnose, variants,
3119
           params, api_versions) in nr.payload:
3120
        if name not in all_os:
3121
          # build a list of nodes for this os containing empty lists
3122
          # for each node in node_list
3123
          all_os[name] = {}
3124
          for nname in good_nodes:
3125
            all_os[name][nname] = []
3126
        # convert params from [name, help] to (name, help)
3127
        params = [tuple(v) for v in params]
3128
        all_os[name][node_name].append((path, status, diagnose,
3129
                                        variants, params, api_versions))
3130
    return all_os
3131

    
3132
  def Exec(self, feedback_fn):
3133
    """Compute the list of OSes.
3134

3135
    """
3136
    valid_nodes = [node for node in self.cfg.GetOnlineNodeList()]
3137
    node_data = self.rpc.call_os_diagnose(valid_nodes)
3138
    pol = self._DiagnoseByOS(node_data)
3139
    output = []
3140

    
3141
    for os_name, os_data in pol.items():
3142
      row = []
3143
      valid = True
3144
      (variants, params, api_versions) = null_state = (set(), set(), set())
3145
      for idx, osl in enumerate(os_data.values()):
3146
        valid = bool(valid and osl and osl[0][1])
3147
        if not valid:
3148
          (variants, params, api_versions) = null_state
3149
          break
3150
        node_variants, node_params, node_api = osl[0][3:6]
3151
        if idx == 0: # first entry
3152
          variants = set(node_variants)
3153
          params = set(node_params)
3154
          api_versions = set(node_api)
3155
        else: # keep consistency
3156
          variants.intersection_update(node_variants)
3157
          params.intersection_update(node_params)
3158
          api_versions.intersection_update(node_api)
3159

    
3160
      for field in self.op.output_fields:
3161
        if field == "name":
3162
          val = os_name
3163
        elif field == "valid":
3164
          val = valid
3165
        elif field == "node_status":
3166
          # this is just a copy of the dict
3167
          val = {}
3168
          for node_name, nos_list in os_data.items():
3169
            val[node_name] = nos_list
3170
        elif field == "variants":
3171
          val = list(variants)
3172
        elif field == "parameters":
3173
          val = list(params)
3174
        elif field == "api_versions":
3175
          val = list(api_versions)
3176
        else:
3177
          raise errors.ParameterError(field)
3178
        row.append(val)
3179
      output.append(row)
3180

    
3181
    return output
3182

    
3183

    
3184
class LURemoveNode(LogicalUnit):
3185
  """Logical unit for removing a node.
3186

3187
  """
3188
  HPATH = "node-remove"
3189
  HTYPE = constants.HTYPE_NODE
3190
  _OP_PARAMS = [
3191
    _PNodeName,
3192
    ]
3193

    
3194
  def BuildHooksEnv(self):
3195
    """Build hooks env.
3196

3197
    This doesn't run on the target node in the pre phase as a failed
3198
    node would then be impossible to remove.
3199

3200
    """
3201
    env = {
3202
      "OP_TARGET": self.op.node_name,
3203
      "NODE_NAME": self.op.node_name,
3204
      }
3205
    all_nodes = self.cfg.GetNodeList()
3206
    try:
3207
      all_nodes.remove(self.op.node_name)
3208
    except ValueError:
3209
      logging.warning("Node %s which is about to be removed not found"
3210
                      " in the all nodes list", self.op.node_name)
3211
    return env, all_nodes, all_nodes
3212

    
3213
  def CheckPrereq(self):
3214
    """Check prerequisites.
3215

3216
    This checks:
3217
     - the node exists in the configuration
3218
     - it does not have primary or secondary instances
3219
     - it's not the master
3220

3221
    Any errors are signaled by raising errors.OpPrereqError.
3222

3223
    """
3224
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3225
    node = self.cfg.GetNodeInfo(self.op.node_name)
3226
    assert node is not None
3227

    
3228
    instance_list = self.cfg.GetInstanceList()
3229

    
3230
    masternode = self.cfg.GetMasterNode()
3231
    if node.name == masternode:
3232
      raise errors.OpPrereqError("Node is the master node,"
3233
                                 " you need to failover first.",
3234
                                 errors.ECODE_INVAL)
3235

    
3236
    for instance_name in instance_list:
3237
      instance = self.cfg.GetInstanceInfo(instance_name)
3238
      if node.name in instance.all_nodes:
3239
        raise errors.OpPrereqError("Instance %s is still running on the node,"
3240
                                   " please remove first." % instance_name,
3241
                                   errors.ECODE_INVAL)
3242
    self.op.node_name = node.name
3243
    self.node = node
3244

    
3245
  def Exec(self, feedback_fn):
3246
    """Removes the node from the cluster.
3247

3248
    """
3249
    node = self.node
3250
    logging.info("Stopping the node daemon and removing configs from node %s",
3251
                 node.name)
3252

    
3253
    modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
3254

    
3255
    # Promote nodes to master candidate as needed
3256
    _AdjustCandidatePool(self, exceptions=[node.name])
3257
    self.context.RemoveNode(node.name)
3258

    
3259
    # Run post hooks on the node before it's removed
3260
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
3261
    try:
3262
      hm.RunPhase(constants.HOOKS_PHASE_POST, [node.name])
3263
    except:
3264
      # pylint: disable-msg=W0702
3265
      self.LogWarning("Errors occurred running hooks on %s" % node.name)
3266

    
3267
    result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
3268
    msg = result.fail_msg
3269
    if msg:
3270
      self.LogWarning("Errors encountered on the remote node while leaving"
3271
                      " the cluster: %s", msg)
3272

    
3273
    # Remove node from our /etc/hosts
3274
    if self.cfg.GetClusterInfo().modify_etc_hosts:
3275
      master_node = self.cfg.GetMasterNode()
3276
      result = self.rpc.call_etc_hosts_modify(master_node,
3277
                                              constants.ETC_HOSTS_REMOVE,
3278
                                              node.name, None)
3279
      result.Raise("Can't update hosts file with new host data")
3280
      _RedistributeAncillaryFiles(self)
3281

    
3282

    
3283
class LUQueryNodes(NoHooksLU):
3284
  """Logical unit for querying nodes.
3285

3286
  """
3287
  # pylint: disable-msg=W0142
3288
  _OP_PARAMS = [
3289
    _POutputFields,
3290
    ("names", _EmptyList, _TListOf(_TNonEmptyString)),
3291
    ("use_locking", False, _TBool),
3292
    ]
3293
  REQ_BGL = False
3294

    
3295
  _SIMPLE_FIELDS = ["name", "serial_no", "ctime", "mtime", "uuid",
3296
                    "master_candidate", "offline", "drained"]
3297

    
3298
  _FIELDS_DYNAMIC = utils.FieldSet(
3299
    "dtotal", "dfree",
3300
    "mtotal", "mnode", "mfree",
3301
    "bootid",
3302
    "ctotal", "cnodes", "csockets",
3303
    )
3304

    
3305
  _FIELDS_STATIC = utils.FieldSet(*[
3306
    "pinst_cnt", "sinst_cnt",
3307
    "pinst_list", "sinst_list",
3308
    "pip", "sip", "tags",
3309
    "master",
3310
    "role"] + _SIMPLE_FIELDS
3311
    )
3312

    
3313
  def CheckArguments(self):
3314
    _CheckOutputFields(static=self._FIELDS_STATIC,
3315
                       dynamic=self._FIELDS_DYNAMIC,
3316
                       selected=self.op.output_fields)
3317

    
3318
  def ExpandNames(self):
3319
    self.needed_locks = {}
3320
    self.share_locks[locking.LEVEL_NODE] = 1
3321

    
3322
    if self.op.names:
3323
      self.wanted = _GetWantedNodes(self, self.op.names)
3324
    else:
3325
      self.wanted = locking.ALL_SET
3326

    
3327
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
3328
    self.do_locking = self.do_node_query and self.op.use_locking
3329
    if self.do_locking:
3330
      # if we don't request only static fields, we need to lock the nodes
3331
      self.needed_locks[locking.LEVEL_NODE] = self.wanted
3332

    
3333
  def Exec(self, feedback_fn):
3334
    """Computes the list of nodes and their attributes.
3335

3336
    """
3337
    all_info = self.cfg.GetAllNodesInfo()
3338
    if self.do_locking:
3339
      nodenames = self.acquired_locks[locking.LEVEL_NODE]
3340
    elif self.wanted != locking.ALL_SET:
3341
      nodenames = self.wanted
3342
      missing = set(nodenames).difference(all_info.keys())
3343
      if missing:
3344
        raise errors.OpExecError(
3345
          "Some nodes were removed before retrieving their data: %s" % missing)
3346
    else:
3347
      nodenames = all_info.keys()
3348

    
3349
    nodenames = utils.NiceSort(nodenames)
3350
    nodelist = [all_info[name] for name in nodenames]
3351

    
3352
    # begin data gathering
3353

    
3354
    if self.do_node_query:
3355
      live_data = {}
3356
      node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
3357
                                          self.cfg.GetHypervisorType())
3358
      for name in nodenames:
3359
        nodeinfo = node_data[name]
3360
        if not nodeinfo.fail_msg and nodeinfo.payload:
3361
          nodeinfo = nodeinfo.payload
3362
          fn = utils.TryConvert
3363
          live_data[name] = {
3364
            "mtotal": fn(int, nodeinfo.get('memory_total', None)),
3365
            "mnode": fn(int, nodeinfo.get('memory_dom0', None)),
3366
            "mfree": fn(int, nodeinfo.get('memory_free', None)),
3367
            "dtotal": fn(int, nodeinfo.get('vg_size', None)),
3368
            "dfree": fn(int, nodeinfo.get('vg_free', None)),
3369
            "ctotal": fn(int, nodeinfo.get('cpu_total', None)),
3370
            "bootid": nodeinfo.get('bootid', None),
3371
            "cnodes": fn(int, nodeinfo.get('cpu_nodes', None)),
3372
            "csockets": fn(int, nodeinfo.get('cpu_sockets', None)),
3373
            }
3374
        else:
3375
          live_data[name] = {}
3376
    else:
3377
      live_data = dict.fromkeys(nodenames, {})
3378

    
3379
    node_to_primary = dict([(name, set()) for name in nodenames])
3380
    node_to_secondary = dict([(name, set()) for name in nodenames])
3381

    
3382
    inst_fields = frozenset(("pinst_cnt", "pinst_list",
3383
                             "sinst_cnt", "sinst_list"))
3384
    if inst_fields & frozenset(self.op.output_fields):
3385
      inst_data = self.cfg.GetAllInstancesInfo()
3386

    
3387
      for inst in inst_data.values():
3388
        if inst.primary_node in node_to_primary:
3389
          node_to_primary[inst.primary_node].add(inst.name)
3390
        for secnode in inst.secondary_nodes:
3391
          if secnode in node_to_secondary:
3392
            node_to_secondary[secnode].add(inst.name)
3393

    
3394
    master_node = self.cfg.GetMasterNode()
3395

    
3396
    # end data gathering
3397

    
3398
    output = []
3399
    for node in nodelist:
3400
      node_output = []
3401
      for field in self.op.output_fields:
3402
        if field in self._SIMPLE_FIELDS:
3403
          val = getattr(node, field)
3404
        elif field == "pinst_list":
3405
          val = list(node_to_primary[node.name])
3406
        elif field == "sinst_list":
3407
          val = list(node_to_secondary[node.name])
3408
        elif field == "pinst_cnt":
3409
          val = len(node_to_primary[node.name])
3410
        elif field == "sinst_cnt":
3411
          val = len(node_to_secondary[node.name])
3412
        elif field == "pip":
3413
          val = node.primary_ip
3414
        elif field == "sip":
3415
          val = node.secondary_ip
3416
        elif field == "tags":
3417
          val = list(node.GetTags())
3418
        elif field == "master":
3419
          val = node.name == master_node
3420
        elif self._FIELDS_DYNAMIC.Matches(field):
3421
          val = live_data[node.name].get(field, None)
3422
        elif field == "role":
3423
          if node.name == master_node:
3424
            val = "M"
3425
          elif node.master_candidate:
3426
            val = "C"
3427
          elif node.drained:
3428
            val = "D"
3429
          elif node.offline:
3430
            val = "O"
3431
          else:
3432
            val = "R"
3433
        else:
3434
          raise errors.ParameterError(field)
3435
        node_output.append(val)
3436
      output.append(node_output)
3437

    
3438
    return output
3439

    
3440

    
3441
class LUQueryNodeVolumes(NoHooksLU):
3442
  """Logical unit for getting volumes on node(s).
3443

3444
  """
3445
  _OP_PARAMS = [
3446
    ("nodes", _EmptyList, _TListOf(_TNonEmptyString)),
3447
    ("output_fields", _NoDefault, _TListOf(_TNonEmptyString)),
3448
    ]
3449
  REQ_BGL = False
3450
  _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
3451
  _FIELDS_STATIC = utils.FieldSet("node")
3452

    
3453
  def CheckArguments(self):
3454
    _CheckOutputFields(static=self._FIELDS_STATIC,
3455
                       dynamic=self._FIELDS_DYNAMIC,
3456
                       selected=self.op.output_fields)
3457

    
3458
  def ExpandNames(self):
3459
    self.needed_locks = {}
3460
    self.share_locks[locking.LEVEL_NODE] = 1
3461
    if not self.op.nodes:
3462
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3463
    else:
3464
      self.needed_locks[locking.LEVEL_NODE] = \
3465
        _GetWantedNodes(self, self.op.nodes)
3466

    
3467
  def Exec(self, feedback_fn):
3468
    """Computes the list of nodes and their attributes.
3469

3470
    """
3471
    nodenames = self.acquired_locks[locking.LEVEL_NODE]
3472
    volumes = self.rpc.call_node_volumes(nodenames)
3473

    
3474
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
3475
             in self.cfg.GetInstanceList()]
3476

    
3477
    lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
3478

    
3479
    output = []
3480
    for node in nodenames:
3481
      nresult = volumes[node]
3482
      if nresult.offline:
3483
        continue
3484
      msg = nresult.fail_msg
3485
      if msg:
3486
        self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
3487
        continue
3488

    
3489
      node_vols = nresult.payload[:]
3490
      node_vols.sort(key=lambda vol: vol['dev'])
3491

    
3492
      for vol in node_vols:
3493
        node_output = []
3494
        for field in self.op.output_fields:
3495
          if field == "node":
3496
            val = node
3497
          elif field == "phys":
3498
            val = vol['dev']
3499
          elif field == "vg":
3500
            val = vol['vg']
3501
          elif field == "name":
3502
            val = vol['name']
3503
          elif field == "size":
3504
            val = int(float(vol['size']))
3505
          elif field == "instance":
3506
            for inst in ilist:
3507
              if node not in lv_by_node[inst]:
3508
                continue
3509
              if vol['name'] in lv_by_node[inst][node]:
3510
                val = inst.name
3511
                break
3512
            else:
3513
              val = '-'
3514
          else:
3515
            raise errors.ParameterError(field)
3516
          node_output.append(str(val))
3517

    
3518
        output.append(node_output)
3519

    
3520
    return output
3521

    
3522

    
3523
class LUQueryNodeStorage(NoHooksLU):
3524
  """Logical unit for getting information on storage units on node(s).
3525

3526
  """
3527
  _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
3528
  _OP_PARAMS = [
3529
    ("nodes", _EmptyList, _TListOf(_TNonEmptyString)),
3530
    ("storage_type", _NoDefault, _CheckStorageType),
3531
    ("output_fields", _NoDefault, _TListOf(_TNonEmptyString)),
3532
    ("name", None, _TMaybeString),
3533
    ]
3534
  REQ_BGL = False
3535

    
3536
  def CheckArguments(self):
3537
    _CheckOutputFields(static=self._FIELDS_STATIC,
3538
                       dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
3539
                       selected=self.op.output_fields)
3540

    
3541
  def ExpandNames(self):
3542
    self.needed_locks = {}
3543
    self.share_locks[locking.LEVEL_NODE] = 1
3544

    
3545
    if self.op.nodes:
3546
      self.needed_locks[locking.LEVEL_NODE] = \
3547
        _GetWantedNodes(self, self.op.nodes)
3548
    else:
3549
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3550

    
3551
  def Exec(self, feedback_fn):
3552
    """Computes the list of nodes and their attributes.
3553

3554
    """
3555
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
3556

    
3557
    # Always get name to sort by
3558
    if constants.SF_NAME in self.op.output_fields:
3559
      fields = self.op.output_fields[:]
3560
    else:
3561
      fields = [constants.SF_NAME] + self.op.output_fields
3562

    
3563
    # Never ask for node or type as it's only known to the LU
3564
    for extra in [constants.SF_NODE, constants.SF_TYPE]:
3565
      while extra in fields:
3566
        fields.remove(extra)
3567

    
3568
    field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
3569
    name_idx = field_idx[constants.SF_NAME]
3570

    
3571
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3572
    data = self.rpc.call_storage_list(self.nodes,
3573
                                      self.op.storage_type, st_args,
3574
                                      self.op.name, fields)
3575

    
3576
    result = []
3577

    
3578
    for node in utils.NiceSort(self.nodes):
3579
      nresult = data[node]
3580
      if nresult.offline:
3581
        continue
3582

    
3583
      msg = nresult.fail_msg
3584
      if msg:
3585
        self.LogWarning("Can't get storage data from node %s: %s", node, msg)
3586
        continue
3587

    
3588
      rows = dict([(row[name_idx], row) for row in nresult.payload])
3589

    
3590
      for name in utils.NiceSort(rows.keys()):
3591
        row = rows[name]
3592

    
3593
        out = []
3594

    
3595
        for field in self.op.output_fields:
3596
          if field == constants.SF_NODE:
3597
            val = node
3598
          elif field == constants.SF_TYPE:
3599
            val = self.op.storage_type
3600
          elif field in field_idx:
3601
            val = row[field_idx[field]]
3602
          else:
3603
            raise errors.ParameterError(field)
3604

    
3605
          out.append(val)
3606

    
3607
        result.append(out)
3608

    
3609
    return result
3610

    
3611

    
3612
class LUModifyNodeStorage(NoHooksLU):
3613
  """Logical unit for modifying a storage volume on a node.
3614

3615
  """
3616
  _OP_PARAMS = [
3617
    _PNodeName,
3618
    ("storage_type", _NoDefault, _CheckStorageType),
3619
    ("name", _NoDefault, _TNonEmptyString),
3620
    ("changes", _NoDefault, _TDict),
3621
    ]
3622
  REQ_BGL = False
3623

    
3624
  def CheckArguments(self):
3625
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3626

    
3627
    storage_type = self.op.storage_type
3628

    
3629
    try:
3630
      modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
3631
    except KeyError:
3632
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
3633
                                 " modified" % storage_type,
3634
                                 errors.ECODE_INVAL)
3635

    
3636
    diff = set(self.op.changes.keys()) - modifiable
3637
    if diff:
3638
      raise errors.OpPrereqError("The following fields can not be modified for"
3639
                                 " storage units of type '%s': %r" %
3640
                                 (storage_type, list(diff)),
3641
                                 errors.ECODE_INVAL)
3642

    
3643
  def ExpandNames(self):
3644
    self.needed_locks = {
3645
      locking.LEVEL_NODE: self.op.node_name,
3646
      }
3647

    
3648
  def Exec(self, feedback_fn):
3649
    """Computes the list of nodes and their attributes.
3650

3651
    """
3652
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3653
    result = self.rpc.call_storage_modify(self.op.node_name,
3654
                                          self.op.storage_type, st_args,
3655
                                          self.op.name, self.op.changes)
3656
    result.Raise("Failed to modify storage unit '%s' on %s" %
3657
                 (self.op.name, self.op.node_name))
3658

    
3659

    
3660
class LUAddNode(LogicalUnit):
3661
  """Logical unit for adding node to the cluster.
3662

3663
  """
3664
  HPATH = "node-add"
3665
  HTYPE = constants.HTYPE_NODE
3666
  _OP_PARAMS = [
3667
    _PNodeName,
3668
    ("primary_ip", None, _NoType),
3669
    ("secondary_ip", None, _TMaybeString),
3670
    ("readd", False, _TBool),
3671
    ("nodegroup", None, _TMaybeString)
3672
    ]
3673

    
3674
  def CheckArguments(self):
3675
    self.primary_ip_family = self.cfg.GetPrimaryIPFamily()
3676
    # validate/normalize the node name
3677
    self.hostname = netutils.GetHostname(name=self.op.node_name,
3678
                                         family=self.primary_ip_family)
3679
    self.op.node_name = self.hostname.name
3680
    if self.op.readd and self.op.nodegroup:
3681
      raise errors.OpPrereqError("Cannot pass a nodegroup when a node is"
3682
                                 " being readded", errors.ECODE_INVAL)
3683

    
3684
  def BuildHooksEnv(self):
3685
    """Build hooks env.
3686

3687
    This will run on all nodes before, and on all nodes + the new node after.
3688

3689
    """
3690
    env = {
3691
      "OP_TARGET": self.op.node_name,
3692
      "NODE_NAME": self.op.node_name,
3693
      "NODE_PIP": self.op.primary_ip,
3694
      "NODE_SIP": self.op.secondary_ip,
3695
      }
3696
    nodes_0 = self.cfg.GetNodeList()
3697
    nodes_1 = nodes_0 + [self.op.node_name, ]
3698
    return env, nodes_0, nodes_1
3699

    
3700
  def CheckPrereq(self):
3701
    """Check prerequisites.
3702

3703
    This checks:
3704
     - the new node is not already in the config
3705
     - it is resolvable
3706
     - its parameters (single/dual homed) matches the cluster
3707

3708
    Any errors are signaled by raising errors.OpPrereqError.
3709

3710
    """
3711
    cfg = self.cfg
3712
    hostname = self.hostname
3713
    node = hostname.name
3714
    primary_ip = self.op.primary_ip = hostname.ip
3715
    if self.op.secondary_ip is None:
3716
      if self.primary_ip_family == netutils.IP6Address.family:
3717
        raise errors.OpPrereqError("When using a IPv6 primary address, a valid"
3718
                                   " IPv4 address must be given as secondary",
3719
                                   errors.ECODE_INVAL)
3720
      self.op.secondary_ip = primary_ip
3721

    
3722
    secondary_ip = self.op.secondary_ip
3723
    if not netutils.IP4Address.IsValid(secondary_ip):
3724
      raise errors.OpPrereqError("Secondary IP (%s) needs to be a valid IPv4"
3725
                                 " address" % secondary_ip, errors.ECODE_INVAL)
3726

    
3727
    node_list = cfg.GetNodeList()
3728
    if not self.op.readd and node in node_list:
3729
      raise errors.OpPrereqError("Node %s is already in the configuration" %
3730
                                 node, errors.ECODE_EXISTS)
3731
    elif self.op.readd and node not in node_list:
3732
      raise errors.OpPrereqError("Node %s is not in the configuration" % node,
3733
                                 errors.ECODE_NOENT)
3734

    
3735
    self.changed_primary_ip = False
3736

    
3737
    for existing_node_name in node_list:
3738
      existing_node = cfg.GetNodeInfo(existing_node_name)
3739

    
3740
      if self.op.readd and node == existing_node_name:
3741
        if existing_node.secondary_ip != secondary_ip:
3742
          raise errors.OpPrereqError("Readded node doesn't have the same IP"
3743
                                     " address configuration as before",
3744
                                     errors.ECODE_INVAL)
3745
        if existing_node.primary_ip != primary_ip:
3746
          self.changed_primary_ip = True
3747

    
3748
        continue
3749

    
3750
      if (existing_node.primary_ip == primary_ip or
3751
          existing_node.secondary_ip == primary_ip or
3752
          existing_node.primary_ip == secondary_ip or
3753
          existing_node.secondary_ip == secondary_ip):
3754
        raise errors.OpPrereqError("New node ip address(es) conflict with"
3755
                                   " existing node %s" % existing_node.name,
3756
                                   errors.ECODE_NOTUNIQUE)
3757

    
3758
    # check that the type of the node (single versus dual homed) is the
3759
    # same as for the master
3760
    myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
3761
    master_singlehomed = myself.secondary_ip == myself.primary_ip
3762
    newbie_singlehomed = secondary_ip == primary_ip
3763
    if master_singlehomed != newbie_singlehomed:
3764
      if master_singlehomed:
3765
        raise errors.OpPrereqError("The master has no private ip but the"
3766
                                   " new node has one",
3767
                                   errors.ECODE_INVAL)
3768
      else:
3769
        raise errors.OpPrereqError("The master has a private ip but the"
3770
                                   " new node doesn't have one",
3771
                                   errors.ECODE_INVAL)
3772

    
3773
    # checks reachability
3774
    if not netutils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
3775
      raise errors.OpPrereqError("Node not reachable by ping",
3776
                                 errors.ECODE_ENVIRON)
3777

    
3778
    if not newbie_singlehomed:
3779
      # check reachability from my secondary ip to newbie's secondary ip
3780
      if not netutils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
3781
                           source=myself.secondary_ip):
3782
        raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
3783
                                   " based ping to noded port",
3784
                                   errors.ECODE_ENVIRON)
3785

    
3786
    if self.op.readd:
3787
      exceptions = [node]
3788
    else:
3789
      exceptions = []
3790

    
3791
    self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
3792

    
3793
    if self.op.readd:
3794
      self.new_node = self.cfg.GetNodeInfo(node)
3795
      assert self.new_node is not None, "Can't retrieve locked node %s" % node
3796
    else:
3797
      nodegroup = cfg.LookupNodeGroup(self.op.nodegroup)
3798
      self.new_node = objects.Node(name=node,
3799
                                   primary_ip=primary_ip,
3800
                                   secondary_ip=secondary_ip,
3801
                                   master_candidate=self.master_candidate,
3802
                                   offline=False, drained=False,
3803
                                   nodegroup=nodegroup)
3804

    
3805
  def Exec(self, feedback_fn):
3806
    """Adds the new node to the cluster.
3807

3808
    """
3809
    new_node = self.new_node
3810
    node = new_node.name
3811

    
3812
    # for re-adds, reset the offline/drained/master-candidate flags;
3813
    # we need to reset here, otherwise offline would prevent RPC calls
3814
    # later in the procedure; this also means that if the re-add
3815
    # fails, we are left with a non-offlined, broken node
3816
    if self.op.readd:
3817
      new_node.drained = new_node.offline = False # pylint: disable-msg=W0201
3818
      self.LogInfo("Readding a node, the offline/drained flags were reset")
3819
      # if we demote the node, we do cleanup later in the procedure
3820
      new_node.master_candidate = self.master_candidate
3821
      if self.changed_primary_ip:
3822
        new_node.primary_ip = self.op.primary_ip
3823

    
3824
    # notify the user about any possible mc promotion
3825
    if new_node.master_candidate:
3826
      self.LogInfo("Node will be a master candidate")
3827

    
3828
    # check connectivity
3829
    result = self.rpc.call_version([node])[node]
3830
    result.Raise("Can't get version information from node %s" % node)
3831
    if constants.PROTOCOL_VERSION == result.payload:
3832
      logging.info("Communication to node %s fine, sw version %s match",
3833
                   node, result.payload)
3834
    else:
3835
      raise errors.OpExecError("Version mismatch master version %s,"
3836
                               " node version %s" %
3837
                               (constants.PROTOCOL_VERSION, result.payload))
3838

    
3839
    # Add node to our /etc/hosts, and add key to known_hosts
3840
    if self.cfg.GetClusterInfo().modify_etc_hosts:
3841
      master_node = self.cfg.GetMasterNode()
3842
      result = self.rpc.call_etc_hosts_modify(master_node,
3843
                                              constants.ETC_HOSTS_ADD,
3844
                                              self.hostname.name,
3845
                                              self.hostname.ip)
3846
      result.Raise("Can't update hosts file with new host data")
3847

    
3848
    if new_node.secondary_ip != new_node.primary_ip:
3849
      result = self.rpc.call_node_has_ip_address(new_node.name,
3850
                                                 new_node.secondary_ip)
3851
      result.Raise("Failure checking secondary ip on node %s" % new_node.name,
3852
                   prereq=True, ecode=errors.ECODE_ENVIRON)
3853
      if not result.payload:
3854
        raise errors.OpExecError("Node claims it doesn't have the secondary ip"
3855
                                 " you gave (%s). Please fix and re-run this"
3856
                                 " command." % new_node.secondary_ip)
3857

    
3858
    node_verify_list = [self.cfg.GetMasterNode()]
3859
    node_verify_param = {
3860
      constants.NV_NODELIST: [node],
3861
      # TODO: do a node-net-test as well?
3862
    }
3863

    
3864
    result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
3865
                                       self.cfg.GetClusterName())
3866
    for verifier in node_verify_list:
3867
      result[verifier].Raise("Cannot communicate with node %s" % verifier)
3868
      nl_payload = result[verifier].payload[constants.NV_NODELIST]
3869
      if nl_payload:
3870
        for failed in nl_payload:
3871
          feedback_fn("ssh/hostname verification failed"
3872
                      " (checking from %s): %s" %
3873
                      (verifier, nl_payload[failed]))
3874
        raise errors.OpExecError("ssh/hostname verification failed.")
3875

    
3876
    if self.op.readd:
3877
      _RedistributeAncillaryFiles(self)
3878
      self.context.ReaddNode(new_node)
3879
      # make sure we redistribute the config
3880
      self.cfg.Update(new_node, feedback_fn)
3881
      # and make sure the new node will not have old files around
3882
      if not new_node.master_candidate:
3883
        result = self.rpc.call_node_demote_from_mc(new_node.name)
3884
        msg = result.fail_msg
3885
        if msg:
3886
          self.LogWarning("Node failed to demote itself from master"
3887
                          " candidate status: %s" % msg)
3888
    else:
3889
      _RedistributeAncillaryFiles(self, additional_nodes=[node])
3890
      self.context.AddNode(new_node, self.proc.GetECId())
3891

    
3892

    
3893
class LUSetNodeParams(LogicalUnit):
3894
  """Modifies the parameters of a node.
3895

3896
  """
3897
  HPATH = "node-modify"
3898
  HTYPE = constants.HTYPE_NODE
3899
  _OP_PARAMS = [
3900
    _PNodeName,
3901
    ("master_candidate", None, _TMaybeBool),
3902
    ("offline", None, _TMaybeBool),
3903
    ("drained", None, _TMaybeBool),
3904
    ("auto_promote", False, _TBool),
3905
    _PForce,
3906
    ]
3907
  REQ_BGL = False
3908

    
3909
  def CheckArguments(self):
3910
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3911
    all_mods = [self.op.offline, self.op.master_candidate, self.op.drained]
3912
    if all_mods.count(None) == 3:
3913
      raise errors.OpPrereqError("Please pass at least one modification",
3914
                                 errors.ECODE_INVAL)
3915
    if all_mods.count(True) > 1:
3916
      raise errors.OpPrereqError("Can't set the node into more than one"
3917
                                 " state at the same time",
3918
                                 errors.ECODE_INVAL)
3919

    
3920
    # Boolean value that tells us whether we're offlining or draining the node
3921
    self.offline_or_drain = (self.op.offline == True or
3922
                             self.op.drained == True)
3923
    self.deoffline_or_drain = (self.op.offline == False or
3924
                               self.op.drained == False)
3925
    self.might_demote = (self.op.master_candidate == False or
3926
                         self.offline_or_drain)
3927

    
3928
    self.lock_all = self.op.auto_promote and self.might_demote
3929

    
3930

    
3931
  def ExpandNames(self):
3932
    if self.lock_all:
3933
      self.needed_locks = {locking.LEVEL_NODE: locking.ALL_SET}
3934
    else:
3935
      self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
3936

    
3937
  def BuildHooksEnv(self):
3938
    """Build hooks env.
3939

3940
    This runs on the master node.
3941

3942
    """
3943
    env = {
3944
      "OP_TARGET": self.op.node_name,
3945
      "MASTER_CANDIDATE": str(self.op.master_candidate),
3946
      "OFFLINE": str(self.op.offline),
3947
      "DRAINED": str(self.op.drained),
3948
      }
3949
    nl = [self.cfg.GetMasterNode(),
3950
          self.op.node_name]
3951
    return env, nl, nl
3952

    
3953
  def CheckPrereq(self):
3954
    """Check prerequisites.
3955

3956
    This only checks the instance list against the existing names.
3957

3958
    """
3959
    node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
3960

    
3961
    if (self.op.master_candidate is not None or
3962
        self.op.drained is not None or
3963
        self.op.offline is not None):
3964
      # we can't change the master's node flags
3965
      if self.op.node_name == self.cfg.GetMasterNode():
3966
        raise errors.OpPrereqError("The master role can be changed"
3967
                                   " only via master-failover",
3968
                                   errors.ECODE_INVAL)
3969

    
3970

    
3971
    if node.master_candidate and self.might_demote and not self.lock_all:
3972
      assert not self.op.auto_promote, "auto-promote set but lock_all not"
3973
      # check if after removing the current node, we're missing master
3974
      # candidates
3975
      (mc_remaining, mc_should, _) = \
3976
          self.cfg.GetMasterCandidateStats(exceptions=[node.name])
3977
      if mc_remaining < mc_should:
3978
        raise errors.OpPrereqError("Not enough master candidates, please"
3979
                                   " pass auto_promote to allow promotion",
3980
                                   errors.ECODE_INVAL)
3981

    
3982
    if (self.op.master_candidate == True and
3983
        ((node.offline and not self.op.offline == False) or
3984
         (node.drained and not self.op.drained == False))):
3985
      raise errors.OpPrereqError("Node '%s' is offline or drained, can't set"
3986
                                 " to master_candidate" % node.name,
3987
                                 errors.ECODE_INVAL)
3988

    
3989
    # If we're being deofflined/drained, we'll MC ourself if needed
3990
    if (self.deoffline_or_drain and not self.offline_or_drain and not
3991
        self.op.master_candidate == True and not node.master_candidate):
3992
      self.op.master_candidate = _DecideSelfPromotion(self)
3993
      if self.op.master_candidate:
3994
        self.LogInfo("Autopromoting node to master candidate")
3995

    
3996
    return
3997

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

4001
    """
4002
    node = self.node
4003

    
4004
    result = []
4005
    changed_mc = False
4006

    
4007
    if self.op.offline is not None:
4008
      node.offline = self.op.offline
4009
      result.append(("offline", str(self.op.offline)))
4010
      if self.op.offline == True:
4011
        if node.master_candidate:
4012
          node.master_candidate = False
4013
          changed_mc = True
4014
          result.append(("master_candidate", "auto-demotion due to offline"))
4015
        if node.drained:
4016
          node.drained = False
4017
          result.append(("drained", "clear drained status due to offline"))
4018

    
4019
    if self.op.master_candidate is not None:
4020
      node.master_candidate = self.op.master_candidate
4021
      changed_mc = True
4022
      result.append(("master_candidate", str(self.op.master_candidate)))
4023
      if self.op.master_candidate == False:
4024
        rrc = self.rpc.call_node_demote_from_mc(node.name)
4025
        msg = rrc.fail_msg
4026
        if msg:
4027
          self.LogWarning("Node failed to demote itself: %s" % msg)
4028

    
4029
    if self.op.drained is not None:
4030
      node.drained = self.op.drained
4031
      result.append(("drained", str(self.op.drained)))
4032
      if self.op.drained == True:
4033
        if node.master_candidate:
4034
          node.master_candidate = False
4035
          changed_mc = True
4036
          result.append(("master_candidate", "auto-demotion due to drain"))
4037
          rrc = self.rpc.call_node_demote_from_mc(node.name)
4038
          msg = rrc.fail_msg
4039
          if msg:
4040
            self.LogWarning("Node failed to demote itself: %s" % msg)
4041
        if node.offline:
4042
          node.offline = False
4043
          result.append(("offline", "clear offline status due to drain"))
4044

    
4045
    # we locked all nodes, we adjust the CP before updating this node
4046
    if self.lock_all:
4047
      _AdjustCandidatePool(self, [node.name])
4048

    
4049
    # this will trigger configuration file update, if needed
4050
    self.cfg.Update(node, feedback_fn)
4051

    
4052
    # this will trigger job queue propagation or cleanup
4053
    if changed_mc:
4054
      self.context.ReaddNode(node)
4055

    
4056
    return result
4057

    
4058

    
4059
class LUPowercycleNode(NoHooksLU):
4060
  """Powercycles a node.
4061

4062
  """
4063
  _OP_PARAMS = [
4064
    _PNodeName,
4065
    _PForce,
4066
    ]
4067
  REQ_BGL = False
4068

    
4069
  def CheckArguments(self):
4070
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4071
    if self.op.node_name == self.cfg.GetMasterNode() and not self.op.force:
4072
      raise errors.OpPrereqError("The node is the master and the force"
4073
                                 " parameter was not set",
4074
                                 errors.ECODE_INVAL)
4075

    
4076
  def ExpandNames(self):
4077
    """Locking for PowercycleNode.
4078

4079
    This is a last-resort option and shouldn't block on other
4080
    jobs. Therefore, we grab no locks.
4081

4082
    """
4083
    self.needed_locks = {}
4084

    
4085
  def Exec(self, feedback_fn):
4086
    """Reboots a node.
4087

4088
    """
4089
    result = self.rpc.call_node_powercycle(self.op.node_name,
4090
                                           self.cfg.GetHypervisorType())
4091
    result.Raise("Failed to schedule the reboot")
4092
    return result.payload
4093

    
4094

    
4095
class LUQueryClusterInfo(NoHooksLU):
4096
  """Query cluster configuration.
4097

4098
  """
4099
  REQ_BGL = False
4100

    
4101
  def ExpandNames(self):
4102
    self.needed_locks = {}
4103

    
4104
  def Exec(self, feedback_fn):
4105
    """Return cluster config.
4106

4107
    """
4108
    cluster = self.cfg.GetClusterInfo()
4109
    os_hvp = {}
4110

    
4111
    # Filter just for enabled hypervisors
4112
    for os_name, hv_dict in cluster.os_hvp.items():
4113
      os_hvp[os_name] = {}
4114
      for hv_name, hv_params in hv_dict.items():
4115
        if hv_name in cluster.enabled_hypervisors:
4116
          os_hvp[os_name][hv_name] = hv_params
4117

    
4118
    # Convert ip_family to ip_version
4119
    primary_ip_version = constants.IP4_VERSION
4120
    if cluster.primary_ip_family == netutils.IP6Address.family:
4121
      primary_ip_version = constants.IP6_VERSION
4122

    
4123
    result = {
4124
      "software_version": constants.RELEASE_VERSION,
4125
      "protocol_version": constants.PROTOCOL_VERSION,
4126
      "config_version": constants.CONFIG_VERSION,
4127
      "os_api_version": max(constants.OS_API_VERSIONS),
4128
      "export_version": constants.EXPORT_VERSION,
4129
      "architecture": (platform.architecture()[0], platform.machine()),
4130
      "name": cluster.cluster_name,
4131
      "master": cluster.master_node,
4132
      "default_hypervisor": cluster.enabled_hypervisors[0],
4133
      "enabled_hypervisors": cluster.enabled_hypervisors,
4134
      "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
4135
                        for hypervisor_name in cluster.enabled_hypervisors]),
4136
      "os_hvp": os_hvp,
4137
      "beparams": cluster.beparams,
4138
      "osparams": cluster.osparams,
4139
      "nicparams": cluster.nicparams,
4140
      "candidate_pool_size": cluster.candidate_pool_size,
4141
      "master_netdev": cluster.master_netdev,
4142
      "volume_group_name": cluster.volume_group_name,
4143
      "drbd_usermode_helper": cluster.drbd_usermode_helper,
4144
      "file_storage_dir": cluster.file_storage_dir,
4145
      "maintain_node_health": cluster.maintain_node_health,
4146
      "ctime": cluster.ctime,
4147
      "mtime": cluster.mtime,
4148
      "uuid": cluster.uuid,
4149
      "tags": list(cluster.GetTags()),
4150
      "uid_pool": cluster.uid_pool,
4151
      "default_iallocator": cluster.default_iallocator,
4152
      "reserved_lvs": cluster.reserved_lvs,
4153
      "primary_ip_version": primary_ip_version,
4154
      }
4155

    
4156
    return result
4157

    
4158

    
4159
class LUQueryConfigValues(NoHooksLU):
4160
  """Return configuration values.
4161

4162
  """
4163
  _OP_PARAMS = [_POutputFields]
4164
  REQ_BGL = False
4165
  _FIELDS_DYNAMIC = utils.FieldSet()
4166
  _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag",
4167
                                  "watcher_pause")
4168

    
4169
  def CheckArguments(self):
4170
    _CheckOutputFields(static=self._FIELDS_STATIC,
4171
                       dynamic=self._FIELDS_DYNAMIC,
4172
                       selected=self.op.output_fields)
4173

    
4174
  def ExpandNames(self):
4175
    self.needed_locks = {}
4176

    
4177
  def Exec(self, feedback_fn):
4178
    """Dump a representation of the cluster config to the standard output.
4179

4180
    """
4181
    values = []
4182
    for field in self.op.output_fields:
4183
      if field == "cluster_name":
4184
        entry = self.cfg.GetClusterName()
4185
      elif field == "master_node":
4186
        entry = self.cfg.GetMasterNode()
4187
      elif field == "drain_flag":
4188
        entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
4189
      elif field == "watcher_pause":
4190
        entry = utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
4191
      else:
4192
        raise errors.ParameterError(field)
4193
      values.append(entry)
4194
    return values
4195

    
4196

    
4197
class LUActivateInstanceDisks(NoHooksLU):
4198
  """Bring up an instance's disks.
4199

4200
  """
4201
  _OP_PARAMS = [
4202
    _PInstanceName,
4203
    ("ignore_size", False, _TBool),
4204
    ]
4205
  REQ_BGL = False
4206

    
4207
  def ExpandNames(self):
4208
    self._ExpandAndLockInstance()
4209
    self.needed_locks[locking.LEVEL_NODE] = []
4210
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4211

    
4212
  def DeclareLocks(self, level):
4213
    if level == locking.LEVEL_NODE:
4214
      self._LockInstancesNodes()
4215

    
4216
  def CheckPrereq(self):
4217
    """Check prerequisites.
4218

4219
    This checks that the instance is in the cluster.
4220

4221
    """
4222
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4223
    assert self.instance is not None, \
4224
      "Cannot retrieve locked instance %s" % self.op.instance_name
4225
    _CheckNodeOnline(self, self.instance.primary_node)
4226

    
4227
  def Exec(self, feedback_fn):
4228
    """Activate the disks.
4229

4230
    """
4231
    disks_ok, disks_info = \
4232
              _AssembleInstanceDisks(self, self.instance,
4233
                                     ignore_size=self.op.ignore_size)
4234
    if not disks_ok:
4235
      raise errors.OpExecError("Cannot activate block devices")
4236

    
4237
    return disks_info
4238

    
4239

    
4240
def _AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False,
4241
                           ignore_size=False):
4242
  """Prepare the block devices for an instance.
4243

4244
  This sets up the block devices on all nodes.
4245

4246
  @type lu: L{LogicalUnit}
4247
  @param lu: the logical unit on whose behalf we execute
4248
  @type instance: L{objects.Instance}
4249
  @param instance: the instance for whose disks we assemble
4250
  @type disks: list of L{objects.Disk} or None
4251
  @param disks: which disks to assemble (or all, if None)
4252
  @type ignore_secondaries: boolean
4253
  @param ignore_secondaries: if true, errors on secondary nodes
4254
      won't result in an error return from the function
4255
  @type ignore_size: boolean
4256
  @param ignore_size: if true, the current known size of the disk
4257
      will not be used during the disk activation, useful for cases
4258
      when the size is wrong
4259
  @return: False if the operation failed, otherwise a list of
4260
      (host, instance_visible_name, node_visible_name)
4261
      with the mapping from node devices to instance devices
4262

4263
  """
4264
  device_info = []
4265
  disks_ok = True
4266
  iname = instance.name
4267
  disks = _ExpandCheckDisks(instance, disks)
4268

    
4269
  # With the two passes mechanism we try to reduce the window of
4270
  # opportunity for the race condition of switching DRBD to primary
4271
  # before handshaking occured, but we do not eliminate it
4272

    
4273
  # The proper fix would be to wait (with some limits) until the
4274
  # connection has been made and drbd transitions from WFConnection
4275
  # into any other network-connected state (Connected, SyncTarget,
4276
  # SyncSource, etc.)
4277

    
4278
  # 1st pass, assemble on all nodes in secondary mode
4279
  for inst_disk in disks:
4280
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
4281
      if ignore_size:
4282
        node_disk = node_disk.Copy()
4283
        node_disk.UnsetSize()
4284
      lu.cfg.SetDiskID(node_disk, node)
4285
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
4286
      msg = result.fail_msg
4287
      if msg:
4288
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
4289
                           " (is_primary=False, pass=1): %s",
4290
                           inst_disk.iv_name, node, msg)
4291
        if not ignore_secondaries:
4292
          disks_ok = False
4293

    
4294
  # FIXME: race condition on drbd migration to primary
4295

    
4296
  # 2nd pass, do only the primary node
4297
  for inst_disk in disks:
4298
    dev_path = None
4299

    
4300
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
4301
      if node != instance.primary_node:
4302
        continue
4303
      if ignore_size:
4304
        node_disk = node_disk.Copy()
4305
        node_disk.UnsetSize()
4306
      lu.cfg.SetDiskID(node_disk, node)
4307
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
4308
      msg = result.fail_msg
4309
      if msg:
4310
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
4311
                           " (is_primary=True, pass=2): %s",
4312
                           inst_disk.iv_name, node, msg)
4313
        disks_ok = False
4314
      else:
4315
        dev_path = result.payload
4316

    
4317
    device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
4318

    
4319
  # leave the disks configured for the primary node
4320
  # this is a workaround that would be fixed better by
4321
  # improving the logical/physical id handling
4322
  for disk in disks:
4323
    lu.cfg.SetDiskID(disk, instance.primary_node)
4324

    
4325
  return disks_ok, device_info
4326

    
4327

    
4328
def _StartInstanceDisks(lu, instance, force):
4329
  """Start the disks of an instance.
4330

4331
  """
4332
  disks_ok, _ = _AssembleInstanceDisks(lu, instance,
4333
                                           ignore_secondaries=force)
4334
  if not disks_ok:
4335
    _ShutdownInstanceDisks(lu, instance)
4336
    if force is not None and not force:
4337
      lu.proc.LogWarning("", hint="If the message above refers to a"
4338
                         " secondary node,"
4339
                         " you can retry the operation using '--force'.")
4340
    raise errors.OpExecError("Disk consistency error")
4341

    
4342

    
4343
class LUDeactivateInstanceDisks(NoHooksLU):
4344
  """Shutdown an instance's disks.
4345

4346
  """
4347
  _OP_PARAMS = [
4348
    _PInstanceName,
4349
    ]
4350
  REQ_BGL = False
4351

    
4352
  def ExpandNames(self):
4353
    self._ExpandAndLockInstance()
4354
    self.needed_locks[locking.LEVEL_NODE] = []
4355
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4356

    
4357
  def DeclareLocks(self, level):
4358
    if level == locking.LEVEL_NODE:
4359
      self._LockInstancesNodes()
4360

    
4361
  def CheckPrereq(self):
4362
    """Check prerequisites.
4363

4364
    This checks that the instance is in the cluster.
4365

4366
    """
4367
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4368
    assert self.instance is not None, \
4369
      "Cannot retrieve locked instance %s" % self.op.instance_name
4370

    
4371
  def Exec(self, feedback_fn):
4372
    """Deactivate the disks
4373

4374
    """
4375
    instance = self.instance
4376
    _SafeShutdownInstanceDisks(self, instance)
4377

    
4378

    
4379
def _SafeShutdownInstanceDisks(lu, instance, disks=None):
4380
  """Shutdown block devices of an instance.
4381

4382
  This function checks if an instance is running, before calling
4383
  _ShutdownInstanceDisks.
4384

4385
  """
4386
  _CheckInstanceDown(lu, instance, "cannot shutdown disks")
4387
  _ShutdownInstanceDisks(lu, instance, disks=disks)
4388

    
4389

    
4390
def _ExpandCheckDisks(instance, disks):
4391
  """Return the instance disks selected by the disks list
4392

4393
  @type disks: list of L{objects.Disk} or None
4394
  @param disks: selected disks
4395
  @rtype: list of L{objects.Disk}
4396
  @return: selected instance disks to act on
4397

4398
  """
4399
  if disks is None:
4400
    return instance.disks
4401
  else:
4402
    if not set(disks).issubset(instance.disks):
4403
      raise errors.ProgrammerError("Can only act on disks belonging to the"
4404
                                   " target instance")
4405
    return disks
4406

    
4407

    
4408
def _ShutdownInstanceDisks(lu, instance, disks=None, ignore_primary=False):
4409
  """Shutdown block devices of an instance.
4410

4411
  This does the shutdown on all nodes of the instance.
4412

4413
  If the ignore_primary is false, errors on the primary node are
4414
  ignored.
4415

4416
  """
4417
  all_result = True
4418
  disks = _ExpandCheckDisks(instance, disks)
4419

    
4420
  for disk in disks:
4421
    for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
4422
      lu.cfg.SetDiskID(top_disk, node)
4423
      result = lu.rpc.call_blockdev_shutdown(node, top_disk)
4424
      msg = result.fail_msg
4425
      if msg:
4426
        lu.LogWarning("Could not shutdown block device %s on node %s: %s",
4427
                      disk.iv_name, node, msg)
4428
        if not ignore_primary or node != instance.primary_node:
4429
          all_result = False
4430
  return all_result
4431

    
4432

    
4433
def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
4434
  """Checks if a node has enough free memory.
4435

4436
  This function check if a given node has the needed amount of free
4437
  memory. In case the node has less memory or we cannot get the
4438
  information from the node, this function raise an OpPrereqError
4439
  exception.
4440

4441
  @type lu: C{LogicalUnit}
4442
  @param lu: a logical unit from which we get configuration data
4443
  @type node: C{str}
4444
  @param node: the node to check
4445
  @type reason: C{str}
4446
  @param reason: string to use in the error message
4447
  @type requested: C{int}
4448
  @param requested: the amount of memory in MiB to check for
4449
  @type hypervisor_name: C{str}
4450
  @param hypervisor_name: the hypervisor to ask for memory stats
4451
  @raise errors.OpPrereqError: if the node doesn't have enough memory, or
4452
      we cannot check the node
4453

4454
  """
4455
  nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
4456
  nodeinfo[node].Raise("Can't get data from node %s" % node,
4457
                       prereq=True, ecode=errors.ECODE_ENVIRON)
4458
  free_mem = nodeinfo[node].payload.get('memory_free', None)
4459
  if not isinstance(free_mem, int):
4460
    raise errors.OpPrereqError("Can't compute free memory on node %s, result"
4461
                               " was '%s'" % (node, free_mem),
4462
                               errors.ECODE_ENVIRON)
4463
  if requested > free_mem:
4464
    raise errors.OpPrereqError("Not enough memory on node %s for %s:"
4465
                               " needed %s MiB, available %s MiB" %
4466
                               (node, reason, requested, free_mem),
4467
                               errors.ECODE_NORES)
4468

    
4469

    
4470
def _CheckNodesFreeDisk(lu, nodenames, requested):
4471
  """Checks if nodes have enough free disk space in the default VG.
4472

4473
  This function check if all given nodes have the needed amount of
4474
  free disk. In case any node has less disk or we cannot get the
4475
  information from the node, this function raise an OpPrereqError
4476
  exception.
4477

4478
  @type lu: C{LogicalUnit}
4479
  @param lu: a logical unit from which we get configuration data
4480
  @type nodenames: C{list}
4481
  @param nodenames: the list of node names to check
4482
  @type requested: C{int}
4483
  @param requested: the amount of disk in MiB to check for
4484
  @raise errors.OpPrereqError: if the node doesn't have enough disk, or
4485
      we cannot check the node
4486

4487
  """
4488
  nodeinfo = lu.rpc.call_node_info(nodenames, lu.cfg.GetVGName(),
4489
                                   lu.cfg.GetHypervisorType())
4490
  for node in nodenames:
4491
    info = nodeinfo[node]
4492
    info.Raise("Cannot get current information from node %s" % node,
4493
               prereq=True, ecode=errors.ECODE_ENVIRON)
4494
    vg_free = info.payload.get("vg_free", None)
4495
    if not isinstance(vg_free, int):
4496
      raise errors.OpPrereqError("Can't compute free disk space on node %s,"
4497
                                 " result was '%s'" % (node, vg_free),
4498
                                 errors.ECODE_ENVIRON)
4499
    if requested > vg_free:
4500
      raise errors.OpPrereqError("Not enough disk space on target node %s:"
4501
                                 " required %d MiB, available %d MiB" %
4502
                                 (node, requested, vg_free),
4503
                                 errors.ECODE_NORES)
4504

    
4505

    
4506
class LUStartupInstance(LogicalUnit):
4507
  """Starts an instance.
4508

4509
  """
4510
  HPATH = "instance-start"
4511
  HTYPE = constants.HTYPE_INSTANCE
4512
  _OP_PARAMS = [
4513
    _PInstanceName,
4514
    _PForce,
4515
    ("hvparams", _EmptyDict, _TDict),
4516
    ("beparams", _EmptyDict, _TDict),
4517
    ]
4518
  REQ_BGL = False
4519

    
4520
  def CheckArguments(self):
4521
    # extra beparams
4522
    if self.op.beparams:
4523
      # fill the beparams dict
4524
      utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
4525

    
4526
  def ExpandNames(self):
4527
    self._ExpandAndLockInstance()
4528

    
4529
  def BuildHooksEnv(self):
4530
    """Build hooks env.
4531

4532
    This runs on master, primary and secondary nodes of the instance.
4533

4534
    """
4535
    env = {
4536
      "FORCE": self.op.force,
4537
      }
4538
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4539
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4540
    return env, nl, nl
4541

    
4542
  def CheckPrereq(self):
4543
    """Check prerequisites.
4544

4545
    This checks that the instance is in the cluster.
4546

4547
    """
4548
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4549
    assert self.instance is not None, \
4550
      "Cannot retrieve locked instance %s" % self.op.instance_name
4551

    
4552
    # extra hvparams
4553
    if self.op.hvparams:
4554
      # check hypervisor parameter syntax (locally)
4555
      cluster = self.cfg.GetClusterInfo()
4556
      utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
4557
      filled_hvp = cluster.FillHV(instance)
4558
      filled_hvp.update(self.op.hvparams)
4559
      hv_type = hypervisor.GetHypervisor(instance.hypervisor)
4560
      hv_type.CheckParameterSyntax(filled_hvp)
4561
      _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
4562

    
4563
    _CheckNodeOnline(self, instance.primary_node)
4564

    
4565
    bep = self.cfg.GetClusterInfo().FillBE(instance)
4566
    # check bridges existence
4567
    _CheckInstanceBridgesExist(self, instance)
4568

    
4569
    remote_info = self.rpc.call_instance_info(instance.primary_node,
4570
                                              instance.name,
4571
                                              instance.hypervisor)
4572
    remote_info.Raise("Error checking node %s" % instance.primary_node,
4573
                      prereq=True, ecode=errors.ECODE_ENVIRON)
4574
    if not remote_info.payload: # not running already
4575
      _CheckNodeFreeMemory(self, instance.primary_node,
4576
                           "starting instance %s" % instance.name,
4577
                           bep[constants.BE_MEMORY], instance.hypervisor)
4578

    
4579
  def Exec(self, feedback_fn):
4580
    """Start the instance.
4581

4582
    """
4583
    instance = self.instance
4584
    force = self.op.force
4585

    
4586
    self.cfg.MarkInstanceUp(instance.name)
4587

    
4588
    node_current = instance.primary_node
4589

    
4590
    _StartInstanceDisks(self, instance, force)
4591

    
4592
    result = self.rpc.call_instance_start(node_current, instance,
4593
                                          self.op.hvparams, self.op.beparams)
4594
    msg = result.fail_msg
4595
    if msg:
4596
      _ShutdownInstanceDisks(self, instance)
4597
      raise errors.OpExecError("Could not start instance: %s" % msg)
4598

    
4599

    
4600
class LURebootInstance(LogicalUnit):
4601
  """Reboot an instance.
4602

4603
  """
4604
  HPATH = "instance-reboot"
4605
  HTYPE = constants.HTYPE_INSTANCE
4606
  _OP_PARAMS = [
4607
    _PInstanceName,
4608
    ("ignore_secondaries", False, _TBool),
4609
    ("reboot_type", _NoDefault, _TElemOf(constants.REBOOT_TYPES)),
4610
    _PShutdownTimeout,
4611
    ]
4612
  REQ_BGL = False
4613

    
4614
  def ExpandNames(self):
4615
    self._ExpandAndLockInstance()
4616

    
4617
  def BuildHooksEnv(self):
4618
    """Build hooks env.
4619

4620
    This runs on master, primary and secondary nodes of the instance.
4621

4622
    """
4623
    env = {
4624
      "IGNORE_SECONDARIES": self.op.ignore_secondaries,
4625
      "REBOOT_TYPE": self.op.reboot_type,
4626
      "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
4627
      }
4628
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4629
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4630
    return env, nl, nl
4631

    
4632
  def CheckPrereq(self):
4633
    """Check prerequisites.
4634

4635
    This checks that the instance is in the cluster.
4636

4637
    """
4638
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4639
    assert self.instance is not None, \
4640
      "Cannot retrieve locked instance %s" % self.op.instance_name
4641

    
4642
    _CheckNodeOnline(self, instance.primary_node)
4643

    
4644
    # check bridges existence
4645
    _CheckInstanceBridgesExist(self, instance)
4646

    
4647
  def Exec(self, feedback_fn):
4648
    """Reboot the instance.
4649

4650
    """
4651
    instance = self.instance
4652
    ignore_secondaries = self.op.ignore_secondaries
4653
    reboot_type = self.op.reboot_type
4654

    
4655
    node_current = instance.primary_node
4656

    
4657
    if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
4658
                       constants.INSTANCE_REBOOT_HARD]:
4659
      for disk in instance.disks:
4660
        self.cfg.SetDiskID(disk, node_current)
4661
      result = self.rpc.call_instance_reboot(node_current, instance,
4662
                                             reboot_type,
4663
                                             self.op.shutdown_timeout)
4664
      result.Raise("Could not reboot instance")
4665
    else:
4666
      result = self.rpc.call_instance_shutdown(node_current, instance,
4667
                                               self.op.shutdown_timeout)
4668
      result.Raise("Could not shutdown instance for full reboot")
4669
      _ShutdownInstanceDisks(self, instance)
4670
      _StartInstanceDisks(self, instance, ignore_secondaries)
4671
      result = self.rpc.call_instance_start(node_current, instance, None, None)
4672
      msg = result.fail_msg
4673
      if msg:
4674
        _ShutdownInstanceDisks(self, instance)
4675
        raise errors.OpExecError("Could not start instance for"
4676
                                 " full reboot: %s" % msg)
4677

    
4678
    self.cfg.MarkInstanceUp(instance.name)
4679

    
4680

    
4681
class LUShutdownInstance(LogicalUnit):
4682
  """Shutdown an instance.
4683

4684
  """
4685
  HPATH = "instance-stop"
4686
  HTYPE = constants.HTYPE_INSTANCE
4687
  _OP_PARAMS = [
4688
    _PInstanceName,
4689
    ("timeout", constants.DEFAULT_SHUTDOWN_TIMEOUT, _TPositiveInt),
4690
    ]
4691
  REQ_BGL = False
4692

    
4693
  def ExpandNames(self):
4694
    self._ExpandAndLockInstance()
4695

    
4696
  def BuildHooksEnv(self):
4697
    """Build hooks env.
4698

4699
    This runs on master, primary and secondary nodes of the instance.
4700

4701
    """
4702
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4703
    env["TIMEOUT"] = self.op.timeout
4704
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4705
    return env, nl, nl
4706

    
4707
  def CheckPrereq(self):
4708
    """Check prerequisites.
4709

4710
    This checks that the instance is in the cluster.
4711

4712
    """
4713
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4714
    assert self.instance is not None, \
4715
      "Cannot retrieve locked instance %s" % self.op.instance_name
4716
    _CheckNodeOnline(self, self.instance.primary_node)
4717

    
4718
  def Exec(self, feedback_fn):
4719
    """Shutdown the instance.
4720

4721
    """
4722
    instance = self.instance
4723
    node_current = instance.primary_node
4724
    timeout = self.op.timeout
4725
    self.cfg.MarkInstanceDown(instance.name)
4726
    result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
4727
    msg = result.fail_msg
4728
    if msg:
4729
      self.proc.LogWarning("Could not shutdown instance: %s" % msg)
4730

    
4731
    _ShutdownInstanceDisks(self, instance)
4732

    
4733

    
4734
class LUReinstallInstance(LogicalUnit):
4735
  """Reinstall an instance.
4736

4737
  """
4738
  HPATH = "instance-reinstall"
4739
  HTYPE = constants.HTYPE_INSTANCE
4740
  _OP_PARAMS = [
4741
    _PInstanceName,
4742
    ("os_type", None, _TMaybeString),
4743
    ("force_variant", False, _TBool),
4744
    ]
4745
  REQ_BGL = False
4746

    
4747
  def ExpandNames(self):
4748
    self._ExpandAndLockInstance()
4749

    
4750
  def BuildHooksEnv(self):
4751
    """Build hooks env.
4752

4753
    This runs on master, primary and secondary nodes of the instance.
4754

4755
    """
4756
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4757
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4758
    return env, nl, nl
4759

    
4760
  def CheckPrereq(self):
4761
    """Check prerequisites.
4762

4763
    This checks that the instance is in the cluster and is not running.
4764

4765
    """
4766
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4767
    assert instance is not None, \
4768
      "Cannot retrieve locked instance %s" % self.op.instance_name
4769
    _CheckNodeOnline(self, instance.primary_node)
4770

    
4771
    if instance.disk_template == constants.DT_DISKLESS:
4772
      raise errors.OpPrereqError("Instance '%s' has no disks" %
4773
                                 self.op.instance_name,
4774
                                 errors.ECODE_INVAL)
4775
    _CheckInstanceDown(self, instance, "cannot reinstall")
4776

    
4777
    if self.op.os_type is not None:
4778
      # OS verification
4779
      pnode = _ExpandNodeName(self.cfg, instance.primary_node)
4780
      _CheckNodeHasOS(self, pnode, self.op.os_type, self.op.force_variant)
4781

    
4782
    self.instance = instance
4783

    
4784
  def Exec(self, feedback_fn):
4785
    """Reinstall the instance.
4786

4787
    """
4788
    inst = self.instance
4789

    
4790
    if self.op.os_type is not None:
4791
      feedback_fn("Changing OS to '%s'..." % self.op.os_type)
4792
      inst.os = self.op.os_type
4793
      self.cfg.Update(inst, feedback_fn)
4794

    
4795
    _StartInstanceDisks(self, inst, None)
4796
    try:
4797
      feedback_fn("Running the instance OS create scripts...")
4798
      # FIXME: pass debug option from opcode to backend
4799
      result = self.rpc.call_instance_os_add(inst.primary_node, inst, True,
4800
                                             self.op.debug_level)
4801
      result.Raise("Could not install OS for instance %s on node %s" %
4802
                   (inst.name, inst.primary_node))
4803
    finally:
4804
      _ShutdownInstanceDisks(self, inst)
4805

    
4806

    
4807
class LURecreateInstanceDisks(LogicalUnit):
4808
  """Recreate an instance's missing disks.
4809

4810
  """
4811
  HPATH = "instance-recreate-disks"
4812
  HTYPE = constants.HTYPE_INSTANCE
4813
  _OP_PARAMS = [
4814
    _PInstanceName,
4815
    ("disks", _EmptyList, _TListOf(_TPositiveInt)),
4816
    ]
4817
  REQ_BGL = False
4818

    
4819
  def ExpandNames(self):
4820
    self._ExpandAndLockInstance()
4821

    
4822
  def BuildHooksEnv(self):
4823
    """Build hooks env.
4824

4825
    This runs on master, primary and secondary nodes of the instance.
4826

4827
    """
4828
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4829
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4830
    return env, nl, nl
4831

    
4832
  def CheckPrereq(self):
4833
    """Check prerequisites.
4834

4835
    This checks that the instance is in the cluster and is not running.
4836

4837
    """
4838
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4839
    assert instance is not None, \
4840
      "Cannot retrieve locked instance %s" % self.op.instance_name
4841
    _CheckNodeOnline(self, instance.primary_node)
4842

    
4843
    if instance.disk_template == constants.DT_DISKLESS:
4844
      raise errors.OpPrereqError("Instance '%s' has no disks" %
4845
                                 self.op.instance_name, errors.ECODE_INVAL)
4846
    _CheckInstanceDown(self, instance, "cannot recreate disks")
4847

    
4848
    if not self.op.disks:
4849
      self.op.disks = range(len(instance.disks))
4850
    else:
4851
      for idx in self.op.disks:
4852
        if idx >= len(instance.disks):
4853
          raise errors.OpPrereqError("Invalid disk index passed '%s'" % idx,
4854
                                     errors.ECODE_INVAL)
4855

    
4856
    self.instance = instance
4857

    
4858
  def Exec(self, feedback_fn):
4859
    """Recreate the disks.
4860

4861
    """
4862
    to_skip = []
4863
    for idx, _ in enumerate(self.instance.disks):
4864
      if idx not in self.op.disks: # disk idx has not been passed in
4865
        to_skip.append(idx)
4866
        continue
4867

    
4868
    _CreateDisks(self, self.instance, to_skip=to_skip)
4869

    
4870

    
4871
class LURenameInstance(LogicalUnit):
4872
  """Rename an instance.
4873

4874
  """
4875
  HPATH = "instance-rename"
4876
  HTYPE = constants.HTYPE_INSTANCE
4877
  _OP_PARAMS = [
4878
    _PInstanceName,
4879
    ("new_name", _NoDefault, _TNonEmptyString),
4880
    ("ip_check", False, _TBool),
4881
    ("name_check", True, _TBool),
4882
    ]
4883

    
4884
  def CheckArguments(self):
4885
    """Check arguments.
4886

4887
    """
4888
    if self.op.ip_check and not self.op.name_check:
4889
      # TODO: make the ip check more flexible and not depend on the name check
4890
      raise errors.OpPrereqError("Cannot do ip check without a name check",
4891
                                 errors.ECODE_INVAL)
4892

    
4893
  def BuildHooksEnv(self):
4894
    """Build hooks env.
4895

4896
    This runs on master, primary and secondary nodes of the instance.
4897

4898
    """
4899
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4900
    env["INSTANCE_NEW_NAME"] = self.op.new_name
4901
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4902
    return env, nl, nl
4903

    
4904
  def CheckPrereq(self):
4905
    """Check prerequisites.
4906

4907
    This checks that the instance is in the cluster and is not running.
4908

4909
    """
4910
    self.op.instance_name = _ExpandInstanceName(self.cfg,
4911
                                                self.op.instance_name)
4912
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4913
    assert instance is not None
4914
    _CheckNodeOnline(self, instance.primary_node)
4915
    _CheckInstanceDown(self, instance, "cannot rename")
4916
    self.instance = instance
4917

    
4918
    new_name = self.op.new_name
4919
    if self.op.name_check:
4920
      hostname = netutils.GetHostname(name=new_name)
4921
      new_name = hostname.name
4922
      if (self.op.ip_check and
4923
          netutils.TcpPing(hostname.ip, constants.DEFAULT_NODED_PORT)):
4924
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
4925
                                   (hostname.ip, new_name),
4926
                                   errors.ECODE_NOTUNIQUE)
4927

    
4928
    instance_list = self.cfg.GetInstanceList()
4929
    if new_name in instance_list:
4930
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
4931
                                 new_name, errors.ECODE_EXISTS)
4932

    
4933
  def Exec(self, feedback_fn):
4934
    """Reinstall the instance.
4935

4936
    """
4937
    inst = self.instance
4938
    old_name = inst.name
4939

    
4940
    if inst.disk_template == constants.DT_FILE:
4941
      old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
4942

    
4943
    self.cfg.RenameInstance(inst.name, self.op.new_name)
4944
    # Change the instance lock. This is definitely safe while we hold the BGL
4945
    self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
4946
    self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
4947

    
4948
    # re-read the instance from the configuration after rename
4949
    inst = self.cfg.GetInstanceInfo(self.op.new_name)
4950

    
4951
    if inst.disk_template == constants.DT_FILE:
4952
      new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
4953
      result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
4954
                                                     old_file_storage_dir,
4955
                                                     new_file_storage_dir)
4956
      result.Raise("Could not rename on node %s directory '%s' to '%s'"
4957
                   " (but the instance has been renamed in Ganeti)" %
4958
                   (inst.primary_node, old_file_storage_dir,
4959
                    new_file_storage_dir))
4960

    
4961
    _StartInstanceDisks(self, inst, None)
4962
    try:
4963
      result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
4964
                                                 old_name, self.op.debug_level)
4965
      msg = result.fail_msg
4966
      if msg:
4967
        msg = ("Could not run OS rename script for instance %s on node %s"
4968
               " (but the instance has been renamed in Ganeti): %s" %
4969
               (inst.name, inst.primary_node, msg))
4970
        self.proc.LogWarning(msg)
4971
    finally:
4972
      _ShutdownInstanceDisks(self, inst)
4973

    
4974
    return inst.name
4975

    
4976

    
4977
class LURemoveInstance(LogicalUnit):
4978
  """Remove an instance.
4979

4980
  """
4981
  HPATH = "instance-remove"
4982
  HTYPE = constants.HTYPE_INSTANCE
4983
  _OP_PARAMS = [
4984
    _PInstanceName,
4985
    ("ignore_failures", False, _TBool),
4986
    _PShutdownTimeout,
4987
    ]
4988
  REQ_BGL = False
4989

    
4990
  def ExpandNames(self):
4991
    self._ExpandAndLockInstance()
4992
    self.needed_locks[locking.LEVEL_NODE] = []
4993
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4994

    
4995
  def DeclareLocks(self, level):
4996
    if level == locking.LEVEL_NODE:
4997
      self._LockInstancesNodes()
4998

    
4999
  def BuildHooksEnv(self):
5000
    """Build hooks env.
5001

5002
    This runs on master, primary and secondary nodes of the instance.
5003

5004
    """
5005
    env = _BuildInstanceHookEnvByObject(self, self.instance)
5006
    env["SHUTDOWN_TIMEOUT"] = self.op.shutdown_timeout
5007
    nl = [self.cfg.GetMasterNode()]
5008
    nl_post = list(self.instance.all_nodes) + nl
5009
    return env, nl, nl_post
5010

    
5011
  def CheckPrereq(self):
5012
    """Check prerequisites.
5013

5014
    This checks that the instance is in the cluster.
5015

5016
    """
5017
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5018
    assert self.instance is not None, \
5019
      "Cannot retrieve locked instance %s" % self.op.instance_name
5020

    
5021
  def Exec(self, feedback_fn):
5022
    """Remove the instance.
5023

5024
    """
5025
    instance = self.instance
5026
    logging.info("Shutting down instance %s on node %s",
5027
                 instance.name, instance.primary_node)
5028

    
5029
    result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
5030
                                             self.op.shutdown_timeout)
5031
    msg = result.fail_msg
5032
    if msg:
5033
      if self.op.ignore_failures:
5034
        feedback_fn("Warning: can't shutdown instance: %s" % msg)
5035
      else:
5036
        raise errors.OpExecError("Could not shutdown instance %s on"
5037
                                 " node %s: %s" %
5038
                                 (instance.name, instance.primary_node, msg))
5039

    
5040
    _RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures)
5041

    
5042

    
5043
def _RemoveInstance(lu, feedback_fn, instance, ignore_failures):
5044
  """Utility function to remove an instance.
5045

5046
  """
5047
  logging.info("Removing block devices for instance %s", instance.name)
5048

    
5049
  if not _RemoveDisks(lu, instance):
5050
    if not ignore_failures:
5051
      raise errors.OpExecError("Can't remove instance's disks")
5052
    feedback_fn("Warning: can't remove instance's disks")
5053

    
5054
  logging.info("Removing instance %s out of cluster config", instance.name)
5055

    
5056
  lu.cfg.RemoveInstance(instance.name)
5057

    
5058
  assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \
5059
    "Instance lock removal conflict"
5060

    
5061
  # Remove lock for the instance
5062
  lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name
5063

    
5064

    
5065
class LUQueryInstances(NoHooksLU):
5066
  """Logical unit for querying instances.
5067

5068
  """
5069
  # pylint: disable-msg=W0142
5070
  _OP_PARAMS = [
5071
    ("output_fields", _NoDefault, _TListOf(_TNonEmptyString)),
5072
    ("names", _EmptyList, _TListOf(_TNonEmptyString)),
5073
    ("use_locking", False, _TBool),
5074
    ]
5075
  REQ_BGL = False
5076
  _SIMPLE_FIELDS = ["name", "os", "network_port", "hypervisor",
5077
                    "serial_no", "ctime", "mtime", "uuid"]
5078
  _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
5079
                                    "admin_state",
5080
                                    "disk_template", "ip", "mac", "bridge",
5081
                                    "nic_mode", "nic_link",
5082
                                    "sda_size", "sdb_size", "vcpus", "tags",
5083
                                    "network_port", "beparams",
5084
                                    r"(disk)\.(size)/([0-9]+)",
5085
                                    r"(disk)\.(sizes)", "disk_usage",
5086
                                    r"(nic)\.(mac|ip|mode|link)/([0-9]+)",
5087
                                    r"(nic)\.(bridge)/([0-9]+)",
5088
                                    r"(nic)\.(macs|ips|modes|links|bridges)",
5089
                                    r"(disk|nic)\.(count)",
5090
                                    "hvparams",
5091
                                    ] + _SIMPLE_FIELDS +
5092
                                  ["hv/%s" % name
5093
                                   for name in constants.HVS_PARAMETERS
5094
                                   if name not in constants.HVC_GLOBALS] +
5095
                                  ["be/%s" % name
5096
                                   for name in constants.BES_PARAMETERS])
5097
  _FIELDS_DYNAMIC = utils.FieldSet("oper_state",
5098
                                   "oper_ram",
5099
                                   "oper_vcpus",
5100
                                   "status")
5101

    
5102

    
5103
  def CheckArguments(self):
5104
    _CheckOutputFields(static=self._FIELDS_STATIC,
5105
                       dynamic=self._FIELDS_DYNAMIC,
5106
                       selected=self.op.output_fields)
5107

    
5108
  def ExpandNames(self):
5109
    self.needed_locks = {}
5110
    self.share_locks[locking.LEVEL_INSTANCE] = 1
5111
    self.share_locks[locking.LEVEL_NODE] = 1
5112

    
5113
    if self.op.names:
5114
      self.wanted = _GetWantedInstances(self, self.op.names)
5115
    else:
5116
      self.wanted = locking.ALL_SET
5117

    
5118
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
5119
    self.do_locking = self.do_node_query and self.op.use_locking
5120
    if self.do_locking:
5121
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
5122
      self.needed_locks[locking.LEVEL_NODE] = []
5123
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5124

    
5125
  def DeclareLocks(self, level):
5126
    if level == locking.LEVEL_NODE and self.do_locking:
5127
      self._LockInstancesNodes()
5128

    
5129
  def Exec(self, feedback_fn):
5130
    """Computes the list of nodes and their attributes.
5131

5132
    """
5133
    # pylint: disable-msg=R0912
5134
    # way too many branches here
5135
    all_info = self.cfg.GetAllInstancesInfo()
5136
    if self.wanted == locking.ALL_SET:
5137
      # caller didn't specify instance names, so ordering is not important
5138
      if self.do_locking:
5139
        instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
5140
      else:
5141
        instance_names = all_info.keys()
5142
      instance_names = utils.NiceSort(instance_names)
5143
    else:
5144
      # caller did specify names, so we must keep the ordering
5145
      if self.do_locking:
5146
        tgt_set = self.acquired_locks[locking.LEVEL_INSTANCE]
5147
      else:
5148
        tgt_set = all_info.keys()
5149
      missing = set(self.wanted).difference(tgt_set)
5150
      if missing:
5151
        raise errors.OpExecError("Some instances were removed before"
5152
                                 " retrieving their data: %s" % missing)
5153
      instance_names = self.wanted
5154

    
5155
    instance_list = [all_info[iname] for iname in instance_names]
5156

    
5157
    # begin data gathering
5158

    
5159
    nodes = frozenset([inst.primary_node for inst in instance_list])
5160
    hv_list = list(set([inst.hypervisor for inst in instance_list]))
5161

    
5162
    bad_nodes = []
5163
    off_nodes = []
5164
    if self.do_node_query:
5165
      live_data = {}
5166
      node_data = self.rpc.call_all_instances_info(nodes, hv_list)
5167
      for name in nodes:
5168
        result = node_data[name]
5169
        if result.offline:
5170
          # offline nodes will be in both lists
5171
          off_nodes.append(name)
5172
        if result.fail_msg:
5173
          bad_nodes.append(name)
5174
        else:
5175
          if result.payload:
5176
            live_data.update(result.payload)
5177
          # else no instance is alive
5178
    else:
5179
      live_data = dict([(name, {}) for name in instance_names])
5180

    
5181
    # end data gathering
5182

    
5183
    HVPREFIX = "hv/"
5184
    BEPREFIX = "be/"
5185
    output = []
5186
    cluster = self.cfg.GetClusterInfo()
5187
    for instance in instance_list:
5188
      iout = []
5189
      i_hv = cluster.FillHV(instance, skip_globals=True)
5190
      i_be = cluster.FillBE(instance)
5191
      i_nicp = [cluster.SimpleFillNIC(nic.nicparams) for nic in instance.nics]
5192
      for field in self.op.output_fields:
5193
        st_match = self._FIELDS_STATIC.Matches(field)
5194
        if field in self._SIMPLE_FIELDS:
5195
          val = getattr(instance, field)
5196
        elif field == "pnode":
5197
          val = instance.primary_node
5198
        elif field == "snodes":
5199
          val = list(instance.secondary_nodes)
5200
        elif field == "admin_state":
5201
          val = instance.admin_up
5202
        elif field == "oper_state":
5203
          if instance.primary_node in bad_nodes:
5204
            val = None
5205
          else:
5206
            val = bool(live_data.get(instance.name))
5207
        elif field == "status":
5208
          if instance.primary_node in off_nodes:
5209
            val = "ERROR_nodeoffline"
5210
          elif instance.primary_node in bad_nodes:
5211
            val = "ERROR_nodedown"
5212
          else:
5213
            running = bool(live_data.get(instance.name))
5214
            if running:
5215
              if instance.admin_up:
5216
                val = "running"
5217
              else:
5218
                val = "ERROR_up"
5219
            else:
5220
              if instance.admin_up:
5221
                val = "ERROR_down"
5222
              else:
5223
                val = "ADMIN_down"
5224
        elif field == "oper_ram":
5225
          if instance.primary_node in bad_nodes:
5226
            val = None
5227
          elif instance.name in live_data:
5228
            val = live_data[instance.name].get("memory", "?")
5229
          else:
5230
            val = "-"
5231
        elif field == "oper_vcpus":
5232
          if instance.primary_node in bad_nodes:
5233
            val = None
5234
          elif instance.name in live_data:
5235
            val = live_data[instance.name].get("vcpus", "?")
5236
          else:
5237
            val = "-"
5238
        elif field == "vcpus":
5239
          val = i_be[constants.BE_VCPUS]
5240
        elif field == "disk_template":
5241
          val = instance.disk_template
5242
        elif field == "ip":
5243
          if instance.nics:
5244
            val = instance.nics[0].ip
5245
          else:
5246
            val = None
5247
        elif field == "nic_mode":
5248
          if instance.nics:
5249
            val = i_nicp[0][constants.NIC_MODE]
5250
          else:
5251
            val = None
5252
        elif field == "nic_link":
5253
          if instance.nics:
5254
            val = i_nicp[0][constants.NIC_LINK]
5255
          else:
5256
            val = None
5257
        elif field == "bridge":
5258
          if (instance.nics and
5259
              i_nicp[0][constants.NIC_MODE] == constants.NIC_MODE_BRIDGED):
5260
            val = i_nicp[0][constants.NIC_LINK]
5261
          else:
5262
            val = None
5263
        elif field == "mac":
5264
          if instance.nics:
5265
            val = instance.nics[0].mac
5266
          else:
5267
            val = None
5268
        elif field == "sda_size" or field == "sdb_size":
5269
          idx = ord(field[2]) - ord('a')
5270
          try:
5271
            val = instance.FindDisk(idx).size
5272
          except errors.OpPrereqError:
5273
            val = None
5274
        elif field == "disk_usage": # total disk usage per node
5275
          disk_sizes = [{'size': disk.size} for disk in instance.disks]
5276
          val = _ComputeDiskSize(instance.disk_template, disk_sizes)
5277
        elif field == "tags":
5278
          val = list(instance.GetTags())
5279
        elif field == "hvparams":
5280
          val = i_hv
5281
        elif (field.startswith(HVPREFIX) and
5282
              field[len(HVPREFIX):] in constants.HVS_PARAMETERS and
5283
              field[len(HVPREFIX):] not in constants.HVC_GLOBALS):
5284
          val = i_hv.get(field[len(HVPREFIX):], None)
5285
        elif field == "beparams":
5286
          val = i_be
5287
        elif (field.startswith(BEPREFIX) and
5288
              field[len(BEPREFIX):] in constants.BES_PARAMETERS):
5289
          val = i_be.get(field[len(BEPREFIX):], None)
5290
        elif st_match and st_match.groups():
5291
          # matches a variable list
5292
          st_groups = st_match.groups()
5293
          if st_groups and st_groups[0] == "disk":
5294
            if st_groups[1] == "count":
5295
              val = len(instance.disks)
5296
            elif st_groups[1] == "sizes":
5297
              val = [disk.size for disk in instance.disks]
5298
            elif st_groups[1] == "size":
5299
              try:
5300
                val = instance.FindDisk(st_groups[2]).size
5301
              except errors.OpPrereqError:
5302
                val = None
5303
            else:
5304
              assert False, "Unhandled disk parameter"
5305
          elif st_groups[0] == "nic":
5306
            if st_groups[1] == "count":
5307
              val = len(instance.nics)
5308
            elif st_groups[1] == "macs":
5309
              val = [nic.mac for nic in instance.nics]
5310
            elif st_groups[1] == "ips":
5311
              val = [nic.ip for nic in instance.nics]
5312
            elif st_groups[1] == "modes":
5313
              val = [nicp[constants.NIC_MODE] for nicp in i_nicp]
5314
            elif st_groups[1] == "links":
5315
              val = [nicp[constants.NIC_LINK] for nicp in i_nicp]
5316
            elif st_groups[1] == "bridges":
5317
              val = []
5318
              for nicp in i_nicp:
5319
                if nicp[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
5320
                  val.append(nicp[constants.NIC_LINK])
5321
                else:
5322
                  val.append(None)
5323
            else:
5324
              # index-based item
5325
              nic_idx = int(st_groups[2])
5326
              if nic_idx >= len(instance.nics):
5327
                val = None
5328
              else:
5329
                if st_groups[1] == "mac":
5330
                  val = instance.nics[nic_idx].mac
5331
                elif st_groups[1] == "ip":
5332
                  val = instance.nics[nic_idx].ip
5333
                elif st_groups[1] == "mode":
5334
                  val = i_nicp[nic_idx][constants.NIC_MODE]
5335
                elif st_groups[1] == "link":
5336
                  val = i_nicp[nic_idx][constants.NIC_LINK]
5337
                elif st_groups[1] == "bridge":
5338
                  nic_mode = i_nicp[nic_idx][constants.NIC_MODE]
5339
                  if nic_mode == constants.NIC_MODE_BRIDGED:
5340
                    val = i_nicp[nic_idx][constants.NIC_LINK]
5341
                  else:
5342
                    val = None
5343
                else:
5344
                  assert False, "Unhandled NIC parameter"
5345
          else:
5346
            assert False, ("Declared but unhandled variable parameter '%s'" %
5347
                           field)
5348
        else:
5349
          assert False, "Declared but unhandled parameter '%s'" % field
5350
        iout.append(val)
5351
      output.append(iout)
5352

    
5353
    return output
5354

    
5355

    
5356
class LUFailoverInstance(LogicalUnit):
5357
  """Failover an instance.
5358

5359
  """
5360
  HPATH = "instance-failover"
5361
  HTYPE = constants.HTYPE_INSTANCE
5362
  _OP_PARAMS = [
5363
    _PInstanceName,
5364
    ("ignore_consistency", False, _TBool),
5365
    _PShutdownTimeout,
5366
    ]
5367
  REQ_BGL = False
5368

    
5369
  def ExpandNames(self):
5370
    self._ExpandAndLockInstance()
5371
    self.needed_locks[locking.LEVEL_NODE] = []
5372
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5373

    
5374
  def DeclareLocks(self, level):
5375
    if level == locking.LEVEL_NODE:
5376
      self._LockInstancesNodes()
5377

    
5378
  def BuildHooksEnv(self):
5379
    """Build hooks env.
5380

5381
    This runs on master, primary and secondary nodes of the instance.
5382

5383
    """
5384
    instance = self.instance
5385
    source_node = instance.primary_node
5386
    target_node = instance.secondary_nodes[0]
5387
    env = {
5388
      "IGNORE_CONSISTENCY": self.op.ignore_consistency,
5389
      "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5390
      "OLD_PRIMARY": source_node,
5391
      "OLD_SECONDARY": target_node,
5392
      "NEW_PRIMARY": target_node,
5393
      "NEW_SECONDARY": source_node,
5394
      }
5395
    env.update(_BuildInstanceHookEnvByObject(self, instance))
5396
    nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5397
    nl_post = list(nl)
5398
    nl_post.append(source_node)
5399
    return env, nl, nl_post
5400

    
5401
  def CheckPrereq(self):
5402
    """Check prerequisites.
5403

5404
    This checks that the instance is in the cluster.
5405

5406
    """
5407
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5408
    assert self.instance is not None, \
5409
      "Cannot retrieve locked instance %s" % self.op.instance_name
5410

    
5411
    bep = self.cfg.GetClusterInfo().FillBE(instance)
5412
    if instance.disk_template not in constants.DTS_NET_MIRROR:
5413
      raise errors.OpPrereqError("Instance's disk layout is not"
5414
                                 " network mirrored, cannot failover.",
5415
                                 errors.ECODE_STATE)
5416

    
5417
    secondary_nodes = instance.secondary_nodes
5418
    if not secondary_nodes:
5419
      raise errors.ProgrammerError("no secondary node but using "
5420
                                   "a mirrored disk template")
5421

    
5422
    target_node = secondary_nodes[0]
5423
    _CheckNodeOnline(self, target_node)
5424
    _CheckNodeNotDrained(self, target_node)
5425
    if instance.admin_up:
5426
      # check memory requirements on the secondary node
5427
      _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5428
                           instance.name, bep[constants.BE_MEMORY],
5429
                           instance.hypervisor)
5430
    else:
5431
      self.LogInfo("Not checking memory on the secondary node as"
5432
                   " instance will not be started")
5433

    
5434
    # check bridge existance
5435
    _CheckInstanceBridgesExist(self, instance, node=target_node)
5436

    
5437
  def Exec(self, feedback_fn):
5438
    """Failover an instance.
5439

5440
    The failover is done by shutting it down on its present node and
5441
    starting it on the secondary.
5442

5443
    """
5444
    instance = self.instance
5445

    
5446
    source_node = instance.primary_node
5447
    target_node = instance.secondary_nodes[0]
5448

    
5449
    if instance.admin_up:
5450
      feedback_fn("* checking disk consistency between source and target")
5451
      for dev in instance.disks:
5452
        # for drbd, these are drbd over lvm
5453
        if not _CheckDiskConsistency(self, dev, target_node, False):
5454
          if not self.op.ignore_consistency:
5455
            raise errors.OpExecError("Disk %s is degraded on target node,"
5456
                                     " aborting failover." % dev.iv_name)
5457
    else:
5458
      feedback_fn("* not checking disk consistency as instance is not running")
5459

    
5460
    feedback_fn("* shutting down instance on source node")
5461
    logging.info("Shutting down instance %s on node %s",
5462
                 instance.name, source_node)
5463

    
5464
    result = self.rpc.call_instance_shutdown(source_node, instance,
5465
                                             self.op.shutdown_timeout)
5466
    msg = result.fail_msg
5467
    if msg:
5468
      if self.op.ignore_consistency:
5469
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
5470
                             " Proceeding anyway. Please make sure node"
5471
                             " %s is down. Error details: %s",
5472
                             instance.name, source_node, source_node, msg)
5473
      else:
5474
        raise errors.OpExecError("Could not shutdown instance %s on"
5475
                                 " node %s: %s" %
5476
                                 (instance.name, source_node, msg))
5477

    
5478
    feedback_fn("* deactivating the instance's disks on source node")
5479
    if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
5480
      raise errors.OpExecError("Can't shut down the instance's disks.")
5481

    
5482
    instance.primary_node = target_node
5483
    # distribute new instance config to the other nodes
5484
    self.cfg.Update(instance, feedback_fn)
5485

    
5486
    # Only start the instance if it's marked as up
5487
    if instance.admin_up:
5488
      feedback_fn("* activating the instance's disks on target node")
5489
      logging.info("Starting instance %s on node %s",
5490
                   instance.name, target_node)
5491

    
5492
      disks_ok, _ = _AssembleInstanceDisks(self, instance,
5493
                                           ignore_secondaries=True)
5494
      if not disks_ok:
5495
        _ShutdownInstanceDisks(self, instance)
5496
        raise errors.OpExecError("Can't activate the instance's disks")
5497

    
5498
      feedback_fn("* starting the instance on the target node")
5499
      result = self.rpc.call_instance_start(target_node, instance, None, None)
5500
      msg = result.fail_msg
5501
      if msg:
5502
        _ShutdownInstanceDisks(self, instance)
5503
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
5504
                                 (instance.name, target_node, msg))
5505

    
5506

    
5507
class LUMigrateInstance(LogicalUnit):
5508
  """Migrate an instance.
5509

5510
  This is migration without shutting down, compared to the failover,
5511
  which is done with shutdown.
5512

5513
  """
5514
  HPATH = "instance-migrate"
5515
  HTYPE = constants.HTYPE_INSTANCE
5516
  _OP_PARAMS = [
5517
    _PInstanceName,
5518
    _PMigrationMode,
5519
    _PMigrationLive,
5520
    ("cleanup", False, _TBool),
5521
    ]
5522

    
5523
  REQ_BGL = False
5524

    
5525
  def ExpandNames(self):
5526
    self._ExpandAndLockInstance()
5527

    
5528
    self.needed_locks[locking.LEVEL_NODE] = []
5529
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5530

    
5531
    self._migrater = TLMigrateInstance(self, self.op.instance_name,
5532
                                       self.op.cleanup)
5533
    self.tasklets = [self._migrater]
5534

    
5535
  def DeclareLocks(self, level):
5536
    if level == locking.LEVEL_NODE:
5537
      self._LockInstancesNodes()
5538

    
5539
  def BuildHooksEnv(self):
5540
    """Build hooks env.
5541

5542
    This runs on master, primary and secondary nodes of the instance.
5543

5544
    """
5545
    instance = self._migrater.instance
5546
    source_node = instance.primary_node
5547
    target_node = instance.secondary_nodes[0]
5548
    env = _BuildInstanceHookEnvByObject(self, instance)
5549
    env["MIGRATE_LIVE"] = self._migrater.live
5550
    env["MIGRATE_CLEANUP"] = self.op.cleanup
5551
    env.update({
5552
        "OLD_PRIMARY": source_node,
5553
        "OLD_SECONDARY": target_node,
5554
        "NEW_PRIMARY": target_node,
5555
        "NEW_SECONDARY": source_node,
5556
        })
5557
    nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5558
    nl_post = list(nl)
5559
    nl_post.append(source_node)
5560
    return env, nl, nl_post
5561

    
5562

    
5563
class LUMoveInstance(LogicalUnit):
5564
  """Move an instance by data-copying.
5565

5566
  """
5567
  HPATH = "instance-move"
5568
  HTYPE = constants.HTYPE_INSTANCE
5569
  _OP_PARAMS = [
5570
    _PInstanceName,
5571
    ("target_node", _NoDefault, _TNonEmptyString),
5572
    _PShutdownTimeout,
5573
    ]
5574
  REQ_BGL = False
5575

    
5576
  def ExpandNames(self):
5577
    self._ExpandAndLockInstance()
5578
    target_node = _ExpandNodeName(self.cfg, self.op.target_node)
5579
    self.op.target_node = target_node
5580
    self.needed_locks[locking.LEVEL_NODE] = [target_node]
5581
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5582

    
5583
  def DeclareLocks(self, level):
5584
    if level == locking.LEVEL_NODE:
5585
      self._LockInstancesNodes(primary_only=True)
5586

    
5587
  def BuildHooksEnv(self):
5588
    """Build hooks env.
5589

5590
    This runs on master, primary and secondary nodes of the instance.
5591

5592
    """
5593
    env = {
5594
      "TARGET_NODE": self.op.target_node,
5595
      "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5596
      }
5597
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5598
    nl = [self.cfg.GetMasterNode()] + [self.instance.primary_node,
5599
                                       self.op.target_node]
5600
    return env, nl, nl
5601

    
5602
  def CheckPrereq(self):
5603
    """Check prerequisites.
5604

5605
    This checks that the instance is in the cluster.
5606

5607
    """
5608
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5609
    assert self.instance is not None, \
5610
      "Cannot retrieve locked instance %s" % self.op.instance_name
5611

    
5612
    node = self.cfg.GetNodeInfo(self.op.target_node)
5613
    assert node is not None, \
5614
      "Cannot retrieve locked node %s" % self.op.target_node
5615

    
5616
    self.target_node = target_node = node.name
5617

    
5618
    if target_node == instance.primary_node:
5619
      raise errors.OpPrereqError("Instance %s is already on the node %s" %
5620
                                 (instance.name, target_node),
5621
                                 errors.ECODE_STATE)
5622

    
5623
    bep = self.cfg.GetClusterInfo().FillBE(instance)
5624

    
5625
    for idx, dsk in enumerate(instance.disks):
5626
      if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
5627
        raise errors.OpPrereqError("Instance disk %d has a complex layout,"
5628
                                   " cannot copy" % idx, errors.ECODE_STATE)
5629

    
5630
    _CheckNodeOnline(self, target_node)
5631
    _CheckNodeNotDrained(self, target_node)
5632

    
5633
    if instance.admin_up:
5634
      # check memory requirements on the secondary node
5635
      _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5636
                           instance.name, bep[constants.BE_MEMORY],
5637
                           instance.hypervisor)
5638
    else:
5639
      self.LogInfo("Not checking memory on the secondary node as"
5640
                   " instance will not be started")
5641

    
5642
    # check bridge existance
5643
    _CheckInstanceBridgesExist(self, instance, node=target_node)
5644

    
5645
  def Exec(self, feedback_fn):
5646
    """Move an instance.
5647

5648
    The move is done by shutting it down on its present node, copying
5649
    the data over (slow) and starting it on the new node.
5650

5651
    """
5652
    instance = self.instance
5653

    
5654
    source_node = instance.primary_node
5655
    target_node = self.target_node
5656

    
5657
    self.LogInfo("Shutting down instance %s on source node %s",
5658
                 instance.name, source_node)
5659

    
5660
    result = self.rpc.call_instance_shutdown(source_node, instance,
5661
                                             self.op.shutdown_timeout)
5662
    msg = result.fail_msg
5663
    if msg:
5664
      if self.op.ignore_consistency:
5665
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
5666
                             " Proceeding anyway. Please make sure node"
5667
                             " %s is down. Error details: %s",
5668
                             instance.name, source_node, source_node, msg)
5669
      else:
5670
        raise errors.OpExecError("Could not shutdown instance %s on"
5671
                                 " node %s: %s" %
5672
                                 (instance.name, source_node, msg))
5673

    
5674
    # create the target disks
5675
    try:
5676
      _CreateDisks(self, instance, target_node=target_node)
5677
    except errors.OpExecError:
5678
      self.LogWarning("Device creation failed, reverting...")
5679
      try:
5680
        _RemoveDisks(self, instance, target_node=target_node)
5681
      finally:
5682
        self.cfg.ReleaseDRBDMinors(instance.name)
5683
        raise
5684

    
5685
    cluster_name = self.cfg.GetClusterInfo().cluster_name
5686

    
5687
    errs = []
5688
    # activate, get path, copy the data over
5689
    for idx, disk in enumerate(instance.disks):
5690
      self.LogInfo("Copying data for disk %d", idx)
5691
      result = self.rpc.call_blockdev_assemble(target_node, disk,
5692
                                               instance.name, True)
5693
      if result.fail_msg:
5694
        self.LogWarning("Can't assemble newly created disk %d: %s",
5695
                        idx, result.fail_msg)
5696
        errs.append(result.fail_msg)
5697
        break
5698
      dev_path = result.payload
5699
      result = self.rpc.call_blockdev_export(source_node, disk,
5700
                                             target_node, dev_path,
5701
                                             cluster_name)
5702
      if result.fail_msg:
5703
        self.LogWarning("Can't copy data over for disk %d: %s",
5704
                        idx, result.fail_msg)
5705
        errs.append(result.fail_msg)
5706
        break
5707

    
5708
    if errs:
5709
      self.LogWarning("Some disks failed to copy, aborting")
5710
      try:
5711
        _RemoveDisks(self, instance, target_node=target_node)
5712
      finally:
5713
        self.cfg.ReleaseDRBDMinors(instance.name)
5714
        raise errors.OpExecError("Errors during disk copy: %s" %
5715
                                 (",".join(errs),))
5716

    
5717
    instance.primary_node = target_node
5718
    self.cfg.Update(instance, feedback_fn)
5719

    
5720
    self.LogInfo("Removing the disks on the original node")
5721
    _RemoveDisks(self, instance, target_node=source_node)
5722

    
5723
    # Only start the instance if it's marked as up
5724
    if instance.admin_up:
5725
      self.LogInfo("Starting instance %s on node %s",
5726
                   instance.name, target_node)
5727

    
5728
      disks_ok, _ = _AssembleInstanceDisks(self, instance,
5729
                                           ignore_secondaries=True)
5730
      if not disks_ok:
5731
        _ShutdownInstanceDisks(self, instance)
5732
        raise errors.OpExecError("Can't activate the instance's disks")
5733

    
5734
      result = self.rpc.call_instance_start(target_node, instance, None, None)
5735
      msg = result.fail_msg
5736
      if msg:
5737
        _ShutdownInstanceDisks(self, instance)
5738
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
5739
                                 (instance.name, target_node, msg))
5740

    
5741

    
5742
class LUMigrateNode(LogicalUnit):
5743
  """Migrate all instances from a node.
5744

5745
  """
5746
  HPATH = "node-migrate"
5747
  HTYPE = constants.HTYPE_NODE
5748
  _OP_PARAMS = [
5749
    _PNodeName,
5750
    _PMigrationMode,
5751
    _PMigrationLive,
5752
    ]
5753
  REQ_BGL = False
5754

    
5755
  def ExpandNames(self):
5756
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
5757

    
5758
    self.needed_locks = {
5759
      locking.LEVEL_NODE: [self.op.node_name],
5760
      }
5761

    
5762
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5763

    
5764
    # Create tasklets for migrating instances for all instances on this node
5765
    names = []
5766
    tasklets = []
5767

    
5768
    for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name):
5769
      logging.debug("Migrating instance %s", inst.name)
5770
      names.append(inst.name)
5771

    
5772
      tasklets.append(TLMigrateInstance(self, inst.name, False))
5773

    
5774
    self.tasklets = tasklets
5775

    
5776
    # Declare instance locks
5777
    self.needed_locks[locking.LEVEL_INSTANCE] = names
5778

    
5779
  def DeclareLocks(self, level):
5780
    if level == locking.LEVEL_NODE:
5781
      self._LockInstancesNodes()
5782

    
5783
  def BuildHooksEnv(self):
5784
    """Build hooks env.
5785

5786
    This runs on the master, the primary and all the secondaries.
5787

5788
    """
5789
    env = {
5790
      "NODE_NAME": self.op.node_name,
5791
      }
5792

    
5793
    nl = [self.cfg.GetMasterNode()]
5794

    
5795
    return (env, nl, nl)
5796

    
5797

    
5798
class TLMigrateInstance(Tasklet):
5799
  """Tasklet class for instance migration.
5800

5801
  @type live: boolean
5802
  @ivar live: whether the migration will be done live or non-live;
5803
      this variable is initalized only after CheckPrereq has run
5804

5805
  """
5806
  def __init__(self, lu, instance_name, cleanup):
5807
    """Initializes this class.
5808

5809
    """
5810
    Tasklet.__init__(self, lu)
5811

    
5812
    # Parameters
5813
    self.instance_name = instance_name
5814
    self.cleanup = cleanup
5815
    self.live = False # will be overridden later
5816

    
5817
  def CheckPrereq(self):
5818
    """Check prerequisites.
5819

5820
    This checks that the instance is in the cluster.
5821

5822
    """
5823
    instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name)
5824
    instance = self.cfg.GetInstanceInfo(instance_name)
5825
    assert instance is not None
5826

    
5827
    if instance.disk_template != constants.DT_DRBD8:
5828
      raise errors.OpPrereqError("Instance's disk layout is not"
5829
                                 " drbd8, cannot migrate.", errors.ECODE_STATE)
5830

    
5831
    secondary_nodes = instance.secondary_nodes
5832
    if not secondary_nodes:
5833
      raise errors.ConfigurationError("No secondary node but using"
5834
                                      " drbd8 disk template")
5835

    
5836
    i_be = self.cfg.GetClusterInfo().FillBE(instance)
5837

    
5838
    target_node = secondary_nodes[0]
5839
    # check memory requirements on the secondary node
5840
    _CheckNodeFreeMemory(self.lu, target_node, "migrating instance %s" %
5841
                         instance.name, i_be[constants.BE_MEMORY],
5842
                         instance.hypervisor)
5843

    
5844
    # check bridge existance
5845
    _CheckInstanceBridgesExist(self.lu, instance, node=target_node)
5846

    
5847
    if not self.cleanup:
5848
      _CheckNodeNotDrained(self.lu, target_node)
5849
      result = self.rpc.call_instance_migratable(instance.primary_node,
5850
                                                 instance)
5851
      result.Raise("Can't migrate, please use failover",
5852
                   prereq=True, ecode=errors.ECODE_STATE)
5853

    
5854
    self.instance = instance
5855

    
5856
    if self.lu.op.live is not None and self.lu.op.mode is not None:
5857
      raise errors.OpPrereqError("Only one of the 'live' and 'mode'"
5858
                                 " parameters are accepted",
5859
                                 errors.ECODE_INVAL)
5860
    if self.lu.op.live is not None:
5861
      if self.lu.op.live:
5862
        self.lu.op.mode = constants.HT_MIGRATION_LIVE
5863
      else:
5864
        self.lu.op.mode = constants.HT_MIGRATION_NONLIVE
5865
      # reset the 'live' parameter to None so that repeated
5866
      # invocations of CheckPrereq do not raise an exception
5867
      self.lu.op.live = None
5868
    elif self.lu.op.mode is None:
5869
      # read the default value from the hypervisor
5870
      i_hv = self.cfg.GetClusterInfo().FillHV(instance, skip_globals=False)
5871
      self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE]
5872

    
5873
    self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE
5874

    
5875
  def _WaitUntilSync(self):
5876
    """Poll with custom rpc for disk sync.
5877

5878
    This uses our own step-based rpc call.
5879

5880
    """
5881
    self.feedback_fn("* wait until resync is done")
5882
    all_done = False
5883
    while not all_done:
5884
      all_done = True
5885
      result = self.rpc.call_drbd_wait_sync(self.all_nodes,
5886
                                            self.nodes_ip,
5887
                                            self.instance.disks)
5888
      min_percent = 100
5889
      for node, nres in result.items():
5890
        nres.Raise("Cannot resync disks on node %s" % node)
5891
        node_done, node_percent = nres.payload
5892
        all_done = all_done and node_done
5893
        if node_percent is not None:
5894
          min_percent = min(min_percent, node_percent)
5895
      if not all_done:
5896
        if min_percent < 100:
5897
          self.feedback_fn("   - progress: %.1f%%" % min_percent)
5898
        time.sleep(2)
5899

    
5900
  def _EnsureSecondary(self, node):
5901
    """Demote a node to secondary.
5902

5903
    """
5904
    self.feedback_fn("* switching node %s to secondary mode" % node)
5905

    
5906
    for dev in self.instance.disks:
5907
      self.cfg.SetDiskID(dev, node)
5908

    
5909
    result = self.rpc.call_blockdev_close(node, self.instance.name,
5910
                                          self.instance.disks)
5911
    result.Raise("Cannot change disk to secondary on node %s" % node)
5912

    
5913
  def _GoStandalone(self):
5914
    """Disconnect from the network.
5915

5916
    """
5917
    self.feedback_fn("* changing into standalone mode")
5918
    result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
5919
                                               self.instance.disks)
5920
    for node, nres in result.items():
5921
      nres.Raise("Cannot disconnect disks node %s" % node)
5922

    
5923
  def _GoReconnect(self, multimaster):
5924
    """Reconnect to the network.
5925

5926
    """
5927
    if multimaster:
5928
      msg = "dual-master"
5929
    else:
5930
      msg = "single-master"
5931
    self.feedback_fn("* changing disks into %s mode" % msg)
5932
    result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
5933
                                           self.instance.disks,
5934
                                           self.instance.name, multimaster)
5935
    for node, nres in result.items():
5936
      nres.Raise("Cannot change disks config on node %s" % node)
5937

    
5938
  def _ExecCleanup(self):
5939
    """Try to cleanup after a failed migration.
5940

5941
    The cleanup is done by:
5942
      - check that the instance is running only on one node
5943
        (and update the config if needed)
5944
      - change disks on its secondary node to secondary
5945
      - wait until disks are fully synchronized
5946
      - disconnect from the network
5947
      - change disks into single-master mode
5948
      - wait again until disks are fully synchronized
5949

5950
    """
5951
    instance = self.instance
5952
    target_node = self.target_node
5953
    source_node = self.source_node
5954

    
5955
    # check running on only one node
5956
    self.feedback_fn("* checking where the instance actually runs"
5957
                     " (if this hangs, the hypervisor might be in"
5958
                     " a bad state)")
5959
    ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
5960
    for node, result in ins_l.items():
5961
      result.Raise("Can't contact node %s" % node)
5962

    
5963
    runningon_source = instance.name in ins_l[source_node].payload
5964
    runningon_target = instance.name in ins_l[target_node].payload
5965

    
5966
    if runningon_source and runningon_target:
5967
      raise errors.OpExecError("Instance seems to be running on two nodes,"
5968
                               " or the hypervisor is confused. You will have"
5969
                               " to ensure manually that it runs only on one"
5970
                               " and restart this operation.")
5971

    
5972
    if not (runningon_source or runningon_target):
5973
      raise errors.OpExecError("Instance does not seem to be running at all."
5974
                               " In this case, it's safer to repair by"
5975
                               " running 'gnt-instance stop' to ensure disk"
5976
                               " shutdown, and then restarting it.")
5977

    
5978
    if runningon_target:
5979
      # the migration has actually succeeded, we need to update the config
5980
      self.feedback_fn("* instance running on secondary node (%s),"
5981
                       " updating config" % target_node)
5982
      instance.primary_node = target_node
5983
      self.cfg.Update(instance, self.feedback_fn)
5984
      demoted_node = source_node
5985
    else:
5986
      self.feedback_fn("* instance confirmed to be running on its"
5987
                       " primary node (%s)" % source_node)
5988
      demoted_node = target_node
5989

    
5990
    self._EnsureSecondary(demoted_node)
5991
    try:
5992
      self._WaitUntilSync()
5993
    except errors.OpExecError:
5994
      # we ignore here errors, since if the device is standalone, it
5995
      # won't be able to sync
5996
      pass
5997
    self._GoStandalone()
5998
    self._GoReconnect(False)
5999
    self._WaitUntilSync()
6000

    
6001
    self.feedback_fn("* done")
6002

    
6003
  def _RevertDiskStatus(self):
6004
    """Try to revert the disk status after a failed migration.
6005

6006
    """
6007
    target_node = self.target_node
6008
    try:
6009
      self._EnsureSecondary(target_node)
6010
      self._GoStandalone()
6011
      self._GoReconnect(False)
6012
      self._WaitUntilSync()
6013
    except errors.OpExecError, err:
6014
      self.lu.LogWarning("Migration failed and I can't reconnect the"
6015
                         " drives: error '%s'\n"
6016
                         "Please look and recover the instance status" %
6017
                         str(err))
6018

    
6019
  def _AbortMigration(self):
6020
    """Call the hypervisor code to abort a started migration.
6021

6022
    """
6023
    instance = self.instance
6024
    target_node = self.target_node
6025
    migration_info = self.migration_info
6026

    
6027
    abort_result = self.rpc.call_finalize_migration(target_node,
6028
                                                    instance,
6029
                                                    migration_info,
6030
                                                    False)
6031
    abort_msg = abort_result.fail_msg
6032
    if abort_msg:
6033
      logging.error("Aborting migration failed on target node %s: %s",
6034
                    target_node, abort_msg)
6035
      # Don't raise an exception here, as we stil have to try to revert the
6036
      # disk status, even if this step failed.
6037

    
6038
  def _ExecMigration(self):
6039
    """Migrate an instance.
6040

6041
    The migrate is done by:
6042
      - change the disks into dual-master mode
6043
      - wait until disks are fully synchronized again
6044
      - migrate the instance
6045
      - change disks on the new secondary node (the old primary) to secondary
6046
      - wait until disks are fully synchronized
6047
      - change disks into single-master mode
6048

6049
    """
6050
    instance = self.instance
6051
    target_node = self.target_node
6052
    source_node = self.source_node
6053

    
6054
    self.feedback_fn("* checking disk consistency between source and target")
6055
    for dev in instance.disks:
6056
      if not _CheckDiskConsistency(self.lu, dev, target_node, False):
6057
        raise errors.OpExecError("Disk %s is degraded or not fully"
6058
                                 " synchronized on target node,"
6059
                                 " aborting migrate." % dev.iv_name)
6060

    
6061
    # First get the migration information from the remote node
6062
    result = self.rpc.call_migration_info(source_node, instance)
6063
    msg = result.fail_msg
6064
    if msg:
6065
      log_err = ("Failed fetching source migration information from %s: %s" %
6066
                 (source_node, msg))
6067
      logging.error(log_err)
6068
      raise errors.OpExecError(log_err)
6069

    
6070
    self.migration_info = migration_info = result.payload
6071

    
6072
    # Then switch the disks to master/master mode
6073
    self._EnsureSecondary(target_node)
6074
    self._GoStandalone()
6075
    self._GoReconnect(True)
6076
    self._WaitUntilSync()
6077

    
6078
    self.feedback_fn("* preparing %s to accept the instance" % target_node)
6079
    result = self.rpc.call_accept_instance(target_node,
6080
                                           instance,
6081
                                           migration_info,
6082
                                           self.nodes_ip[target_node])
6083

    
6084
    msg = result.fail_msg
6085
    if msg:
6086
      logging.error("Instance pre-migration failed, trying to revert"
6087
                    " disk status: %s", msg)
6088
      self.feedback_fn("Pre-migration failed, aborting")
6089
      self._AbortMigration()
6090
      self._RevertDiskStatus()
6091
      raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
6092
                               (instance.name, msg))
6093

    
6094
    self.feedback_fn("* migrating instance to %s" % target_node)
6095
    time.sleep(10)
6096
    result = self.rpc.call_instance_migrate(source_node, instance,
6097
                                            self.nodes_ip[target_node],
6098
                                            self.live)
6099
    msg = result.fail_msg
6100
    if msg:
6101
      logging.error("Instance migration failed, trying to revert"
6102
                    " disk status: %s", msg)
6103
      self.feedback_fn("Migration failed, aborting")
6104
      self._AbortMigration()
6105
      self._RevertDiskStatus()
6106
      raise errors.OpExecError("Could not migrate instance %s: %s" %
6107
                               (instance.name, msg))
6108
    time.sleep(10)
6109

    
6110
    instance.primary_node = target_node
6111
    # distribute new instance config to the other nodes
6112
    self.cfg.Update(instance, self.feedback_fn)
6113

    
6114
    result = self.rpc.call_finalize_migration(target_node,
6115
                                              instance,
6116
                                              migration_info,
6117
                                              True)
6118
    msg = result.fail_msg
6119
    if msg:
6120
      logging.error("Instance migration succeeded, but finalization failed:"
6121
                    " %s", msg)
6122
      raise errors.OpExecError("Could not finalize instance migration: %s" %
6123
                               msg)
6124

    
6125
    self._EnsureSecondary(source_node)
6126
    self._WaitUntilSync()
6127
    self._GoStandalone()
6128
    self._GoReconnect(False)
6129
    self._WaitUntilSync()
6130

    
6131
    self.feedback_fn("* done")
6132

    
6133
  def Exec(self, feedback_fn):
6134
    """Perform the migration.
6135

6136
    """
6137
    feedback_fn("Migrating instance %s" % self.instance.name)
6138

    
6139
    self.feedback_fn = feedback_fn
6140

    
6141
    self.source_node = self.instance.primary_node
6142
    self.target_node = self.instance.secondary_nodes[0]
6143
    self.all_nodes = [self.source_node, self.target_node]
6144
    self.nodes_ip = {
6145
      self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
6146
      self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
6147
      }
6148

    
6149
    if self.cleanup:
6150
      return self._ExecCleanup()
6151
    else:
6152
      return self._ExecMigration()
6153

    
6154

    
6155
def _CreateBlockDev(lu, node, instance, device, force_create,
6156
                    info, force_open):
6157
  """Create a tree of block devices on a given node.
6158

6159
  If this device type has to be created on secondaries, create it and
6160
  all its children.
6161

6162
  If not, just recurse to children keeping the same 'force' value.
6163

6164
  @param lu: the lu on whose behalf we execute
6165
  @param node: the node on which to create the device
6166
  @type instance: L{objects.Instance}
6167
  @param instance: the instance which owns the device
6168
  @type device: L{objects.Disk}
6169
  @param device: the device to create
6170
  @type force_create: boolean
6171
  @param force_create: whether to force creation of this device; this
6172
      will be change to True whenever we find a device which has
6173
      CreateOnSecondary() attribute
6174
  @param info: the extra 'metadata' we should attach to the device
6175
      (this will be represented as a LVM tag)
6176
  @type force_open: boolean
6177
  @param force_open: this parameter will be passes to the
6178
      L{backend.BlockdevCreate} function where it specifies
6179
      whether we run on primary or not, and it affects both
6180
      the child assembly and the device own Open() execution
6181

6182
  """
6183
  if device.CreateOnSecondary():
6184
    force_create = True
6185

    
6186
  if device.children:
6187
    for child in device.children:
6188
      _CreateBlockDev(lu, node, instance, child, force_create,
6189
                      info, force_open)
6190

    
6191
  if not force_create:
6192
    return
6193

    
6194
  _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
6195

    
6196

    
6197
def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
6198
  """Create a single block device on a given node.
6199

6200
  This will not recurse over children of the device, so they must be
6201
  created in advance.
6202

6203
  @param lu: the lu on whose behalf we execute
6204
  @param node: the node on which to create the device
6205
  @type instance: L{objects.Instance}
6206
  @param instance: the instance which owns the device
6207
  @type device: L{objects.Disk}
6208
  @param device: the device to create
6209
  @param info: the extra 'metadata' we should attach to the device
6210
      (this will be represented as a LVM tag)
6211
  @type force_open: boolean
6212
  @param force_open: this parameter will be passes to the
6213
      L{backend.BlockdevCreate} function where it specifies
6214
      whether we run on primary or not, and it affects both
6215
      the child assembly and the device own Open() execution
6216

6217
  """
6218
  lu.cfg.SetDiskID(device, node)
6219
  result = lu.rpc.call_blockdev_create(node, device, device.size,
6220
                                       instance.name, force_open, info)
6221
  result.Raise("Can't create block device %s on"
6222
               " node %s for instance %s" % (device, node, instance.name))
6223
  if device.physical_id is None:
6224
    device.physical_id = result.payload
6225

    
6226

    
6227
def _GenerateUniqueNames(lu, exts):
6228
  """Generate a suitable LV name.
6229

6230
  This will generate a logical volume name for the given instance.
6231

6232
  """
6233
  results = []
6234
  for val in exts:
6235
    new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
6236
    results.append("%s%s" % (new_id, val))
6237
  return results
6238

    
6239

    
6240
def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
6241
                         p_minor, s_minor):
6242
  """Generate a drbd8 device complete with its children.
6243

6244
  """
6245
  port = lu.cfg.AllocatePort()
6246
  vgname = lu.cfg.GetVGName()
6247
  shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
6248
  dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
6249
                          logical_id=(vgname, names[0]))
6250
  dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
6251
                          logical_id=(vgname, names[1]))
6252
  drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
6253
                          logical_id=(primary, secondary, port,
6254
                                      p_minor, s_minor,
6255
                                      shared_secret),
6256
                          children=[dev_data, dev_meta],
6257
                          iv_name=iv_name)
6258
  return drbd_dev
6259

    
6260

    
6261
def _GenerateDiskTemplate(lu, template_name,
6262
                          instance_name, primary_node,
6263
                          secondary_nodes, disk_info,
6264
                          file_storage_dir, file_driver,
6265
                          base_index):
6266
  """Generate the entire disk layout for a given template type.
6267

6268
  """
6269
  #TODO: compute space requirements
6270

    
6271
  vgname = lu.cfg.GetVGName()
6272
  disk_count = len(disk_info)
6273
  disks = []
6274
  if template_name == constants.DT_DISKLESS:
6275
    pass
6276
  elif template_name == constants.DT_PLAIN:
6277
    if len(secondary_nodes) != 0:
6278
      raise errors.ProgrammerError("Wrong template configuration")
6279

    
6280
    names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
6281
                                      for i in range(disk_count)])
6282
    for idx, disk in enumerate(disk_info):
6283
      disk_index = idx + base_index
6284
      disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
6285
                              logical_id=(vgname, names[idx]),
6286
                              iv_name="disk/%d" % disk_index,
6287
                              mode=disk["mode"])
6288
      disks.append(disk_dev)
6289
  elif template_name == constants.DT_DRBD8:
6290
    if len(secondary_nodes) != 1:
6291
      raise errors.ProgrammerError("Wrong template configuration")
6292
    remote_node = secondary_nodes[0]
6293
    minors = lu.cfg.AllocateDRBDMinor(
6294
      [primary_node, remote_node] * len(disk_info), instance_name)
6295

    
6296
    names = []
6297
    for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
6298
                                               for i in range(disk_count)]):
6299
      names.append(lv_prefix + "_data")
6300
      names.append(lv_prefix + "_meta")
6301
    for idx, disk in enumerate(disk_info):
6302
      disk_index = idx + base_index
6303
      disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
6304
                                      disk["size"], names[idx*2:idx*2+2],
6305
                                      "disk/%d" % disk_index,
6306
                                      minors[idx*2], minors[idx*2+1])
6307
      disk_dev.mode = disk["mode"]
6308
      disks.append(disk_dev)
6309
  elif template_name == constants.DT_FILE:
6310
    if len(secondary_nodes) != 0:
6311
      raise errors.ProgrammerError("Wrong template configuration")
6312

    
6313
    _RequireFileStorage()
6314

    
6315
    for idx, disk in enumerate(disk_info):
6316
      disk_index = idx + base_index
6317
      disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
6318
                              iv_name="disk/%d" % disk_index,
6319
                              logical_id=(file_driver,
6320
                                          "%s/disk%d" % (file_storage_dir,
6321
                                                         disk_index)),
6322
                              mode=disk["mode"])
6323
      disks.append(disk_dev)
6324
  else:
6325
    raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
6326
  return disks
6327

    
6328

    
6329
def _GetInstanceInfoText(instance):
6330
  """Compute that text that should be added to the disk's metadata.
6331

6332
  """
6333
  return "originstname+%s" % instance.name
6334

    
6335

    
6336
def _CreateDisks(lu, instance, to_skip=None, target_node=None):
6337
  """Create all disks for an instance.
6338

6339
  This abstracts away some work from AddInstance.
6340

6341
  @type lu: L{LogicalUnit}
6342
  @param lu: the logical unit on whose behalf we execute
6343
  @type instance: L{objects.Instance}
6344
  @param instance: the instance whose disks we should create
6345
  @type to_skip: list
6346
  @param to_skip: list of indices to skip
6347
  @type target_node: string
6348
  @param target_node: if passed, overrides the target node for creation
6349
  @rtype: boolean
6350
  @return: the success of the creation
6351

6352
  """
6353
  info = _GetInstanceInfoText(instance)
6354
  if target_node is None:
6355
    pnode = instance.primary_node
6356
    all_nodes = instance.all_nodes
6357
  else:
6358
    pnode = target_node
6359
    all_nodes = [pnode]
6360

    
6361
  if instance.disk_template == constants.DT_FILE:
6362
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
6363
    result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
6364

    
6365
    result.Raise("Failed to create directory '%s' on"
6366
                 " node %s" % (file_storage_dir, pnode))
6367

    
6368
  # Note: this needs to be kept in sync with adding of disks in
6369
  # LUSetInstanceParams
6370
  for idx, device in enumerate(instance.disks):
6371
    if to_skip and idx in to_skip:
6372
      continue
6373
    logging.info("Creating volume %s for instance %s",
6374
                 device.iv_name, instance.name)
6375
    #HARDCODE
6376
    for node in all_nodes:
6377
      f_create = node == pnode
6378
      _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
6379

    
6380

    
6381
def _RemoveDisks(lu, instance, target_node=None):
6382
  """Remove all disks for an instance.
6383

6384
  This abstracts away some work from `AddInstance()` and
6385
  `RemoveInstance()`. Note that in case some of the devices couldn't
6386
  be removed, the removal will continue with the other ones (compare
6387
  with `_CreateDisks()`).
6388

6389
  @type lu: L{LogicalUnit}
6390
  @param lu: the logical unit on whose behalf we execute
6391
  @type instance: L{objects.Instance}
6392
  @param instance: the instance whose disks we should remove
6393
  @type target_node: string
6394
  @param target_node: used to override the node on which to remove the disks
6395
  @rtype: boolean
6396
  @return: the success of the removal
6397

6398
  """
6399
  logging.info("Removing block devices for instance %s", instance.name)
6400

    
6401
  all_result = True
6402
  for device in instance.disks:
6403
    if target_node:
6404
      edata = [(target_node, device)]
6405
    else:
6406
      edata = device.ComputeNodeTree(instance.primary_node)
6407
    for node, disk in edata:
6408
      lu.cfg.SetDiskID(disk, node)
6409
      msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
6410
      if msg:
6411
        lu.LogWarning("Could not remove block device %s on node %s,"
6412
                      " continuing anyway: %s", device.iv_name, node, msg)
6413
        all_result = False
6414

    
6415
  if instance.disk_template == constants.DT_FILE:
6416
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
6417
    if target_node:
6418
      tgt = target_node
6419
    else:
6420
      tgt = instance.primary_node
6421
    result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
6422
    if result.fail_msg:
6423
      lu.LogWarning("Could not remove directory '%s' on node %s: %s",
6424
                    file_storage_dir, instance.primary_node, result.fail_msg)
6425
      all_result = False
6426

    
6427
  return all_result
6428

    
6429

    
6430
def _ComputeDiskSize(disk_template, disks):
6431
  """Compute disk size requirements in the volume group
6432

6433
  """
6434
  # Required free disk space as a function of disk and swap space
6435
  req_size_dict = {
6436
    constants.DT_DISKLESS: None,
6437
    constants.DT_PLAIN: sum(d["size"] for d in disks),
6438
    # 128 MB are added for drbd metadata for each disk
6439
    constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
6440
    constants.DT_FILE: None,
6441
  }
6442

    
6443
  if disk_template not in req_size_dict:
6444
    raise errors.ProgrammerError("Disk template '%s' size requirement"
6445
                                 " is unknown" %  disk_template)
6446

    
6447
  return req_size_dict[disk_template]
6448

    
6449

    
6450
def _CheckHVParams(lu, nodenames, hvname, hvparams):
6451
  """Hypervisor parameter validation.
6452

6453
  This function abstract the hypervisor parameter validation to be
6454
  used in both instance create and instance modify.
6455

6456
  @type lu: L{LogicalUnit}
6457
  @param lu: the logical unit for which we check
6458
  @type nodenames: list
6459
  @param nodenames: the list of nodes on which we should check
6460
  @type hvname: string
6461
  @param hvname: the name of the hypervisor we should use
6462
  @type hvparams: dict
6463
  @param hvparams: the parameters which we need to check
6464
  @raise errors.OpPrereqError: if the parameters are not valid
6465

6466
  """
6467
  hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
6468
                                                  hvname,
6469
                                                  hvparams)
6470
  for node in nodenames:
6471
    info = hvinfo[node]
6472
    if info.offline:
6473
      continue
6474
    info.Raise("Hypervisor parameter validation failed on node %s" % node)
6475

    
6476

    
6477
def _CheckOSParams(lu, required, nodenames, osname, osparams):
6478
  """OS parameters validation.
6479

6480
  @type lu: L{LogicalUnit}
6481
  @param lu: the logical unit for which we check
6482
  @type required: boolean
6483
  @param required: whether the validation should fail if the OS is not
6484
      found
6485
  @type nodenames: list
6486
  @param nodenames: the list of nodes on which we should check
6487
  @type osname: string
6488
  @param osname: the name of the hypervisor we should use
6489
  @type osparams: dict
6490
  @param osparams: the parameters which we need to check
6491
  @raise errors.OpPrereqError: if the parameters are not valid
6492

6493
  """
6494
  result = lu.rpc.call_os_validate(required, nodenames, osname,
6495
                                   [constants.OS_VALIDATE_PARAMETERS],
6496
                                   osparams)
6497
  for node, nres in result.items():
6498
    # we don't check for offline cases since this should be run only
6499
    # against the master node and/or an instance's nodes
6500
    nres.Raise("OS Parameters validation failed on node %s" % node)
6501
    if not nres.payload:
6502
      lu.LogInfo("OS %s not found on node %s, validation skipped",
6503
                 osname, node)
6504

    
6505

    
6506
class LUCreateInstance(LogicalUnit):
6507
  """Create an instance.
6508

6509
  """
6510
  HPATH = "instance-add"
6511
  HTYPE = constants.HTYPE_INSTANCE
6512
  _OP_PARAMS = [
6513
    _PInstanceName,
6514
    ("mode", _NoDefault, _TElemOf(constants.INSTANCE_CREATE_MODES)),
6515
    ("start", True, _TBool),
6516
    ("wait_for_sync", True, _TBool),
6517
    ("ip_check", True, _TBool),
6518
    ("name_check", True, _TBool),
6519
    ("disks", _NoDefault, _TListOf(_TDict)),
6520
    ("nics", _NoDefault, _TListOf(_TDict)),
6521
    ("hvparams", _EmptyDict, _TDict),
6522
    ("beparams", _EmptyDict, _TDict),
6523
    ("osparams", _EmptyDict, _TDict),
6524
    ("no_install", None, _TMaybeBool),
6525
    ("os_type", None, _TMaybeString),
6526
    ("force_variant", False, _TBool),
6527
    ("source_handshake", None, _TOr(_TList, _TNone)),
6528
    ("source_x509_ca", None, _TMaybeString),
6529
    ("source_instance_name", None, _TMaybeString),
6530
    ("src_node", None, _TMaybeString),
6531
    ("src_path", None, _TMaybeString),
6532
    ("pnode", None, _TMaybeString),
6533
    ("snode", None, _TMaybeString),
6534
    ("iallocator", None, _TMaybeString),
6535
    ("hypervisor", None, _TMaybeString),
6536
    ("disk_template", _NoDefault, _CheckDiskTemplate),
6537
    ("identify_defaults", False, _TBool),
6538
    ("file_driver", None, _TOr(_TNone, _TElemOf(constants.FILE_DRIVER))),
6539
    ("file_storage_dir", None, _TMaybeString),
6540
    ]
6541
  REQ_BGL = False
6542

    
6543
  def CheckArguments(self):
6544
    """Check arguments.
6545

6546
    """
6547
    # do not require name_check to ease forward/backward compatibility
6548
    # for tools
6549
    if self.op.no_install and self.op.start:
6550
      self.LogInfo("No-installation mode selected, disabling startup")
6551
      self.op.start = False
6552
    # validate/normalize the instance name
6553
    self.op.instance_name = \
6554
      netutils.Hostname.GetNormalizedName(self.op.instance_name)
6555

    
6556
    if self.op.ip_check and not self.op.name_check:
6557
      # TODO: make the ip check more flexible and not depend on the name check
6558
      raise errors.OpPrereqError("Cannot do ip check without a name check",
6559
                                 errors.ECODE_INVAL)
6560

    
6561
    # check nics' parameter names
6562
    for nic in self.op.nics:
6563
      utils.ForceDictType(nic, constants.INIC_PARAMS_TYPES)
6564

    
6565
    # check disks. parameter names and consistent adopt/no-adopt strategy
6566
    has_adopt = has_no_adopt = False
6567
    for disk in self.op.disks:
6568
      utils.ForceDictType(disk, constants.IDISK_PARAMS_TYPES)
6569
      if "adopt" in disk:
6570
        has_adopt = True
6571
      else:
6572
        has_no_adopt = True
6573
    if has_adopt and has_no_adopt:
6574
      raise errors.OpPrereqError("Either all disks are adopted or none is",
6575
                                 errors.ECODE_INVAL)
6576
    if has_adopt:
6577
      if self.op.disk_template not in constants.DTS_MAY_ADOPT:
6578
        raise errors.OpPrereqError("Disk adoption is not supported for the"
6579
                                   " '%s' disk template" %
6580
                                   self.op.disk_template,
6581
                                   errors.ECODE_INVAL)
6582
      if self.op.iallocator is not None:
6583
        raise errors.OpPrereqError("Disk adoption not allowed with an"
6584
                                   " iallocator script", errors.ECODE_INVAL)
6585
      if self.op.mode == constants.INSTANCE_IMPORT:
6586
        raise errors.OpPrereqError("Disk adoption not allowed for"
6587
                                   " instance import", errors.ECODE_INVAL)
6588

    
6589
    self.adopt_disks = has_adopt
6590

    
6591
    # instance name verification
6592
    if self.op.name_check:
6593
      self.hostname1 = netutils.GetHostname(name=self.op.instance_name)
6594
      self.op.instance_name = self.hostname1.name
6595
      # used in CheckPrereq for ip ping check
6596
      self.check_ip = self.hostname1.ip
6597
    elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
6598
      raise errors.OpPrereqError("Remote imports require names to be checked" %
6599
                                 errors.ECODE_INVAL)
6600
    else:
6601
      self.check_ip = None
6602

    
6603
    # file storage checks
6604
    if (self.op.file_driver and
6605
        not self.op.file_driver in constants.FILE_DRIVER):
6606
      raise errors.OpPrereqError("Invalid file driver name '%s'" %
6607
                                 self.op.file_driver, errors.ECODE_INVAL)
6608

    
6609
    if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
6610
      raise errors.OpPrereqError("File storage directory path not absolute",
6611
                                 errors.ECODE_INVAL)
6612

    
6613
    ### Node/iallocator related checks
6614
    _CheckIAllocatorOrNode(self, "iallocator", "pnode")
6615

    
6616
    if self.op.pnode is not None:
6617
      if self.op.disk_template in constants.DTS_NET_MIRROR:
6618
        if self.op.snode is None:
6619
          raise errors.OpPrereqError("The networked disk templates need"
6620
                                     " a mirror node", errors.ECODE_INVAL)
6621
      elif self.op.snode:
6622
        self.LogWarning("Secondary node will be ignored on non-mirrored disk"
6623
                        " template")
6624
        self.op.snode = None
6625

    
6626
    self._cds = _GetClusterDomainSecret()
6627

    
6628
    if self.op.mode == constants.INSTANCE_IMPORT:
6629
      # On import force_variant must be True, because if we forced it at
6630
      # initial install, our only chance when importing it back is that it
6631
      # works again!
6632
      self.op.force_variant = True
6633

    
6634
      if self.op.no_install:
6635
        self.LogInfo("No-installation mode has no effect during import")
6636

    
6637
    elif self.op.mode == constants.INSTANCE_CREATE:
6638
      if self.op.os_type is None:
6639
        raise errors.OpPrereqError("No guest OS specified",
6640
                                   errors.ECODE_INVAL)
6641
      if self.op.disk_template is None:
6642
        raise errors.OpPrereqError("No disk template specified",
6643
                                   errors.ECODE_INVAL)
6644

    
6645
    elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
6646
      # Check handshake to ensure both clusters have the same domain secret
6647
      src_handshake = self.op.source_handshake
6648
      if not src_handshake:
6649
        raise errors.OpPrereqError("Missing source handshake",
6650
                                   errors.ECODE_INVAL)
6651

    
6652
      errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds,
6653
                                                           src_handshake)
6654
      if errmsg:
6655
        raise errors.OpPrereqError("Invalid handshake: %s" % errmsg,
6656
                                   errors.ECODE_INVAL)
6657

    
6658
      # Load and check source CA
6659
      self.source_x509_ca_pem = self.op.source_x509_ca
6660
      if not self.source_x509_ca_pem:
6661
        raise errors.OpPrereqError("Missing source X509 CA",
6662
                                   errors.ECODE_INVAL)
6663

    
6664
      try:
6665
        (cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem,
6666
                                                    self._cds)
6667
      except OpenSSL.crypto.Error, err:
6668
        raise errors.OpPrereqError("Unable to load source X509 CA (%s)" %
6669
                                   (err, ), errors.ECODE_INVAL)
6670

    
6671
      (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
6672
      if errcode is not None:
6673
        raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ),
6674
                                   errors.ECODE_INVAL)
6675

    
6676
      self.source_x509_ca = cert
6677

    
6678
      src_instance_name = self.op.source_instance_name
6679
      if not src_instance_name:
6680
        raise errors.OpPrereqError("Missing source instance name",
6681
                                   errors.ECODE_INVAL)
6682

    
6683
      self.source_instance_name = \
6684
          netutils.GetHostname(name=src_instance_name).name
6685

    
6686
    else:
6687
      raise errors.OpPrereqError("Invalid instance creation mode %r" %
6688
                                 self.op.mode, errors.ECODE_INVAL)
6689

    
6690
  def ExpandNames(self):
6691
    """ExpandNames for CreateInstance.
6692

6693
    Figure out the right locks for instance creation.
6694

6695
    """
6696
    self.needed_locks = {}
6697

    
6698
    instance_name = self.op.instance_name
6699
    # this is just a preventive check, but someone might still add this
6700
    # instance in the meantime, and creation will fail at lock-add time
6701
    if instance_name in self.cfg.GetInstanceList():
6702
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
6703
                                 instance_name, errors.ECODE_EXISTS)
6704

    
6705
    self.add_locks[locking.LEVEL_INSTANCE] = instance_name
6706

    
6707
    if self.op.iallocator:
6708
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6709
    else:
6710
      self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode)
6711
      nodelist = [self.op.pnode]
6712
      if self.op.snode is not None:
6713
        self.op.snode = _ExpandNodeName(self.cfg, self.op.snode)
6714
        nodelist.append(self.op.snode)
6715
      self.needed_locks[locking.LEVEL_NODE] = nodelist
6716

    
6717
    # in case of import lock the source node too
6718
    if self.op.mode == constants.INSTANCE_IMPORT:
6719
      src_node = self.op.src_node
6720
      src_path = self.op.src_path
6721

    
6722
      if src_path is None:
6723
        self.op.src_path = src_path = self.op.instance_name
6724

    
6725
      if src_node is None:
6726
        self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6727
        self.op.src_node = None
6728
        if os.path.isabs(src_path):
6729
          raise errors.OpPrereqError("Importing an instance from an absolute"
6730
                                     " path requires a source node option.",
6731
                                     errors.ECODE_INVAL)
6732
      else:
6733
        self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node)
6734
        if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
6735
          self.needed_locks[locking.LEVEL_NODE].append(src_node)
6736
        if not os.path.isabs(src_path):
6737
          self.op.src_path = src_path = \
6738
            utils.PathJoin(constants.EXPORT_DIR, src_path)
6739

    
6740
  def _RunAllocator(self):
6741
    """Run the allocator based on input opcode.
6742

6743
    """
6744
    nics = [n.ToDict() for n in self.nics]
6745
    ial = IAllocator(self.cfg, self.rpc,
6746
                     mode=constants.IALLOCATOR_MODE_ALLOC,
6747
                     name=self.op.instance_name,
6748
                     disk_template=self.op.disk_template,
6749
                     tags=[],
6750
                     os=self.op.os_type,
6751
                     vcpus=self.be_full[constants.BE_VCPUS],
6752
                     mem_size=self.be_full[constants.BE_MEMORY],
6753
                     disks=self.disks,
6754
                     nics=nics,
6755
                     hypervisor=self.op.hypervisor,
6756
                     )
6757

    
6758
    ial.Run(self.op.iallocator)
6759

    
6760
    if not ial.success:
6761
      raise errors.OpPrereqError("Can't compute nodes using"
6762
                                 " iallocator '%s': %s" %
6763
                                 (self.op.iallocator, ial.info),
6764
                                 errors.ECODE_NORES)
6765
    if len(ial.result) != ial.required_nodes:
6766
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
6767
                                 " of nodes (%s), required %s" %
6768
                                 (self.op.iallocator, len(ial.result),
6769
                                  ial.required_nodes), errors.ECODE_FAULT)
6770
    self.op.pnode = ial.result[0]
6771
    self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
6772
                 self.op.instance_name, self.op.iallocator,
6773
                 utils.CommaJoin(ial.result))
6774
    if ial.required_nodes == 2:
6775
      self.op.snode = ial.result[1]
6776

    
6777
  def BuildHooksEnv(self):
6778
    """Build hooks env.
6779

6780
    This runs on master, primary and secondary nodes of the instance.
6781

6782
    """
6783
    env = {
6784
      "ADD_MODE": self.op.mode,
6785
      }
6786
    if self.op.mode == constants.INSTANCE_IMPORT:
6787
      env["SRC_NODE"] = self.op.src_node
6788
      env["SRC_PATH"] = self.op.src_path
6789
      env["SRC_IMAGES"] = self.src_images
6790

    
6791
    env.update(_BuildInstanceHookEnv(
6792
      name=self.op.instance_name,
6793
      primary_node=self.op.pnode,
6794
      secondary_nodes=self.secondaries,
6795
      status=self.op.start,
6796
      os_type=self.op.os_type,
6797
      memory=self.be_full[constants.BE_MEMORY],
6798
      vcpus=self.be_full[constants.BE_VCPUS],
6799
      nics=_NICListToTuple(self, self.nics),
6800
      disk_template=self.op.disk_template,
6801
      disks=[(d["size"], d["mode"]) for d in self.disks],
6802
      bep=self.be_full,
6803
      hvp=self.hv_full,
6804
      hypervisor_name=self.op.hypervisor,
6805
    ))
6806

    
6807
    nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
6808
          self.secondaries)
6809
    return env, nl, nl
6810

    
6811
  def _ReadExportInfo(self):
6812
    """Reads the export information from disk.
6813

6814
    It will override the opcode source node and path with the actual
6815
    information, if these two were not specified before.
6816

6817
    @return: the export information
6818

6819
    """
6820
    assert self.op.mode == constants.INSTANCE_IMPORT
6821

    
6822
    src_node = self.op.src_node
6823
    src_path = self.op.src_path
6824

    
6825
    if src_node is None:
6826
      locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
6827
      exp_list = self.rpc.call_export_list(locked_nodes)
6828
      found = False
6829
      for node in exp_list:
6830
        if exp_list[node].fail_msg:
6831
          continue
6832
        if src_path in exp_list[node].payload:
6833
          found = True
6834
          self.op.src_node = src_node = node
6835
          self.op.src_path = src_path = utils.PathJoin(constants.EXPORT_DIR,
6836
                                                       src_path)
6837
          break
6838
      if not found:
6839
        raise errors.OpPrereqError("No export found for relative path %s" %
6840
                                    src_path, errors.ECODE_INVAL)
6841

    
6842
    _CheckNodeOnline(self, src_node)
6843
    result = self.rpc.call_export_info(src_node, src_path)
6844
    result.Raise("No export or invalid export found in dir %s" % src_path)
6845

    
6846
    export_info = objects.SerializableConfigParser.Loads(str(result.payload))
6847
    if not export_info.has_section(constants.INISECT_EXP):
6848
      raise errors.ProgrammerError("Corrupted export config",
6849
                                   errors.ECODE_ENVIRON)
6850

    
6851
    ei_version = export_info.get(constants.INISECT_EXP, "version")
6852
    if (int(ei_version) != constants.EXPORT_VERSION):
6853
      raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
6854
                                 (ei_version, constants.EXPORT_VERSION),
6855
                                 errors.ECODE_ENVIRON)
6856
    return export_info
6857

    
6858
  def _ReadExportParams(self, einfo):
6859
    """Use export parameters as defaults.
6860

6861
    In case the opcode doesn't specify (as in override) some instance
6862
    parameters, then try to use them from the export information, if
6863
    that declares them.
6864

6865
    """
6866
    self.op.os_type = einfo.get(constants.INISECT_EXP, "os")
6867

    
6868
    if self.op.disk_template is None:
6869
      if einfo.has_option(constants.INISECT_INS, "disk_template"):
6870
        self.op.disk_template = einfo.get(constants.INISECT_INS,
6871
                                          "disk_template")
6872
      else:
6873
        raise errors.OpPrereqError("No disk template specified and the export"
6874
                                   " is missing the disk_template information",
6875
                                   errors.ECODE_INVAL)
6876

    
6877
    if not self.op.disks:
6878
      if einfo.has_option(constants.INISECT_INS, "disk_count"):
6879
        disks = []
6880
        # TODO: import the disk iv_name too
6881
        for idx in range(einfo.getint(constants.INISECT_INS, "disk_count")):
6882
          disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
6883
          disks.append({"size": disk_sz})
6884
        self.op.disks = disks
6885
      else:
6886
        raise errors.OpPrereqError("No disk info specified and the export"
6887
                                   " is missing the disk information",
6888
                                   errors.ECODE_INVAL)
6889

    
6890
    if (not self.op.nics and
6891
        einfo.has_option(constants.INISECT_INS, "nic_count")):
6892
      nics = []
6893
      for idx in range(einfo.getint(constants.INISECT_INS, "nic_count")):
6894
        ndict = {}
6895
        for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]:
6896
          v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
6897
          ndict[name] = v
6898
        nics.append(ndict)
6899
      self.op.nics = nics
6900

    
6901
    if (self.op.hypervisor is None and
6902
        einfo.has_option(constants.INISECT_INS, "hypervisor")):
6903
      self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor")
6904
    if einfo.has_section(constants.INISECT_HYP):
6905
      # use the export parameters but do not override the ones
6906
      # specified by the user
6907
      for name, value in einfo.items(constants.INISECT_HYP):
6908
        if name not in self.op.hvparams:
6909
          self.op.hvparams[name] = value
6910

    
6911
    if einfo.has_section(constants.INISECT_BEP):
6912
      # use the parameters, without overriding
6913
      for name, value in einfo.items(constants.INISECT_BEP):
6914
        if name not in self.op.beparams:
6915
          self.op.beparams[name] = value
6916
    else:
6917
      # try to read the parameters old style, from the main section
6918
      for name in constants.BES_PARAMETERS:
6919
        if (name not in self.op.beparams and
6920
            einfo.has_option(constants.INISECT_INS, name)):
6921
          self.op.beparams[name] = einfo.get(constants.INISECT_INS, name)
6922

    
6923
    if einfo.has_section(constants.INISECT_OSP):
6924
      # use the parameters, without overriding
6925
      for name, value in einfo.items(constants.INISECT_OSP):
6926
        if name not in self.op.osparams:
6927
          self.op.osparams[name] = value
6928

    
6929
  def _RevertToDefaults(self, cluster):
6930
    """Revert the instance parameters to the default values.
6931

6932
    """
6933
    # hvparams
6934
    hv_defs = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, {})
6935
    for name in self.op.hvparams.keys():
6936
      if name in hv_defs and hv_defs[name] == self.op.hvparams[name]:
6937
        del self.op.hvparams[name]
6938
    # beparams
6939
    be_defs = cluster.SimpleFillBE({})
6940
    for name in self.op.beparams.keys():
6941
      if name in be_defs and be_defs[name] == self.op.beparams[name]:
6942
        del self.op.beparams[name]
6943
    # nic params
6944
    nic_defs = cluster.SimpleFillNIC({})
6945
    for nic in self.op.nics:
6946
      for name in constants.NICS_PARAMETERS:
6947
        if name in nic and name in nic_defs and nic[name] == nic_defs[name]:
6948
          del nic[name]
6949
    # osparams
6950
    os_defs = cluster.SimpleFillOS(self.op.os_type, {})
6951
    for name in self.op.osparams.keys():
6952
      if name in os_defs and os_defs[name] == self.op.osparams[name]:
6953
        del self.op.osparams[name]
6954

    
6955
  def CheckPrereq(self):
6956
    """Check prerequisites.
6957

6958
    """
6959
    if self.op.mode == constants.INSTANCE_IMPORT:
6960
      export_info = self._ReadExportInfo()
6961
      self._ReadExportParams(export_info)
6962

    
6963
    _CheckDiskTemplate(self.op.disk_template)
6964

    
6965
    if (not self.cfg.GetVGName() and
6966
        self.op.disk_template not in constants.DTS_NOT_LVM):
6967
      raise errors.OpPrereqError("Cluster does not support lvm-based"
6968
                                 " instances", errors.ECODE_STATE)
6969

    
6970
    if self.op.hypervisor is None:
6971
      self.op.hypervisor = self.cfg.GetHypervisorType()
6972

    
6973
    cluster = self.cfg.GetClusterInfo()
6974
    enabled_hvs = cluster.enabled_hypervisors
6975
    if self.op.hypervisor not in enabled_hvs:
6976
      raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
6977
                                 " cluster (%s)" % (self.op.hypervisor,
6978
                                  ",".join(enabled_hvs)),
6979
                                 errors.ECODE_STATE)
6980

    
6981
    # check hypervisor parameter syntax (locally)
6982
    utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
6983
    filled_hvp = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type,
6984
                                      self.op.hvparams)
6985
    hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
6986
    hv_type.CheckParameterSyntax(filled_hvp)
6987
    self.hv_full = filled_hvp
6988
    # check that we don't specify global parameters on an instance
6989
    _CheckGlobalHvParams(self.op.hvparams)
6990

    
6991
    # fill and remember the beparams dict
6992
    utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
6993
    self.be_full = cluster.SimpleFillBE(self.op.beparams)
6994

    
6995
    # build os parameters
6996
    self.os_full = cluster.SimpleFillOS(self.op.os_type, self.op.osparams)
6997

    
6998
    # now that hvp/bep are in final format, let's reset to defaults,
6999
    # if told to do so
7000
    if self.op.identify_defaults:
7001
      self._RevertToDefaults(cluster)
7002

    
7003
    # NIC buildup
7004
    self.nics = []
7005
    for idx, nic in enumerate(self.op.nics):
7006
      nic_mode_req = nic.get("mode", None)
7007
      nic_mode = nic_mode_req
7008
      if nic_mode is None:
7009
        nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
7010

    
7011
      # in routed mode, for the first nic, the default ip is 'auto'
7012
      if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
7013
        default_ip_mode = constants.VALUE_AUTO
7014
      else:
7015
        default_ip_mode = constants.VALUE_NONE
7016

    
7017
      # ip validity checks
7018
      ip = nic.get("ip", default_ip_mode)
7019
      if ip is None or ip.lower() == constants.VALUE_NONE:
7020
        nic_ip = None
7021
      elif ip.lower() == constants.VALUE_AUTO:
7022
        if not self.op.name_check:
7023
          raise errors.OpPrereqError("IP address set to auto but name checks"
7024
                                     " have been skipped",
7025
                                     errors.ECODE_INVAL)
7026
        nic_ip = self.hostname1.ip
7027
      else:
7028
        if not netutils.IPAddress.IsValid(ip):
7029
          raise errors.OpPrereqError("Invalid IP address '%s'" % ip,
7030
                                     errors.ECODE_INVAL)
7031
        nic_ip = ip
7032

    
7033
      # TODO: check the ip address for uniqueness
7034
      if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
7035
        raise errors.OpPrereqError("Routed nic mode requires an ip address",
7036
                                   errors.ECODE_INVAL)
7037

    
7038
      # MAC address verification
7039
      mac = nic.get("mac", constants.VALUE_AUTO)
7040
      if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7041
        mac = utils.NormalizeAndValidateMac(mac)
7042

    
7043
        try:
7044
          self.cfg.ReserveMAC(mac, self.proc.GetECId())
7045
        except errors.ReservationError:
7046
          raise errors.OpPrereqError("MAC address %s already in use"
7047
                                     " in cluster" % mac,
7048
                                     errors.ECODE_NOTUNIQUE)
7049

    
7050
      # bridge verification
7051
      bridge = nic.get("bridge", None)
7052
      link = nic.get("link", None)
7053
      if bridge and link:
7054
        raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
7055
                                   " at the same time", errors.ECODE_INVAL)
7056
      elif bridge and nic_mode == constants.NIC_MODE_ROUTED:
7057
        raise errors.OpPrereqError("Cannot pass 'bridge' on a routed nic",
7058
                                   errors.ECODE_INVAL)
7059
      elif bridge:
7060
        link = bridge
7061

    
7062
      nicparams = {}
7063
      if nic_mode_req:
7064
        nicparams[constants.NIC_MODE] = nic_mode_req
7065
      if link:
7066
        nicparams[constants.NIC_LINK] = link
7067

    
7068
      check_params = cluster.SimpleFillNIC(nicparams)
7069
      objects.NIC.CheckParameterSyntax(check_params)
7070
      self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
7071

    
7072
    # disk checks/pre-build
7073
    self.disks = []
7074
    for disk in self.op.disks:
7075
      mode = disk.get("mode", constants.DISK_RDWR)
7076
      if mode not in constants.DISK_ACCESS_SET:
7077
        raise errors.OpPrereqError("Invalid disk access mode '%s'" %
7078
                                   mode, errors.ECODE_INVAL)
7079
      size = disk.get("size", None)
7080
      if size is None:
7081
        raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
7082
      try:
7083
        size = int(size)
7084
      except (TypeError, ValueError):
7085
        raise errors.OpPrereqError("Invalid disk size '%s'" % size,
7086
                                   errors.ECODE_INVAL)
7087
      new_disk = {"size": size, "mode": mode}
7088
      if "adopt" in disk:
7089
        new_disk["adopt"] = disk["adopt"]
7090
      self.disks.append(new_disk)
7091

    
7092
    if self.op.mode == constants.INSTANCE_IMPORT:
7093

    
7094
      # Check that the new instance doesn't have less disks than the export
7095
      instance_disks = len(self.disks)
7096
      export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
7097
      if instance_disks < export_disks:
7098
        raise errors.OpPrereqError("Not enough disks to import."
7099
                                   " (instance: %d, export: %d)" %
7100
                                   (instance_disks, export_disks),
7101
                                   errors.ECODE_INVAL)
7102

    
7103
      disk_images = []
7104
      for idx in range(export_disks):
7105
        option = 'disk%d_dump' % idx
7106
        if export_info.has_option(constants.INISECT_INS, option):
7107
          # FIXME: are the old os-es, disk sizes, etc. useful?
7108
          export_name = export_info.get(constants.INISECT_INS, option)
7109
          image = utils.PathJoin(self.op.src_path, export_name)
7110
          disk_images.append(image)
7111
        else:
7112
          disk_images.append(False)
7113

    
7114
      self.src_images = disk_images
7115

    
7116
      old_name = export_info.get(constants.INISECT_INS, 'name')
7117
      try:
7118
        exp_nic_count = export_info.getint(constants.INISECT_INS, 'nic_count')
7119
      except (TypeError, ValueError), err:
7120
        raise errors.OpPrereqError("Invalid export file, nic_count is not"
7121
                                   " an integer: %s" % str(err),
7122
                                   errors.ECODE_STATE)
7123
      if self.op.instance_name == old_name:
7124
        for idx, nic in enumerate(self.nics):
7125
          if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
7126
            nic_mac_ini = 'nic%d_mac' % idx
7127
            nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
7128

    
7129
    # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
7130

    
7131
    # ip ping checks (we use the same ip that was resolved in ExpandNames)
7132
    if self.op.ip_check:
7133
      if netutils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
7134
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
7135
                                   (self.check_ip, self.op.instance_name),
7136
                                   errors.ECODE_NOTUNIQUE)
7137

    
7138
    #### mac address generation
7139
    # By generating here the mac address both the allocator and the hooks get
7140
    # the real final mac address rather than the 'auto' or 'generate' value.
7141
    # There is a race condition between the generation and the instance object
7142
    # creation, which means that we know the mac is valid now, but we're not
7143
    # sure it will be when we actually add the instance. If things go bad
7144
    # adding the instance will abort because of a duplicate mac, and the
7145
    # creation job will fail.
7146
    for nic in self.nics:
7147
      if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7148
        nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
7149

    
7150
    #### allocator run
7151

    
7152
    if self.op.iallocator is not None:
7153
      self._RunAllocator()
7154

    
7155
    #### node related checks
7156

    
7157
    # check primary node
7158
    self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
7159
    assert self.pnode is not None, \
7160
      "Cannot retrieve locked node %s" % self.op.pnode
7161
    if pnode.offline:
7162
      raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
7163
                                 pnode.name, errors.ECODE_STATE)
7164
    if pnode.drained:
7165
      raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
7166
                                 pnode.name, errors.ECODE_STATE)
7167

    
7168
    self.secondaries = []
7169

    
7170
    # mirror node verification
7171
    if self.op.disk_template in constants.DTS_NET_MIRROR:
7172
      if self.op.snode == pnode.name:
7173
        raise errors.OpPrereqError("The secondary node cannot be the"
7174
                                   " primary node.", errors.ECODE_INVAL)
7175
      _CheckNodeOnline(self, self.op.snode)
7176
      _CheckNodeNotDrained(self, self.op.snode)
7177
      self.secondaries.append(self.op.snode)
7178

    
7179
    nodenames = [pnode.name] + self.secondaries
7180

    
7181
    req_size = _ComputeDiskSize(self.op.disk_template,
7182
                                self.disks)
7183

    
7184
    # Check lv size requirements, if not adopting
7185
    if req_size is not None and not self.adopt_disks:
7186
      _CheckNodesFreeDisk(self, nodenames, req_size)
7187

    
7188
    if self.adopt_disks: # instead, we must check the adoption data
7189
      all_lvs = set([i["adopt"] for i in self.disks])
7190
      if len(all_lvs) != len(self.disks):
7191
        raise errors.OpPrereqError("Duplicate volume names given for adoption",
7192
                                   errors.ECODE_INVAL)
7193
      for lv_name in all_lvs:
7194
        try:
7195
          self.cfg.ReserveLV(lv_name, self.proc.GetECId())
7196
        except errors.ReservationError:
7197
          raise errors.OpPrereqError("LV named %s used by another instance" %
7198
                                     lv_name, errors.ECODE_NOTUNIQUE)
7199

    
7200
      node_lvs = self.rpc.call_lv_list([pnode.name],
7201
                                       self.cfg.GetVGName())[pnode.name]
7202
      node_lvs.Raise("Cannot get LV information from node %s" % pnode.name)
7203
      node_lvs = node_lvs.payload
7204
      delta = all_lvs.difference(node_lvs.keys())
7205
      if delta:
7206
        raise errors.OpPrereqError("Missing logical volume(s): %s" %
7207
                                   utils.CommaJoin(delta),
7208
                                   errors.ECODE_INVAL)
7209
      online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]]
7210
      if online_lvs:
7211
        raise errors.OpPrereqError("Online logical volumes found, cannot"
7212
                                   " adopt: %s" % utils.CommaJoin(online_lvs),
7213
                                   errors.ECODE_STATE)
7214
      # update the size of disk based on what is found
7215
      for dsk in self.disks:
7216
        dsk["size"] = int(float(node_lvs[dsk["adopt"]][0]))
7217

    
7218
    _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
7219

    
7220
    _CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant)
7221
    # check OS parameters (remotely)
7222
    _CheckOSParams(self, True, nodenames, self.op.os_type, self.os_full)
7223

    
7224
    _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
7225

    
7226
    # memory check on primary node
7227
    if self.op.start:
7228
      _CheckNodeFreeMemory(self, self.pnode.name,
7229
                           "creating instance %s" % self.op.instance_name,
7230
                           self.be_full[constants.BE_MEMORY],
7231
                           self.op.hypervisor)
7232

    
7233
    self.dry_run_result = list(nodenames)
7234

    
7235
  def Exec(self, feedback_fn):
7236
    """Create and add the instance to the cluster.
7237

7238
    """
7239
    instance = self.op.instance_name
7240
    pnode_name = self.pnode.name
7241

    
7242
    ht_kind = self.op.hypervisor
7243
    if ht_kind in constants.HTS_REQ_PORT:
7244
      network_port = self.cfg.AllocatePort()
7245
    else:
7246
      network_port = None
7247

    
7248
    if constants.ENABLE_FILE_STORAGE:
7249
      # this is needed because os.path.join does not accept None arguments
7250
      if self.op.file_storage_dir is None:
7251
        string_file_storage_dir = ""
7252
      else:
7253
        string_file_storage_dir = self.op.file_storage_dir
7254

    
7255
      # build the full file storage dir path
7256
      file_storage_dir = utils.PathJoin(self.cfg.GetFileStorageDir(),
7257
                                        string_file_storage_dir, instance)
7258
    else:
7259
      file_storage_dir = ""
7260

    
7261
    disks = _GenerateDiskTemplate(self,
7262
                                  self.op.disk_template,
7263
                                  instance, pnode_name,
7264
                                  self.secondaries,
7265
                                  self.disks,
7266
                                  file_storage_dir,
7267
                                  self.op.file_driver,
7268
                                  0)
7269

    
7270
    iobj = objects.Instance(name=instance, os=self.op.os_type,
7271
                            primary_node=pnode_name,
7272
                            nics=self.nics, disks=disks,
7273
                            disk_template=self.op.disk_template,
7274
                            admin_up=False,
7275
                            network_port=network_port,
7276
                            beparams=self.op.beparams,
7277
                            hvparams=self.op.hvparams,
7278
                            hypervisor=self.op.hypervisor,
7279
                            osparams=self.op.osparams,
7280
                            )
7281

    
7282
    if self.adopt_disks:
7283
      # rename LVs to the newly-generated names; we need to construct
7284
      # 'fake' LV disks with the old data, plus the new unique_id
7285
      tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks]
7286
      rename_to = []
7287
      for t_dsk, a_dsk in zip (tmp_disks, self.disks):
7288
        rename_to.append(t_dsk.logical_id)
7289
        t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk["adopt"])
7290
        self.cfg.SetDiskID(t_dsk, pnode_name)
7291
      result = self.rpc.call_blockdev_rename(pnode_name,
7292
                                             zip(tmp_disks, rename_to))
7293
      result.Raise("Failed to rename adoped LVs")
7294
    else:
7295
      feedback_fn("* creating instance disks...")
7296
      try:
7297
        _CreateDisks(self, iobj)
7298
      except errors.OpExecError:
7299
        self.LogWarning("Device creation failed, reverting...")
7300
        try:
7301
          _RemoveDisks(self, iobj)
7302
        finally:
7303
          self.cfg.ReleaseDRBDMinors(instance)
7304
          raise
7305

    
7306
    feedback_fn("adding instance %s to cluster config" % instance)
7307

    
7308
    self.cfg.AddInstance(iobj, self.proc.GetECId())
7309

    
7310
    # Declare that we don't want to remove the instance lock anymore, as we've
7311
    # added the instance to the config
7312
    del self.remove_locks[locking.LEVEL_INSTANCE]
7313
    # Unlock all the nodes
7314
    if self.op.mode == constants.INSTANCE_IMPORT:
7315
      nodes_keep = [self.op.src_node]
7316
      nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
7317
                       if node != self.op.src_node]
7318
      self.context.glm.release(locking.LEVEL_NODE, nodes_release)
7319
      self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
7320
    else:
7321
      self.context.glm.release(locking.LEVEL_NODE)
7322
      del self.acquired_locks[locking.LEVEL_NODE]
7323

    
7324
    if self.op.wait_for_sync:
7325
      disk_abort = not _WaitForSync(self, iobj)
7326
    elif iobj.disk_template in constants.DTS_NET_MIRROR:
7327
      # make sure the disks are not degraded (still sync-ing is ok)
7328
      time.sleep(15)
7329
      feedback_fn("* checking mirrors status")
7330
      disk_abort = not _WaitForSync(self, iobj, oneshot=True)
7331
    else:
7332
      disk_abort = False
7333

    
7334
    if disk_abort:
7335
      _RemoveDisks(self, iobj)
7336
      self.cfg.RemoveInstance(iobj.name)
7337
      # Make sure the instance lock gets removed
7338
      self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
7339
      raise errors.OpExecError("There are some degraded disks for"
7340
                               " this instance")
7341

    
7342
    if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks:
7343
      if self.op.mode == constants.INSTANCE_CREATE:
7344
        if not self.op.no_install:
7345
          feedback_fn("* running the instance OS create scripts...")
7346
          # FIXME: pass debug option from opcode to backend
7347
          result = self.rpc.call_instance_os_add(pnode_name, iobj, False,
7348
                                                 self.op.debug_level)
7349
          result.Raise("Could not add os for instance %s"
7350
                       " on node %s" % (instance, pnode_name))
7351

    
7352
      elif self.op.mode == constants.INSTANCE_IMPORT:
7353
        feedback_fn("* running the instance OS import scripts...")
7354

    
7355
        transfers = []
7356

    
7357
        for idx, image in enumerate(self.src_images):
7358
          if not image:
7359
            continue
7360

    
7361
          # FIXME: pass debug option from opcode to backend
7362
          dt = masterd.instance.DiskTransfer("disk/%s" % idx,
7363
                                             constants.IEIO_FILE, (image, ),
7364
                                             constants.IEIO_SCRIPT,
7365
                                             (iobj.disks[idx], idx),
7366
                                             None)
7367
          transfers.append(dt)
7368

    
7369
        import_result = \
7370
          masterd.instance.TransferInstanceData(self, feedback_fn,
7371
                                                self.op.src_node, pnode_name,
7372
                                                self.pnode.secondary_ip,
7373
                                                iobj, transfers)
7374
        if not compat.all(import_result):
7375
          self.LogWarning("Some disks for instance %s on node %s were not"
7376
                          " imported successfully" % (instance, pnode_name))
7377

    
7378
      elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
7379
        feedback_fn("* preparing remote import...")
7380
        connect_timeout = constants.RIE_CONNECT_TIMEOUT
7381
        timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
7382

    
7383
        disk_results = masterd.instance.RemoteImport(self, feedback_fn, iobj,
7384
                                                     self.source_x509_ca,
7385
                                                     self._cds, timeouts)
7386
        if not compat.all(disk_results):
7387
          # TODO: Should the instance still be started, even if some disks
7388
          # failed to import (valid for local imports, too)?
7389
          self.LogWarning("Some disks for instance %s on node %s were not"
7390
                          " imported successfully" % (instance, pnode_name))
7391

    
7392
        # Run rename script on newly imported instance
7393
        assert iobj.name == instance
7394
        feedback_fn("Running rename script for %s" % instance)
7395
        result = self.rpc.call_instance_run_rename(pnode_name, iobj,
7396
                                                   self.source_instance_name,
7397
                                                   self.op.debug_level)
7398
        if result.fail_msg:
7399
          self.LogWarning("Failed to run rename script for %s on node"
7400
                          " %s: %s" % (instance, pnode_name, result.fail_msg))
7401

    
7402
      else:
7403
        # also checked in the prereq part
7404
        raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
7405
                                     % self.op.mode)
7406

    
7407
    if self.op.start:
7408
      iobj.admin_up = True
7409
      self.cfg.Update(iobj, feedback_fn)
7410
      logging.info("Starting instance %s on node %s", instance, pnode_name)
7411
      feedback_fn("* starting instance...")
7412
      result = self.rpc.call_instance_start(pnode_name, iobj, None, None)
7413
      result.Raise("Could not start instance")
7414

    
7415
    return list(iobj.all_nodes)
7416

    
7417

    
7418
class LUConnectConsole(NoHooksLU):
7419
  """Connect to an instance's console.
7420

7421
  This is somewhat special in that it returns the command line that
7422
  you need to run on the master node in order to connect to the
7423
  console.
7424

7425
  """
7426
  _OP_PARAMS = [
7427
    _PInstanceName
7428
    ]
7429
  REQ_BGL = False
7430

    
7431
  def ExpandNames(self):
7432
    self._ExpandAndLockInstance()
7433

    
7434
  def CheckPrereq(self):
7435
    """Check prerequisites.
7436

7437
    This checks that the instance is in the cluster.
7438

7439
    """
7440
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7441
    assert self.instance is not None, \
7442
      "Cannot retrieve locked instance %s" % self.op.instance_name
7443
    _CheckNodeOnline(self, self.instance.primary_node)
7444

    
7445
  def Exec(self, feedback_fn):
7446
    """Connect to the console of an instance
7447

7448
    """
7449
    instance = self.instance
7450
    node = instance.primary_node
7451

    
7452
    node_insts = self.rpc.call_instance_list([node],
7453
                                             [instance.hypervisor])[node]
7454
    node_insts.Raise("Can't get node information from %s" % node)
7455

    
7456
    if instance.name not in node_insts.payload:
7457
      raise errors.OpExecError("Instance %s is not running." % instance.name)
7458

    
7459
    logging.debug("Connecting to console of %s on %s", instance.name, node)
7460

    
7461
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
7462
    cluster = self.cfg.GetClusterInfo()
7463
    # beparams and hvparams are passed separately, to avoid editing the
7464
    # instance and then saving the defaults in the instance itself.
7465
    hvparams = cluster.FillHV(instance)
7466
    beparams = cluster.FillBE(instance)
7467
    console_cmd = hyper.GetShellCommandForConsole(instance, hvparams, beparams)
7468

    
7469
    # build ssh cmdline
7470
    return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
7471

    
7472

    
7473
class LUReplaceDisks(LogicalUnit):
7474
  """Replace the disks of an instance.
7475

7476
  """
7477
  HPATH = "mirrors-replace"
7478
  HTYPE = constants.HTYPE_INSTANCE
7479
  _OP_PARAMS = [
7480
    _PInstanceName,
7481
    ("mode", _NoDefault, _TElemOf(constants.REPLACE_MODES)),
7482
    ("disks", _EmptyList, _TListOf(_TPositiveInt)),
7483
    ("remote_node", None, _TMaybeString),
7484
    ("iallocator", None, _TMaybeString),
7485
    ("early_release", False, _TBool),
7486
    ]
7487
  REQ_BGL = False
7488

    
7489
  def CheckArguments(self):
7490
    TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
7491
                                  self.op.iallocator)
7492

    
7493
  def ExpandNames(self):
7494
    self._ExpandAndLockInstance()
7495

    
7496
    if self.op.iallocator is not None:
7497
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7498

    
7499
    elif self.op.remote_node is not None:
7500
      remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
7501
      self.op.remote_node = remote_node
7502

    
7503
      # Warning: do not remove the locking of the new secondary here
7504
      # unless DRBD8.AddChildren is changed to work in parallel;
7505
      # currently it doesn't since parallel invocations of
7506
      # FindUnusedMinor will conflict
7507
      self.needed_locks[locking.LEVEL_NODE] = [remote_node]
7508
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
7509

    
7510
    else:
7511
      self.needed_locks[locking.LEVEL_NODE] = []
7512
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7513

    
7514
    self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
7515
                                   self.op.iallocator, self.op.remote_node,
7516
                                   self.op.disks, False, self.op.early_release)
7517

    
7518
    self.tasklets = [self.replacer]
7519

    
7520
  def DeclareLocks(self, level):
7521
    # If we're not already locking all nodes in the set we have to declare the
7522
    # instance's primary/secondary nodes.
7523
    if (level == locking.LEVEL_NODE and
7524
        self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
7525
      self._LockInstancesNodes()
7526

    
7527
  def BuildHooksEnv(self):
7528
    """Build hooks env.
7529

7530
    This runs on the master, the primary and all the secondaries.
7531

7532
    """
7533
    instance = self.replacer.instance
7534
    env = {
7535
      "MODE": self.op.mode,
7536
      "NEW_SECONDARY": self.op.remote_node,
7537
      "OLD_SECONDARY": instance.secondary_nodes[0],
7538
      }
7539
    env.update(_BuildInstanceHookEnvByObject(self, instance))
7540
    nl = [
7541
      self.cfg.GetMasterNode(),
7542
      instance.primary_node,
7543
      ]
7544
    if self.op.remote_node is not None:
7545
      nl.append(self.op.remote_node)
7546
    return env, nl, nl
7547

    
7548

    
7549
class TLReplaceDisks(Tasklet):
7550
  """Replaces disks for an instance.
7551

7552
  Note: Locking is not within the scope of this class.
7553

7554
  """
7555
  def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
7556
               disks, delay_iallocator, early_release):
7557
    """Initializes this class.
7558

7559
    """
7560
    Tasklet.__init__(self, lu)
7561

    
7562
    # Parameters
7563
    self.instance_name = instance_name
7564
    self.mode = mode
7565
    self.iallocator_name = iallocator_name
7566
    self.remote_node = remote_node
7567
    self.disks = disks
7568
    self.delay_iallocator = delay_iallocator
7569
    self.early_release = early_release
7570

    
7571
    # Runtime data
7572
    self.instance = None
7573
    self.new_node = None
7574
    self.target_node = None
7575
    self.other_node = None
7576
    self.remote_node_info = None
7577
    self.node_secondary_ip = None
7578

    
7579
  @staticmethod
7580
  def CheckArguments(mode, remote_node, iallocator):
7581
    """Helper function for users of this class.
7582

7583
    """
7584
    # check for valid parameter combination
7585
    if mode == constants.REPLACE_DISK_CHG:
7586
      if remote_node is None and iallocator is None:
7587
        raise errors.OpPrereqError("When changing the secondary either an"
7588
                                   " iallocator script must be used or the"
7589
                                   " new node given", errors.ECODE_INVAL)
7590

    
7591
      if remote_node is not None and iallocator is not None:
7592
        raise errors.OpPrereqError("Give either the iallocator or the new"
7593
                                   " secondary, not both", errors.ECODE_INVAL)
7594

    
7595
    elif remote_node is not None or iallocator is not None:
7596
      # Not replacing the secondary
7597
      raise errors.OpPrereqError("The iallocator and new node options can"
7598
                                 " only be used when changing the"
7599
                                 " secondary node", errors.ECODE_INVAL)
7600

    
7601
  @staticmethod
7602
  def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
7603
    """Compute a new secondary node using an IAllocator.
7604

7605
    """
7606
    ial = IAllocator(lu.cfg, lu.rpc,
7607
                     mode=constants.IALLOCATOR_MODE_RELOC,
7608
                     name=instance_name,
7609
                     relocate_from=relocate_from)
7610

    
7611
    ial.Run(iallocator_name)
7612

    
7613
    if not ial.success:
7614
      raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
7615
                                 " %s" % (iallocator_name, ial.info),
7616
                                 errors.ECODE_NORES)
7617

    
7618
    if len(ial.result) != ial.required_nodes:
7619
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
7620
                                 " of nodes (%s), required %s" %
7621
                                 (iallocator_name,
7622
                                  len(ial.result), ial.required_nodes),
7623
                                 errors.ECODE_FAULT)
7624

    
7625
    remote_node_name = ial.result[0]
7626

    
7627
    lu.LogInfo("Selected new secondary for instance '%s': %s",
7628
               instance_name, remote_node_name)
7629

    
7630
    return remote_node_name
7631

    
7632
  def _FindFaultyDisks(self, node_name):
7633
    return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
7634
                                    node_name, True)
7635

    
7636
  def CheckPrereq(self):
7637
    """Check prerequisites.
7638

7639
    This checks that the instance is in the cluster.
7640

7641
    """
7642
    self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
7643
    assert instance is not None, \
7644
      "Cannot retrieve locked instance %s" % self.instance_name
7645

    
7646
    if instance.disk_template != constants.DT_DRBD8:
7647
      raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
7648
                                 " instances", errors.ECODE_INVAL)
7649

    
7650
    if len(instance.secondary_nodes) != 1:
7651
      raise errors.OpPrereqError("The instance has a strange layout,"
7652
                                 " expected one secondary but found %d" %
7653
                                 len(instance.secondary_nodes),
7654
                                 errors.ECODE_FAULT)
7655

    
7656
    if not self.delay_iallocator:
7657
      self._CheckPrereq2()
7658

    
7659
  def _CheckPrereq2(self):
7660
    """Check prerequisites, second part.
7661

7662
    This function should always be part of CheckPrereq. It was separated and is
7663
    now called from Exec because during node evacuation iallocator was only
7664
    called with an unmodified cluster model, not taking planned changes into
7665
    account.
7666

7667
    """
7668
    instance = self.instance
7669
    secondary_node = instance.secondary_nodes[0]
7670

    
7671
    if self.iallocator_name is None:
7672
      remote_node = self.remote_node
7673
    else:
7674
      remote_node = self._RunAllocator(self.lu, self.iallocator_name,
7675
                                       instance.name, instance.secondary_nodes)
7676

    
7677
    if remote_node is not None:
7678
      self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
7679
      assert self.remote_node_info is not None, \
7680
        "Cannot retrieve locked node %s" % remote_node
7681
    else:
7682
      self.remote_node_info = None
7683

    
7684
    if remote_node == self.instance.primary_node:
7685
      raise errors.OpPrereqError("The specified node is the primary node of"
7686
                                 " the instance.", errors.ECODE_INVAL)
7687

    
7688
    if remote_node == secondary_node:
7689
      raise errors.OpPrereqError("The specified node is already the"
7690
                                 " secondary node of the instance.",
7691
                                 errors.ECODE_INVAL)
7692

    
7693
    if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
7694
                                    constants.REPLACE_DISK_CHG):
7695
      raise errors.OpPrereqError("Cannot specify disks to be replaced",
7696
                                 errors.ECODE_INVAL)
7697

    
7698
    if self.mode == constants.REPLACE_DISK_AUTO:
7699
      faulty_primary = self._FindFaultyDisks(instance.primary_node)
7700
      faulty_secondary = self._FindFaultyDisks(secondary_node)
7701

    
7702
      if faulty_primary and faulty_secondary:
7703
        raise errors.OpPrereqError("Instance %s has faulty disks on more than"
7704
                                   " one node and can not be repaired"
7705
                                   " automatically" % self.instance_name,
7706
                                   errors.ECODE_STATE)
7707

    
7708
      if faulty_primary:
7709
        self.disks = faulty_primary
7710
        self.target_node = instance.primary_node
7711
        self.other_node = secondary_node
7712
        check_nodes = [self.target_node, self.other_node]
7713
      elif faulty_secondary:
7714
        self.disks = faulty_secondary
7715
        self.target_node = secondary_node
7716
        self.other_node = instance.primary_node
7717
        check_nodes = [self.target_node, self.other_node]
7718
      else:
7719
        self.disks = []
7720
        check_nodes = []
7721

    
7722
    else:
7723
      # Non-automatic modes
7724
      if self.mode == constants.REPLACE_DISK_PRI:
7725
        self.target_node = instance.primary_node
7726
        self.other_node = secondary_node
7727
        check_nodes = [self.target_node, self.other_node]
7728

    
7729
      elif self.mode == constants.REPLACE_DISK_SEC:
7730
        self.target_node = secondary_node
7731
        self.other_node = instance.primary_node
7732
        check_nodes = [self.target_node, self.other_node]
7733

    
7734
      elif self.mode == constants.REPLACE_DISK_CHG:
7735
        self.new_node = remote_node
7736
        self.other_node = instance.primary_node
7737
        self.target_node = secondary_node
7738
        check_nodes = [self.new_node, self.other_node]
7739

    
7740
        _CheckNodeNotDrained(self.lu, remote_node)
7741

    
7742
        old_node_info = self.cfg.GetNodeInfo(secondary_node)
7743
        assert old_node_info is not None
7744
        if old_node_info.offline and not self.early_release:
7745
          # doesn't make sense to delay the release
7746
          self.early_release = True
7747
          self.lu.LogInfo("Old secondary %s is offline, automatically enabling"
7748
                          " early-release mode", secondary_node)
7749

    
7750
      else:
7751
        raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
7752
                                     self.mode)
7753

    
7754
      # If not specified all disks should be replaced
7755
      if not self.disks:
7756
        self.disks = range(len(self.instance.disks))
7757

    
7758
    for node in check_nodes:
7759
      _CheckNodeOnline(self.lu, node)
7760

    
7761
    # Check whether disks are valid
7762
    for disk_idx in self.disks:
7763
      instance.FindDisk(disk_idx)
7764

    
7765
    # Get secondary node IP addresses
7766
    node_2nd_ip = {}
7767

    
7768
    for node_name in [self.target_node, self.other_node, self.new_node]:
7769
      if node_name is not None:
7770
        node_2nd_ip[node_name] = self.cfg.GetNodeInfo(node_name).secondary_ip
7771

    
7772
    self.node_secondary_ip = node_2nd_ip
7773

    
7774
  def Exec(self, feedback_fn):
7775
    """Execute disk replacement.
7776

7777
    This dispatches the disk replacement to the appropriate handler.
7778

7779
    """
7780
    if self.delay_iallocator:
7781
      self._CheckPrereq2()
7782

    
7783
    if not self.disks:
7784
      feedback_fn("No disks need replacement")
7785
      return
7786

    
7787
    feedback_fn("Replacing disk(s) %s for %s" %
7788
                (utils.CommaJoin(self.disks), self.instance.name))
7789

    
7790
    activate_disks = (not self.instance.admin_up)
7791

    
7792
    # Activate the instance disks if we're replacing them on a down instance
7793
    if activate_disks:
7794
      _StartInstanceDisks(self.lu, self.instance, True)
7795

    
7796
    try:
7797
      # Should we replace the secondary node?
7798
      if self.new_node is not None:
7799
        fn = self._ExecDrbd8Secondary
7800
      else:
7801
        fn = self._ExecDrbd8DiskOnly
7802

    
7803
      return fn(feedback_fn)
7804

    
7805
    finally:
7806
      # Deactivate the instance disks if we're replacing them on a
7807
      # down instance
7808
      if activate_disks:
7809
        _SafeShutdownInstanceDisks(self.lu, self.instance)
7810

    
7811
  def _CheckVolumeGroup(self, nodes):
7812
    self.lu.LogInfo("Checking volume groups")
7813

    
7814
    vgname = self.cfg.GetVGName()
7815

    
7816
    # Make sure volume group exists on all involved nodes
7817
    results = self.rpc.call_vg_list(nodes)
7818
    if not results:
7819
      raise errors.OpExecError("Can't list volume groups on the nodes")
7820

    
7821
    for node in nodes:
7822
      res = results[node]
7823
      res.Raise("Error checking node %s" % node)
7824
      if vgname not in res.payload:
7825
        raise errors.OpExecError("Volume group '%s' not found on node %s" %
7826
                                 (vgname, node))
7827

    
7828
  def _CheckDisksExistence(self, nodes):
7829
    # Check disk existence
7830
    for idx, dev in enumerate(self.instance.disks):
7831
      if idx not in self.disks:
7832
        continue
7833

    
7834
      for node in nodes:
7835
        self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
7836
        self.cfg.SetDiskID(dev, node)
7837

    
7838
        result = self.rpc.call_blockdev_find(node, dev)
7839

    
7840
        msg = result.fail_msg
7841
        if msg or not result.payload:
7842
          if not msg:
7843
            msg = "disk not found"
7844
          raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
7845
                                   (idx, node, msg))
7846

    
7847
  def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
7848
    for idx, dev in enumerate(self.instance.disks):
7849
      if idx not in self.disks:
7850
        continue
7851

    
7852
      self.lu.LogInfo("Checking disk/%d consistency on node %s" %
7853
                      (idx, node_name))
7854

    
7855
      if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
7856
                                   ldisk=ldisk):
7857
        raise errors.OpExecError("Node %s has degraded storage, unsafe to"
7858
                                 " replace disks for instance %s" %
7859
                                 (node_name, self.instance.name))
7860

    
7861
  def _CreateNewStorage(self, node_name):
7862
    vgname = self.cfg.GetVGName()
7863
    iv_names = {}
7864

    
7865
    for idx, dev in enumerate(self.instance.disks):
7866
      if idx not in self.disks:
7867
        continue
7868

    
7869
      self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
7870

    
7871
      self.cfg.SetDiskID(dev, node_name)
7872

    
7873
      lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
7874
      names = _GenerateUniqueNames(self.lu, lv_names)
7875

    
7876
      lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
7877
                             logical_id=(vgname, names[0]))
7878
      lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
7879
                             logical_id=(vgname, names[1]))
7880

    
7881
      new_lvs = [lv_data, lv_meta]
7882
      old_lvs = dev.children
7883
      iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
7884

    
7885
      # we pass force_create=True to force the LVM creation
7886
      for new_lv in new_lvs:
7887
        _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
7888
                        _GetInstanceInfoText(self.instance), False)
7889

    
7890
    return iv_names
7891

    
7892
  def _CheckDevices(self, node_name, iv_names):
7893
    for name, (dev, _, _) in iv_names.iteritems():
7894
      self.cfg.SetDiskID(dev, node_name)
7895

    
7896
      result = self.rpc.call_blockdev_find(node_name, dev)
7897

    
7898
      msg = result.fail_msg
7899
      if msg or not result.payload:
7900
        if not msg:
7901
          msg = "disk not found"
7902
        raise errors.OpExecError("Can't find DRBD device %s: %s" %
7903
                                 (name, msg))
7904

    
7905
      if result.payload.is_degraded:
7906
        raise errors.OpExecError("DRBD device %s is degraded!" % name)
7907

    
7908
  def _RemoveOldStorage(self, node_name, iv_names):
7909
    for name, (_, old_lvs, _) in iv_names.iteritems():
7910
      self.lu.LogInfo("Remove logical volumes for %s" % name)
7911

    
7912
      for lv in old_lvs:
7913
        self.cfg.SetDiskID(lv, node_name)
7914

    
7915
        msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
7916
        if msg:
7917
          self.lu.LogWarning("Can't remove old LV: %s" % msg,
7918
                             hint="remove unused LVs manually")
7919

    
7920
  def _ReleaseNodeLock(self, node_name):
7921
    """Releases the lock for a given node."""
7922
    self.lu.context.glm.release(locking.LEVEL_NODE, node_name)
7923

    
7924
  def _ExecDrbd8DiskOnly(self, feedback_fn):
7925
    """Replace a disk on the primary or secondary for DRBD 8.
7926

7927
    The algorithm for replace is quite complicated:
7928

7929
      1. for each disk to be replaced:
7930

7931
        1. create new LVs on the target node with unique names
7932
        1. detach old LVs from the drbd device
7933
        1. rename old LVs to name_replaced.<time_t>
7934
        1. rename new LVs to old LVs
7935
        1. attach the new LVs (with the old names now) to the drbd device
7936

7937
      1. wait for sync across all devices
7938

7939
      1. for each modified disk:
7940

7941
        1. remove old LVs (which have the name name_replaces.<time_t>)
7942

7943
    Failures are not very well handled.
7944

7945
    """
7946
    steps_total = 6
7947

    
7948
    # Step: check device activation
7949
    self.lu.LogStep(1, steps_total, "Check device existence")
7950
    self._CheckDisksExistence([self.other_node, self.target_node])
7951
    self._CheckVolumeGroup([self.target_node, self.other_node])
7952

    
7953
    # Step: check other node consistency
7954
    self.lu.LogStep(2, steps_total, "Check peer consistency")
7955
    self._CheckDisksConsistency(self.other_node,
7956
                                self.other_node == self.instance.primary_node,
7957
                                False)
7958

    
7959
    # Step: create new storage
7960
    self.lu.LogStep(3, steps_total, "Allocate new storage")
7961
    iv_names = self._CreateNewStorage(self.target_node)
7962

    
7963
    # Step: for each lv, detach+rename*2+attach
7964
    self.lu.LogStep(4, steps_total, "Changing drbd configuration")
7965
    for dev, old_lvs, new_lvs in iv_names.itervalues():
7966
      self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
7967

    
7968
      result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
7969
                                                     old_lvs)
7970
      result.Raise("Can't detach drbd from local storage on node"
7971
                   " %s for device %s" % (self.target_node, dev.iv_name))
7972
      #dev.children = []
7973
      #cfg.Update(instance)
7974

    
7975
      # ok, we created the new LVs, so now we know we have the needed
7976
      # storage; as such, we proceed on the target node to rename
7977
      # old_lv to _old, and new_lv to old_lv; note that we rename LVs
7978
      # using the assumption that logical_id == physical_id (which in
7979
      # turn is the unique_id on that node)
7980

    
7981
      # FIXME(iustin): use a better name for the replaced LVs
7982
      temp_suffix = int(time.time())
7983
      ren_fn = lambda d, suff: (d.physical_id[0],
7984
                                d.physical_id[1] + "_replaced-%s" % suff)
7985

    
7986
      # Build the rename list based on what LVs exist on the node
7987
      rename_old_to_new = []
7988
      for to_ren in old_lvs:
7989
        result = self.rpc.call_blockdev_find(self.target_node, to_ren)
7990
        if not result.fail_msg and result.payload:
7991
          # device exists
7992
          rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
7993

    
7994
      self.lu.LogInfo("Renaming the old LVs on the target node")
7995
      result = self.rpc.call_blockdev_rename(self.target_node,
7996
                                             rename_old_to_new)
7997
      result.Raise("Can't rename old LVs on node %s" % self.target_node)
7998

    
7999
      # Now we rename the new LVs to the old LVs
8000
      self.lu.LogInfo("Renaming the new LVs on the target node")
8001
      rename_new_to_old = [(new, old.physical_id)
8002
                           for old, new in zip(old_lvs, new_lvs)]
8003
      result = self.rpc.call_blockdev_rename(self.target_node,
8004
                                             rename_new_to_old)
8005
      result.Raise("Can't rename new LVs on node %s" % self.target_node)
8006

    
8007
      for old, new in zip(old_lvs, new_lvs):
8008
        new.logical_id = old.logical_id
8009
        self.cfg.SetDiskID(new, self.target_node)
8010

    
8011
      for disk in old_lvs:
8012
        disk.logical_id = ren_fn(disk, temp_suffix)
8013
        self.cfg.SetDiskID(disk, self.target_node)
8014

    
8015
      # Now that the new lvs have the old name, we can add them to the device
8016
      self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
8017
      result = self.rpc.call_blockdev_addchildren(self.target_node, dev,
8018
                                                  new_lvs)
8019
      msg = result.fail_msg
8020
      if msg:
8021
        for new_lv in new_lvs:
8022
          msg2 = self.rpc.call_blockdev_remove(self.target_node,
8023
                                               new_lv).fail_msg
8024
          if msg2:
8025
            self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
8026
                               hint=("cleanup manually the unused logical"
8027
                                     "volumes"))
8028
        raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
8029

    
8030
      dev.children = new_lvs
8031

    
8032
      self.cfg.Update(self.instance, feedback_fn)
8033

    
8034
    cstep = 5
8035
    if self.early_release:
8036
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
8037
      cstep += 1
8038
      self._RemoveOldStorage(self.target_node, iv_names)
8039
      # WARNING: we release both node locks here, do not do other RPCs
8040
      # than WaitForSync to the primary node
8041
      self._ReleaseNodeLock([self.target_node, self.other_node])
8042

    
8043
    # Wait for sync
8044
    # This can fail as the old devices are degraded and _WaitForSync
8045
    # does a combined result over all disks, so we don't check its return value
8046
    self.lu.LogStep(cstep, steps_total, "Sync devices")
8047
    cstep += 1
8048
    _WaitForSync(self.lu, self.instance)
8049

    
8050
    # Check all devices manually
8051
    self._CheckDevices(self.instance.primary_node, iv_names)
8052

    
8053
    # Step: remove old storage
8054
    if not self.early_release:
8055
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
8056
      cstep += 1
8057
      self._RemoveOldStorage(self.target_node, iv_names)
8058

    
8059
  def _ExecDrbd8Secondary(self, feedback_fn):
8060
    """Replace the secondary node for DRBD 8.
8061

8062
    The algorithm for replace is quite complicated:
8063
      - for all disks of the instance:
8064
        - create new LVs on the new node with same names
8065
        - shutdown the drbd device on the old secondary
8066
        - disconnect the drbd network on the primary
8067
        - create the drbd device on the new secondary
8068
        - network attach the drbd on the primary, using an artifice:
8069
          the drbd code for Attach() will connect to the network if it
8070
          finds a device which is connected to the good local disks but
8071
          not network enabled
8072
      - wait for sync across all devices
8073
      - remove all disks from the old secondary
8074

8075
    Failures are not very well handled.
8076

8077
    """
8078
    steps_total = 6
8079

    
8080
    # Step: check device activation
8081
    self.lu.LogStep(1, steps_total, "Check device existence")
8082
    self._CheckDisksExistence([self.instance.primary_node])
8083
    self._CheckVolumeGroup([self.instance.primary_node])
8084

    
8085
    # Step: check other node consistency
8086
    self.lu.LogStep(2, steps_total, "Check peer consistency")
8087
    self._CheckDisksConsistency(self.instance.primary_node, True, True)
8088

    
8089
    # Step: create new storage
8090
    self.lu.LogStep(3, steps_total, "Allocate new storage")
8091
    for idx, dev in enumerate(self.instance.disks):
8092
      self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
8093
                      (self.new_node, idx))
8094
      # we pass force_create=True to force LVM creation
8095
      for new_lv in dev.children:
8096
        _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
8097
                        _GetInstanceInfoText(self.instance), False)
8098

    
8099
    # Step 4: dbrd minors and drbd setups changes
8100
    # after this, we must manually remove the drbd minors on both the
8101
    # error and the success paths
8102
    self.lu.LogStep(4, steps_total, "Changing drbd configuration")
8103
    minors = self.cfg.AllocateDRBDMinor([self.new_node
8104
                                         for dev in self.instance.disks],
8105
                                        self.instance.name)
8106
    logging.debug("Allocated minors %r", minors)
8107

    
8108
    iv_names = {}
8109
    for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
8110
      self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
8111
                      (self.new_node, idx))
8112
      # create new devices on new_node; note that we create two IDs:
8113
      # one without port, so the drbd will be activated without
8114
      # networking information on the new node at this stage, and one
8115
      # with network, for the latter activation in step 4
8116
      (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
8117
      if self.instance.primary_node == o_node1:
8118
        p_minor = o_minor1
8119
      else:
8120
        assert self.instance.primary_node == o_node2, "Three-node instance?"
8121
        p_minor = o_minor2
8122

    
8123
      new_alone_id = (self.instance.primary_node, self.new_node, None,
8124
                      p_minor, new_minor, o_secret)
8125
      new_net_id = (self.instance.primary_node, self.new_node, o_port,
8126
                    p_minor, new_minor, o_secret)
8127

    
8128
      iv_names[idx] = (dev, dev.children, new_net_id)
8129
      logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
8130
                    new_net_id)
8131
      new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
8132
                              logical_id=new_alone_id,
8133
                              children=dev.children,
8134
                              size=dev.size)
8135
      try:
8136
        _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
8137
                              _GetInstanceInfoText(self.instance), False)
8138
      except errors.GenericError:
8139
        self.cfg.ReleaseDRBDMinors(self.instance.name)
8140
        raise
8141

    
8142
    # We have new devices, shutdown the drbd on the old secondary
8143
    for idx, dev in enumerate(self.instance.disks):
8144
      self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
8145
      self.cfg.SetDiskID(dev, self.target_node)
8146
      msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
8147
      if msg:
8148
        self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
8149
                           "node: %s" % (idx, msg),
8150
                           hint=("Please cleanup this device manually as"
8151
                                 " soon as possible"))
8152

    
8153
    self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
8154
    result = self.rpc.call_drbd_disconnect_net([self.instance.primary_node],
8155
                                               self.node_secondary_ip,
8156
                                               self.instance.disks)\
8157
                                              [self.instance.primary_node]
8158

    
8159
    msg = result.fail_msg
8160
    if msg:
8161
      # detaches didn't succeed (unlikely)
8162
      self.cfg.ReleaseDRBDMinors(self.instance.name)
8163
      raise errors.OpExecError("Can't detach the disks from the network on"
8164
                               " old node: %s" % (msg,))
8165

    
8166
    # if we managed to detach at least one, we update all the disks of
8167
    # the instance to point to the new secondary
8168
    self.lu.LogInfo("Updating instance configuration")
8169
    for dev, _, new_logical_id in iv_names.itervalues():
8170
      dev.logical_id = new_logical_id
8171
      self.cfg.SetDiskID(dev, self.instance.primary_node)
8172

    
8173
    self.cfg.Update(self.instance, feedback_fn)
8174

    
8175
    # and now perform the drbd attach
8176
    self.lu.LogInfo("Attaching primary drbds to new secondary"
8177
                    " (standalone => connected)")
8178
    result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
8179
                                            self.new_node],
8180
                                           self.node_secondary_ip,
8181
                                           self.instance.disks,
8182
                                           self.instance.name,
8183
                                           False)
8184
    for to_node, to_result in result.items():
8185
      msg = to_result.fail_msg
8186
      if msg:
8187
        self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
8188
                           to_node, msg,
8189
                           hint=("please do a gnt-instance info to see the"
8190
                                 " status of disks"))
8191
    cstep = 5
8192
    if self.early_release:
8193
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
8194
      cstep += 1
8195
      self._RemoveOldStorage(self.target_node, iv_names)
8196
      # WARNING: we release all node locks here, do not do other RPCs
8197
      # than WaitForSync to the primary node
8198
      self._ReleaseNodeLock([self.instance.primary_node,
8199
                             self.target_node,
8200
                             self.new_node])
8201

    
8202
    # Wait for sync
8203
    # This can fail as the old devices are degraded and _WaitForSync
8204
    # does a combined result over all disks, so we don't check its return value
8205
    self.lu.LogStep(cstep, steps_total, "Sync devices")
8206
    cstep += 1
8207
    _WaitForSync(self.lu, self.instance)
8208

    
8209
    # Check all devices manually
8210
    self._CheckDevices(self.instance.primary_node, iv_names)
8211

    
8212
    # Step: remove old storage
8213
    if not self.early_release:
8214
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
8215
      self._RemoveOldStorage(self.target_node, iv_names)
8216

    
8217

    
8218
class LURepairNodeStorage(NoHooksLU):
8219
  """Repairs the volume group on a node.
8220

8221
  """
8222
  _OP_PARAMS = [
8223
    _PNodeName,
8224
    ("storage_type", _NoDefault, _CheckStorageType),
8225
    ("name", _NoDefault, _TNonEmptyString),
8226
    ("ignore_consistency", False, _TBool),
8227
    ]
8228
  REQ_BGL = False
8229

    
8230
  def CheckArguments(self):
8231
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
8232

    
8233
    storage_type = self.op.storage_type
8234

    
8235
    if (constants.SO_FIX_CONSISTENCY not in
8236
        constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
8237
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
8238
                                 " repaired" % storage_type,
8239
                                 errors.ECODE_INVAL)
8240

    
8241
  def ExpandNames(self):
8242
    self.needed_locks = {
8243
      locking.LEVEL_NODE: [self.op.node_name],
8244
      }
8245

    
8246
  def _CheckFaultyDisks(self, instance, node_name):
8247
    """Ensure faulty disks abort the opcode or at least warn."""
8248
    try:
8249
      if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
8250
                                  node_name, True):
8251
        raise errors.OpPrereqError("Instance '%s' has faulty disks on"
8252
                                   " node '%s'" % (instance.name, node_name),
8253
                                   errors.ECODE_STATE)
8254
    except errors.OpPrereqError, err:
8255
      if self.op.ignore_consistency:
8256
        self.proc.LogWarning(str(err.args[0]))
8257
      else:
8258
        raise
8259

    
8260
  def CheckPrereq(self):
8261
    """Check prerequisites.
8262

8263
    """
8264
    # Check whether any instance on this node has faulty disks
8265
    for inst in _GetNodeInstances(self.cfg, self.op.node_name):
8266
      if not inst.admin_up:
8267
        continue
8268
      check_nodes = set(inst.all_nodes)
8269
      check_nodes.discard(self.op.node_name)
8270
      for inst_node_name in check_nodes:
8271
        self._CheckFaultyDisks(inst, inst_node_name)
8272

    
8273
  def Exec(self, feedback_fn):
8274
    feedback_fn("Repairing storage unit '%s' on %s ..." %
8275
                (self.op.name, self.op.node_name))
8276

    
8277
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
8278
    result = self.rpc.call_storage_execute(self.op.node_name,
8279
                                           self.op.storage_type, st_args,
8280
                                           self.op.name,
8281
                                           constants.SO_FIX_CONSISTENCY)
8282
    result.Raise("Failed to repair storage unit '%s' on %s" %
8283
                 (self.op.name, self.op.node_name))
8284

    
8285

    
8286
class LUNodeEvacuationStrategy(NoHooksLU):
8287
  """Computes the node evacuation strategy.
8288

8289
  """
8290
  _OP_PARAMS = [
8291
    ("nodes", _NoDefault, _TListOf(_TNonEmptyString)),
8292
    ("remote_node", None, _TMaybeString),
8293
    ("iallocator", None, _TMaybeString),
8294
    ]
8295
  REQ_BGL = False
8296

    
8297
  def CheckArguments(self):
8298
    _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
8299

    
8300
  def ExpandNames(self):
8301
    self.op.nodes = _GetWantedNodes(self, self.op.nodes)
8302
    self.needed_locks = locks = {}
8303
    if self.op.remote_node is None:
8304
      locks[locking.LEVEL_NODE] = locking.ALL_SET
8305
    else:
8306
      self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8307
      locks[locking.LEVEL_NODE] = self.op.nodes + [self.op.remote_node]
8308

    
8309
  def Exec(self, feedback_fn):
8310
    if self.op.remote_node is not None:
8311
      instances = []
8312
      for node in self.op.nodes:
8313
        instances.extend(_GetNodeSecondaryInstances(self.cfg, node))
8314
      result = []
8315
      for i in instances:
8316
        if i.primary_node == self.op.remote_node:
8317
          raise errors.OpPrereqError("Node %s is the primary node of"
8318
                                     " instance %s, cannot use it as"
8319
                                     " secondary" %
8320
                                     (self.op.remote_node, i.name),
8321
                                     errors.ECODE_INVAL)
8322
        result.append([i.name, self.op.remote_node])
8323
    else:
8324
      ial = IAllocator(self.cfg, self.rpc,
8325
                       mode=constants.IALLOCATOR_MODE_MEVAC,
8326
                       evac_nodes=self.op.nodes)
8327
      ial.Run(self.op.iallocator, validate=True)
8328
      if not ial.success:
8329
        raise errors.OpExecError("No valid evacuation solution: %s" % ial.info,
8330
                                 errors.ECODE_NORES)
8331
      result = ial.result
8332
    return result
8333

    
8334

    
8335
class LUGrowDisk(LogicalUnit):
8336
  """Grow a disk of an instance.
8337

8338
  """
8339
  HPATH = "disk-grow"
8340
  HTYPE = constants.HTYPE_INSTANCE
8341
  _OP_PARAMS = [
8342
    _PInstanceName,
8343
    ("disk", _NoDefault, _TInt),
8344
    ("amount", _NoDefault, _TInt),
8345
    ("wait_for_sync", True, _TBool),
8346
    ]
8347
  REQ_BGL = False
8348

    
8349
  def ExpandNames(self):
8350
    self._ExpandAndLockInstance()
8351
    self.needed_locks[locking.LEVEL_NODE] = []
8352
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8353

    
8354
  def DeclareLocks(self, level):
8355
    if level == locking.LEVEL_NODE:
8356
      self._LockInstancesNodes()
8357

    
8358
  def BuildHooksEnv(self):
8359
    """Build hooks env.
8360

8361
    This runs on the master, the primary and all the secondaries.
8362

8363
    """
8364
    env = {
8365
      "DISK": self.op.disk,
8366
      "AMOUNT": self.op.amount,
8367
      }
8368
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
8369
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
8370
    return env, nl, nl
8371

    
8372
  def CheckPrereq(self):
8373
    """Check prerequisites.
8374

8375
    This checks that the instance is in the cluster.
8376

8377
    """
8378
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
8379
    assert instance is not None, \
8380
      "Cannot retrieve locked instance %s" % self.op.instance_name
8381
    nodenames = list(instance.all_nodes)
8382
    for node in nodenames:
8383
      _CheckNodeOnline(self, node)
8384

    
8385
    self.instance = instance
8386

    
8387
    if instance.disk_template not in constants.DTS_GROWABLE:
8388
      raise errors.OpPrereqError("Instance's disk layout does not support"
8389
                                 " growing.", errors.ECODE_INVAL)
8390

    
8391
    self.disk = instance.FindDisk(self.op.disk)
8392

    
8393
    if instance.disk_template != constants.DT_FILE:
8394
      # TODO: check the free disk space for file, when that feature will be
8395
      # supported
8396
      _CheckNodesFreeDisk(self, nodenames, self.op.amount)
8397

    
8398
  def Exec(self, feedback_fn):
8399
    """Execute disk grow.
8400

8401
    """
8402
    instance = self.instance
8403
    disk = self.disk
8404

    
8405
    disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk])
8406
    if not disks_ok:
8407
      raise errors.OpExecError("Cannot activate block device to grow")
8408

    
8409
    for node in instance.all_nodes:
8410
      self.cfg.SetDiskID(disk, node)
8411
      result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
8412
      result.Raise("Grow request failed to node %s" % node)
8413

    
8414
      # TODO: Rewrite code to work properly
8415
      # DRBD goes into sync mode for a short amount of time after executing the
8416
      # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
8417
      # calling "resize" in sync mode fails. Sleeping for a short amount of
8418
      # time is a work-around.
8419
      time.sleep(5)
8420

    
8421
    disk.RecordGrow(self.op.amount)
8422
    self.cfg.Update(instance, feedback_fn)
8423
    if self.op.wait_for_sync:
8424
      disk_abort = not _WaitForSync(self, instance, disks=[disk])
8425
      if disk_abort:
8426
        self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
8427
                             " status.\nPlease check the instance.")
8428
      if not instance.admin_up:
8429
        _SafeShutdownInstanceDisks(self, instance, disks=[disk])
8430
    elif not instance.admin_up:
8431
      self.proc.LogWarning("Not shutting down the disk even if the instance is"
8432
                           " not supposed to be running because no wait for"
8433
                           " sync mode was requested.")
8434

    
8435

    
8436
class LUQueryInstanceData(NoHooksLU):
8437
  """Query runtime instance data.
8438

8439
  """
8440
  _OP_PARAMS = [
8441
    ("instances", _EmptyList, _TListOf(_TNonEmptyString)),
8442
    ("static", False, _TBool),
8443
    ]
8444
  REQ_BGL = False
8445

    
8446
  def ExpandNames(self):
8447
    self.needed_locks = {}
8448
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
8449

    
8450
    if self.op.instances:
8451
      self.wanted_names = []
8452
      for name in self.op.instances:
8453
        full_name = _ExpandInstanceName(self.cfg, name)
8454
        self.wanted_names.append(full_name)
8455
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
8456
    else:
8457
      self.wanted_names = None
8458
      self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
8459

    
8460
    self.needed_locks[locking.LEVEL_NODE] = []
8461
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8462

    
8463
  def DeclareLocks(self, level):
8464
    if level == locking.LEVEL_NODE:
8465
      self._LockInstancesNodes()
8466

    
8467
  def CheckPrereq(self):
8468
    """Check prerequisites.
8469

8470
    This only checks the optional instance list against the existing names.
8471

8472
    """
8473
    if self.wanted_names is None:
8474
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
8475

    
8476
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
8477
                             in self.wanted_names]
8478

    
8479
  def _ComputeBlockdevStatus(self, node, instance_name, dev):
8480
    """Returns the status of a block device
8481

8482
    """
8483
    if self.op.static or not node:
8484
      return None
8485

    
8486
    self.cfg.SetDiskID(dev, node)
8487

    
8488
    result = self.rpc.call_blockdev_find(node, dev)
8489
    if result.offline:
8490
      return None
8491

    
8492
    result.Raise("Can't compute disk status for %s" % instance_name)
8493

    
8494
    status = result.payload
8495
    if status is None:
8496
      return None
8497

    
8498
    return (status.dev_path, status.major, status.minor,
8499
            status.sync_percent, status.estimated_time,
8500
            status.is_degraded, status.ldisk_status)
8501

    
8502
  def _ComputeDiskStatus(self, instance, snode, dev):
8503
    """Compute block device status.
8504

8505
    """
8506
    if dev.dev_type in constants.LDS_DRBD:
8507
      # we change the snode then (otherwise we use the one passed in)
8508
      if dev.logical_id[0] == instance.primary_node:
8509
        snode = dev.logical_id[1]
8510
      else:
8511
        snode = dev.logical_id[0]
8512

    
8513
    dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
8514
                                              instance.name, dev)
8515
    dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
8516

    
8517
    if dev.children:
8518
      dev_children = [self._ComputeDiskStatus(instance, snode, child)
8519
                      for child in dev.children]
8520
    else:
8521
      dev_children = []
8522

    
8523
    data = {
8524
      "iv_name": dev.iv_name,
8525
      "dev_type": dev.dev_type,
8526
      "logical_id": dev.logical_id,
8527
      "physical_id": dev.physical_id,
8528
      "pstatus": dev_pstatus,
8529
      "sstatus": dev_sstatus,
8530
      "children": dev_children,
8531
      "mode": dev.mode,
8532
      "size": dev.size,
8533
      }
8534

    
8535
    return data
8536

    
8537
  def Exec(self, feedback_fn):
8538
    """Gather and return data"""
8539
    result = {}
8540

    
8541
    cluster = self.cfg.GetClusterInfo()
8542

    
8543
    for instance in self.wanted_instances:
8544
      if not self.op.static:
8545
        remote_info = self.rpc.call_instance_info(instance.primary_node,
8546
                                                  instance.name,
8547
                                                  instance.hypervisor)
8548
        remote_info.Raise("Error checking node %s" % instance.primary_node)
8549
        remote_info = remote_info.payload
8550
        if remote_info and "state" in remote_info:
8551
          remote_state = "up"
8552
        else:
8553
          remote_state = "down"
8554
      else:
8555
        remote_state = None
8556
      if instance.admin_up:
8557
        config_state = "up"
8558
      else:
8559
        config_state = "down"
8560

    
8561
      disks = [self._ComputeDiskStatus(instance, None, device)
8562
               for device in instance.disks]
8563

    
8564
      idict = {
8565
        "name": instance.name,
8566
        "config_state": config_state,
8567
        "run_state": remote_state,
8568
        "pnode": instance.primary_node,
8569
        "snodes": instance.secondary_nodes,
8570
        "os": instance.os,
8571
        # this happens to be the same format used for hooks
8572
        "nics": _NICListToTuple(self, instance.nics),
8573
        "disk_template": instance.disk_template,
8574
        "disks": disks,
8575
        "hypervisor": instance.hypervisor,
8576
        "network_port": instance.network_port,
8577
        "hv_instance": instance.hvparams,
8578
        "hv_actual": cluster.FillHV(instance, skip_globals=True),
8579
        "be_instance": instance.beparams,
8580
        "be_actual": cluster.FillBE(instance),
8581
        "os_instance": instance.osparams,
8582
        "os_actual": cluster.SimpleFillOS(instance.os, instance.osparams),
8583
        "serial_no": instance.serial_no,
8584
        "mtime": instance.mtime,
8585
        "ctime": instance.ctime,
8586
        "uuid": instance.uuid,
8587
        }
8588

    
8589
      result[instance.name] = idict
8590

    
8591
    return result
8592

    
8593

    
8594
class LUSetInstanceParams(LogicalUnit):
8595
  """Modifies an instances's parameters.
8596

8597
  """
8598
  HPATH = "instance-modify"
8599
  HTYPE = constants.HTYPE_INSTANCE
8600
  _OP_PARAMS = [
8601
    _PInstanceName,
8602
    ("nics", _EmptyList, _TList),
8603
    ("disks", _EmptyList, _TList),
8604
    ("beparams", _EmptyDict, _TDict),
8605
    ("hvparams", _EmptyDict, _TDict),
8606
    ("disk_template", None, _TMaybeString),
8607
    ("remote_node", None, _TMaybeString),
8608
    ("os_name", None, _TMaybeString),
8609
    ("force_variant", False, _TBool),
8610
    ("osparams", None, _TOr(_TDict, _TNone)),
8611
    _PForce,
8612
    ]
8613
  REQ_BGL = False
8614

    
8615
  def CheckArguments(self):
8616
    if not (self.op.nics or self.op.disks or self.op.disk_template or
8617
            self.op.hvparams or self.op.beparams or self.op.os_name):
8618
      raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
8619

    
8620
    if self.op.hvparams:
8621
      _CheckGlobalHvParams(self.op.hvparams)
8622

    
8623
    # Disk validation
8624
    disk_addremove = 0
8625
    for disk_op, disk_dict in self.op.disks:
8626
      utils.ForceDictType(disk_dict, constants.IDISK_PARAMS_TYPES)
8627
      if disk_op == constants.DDM_REMOVE:
8628
        disk_addremove += 1
8629
        continue
8630
      elif disk_op == constants.DDM_ADD:
8631
        disk_addremove += 1
8632
      else:
8633
        if not isinstance(disk_op, int):
8634
          raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL)
8635
        if not isinstance(disk_dict, dict):
8636
          msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
8637
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
8638

    
8639
      if disk_op == constants.DDM_ADD:
8640
        mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
8641
        if mode not in constants.DISK_ACCESS_SET:
8642
          raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
8643
                                     errors.ECODE_INVAL)
8644
        size = disk_dict.get('size', None)
8645
        if size is None:
8646
          raise errors.OpPrereqError("Required disk parameter size missing",
8647
                                     errors.ECODE_INVAL)
8648
        try:
8649
          size = int(size)
8650
        except (TypeError, ValueError), err:
8651
          raise errors.OpPrereqError("Invalid disk size parameter: %s" %
8652
                                     str(err), errors.ECODE_INVAL)
8653
        disk_dict['size'] = size
8654
      else:
8655
        # modification of disk
8656
        if 'size' in disk_dict:
8657
          raise errors.OpPrereqError("Disk size change not possible, use"
8658
                                     " grow-disk", errors.ECODE_INVAL)
8659

    
8660
    if disk_addremove > 1:
8661
      raise errors.OpPrereqError("Only one disk add or remove operation"
8662
                                 " supported at a time", errors.ECODE_INVAL)
8663

    
8664
    if self.op.disks and self.op.disk_template is not None:
8665
      raise errors.OpPrereqError("Disk template conversion and other disk"
8666
                                 " changes not supported at the same time",
8667
                                 errors.ECODE_INVAL)
8668

    
8669
    if self.op.disk_template:
8670
      _CheckDiskTemplate(self.op.disk_template)
8671
      if (self.op.disk_template in constants.DTS_NET_MIRROR and
8672
          self.op.remote_node is None):
8673
        raise errors.OpPrereqError("Changing the disk template to a mirrored"
8674
                                   " one requires specifying a secondary node",
8675
                                   errors.ECODE_INVAL)
8676

    
8677
    # NIC validation
8678
    nic_addremove = 0
8679
    for nic_op, nic_dict in self.op.nics:
8680
      utils.ForceDictType(nic_dict, constants.INIC_PARAMS_TYPES)
8681
      if nic_op == constants.DDM_REMOVE:
8682
        nic_addremove += 1
8683
        continue
8684
      elif nic_op == constants.DDM_ADD:
8685
        nic_addremove += 1
8686
      else:
8687
        if not isinstance(nic_op, int):
8688
          raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL)
8689
        if not isinstance(nic_dict, dict):
8690
          msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
8691
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
8692

    
8693
      # nic_dict should be a dict
8694
      nic_ip = nic_dict.get('ip', None)
8695
      if nic_ip is not None:
8696
        if nic_ip.lower() == constants.VALUE_NONE:
8697
          nic_dict['ip'] = None
8698
        else:
8699
          if not netutils.IPAddress.IsValid(nic_ip):
8700
            raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip,
8701
                                       errors.ECODE_INVAL)
8702

    
8703
      nic_bridge = nic_dict.get('bridge', None)
8704
      nic_link = nic_dict.get('link', None)
8705
      if nic_bridge and nic_link:
8706
        raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
8707
                                   " at the same time", errors.ECODE_INVAL)
8708
      elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
8709
        nic_dict['bridge'] = None
8710
      elif nic_link and nic_link.lower() == constants.VALUE_NONE:
8711
        nic_dict['link'] = None
8712

    
8713
      if nic_op == constants.DDM_ADD:
8714
        nic_mac = nic_dict.get('mac', None)
8715
        if nic_mac is None:
8716
          nic_dict['mac'] = constants.VALUE_AUTO
8717

    
8718
      if 'mac' in nic_dict:
8719
        nic_mac = nic_dict['mac']
8720
        if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
8721
          nic_mac = utils.NormalizeAndValidateMac(nic_mac)
8722

    
8723
        if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
8724
          raise errors.OpPrereqError("'auto' is not a valid MAC address when"
8725
                                     " modifying an existing nic",
8726
                                     errors.ECODE_INVAL)
8727

    
8728
    if nic_addremove > 1:
8729
      raise errors.OpPrereqError("Only one NIC add or remove operation"
8730
                                 " supported at a time", errors.ECODE_INVAL)
8731

    
8732
  def ExpandNames(self):
8733
    self._ExpandAndLockInstance()
8734
    self.needed_locks[locking.LEVEL_NODE] = []
8735
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8736

    
8737
  def DeclareLocks(self, level):
8738
    if level == locking.LEVEL_NODE:
8739
      self._LockInstancesNodes()
8740
      if self.op.disk_template and self.op.remote_node:
8741
        self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8742
        self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)
8743

    
8744
  def BuildHooksEnv(self):
8745
    """Build hooks env.
8746

8747
    This runs on the master, primary and secondaries.
8748

8749
    """
8750
    args = dict()
8751
    if constants.BE_MEMORY in self.be_new:
8752
      args['memory'] = self.be_new[constants.BE_MEMORY]
8753
    if constants.BE_VCPUS in self.be_new:
8754
      args['vcpus'] = self.be_new[constants.BE_VCPUS]
8755
    # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
8756
    # information at all.
8757
    if self.op.nics:
8758
      args['nics'] = []
8759
      nic_override = dict(self.op.nics)
8760
      for idx, nic in enumerate(self.instance.nics):
8761
        if idx in nic_override:
8762
          this_nic_override = nic_override[idx]
8763
        else:
8764
          this_nic_override = {}
8765
        if 'ip' in this_nic_override:
8766
          ip = this_nic_override['ip']
8767
        else:
8768
          ip = nic.ip
8769
        if 'mac' in this_nic_override:
8770
          mac = this_nic_override['mac']
8771
        else:
8772
          mac = nic.mac
8773
        if idx in self.nic_pnew:
8774
          nicparams = self.nic_pnew[idx]
8775
        else:
8776
          nicparams = self.cluster.SimpleFillNIC(nic.nicparams)
8777
        mode = nicparams[constants.NIC_MODE]
8778
        link = nicparams[constants.NIC_LINK]
8779
        args['nics'].append((ip, mac, mode, link))
8780
      if constants.DDM_ADD in nic_override:
8781
        ip = nic_override[constants.DDM_ADD].get('ip', None)
8782
        mac = nic_override[constants.DDM_ADD]['mac']
8783
        nicparams = self.nic_pnew[constants.DDM_ADD]
8784
        mode = nicparams[constants.NIC_MODE]
8785
        link = nicparams[constants.NIC_LINK]
8786
        args['nics'].append((ip, mac, mode, link))
8787
      elif constants.DDM_REMOVE in nic_override:
8788
        del args['nics'][-1]
8789

    
8790
    env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
8791
    if self.op.disk_template:
8792
      env["NEW_DISK_TEMPLATE"] = self.op.disk_template
8793
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
8794
    return env, nl, nl
8795

    
8796
  def CheckPrereq(self):
8797
    """Check prerequisites.
8798

8799
    This only checks the instance list against the existing names.
8800

8801
    """
8802
    # checking the new params on the primary/secondary nodes
8803

    
8804
    instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
8805
    cluster = self.cluster = self.cfg.GetClusterInfo()
8806
    assert self.instance is not None, \
8807
      "Cannot retrieve locked instance %s" % self.op.instance_name
8808
    pnode = instance.primary_node
8809
    nodelist = list(instance.all_nodes)
8810

    
8811
    # OS change
8812
    if self.op.os_name and not self.op.force:
8813
      _CheckNodeHasOS(self, instance.primary_node, self.op.os_name,
8814
                      self.op.force_variant)
8815
      instance_os = self.op.os_name
8816
    else:
8817
      instance_os = instance.os
8818

    
8819
    if self.op.disk_template:
8820
      if instance.disk_template == self.op.disk_template:
8821
        raise errors.OpPrereqError("Instance already has disk template %s" %
8822
                                   instance.disk_template, errors.ECODE_INVAL)
8823

    
8824
      if (instance.disk_template,
8825
          self.op.disk_template) not in self._DISK_CONVERSIONS:
8826
        raise errors.OpPrereqError("Unsupported disk template conversion from"
8827
                                   " %s to %s" % (instance.disk_template,
8828
                                                  self.op.disk_template),
8829
                                   errors.ECODE_INVAL)
8830
      _CheckInstanceDown(self, instance, "cannot change disk template")
8831
      if self.op.disk_template in constants.DTS_NET_MIRROR:
8832
        if self.op.remote_node == pnode:
8833
          raise errors.OpPrereqError("Given new secondary node %s is the same"
8834
                                     " as the primary node of the instance" %
8835
                                     self.op.remote_node, errors.ECODE_STATE)
8836
        _CheckNodeOnline(self, self.op.remote_node)
8837
        _CheckNodeNotDrained(self, self.op.remote_node)
8838
        disks = [{"size": d.size} for d in instance.disks]
8839
        required = _ComputeDiskSize(self.op.disk_template, disks)
8840
        _CheckNodesFreeDisk(self, [self.op.remote_node], required)
8841

    
8842
    # hvparams processing
8843
    if self.op.hvparams:
8844
      hv_type = instance.hypervisor
8845
      i_hvdict = _GetUpdatedParams(instance.hvparams, self.op.hvparams)
8846
      utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES)
8847
      hv_new = cluster.SimpleFillHV(hv_type, instance.os, i_hvdict)
8848

    
8849
      # local check
8850
      hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new)
8851
      _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
8852
      self.hv_new = hv_new # the new actual values
8853
      self.hv_inst = i_hvdict # the new dict (without defaults)
8854
    else:
8855
      self.hv_new = self.hv_inst = {}
8856

    
8857
    # beparams processing
8858
    if self.op.beparams:
8859
      i_bedict = _GetUpdatedParams(instance.beparams, self.op.beparams,
8860
                                   use_none=True)
8861
      utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
8862
      be_new = cluster.SimpleFillBE(i_bedict)
8863
      self.be_new = be_new # the new actual values
8864
      self.be_inst = i_bedict # the new dict (without defaults)
8865
    else:
8866
      self.be_new = self.be_inst = {}
8867

    
8868
    # osparams processing
8869
    if self.op.osparams:
8870
      i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
8871
      _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
8872
      self.os_new = cluster.SimpleFillOS(instance_os, i_osdict)
8873
      self.os_inst = i_osdict # the new dict (without defaults)
8874
    else:
8875
      self.os_new = self.os_inst = {}
8876

    
8877
    self.warn = []
8878

    
8879
    if constants.BE_MEMORY in self.op.beparams and not self.op.force:
8880
      mem_check_list = [pnode]
8881
      if be_new[constants.BE_AUTO_BALANCE]:
8882
        # either we changed auto_balance to yes or it was from before
8883
        mem_check_list.extend(instance.secondary_nodes)
8884
      instance_info = self.rpc.call_instance_info(pnode, instance.name,
8885
                                                  instance.hypervisor)
8886
      nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
8887
                                         instance.hypervisor)
8888
      pninfo = nodeinfo[pnode]
8889
      msg = pninfo.fail_msg
8890
      if msg:
8891
        # Assume the primary node is unreachable and go ahead
8892
        self.warn.append("Can't get info from primary node %s: %s" %
8893
                         (pnode,  msg))
8894
      elif not isinstance(pninfo.payload.get('memory_free', None), int):
8895
        self.warn.append("Node data from primary node %s doesn't contain"
8896
                         " free memory information" % pnode)
8897
      elif instance_info.fail_msg:
8898
        self.warn.append("Can't get instance runtime information: %s" %
8899
                        instance_info.fail_msg)
8900
      else:
8901
        if instance_info.payload:
8902
          current_mem = int(instance_info.payload['memory'])
8903
        else:
8904
          # Assume instance not running
8905
          # (there is a slight race condition here, but it's not very probable,
8906
          # and we have no other way to check)
8907
          current_mem = 0
8908
        miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
8909
                    pninfo.payload['memory_free'])
8910
        if miss_mem > 0:
8911
          raise errors.OpPrereqError("This change will prevent the instance"
8912
                                     " from starting, due to %d MB of memory"
8913
                                     " missing on its primary node" % miss_mem,
8914
                                     errors.ECODE_NORES)
8915

    
8916
      if be_new[constants.BE_AUTO_BALANCE]:
8917
        for node, nres in nodeinfo.items():
8918
          if node not in instance.secondary_nodes:
8919
            continue
8920
          msg = nres.fail_msg
8921
          if msg:
8922
            self.warn.append("Can't get info from secondary node %s: %s" %
8923
                             (node, msg))
8924
          elif not isinstance(nres.payload.get('memory_free', None), int):
8925
            self.warn.append("Secondary node %s didn't return free"
8926
                             " memory information" % node)
8927
          elif be_new[constants.BE_MEMORY] > nres.payload['memory_free']:
8928
            self.warn.append("Not enough memory to failover instance to"
8929
                             " secondary node %s" % node)
8930

    
8931
    # NIC processing
8932
    self.nic_pnew = {}
8933
    self.nic_pinst = {}
8934
    for nic_op, nic_dict in self.op.nics:
8935
      if nic_op == constants.DDM_REMOVE:
8936
        if not instance.nics:
8937
          raise errors.OpPrereqError("Instance has no NICs, cannot remove",
8938
                                     errors.ECODE_INVAL)
8939
        continue
8940
      if nic_op != constants.DDM_ADD:
8941
        # an existing nic
8942
        if not instance.nics:
8943
          raise errors.OpPrereqError("Invalid NIC index %s, instance has"
8944
                                     " no NICs" % nic_op,
8945
                                     errors.ECODE_INVAL)
8946
        if nic_op < 0 or nic_op >= len(instance.nics):
8947
          raise errors.OpPrereqError("Invalid NIC index %s, valid values"
8948
                                     " are 0 to %d" %
8949
                                     (nic_op, len(instance.nics) - 1),
8950
                                     errors.ECODE_INVAL)
8951
        old_nic_params = instance.nics[nic_op].nicparams
8952
        old_nic_ip = instance.nics[nic_op].ip
8953
      else:
8954
        old_nic_params = {}
8955
        old_nic_ip = None
8956

    
8957
      update_params_dict = dict([(key, nic_dict[key])
8958
                                 for key in constants.NICS_PARAMETERS
8959
                                 if key in nic_dict])
8960

    
8961
      if 'bridge' in nic_dict:
8962
        update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
8963

    
8964
      new_nic_params = _GetUpdatedParams(old_nic_params,
8965
                                         update_params_dict)
8966
      utils.ForceDictType(new_nic_params, constants.NICS_PARAMETER_TYPES)
8967
      new_filled_nic_params = cluster.SimpleFillNIC(new_nic_params)
8968
      objects.NIC.CheckParameterSyntax(new_filled_nic_params)
8969
      self.nic_pinst[nic_op] = new_nic_params
8970
      self.nic_pnew[nic_op] = new_filled_nic_params
8971
      new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
8972

    
8973
      if new_nic_mode == constants.NIC_MODE_BRIDGED:
8974
        nic_bridge = new_filled_nic_params[constants.NIC_LINK]
8975
        msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
8976
        if msg:
8977
          msg = "Error checking bridges on node %s: %s" % (pnode, msg)
8978
          if self.op.force:
8979
            self.warn.append(msg)
8980
          else:
8981
            raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
8982
      if new_nic_mode == constants.NIC_MODE_ROUTED:
8983
        if 'ip' in nic_dict:
8984
          nic_ip = nic_dict['ip']
8985
        else:
8986
          nic_ip = old_nic_ip
8987
        if nic_ip is None:
8988
          raise errors.OpPrereqError('Cannot set the nic ip to None'
8989
                                     ' on a routed nic', errors.ECODE_INVAL)
8990
      if 'mac' in nic_dict:
8991
        nic_mac = nic_dict['mac']
8992
        if nic_mac is None:
8993
          raise errors.OpPrereqError('Cannot set the nic mac to None',
8994
                                     errors.ECODE_INVAL)
8995
        elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
8996
          # otherwise generate the mac
8997
          nic_dict['mac'] = self.cfg.GenerateMAC(self.proc.GetECId())
8998
        else:
8999
          # or validate/reserve the current one
9000
          try:
9001
            self.cfg.ReserveMAC(nic_mac, self.proc.GetECId())
9002
          except errors.ReservationError:
9003
            raise errors.OpPrereqError("MAC address %s already in use"
9004
                                       " in cluster" % nic_mac,
9005
                                       errors.ECODE_NOTUNIQUE)
9006

    
9007
    # DISK processing
9008
    if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
9009
      raise errors.OpPrereqError("Disk operations not supported for"
9010
                                 " diskless instances",
9011
                                 errors.ECODE_INVAL)
9012
    for disk_op, _ in self.op.disks:
9013
      if disk_op == constants.DDM_REMOVE:
9014
        if len(instance.disks) == 1:
9015
          raise errors.OpPrereqError("Cannot remove the last disk of"
9016
                                     " an instance", errors.ECODE_INVAL)
9017
        _CheckInstanceDown(self, instance, "cannot remove disks")
9018

    
9019
      if (disk_op == constants.DDM_ADD and
9020
          len(instance.nics) >= constants.MAX_DISKS):
9021
        raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
9022
                                   " add more" % constants.MAX_DISKS,
9023
                                   errors.ECODE_STATE)
9024
      if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
9025
        # an existing disk
9026
        if disk_op < 0 or disk_op >= len(instance.disks):
9027
          raise errors.OpPrereqError("Invalid disk index %s, valid values"
9028
                                     " are 0 to %d" %
9029
                                     (disk_op, len(instance.disks)),
9030
                                     errors.ECODE_INVAL)
9031

    
9032
    return
9033

    
9034
  def _ConvertPlainToDrbd(self, feedback_fn):
9035
    """Converts an instance from plain to drbd.
9036

9037
    """
9038
    feedback_fn("Converting template to drbd")
9039
    instance = self.instance
9040
    pnode = instance.primary_node
9041
    snode = self.op.remote_node
9042

    
9043
    # create a fake disk info for _GenerateDiskTemplate
9044
    disk_info = [{"size": d.size, "mode": d.mode} for d in instance.disks]
9045
    new_disks = _GenerateDiskTemplate(self, self.op.disk_template,
9046
                                      instance.name, pnode, [snode],
9047
                                      disk_info, None, None, 0)
9048
    info = _GetInstanceInfoText(instance)
9049
    feedback_fn("Creating aditional volumes...")
9050
    # first, create the missing data and meta devices
9051
    for disk in new_disks:
9052
      # unfortunately this is... not too nice
9053
      _CreateSingleBlockDev(self, pnode, instance, disk.children[1],
9054
                            info, True)
9055
      for child in disk.children:
9056
        _CreateSingleBlockDev(self, snode, instance, child, info, True)
9057
    # at this stage, all new LVs have been created, we can rename the
9058
    # old ones
9059
    feedback_fn("Renaming original volumes...")
9060
    rename_list = [(o, n.children[0].logical_id)
9061
                   for (o, n) in zip(instance.disks, new_disks)]
9062
    result = self.rpc.call_blockdev_rename(pnode, rename_list)
9063
    result.Raise("Failed to rename original LVs")
9064

    
9065
    feedback_fn("Initializing DRBD devices...")
9066
    # all child devices are in place, we can now create the DRBD devices
9067
    for disk in new_disks:
9068
      for node in [pnode, snode]:
9069
        f_create = node == pnode
9070
        _CreateSingleBlockDev(self, node, instance, disk, info, f_create)
9071

    
9072
    # at this point, the instance has been modified
9073
    instance.disk_template = constants.DT_DRBD8
9074
    instance.disks = new_disks
9075
    self.cfg.Update(instance, feedback_fn)
9076

    
9077
    # disks are created, waiting for sync
9078
    disk_abort = not _WaitForSync(self, instance)
9079
    if disk_abort:
9080
      raise errors.OpExecError("There are some degraded disks for"
9081
                               " this instance, please cleanup manually")
9082

    
9083
  def _ConvertDrbdToPlain(self, feedback_fn):
9084
    """Converts an instance from drbd to plain.
9085

9086
    """
9087
    instance = self.instance
9088
    assert len(instance.secondary_nodes) == 1
9089
    pnode = instance.primary_node
9090
    snode = instance.secondary_nodes[0]
9091
    feedback_fn("Converting template to plain")
9092

    
9093
    old_disks = instance.disks
9094
    new_disks = [d.children[0] for d in old_disks]
9095

    
9096
    # copy over size and mode
9097
    for parent, child in zip(old_disks, new_disks):
9098
      child.size = parent.size
9099
      child.mode = parent.mode
9100

    
9101
    # update instance structure
9102
    instance.disks = new_disks
9103
    instance.disk_template = constants.DT_PLAIN
9104
    self.cfg.Update(instance, feedback_fn)
9105

    
9106
    feedback_fn("Removing volumes on the secondary node...")
9107
    for disk in old_disks:
9108
      self.cfg.SetDiskID(disk, snode)
9109
      msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg
9110
      if msg:
9111
        self.LogWarning("Could not remove block device %s on node %s,"
9112
                        " continuing anyway: %s", disk.iv_name, snode, msg)
9113

    
9114
    feedback_fn("Removing unneeded volumes on the primary node...")
9115
    for idx, disk in enumerate(old_disks):
9116
      meta = disk.children[1]
9117
      self.cfg.SetDiskID(meta, pnode)
9118
      msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg
9119
      if msg:
9120
        self.LogWarning("Could not remove metadata for disk %d on node %s,"
9121
                        " continuing anyway: %s", idx, pnode, msg)
9122

    
9123

    
9124
  def Exec(self, feedback_fn):
9125
    """Modifies an instance.
9126

9127
    All parameters take effect only at the next restart of the instance.
9128

9129
    """
9130
    # Process here the warnings from CheckPrereq, as we don't have a
9131
    # feedback_fn there.
9132
    for warn in self.warn:
9133
      feedback_fn("WARNING: %s" % warn)
9134

    
9135
    result = []
9136
    instance = self.instance
9137
    # disk changes
9138
    for disk_op, disk_dict in self.op.disks:
9139
      if disk_op == constants.DDM_REMOVE:
9140
        # remove the last disk
9141
        device = instance.disks.pop()
9142
        device_idx = len(instance.disks)
9143
        for node, disk in device.ComputeNodeTree(instance.primary_node):
9144
          self.cfg.SetDiskID(disk, node)
9145
          msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
9146
          if msg:
9147
            self.LogWarning("Could not remove disk/%d on node %s: %s,"
9148
                            " continuing anyway", device_idx, node, msg)
9149
        result.append(("disk/%d" % device_idx, "remove"))
9150
      elif disk_op == constants.DDM_ADD:
9151
        # add a new disk
9152
        if instance.disk_template == constants.DT_FILE:
9153
          file_driver, file_path = instance.disks[0].logical_id
9154
          file_path = os.path.dirname(file_path)
9155
        else:
9156
          file_driver = file_path = None
9157
        disk_idx_base = len(instance.disks)
9158
        new_disk = _GenerateDiskTemplate(self,
9159
                                         instance.disk_template,
9160
                                         instance.name, instance.primary_node,
9161
                                         instance.secondary_nodes,
9162
                                         [disk_dict],
9163
                                         file_path,
9164
                                         file_driver,
9165
                                         disk_idx_base)[0]
9166
        instance.disks.append(new_disk)
9167
        info = _GetInstanceInfoText(instance)
9168

    
9169
        logging.info("Creating volume %s for instance %s",
9170
                     new_disk.iv_name, instance.name)
9171
        # Note: this needs to be kept in sync with _CreateDisks
9172
        #HARDCODE
9173
        for node in instance.all_nodes:
9174
          f_create = node == instance.primary_node
9175
          try:
9176
            _CreateBlockDev(self, node, instance, new_disk,
9177
                            f_create, info, f_create)
9178
          except errors.OpExecError, err:
9179
            self.LogWarning("Failed to create volume %s (%s) on"
9180
                            " node %s: %s",
9181
                            new_disk.iv_name, new_disk, node, err)
9182
        result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
9183
                       (new_disk.size, new_disk.mode)))
9184
      else:
9185
        # change a given disk
9186
        instance.disks[disk_op].mode = disk_dict['mode']
9187
        result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
9188

    
9189
    if self.op.disk_template:
9190
      r_shut = _ShutdownInstanceDisks(self, instance)
9191
      if not r_shut:
9192
        raise errors.OpExecError("Cannot shutdow instance disks, unable to"
9193
                                 " proceed with disk template conversion")
9194
      mode = (instance.disk_template, self.op.disk_template)
9195
      try:
9196
        self._DISK_CONVERSIONS[mode](self, feedback_fn)
9197
      except:
9198
        self.cfg.ReleaseDRBDMinors(instance.name)
9199
        raise
9200
      result.append(("disk_template", self.op.disk_template))
9201

    
9202
    # NIC changes
9203
    for nic_op, nic_dict in self.op.nics:
9204
      if nic_op == constants.DDM_REMOVE:
9205
        # remove the last nic
9206
        del instance.nics[-1]
9207
        result.append(("nic.%d" % len(instance.nics), "remove"))
9208
      elif nic_op == constants.DDM_ADD:
9209
        # mac and bridge should be set, by now
9210
        mac = nic_dict['mac']
9211
        ip = nic_dict.get('ip', None)
9212
        nicparams = self.nic_pinst[constants.DDM_ADD]
9213
        new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
9214
        instance.nics.append(new_nic)
9215
        result.append(("nic.%d" % (len(instance.nics) - 1),
9216
                       "add:mac=%s,ip=%s,mode=%s,link=%s" %
9217
                       (new_nic.mac, new_nic.ip,
9218
                        self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
9219
                        self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
9220
                       )))
9221
      else:
9222
        for key in 'mac', 'ip':
9223
          if key in nic_dict:
9224
            setattr(instance.nics[nic_op], key, nic_dict[key])
9225
        if nic_op in self.nic_pinst:
9226
          instance.nics[nic_op].nicparams = self.nic_pinst[nic_op]
9227
        for key, val in nic_dict.iteritems():
9228
          result.append(("nic.%s/%d" % (key, nic_op), val))
9229

    
9230
    # hvparams changes
9231
    if self.op.hvparams:
9232
      instance.hvparams = self.hv_inst
9233
      for key, val in self.op.hvparams.iteritems():
9234
        result.append(("hv/%s" % key, val))
9235

    
9236
    # beparams changes
9237
    if self.op.beparams:
9238
      instance.beparams = self.be_inst
9239
      for key, val in self.op.beparams.iteritems():
9240
        result.append(("be/%s" % key, val))
9241

    
9242
    # OS change
9243
    if self.op.os_name:
9244
      instance.os = self.op.os_name
9245

    
9246
    # osparams changes
9247
    if self.op.osparams:
9248
      instance.osparams = self.os_inst
9249
      for key, val in self.op.osparams.iteritems():
9250
        result.append(("os/%s" % key, val))
9251

    
9252
    self.cfg.Update(instance, feedback_fn)
9253

    
9254
    return result
9255

    
9256
  _DISK_CONVERSIONS = {
9257
    (constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd,
9258
    (constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain,
9259
    }
9260

    
9261

    
9262
class LUQueryExports(NoHooksLU):
9263
  """Query the exports list
9264

9265
  """
9266
  _OP_PARAMS = [
9267
    ("nodes", _EmptyList, _TListOf(_TNonEmptyString)),
9268
    ("use_locking", False, _TBool),
9269
    ]
9270
  REQ_BGL = False
9271

    
9272
  def ExpandNames(self):
9273
    self.needed_locks = {}
9274
    self.share_locks[locking.LEVEL_NODE] = 1
9275
    if not self.op.nodes:
9276
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9277
    else:
9278
      self.needed_locks[locking.LEVEL_NODE] = \
9279
        _GetWantedNodes(self, self.op.nodes)
9280

    
9281
  def Exec(self, feedback_fn):
9282
    """Compute the list of all the exported system images.
9283

9284
    @rtype: dict
9285
    @return: a dictionary with the structure node->(export-list)
9286
        where export-list is a list of the instances exported on
9287
        that node.
9288

9289
    """
9290
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
9291
    rpcresult = self.rpc.call_export_list(self.nodes)
9292
    result = {}
9293
    for node in rpcresult:
9294
      if rpcresult[node].fail_msg:
9295
        result[node] = False
9296
      else:
9297
        result[node] = rpcresult[node].payload
9298

    
9299
    return result
9300

    
9301

    
9302
class LUPrepareExport(NoHooksLU):
9303
  """Prepares an instance for an export and returns useful information.
9304

9305
  """
9306
  _OP_PARAMS = [
9307
    _PInstanceName,
9308
    ("mode", _NoDefault, _TElemOf(constants.EXPORT_MODES)),
9309
    ]
9310
  REQ_BGL = False
9311

    
9312
  def ExpandNames(self):
9313
    self._ExpandAndLockInstance()
9314

    
9315
  def CheckPrereq(self):
9316
    """Check prerequisites.
9317

9318
    """
9319
    instance_name = self.op.instance_name
9320

    
9321
    self.instance = self.cfg.GetInstanceInfo(instance_name)
9322
    assert self.instance is not None, \
9323
          "Cannot retrieve locked instance %s" % self.op.instance_name
9324
    _CheckNodeOnline(self, self.instance.primary_node)
9325

    
9326
    self._cds = _GetClusterDomainSecret()
9327

    
9328
  def Exec(self, feedback_fn):
9329
    """Prepares an instance for an export.
9330

9331
    """
9332
    instance = self.instance
9333

    
9334
    if self.op.mode == constants.EXPORT_MODE_REMOTE:
9335
      salt = utils.GenerateSecret(8)
9336

    
9337
      feedback_fn("Generating X509 certificate on %s" % instance.primary_node)
9338
      result = self.rpc.call_x509_cert_create(instance.primary_node,
9339
                                              constants.RIE_CERT_VALIDITY)
9340
      result.Raise("Can't create X509 key and certificate on %s" % result.node)
9341

    
9342
      (name, cert_pem) = result.payload
9343

    
9344
      cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
9345
                                             cert_pem)
9346

    
9347
      return {
9348
        "handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds),
9349
        "x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt),
9350
                          salt),
9351
        "x509_ca": utils.SignX509Certificate(cert, self._cds, salt),
9352
        }
9353

    
9354
    return None
9355

    
9356

    
9357
class LUExportInstance(LogicalUnit):
9358
  """Export an instance to an image in the cluster.
9359

9360
  """
9361
  HPATH = "instance-export"
9362
  HTYPE = constants.HTYPE_INSTANCE
9363
  _OP_PARAMS = [
9364
    _PInstanceName,
9365
    ("target_node", _NoDefault, _TOr(_TNonEmptyString, _TList)),
9366
    ("shutdown", True, _TBool),
9367
    _PShutdownTimeout,
9368
    ("remove_instance", False, _TBool),
9369
    ("ignore_remove_failures", False, _TBool),
9370
    ("mode", constants.EXPORT_MODE_LOCAL, _TElemOf(constants.EXPORT_MODES)),
9371
    ("x509_key_name", None, _TOr(_TList, _TNone)),
9372
    ("destination_x509_ca", None, _TMaybeString),
9373
    ]
9374
  REQ_BGL = False
9375

    
9376
  def CheckArguments(self):
9377
    """Check the arguments.
9378

9379
    """
9380
    self.x509_key_name = self.op.x509_key_name
9381
    self.dest_x509_ca_pem = self.op.destination_x509_ca
9382

    
9383
    if self.op.remove_instance and not self.op.shutdown:
9384
      raise errors.OpPrereqError("Can not remove instance without shutting it"
9385
                                 " down before")
9386

    
9387
    if self.op.mode == constants.EXPORT_MODE_REMOTE:
9388
      if not self.x509_key_name:
9389
        raise errors.OpPrereqError("Missing X509 key name for encryption",
9390
                                   errors.ECODE_INVAL)
9391

    
9392
      if not self.dest_x509_ca_pem:
9393
        raise errors.OpPrereqError("Missing destination X509 CA",
9394
                                   errors.ECODE_INVAL)
9395

    
9396
  def ExpandNames(self):
9397
    self._ExpandAndLockInstance()
9398

    
9399
    # Lock all nodes for local exports
9400
    if self.op.mode == constants.EXPORT_MODE_LOCAL:
9401
      # FIXME: lock only instance primary and destination node
9402
      #
9403
      # Sad but true, for now we have do lock all nodes, as we don't know where
9404
      # the previous export might be, and in this LU we search for it and
9405
      # remove it from its current node. In the future we could fix this by:
9406
      #  - making a tasklet to search (share-lock all), then create the
9407
      #    new one, then one to remove, after
9408
      #  - removing the removal operation altogether
9409
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9410

    
9411
  def DeclareLocks(self, level):
9412
    """Last minute lock declaration."""
9413
    # All nodes are locked anyway, so nothing to do here.
9414

    
9415
  def BuildHooksEnv(self):
9416
    """Build hooks env.
9417

9418
    This will run on the master, primary node and target node.
9419

9420
    """
9421
    env = {
9422
      "EXPORT_MODE": self.op.mode,
9423
      "EXPORT_NODE": self.op.target_node,
9424
      "EXPORT_DO_SHUTDOWN": self.op.shutdown,
9425
      "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
9426
      # TODO: Generic function for boolean env variables
9427
      "REMOVE_INSTANCE": str(bool(self.op.remove_instance)),
9428
      }
9429

    
9430
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
9431

    
9432
    nl = [self.cfg.GetMasterNode(), self.instance.primary_node]
9433

    
9434
    if self.op.mode == constants.EXPORT_MODE_LOCAL:
9435
      nl.append(self.op.target_node)
9436

    
9437
    return env, nl, nl
9438

    
9439
  def CheckPrereq(self):
9440
    """Check prerequisites.
9441

9442
    This checks that the instance and node names are valid.
9443

9444
    """
9445
    instance_name = self.op.instance_name
9446

    
9447
    self.instance = self.cfg.GetInstanceInfo(instance_name)
9448
    assert self.instance is not None, \
9449
          "Cannot retrieve locked instance %s" % self.op.instance_name
9450
    _CheckNodeOnline(self, self.instance.primary_node)
9451

    
9452
    if self.op.mode == constants.EXPORT_MODE_LOCAL:
9453
      self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
9454
      self.dst_node = self.cfg.GetNodeInfo(self.op.target_node)
9455
      assert self.dst_node is not None
9456

    
9457
      _CheckNodeOnline(self, self.dst_node.name)
9458
      _CheckNodeNotDrained(self, self.dst_node.name)
9459

    
9460
      self._cds = None
9461
      self.dest_disk_info = None
9462
      self.dest_x509_ca = None
9463

    
9464
    elif self.op.mode == constants.EXPORT_MODE_REMOTE:
9465
      self.dst_node = None
9466

    
9467
      if len(self.op.target_node) != len(self.instance.disks):
9468
        raise errors.OpPrereqError(("Received destination information for %s"
9469
                                    " disks, but instance %s has %s disks") %
9470
                                   (len(self.op.target_node), instance_name,
9471
                                    len(self.instance.disks)),
9472
                                   errors.ECODE_INVAL)
9473

    
9474
      cds = _GetClusterDomainSecret()
9475

    
9476
      # Check X509 key name
9477
      try:
9478
        (key_name, hmac_digest, hmac_salt) = self.x509_key_name
9479
      except (TypeError, ValueError), err:
9480
        raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err)
9481

    
9482
      if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt):
9483
        raise errors.OpPrereqError("HMAC for X509 key name is wrong",
9484
                                   errors.ECODE_INVAL)
9485

    
9486
      # Load and verify CA
9487
      try:
9488
        (cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds)
9489
      except OpenSSL.crypto.Error, err:
9490
        raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" %
9491
                                   (err, ), errors.ECODE_INVAL)
9492

    
9493
      (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
9494
      if errcode is not None:
9495
        raise errors.OpPrereqError("Invalid destination X509 CA (%s)" %
9496
                                   (msg, ), errors.ECODE_INVAL)
9497

    
9498
      self.dest_x509_ca = cert
9499

    
9500
      # Verify target information
9501
      disk_info = []
9502
      for idx, disk_data in enumerate(self.op.target_node):
9503
        try:
9504
          (host, port, magic) = \
9505
            masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data)
9506
        except errors.GenericError, err:
9507
          raise errors.OpPrereqError("Target info for disk %s: %s" %
9508
                                     (idx, err), errors.ECODE_INVAL)
9509

    
9510
        disk_info.append((host, port, magic))
9511

    
9512
      assert len(disk_info) == len(self.op.target_node)
9513
      self.dest_disk_info = disk_info
9514

    
9515
    else:
9516
      raise errors.ProgrammerError("Unhandled export mode %r" %
9517
                                   self.op.mode)
9518

    
9519
    # instance disk type verification
9520
    # TODO: Implement export support for file-based disks
9521
    for disk in self.instance.disks:
9522
      if disk.dev_type == constants.LD_FILE:
9523
        raise errors.OpPrereqError("Export not supported for instances with"
9524
                                   " file-based disks", errors.ECODE_INVAL)
9525

    
9526
  def _CleanupExports(self, feedback_fn):
9527
    """Removes exports of current instance from all other nodes.
9528

9529
    If an instance in a cluster with nodes A..D was exported to node C, its
9530
    exports will be removed from the nodes A, B and D.
9531

9532
    """
9533
    assert self.op.mode != constants.EXPORT_MODE_REMOTE
9534

    
9535
    nodelist = self.cfg.GetNodeList()
9536
    nodelist.remove(self.dst_node.name)
9537

    
9538
    # on one-node clusters nodelist will be empty after the removal
9539
    # if we proceed the backup would be removed because OpQueryExports
9540
    # substitutes an empty list with the full cluster node list.
9541
    iname = self.instance.name
9542
    if nodelist:
9543
      feedback_fn("Removing old exports for instance %s" % iname)
9544
      exportlist = self.rpc.call_export_list(nodelist)
9545
      for node in exportlist:
9546
        if exportlist[node].fail_msg:
9547
          continue
9548
        if iname in exportlist[node].payload:
9549
          msg = self.rpc.call_export_remove(node, iname).fail_msg
9550
          if msg:
9551
            self.LogWarning("Could not remove older export for instance %s"
9552
                            " on node %s: %s", iname, node, msg)
9553

    
9554
  def Exec(self, feedback_fn):
9555
    """Export an instance to an image in the cluster.
9556

9557
    """
9558
    assert self.op.mode in constants.EXPORT_MODES
9559

    
9560
    instance = self.instance
9561
    src_node = instance.primary_node
9562

    
9563
    if self.op.shutdown:
9564
      # shutdown the instance, but not the disks
9565
      feedback_fn("Shutting down instance %s" % instance.name)
9566
      result = self.rpc.call_instance_shutdown(src_node, instance,
9567
                                               self.op.shutdown_timeout)
9568
      # TODO: Maybe ignore failures if ignore_remove_failures is set
9569
      result.Raise("Could not shutdown instance %s on"
9570
                   " node %s" % (instance.name, src_node))
9571

    
9572
    # set the disks ID correctly since call_instance_start needs the
9573
    # correct drbd minor to create the symlinks
9574
    for disk in instance.disks:
9575
      self.cfg.SetDiskID(disk, src_node)
9576

    
9577
    activate_disks = (not instance.admin_up)
9578

    
9579
    if activate_disks:
9580
      # Activate the instance disks if we'exporting a stopped instance
9581
      feedback_fn("Activating disks for %s" % instance.name)
9582
      _StartInstanceDisks(self, instance, None)
9583

    
9584
    try:
9585
      helper = masterd.instance.ExportInstanceHelper(self, feedback_fn,
9586
                                                     instance)
9587

    
9588
      helper.CreateSnapshots()
9589
      try:
9590
        if (self.op.shutdown and instance.admin_up and
9591
            not self.op.remove_instance):
9592
          assert not activate_disks
9593
          feedback_fn("Starting instance %s" % instance.name)
9594
          result = self.rpc.call_instance_start(src_node, instance, None, None)
9595
          msg = result.fail_msg
9596
          if msg:
9597
            feedback_fn("Failed to start instance: %s" % msg)
9598
            _ShutdownInstanceDisks(self, instance)
9599
            raise errors.OpExecError("Could not start instance: %s" % msg)
9600

    
9601
        if self.op.mode == constants.EXPORT_MODE_LOCAL:
9602
          (fin_resu, dresults) = helper.LocalExport(self.dst_node)
9603
        elif self.op.mode == constants.EXPORT_MODE_REMOTE:
9604
          connect_timeout = constants.RIE_CONNECT_TIMEOUT
9605
          timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
9606

    
9607
          (key_name, _, _) = self.x509_key_name
9608

    
9609
          dest_ca_pem = \
9610
            OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM,
9611
                                            self.dest_x509_ca)
9612

    
9613
          (fin_resu, dresults) = helper.RemoteExport(self.dest_disk_info,
9614
                                                     key_name, dest_ca_pem,
9615
                                                     timeouts)
9616
      finally:
9617
        helper.Cleanup()
9618

    
9619
      # Check for backwards compatibility
9620
      assert len(dresults) == len(instance.disks)
9621
      assert compat.all(isinstance(i, bool) for i in dresults), \
9622
             "Not all results are boolean: %r" % dresults
9623

    
9624
    finally:
9625
      if activate_disks:
9626
        feedback_fn("Deactivating disks for %s" % instance.name)
9627
        _ShutdownInstanceDisks(self, instance)
9628

    
9629
    if not (compat.all(dresults) and fin_resu):
9630
      failures = []
9631
      if not fin_resu:
9632
        failures.append("export finalization")
9633
      if not compat.all(dresults):
9634
        fdsk = utils.CommaJoin(idx for (idx, dsk) in enumerate(dresults)
9635
                               if not dsk)
9636
        failures.append("disk export: disk(s) %s" % fdsk)
9637

    
9638
      raise errors.OpExecError("Export failed, errors in %s" %
9639
                               utils.CommaJoin(failures))
9640

    
9641
    # At this point, the export was successful, we can cleanup/finish
9642

    
9643
    # Remove instance if requested
9644
    if self.op.remove_instance:
9645
      feedback_fn("Removing instance %s" % instance.name)
9646
      _RemoveInstance(self, feedback_fn, instance,
9647
                      self.op.ignore_remove_failures)
9648

    
9649
    if self.op.mode == constants.EXPORT_MODE_LOCAL:
9650
      self._CleanupExports(feedback_fn)
9651

    
9652
    return fin_resu, dresults
9653

    
9654

    
9655
class LURemoveExport(NoHooksLU):
9656
  """Remove exports related to the named instance.
9657

9658
  """
9659
  _OP_PARAMS = [
9660
    _PInstanceName,
9661
    ]
9662
  REQ_BGL = False
9663

    
9664
  def ExpandNames(self):
9665
    self.needed_locks = {}
9666
    # We need all nodes to be locked in order for RemoveExport to work, but we
9667
    # don't need to lock the instance itself, as nothing will happen to it (and
9668
    # we can remove exports also for a removed instance)
9669
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9670

    
9671
  def Exec(self, feedback_fn):
9672
    """Remove any export.
9673

9674
    """
9675
    instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
9676
    # If the instance was not found we'll try with the name that was passed in.
9677
    # This will only work if it was an FQDN, though.
9678
    fqdn_warn = False
9679
    if not instance_name:
9680
      fqdn_warn = True
9681
      instance_name = self.op.instance_name
9682

    
9683
    locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
9684
    exportlist = self.rpc.call_export_list(locked_nodes)
9685
    found = False
9686
    for node in exportlist:
9687
      msg = exportlist[node].fail_msg
9688
      if msg:
9689
        self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
9690
        continue
9691
      if instance_name in exportlist[node].payload:
9692
        found = True
9693
        result = self.rpc.call_export_remove(node, instance_name)
9694
        msg = result.fail_msg
9695
        if msg:
9696
          logging.error("Could not remove export for instance %s"
9697
                        " on node %s: %s", instance_name, node, msg)
9698

    
9699
    if fqdn_warn and not found:
9700
      feedback_fn("Export not found. If trying to remove an export belonging"
9701
                  " to a deleted instance please use its Fully Qualified"
9702
                  " Domain Name.")
9703

    
9704

    
9705
class TagsLU(NoHooksLU): # pylint: disable-msg=W0223
9706
  """Generic tags LU.
9707

9708
  This is an abstract class which is the parent of all the other tags LUs.
9709

9710
  """
9711

    
9712
  def ExpandNames(self):
9713
    self.needed_locks = {}
9714
    if self.op.kind == constants.TAG_NODE:
9715
      self.op.name = _ExpandNodeName(self.cfg, self.op.name)
9716
      self.needed_locks[locking.LEVEL_NODE] = self.op.name
9717
    elif self.op.kind == constants.TAG_INSTANCE:
9718
      self.op.name = _ExpandInstanceName(self.cfg, self.op.name)
9719
      self.needed_locks[locking.LEVEL_INSTANCE] = self.op.name
9720

    
9721
  def CheckPrereq(self):
9722
    """Check prerequisites.
9723

9724
    """
9725
    if self.op.kind == constants.TAG_CLUSTER:
9726
      self.target = self.cfg.GetClusterInfo()
9727
    elif self.op.kind == constants.TAG_NODE:
9728
      self.target = self.cfg.GetNodeInfo(self.op.name)
9729
    elif self.op.kind == constants.TAG_INSTANCE:
9730
      self.target = self.cfg.GetInstanceInfo(self.op.name)
9731
    else:
9732
      raise errors.OpPrereqError("Wrong tag type requested (%s)" %
9733
                                 str(self.op.kind), errors.ECODE_INVAL)
9734

    
9735

    
9736
class LUGetTags(TagsLU):
9737
  """Returns the tags of a given object.
9738

9739
  """
9740
  _OP_PARAMS = [
9741
    ("kind", _NoDefault, _TElemOf(constants.VALID_TAG_TYPES)),
9742
    ("name", _NoDefault, _TNonEmptyString),
9743
    ]
9744
  REQ_BGL = False
9745

    
9746
  def Exec(self, feedback_fn):
9747
    """Returns the tag list.
9748

9749
    """
9750
    return list(self.target.GetTags())
9751

    
9752

    
9753
class LUSearchTags(NoHooksLU):
9754
  """Searches the tags for a given pattern.
9755

9756
  """
9757
  _OP_PARAMS = [
9758
    ("pattern", _NoDefault, _TNonEmptyString),
9759
    ]
9760
  REQ_BGL = False
9761

    
9762
  def ExpandNames(self):
9763
    self.needed_locks = {}
9764

    
9765
  def CheckPrereq(self):
9766
    """Check prerequisites.
9767

9768
    This checks the pattern passed for validity by compiling it.
9769

9770
    """
9771
    try:
9772
      self.re = re.compile(self.op.pattern)
9773
    except re.error, err:
9774
      raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
9775
                                 (self.op.pattern, err), errors.ECODE_INVAL)
9776

    
9777
  def Exec(self, feedback_fn):
9778
    """Returns the tag list.
9779

9780
    """
9781
    cfg = self.cfg
9782
    tgts = [("/cluster", cfg.GetClusterInfo())]
9783
    ilist = cfg.GetAllInstancesInfo().values()
9784
    tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
9785
    nlist = cfg.GetAllNodesInfo().values()
9786
    tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
9787
    results = []
9788
    for path, target in tgts:
9789
      for tag in target.GetTags():
9790
        if self.re.search(tag):
9791
          results.append((path, tag))
9792
    return results
9793

    
9794

    
9795
class LUAddTags(TagsLU):
9796
  """Sets a tag on a given object.
9797

9798
  """
9799
  _OP_PARAMS = [
9800
    ("kind", _NoDefault, _TElemOf(constants.VALID_TAG_TYPES)),
9801
    ("name", _NoDefault, _TNonEmptyString),
9802
    ("tags", _NoDefault, _TListOf(_TNonEmptyString)),
9803
    ]
9804
  REQ_BGL = False
9805

    
9806
  def CheckPrereq(self):
9807
    """Check prerequisites.
9808

9809
    This checks the type and length of the tag name and value.
9810

9811
    """
9812
    TagsLU.CheckPrereq(self)
9813
    for tag in self.op.tags:
9814
      objects.TaggableObject.ValidateTag(tag)
9815

    
9816
  def Exec(self, feedback_fn):
9817
    """Sets the tag.
9818

9819
    """
9820
    try:
9821
      for tag in self.op.tags:
9822
        self.target.AddTag(tag)
9823
    except errors.TagError, err:
9824
      raise errors.OpExecError("Error while setting tag: %s" % str(err))
9825
    self.cfg.Update(self.target, feedback_fn)
9826

    
9827

    
9828
class LUDelTags(TagsLU):
9829
  """Delete a list of tags from a given object.
9830

9831
  """
9832
  _OP_PARAMS = [
9833
    ("kind", _NoDefault, _TElemOf(constants.VALID_TAG_TYPES)),
9834
    ("name", _NoDefault, _TNonEmptyString),
9835
    ("tags", _NoDefault, _TListOf(_TNonEmptyString)),
9836
    ]
9837
  REQ_BGL = False
9838

    
9839
  def CheckPrereq(self):
9840
    """Check prerequisites.
9841

9842
    This checks that we have the given tag.
9843

9844
    """
9845
    TagsLU.CheckPrereq(self)
9846
    for tag in self.op.tags:
9847
      objects.TaggableObject.ValidateTag(tag)
9848
    del_tags = frozenset(self.op.tags)
9849
    cur_tags = self.target.GetTags()
9850
    if not del_tags <= cur_tags:
9851
      diff_tags = del_tags - cur_tags
9852
      diff_names = ["'%s'" % tag for tag in diff_tags]
9853
      diff_names.sort()
9854
      raise errors.OpPrereqError("Tag(s) %s not found" %
9855
                                 (",".join(diff_names)), errors.ECODE_NOENT)
9856

    
9857
  def Exec(self, feedback_fn):
9858
    """Remove the tag from the object.
9859

9860
    """
9861
    for tag in self.op.tags:
9862
      self.target.RemoveTag(tag)
9863
    self.cfg.Update(self.target, feedback_fn)
9864

    
9865

    
9866
class LUTestDelay(NoHooksLU):
9867
  """Sleep for a specified amount of time.
9868

9869
  This LU sleeps on the master and/or nodes for a specified amount of
9870
  time.
9871

9872
  """
9873
  _OP_PARAMS = [
9874
    ("duration", _NoDefault, _TFloat),
9875
    ("on_master", True, _TBool),
9876
    ("on_nodes", _EmptyList, _TListOf(_TNonEmptyString)),
9877
    ("repeat", 0, _TPositiveInt)
9878
    ]
9879
  REQ_BGL = False
9880

    
9881
  def ExpandNames(self):
9882
    """Expand names and set required locks.
9883

9884
    This expands the node list, if any.
9885

9886
    """
9887
    self.needed_locks = {}
9888
    if self.op.on_nodes:
9889
      # _GetWantedNodes can be used here, but is not always appropriate to use
9890
      # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
9891
      # more information.
9892
      self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
9893
      self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
9894

    
9895
  def _TestDelay(self):
9896
    """Do the actual sleep.
9897

9898
    """
9899
    if self.op.on_master:
9900
      if not utils.TestDelay(self.op.duration):
9901
        raise errors.OpExecError("Error during master delay test")
9902
    if self.op.on_nodes:
9903
      result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
9904
      for node, node_result in result.items():
9905
        node_result.Raise("Failure during rpc call to node %s" % node)
9906

    
9907
  def Exec(self, feedback_fn):
9908
    """Execute the test delay opcode, with the wanted repetitions.
9909

9910
    """
9911
    if self.op.repeat == 0:
9912
      self._TestDelay()
9913
    else:
9914
      top_value = self.op.repeat - 1
9915
      for i in range(self.op.repeat):
9916
        self.LogInfo("Test delay iteration %d/%d" % (i, top_value))
9917
        self._TestDelay()
9918

    
9919

    
9920
class LUTestJobqueue(NoHooksLU):
9921
  """Utility LU to test some aspects of the job queue.
9922

9923
  """
9924
  _OP_PARAMS = [
9925
    ("notify_waitlock", False, _TBool),
9926
    ("notify_exec", False, _TBool),
9927
    ("log_messages", _EmptyList, _TListOf(_TString)),
9928
    ("fail", False, _TBool),
9929
    ]
9930
  REQ_BGL = False
9931

    
9932
  # Must be lower than default timeout for WaitForJobChange to see whether it
9933
  # notices changed jobs
9934
  _CLIENT_CONNECT_TIMEOUT = 20.0
9935
  _CLIENT_CONFIRM_TIMEOUT = 60.0
9936

    
9937
  @classmethod
9938
  def _NotifyUsingSocket(cls, cb, errcls):
9939
    """Opens a Unix socket and waits for another program to connect.
9940

9941
    @type cb: callable
9942
    @param cb: Callback to send socket name to client
9943
    @type errcls: class
9944
    @param errcls: Exception class to use for errors
9945

9946
    """
9947
    # Using a temporary directory as there's no easy way to create temporary
9948
    # sockets without writing a custom loop around tempfile.mktemp and
9949
    # socket.bind
9950
    tmpdir = tempfile.mkdtemp()
9951
    try:
9952
      tmpsock = utils.PathJoin(tmpdir, "sock")
9953

    
9954
      logging.debug("Creating temporary socket at %s", tmpsock)
9955
      sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
9956
      try:
9957
        sock.bind(tmpsock)
9958
        sock.listen(1)
9959

    
9960
        # Send details to client
9961
        cb(tmpsock)
9962

    
9963
        # Wait for client to connect before continuing
9964
        sock.settimeout(cls._CLIENT_CONNECT_TIMEOUT)
9965
        try:
9966
          (conn, _) = sock.accept()
9967
        except socket.error, err:
9968
          raise errcls("Client didn't connect in time (%s)" % err)
9969
      finally:
9970
        sock.close()
9971
    finally:
9972
      # Remove as soon as client is connected
9973
      shutil.rmtree(tmpdir)
9974

    
9975
    # Wait for client to close
9976
    try:
9977
      try:
9978
        # pylint: disable-msg=E1101
9979
        # Instance of '_socketobject' has no ... member
9980
        conn.settimeout(cls._CLIENT_CONFIRM_TIMEOUT)
9981
        conn.recv(1)
9982
      except socket.error, err:
9983
        raise errcls("Client failed to confirm notification (%s)" % err)
9984
    finally:
9985
      conn.close()
9986

    
9987
  def _SendNotification(self, test, arg, sockname):
9988
    """Sends a notification to the client.
9989

9990
    @type test: string
9991
    @param test: Test name
9992
    @param arg: Test argument (depends on test)
9993
    @type sockname: string
9994
    @param sockname: Socket path
9995

9996
    """
9997
    self.Log(constants.ELOG_JQUEUE_TEST, (sockname, test, arg))
9998

    
9999
  def _Notify(self, prereq, test, arg):
10000
    """Notifies the client of a test.
10001

10002
    @type prereq: bool
10003
    @param prereq: Whether this is a prereq-phase test
10004
    @type test: string
10005
    @param test: Test name
10006
    @param arg: Test argument (depends on test)
10007

10008
    """
10009
    if prereq:
10010
      errcls = errors.OpPrereqError
10011
    else:
10012
      errcls = errors.OpExecError
10013

    
10014
    return self._NotifyUsingSocket(compat.partial(self._SendNotification,
10015
                                                  test, arg),
10016
                                   errcls)
10017

    
10018
  def CheckArguments(self):
10019
    self.checkargs_calls = getattr(self, "checkargs_calls", 0) + 1
10020
    self.expandnames_calls = 0
10021

    
10022
  def ExpandNames(self):
10023
    checkargs_calls = getattr(self, "checkargs_calls", 0)
10024
    if checkargs_calls < 1:
10025
      raise errors.ProgrammerError("CheckArguments was not called")
10026

    
10027
    self.expandnames_calls += 1
10028

    
10029
    if self.op.notify_waitlock:
10030
      self._Notify(True, constants.JQT_EXPANDNAMES, None)
10031

    
10032
    self.LogInfo("Expanding names")
10033

    
10034
    # Get lock on master node (just to get a lock, not for a particular reason)
10035
    self.needed_locks = {
10036
      locking.LEVEL_NODE: self.cfg.GetMasterNode(),
10037
      }
10038

    
10039
  def Exec(self, feedback_fn):
10040
    if self.expandnames_calls < 1:
10041
      raise errors.ProgrammerError("ExpandNames was not called")
10042

    
10043
    if self.op.notify_exec:
10044
      self._Notify(False, constants.JQT_EXEC, None)
10045

    
10046
    self.LogInfo("Executing")
10047

    
10048
    if self.op.log_messages:
10049
      self._Notify(False, constants.JQT_STARTMSG, len(self.op.log_messages))
10050
      for idx, msg in enumerate(self.op.log_messages):
10051
        self.LogInfo("Sending log message %s", idx + 1)
10052
        feedback_fn(constants.JQT_MSGPREFIX + msg)
10053
        # Report how many test messages have been sent
10054
        self._Notify(False, constants.JQT_LOGMSG, idx + 1)
10055

    
10056
    if self.op.fail:
10057
      raise errors.OpExecError("Opcode failure was requested")
10058

    
10059
    return True
10060

    
10061

    
10062
class IAllocator(object):
10063
  """IAllocator framework.
10064

10065
  An IAllocator instance has three sets of attributes:
10066
    - cfg that is needed to query the cluster
10067
    - input data (all members of the _KEYS class attribute are required)
10068
    - four buffer attributes (in|out_data|text), that represent the
10069
      input (to the external script) in text and data structure format,
10070
      and the output from it, again in two formats
10071
    - the result variables from the script (success, info, nodes) for
10072
      easy usage
10073

10074
  """
10075
  # pylint: disable-msg=R0902
10076
  # lots of instance attributes
10077
  _ALLO_KEYS = [
10078
    "name", "mem_size", "disks", "disk_template",
10079
    "os", "tags", "nics", "vcpus", "hypervisor",
10080
    ]
10081
  _RELO_KEYS = [
10082
    "name", "relocate_from",
10083
    ]
10084
  _EVAC_KEYS = [
10085
    "evac_nodes",
10086
    ]
10087

    
10088
  def __init__(self, cfg, rpc, mode, **kwargs):
10089
    self.cfg = cfg
10090
    self.rpc = rpc
10091
    # init buffer variables
10092
    self.in_text = self.out_text = self.in_data = self.out_data = None
10093
    # init all input fields so that pylint is happy
10094
    self.mode = mode
10095
    self.mem_size = self.disks = self.disk_template = None
10096
    self.os = self.tags = self.nics = self.vcpus = None
10097
    self.hypervisor = None
10098
    self.relocate_from = None
10099
    self.name = None
10100
    self.evac_nodes = None
10101
    # computed fields
10102
    self.required_nodes = None
10103
    # init result fields
10104
    self.success = self.info = self.result = None
10105
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
10106
      keyset = self._ALLO_KEYS
10107
      fn = self._AddNewInstance
10108
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
10109
      keyset = self._RELO_KEYS
10110
      fn = self._AddRelocateInstance
10111
    elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
10112
      keyset = self._EVAC_KEYS
10113
      fn = self._AddEvacuateNodes
10114
    else:
10115
      raise errors.ProgrammerError("Unknown mode '%s' passed to the"
10116
                                   " IAllocator" % self.mode)
10117
    for key in kwargs:
10118
      if key not in keyset:
10119
        raise errors.ProgrammerError("Invalid input parameter '%s' to"
10120
                                     " IAllocator" % key)
10121
      setattr(self, key, kwargs[key])
10122

    
10123
    for key in keyset:
10124
      if key not in kwargs:
10125
        raise errors.ProgrammerError("Missing input parameter '%s' to"
10126
                                     " IAllocator" % key)
10127
    self._BuildInputData(fn)
10128

    
10129
  def _ComputeClusterData(self):
10130
    """Compute the generic allocator input data.
10131

10132
    This is the data that is independent of the actual operation.
10133

10134
    """
10135
    cfg = self.cfg
10136
    cluster_info = cfg.GetClusterInfo()
10137
    # cluster data
10138
    data = {
10139
      "version": constants.IALLOCATOR_VERSION,
10140
      "cluster_name": cfg.GetClusterName(),
10141
      "cluster_tags": list(cluster_info.GetTags()),
10142
      "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
10143
      # we don't have job IDs
10144
      }
10145
    iinfo = cfg.GetAllInstancesInfo().values()
10146
    i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
10147

    
10148
    # node data
10149
    node_results = {}
10150
    node_list = cfg.GetNodeList()
10151

    
10152
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
10153
      hypervisor_name = self.hypervisor
10154
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
10155
      hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
10156
    elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
10157
      hypervisor_name = cluster_info.enabled_hypervisors[0]
10158

    
10159
    node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(),
10160
                                        hypervisor_name)
10161
    node_iinfo = \
10162
      self.rpc.call_all_instances_info(node_list,
10163
                                       cluster_info.enabled_hypervisors)
10164
    for nname, nresult in node_data.items():
10165
      # first fill in static (config-based) values
10166
      ninfo = cfg.GetNodeInfo(nname)
10167
      pnr = {
10168
        "tags": list(ninfo.GetTags()),
10169
        "primary_ip": ninfo.primary_ip,
10170
        "secondary_ip": ninfo.secondary_ip,
10171
        "offline": ninfo.offline,
10172
        "drained": ninfo.drained,
10173
        "master_candidate": ninfo.master_candidate,
10174
        }
10175

    
10176
      if not (ninfo.offline or ninfo.drained):
10177
        nresult.Raise("Can't get data for node %s" % nname)
10178
        node_iinfo[nname].Raise("Can't get node instance info from node %s" %
10179
                                nname)
10180
        remote_info = nresult.payload
10181

    
10182
        for attr in ['memory_total', 'memory_free', 'memory_dom0',
10183
                     'vg_size', 'vg_free', 'cpu_total']:
10184
          if attr not in remote_info:
10185
            raise errors.OpExecError("Node '%s' didn't return attribute"
10186
                                     " '%s'" % (nname, attr))
10187
          if not isinstance(remote_info[attr], int):
10188
            raise errors.OpExecError("Node '%s' returned invalid value"
10189
                                     " for '%s': %s" %
10190
                                     (nname, attr, remote_info[attr]))
10191
        # compute memory used by primary instances
10192
        i_p_mem = i_p_up_mem = 0
10193
        for iinfo, beinfo in i_list:
10194
          if iinfo.primary_node == nname:
10195
            i_p_mem += beinfo[constants.BE_MEMORY]
10196
            if iinfo.name not in node_iinfo[nname].payload:
10197
              i_used_mem = 0
10198
            else:
10199
              i_used_mem = int(node_iinfo[nname].payload[iinfo.name]['memory'])
10200
            i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
10201
            remote_info['memory_free'] -= max(0, i_mem_diff)
10202

    
10203
            if iinfo.admin_up:
10204
              i_p_up_mem += beinfo[constants.BE_MEMORY]
10205

    
10206
        # compute memory used by instances
10207
        pnr_dyn = {
10208
          "total_memory": remote_info['memory_total'],
10209
          "reserved_memory": remote_info['memory_dom0'],
10210
          "free_memory": remote_info['memory_free'],
10211
          "total_disk": remote_info['vg_size'],
10212
          "free_disk": remote_info['vg_free'],
10213
          "total_cpus": remote_info['cpu_total'],
10214
          "i_pri_memory": i_p_mem,
10215
          "i_pri_up_memory": i_p_up_mem,
10216
          }
10217
        pnr.update(pnr_dyn)
10218

    
10219
      node_results[nname] = pnr
10220
    data["nodes"] = node_results
10221

    
10222
    # instance data
10223
    instance_data = {}
10224
    for iinfo, beinfo in i_list:
10225
      nic_data = []
10226
      for nic in iinfo.nics:
10227
        filled_params = cluster_info.SimpleFillNIC(nic.nicparams)
10228
        nic_dict = {"mac": nic.mac,
10229
                    "ip": nic.ip,
10230
                    "mode": filled_params[constants.NIC_MODE],
10231
                    "link": filled_params[constants.NIC_LINK],
10232
                   }
10233
        if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
10234
          nic_dict["bridge"] = filled_params[constants.NIC_LINK]
10235
        nic_data.append(nic_dict)
10236
      pir = {
10237
        "tags": list(iinfo.GetTags()),
10238
        "admin_up": iinfo.admin_up,
10239
        "vcpus": beinfo[constants.BE_VCPUS],
10240
        "memory": beinfo[constants.BE_MEMORY],
10241
        "os": iinfo.os,
10242
        "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
10243
        "nics": nic_data,
10244
        "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
10245
        "disk_template": iinfo.disk_template,
10246
        "hypervisor": iinfo.hypervisor,
10247
        }
10248
      pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
10249
                                                 pir["disks"])
10250
      instance_data[iinfo.name] = pir
10251

    
10252
    data["instances"] = instance_data
10253

    
10254
    self.in_data = data
10255

    
10256
  def _AddNewInstance(self):
10257
    """Add new instance data to allocator structure.
10258

10259
    This in combination with _AllocatorGetClusterData will create the
10260
    correct structure needed as input for the allocator.
10261

10262
    The checks for the completeness of the opcode must have already been
10263
    done.
10264

10265
    """
10266
    disk_space = _ComputeDiskSize(self.disk_template, self.disks)
10267

    
10268
    if self.disk_template in constants.DTS_NET_MIRROR:
10269
      self.required_nodes = 2
10270
    else:
10271
      self.required_nodes = 1
10272
    request = {
10273
      "name": self.name,
10274
      "disk_template": self.disk_template,
10275
      "tags": self.tags,
10276
      "os": self.os,
10277
      "vcpus": self.vcpus,
10278
      "memory": self.mem_size,
10279
      "disks": self.disks,
10280
      "disk_space_total": disk_space,
10281
      "nics": self.nics,
10282
      "required_nodes": self.required_nodes,
10283
      }
10284
    return request
10285

    
10286
  def _AddRelocateInstance(self):
10287
    """Add relocate instance data to allocator structure.
10288

10289
    This in combination with _IAllocatorGetClusterData will create the
10290
    correct structure needed as input for the allocator.
10291

10292
    The checks for the completeness of the opcode must have already been
10293
    done.
10294

10295
    """
10296
    instance = self.cfg.GetInstanceInfo(self.name)
10297
    if instance is None:
10298
      raise errors.ProgrammerError("Unknown instance '%s' passed to"
10299
                                   " IAllocator" % self.name)
10300

    
10301
    if instance.disk_template not in constants.DTS_NET_MIRROR:
10302
      raise errors.OpPrereqError("Can't relocate non-mirrored instances",
10303
                                 errors.ECODE_INVAL)
10304

    
10305
    if len(instance.secondary_nodes) != 1:
10306
      raise errors.OpPrereqError("Instance has not exactly one secondary node",
10307
                                 errors.ECODE_STATE)
10308

    
10309
    self.required_nodes = 1
10310
    disk_sizes = [{'size': disk.size} for disk in instance.disks]
10311
    disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
10312

    
10313
    request = {
10314
      "name": self.name,
10315
      "disk_space_total": disk_space,
10316
      "required_nodes": self.required_nodes,
10317
      "relocate_from": self.relocate_from,
10318
      }
10319
    return request
10320

    
10321
  def _AddEvacuateNodes(self):
10322
    """Add evacuate nodes data to allocator structure.
10323

10324
    """
10325
    request = {
10326
      "evac_nodes": self.evac_nodes
10327
      }
10328
    return request
10329

    
10330
  def _BuildInputData(self, fn):
10331
    """Build input data structures.
10332

10333
    """
10334
    self._ComputeClusterData()
10335

    
10336
    request = fn()
10337
    request["type"] = self.mode
10338
    self.in_data["request"] = request
10339

    
10340
    self.in_text = serializer.Dump(self.in_data)
10341

    
10342
  def Run(self, name, validate=True, call_fn=None):
10343
    """Run an instance allocator and return the results.
10344

10345
    """
10346
    if call_fn is None:
10347
      call_fn = self.rpc.call_iallocator_runner
10348

    
10349
    result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
10350
    result.Raise("Failure while running the iallocator script")
10351

    
10352
    self.out_text = result.payload
10353
    if validate:
10354
      self._ValidateResult()
10355

    
10356
  def _ValidateResult(self):
10357
    """Process the allocator results.
10358

10359
    This will process and if successful save the result in
10360
    self.out_data and the other parameters.
10361

10362
    """
10363
    try:
10364
      rdict = serializer.Load(self.out_text)
10365
    except Exception, err:
10366
      raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
10367

    
10368
    if not isinstance(rdict, dict):
10369
      raise errors.OpExecError("Can't parse iallocator results: not a dict")
10370

    
10371
    # TODO: remove backwards compatiblity in later versions
10372
    if "nodes" in rdict and "result" not in rdict:
10373
      rdict["result"] = rdict["nodes"]
10374
      del rdict["nodes"]
10375

    
10376
    for key in "success", "info", "result":
10377
      if key not in rdict:
10378
        raise errors.OpExecError("Can't parse iallocator results:"
10379
                                 " missing key '%s'" % key)
10380
      setattr(self, key, rdict[key])
10381

    
10382
    if not isinstance(rdict["result"], list):
10383
      raise errors.OpExecError("Can't parse iallocator results: 'result' key"
10384
                               " is not a list")
10385
    self.out_data = rdict
10386

    
10387

    
10388
class LUTestAllocator(NoHooksLU):
10389
  """Run allocator tests.
10390

10391
  This LU runs the allocator tests
10392

10393
  """
10394
  _OP_PARAMS = [
10395
    ("direction", _NoDefault, _TElemOf(constants.VALID_IALLOCATOR_DIRECTIONS)),
10396
    ("mode", _NoDefault, _TElemOf(constants.VALID_IALLOCATOR_MODES)),
10397
    ("name", _NoDefault, _TNonEmptyString),
10398
    ("nics", _NoDefault, _TOr(_TNone, _TListOf(
10399
      _TDictOf(_TElemOf(["mac", "ip", "bridge"]),
10400
               _TOr(_TNone, _TNonEmptyString))))),
10401
    ("disks", _NoDefault, _TOr(_TNone, _TList)),
10402
    ("hypervisor", None, _TMaybeString),
10403
    ("allocator", None, _TMaybeString),
10404
    ("tags", _EmptyList, _TListOf(_TNonEmptyString)),
10405
    ("mem_size", None, _TOr(_TNone, _TPositiveInt)),
10406
    ("vcpus", None, _TOr(_TNone, _TPositiveInt)),
10407
    ("os", None, _TMaybeString),
10408
    ("disk_template", None, _TMaybeString),
10409
    ("evac_nodes", None, _TOr(_TNone, _TListOf(_TNonEmptyString))),
10410
    ]
10411

    
10412
  def CheckPrereq(self):
10413
    """Check prerequisites.
10414

10415
    This checks the opcode parameters depending on the director and mode test.
10416

10417
    """
10418
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
10419
      for attr in ["mem_size", "disks", "disk_template",
10420
                   "os", "tags", "nics", "vcpus"]:
10421
        if not hasattr(self.op, attr):
10422
          raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
10423
                                     attr, errors.ECODE_INVAL)
10424
      iname = self.cfg.ExpandInstanceName(self.op.name)
10425
      if iname is not None:
10426
        raise errors.OpPrereqError("Instance '%s' already in the cluster" %
10427
                                   iname, errors.ECODE_EXISTS)
10428
      if not isinstance(self.op.nics, list):
10429
        raise errors.OpPrereqError("Invalid parameter 'nics'",
10430
                                   errors.ECODE_INVAL)
10431
      if not isinstance(self.op.disks, list):
10432
        raise errors.OpPrereqError("Invalid parameter 'disks'",
10433
                                   errors.ECODE_INVAL)
10434
      for row in self.op.disks:
10435
        if (not isinstance(row, dict) or
10436
            "size" not in row or
10437
            not isinstance(row["size"], int) or
10438
            "mode" not in row or
10439
            row["mode"] not in ['r', 'w']):
10440
          raise errors.OpPrereqError("Invalid contents of the 'disks'"
10441
                                     " parameter", errors.ECODE_INVAL)
10442
      if self.op.hypervisor is None:
10443
        self.op.hypervisor = self.cfg.GetHypervisorType()
10444
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
10445
      fname = _ExpandInstanceName(self.cfg, self.op.name)
10446
      self.op.name = fname
10447
      self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
10448
    elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
10449
      if not hasattr(self.op, "evac_nodes"):
10450
        raise errors.OpPrereqError("Missing attribute 'evac_nodes' on"
10451
                                   " opcode input", errors.ECODE_INVAL)
10452
    else:
10453
      raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
10454
                                 self.op.mode, errors.ECODE_INVAL)
10455

    
10456
    if self.op.direction == constants.IALLOCATOR_DIR_OUT:
10457
      if self.op.allocator is None:
10458
        raise errors.OpPrereqError("Missing allocator name",
10459
                                   errors.ECODE_INVAL)
10460
    elif self.op.direction != constants.IALLOCATOR_DIR_IN:
10461
      raise errors.OpPrereqError("Wrong allocator test '%s'" %
10462
                                 self.op.direction, errors.ECODE_INVAL)
10463

    
10464
  def Exec(self, feedback_fn):
10465
    """Run the allocator test.
10466

10467
    """
10468
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
10469
      ial = IAllocator(self.cfg, self.rpc,
10470
                       mode=self.op.mode,
10471
                       name=self.op.name,
10472
                       mem_size=self.op.mem_size,
10473
                       disks=self.op.disks,
10474
                       disk_template=self.op.disk_template,
10475
                       os=self.op.os,
10476
                       tags=self.op.tags,
10477
                       nics=self.op.nics,
10478
                       vcpus=self.op.vcpus,
10479
                       hypervisor=self.op.hypervisor,
10480
                       )
10481
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
10482
      ial = IAllocator(self.cfg, self.rpc,
10483
                       mode=self.op.mode,
10484
                       name=self.op.name,
10485
                       relocate_from=list(self.relocate_from),
10486
                       )
10487
    elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
10488
      ial = IAllocator(self.cfg, self.rpc,
10489
                       mode=self.op.mode,
10490
                       evac_nodes=self.op.evac_nodes)
10491
    else:
10492
      raise errors.ProgrammerError("Uncatched mode %s in"
10493
                                   " LUTestAllocator.Exec", self.op.mode)
10494

    
10495
    if self.op.direction == constants.IALLOCATOR_DIR_IN:
10496
      result = ial.in_text
10497
    else:
10498
      ial.Run(self.op.allocator, validate=False)
10499
      result = ial.out_text
10500
    return result