Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ 3f2ad566

History | View | Annotate | Download (368.3 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
def _TIsLength(size):
158
  """Check is the given container is of the given size.
159

160
  """
161
  return lambda container: len(container) == size
162

    
163

    
164
# Combinator types
165
def _TAnd(*args):
166
  """Combine multiple functions using an AND operation.
167

168
  """
169
  def fn(val):
170
    return compat.all(t(val) for t in args)
171
  return fn
172

    
173

    
174
def _TOr(*args):
175
  """Combine multiple functions using an AND operation.
176

177
  """
178
  def fn(val):
179
    return compat.any(t(val) for t in args)
180
  return fn
181

    
182

    
183
def _TMap(fn, test):
184
  """Checks that a modified version of the argument passes the given test.
185

186
  """
187
  return lambda val: test(fn(val))
188

    
189

    
190
# Type aliases
191

    
192
#: a non-empty string
193
_TNonEmptyString = _TAnd(_TString, _TTrue)
194

    
195

    
196
#: a maybe non-empty string
197
_TMaybeString = _TOr(_TNonEmptyString, _TNone)
198

    
199

    
200
#: a maybe boolean (bool or none)
201
_TMaybeBool = _TOr(_TBool, _TNone)
202

    
203

    
204
#: a positive integer
205
_TPositiveInt = _TAnd(_TInt, lambda v: v >= 0)
206

    
207
#: a strictly positive integer
208
_TStrictPositiveInt = _TAnd(_TInt, lambda v: v > 0)
209

    
210

    
211
def _TListOf(my_type):
212
  """Checks if a given value is a list with all elements of the same type.
213

214
  """
215
  return _TAnd(_TList,
216
               lambda lst: compat.all(my_type(v) for v in lst))
217

    
218

    
219
def _TDictOf(key_type, val_type):
220
  """Checks a dict type for the type of its key/values.
221

222
  """
223
  return _TAnd(_TDict,
224
               lambda my_dict: (compat.all(key_type(v) for v in my_dict.keys())
225
                                and compat.all(val_type(v)
226
                                               for v in my_dict.values())))
227

    
228

    
229
# Common opcode attributes
230

    
231
#: output fields for a query operation
232
_POutputFields = ("output_fields", _NoDefault, _TListOf(_TNonEmptyString))
233

    
234

    
235
#: the shutdown timeout
236
_PShutdownTimeout = ("shutdown_timeout", constants.DEFAULT_SHUTDOWN_TIMEOUT,
237
                     _TPositiveInt)
238

    
239
#: the force parameter
240
_PForce = ("force", False, _TBool)
241

    
242
#: a required instance name (for single-instance LUs)
243
_PInstanceName = ("instance_name", _NoDefault, _TNonEmptyString)
244

    
245

    
246
#: a required node name (for single-node LUs)
247
_PNodeName = ("node_name", _NoDefault, _TNonEmptyString)
248

    
249
#: the migration type (live/non-live)
250
_PMigrationMode = ("mode", None, _TOr(_TNone,
251
                                      _TElemOf(constants.HT_MIGRATION_MODES)))
252

    
253
#: the obsolete 'live' mode (boolean)
254
_PMigrationLive = ("live", None, _TMaybeBool)
255

    
256

    
257
# End types
258
class LogicalUnit(object):
259
  """Logical Unit base class.
260

261
  Subclasses must follow these rules:
262
    - implement ExpandNames
263
    - implement CheckPrereq (except when tasklets are used)
264
    - implement Exec (except when tasklets are used)
265
    - implement BuildHooksEnv
266
    - redefine HPATH and HTYPE
267
    - optionally redefine their run requirements:
268
        REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
269

270
  Note that all commands require root permissions.
271

272
  @ivar dry_run_result: the value (if any) that will be returned to the caller
273
      in dry-run mode (signalled by opcode dry_run parameter)
274
  @cvar _OP_PARAMS: a list of opcode attributes, their defaults values
275
      they should get if not already defined, and types they must match
276

277
  """
278
  HPATH = None
279
  HTYPE = None
280
  _OP_PARAMS = []
281
  REQ_BGL = True
282

    
283
  def __init__(self, processor, op, context, rpc):
284
    """Constructor for LogicalUnit.
285

286
    This needs to be overridden in derived classes in order to check op
287
    validity.
288

289
    """
290
    self.proc = processor
291
    self.op = op
292
    self.cfg = context.cfg
293
    self.context = context
294
    self.rpc = rpc
295
    # Dicts used to declare locking needs to mcpu
296
    self.needed_locks = None
297
    self.acquired_locks = {}
298
    self.share_locks = dict.fromkeys(locking.LEVELS, 0)
299
    self.add_locks = {}
300
    self.remove_locks = {}
301
    # Used to force good behavior when calling helper functions
302
    self.recalculate_locks = {}
303
    self.__ssh = None
304
    # logging
305
    self.Log = processor.Log # pylint: disable-msg=C0103
306
    self.LogWarning = processor.LogWarning # pylint: disable-msg=C0103
307
    self.LogInfo = processor.LogInfo # pylint: disable-msg=C0103
308
    self.LogStep = processor.LogStep # pylint: disable-msg=C0103
309
    # support for dry-run
310
    self.dry_run_result = None
311
    # support for generic debug attribute
312
    if (not hasattr(self.op, "debug_level") or
313
        not isinstance(self.op.debug_level, int)):
314
      self.op.debug_level = 0
315

    
316
    # Tasklets
317
    self.tasklets = None
318

    
319
    # The new kind-of-type-system
320
    op_id = self.op.OP_ID
321
    for attr_name, aval, test in self._OP_PARAMS:
322
      if not hasattr(op, attr_name):
323
        if aval == _NoDefault:
324
          raise errors.OpPrereqError("Required parameter '%s.%s' missing" %
325
                                     (op_id, attr_name), errors.ECODE_INVAL)
326
        else:
327
          if callable(aval):
328
            dval = aval()
329
          else:
330
            dval = aval
331
          setattr(self.op, attr_name, dval)
332
      attr_val = getattr(op, attr_name)
333
      if test == _NoType:
334
        # no tests here
335
        continue
336
      if not callable(test):
337
        raise errors.ProgrammerError("Validation for parameter '%s.%s' failed,"
338
                                     " given type is not a proper type (%s)" %
339
                                     (op_id, attr_name, test))
340
      if not test(attr_val):
341
        logging.error("OpCode %s, parameter %s, has invalid type %s/value %s",
342
                      self.op.OP_ID, attr_name, type(attr_val), attr_val)
343
        raise errors.OpPrereqError("Parameter '%s.%s' fails validation" %
344
                                   (op_id, attr_name), errors.ECODE_INVAL)
345

    
346
    self.CheckArguments()
347

    
348
  def __GetSSH(self):
349
    """Returns the SshRunner object
350

351
    """
352
    if not self.__ssh:
353
      self.__ssh = ssh.SshRunner(self.cfg.GetClusterName())
354
    return self.__ssh
355

    
356
  ssh = property(fget=__GetSSH)
357

    
358
  def CheckArguments(self):
359
    """Check syntactic validity for the opcode arguments.
360

361
    This method is for doing a simple syntactic check and ensure
362
    validity of opcode parameters, without any cluster-related
363
    checks. While the same can be accomplished in ExpandNames and/or
364
    CheckPrereq, doing these separate is better because:
365

366
      - ExpandNames is left as as purely a lock-related function
367
      - CheckPrereq is run after we have acquired locks (and possible
368
        waited for them)
369

370
    The function is allowed to change the self.op attribute so that
371
    later methods can no longer worry about missing parameters.
372

373
    """
374
    pass
375

    
376
  def ExpandNames(self):
377
    """Expand names for this LU.
378

379
    This method is called before starting to execute the opcode, and it should
380
    update all the parameters of the opcode to their canonical form (e.g. a
381
    short node name must be fully expanded after this method has successfully
382
    completed). This way locking, hooks, logging, ecc. can work correctly.
383

384
    LUs which implement this method must also populate the self.needed_locks
385
    member, as a dict with lock levels as keys, and a list of needed lock names
386
    as values. Rules:
387

388
      - use an empty dict if you don't need any lock
389
      - if you don't need any lock at a particular level omit that level
390
      - don't put anything for the BGL level
391
      - if you want all locks at a level use locking.ALL_SET as a value
392

393
    If you need to share locks (rather than acquire them exclusively) at one
394
    level you can modify self.share_locks, setting a true value (usually 1) for
395
    that level. By default locks are not shared.
396

397
    This function can also define a list of tasklets, which then will be
398
    executed in order instead of the usual LU-level CheckPrereq and Exec
399
    functions, if those are not defined by the LU.
400

401
    Examples::
402

403
      # Acquire all nodes and one instance
404
      self.needed_locks = {
405
        locking.LEVEL_NODE: locking.ALL_SET,
406
        locking.LEVEL_INSTANCE: ['instance1.example.com'],
407
      }
408
      # Acquire just two nodes
409
      self.needed_locks = {
410
        locking.LEVEL_NODE: ['node1.example.com', 'node2.example.com'],
411
      }
412
      # Acquire no locks
413
      self.needed_locks = {} # No, you can't leave it to the default value None
414

415
    """
416
    # The implementation of this method is mandatory only if the new LU is
417
    # concurrent, so that old LUs don't need to be changed all at the same
418
    # time.
419
    if self.REQ_BGL:
420
      self.needed_locks = {} # Exclusive LUs don't need locks.
421
    else:
422
      raise NotImplementedError
423

    
424
  def DeclareLocks(self, level):
425
    """Declare LU locking needs for a level
426

427
    While most LUs can just declare their locking needs at ExpandNames time,
428
    sometimes there's the need to calculate some locks after having acquired
429
    the ones before. This function is called just before acquiring locks at a
430
    particular level, but after acquiring the ones at lower levels, and permits
431
    such calculations. It can be used to modify self.needed_locks, and by
432
    default it does nothing.
433

434
    This function is only called if you have something already set in
435
    self.needed_locks for the level.
436

437
    @param level: Locking level which is going to be locked
438
    @type level: member of ganeti.locking.LEVELS
439

440
    """
441

    
442
  def CheckPrereq(self):
443
    """Check prerequisites for this LU.
444

445
    This method should check that the prerequisites for the execution
446
    of this LU are fulfilled. It can do internode communication, but
447
    it should be idempotent - no cluster or system changes are
448
    allowed.
449

450
    The method should raise errors.OpPrereqError in case something is
451
    not fulfilled. Its return value is ignored.
452

453
    This method should also update all the parameters of the opcode to
454
    their canonical form if it hasn't been done by ExpandNames before.
455

456
    """
457
    if self.tasklets is not None:
458
      for (idx, tl) in enumerate(self.tasklets):
459
        logging.debug("Checking prerequisites for tasklet %s/%s",
460
                      idx + 1, len(self.tasklets))
461
        tl.CheckPrereq()
462
    else:
463
      pass
464

    
465
  def Exec(self, feedback_fn):
466
    """Execute the LU.
467

468
    This method should implement the actual work. It should raise
469
    errors.OpExecError for failures that are somewhat dealt with in
470
    code, or expected.
471

472
    """
473
    if self.tasklets is not None:
474
      for (idx, tl) in enumerate(self.tasklets):
475
        logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets))
476
        tl.Exec(feedback_fn)
477
    else:
478
      raise NotImplementedError
479

    
480
  def BuildHooksEnv(self):
481
    """Build hooks environment for this LU.
482

483
    This method should return a three-node tuple consisting of: a dict
484
    containing the environment that will be used for running the
485
    specific hook for this LU, a list of node names on which the hook
486
    should run before the execution, and a list of node names on which
487
    the hook should run after the execution.
488

489
    The keys of the dict must not have 'GANETI_' prefixed as this will
490
    be handled in the hooks runner. Also note additional keys will be
491
    added by the hooks runner. If the LU doesn't define any
492
    environment, an empty dict (and not None) should be returned.
493

494
    No nodes should be returned as an empty list (and not None).
495

496
    Note that if the HPATH for a LU class is None, this function will
497
    not be called.
498

499
    """
500
    raise NotImplementedError
501

    
502
  def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
503
    """Notify the LU about the results of its hooks.
504

505
    This method is called every time a hooks phase is executed, and notifies
506
    the Logical Unit about the hooks' result. The LU can then use it to alter
507
    its result based on the hooks.  By default the method does nothing and the
508
    previous result is passed back unchanged but any LU can define it if it
509
    wants to use the local cluster hook-scripts somehow.
510

511
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
512
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
513
    @param hook_results: the results of the multi-node hooks rpc call
514
    @param feedback_fn: function used send feedback back to the caller
515
    @param lu_result: the previous Exec result this LU had, or None
516
        in the PRE phase
517
    @return: the new Exec result, based on the previous result
518
        and hook results
519

520
    """
521
    # API must be kept, thus we ignore the unused argument and could
522
    # be a function warnings
523
    # pylint: disable-msg=W0613,R0201
524
    return lu_result
525

    
526
  def _ExpandAndLockInstance(self):
527
    """Helper function to expand and lock an instance.
528

529
    Many LUs that work on an instance take its name in self.op.instance_name
530
    and need to expand it and then declare the expanded name for locking. This
531
    function does it, and then updates self.op.instance_name to the expanded
532
    name. It also initializes needed_locks as a dict, if this hasn't been done
533
    before.
534

535
    """
536
    if self.needed_locks is None:
537
      self.needed_locks = {}
538
    else:
539
      assert locking.LEVEL_INSTANCE not in self.needed_locks, \
540
        "_ExpandAndLockInstance called with instance-level locks set"
541
    self.op.instance_name = _ExpandInstanceName(self.cfg,
542
                                                self.op.instance_name)
543
    self.needed_locks[locking.LEVEL_INSTANCE] = self.op.instance_name
544

    
545
  def _LockInstancesNodes(self, primary_only=False):
546
    """Helper function to declare instances' nodes for locking.
547

548
    This function should be called after locking one or more instances to lock
549
    their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
550
    with all primary or secondary nodes for instances already locked and
551
    present in self.needed_locks[locking.LEVEL_INSTANCE].
552

553
    It should be called from DeclareLocks, and for safety only works if
554
    self.recalculate_locks[locking.LEVEL_NODE] is set.
555

556
    In the future it may grow parameters to just lock some instance's nodes, or
557
    to just lock primaries or secondary nodes, if needed.
558

559
    If should be called in DeclareLocks in a way similar to::
560

561
      if level == locking.LEVEL_NODE:
562
        self._LockInstancesNodes()
563

564
    @type primary_only: boolean
565
    @param primary_only: only lock primary nodes of locked instances
566

567
    """
568
    assert locking.LEVEL_NODE in self.recalculate_locks, \
569
      "_LockInstancesNodes helper function called with no nodes to recalculate"
570

    
571
    # TODO: check if we're really been called with the instance locks held
572

    
573
    # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
574
    # future we might want to have different behaviors depending on the value
575
    # of self.recalculate_locks[locking.LEVEL_NODE]
576
    wanted_nodes = []
577
    for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
578
      instance = self.context.cfg.GetInstanceInfo(instance_name)
579
      wanted_nodes.append(instance.primary_node)
580
      if not primary_only:
581
        wanted_nodes.extend(instance.secondary_nodes)
582

    
583
    if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE:
584
      self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
585
    elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND:
586
      self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
587

    
588
    del self.recalculate_locks[locking.LEVEL_NODE]
589

    
590

    
591
class NoHooksLU(LogicalUnit): # pylint: disable-msg=W0223
592
  """Simple LU which runs no hooks.
593

594
  This LU is intended as a parent for other LogicalUnits which will
595
  run no hooks, in order to reduce duplicate code.
596

597
  """
598
  HPATH = None
599
  HTYPE = None
600

    
601
  def BuildHooksEnv(self):
602
    """Empty BuildHooksEnv for NoHooksLu.
603

604
    This just raises an error.
605

606
    """
607
    assert False, "BuildHooksEnv called for NoHooksLUs"
608

    
609

    
610
class Tasklet:
611
  """Tasklet base class.
612

613
  Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
614
  they can mix legacy code with tasklets. Locking needs to be done in the LU,
615
  tasklets know nothing about locks.
616

617
  Subclasses must follow these rules:
618
    - Implement CheckPrereq
619
    - Implement Exec
620

621
  """
622
  def __init__(self, lu):
623
    self.lu = lu
624

    
625
    # Shortcuts
626
    self.cfg = lu.cfg
627
    self.rpc = lu.rpc
628

    
629
  def CheckPrereq(self):
630
    """Check prerequisites for this tasklets.
631

632
    This method should check whether the prerequisites for the execution of
633
    this tasklet are fulfilled. It can do internode communication, but it
634
    should be idempotent - no cluster or system changes are allowed.
635

636
    The method should raise errors.OpPrereqError in case something is not
637
    fulfilled. Its return value is ignored.
638

639
    This method should also update all parameters to their canonical form if it
640
    hasn't been done before.
641

642
    """
643
    pass
644

    
645
  def Exec(self, feedback_fn):
646
    """Execute the tasklet.
647

648
    This method should implement the actual work. It should raise
649
    errors.OpExecError for failures that are somewhat dealt with in code, or
650
    expected.
651

652
    """
653
    raise NotImplementedError
654

    
655

    
656
def _GetWantedNodes(lu, nodes):
657
  """Returns list of checked and expanded node names.
658

659
  @type lu: L{LogicalUnit}
660
  @param lu: the logical unit on whose behalf we execute
661
  @type nodes: list
662
  @param nodes: list of node names or None for all nodes
663
  @rtype: list
664
  @return: the list of nodes, sorted
665
  @raise errors.ProgrammerError: if the nodes parameter is wrong type
666

667
  """
668
  if not nodes:
669
    raise errors.ProgrammerError("_GetWantedNodes should only be called with a"
670
      " non-empty list of nodes whose name is to be expanded.")
671

    
672
  wanted = [_ExpandNodeName(lu.cfg, name) for name in nodes]
673
  return utils.NiceSort(wanted)
674

    
675

    
676
def _GetWantedInstances(lu, instances):
677
  """Returns list of checked and expanded instance names.
678

679
  @type lu: L{LogicalUnit}
680
  @param lu: the logical unit on whose behalf we execute
681
  @type instances: list
682
  @param instances: list of instance names or None for all instances
683
  @rtype: list
684
  @return: the list of instances, sorted
685
  @raise errors.OpPrereqError: if the instances parameter is wrong type
686
  @raise errors.OpPrereqError: if any of the passed instances is not found
687

688
  """
689
  if instances:
690
    wanted = [_ExpandInstanceName(lu.cfg, name) for name in instances]
691
  else:
692
    wanted = utils.NiceSort(lu.cfg.GetInstanceList())
693
  return wanted
694

    
695

    
696
def _GetUpdatedParams(old_params, update_dict,
697
                      use_default=True, use_none=False):
698
  """Return the new version of a parameter dictionary.
699

700
  @type old_params: dict
701
  @param old_params: old parameters
702
  @type update_dict: dict
703
  @param update_dict: dict containing new parameter values, or
704
      constants.VALUE_DEFAULT to reset the parameter to its default
705
      value
706
  @param use_default: boolean
707
  @type use_default: whether to recognise L{constants.VALUE_DEFAULT}
708
      values as 'to be deleted' values
709
  @param use_none: boolean
710
  @type use_none: whether to recognise C{None} values as 'to be
711
      deleted' values
712
  @rtype: dict
713
  @return: the new parameter dictionary
714

715
  """
716
  params_copy = copy.deepcopy(old_params)
717
  for key, val in update_dict.iteritems():
718
    if ((use_default and val == constants.VALUE_DEFAULT) or
719
        (use_none and val is None)):
720
      try:
721
        del params_copy[key]
722
      except KeyError:
723
        pass
724
    else:
725
      params_copy[key] = val
726
  return params_copy
727

    
728

    
729
def _CheckOutputFields(static, dynamic, selected):
730
  """Checks whether all selected fields are valid.
731

732
  @type static: L{utils.FieldSet}
733
  @param static: static fields set
734
  @type dynamic: L{utils.FieldSet}
735
  @param dynamic: dynamic fields set
736

737
  """
738
  f = utils.FieldSet()
739
  f.Extend(static)
740
  f.Extend(dynamic)
741

    
742
  delta = f.NonMatching(selected)
743
  if delta:
744
    raise errors.OpPrereqError("Unknown output fields selected: %s"
745
                               % ",".join(delta), errors.ECODE_INVAL)
746

    
747

    
748
def _CheckGlobalHvParams(params):
749
  """Validates that given hypervisor params are not global ones.
750

751
  This will ensure that instances don't get customised versions of
752
  global params.
753

754
  """
755
  used_globals = constants.HVC_GLOBALS.intersection(params)
756
  if used_globals:
757
    msg = ("The following hypervisor parameters are global and cannot"
758
           " be customized at instance level, please modify them at"
759
           " cluster level: %s" % utils.CommaJoin(used_globals))
760
    raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
761

    
762

    
763
def _CheckNodeOnline(lu, node):
764
  """Ensure that a given node is online.
765

766
  @param lu: the LU on behalf of which we make the check
767
  @param node: the node to check
768
  @raise errors.OpPrereqError: if the node is offline
769

770
  """
771
  if lu.cfg.GetNodeInfo(node).offline:
772
    raise errors.OpPrereqError("Can't use offline node %s" % node,
773
                               errors.ECODE_INVAL)
774

    
775

    
776
def _CheckNodeNotDrained(lu, node):
777
  """Ensure that a given node is not drained.
778

779
  @param lu: the LU on behalf of which we make the check
780
  @param node: the node to check
781
  @raise errors.OpPrereqError: if the node is drained
782

783
  """
784
  if lu.cfg.GetNodeInfo(node).drained:
785
    raise errors.OpPrereqError("Can't use drained node %s" % node,
786
                               errors.ECODE_INVAL)
787

    
788

    
789
def _CheckNodeHasOS(lu, node, os_name, force_variant):
790
  """Ensure that a node supports a given OS.
791

792
  @param lu: the LU on behalf of which we make the check
793
  @param node: the node to check
794
  @param os_name: the OS to query about
795
  @param force_variant: whether to ignore variant errors
796
  @raise errors.OpPrereqError: if the node is not supporting the OS
797

798
  """
799
  result = lu.rpc.call_os_get(node, os_name)
800
  result.Raise("OS '%s' not in supported OS list for node %s" %
801
               (os_name, node),
802
               prereq=True, ecode=errors.ECODE_INVAL)
803
  if not force_variant:
804
    _CheckOSVariant(result.payload, os_name)
805

    
806

    
807
def _RequireFileStorage():
808
  """Checks that file storage is enabled.
809

810
  @raise errors.OpPrereqError: when file storage is disabled
811

812
  """
813
  if not constants.ENABLE_FILE_STORAGE:
814
    raise errors.OpPrereqError("File storage disabled at configure time",
815
                               errors.ECODE_INVAL)
816

    
817

    
818
def _CheckDiskTemplate(template):
819
  """Ensure a given disk template is valid.
820

821
  """
822
  if template not in constants.DISK_TEMPLATES:
823
    msg = ("Invalid disk template name '%s', valid templates are: %s" %
824
           (template, utils.CommaJoin(constants.DISK_TEMPLATES)))
825
    raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
826
  if template == constants.DT_FILE:
827
    _RequireFileStorage()
828
  return True
829

    
830

    
831
def _CheckStorageType(storage_type):
832
  """Ensure a given storage type is valid.
833

834
  """
835
  if storage_type not in constants.VALID_STORAGE_TYPES:
836
    raise errors.OpPrereqError("Unknown storage type: %s" % storage_type,
837
                               errors.ECODE_INVAL)
838
  if storage_type == constants.ST_FILE:
839
    _RequireFileStorage()
840
  return True
841

    
842

    
843
def _GetClusterDomainSecret():
844
  """Reads the cluster domain secret.
845

846
  """
847
  return utils.ReadOneLineFile(constants.CLUSTER_DOMAIN_SECRET_FILE,
848
                               strict=True)
849

    
850

    
851
def _CheckInstanceDown(lu, instance, reason):
852
  """Ensure that an instance is not running."""
853
  if instance.admin_up:
854
    raise errors.OpPrereqError("Instance %s is marked to be up, %s" %
855
                               (instance.name, reason), errors.ECODE_STATE)
856

    
857
  pnode = instance.primary_node
858
  ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
859
  ins_l.Raise("Can't contact node %s for instance information" % pnode,
860
              prereq=True, ecode=errors.ECODE_ENVIRON)
861

    
862
  if instance.name in ins_l.payload:
863
    raise errors.OpPrereqError("Instance %s is running, %s" %
864
                               (instance.name, reason), errors.ECODE_STATE)
865

    
866

    
867
def _ExpandItemName(fn, name, kind):
868
  """Expand an item name.
869

870
  @param fn: the function to use for expansion
871
  @param name: requested item name
872
  @param kind: text description ('Node' or 'Instance')
873
  @return: the resolved (full) name
874
  @raise errors.OpPrereqError: if the item is not found
875

876
  """
877
  full_name = fn(name)
878
  if full_name is None:
879
    raise errors.OpPrereqError("%s '%s' not known" % (kind, name),
880
                               errors.ECODE_NOENT)
881
  return full_name
882

    
883

    
884
def _ExpandNodeName(cfg, name):
885
  """Wrapper over L{_ExpandItemName} for nodes."""
886
  return _ExpandItemName(cfg.ExpandNodeName, name, "Node")
887

    
888

    
889
def _ExpandInstanceName(cfg, name):
890
  """Wrapper over L{_ExpandItemName} for instance."""
891
  return _ExpandItemName(cfg.ExpandInstanceName, name, "Instance")
892

    
893

    
894
def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
895
                          memory, vcpus, nics, disk_template, disks,
896
                          bep, hvp, hypervisor_name):
897
  """Builds instance related env variables for hooks
898

899
  This builds the hook environment from individual variables.
900

901
  @type name: string
902
  @param name: the name of the instance
903
  @type primary_node: string
904
  @param primary_node: the name of the instance's primary node
905
  @type secondary_nodes: list
906
  @param secondary_nodes: list of secondary nodes as strings
907
  @type os_type: string
908
  @param os_type: the name of the instance's OS
909
  @type status: boolean
910
  @param status: the should_run status of the instance
911
  @type memory: string
912
  @param memory: the memory size of the instance
913
  @type vcpus: string
914
  @param vcpus: the count of VCPUs the instance has
915
  @type nics: list
916
  @param nics: list of tuples (ip, mac, mode, link) representing
917
      the NICs the instance has
918
  @type disk_template: string
919
  @param disk_template: the disk template of the instance
920
  @type disks: list
921
  @param disks: the list of (size, mode) pairs
922
  @type bep: dict
923
  @param bep: the backend parameters for the instance
924
  @type hvp: dict
925
  @param hvp: the hypervisor parameters for the instance
926
  @type hypervisor_name: string
927
  @param hypervisor_name: the hypervisor for the instance
928
  @rtype: dict
929
  @return: the hook environment for this instance
930

931
  """
932
  if status:
933
    str_status = "up"
934
  else:
935
    str_status = "down"
936
  env = {
937
    "OP_TARGET": name,
938
    "INSTANCE_NAME": name,
939
    "INSTANCE_PRIMARY": primary_node,
940
    "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
941
    "INSTANCE_OS_TYPE": os_type,
942
    "INSTANCE_STATUS": str_status,
943
    "INSTANCE_MEMORY": memory,
944
    "INSTANCE_VCPUS": vcpus,
945
    "INSTANCE_DISK_TEMPLATE": disk_template,
946
    "INSTANCE_HYPERVISOR": hypervisor_name,
947
  }
948

    
949
  if nics:
950
    nic_count = len(nics)
951
    for idx, (ip, mac, mode, link) in enumerate(nics):
952
      if ip is None:
953
        ip = ""
954
      env["INSTANCE_NIC%d_IP" % idx] = ip
955
      env["INSTANCE_NIC%d_MAC" % idx] = mac
956
      env["INSTANCE_NIC%d_MODE" % idx] = mode
957
      env["INSTANCE_NIC%d_LINK" % idx] = link
958
      if mode == constants.NIC_MODE_BRIDGED:
959
        env["INSTANCE_NIC%d_BRIDGE" % idx] = link
960
  else:
961
    nic_count = 0
962

    
963
  env["INSTANCE_NIC_COUNT"] = nic_count
964

    
965
  if disks:
966
    disk_count = len(disks)
967
    for idx, (size, mode) in enumerate(disks):
968
      env["INSTANCE_DISK%d_SIZE" % idx] = size
969
      env["INSTANCE_DISK%d_MODE" % idx] = mode
970
  else:
971
    disk_count = 0
972

    
973
  env["INSTANCE_DISK_COUNT"] = disk_count
974

    
975
  for source, kind in [(bep, "BE"), (hvp, "HV")]:
976
    for key, value in source.items():
977
      env["INSTANCE_%s_%s" % (kind, key)] = value
978

    
979
  return env
980

    
981

    
982
def _NICListToTuple(lu, nics):
983
  """Build a list of nic information tuples.
984

985
  This list is suitable to be passed to _BuildInstanceHookEnv or as a return
986
  value in LUQueryInstanceData.
987

988
  @type lu:  L{LogicalUnit}
989
  @param lu: the logical unit on whose behalf we execute
990
  @type nics: list of L{objects.NIC}
991
  @param nics: list of nics to convert to hooks tuples
992

993
  """
994
  hooks_nics = []
995
  cluster = lu.cfg.GetClusterInfo()
996
  for nic in nics:
997
    ip = nic.ip
998
    mac = nic.mac
999
    filled_params = cluster.SimpleFillNIC(nic.nicparams)
1000
    mode = filled_params[constants.NIC_MODE]
1001
    link = filled_params[constants.NIC_LINK]
1002
    hooks_nics.append((ip, mac, mode, link))
1003
  return hooks_nics
1004

    
1005

    
1006
def _BuildInstanceHookEnvByObject(lu, instance, override=None):
1007
  """Builds instance related env variables for hooks from an object.
1008

1009
  @type lu: L{LogicalUnit}
1010
  @param lu: the logical unit on whose behalf we execute
1011
  @type instance: L{objects.Instance}
1012
  @param instance: the instance for which we should build the
1013
      environment
1014
  @type override: dict
1015
  @param override: dictionary with key/values that will override
1016
      our values
1017
  @rtype: dict
1018
  @return: the hook environment dictionary
1019

1020
  """
1021
  cluster = lu.cfg.GetClusterInfo()
1022
  bep = cluster.FillBE(instance)
1023
  hvp = cluster.FillHV(instance)
1024
  args = {
1025
    'name': instance.name,
1026
    'primary_node': instance.primary_node,
1027
    'secondary_nodes': instance.secondary_nodes,
1028
    'os_type': instance.os,
1029
    'status': instance.admin_up,
1030
    'memory': bep[constants.BE_MEMORY],
1031
    'vcpus': bep[constants.BE_VCPUS],
1032
    'nics': _NICListToTuple(lu, instance.nics),
1033
    'disk_template': instance.disk_template,
1034
    'disks': [(disk.size, disk.mode) for disk in instance.disks],
1035
    'bep': bep,
1036
    'hvp': hvp,
1037
    'hypervisor_name': instance.hypervisor,
1038
  }
1039
  if override:
1040
    args.update(override)
1041
  return _BuildInstanceHookEnv(**args) # pylint: disable-msg=W0142
1042

    
1043

    
1044
def _AdjustCandidatePool(lu, exceptions):
1045
  """Adjust the candidate pool after node operations.
1046

1047
  """
1048
  mod_list = lu.cfg.MaintainCandidatePool(exceptions)
1049
  if mod_list:
1050
    lu.LogInfo("Promoted nodes to master candidate role: %s",
1051
               utils.CommaJoin(node.name for node in mod_list))
1052
    for name in mod_list:
1053
      lu.context.ReaddNode(name)
1054
  mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
1055
  if mc_now > mc_max:
1056
    lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
1057
               (mc_now, mc_max))
1058

    
1059

    
1060
def _DecideSelfPromotion(lu, exceptions=None):
1061
  """Decide whether I should promote myself as a master candidate.
1062

1063
  """
1064
  cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
1065
  mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
1066
  # the new node will increase mc_max with one, so:
1067
  mc_should = min(mc_should + 1, cp_size)
1068
  return mc_now < mc_should
1069

    
1070

    
1071
def _CheckNicsBridgesExist(lu, target_nics, target_node):
1072
  """Check that the brigdes needed by a list of nics exist.
1073

1074
  """
1075
  cluster = lu.cfg.GetClusterInfo()
1076
  paramslist = [cluster.SimpleFillNIC(nic.nicparams) for nic in target_nics]
1077
  brlist = [params[constants.NIC_LINK] for params in paramslist
1078
            if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
1079
  if brlist:
1080
    result = lu.rpc.call_bridges_exist(target_node, brlist)
1081
    result.Raise("Error checking bridges on destination node '%s'" %
1082
                 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
1083

    
1084

    
1085
def _CheckInstanceBridgesExist(lu, instance, node=None):
1086
  """Check that the brigdes needed by an instance exist.
1087

1088
  """
1089
  if node is None:
1090
    node = instance.primary_node
1091
  _CheckNicsBridgesExist(lu, instance.nics, node)
1092

    
1093

    
1094
def _CheckOSVariant(os_obj, name):
1095
  """Check whether an OS name conforms to the os variants specification.
1096

1097
  @type os_obj: L{objects.OS}
1098
  @param os_obj: OS object to check
1099
  @type name: string
1100
  @param name: OS name passed by the user, to check for validity
1101

1102
  """
1103
  if not os_obj.supported_variants:
1104
    return
1105
  variant = objects.OS.GetVariant(name)
1106
  if not variant:
1107
    raise errors.OpPrereqError("OS name must include a variant",
1108
                               errors.ECODE_INVAL)
1109

    
1110
  if variant not in os_obj.supported_variants:
1111
    raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
1112

    
1113

    
1114
def _GetNodeInstancesInner(cfg, fn):
1115
  return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
1116

    
1117

    
1118
def _GetNodeInstances(cfg, node_name):
1119
  """Returns a list of all primary and secondary instances on a node.
1120

1121
  """
1122

    
1123
  return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
1124

    
1125

    
1126
def _GetNodePrimaryInstances(cfg, node_name):
1127
  """Returns primary instances on a node.
1128

1129
  """
1130
  return _GetNodeInstancesInner(cfg,
1131
                                lambda inst: node_name == inst.primary_node)
1132

    
1133

    
1134
def _GetNodeSecondaryInstances(cfg, node_name):
1135
  """Returns secondary instances on a node.
1136

1137
  """
1138
  return _GetNodeInstancesInner(cfg,
1139
                                lambda inst: node_name in inst.secondary_nodes)
1140

    
1141

    
1142
def _GetStorageTypeArgs(cfg, storage_type):
1143
  """Returns the arguments for a storage type.
1144

1145
  """
1146
  # Special case for file storage
1147
  if storage_type == constants.ST_FILE:
1148
    # storage.FileStorage wants a list of storage directories
1149
    return [[cfg.GetFileStorageDir()]]
1150

    
1151
  return []
1152

    
1153

    
1154
def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq):
1155
  faulty = []
1156

    
1157
  for dev in instance.disks:
1158
    cfg.SetDiskID(dev, node_name)
1159

    
1160
  result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks)
1161
  result.Raise("Failed to get disk status from node %s" % node_name,
1162
               prereq=prereq, ecode=errors.ECODE_ENVIRON)
1163

    
1164
  for idx, bdev_status in enumerate(result.payload):
1165
    if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
1166
      faulty.append(idx)
1167

    
1168
  return faulty
1169

    
1170

    
1171
def _CheckIAllocatorOrNode(lu, iallocator_slot, node_slot):
1172
  """Check the sanity of iallocator and node arguments and use the
1173
  cluster-wide iallocator if appropriate.
1174

1175
  Check that at most one of (iallocator, node) is specified. If none is
1176
  specified, then the LU's opcode's iallocator slot is filled with the
1177
  cluster-wide default iallocator.
1178

1179
  @type iallocator_slot: string
1180
  @param iallocator_slot: the name of the opcode iallocator slot
1181
  @type node_slot: string
1182
  @param node_slot: the name of the opcode target node slot
1183

1184
  """
1185
  node = getattr(lu.op, node_slot, None)
1186
  iallocator = getattr(lu.op, iallocator_slot, None)
1187

    
1188
  if node is not None and iallocator is not None:
1189
    raise errors.OpPrereqError("Do not specify both, iallocator and node.",
1190
                               errors.ECODE_INVAL)
1191
  elif node is None and iallocator is None:
1192
    default_iallocator = lu.cfg.GetDefaultIAllocator()
1193
    if default_iallocator:
1194
      setattr(lu.op, iallocator_slot, default_iallocator)
1195
    else:
1196
      raise errors.OpPrereqError("No iallocator or node given and no"
1197
                                 " cluster-wide default iallocator found."
1198
                                 " Please specify either an iallocator or a"
1199
                                 " node, or set a cluster-wide default"
1200
                                 " iallocator.")
1201

    
1202

    
1203
class LUPostInitCluster(LogicalUnit):
1204
  """Logical unit for running hooks after cluster initialization.
1205

1206
  """
1207
  HPATH = "cluster-init"
1208
  HTYPE = constants.HTYPE_CLUSTER
1209

    
1210
  def BuildHooksEnv(self):
1211
    """Build hooks env.
1212

1213
    """
1214
    env = {"OP_TARGET": self.cfg.GetClusterName()}
1215
    mn = self.cfg.GetMasterNode()
1216
    return env, [], [mn]
1217

    
1218
  def Exec(self, feedback_fn):
1219
    """Nothing to do.
1220

1221
    """
1222
    return True
1223

    
1224

    
1225
class LUDestroyCluster(LogicalUnit):
1226
  """Logical unit for destroying the cluster.
1227

1228
  """
1229
  HPATH = "cluster-destroy"
1230
  HTYPE = constants.HTYPE_CLUSTER
1231

    
1232
  def BuildHooksEnv(self):
1233
    """Build hooks env.
1234

1235
    """
1236
    env = {"OP_TARGET": self.cfg.GetClusterName()}
1237
    return env, [], []
1238

    
1239
  def CheckPrereq(self):
1240
    """Check prerequisites.
1241

1242
    This checks whether the cluster is empty.
1243

1244
    Any errors are signaled by raising errors.OpPrereqError.
1245

1246
    """
1247
    master = self.cfg.GetMasterNode()
1248

    
1249
    nodelist = self.cfg.GetNodeList()
1250
    if len(nodelist) != 1 or nodelist[0] != master:
1251
      raise errors.OpPrereqError("There are still %d node(s) in"
1252
                                 " this cluster." % (len(nodelist) - 1),
1253
                                 errors.ECODE_INVAL)
1254
    instancelist = self.cfg.GetInstanceList()
1255
    if instancelist:
1256
      raise errors.OpPrereqError("There are still %d instance(s) in"
1257
                                 " this cluster." % len(instancelist),
1258
                                 errors.ECODE_INVAL)
1259

    
1260
  def Exec(self, feedback_fn):
1261
    """Destroys the cluster.
1262

1263
    """
1264
    master = self.cfg.GetMasterNode()
1265
    modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
1266

    
1267
    # Run post hooks on master node before it's removed
1268
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
1269
    try:
1270
      hm.RunPhase(constants.HOOKS_PHASE_POST, [master])
1271
    except:
1272
      # pylint: disable-msg=W0702
1273
      self.LogWarning("Errors occurred running hooks on %s" % master)
1274

    
1275
    result = self.rpc.call_node_stop_master(master, False)
1276
    result.Raise("Could not disable the master role")
1277

    
1278
    if modify_ssh_setup:
1279
      priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
1280
      utils.CreateBackup(priv_key)
1281
      utils.CreateBackup(pub_key)
1282

    
1283
    return master
1284

    
1285

    
1286
def _VerifyCertificate(filename):
1287
  """Verifies a certificate for LUVerifyCluster.
1288

1289
  @type filename: string
1290
  @param filename: Path to PEM file
1291

1292
  """
1293
  try:
1294
    cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
1295
                                           utils.ReadFile(filename))
1296
  except Exception, err: # pylint: disable-msg=W0703
1297
    return (LUVerifyCluster.ETYPE_ERROR,
1298
            "Failed to load X509 certificate %s: %s" % (filename, err))
1299

    
1300
  (errcode, msg) = \
1301
    utils.VerifyX509Certificate(cert, constants.SSL_CERT_EXPIRATION_WARN,
1302
                                constants.SSL_CERT_EXPIRATION_ERROR)
1303

    
1304
  if msg:
1305
    fnamemsg = "While verifying %s: %s" % (filename, msg)
1306
  else:
1307
    fnamemsg = None
1308

    
1309
  if errcode is None:
1310
    return (None, fnamemsg)
1311
  elif errcode == utils.CERT_WARNING:
1312
    return (LUVerifyCluster.ETYPE_WARNING, fnamemsg)
1313
  elif errcode == utils.CERT_ERROR:
1314
    return (LUVerifyCluster.ETYPE_ERROR, fnamemsg)
1315

    
1316
  raise errors.ProgrammerError("Unhandled certificate error code %r" % errcode)
1317

    
1318

    
1319
class LUVerifyCluster(LogicalUnit):
1320
  """Verifies the cluster status.
1321

1322
  """
1323
  HPATH = "cluster-verify"
1324
  HTYPE = constants.HTYPE_CLUSTER
1325
  _OP_PARAMS = [
1326
    ("skip_checks", _EmptyList,
1327
     _TListOf(_TElemOf(constants.VERIFY_OPTIONAL_CHECKS))),
1328
    ("verbose", False, _TBool),
1329
    ("error_codes", False, _TBool),
1330
    ("debug_simulate_errors", False, _TBool),
1331
    ]
1332
  REQ_BGL = False
1333

    
1334
  TCLUSTER = "cluster"
1335
  TNODE = "node"
1336
  TINSTANCE = "instance"
1337

    
1338
  ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
1339
  ECLUSTERCERT = (TCLUSTER, "ECLUSTERCERT")
1340
  EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
1341
  EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
1342
  EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
1343
  EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
1344
  EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
1345
  EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE")
1346
  ENODEDRBD = (TNODE, "ENODEDRBD")
1347
  ENODEDRBDHELPER = (TNODE, "ENODEDRBDHELPER")
1348
  ENODEFILECHECK = (TNODE, "ENODEFILECHECK")
1349
  ENODEHOOKS = (TNODE, "ENODEHOOKS")
1350
  ENODEHV = (TNODE, "ENODEHV")
1351
  ENODELVM = (TNODE, "ENODELVM")
1352
  ENODEN1 = (TNODE, "ENODEN1")
1353
  ENODENET = (TNODE, "ENODENET")
1354
  ENODEOS = (TNODE, "ENODEOS")
1355
  ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE")
1356
  ENODEORPHANLV = (TNODE, "ENODEORPHANLV")
1357
  ENODERPC = (TNODE, "ENODERPC")
1358
  ENODESSH = (TNODE, "ENODESSH")
1359
  ENODEVERSION = (TNODE, "ENODEVERSION")
1360
  ENODESETUP = (TNODE, "ENODESETUP")
1361
  ENODETIME = (TNODE, "ENODETIME")
1362

    
1363
  ETYPE_FIELD = "code"
1364
  ETYPE_ERROR = "ERROR"
1365
  ETYPE_WARNING = "WARNING"
1366

    
1367
  class NodeImage(object):
1368
    """A class representing the logical and physical status of a node.
1369

1370
    @type name: string
1371
    @ivar name: the node name to which this object refers
1372
    @ivar volumes: a structure as returned from
1373
        L{ganeti.backend.GetVolumeList} (runtime)
1374
    @ivar instances: a list of running instances (runtime)
1375
    @ivar pinst: list of configured primary instances (config)
1376
    @ivar sinst: list of configured secondary instances (config)
1377
    @ivar sbp: diction of {secondary-node: list of instances} of all peers
1378
        of this node (config)
1379
    @ivar mfree: free memory, as reported by hypervisor (runtime)
1380
    @ivar dfree: free disk, as reported by the node (runtime)
1381
    @ivar offline: the offline status (config)
1382
    @type rpc_fail: boolean
1383
    @ivar rpc_fail: whether the RPC verify call was successfull (overall,
1384
        not whether the individual keys were correct) (runtime)
1385
    @type lvm_fail: boolean
1386
    @ivar lvm_fail: whether the RPC call didn't return valid LVM data
1387
    @type hyp_fail: boolean
1388
    @ivar hyp_fail: whether the RPC call didn't return the instance list
1389
    @type ghost: boolean
1390
    @ivar ghost: whether this is a known node or not (config)
1391
    @type os_fail: boolean
1392
    @ivar os_fail: whether the RPC call didn't return valid OS data
1393
    @type oslist: list
1394
    @ivar oslist: list of OSes as diagnosed by DiagnoseOS
1395

1396
    """
1397
    def __init__(self, offline=False, name=None):
1398
      self.name = name
1399
      self.volumes = {}
1400
      self.instances = []
1401
      self.pinst = []
1402
      self.sinst = []
1403
      self.sbp = {}
1404
      self.mfree = 0
1405
      self.dfree = 0
1406
      self.offline = offline
1407
      self.rpc_fail = False
1408
      self.lvm_fail = False
1409
      self.hyp_fail = False
1410
      self.ghost = False
1411
      self.os_fail = False
1412
      self.oslist = {}
1413

    
1414
  def ExpandNames(self):
1415
    self.needed_locks = {
1416
      locking.LEVEL_NODE: locking.ALL_SET,
1417
      locking.LEVEL_INSTANCE: locking.ALL_SET,
1418
    }
1419
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
1420

    
1421
  def _Error(self, ecode, item, msg, *args, **kwargs):
1422
    """Format an error message.
1423

1424
    Based on the opcode's error_codes parameter, either format a
1425
    parseable error code, or a simpler error string.
1426

1427
    This must be called only from Exec and functions called from Exec.
1428

1429
    """
1430
    ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
1431
    itype, etxt = ecode
1432
    # first complete the msg
1433
    if args:
1434
      msg = msg % args
1435
    # then format the whole message
1436
    if self.op.error_codes:
1437
      msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
1438
    else:
1439
      if item:
1440
        item = " " + item
1441
      else:
1442
        item = ""
1443
      msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
1444
    # and finally report it via the feedback_fn
1445
    self._feedback_fn("  - %s" % msg)
1446

    
1447
  def _ErrorIf(self, cond, *args, **kwargs):
1448
    """Log an error message if the passed condition is True.
1449

1450
    """
1451
    cond = bool(cond) or self.op.debug_simulate_errors
1452
    if cond:
1453
      self._Error(*args, **kwargs)
1454
    # do not mark the operation as failed for WARN cases only
1455
    if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
1456
      self.bad = self.bad or cond
1457

    
1458
  def _VerifyNode(self, ninfo, nresult):
1459
    """Perform some basic validation on data returned from a node.
1460

1461
      - check the result data structure is well formed and has all the
1462
        mandatory fields
1463
      - check ganeti version
1464

1465
    @type ninfo: L{objects.Node}
1466
    @param ninfo: the node to check
1467
    @param nresult: the results from the node
1468
    @rtype: boolean
1469
    @return: whether overall this call was successful (and we can expect
1470
         reasonable values in the respose)
1471

1472
    """
1473
    node = ninfo.name
1474
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1475

    
1476
    # main result, nresult should be a non-empty dict
1477
    test = not nresult or not isinstance(nresult, dict)
1478
    _ErrorIf(test, self.ENODERPC, node,
1479
                  "unable to verify node: no data returned")
1480
    if test:
1481
      return False
1482

    
1483
    # compares ganeti version
1484
    local_version = constants.PROTOCOL_VERSION
1485
    remote_version = nresult.get("version", None)
1486
    test = not (remote_version and
1487
                isinstance(remote_version, (list, tuple)) and
1488
                len(remote_version) == 2)
1489
    _ErrorIf(test, self.ENODERPC, node,
1490
             "connection to node returned invalid data")
1491
    if test:
1492
      return False
1493

    
1494
    test = local_version != remote_version[0]
1495
    _ErrorIf(test, self.ENODEVERSION, node,
1496
             "incompatible protocol versions: master %s,"
1497
             " node %s", local_version, remote_version[0])
1498
    if test:
1499
      return False
1500

    
1501
    # node seems compatible, we can actually try to look into its results
1502

    
1503
    # full package version
1504
    self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
1505
                  self.ENODEVERSION, node,
1506
                  "software version mismatch: master %s, node %s",
1507
                  constants.RELEASE_VERSION, remote_version[1],
1508
                  code=self.ETYPE_WARNING)
1509

    
1510
    hyp_result = nresult.get(constants.NV_HYPERVISOR, None)
1511
    if isinstance(hyp_result, dict):
1512
      for hv_name, hv_result in hyp_result.iteritems():
1513
        test = hv_result is not None
1514
        _ErrorIf(test, self.ENODEHV, node,
1515
                 "hypervisor %s verify failure: '%s'", hv_name, hv_result)
1516

    
1517

    
1518
    test = nresult.get(constants.NV_NODESETUP,
1519
                           ["Missing NODESETUP results"])
1520
    _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s",
1521
             "; ".join(test))
1522

    
1523
    return True
1524

    
1525
  def _VerifyNodeTime(self, ninfo, nresult,
1526
                      nvinfo_starttime, nvinfo_endtime):
1527
    """Check the node time.
1528

1529
    @type ninfo: L{objects.Node}
1530
    @param ninfo: the node to check
1531
    @param nresult: the remote results for the node
1532
    @param nvinfo_starttime: the start time of the RPC call
1533
    @param nvinfo_endtime: the end time of the RPC call
1534

1535
    """
1536
    node = ninfo.name
1537
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1538

    
1539
    ntime = nresult.get(constants.NV_TIME, None)
1540
    try:
1541
      ntime_merged = utils.MergeTime(ntime)
1542
    except (ValueError, TypeError):
1543
      _ErrorIf(True, self.ENODETIME, node, "Node returned invalid time")
1544
      return
1545

    
1546
    if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
1547
      ntime_diff = "%.01fs" % abs(nvinfo_starttime - ntime_merged)
1548
    elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
1549
      ntime_diff = "%.01fs" % abs(ntime_merged - nvinfo_endtime)
1550
    else:
1551
      ntime_diff = None
1552

    
1553
    _ErrorIf(ntime_diff is not None, self.ENODETIME, node,
1554
             "Node time diverges by at least %s from master node time",
1555
             ntime_diff)
1556

    
1557
  def _VerifyNodeLVM(self, ninfo, nresult, vg_name):
1558
    """Check the node time.
1559

1560
    @type ninfo: L{objects.Node}
1561
    @param ninfo: the node to check
1562
    @param nresult: the remote results for the node
1563
    @param vg_name: the configured VG name
1564

1565
    """
1566
    if vg_name is None:
1567
      return
1568

    
1569
    node = ninfo.name
1570
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1571

    
1572
    # checks vg existence and size > 20G
1573
    vglist = nresult.get(constants.NV_VGLIST, None)
1574
    test = not vglist
1575
    _ErrorIf(test, self.ENODELVM, node, "unable to check volume groups")
1576
    if not test:
1577
      vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
1578
                                            constants.MIN_VG_SIZE)
1579
      _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
1580

    
1581
    # check pv names
1582
    pvlist = nresult.get(constants.NV_PVLIST, None)
1583
    test = pvlist is None
1584
    _ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node")
1585
    if not test:
1586
      # check that ':' is not present in PV names, since it's a
1587
      # special character for lvcreate (denotes the range of PEs to
1588
      # use on the PV)
1589
      for _, pvname, owner_vg in pvlist:
1590
        test = ":" in pvname
1591
        _ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV"
1592
                 " '%s' of VG '%s'", pvname, owner_vg)
1593

    
1594
  def _VerifyNodeNetwork(self, ninfo, nresult):
1595
    """Check the node time.
1596

1597
    @type ninfo: L{objects.Node}
1598
    @param ninfo: the node to check
1599
    @param nresult: the remote results for the node
1600

1601
    """
1602
    node = ninfo.name
1603
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1604

    
1605
    test = constants.NV_NODELIST not in nresult
1606
    _ErrorIf(test, self.ENODESSH, node,
1607
             "node hasn't returned node ssh connectivity data")
1608
    if not test:
1609
      if nresult[constants.NV_NODELIST]:
1610
        for a_node, a_msg in nresult[constants.NV_NODELIST].items():
1611
          _ErrorIf(True, self.ENODESSH, node,
1612
                   "ssh communication with node '%s': %s", a_node, a_msg)
1613

    
1614
    test = constants.NV_NODENETTEST not in nresult
1615
    _ErrorIf(test, self.ENODENET, node,
1616
             "node hasn't returned node tcp connectivity data")
1617
    if not test:
1618
      if nresult[constants.NV_NODENETTEST]:
1619
        nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys())
1620
        for anode in nlist:
1621
          _ErrorIf(True, self.ENODENET, node,
1622
                   "tcp communication with node '%s': %s",
1623
                   anode, nresult[constants.NV_NODENETTEST][anode])
1624

    
1625
    test = constants.NV_MASTERIP not in nresult
1626
    _ErrorIf(test, self.ENODENET, node,
1627
             "node hasn't returned node master IP reachability data")
1628
    if not test:
1629
      if not nresult[constants.NV_MASTERIP]:
1630
        if node == self.master_node:
1631
          msg = "the master node cannot reach the master IP (not configured?)"
1632
        else:
1633
          msg = "cannot reach the master IP"
1634
        _ErrorIf(True, self.ENODENET, node, msg)
1635

    
1636

    
1637
  def _VerifyInstance(self, instance, instanceconfig, node_image):
1638
    """Verify an instance.
1639

1640
    This function checks to see if the required block devices are
1641
    available on the instance's node.
1642

1643
    """
1644
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1645
    node_current = instanceconfig.primary_node
1646

    
1647
    node_vol_should = {}
1648
    instanceconfig.MapLVsByNode(node_vol_should)
1649

    
1650
    for node in node_vol_should:
1651
      n_img = node_image[node]
1652
      if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1653
        # ignore missing volumes on offline or broken nodes
1654
        continue
1655
      for volume in node_vol_should[node]:
1656
        test = volume not in n_img.volumes
1657
        _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
1658
                 "volume %s missing on node %s", volume, node)
1659

    
1660
    if instanceconfig.admin_up:
1661
      pri_img = node_image[node_current]
1662
      test = instance not in pri_img.instances and not pri_img.offline
1663
      _ErrorIf(test, self.EINSTANCEDOWN, instance,
1664
               "instance not running on its primary node %s",
1665
               node_current)
1666

    
1667
    for node, n_img in node_image.items():
1668
      if (not node == node_current):
1669
        test = instance in n_img.instances
1670
        _ErrorIf(test, self.EINSTANCEWRONGNODE, instance,
1671
                 "instance should not run on node %s", node)
1672

    
1673
  def _VerifyOrphanVolumes(self, node_vol_should, node_image, reserved):
1674
    """Verify if there are any unknown volumes in the cluster.
1675

1676
    The .os, .swap and backup volumes are ignored. All other volumes are
1677
    reported as unknown.
1678

1679
    @type reserved: L{ganeti.utils.FieldSet}
1680
    @param reserved: a FieldSet of reserved volume names
1681

1682
    """
1683
    for node, n_img in node_image.items():
1684
      if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1685
        # skip non-healthy nodes
1686
        continue
1687
      for volume in n_img.volumes:
1688
        test = ((node not in node_vol_should or
1689
                volume not in node_vol_should[node]) and
1690
                not reserved.Matches(volume))
1691
        self._ErrorIf(test, self.ENODEORPHANLV, node,
1692
                      "volume %s is unknown", volume)
1693

    
1694
  def _VerifyOrphanInstances(self, instancelist, node_image):
1695
    """Verify the list of running instances.
1696

1697
    This checks what instances are running but unknown to the cluster.
1698

1699
    """
1700
    for node, n_img in node_image.items():
1701
      for o_inst in n_img.instances:
1702
        test = o_inst not in instancelist
1703
        self._ErrorIf(test, self.ENODEORPHANINSTANCE, node,
1704
                      "instance %s on node %s should not exist", o_inst, node)
1705

    
1706
  def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
1707
    """Verify N+1 Memory Resilience.
1708

1709
    Check that if one single node dies we can still start all the
1710
    instances it was primary for.
1711

1712
    """
1713
    for node, n_img in node_image.items():
1714
      # This code checks that every node which is now listed as
1715
      # secondary has enough memory to host all instances it is
1716
      # supposed to should a single other node in the cluster fail.
1717
      # FIXME: not ready for failover to an arbitrary node
1718
      # FIXME: does not support file-backed instances
1719
      # WARNING: we currently take into account down instances as well
1720
      # as up ones, considering that even if they're down someone
1721
      # might want to start them even in the event of a node failure.
1722
      for prinode, instances in n_img.sbp.items():
1723
        needed_mem = 0
1724
        for instance in instances:
1725
          bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
1726
          if bep[constants.BE_AUTO_BALANCE]:
1727
            needed_mem += bep[constants.BE_MEMORY]
1728
        test = n_img.mfree < needed_mem
1729
        self._ErrorIf(test, self.ENODEN1, node,
1730
                      "not enough memory on to accommodate"
1731
                      " failovers should peer node %s fail", prinode)
1732

    
1733
  def _VerifyNodeFiles(self, ninfo, nresult, file_list, local_cksum,
1734
                       master_files):
1735
    """Verifies and computes the node required file checksums.
1736

1737
    @type ninfo: L{objects.Node}
1738
    @param ninfo: the node to check
1739
    @param nresult: the remote results for the node
1740
    @param file_list: required list of files
1741
    @param local_cksum: dictionary of local files and their checksums
1742
    @param master_files: list of files that only masters should have
1743

1744
    """
1745
    node = ninfo.name
1746
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1747

    
1748
    remote_cksum = nresult.get(constants.NV_FILELIST, None)
1749
    test = not isinstance(remote_cksum, dict)
1750
    _ErrorIf(test, self.ENODEFILECHECK, node,
1751
             "node hasn't returned file checksum data")
1752
    if test:
1753
      return
1754

    
1755
    for file_name in file_list:
1756
      node_is_mc = ninfo.master_candidate
1757
      must_have = (file_name not in master_files) or node_is_mc
1758
      # missing
1759
      test1 = file_name not in remote_cksum
1760
      # invalid checksum
1761
      test2 = not test1 and remote_cksum[file_name] != local_cksum[file_name]
1762
      # existing and good
1763
      test3 = not test1 and remote_cksum[file_name] == local_cksum[file_name]
1764
      _ErrorIf(test1 and must_have, self.ENODEFILECHECK, node,
1765
               "file '%s' missing", file_name)
1766
      _ErrorIf(test2 and must_have, self.ENODEFILECHECK, node,
1767
               "file '%s' has wrong checksum", file_name)
1768
      # not candidate and this is not a must-have file
1769
      _ErrorIf(test2 and not must_have, self.ENODEFILECHECK, node,
1770
               "file '%s' should not exist on non master"
1771
               " candidates (and the file is outdated)", file_name)
1772
      # all good, except non-master/non-must have combination
1773
      _ErrorIf(test3 and not must_have, self.ENODEFILECHECK, node,
1774
               "file '%s' should not exist"
1775
               " on non master candidates", file_name)
1776

    
1777
  def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_helper,
1778
                      drbd_map):
1779
    """Verifies and the node DRBD status.
1780

1781
    @type ninfo: L{objects.Node}
1782
    @param ninfo: the node to check
1783
    @param nresult: the remote results for the node
1784
    @param instanceinfo: the dict of instances
1785
    @param drbd_helper: the configured DRBD usermode helper
1786
    @param drbd_map: the DRBD map as returned by
1787
        L{ganeti.config.ConfigWriter.ComputeDRBDMap}
1788

1789
    """
1790
    node = ninfo.name
1791
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1792

    
1793
    if drbd_helper:
1794
      helper_result = nresult.get(constants.NV_DRBDHELPER, None)
1795
      test = (helper_result == None)
1796
      _ErrorIf(test, self.ENODEDRBDHELPER, node,
1797
               "no drbd usermode helper returned")
1798
      if helper_result:
1799
        status, payload = helper_result
1800
        test = not status
1801
        _ErrorIf(test, self.ENODEDRBDHELPER, node,
1802
                 "drbd usermode helper check unsuccessful: %s", payload)
1803
        test = status and (payload != drbd_helper)
1804
        _ErrorIf(test, self.ENODEDRBDHELPER, node,
1805
                 "wrong drbd usermode helper: %s", payload)
1806

    
1807
    # compute the DRBD minors
1808
    node_drbd = {}
1809
    for minor, instance in drbd_map[node].items():
1810
      test = instance not in instanceinfo
1811
      _ErrorIf(test, self.ECLUSTERCFG, None,
1812
               "ghost instance '%s' in temporary DRBD map", instance)
1813
        # ghost instance should not be running, but otherwise we
1814
        # don't give double warnings (both ghost instance and
1815
        # unallocated minor in use)
1816
      if test:
1817
        node_drbd[minor] = (instance, False)
1818
      else:
1819
        instance = instanceinfo[instance]
1820
        node_drbd[minor] = (instance.name, instance.admin_up)
1821

    
1822
    # and now check them
1823
    used_minors = nresult.get(constants.NV_DRBDLIST, [])
1824
    test = not isinstance(used_minors, (tuple, list))
1825
    _ErrorIf(test, self.ENODEDRBD, node,
1826
             "cannot parse drbd status file: %s", str(used_minors))
1827
    if test:
1828
      # we cannot check drbd status
1829
      return
1830

    
1831
    for minor, (iname, must_exist) in node_drbd.items():
1832
      test = minor not in used_minors and must_exist
1833
      _ErrorIf(test, self.ENODEDRBD, node,
1834
               "drbd minor %d of instance %s is not active", minor, iname)
1835
    for minor in used_minors:
1836
      test = minor not in node_drbd
1837
      _ErrorIf(test, self.ENODEDRBD, node,
1838
               "unallocated drbd minor %d is in use", minor)
1839

    
1840
  def _UpdateNodeOS(self, ninfo, nresult, nimg):
1841
    """Builds the node OS structures.
1842

1843
    @type ninfo: L{objects.Node}
1844
    @param ninfo: the node to check
1845
    @param nresult: the remote results for the node
1846
    @param nimg: the node image object
1847

1848
    """
1849
    node = ninfo.name
1850
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1851

    
1852
    remote_os = nresult.get(constants.NV_OSLIST, None)
1853
    test = (not isinstance(remote_os, list) or
1854
            not compat.all(isinstance(v, list) and len(v) == 7
1855
                           for v in remote_os))
1856

    
1857
    _ErrorIf(test, self.ENODEOS, node,
1858
             "node hasn't returned valid OS data")
1859

    
1860
    nimg.os_fail = test
1861

    
1862
    if test:
1863
      return
1864

    
1865
    os_dict = {}
1866

    
1867
    for (name, os_path, status, diagnose,
1868
         variants, parameters, api_ver) in nresult[constants.NV_OSLIST]:
1869

    
1870
      if name not in os_dict:
1871
        os_dict[name] = []
1872

    
1873
      # parameters is a list of lists instead of list of tuples due to
1874
      # JSON lacking a real tuple type, fix it:
1875
      parameters = [tuple(v) for v in parameters]
1876
      os_dict[name].append((os_path, status, diagnose,
1877
                            set(variants), set(parameters), set(api_ver)))
1878

    
1879
    nimg.oslist = os_dict
1880

    
1881
  def _VerifyNodeOS(self, ninfo, nimg, base):
1882
    """Verifies the node OS list.
1883

1884
    @type ninfo: L{objects.Node}
1885
    @param ninfo: the node to check
1886
    @param nimg: the node image object
1887
    @param base: the 'template' node we match against (e.g. from the master)
1888

1889
    """
1890
    node = ninfo.name
1891
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1892

    
1893
    assert not nimg.os_fail, "Entered _VerifyNodeOS with failed OS rpc?"
1894

    
1895
    for os_name, os_data in nimg.oslist.items():
1896
      assert os_data, "Empty OS status for OS %s?!" % os_name
1897
      f_path, f_status, f_diag, f_var, f_param, f_api = os_data[0]
1898
      _ErrorIf(not f_status, self.ENODEOS, node,
1899
               "Invalid OS %s (located at %s): %s", os_name, f_path, f_diag)
1900
      _ErrorIf(len(os_data) > 1, self.ENODEOS, node,
1901
               "OS '%s' has multiple entries (first one shadows the rest): %s",
1902
               os_name, utils.CommaJoin([v[0] for v in os_data]))
1903
      # this will catched in backend too
1904
      _ErrorIf(compat.any(v >= constants.OS_API_V15 for v in f_api)
1905
               and not f_var, self.ENODEOS, node,
1906
               "OS %s with API at least %d does not declare any variant",
1907
               os_name, constants.OS_API_V15)
1908
      # comparisons with the 'base' image
1909
      test = os_name not in base.oslist
1910
      _ErrorIf(test, self.ENODEOS, node,
1911
               "Extra OS %s not present on reference node (%s)",
1912
               os_name, base.name)
1913
      if test:
1914
        continue
1915
      assert base.oslist[os_name], "Base node has empty OS status?"
1916
      _, b_status, _, b_var, b_param, b_api = base.oslist[os_name][0]
1917
      if not b_status:
1918
        # base OS is invalid, skipping
1919
        continue
1920
      for kind, a, b in [("API version", f_api, b_api),
1921
                         ("variants list", f_var, b_var),
1922
                         ("parameters", f_param, b_param)]:
1923
        _ErrorIf(a != b, self.ENODEOS, node,
1924
                 "OS %s %s differs from reference node %s: %s vs. %s",
1925
                 kind, os_name, base.name,
1926
                 utils.CommaJoin(a), utils.CommaJoin(b))
1927

    
1928
    # check any missing OSes
1929
    missing = set(base.oslist.keys()).difference(nimg.oslist.keys())
1930
    _ErrorIf(missing, self.ENODEOS, node,
1931
             "OSes present on reference node %s but missing on this node: %s",
1932
             base.name, utils.CommaJoin(missing))
1933

    
1934
  def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
1935
    """Verifies and updates the node volume data.
1936

1937
    This function will update a L{NodeImage}'s internal structures
1938
    with data from the remote call.
1939

1940
    @type ninfo: L{objects.Node}
1941
    @param ninfo: the node to check
1942
    @param nresult: the remote results for the node
1943
    @param nimg: the node image object
1944
    @param vg_name: the configured VG name
1945

1946
    """
1947
    node = ninfo.name
1948
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1949

    
1950
    nimg.lvm_fail = True
1951
    lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
1952
    if vg_name is None:
1953
      pass
1954
    elif isinstance(lvdata, basestring):
1955
      _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
1956
               utils.SafeEncode(lvdata))
1957
    elif not isinstance(lvdata, dict):
1958
      _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
1959
    else:
1960
      nimg.volumes = lvdata
1961
      nimg.lvm_fail = False
1962

    
1963
  def _UpdateNodeInstances(self, ninfo, nresult, nimg):
1964
    """Verifies and updates the node instance list.
1965

1966
    If the listing was successful, then updates this node's instance
1967
    list. Otherwise, it marks the RPC call as failed for the instance
1968
    list key.
1969

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

1975
    """
1976
    idata = nresult.get(constants.NV_INSTANCELIST, None)
1977
    test = not isinstance(idata, list)
1978
    self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed"
1979
                  " (instancelist): %s", utils.SafeEncode(str(idata)))
1980
    if test:
1981
      nimg.hyp_fail = True
1982
    else:
1983
      nimg.instances = idata
1984

    
1985
  def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
1986
    """Verifies and computes a node information map
1987

1988
    @type ninfo: L{objects.Node}
1989
    @param ninfo: the node to check
1990
    @param nresult: the remote results for the node
1991
    @param nimg: the node image object
1992
    @param vg_name: the configured VG name
1993

1994
    """
1995
    node = ninfo.name
1996
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1997

    
1998
    # try to read free memory (from the hypervisor)
1999
    hv_info = nresult.get(constants.NV_HVINFO, None)
2000
    test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
2001
    _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
2002
    if not test:
2003
      try:
2004
        nimg.mfree = int(hv_info["memory_free"])
2005
      except (ValueError, TypeError):
2006
        _ErrorIf(True, self.ENODERPC, node,
2007
                 "node returned invalid nodeinfo, check hypervisor")
2008

    
2009
    # FIXME: devise a free space model for file based instances as well
2010
    if vg_name is not None:
2011
      test = (constants.NV_VGLIST not in nresult or
2012
              vg_name not in nresult[constants.NV_VGLIST])
2013
      _ErrorIf(test, self.ENODELVM, node,
2014
               "node didn't return data for the volume group '%s'"
2015
               " - it is either missing or broken", vg_name)
2016
      if not test:
2017
        try:
2018
          nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
2019
        except (ValueError, TypeError):
2020
          _ErrorIf(True, self.ENODERPC, node,
2021
                   "node returned invalid LVM info, check LVM status")
2022

    
2023
  def BuildHooksEnv(self):
2024
    """Build hooks env.
2025

2026
    Cluster-Verify hooks just ran in the post phase and their failure makes
2027
    the output be logged in the verify output and the verification to fail.
2028

2029
    """
2030
    all_nodes = self.cfg.GetNodeList()
2031
    env = {
2032
      "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
2033
      }
2034
    for node in self.cfg.GetAllNodesInfo().values():
2035
      env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
2036

    
2037
    return env, [], all_nodes
2038

    
2039
  def Exec(self, feedback_fn):
2040
    """Verify integrity of cluster, performing various test on nodes.
2041

2042
    """
2043
    self.bad = False
2044
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
2045
    verbose = self.op.verbose
2046
    self._feedback_fn = feedback_fn
2047
    feedback_fn("* Verifying global settings")
2048
    for msg in self.cfg.VerifyConfig():
2049
      _ErrorIf(True, self.ECLUSTERCFG, None, msg)
2050

    
2051
    # Check the cluster certificates
2052
    for cert_filename in constants.ALL_CERT_FILES:
2053
      (errcode, msg) = _VerifyCertificate(cert_filename)
2054
      _ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
2055

    
2056
    vg_name = self.cfg.GetVGName()
2057
    drbd_helper = self.cfg.GetDRBDHelper()
2058
    hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
2059
    cluster = self.cfg.GetClusterInfo()
2060
    nodelist = utils.NiceSort(self.cfg.GetNodeList())
2061
    nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
2062
    instancelist = utils.NiceSort(self.cfg.GetInstanceList())
2063
    instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
2064
                        for iname in instancelist)
2065
    i_non_redundant = [] # Non redundant instances
2066
    i_non_a_balanced = [] # Non auto-balanced instances
2067
    n_offline = 0 # Count of offline nodes
2068
    n_drained = 0 # Count of nodes being drained
2069
    node_vol_should = {}
2070

    
2071
    # FIXME: verify OS list
2072
    # do local checksums
2073
    master_files = [constants.CLUSTER_CONF_FILE]
2074
    master_node = self.master_node = self.cfg.GetMasterNode()
2075
    master_ip = self.cfg.GetMasterIP()
2076

    
2077
    file_names = ssconf.SimpleStore().GetFileList()
2078
    file_names.extend(constants.ALL_CERT_FILES)
2079
    file_names.extend(master_files)
2080
    if cluster.modify_etc_hosts:
2081
      file_names.append(constants.ETC_HOSTS)
2082

    
2083
    local_checksums = utils.FingerprintFiles(file_names)
2084

    
2085
    feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
2086
    node_verify_param = {
2087
      constants.NV_FILELIST: file_names,
2088
      constants.NV_NODELIST: [node.name for node in nodeinfo
2089
                              if not node.offline],
2090
      constants.NV_HYPERVISOR: hypervisors,
2091
      constants.NV_NODENETTEST: [(node.name, node.primary_ip,
2092
                                  node.secondary_ip) for node in nodeinfo
2093
                                 if not node.offline],
2094
      constants.NV_INSTANCELIST: hypervisors,
2095
      constants.NV_VERSION: None,
2096
      constants.NV_HVINFO: self.cfg.GetHypervisorType(),
2097
      constants.NV_NODESETUP: None,
2098
      constants.NV_TIME: None,
2099
      constants.NV_MASTERIP: (master_node, master_ip),
2100
      constants.NV_OSLIST: None,
2101
      }
2102

    
2103
    if vg_name is not None:
2104
      node_verify_param[constants.NV_VGLIST] = None
2105
      node_verify_param[constants.NV_LVLIST] = vg_name
2106
      node_verify_param[constants.NV_PVLIST] = [vg_name]
2107
      node_verify_param[constants.NV_DRBDLIST] = None
2108

    
2109
    if drbd_helper:
2110
      node_verify_param[constants.NV_DRBDHELPER] = drbd_helper
2111

    
2112
    # Build our expected cluster state
2113
    node_image = dict((node.name, self.NodeImage(offline=node.offline,
2114
                                                 name=node.name))
2115
                      for node in nodeinfo)
2116

    
2117
    for instance in instancelist:
2118
      inst_config = instanceinfo[instance]
2119

    
2120
      for nname in inst_config.all_nodes:
2121
        if nname not in node_image:
2122
          # ghost node
2123
          gnode = self.NodeImage(name=nname)
2124
          gnode.ghost = True
2125
          node_image[nname] = gnode
2126

    
2127
      inst_config.MapLVsByNode(node_vol_should)
2128

    
2129
      pnode = inst_config.primary_node
2130
      node_image[pnode].pinst.append(instance)
2131

    
2132
      for snode in inst_config.secondary_nodes:
2133
        nimg = node_image[snode]
2134
        nimg.sinst.append(instance)
2135
        if pnode not in nimg.sbp:
2136
          nimg.sbp[pnode] = []
2137
        nimg.sbp[pnode].append(instance)
2138

    
2139
    # At this point, we have the in-memory data structures complete,
2140
    # except for the runtime information, which we'll gather next
2141

    
2142
    # Due to the way our RPC system works, exact response times cannot be
2143
    # guaranteed (e.g. a broken node could run into a timeout). By keeping the
2144
    # time before and after executing the request, we can at least have a time
2145
    # window.
2146
    nvinfo_starttime = time.time()
2147
    all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
2148
                                           self.cfg.GetClusterName())
2149
    nvinfo_endtime = time.time()
2150

    
2151
    all_drbd_map = self.cfg.ComputeDRBDMap()
2152

    
2153
    feedback_fn("* Verifying node status")
2154

    
2155
    refos_img = None
2156

    
2157
    for node_i in nodeinfo:
2158
      node = node_i.name
2159
      nimg = node_image[node]
2160

    
2161
      if node_i.offline:
2162
        if verbose:
2163
          feedback_fn("* Skipping offline node %s" % (node,))
2164
        n_offline += 1
2165
        continue
2166

    
2167
      if node == master_node:
2168
        ntype = "master"
2169
      elif node_i.master_candidate:
2170
        ntype = "master candidate"
2171
      elif node_i.drained:
2172
        ntype = "drained"
2173
        n_drained += 1
2174
      else:
2175
        ntype = "regular"
2176
      if verbose:
2177
        feedback_fn("* Verifying node %s (%s)" % (node, ntype))
2178

    
2179
      msg = all_nvinfo[node].fail_msg
2180
      _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
2181
      if msg:
2182
        nimg.rpc_fail = True
2183
        continue
2184

    
2185
      nresult = all_nvinfo[node].payload
2186

    
2187
      nimg.call_ok = self._VerifyNode(node_i, nresult)
2188
      self._VerifyNodeNetwork(node_i, nresult)
2189
      self._VerifyNodeLVM(node_i, nresult, vg_name)
2190
      self._VerifyNodeFiles(node_i, nresult, file_names, local_checksums,
2191
                            master_files)
2192
      self._VerifyNodeDrbd(node_i, nresult, instanceinfo, drbd_helper,
2193
                           all_drbd_map)
2194
      self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
2195

    
2196
      self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
2197
      self._UpdateNodeInstances(node_i, nresult, nimg)
2198
      self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
2199
      self._UpdateNodeOS(node_i, nresult, nimg)
2200
      if not nimg.os_fail:
2201
        if refos_img is None:
2202
          refos_img = nimg
2203
        self._VerifyNodeOS(node_i, nimg, refos_img)
2204

    
2205
    feedback_fn("* Verifying instance status")
2206
    for instance in instancelist:
2207
      if verbose:
2208
        feedback_fn("* Verifying instance %s" % instance)
2209
      inst_config = instanceinfo[instance]
2210
      self._VerifyInstance(instance, inst_config, node_image)
2211
      inst_nodes_offline = []
2212

    
2213
      pnode = inst_config.primary_node
2214
      pnode_img = node_image[pnode]
2215
      _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
2216
               self.ENODERPC, pnode, "instance %s, connection to"
2217
               " primary node failed", instance)
2218

    
2219
      if pnode_img.offline:
2220
        inst_nodes_offline.append(pnode)
2221

    
2222
      # If the instance is non-redundant we cannot survive losing its primary
2223
      # node, so we are not N+1 compliant. On the other hand we have no disk
2224
      # templates with more than one secondary so that situation is not well
2225
      # supported either.
2226
      # FIXME: does not support file-backed instances
2227
      if not inst_config.secondary_nodes:
2228
        i_non_redundant.append(instance)
2229
      _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT,
2230
               instance, "instance has multiple secondary nodes: %s",
2231
               utils.CommaJoin(inst_config.secondary_nodes),
2232
               code=self.ETYPE_WARNING)
2233

    
2234
      if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
2235
        i_non_a_balanced.append(instance)
2236

    
2237
      for snode in inst_config.secondary_nodes:
2238
        s_img = node_image[snode]
2239
        _ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode,
2240
                 "instance %s, connection to secondary node failed", instance)
2241

    
2242
        if s_img.offline:
2243
          inst_nodes_offline.append(snode)
2244

    
2245
      # warn that the instance lives on offline nodes
2246
      _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
2247
               "instance lives on offline node(s) %s",
2248
               utils.CommaJoin(inst_nodes_offline))
2249
      # ... or ghost nodes
2250
      for node in inst_config.all_nodes:
2251
        _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
2252
                 "instance lives on ghost node %s", node)
2253

    
2254
    feedback_fn("* Verifying orphan volumes")
2255
    reserved = utils.FieldSet(*cluster.reserved_lvs)
2256
    self._VerifyOrphanVolumes(node_vol_should, node_image, reserved)
2257

    
2258
    feedback_fn("* Verifying orphan instances")
2259
    self._VerifyOrphanInstances(instancelist, node_image)
2260

    
2261
    if constants.VERIFY_NPLUSONE_MEM not in self.op.skip_checks:
2262
      feedback_fn("* Verifying N+1 Memory redundancy")
2263
      self._VerifyNPlusOneMemory(node_image, instanceinfo)
2264

    
2265
    feedback_fn("* Other Notes")
2266
    if i_non_redundant:
2267
      feedback_fn("  - NOTICE: %d non-redundant instance(s) found."
2268
                  % len(i_non_redundant))
2269

    
2270
    if i_non_a_balanced:
2271
      feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
2272
                  % len(i_non_a_balanced))
2273

    
2274
    if n_offline:
2275
      feedback_fn("  - NOTICE: %d offline node(s) found." % n_offline)
2276

    
2277
    if n_drained:
2278
      feedback_fn("  - NOTICE: %d drained node(s) found." % n_drained)
2279

    
2280
    return not self.bad
2281

    
2282
  def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
2283
    """Analyze the post-hooks' result
2284

2285
    This method analyses the hook result, handles it, and sends some
2286
    nicely-formatted feedback back to the user.
2287

2288
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
2289
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
2290
    @param hooks_results: the results of the multi-node hooks rpc call
2291
    @param feedback_fn: function used send feedback back to the caller
2292
    @param lu_result: previous Exec result
2293
    @return: the new Exec result, based on the previous result
2294
        and hook results
2295

2296
    """
2297
    # We only really run POST phase hooks, and are only interested in
2298
    # their results
2299
    if phase == constants.HOOKS_PHASE_POST:
2300
      # Used to change hooks' output to proper indentation
2301
      indent_re = re.compile('^', re.M)
2302
      feedback_fn("* Hooks Results")
2303
      assert hooks_results, "invalid result from hooks"
2304

    
2305
      for node_name in hooks_results:
2306
        res = hooks_results[node_name]
2307
        msg = res.fail_msg
2308
        test = msg and not res.offline
2309
        self._ErrorIf(test, self.ENODEHOOKS, node_name,
2310
                      "Communication failure in hooks execution: %s", msg)
2311
        if res.offline or msg:
2312
          # No need to investigate payload if node is offline or gave an error.
2313
          # override manually lu_result here as _ErrorIf only
2314
          # overrides self.bad
2315
          lu_result = 1
2316
          continue
2317
        for script, hkr, output in res.payload:
2318
          test = hkr == constants.HKR_FAIL
2319
          self._ErrorIf(test, self.ENODEHOOKS, node_name,
2320
                        "Script %s failed, output:", script)
2321
          if test:
2322
            output = indent_re.sub('      ', output)
2323
            feedback_fn("%s" % output)
2324
            lu_result = 0
2325

    
2326
      return lu_result
2327

    
2328

    
2329
class LUVerifyDisks(NoHooksLU):
2330
  """Verifies the cluster disks status.
2331

2332
  """
2333
  REQ_BGL = False
2334

    
2335
  def ExpandNames(self):
2336
    self.needed_locks = {
2337
      locking.LEVEL_NODE: locking.ALL_SET,
2338
      locking.LEVEL_INSTANCE: locking.ALL_SET,
2339
    }
2340
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
2341

    
2342
  def Exec(self, feedback_fn):
2343
    """Verify integrity of cluster disks.
2344

2345
    @rtype: tuple of three items
2346
    @return: a tuple of (dict of node-to-node_error, list of instances
2347
        which need activate-disks, dict of instance: (node, volume) for
2348
        missing volumes
2349

2350
    """
2351
    result = res_nodes, res_instances, res_missing = {}, [], {}
2352

    
2353
    vg_name = self.cfg.GetVGName()
2354
    nodes = utils.NiceSort(self.cfg.GetNodeList())
2355
    instances = [self.cfg.GetInstanceInfo(name)
2356
                 for name in self.cfg.GetInstanceList()]
2357

    
2358
    nv_dict = {}
2359
    for inst in instances:
2360
      inst_lvs = {}
2361
      if (not inst.admin_up or
2362
          inst.disk_template not in constants.DTS_NET_MIRROR):
2363
        continue
2364
      inst.MapLVsByNode(inst_lvs)
2365
      # transform { iname: {node: [vol,],},} to {(node, vol): iname}
2366
      for node, vol_list in inst_lvs.iteritems():
2367
        for vol in vol_list:
2368
          nv_dict[(node, vol)] = inst
2369

    
2370
    if not nv_dict:
2371
      return result
2372

    
2373
    node_lvs = self.rpc.call_lv_list(nodes, vg_name)
2374

    
2375
    for node in nodes:
2376
      # node_volume
2377
      node_res = node_lvs[node]
2378
      if node_res.offline:
2379
        continue
2380
      msg = node_res.fail_msg
2381
      if msg:
2382
        logging.warning("Error enumerating LVs on node %s: %s", node, msg)
2383
        res_nodes[node] = msg
2384
        continue
2385

    
2386
      lvs = node_res.payload
2387
      for lv_name, (_, _, lv_online) in lvs.items():
2388
        inst = nv_dict.pop((node, lv_name), None)
2389
        if (not lv_online and inst is not None
2390
            and inst.name not in res_instances):
2391
          res_instances.append(inst.name)
2392

    
2393
    # any leftover items in nv_dict are missing LVs, let's arrange the
2394
    # data better
2395
    for key, inst in nv_dict.iteritems():
2396
      if inst.name not in res_missing:
2397
        res_missing[inst.name] = []
2398
      res_missing[inst.name].append(key)
2399

    
2400
    return result
2401

    
2402

    
2403
class LURepairDiskSizes(NoHooksLU):
2404
  """Verifies the cluster disks sizes.
2405

2406
  """
2407
  _OP_PARAMS = [("instances", _EmptyList, _TListOf(_TNonEmptyString))]
2408
  REQ_BGL = False
2409

    
2410
  def ExpandNames(self):
2411
    if self.op.instances:
2412
      self.wanted_names = []
2413
      for name in self.op.instances:
2414
        full_name = _ExpandInstanceName(self.cfg, name)
2415
        self.wanted_names.append(full_name)
2416
      self.needed_locks = {
2417
        locking.LEVEL_NODE: [],
2418
        locking.LEVEL_INSTANCE: self.wanted_names,
2419
        }
2420
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2421
    else:
2422
      self.wanted_names = None
2423
      self.needed_locks = {
2424
        locking.LEVEL_NODE: locking.ALL_SET,
2425
        locking.LEVEL_INSTANCE: locking.ALL_SET,
2426
        }
2427
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
2428

    
2429
  def DeclareLocks(self, level):
2430
    if level == locking.LEVEL_NODE and self.wanted_names is not None:
2431
      self._LockInstancesNodes(primary_only=True)
2432

    
2433
  def CheckPrereq(self):
2434
    """Check prerequisites.
2435

2436
    This only checks the optional instance list against the existing names.
2437

2438
    """
2439
    if self.wanted_names is None:
2440
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
2441

    
2442
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
2443
                             in self.wanted_names]
2444

    
2445
  def _EnsureChildSizes(self, disk):
2446
    """Ensure children of the disk have the needed disk size.
2447

2448
    This is valid mainly for DRBD8 and fixes an issue where the
2449
    children have smaller disk size.
2450

2451
    @param disk: an L{ganeti.objects.Disk} object
2452

2453
    """
2454
    if disk.dev_type == constants.LD_DRBD8:
2455
      assert disk.children, "Empty children for DRBD8?"
2456
      fchild = disk.children[0]
2457
      mismatch = fchild.size < disk.size
2458
      if mismatch:
2459
        self.LogInfo("Child disk has size %d, parent %d, fixing",
2460
                     fchild.size, disk.size)
2461
        fchild.size = disk.size
2462

    
2463
      # and we recurse on this child only, not on the metadev
2464
      return self._EnsureChildSizes(fchild) or mismatch
2465
    else:
2466
      return False
2467

    
2468
  def Exec(self, feedback_fn):
2469
    """Verify the size of cluster disks.
2470

2471
    """
2472
    # TODO: check child disks too
2473
    # TODO: check differences in size between primary/secondary nodes
2474
    per_node_disks = {}
2475
    for instance in self.wanted_instances:
2476
      pnode = instance.primary_node
2477
      if pnode not in per_node_disks:
2478
        per_node_disks[pnode] = []
2479
      for idx, disk in enumerate(instance.disks):
2480
        per_node_disks[pnode].append((instance, idx, disk))
2481

    
2482
    changed = []
2483
    for node, dskl in per_node_disks.items():
2484
      newl = [v[2].Copy() for v in dskl]
2485
      for dsk in newl:
2486
        self.cfg.SetDiskID(dsk, node)
2487
      result = self.rpc.call_blockdev_getsizes(node, newl)
2488
      if result.fail_msg:
2489
        self.LogWarning("Failure in blockdev_getsizes call to node"
2490
                        " %s, ignoring", node)
2491
        continue
2492
      if len(result.data) != len(dskl):
2493
        self.LogWarning("Invalid result from node %s, ignoring node results",
2494
                        node)
2495
        continue
2496
      for ((instance, idx, disk), size) in zip(dskl, result.data):
2497
        if size is None:
2498
          self.LogWarning("Disk %d of instance %s did not return size"
2499
                          " information, ignoring", idx, instance.name)
2500
          continue
2501
        if not isinstance(size, (int, long)):
2502
          self.LogWarning("Disk %d of instance %s did not return valid"
2503
                          " size information, ignoring", idx, instance.name)
2504
          continue
2505
        size = size >> 20
2506
        if size != disk.size:
2507
          self.LogInfo("Disk %d of instance %s has mismatched size,"
2508
                       " correcting: recorded %d, actual %d", idx,
2509
                       instance.name, disk.size, size)
2510
          disk.size = size
2511
          self.cfg.Update(instance, feedback_fn)
2512
          changed.append((instance.name, idx, size))
2513
        if self._EnsureChildSizes(disk):
2514
          self.cfg.Update(instance, feedback_fn)
2515
          changed.append((instance.name, idx, disk.size))
2516
    return changed
2517

    
2518

    
2519
class LURenameCluster(LogicalUnit):
2520
  """Rename the cluster.
2521

2522
  """
2523
  HPATH = "cluster-rename"
2524
  HTYPE = constants.HTYPE_CLUSTER
2525
  _OP_PARAMS = [("name", _NoDefault, _TNonEmptyString)]
2526

    
2527
  def BuildHooksEnv(self):
2528
    """Build hooks env.
2529

2530
    """
2531
    env = {
2532
      "OP_TARGET": self.cfg.GetClusterName(),
2533
      "NEW_NAME": self.op.name,
2534
      }
2535
    mn = self.cfg.GetMasterNode()
2536
    all_nodes = self.cfg.GetNodeList()
2537
    return env, [mn], all_nodes
2538

    
2539
  def CheckPrereq(self):
2540
    """Verify that the passed name is a valid one.
2541

2542
    """
2543
    hostname = netutils.GetHostInfo(self.op.name)
2544

    
2545
    new_name = hostname.name
2546
    self.ip = new_ip = hostname.ip
2547
    old_name = self.cfg.GetClusterName()
2548
    old_ip = self.cfg.GetMasterIP()
2549
    if new_name == old_name and new_ip == old_ip:
2550
      raise errors.OpPrereqError("Neither the name nor the IP address of the"
2551
                                 " cluster has changed",
2552
                                 errors.ECODE_INVAL)
2553
    if new_ip != old_ip:
2554
      if netutils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
2555
        raise errors.OpPrereqError("The given cluster IP address (%s) is"
2556
                                   " reachable on the network. Aborting." %
2557
                                   new_ip, errors.ECODE_NOTUNIQUE)
2558

    
2559
    self.op.name = new_name
2560

    
2561
  def Exec(self, feedback_fn):
2562
    """Rename the cluster.
2563

2564
    """
2565
    clustername = self.op.name
2566
    ip = self.ip
2567

    
2568
    # shutdown the master IP
2569
    master = self.cfg.GetMasterNode()
2570
    result = self.rpc.call_node_stop_master(master, False)
2571
    result.Raise("Could not disable the master role")
2572

    
2573
    try:
2574
      cluster = self.cfg.GetClusterInfo()
2575
      cluster.cluster_name = clustername
2576
      cluster.master_ip = ip
2577
      self.cfg.Update(cluster, feedback_fn)
2578

    
2579
      # update the known hosts file
2580
      ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
2581
      node_list = self.cfg.GetNodeList()
2582
      try:
2583
        node_list.remove(master)
2584
      except ValueError:
2585
        pass
2586
      result = self.rpc.call_upload_file(node_list,
2587
                                         constants.SSH_KNOWN_HOSTS_FILE)
2588
      for to_node, to_result in result.iteritems():
2589
        msg = to_result.fail_msg
2590
        if msg:
2591
          msg = ("Copy of file %s to node %s failed: %s" %
2592
                 (constants.SSH_KNOWN_HOSTS_FILE, to_node, msg))
2593
          self.proc.LogWarning(msg)
2594

    
2595
    finally:
2596
      result = self.rpc.call_node_start_master(master, False, False)
2597
      msg = result.fail_msg
2598
      if msg:
2599
        self.LogWarning("Could not re-enable the master role on"
2600
                        " the master, please restart manually: %s", msg)
2601

    
2602
    return clustername
2603

    
2604

    
2605
class LUSetClusterParams(LogicalUnit):
2606
  """Change the parameters of the cluster.
2607

2608
  """
2609
  HPATH = "cluster-modify"
2610
  HTYPE = constants.HTYPE_CLUSTER
2611
  _OP_PARAMS = [
2612
    ("vg_name", None, _TMaybeString),
2613
    ("enabled_hypervisors", None,
2614
     _TOr(_TAnd(_TListOf(_TElemOf(constants.HYPER_TYPES)), _TTrue), _TNone)),
2615
    ("hvparams", None, _TOr(_TDictOf(_TNonEmptyString, _TDict), _TNone)),
2616
    ("beparams", None, _TOr(_TDictOf(_TNonEmptyString, _TDict), _TNone)),
2617
    ("os_hvp", None, _TOr(_TDictOf(_TNonEmptyString, _TDict), _TNone)),
2618
    ("osparams", None, _TOr(_TDictOf(_TNonEmptyString, _TDict), _TNone)),
2619
    ("candidate_pool_size", None, _TOr(_TStrictPositiveInt, _TNone)),
2620
    ("uid_pool", None, _NoType),
2621
    ("add_uids", None, _NoType),
2622
    ("remove_uids", None, _NoType),
2623
    ("maintain_node_health", None, _TMaybeBool),
2624
    ("nicparams", None, _TOr(_TDict, _TNone)),
2625
    ("drbd_helper", None, _TOr(_TString, _TNone)),
2626
    ("default_iallocator", None, _TMaybeString),
2627
    ("reserved_lvs", None, _TOr(_TListOf(_TNonEmptyString), _TNone)),
2628
    ("hidden_os", None, _TOr(_TListOf(\
2629
          _TAnd(_TList,
2630
                _TIsLength(2),
2631
                _TMap(lambda v: v[0], _TElemOf(constants.DDMS_VALUES)))),
2632
          _TNone)),
2633
    ("blacklisted_os", None, _TOr(_TListOf(\
2634
          _TAnd(_TList,
2635
                _TIsLength(2),
2636
                _TMap(lambda v: v[0], _TElemOf(constants.DDMS_VALUES)))),
2637
          _TNone)),
2638
    ]
2639
  REQ_BGL = False
2640

    
2641
  def CheckArguments(self):
2642
    """Check parameters
2643

2644
    """
2645
    if self.op.uid_pool:
2646
      uidpool.CheckUidPool(self.op.uid_pool)
2647

    
2648
    if self.op.add_uids:
2649
      uidpool.CheckUidPool(self.op.add_uids)
2650

    
2651
    if self.op.remove_uids:
2652
      uidpool.CheckUidPool(self.op.remove_uids)
2653

    
2654
  def ExpandNames(self):
2655
    # FIXME: in the future maybe other cluster params won't require checking on
2656
    # all nodes to be modified.
2657
    self.needed_locks = {
2658
      locking.LEVEL_NODE: locking.ALL_SET,
2659
    }
2660
    self.share_locks[locking.LEVEL_NODE] = 1
2661

    
2662
  def BuildHooksEnv(self):
2663
    """Build hooks env.
2664

2665
    """
2666
    env = {
2667
      "OP_TARGET": self.cfg.GetClusterName(),
2668
      "NEW_VG_NAME": self.op.vg_name,
2669
      }
2670
    mn = self.cfg.GetMasterNode()
2671
    return env, [mn], [mn]
2672

    
2673
  def CheckPrereq(self):
2674
    """Check prerequisites.
2675

2676
    This checks whether the given params don't conflict and
2677
    if the given volume group is valid.
2678

2679
    """
2680
    if self.op.vg_name is not None and not self.op.vg_name:
2681
      if self.cfg.HasAnyDiskOfType(constants.LD_LV):
2682
        raise errors.OpPrereqError("Cannot disable lvm storage while lvm-based"
2683
                                   " instances exist", errors.ECODE_INVAL)
2684

    
2685
    if self.op.drbd_helper is not None and not self.op.drbd_helper:
2686
      if self.cfg.HasAnyDiskOfType(constants.LD_DRBD8):
2687
        raise errors.OpPrereqError("Cannot disable drbd helper while"
2688
                                   " drbd-based instances exist",
2689
                                   errors.ECODE_INVAL)
2690

    
2691
    node_list = self.acquired_locks[locking.LEVEL_NODE]
2692

    
2693
    # if vg_name not None, checks given volume group on all nodes
2694
    if self.op.vg_name:
2695
      vglist = self.rpc.call_vg_list(node_list)
2696
      for node in node_list:
2697
        msg = vglist[node].fail_msg
2698
        if msg:
2699
          # ignoring down node
2700
          self.LogWarning("Error while gathering data on node %s"
2701
                          " (ignoring node): %s", node, msg)
2702
          continue
2703
        vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
2704
                                              self.op.vg_name,
2705
                                              constants.MIN_VG_SIZE)
2706
        if vgstatus:
2707
          raise errors.OpPrereqError("Error on node '%s': %s" %
2708
                                     (node, vgstatus), errors.ECODE_ENVIRON)
2709

    
2710
    if self.op.drbd_helper:
2711
      # checks given drbd helper on all nodes
2712
      helpers = self.rpc.call_drbd_helper(node_list)
2713
      for node in node_list:
2714
        ninfo = self.cfg.GetNodeInfo(node)
2715
        if ninfo.offline:
2716
          self.LogInfo("Not checking drbd helper on offline node %s", node)
2717
          continue
2718
        msg = helpers[node].fail_msg
2719
        if msg:
2720
          raise errors.OpPrereqError("Error checking drbd helper on node"
2721
                                     " '%s': %s" % (node, msg),
2722
                                     errors.ECODE_ENVIRON)
2723
        node_helper = helpers[node].payload
2724
        if node_helper != self.op.drbd_helper:
2725
          raise errors.OpPrereqError("Error on node '%s': drbd helper is %s" %
2726
                                     (node, node_helper), errors.ECODE_ENVIRON)
2727

    
2728
    self.cluster = cluster = self.cfg.GetClusterInfo()
2729
    # validate params changes
2730
    if self.op.beparams:
2731
      utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
2732
      self.new_beparams = cluster.SimpleFillBE(self.op.beparams)
2733

    
2734
    if self.op.nicparams:
2735
      utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
2736
      self.new_nicparams = cluster.SimpleFillNIC(self.op.nicparams)
2737
      objects.NIC.CheckParameterSyntax(self.new_nicparams)
2738
      nic_errors = []
2739

    
2740
      # check all instances for consistency
2741
      for instance in self.cfg.GetAllInstancesInfo().values():
2742
        for nic_idx, nic in enumerate(instance.nics):
2743
          params_copy = copy.deepcopy(nic.nicparams)
2744
          params_filled = objects.FillDict(self.new_nicparams, params_copy)
2745

    
2746
          # check parameter syntax
2747
          try:
2748
            objects.NIC.CheckParameterSyntax(params_filled)
2749
          except errors.ConfigurationError, err:
2750
            nic_errors.append("Instance %s, nic/%d: %s" %
2751
                              (instance.name, nic_idx, err))
2752

    
2753
          # if we're moving instances to routed, check that they have an ip
2754
          target_mode = params_filled[constants.NIC_MODE]
2755
          if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
2756
            nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
2757
                              (instance.name, nic_idx))
2758
      if nic_errors:
2759
        raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
2760
                                   "\n".join(nic_errors))
2761

    
2762
    # hypervisor list/parameters
2763
    self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
2764
    if self.op.hvparams:
2765
      for hv_name, hv_dict in self.op.hvparams.items():
2766
        if hv_name not in self.new_hvparams:
2767
          self.new_hvparams[hv_name] = hv_dict
2768
        else:
2769
          self.new_hvparams[hv_name].update(hv_dict)
2770

    
2771
    # os hypervisor parameters
2772
    self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
2773
    if self.op.os_hvp:
2774
      for os_name, hvs in self.op.os_hvp.items():
2775
        if os_name not in self.new_os_hvp:
2776
          self.new_os_hvp[os_name] = hvs
2777
        else:
2778
          for hv_name, hv_dict in hvs.items():
2779
            if hv_name not in self.new_os_hvp[os_name]:
2780
              self.new_os_hvp[os_name][hv_name] = hv_dict
2781
            else:
2782
              self.new_os_hvp[os_name][hv_name].update(hv_dict)
2783

    
2784
    # os parameters
2785
    self.new_osp = objects.FillDict(cluster.osparams, {})
2786
    if self.op.osparams:
2787
      for os_name, osp in self.op.osparams.items():
2788
        if os_name not in self.new_osp:
2789
          self.new_osp[os_name] = {}
2790

    
2791
        self.new_osp[os_name] = _GetUpdatedParams(self.new_osp[os_name], osp,
2792
                                                  use_none=True)
2793

    
2794
        if not self.new_osp[os_name]:
2795
          # we removed all parameters
2796
          del self.new_osp[os_name]
2797
        else:
2798
          # check the parameter validity (remote check)
2799
          _CheckOSParams(self, False, [self.cfg.GetMasterNode()],
2800
                         os_name, self.new_osp[os_name])
2801

    
2802
    # changes to the hypervisor list
2803
    if self.op.enabled_hypervisors is not None:
2804
      self.hv_list = self.op.enabled_hypervisors
2805
      for hv in self.hv_list:
2806
        # if the hypervisor doesn't already exist in the cluster
2807
        # hvparams, we initialize it to empty, and then (in both
2808
        # cases) we make sure to fill the defaults, as we might not
2809
        # have a complete defaults list if the hypervisor wasn't
2810
        # enabled before
2811
        if hv not in new_hvp:
2812
          new_hvp[hv] = {}
2813
        new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
2814
        utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
2815
    else:
2816
      self.hv_list = cluster.enabled_hypervisors
2817

    
2818
    if self.op.hvparams or self.op.enabled_hypervisors is not None:
2819
      # either the enabled list has changed, or the parameters have, validate
2820
      for hv_name, hv_params in self.new_hvparams.items():
2821
        if ((self.op.hvparams and hv_name in self.op.hvparams) or
2822
            (self.op.enabled_hypervisors and
2823
             hv_name in self.op.enabled_hypervisors)):
2824
          # either this is a new hypervisor, or its parameters have changed
2825
          hv_class = hypervisor.GetHypervisor(hv_name)
2826
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2827
          hv_class.CheckParameterSyntax(hv_params)
2828
          _CheckHVParams(self, node_list, hv_name, hv_params)
2829

    
2830
    if self.op.os_hvp:
2831
      # no need to check any newly-enabled hypervisors, since the
2832
      # defaults have already been checked in the above code-block
2833
      for os_name, os_hvp in self.new_os_hvp.items():
2834
        for hv_name, hv_params in os_hvp.items():
2835
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2836
          # we need to fill in the new os_hvp on top of the actual hv_p
2837
          cluster_defaults = self.new_hvparams.get(hv_name, {})
2838
          new_osp = objects.FillDict(cluster_defaults, hv_params)
2839
          hv_class = hypervisor.GetHypervisor(hv_name)
2840
          hv_class.CheckParameterSyntax(new_osp)
2841
          _CheckHVParams(self, node_list, hv_name, new_osp)
2842

    
2843
    if self.op.default_iallocator:
2844
      alloc_script = utils.FindFile(self.op.default_iallocator,
2845
                                    constants.IALLOCATOR_SEARCH_PATH,
2846
                                    os.path.isfile)
2847
      if alloc_script is None:
2848
        raise errors.OpPrereqError("Invalid default iallocator script '%s'"
2849
                                   " specified" % self.op.default_iallocator,
2850
                                   errors.ECODE_INVAL)
2851

    
2852
  def Exec(self, feedback_fn):
2853
    """Change the parameters of the cluster.
2854

2855
    """
2856
    if self.op.vg_name is not None:
2857
      new_volume = self.op.vg_name
2858
      if not new_volume:
2859
        new_volume = None
2860
      if new_volume != self.cfg.GetVGName():
2861
        self.cfg.SetVGName(new_volume)
2862
      else:
2863
        feedback_fn("Cluster LVM configuration already in desired"
2864
                    " state, not changing")
2865
    if self.op.drbd_helper is not None:
2866
      new_helper = self.op.drbd_helper
2867
      if not new_helper:
2868
        new_helper = None
2869
      if new_helper != self.cfg.GetDRBDHelper():
2870
        self.cfg.SetDRBDHelper(new_helper)
2871
      else:
2872
        feedback_fn("Cluster DRBD helper already in desired state,"
2873
                    " not changing")
2874
    if self.op.hvparams:
2875
      self.cluster.hvparams = self.new_hvparams
2876
    if self.op.os_hvp:
2877
      self.cluster.os_hvp = self.new_os_hvp
2878
    if self.op.enabled_hypervisors is not None:
2879
      self.cluster.hvparams = self.new_hvparams
2880
      self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
2881
    if self.op.beparams:
2882
      self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
2883
    if self.op.nicparams:
2884
      self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
2885
    if self.op.osparams:
2886
      self.cluster.osparams = self.new_osp
2887

    
2888
    if self.op.candidate_pool_size is not None:
2889
      self.cluster.candidate_pool_size = self.op.candidate_pool_size
2890
      # we need to update the pool size here, otherwise the save will fail
2891
      _AdjustCandidatePool(self, [])
2892

    
2893
    if self.op.maintain_node_health is not None:
2894
      self.cluster.maintain_node_health = self.op.maintain_node_health
2895

    
2896
    if self.op.add_uids is not None:
2897
      uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
2898

    
2899
    if self.op.remove_uids is not None:
2900
      uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
2901

    
2902
    if self.op.uid_pool is not None:
2903
      self.cluster.uid_pool = self.op.uid_pool
2904

    
2905
    if self.op.default_iallocator is not None:
2906
      self.cluster.default_iallocator = self.op.default_iallocator
2907

    
2908
    if self.op.reserved_lvs is not None:
2909
      self.cluster.reserved_lvs = self.op.reserved_lvs
2910

    
2911
    def helper_os(aname, mods, desc):
2912
      desc += " OS list"
2913
      lst = getattr(self.cluster, aname)
2914
      for key, val in mods:
2915
        if key == constants.DDM_ADD:
2916
          if val in lst:
2917
            feedback_fn("OS %s already in %s, ignoring", val, desc)
2918
          else:
2919
            lst.append(val)
2920
        elif key == constants.DDM_REMOVE:
2921
          if val in lst:
2922
            lst.remove(val)
2923
          else:
2924
            feedback_fn("OS %s not found in %s, ignoring", val, desc)
2925
        else:
2926
          raise errors.ProgrammerError("Invalid modification '%s'" % key)
2927

    
2928
    if self.op.hidden_os:
2929
      helper_os("hidden_os", self.op.hidden_os, "hidden")
2930

    
2931
    if self.op.blacklisted_os:
2932
      helper_os("blacklisted_os", self.op.blacklisted_os, "blacklisted")
2933

    
2934
    self.cfg.Update(self.cluster, feedback_fn)
2935

    
2936

    
2937
def _RedistributeAncillaryFiles(lu, additional_nodes=None):
2938
  """Distribute additional files which are part of the cluster configuration.
2939

2940
  ConfigWriter takes care of distributing the config and ssconf files, but
2941
  there are more files which should be distributed to all nodes. This function
2942
  makes sure those are copied.
2943

2944
  @param lu: calling logical unit
2945
  @param additional_nodes: list of nodes not in the config to distribute to
2946

2947
  """
2948
  # 1. Gather target nodes
2949
  myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
2950
  dist_nodes = lu.cfg.GetOnlineNodeList()
2951
  if additional_nodes is not None:
2952
    dist_nodes.extend(additional_nodes)
2953
  if myself.name in dist_nodes:
2954
    dist_nodes.remove(myself.name)
2955

    
2956
  # 2. Gather files to distribute
2957
  dist_files = set([constants.ETC_HOSTS,
2958
                    constants.SSH_KNOWN_HOSTS_FILE,
2959
                    constants.RAPI_CERT_FILE,
2960
                    constants.RAPI_USERS_FILE,
2961
                    constants.CONFD_HMAC_KEY,
2962
                    constants.CLUSTER_DOMAIN_SECRET_FILE,
2963
                   ])
2964

    
2965
  enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
2966
  for hv_name in enabled_hypervisors:
2967
    hv_class = hypervisor.GetHypervisor(hv_name)
2968
    dist_files.update(hv_class.GetAncillaryFiles())
2969

    
2970
  # 3. Perform the files upload
2971
  for fname in dist_files:
2972
    if os.path.exists(fname):
2973
      result = lu.rpc.call_upload_file(dist_nodes, fname)
2974
      for to_node, to_result in result.items():
2975
        msg = to_result.fail_msg
2976
        if msg:
2977
          msg = ("Copy of file %s to node %s failed: %s" %
2978
                 (fname, to_node, msg))
2979
          lu.proc.LogWarning(msg)
2980

    
2981

    
2982
class LURedistributeConfig(NoHooksLU):
2983
  """Force the redistribution of cluster configuration.
2984

2985
  This is a very simple LU.
2986

2987
  """
2988
  REQ_BGL = False
2989

    
2990
  def ExpandNames(self):
2991
    self.needed_locks = {
2992
      locking.LEVEL_NODE: locking.ALL_SET,
2993
    }
2994
    self.share_locks[locking.LEVEL_NODE] = 1
2995

    
2996
  def Exec(self, feedback_fn):
2997
    """Redistribute the configuration.
2998

2999
    """
3000
    self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
3001
    _RedistributeAncillaryFiles(self)
3002

    
3003

    
3004
def _WaitForSync(lu, instance, disks=None, oneshot=False):
3005
  """Sleep and poll for an instance's disk to sync.
3006

3007
  """
3008
  if not instance.disks or disks is not None and not disks:
3009
    return True
3010

    
3011
  disks = _ExpandCheckDisks(instance, disks)
3012

    
3013
  if not oneshot:
3014
    lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
3015

    
3016
  node = instance.primary_node
3017

    
3018
  for dev in disks:
3019
    lu.cfg.SetDiskID(dev, node)
3020

    
3021
  # TODO: Convert to utils.Retry
3022

    
3023
  retries = 0
3024
  degr_retries = 10 # in seconds, as we sleep 1 second each time
3025
  while True:
3026
    max_time = 0
3027
    done = True
3028
    cumul_degraded = False
3029
    rstats = lu.rpc.call_blockdev_getmirrorstatus(node, disks)
3030
    msg = rstats.fail_msg
3031
    if msg:
3032
      lu.LogWarning("Can't get any data from node %s: %s", node, msg)
3033
      retries += 1
3034
      if retries >= 10:
3035
        raise errors.RemoteError("Can't contact node %s for mirror data,"
3036
                                 " aborting." % node)
3037
      time.sleep(6)
3038
      continue
3039
    rstats = rstats.payload
3040
    retries = 0
3041
    for i, mstat in enumerate(rstats):
3042
      if mstat is None:
3043
        lu.LogWarning("Can't compute data for node %s/%s",
3044
                           node, disks[i].iv_name)
3045
        continue
3046

    
3047
      cumul_degraded = (cumul_degraded or
3048
                        (mstat.is_degraded and mstat.sync_percent is None))
3049
      if mstat.sync_percent is not None:
3050
        done = False
3051
        if mstat.estimated_time is not None:
3052
          rem_time = ("%s remaining (estimated)" %
3053
                      utils.FormatSeconds(mstat.estimated_time))
3054
          max_time = mstat.estimated_time
3055
        else:
3056
          rem_time = "no time estimate"
3057
        lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
3058
                        (disks[i].iv_name, mstat.sync_percent, rem_time))
3059

    
3060
    # if we're done but degraded, let's do a few small retries, to
3061
    # make sure we see a stable and not transient situation; therefore
3062
    # we force restart of the loop
3063
    if (done or oneshot) and cumul_degraded and degr_retries > 0:
3064
      logging.info("Degraded disks found, %d retries left", degr_retries)
3065
      degr_retries -= 1
3066
      time.sleep(1)
3067
      continue
3068

    
3069
    if done or oneshot:
3070
      break
3071

    
3072
    time.sleep(min(60, max_time))
3073

    
3074
  if done:
3075
    lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
3076
  return not cumul_degraded
3077

    
3078

    
3079
def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
3080
  """Check that mirrors are not degraded.
3081

3082
  The ldisk parameter, if True, will change the test from the
3083
  is_degraded attribute (which represents overall non-ok status for
3084
  the device(s)) to the ldisk (representing the local storage status).
3085

3086
  """
3087
  lu.cfg.SetDiskID(dev, node)
3088

    
3089
  result = True
3090

    
3091
  if on_primary or dev.AssembleOnSecondary():
3092
    rstats = lu.rpc.call_blockdev_find(node, dev)
3093
    msg = rstats.fail_msg
3094
    if msg:
3095
      lu.LogWarning("Can't find disk on node %s: %s", node, msg)
3096
      result = False
3097
    elif not rstats.payload:
3098
      lu.LogWarning("Can't find disk on node %s", node)
3099
      result = False
3100
    else:
3101
      if ldisk:
3102
        result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
3103
      else:
3104
        result = result and not rstats.payload.is_degraded
3105

    
3106
  if dev.children:
3107
    for child in dev.children:
3108
      result = result and _CheckDiskConsistency(lu, child, node, on_primary)
3109

    
3110
  return result
3111

    
3112

    
3113
class LUDiagnoseOS(NoHooksLU):
3114
  """Logical unit for OS diagnose/query.
3115

3116
  """
3117
  _OP_PARAMS = [
3118
    _POutputFields,
3119
    ("names", _EmptyList, _TListOf(_TNonEmptyString)),
3120
    ]
3121
  REQ_BGL = False
3122
  _HID = "hidden"
3123
  _BLK = "blacklisted"
3124
  _VLD = "valid"
3125
  _FIELDS_STATIC = utils.FieldSet()
3126
  _FIELDS_DYNAMIC = utils.FieldSet("name", _VLD, "node_status", "variants",
3127
                                   "parameters", "api_versions", _HID, _BLK)
3128

    
3129
  def CheckArguments(self):
3130
    if self.op.names:
3131
      raise errors.OpPrereqError("Selective OS query not supported",
3132
                                 errors.ECODE_INVAL)
3133

    
3134
    _CheckOutputFields(static=self._FIELDS_STATIC,
3135
                       dynamic=self._FIELDS_DYNAMIC,
3136
                       selected=self.op.output_fields)
3137

    
3138
  def ExpandNames(self):
3139
    # Lock all nodes, in shared mode
3140
    # Temporary removal of locks, should be reverted later
3141
    # TODO: reintroduce locks when they are lighter-weight
3142
    self.needed_locks = {}
3143
    #self.share_locks[locking.LEVEL_NODE] = 1
3144
    #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3145

    
3146
  @staticmethod
3147
  def _DiagnoseByOS(rlist):
3148
    """Remaps a per-node return list into an a per-os per-node dictionary
3149

3150
    @param rlist: a map with node names as keys and OS objects as values
3151

3152
    @rtype: dict
3153
    @return: a dictionary with osnames as keys and as value another
3154
        map, with nodes as keys and tuples of (path, status, diagnose,
3155
        variants, parameters, api_versions) as values, eg::
3156

3157
          {"debian-etch": {"node1": [(/usr/lib/..., True, "", [], []),
3158
                                     (/srv/..., False, "invalid api")],
3159
                           "node2": [(/srv/..., True, "", [], [])]}
3160
          }
3161

3162
    """
3163
    all_os = {}
3164
    # we build here the list of nodes that didn't fail the RPC (at RPC
3165
    # level), so that nodes with a non-responding node daemon don't
3166
    # make all OSes invalid
3167
    good_nodes = [node_name for node_name in rlist
3168
                  if not rlist[node_name].fail_msg]
3169
    for node_name, nr in rlist.items():
3170
      if nr.fail_msg or not nr.payload:
3171
        continue
3172
      for (name, path, status, diagnose, variants,
3173
           params, api_versions) in nr.payload:
3174
        if name not in all_os:
3175
          # build a list of nodes for this os containing empty lists
3176
          # for each node in node_list
3177
          all_os[name] = {}
3178
          for nname in good_nodes:
3179
            all_os[name][nname] = []
3180
        # convert params from [name, help] to (name, help)
3181
        params = [tuple(v) for v in params]
3182
        all_os[name][node_name].append((path, status, diagnose,
3183
                                        variants, params, api_versions))
3184
    return all_os
3185

    
3186
  def Exec(self, feedback_fn):
3187
    """Compute the list of OSes.
3188

3189
    """
3190
    valid_nodes = [node for node in self.cfg.GetOnlineNodeList()]
3191
    node_data = self.rpc.call_os_diagnose(valid_nodes)
3192
    pol = self._DiagnoseByOS(node_data)
3193
    output = []
3194
    cluster = self.cfg.GetClusterInfo()
3195

    
3196
    for os_name in utils.NiceSort(pol.keys()):
3197
      os_data = pol[os_name]
3198
      row = []
3199
      valid = True
3200
      (variants, params, api_versions) = null_state = (set(), set(), set())
3201
      for idx, osl in enumerate(os_data.values()):
3202
        valid = bool(valid and osl and osl[0][1])
3203
        if not valid:
3204
          (variants, params, api_versions) = null_state
3205
          break
3206
        node_variants, node_params, node_api = osl[0][3:6]
3207
        if idx == 0: # first entry
3208
          variants = set(node_variants)
3209
          params = set(node_params)
3210
          api_versions = set(node_api)
3211
        else: # keep consistency
3212
          variants.intersection_update(node_variants)
3213
          params.intersection_update(node_params)
3214
          api_versions.intersection_update(node_api)
3215

    
3216
      is_hid = os_name in cluster.hidden_os
3217
      is_blk = os_name in cluster.blacklisted_os
3218
      if ((self._HID not in self.op.output_fields and is_hid) or
3219
          (self._BLK not in self.op.output_fields and is_blk) or
3220
          (self._VLD not in self.op.output_fields and not valid)):
3221
        continue
3222

    
3223
      for field in self.op.output_fields:
3224
        if field == "name":
3225
          val = os_name
3226
        elif field == self._VLD:
3227
          val = valid
3228
        elif field == "node_status":
3229
          # this is just a copy of the dict
3230
          val = {}
3231
          for node_name, nos_list in os_data.items():
3232
            val[node_name] = nos_list
3233
        elif field == "variants":
3234
          val = utils.NiceSort(list(variants))
3235
        elif field == "parameters":
3236
          val = list(params)
3237
        elif field == "api_versions":
3238
          val = list(api_versions)
3239
        elif field == self._HID:
3240
          val = is_hid
3241
        elif field == self._BLK:
3242
          val = is_blk
3243
        else:
3244
          raise errors.ParameterError(field)
3245
        row.append(val)
3246
      output.append(row)
3247

    
3248
    return output
3249

    
3250

    
3251
class LURemoveNode(LogicalUnit):
3252
  """Logical unit for removing a node.
3253

3254
  """
3255
  HPATH = "node-remove"
3256
  HTYPE = constants.HTYPE_NODE
3257
  _OP_PARAMS = [
3258
    _PNodeName,
3259
    ]
3260

    
3261
  def BuildHooksEnv(self):
3262
    """Build hooks env.
3263

3264
    This doesn't run on the target node in the pre phase as a failed
3265
    node would then be impossible to remove.
3266

3267
    """
3268
    env = {
3269
      "OP_TARGET": self.op.node_name,
3270
      "NODE_NAME": self.op.node_name,
3271
      }
3272
    all_nodes = self.cfg.GetNodeList()
3273
    try:
3274
      all_nodes.remove(self.op.node_name)
3275
    except ValueError:
3276
      logging.warning("Node %s which is about to be removed not found"
3277
                      " in the all nodes list", self.op.node_name)
3278
    return env, all_nodes, all_nodes
3279

    
3280
  def CheckPrereq(self):
3281
    """Check prerequisites.
3282

3283
    This checks:
3284
     - the node exists in the configuration
3285
     - it does not have primary or secondary instances
3286
     - it's not the master
3287

3288
    Any errors are signaled by raising errors.OpPrereqError.
3289

3290
    """
3291
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3292
    node = self.cfg.GetNodeInfo(self.op.node_name)
3293
    assert node is not None
3294

    
3295
    instance_list = self.cfg.GetInstanceList()
3296

    
3297
    masternode = self.cfg.GetMasterNode()
3298
    if node.name == masternode:
3299
      raise errors.OpPrereqError("Node is the master node,"
3300
                                 " you need to failover first.",
3301
                                 errors.ECODE_INVAL)
3302

    
3303
    for instance_name in instance_list:
3304
      instance = self.cfg.GetInstanceInfo(instance_name)
3305
      if node.name in instance.all_nodes:
3306
        raise errors.OpPrereqError("Instance %s is still running on the node,"
3307
                                   " please remove first." % instance_name,
3308
                                   errors.ECODE_INVAL)
3309
    self.op.node_name = node.name
3310
    self.node = node
3311

    
3312
  def Exec(self, feedback_fn):
3313
    """Removes the node from the cluster.
3314

3315
    """
3316
    node = self.node
3317
    logging.info("Stopping the node daemon and removing configs from node %s",
3318
                 node.name)
3319

    
3320
    modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
3321

    
3322
    # Promote nodes to master candidate as needed
3323
    _AdjustCandidatePool(self, exceptions=[node.name])
3324
    self.context.RemoveNode(node.name)
3325

    
3326
    # Run post hooks on the node before it's removed
3327
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
3328
    try:
3329
      hm.RunPhase(constants.HOOKS_PHASE_POST, [node.name])
3330
    except:
3331
      # pylint: disable-msg=W0702
3332
      self.LogWarning("Errors occurred running hooks on %s" % node.name)
3333

    
3334
    result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
3335
    msg = result.fail_msg
3336
    if msg:
3337
      self.LogWarning("Errors encountered on the remote node while leaving"
3338
                      " the cluster: %s", msg)
3339

    
3340
    # Remove node from our /etc/hosts
3341
    if self.cfg.GetClusterInfo().modify_etc_hosts:
3342
      # FIXME: this should be done via an rpc call to node daemon
3343
      utils.RemoveHostFromEtcHosts(node.name)
3344
      _RedistributeAncillaryFiles(self)
3345

    
3346

    
3347
class LUQueryNodes(NoHooksLU):
3348
  """Logical unit for querying nodes.
3349

3350
  """
3351
  # pylint: disable-msg=W0142
3352
  _OP_PARAMS = [
3353
    _POutputFields,
3354
    ("names", _EmptyList, _TListOf(_TNonEmptyString)),
3355
    ("use_locking", False, _TBool),
3356
    ]
3357
  REQ_BGL = False
3358

    
3359
  _SIMPLE_FIELDS = ["name", "serial_no", "ctime", "mtime", "uuid",
3360
                    "master_candidate", "offline", "drained"]
3361

    
3362
  _FIELDS_DYNAMIC = utils.FieldSet(
3363
    "dtotal", "dfree",
3364
    "mtotal", "mnode", "mfree",
3365
    "bootid",
3366
    "ctotal", "cnodes", "csockets",
3367
    )
3368

    
3369
  _FIELDS_STATIC = utils.FieldSet(*[
3370
    "pinst_cnt", "sinst_cnt",
3371
    "pinst_list", "sinst_list",
3372
    "pip", "sip", "tags",
3373
    "master",
3374
    "role"] + _SIMPLE_FIELDS
3375
    )
3376

    
3377
  def CheckArguments(self):
3378
    _CheckOutputFields(static=self._FIELDS_STATIC,
3379
                       dynamic=self._FIELDS_DYNAMIC,
3380
                       selected=self.op.output_fields)
3381

    
3382
  def ExpandNames(self):
3383
    self.needed_locks = {}
3384
    self.share_locks[locking.LEVEL_NODE] = 1
3385

    
3386
    if self.op.names:
3387
      self.wanted = _GetWantedNodes(self, self.op.names)
3388
    else:
3389
      self.wanted = locking.ALL_SET
3390

    
3391
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
3392
    self.do_locking = self.do_node_query and self.op.use_locking
3393
    if self.do_locking:
3394
      # if we don't request only static fields, we need to lock the nodes
3395
      self.needed_locks[locking.LEVEL_NODE] = self.wanted
3396

    
3397
  def Exec(self, feedback_fn):
3398
    """Computes the list of nodes and their attributes.
3399

3400
    """
3401
    all_info = self.cfg.GetAllNodesInfo()
3402
    if self.do_locking:
3403
      nodenames = self.acquired_locks[locking.LEVEL_NODE]
3404
    elif self.wanted != locking.ALL_SET:
3405
      nodenames = self.wanted
3406
      missing = set(nodenames).difference(all_info.keys())
3407
      if missing:
3408
        raise errors.OpExecError(
3409
          "Some nodes were removed before retrieving their data: %s" % missing)
3410
    else:
3411
      nodenames = all_info.keys()
3412

    
3413
    nodenames = utils.NiceSort(nodenames)
3414
    nodelist = [all_info[name] for name in nodenames]
3415

    
3416
    # begin data gathering
3417

    
3418
    if self.do_node_query:
3419
      live_data = {}
3420
      node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
3421
                                          self.cfg.GetHypervisorType())
3422
      for name in nodenames:
3423
        nodeinfo = node_data[name]
3424
        if not nodeinfo.fail_msg and nodeinfo.payload:
3425
          nodeinfo = nodeinfo.payload
3426
          fn = utils.TryConvert
3427
          live_data[name] = {
3428
            "mtotal": fn(int, nodeinfo.get('memory_total', None)),
3429
            "mnode": fn(int, nodeinfo.get('memory_dom0', None)),
3430
            "mfree": fn(int, nodeinfo.get('memory_free', None)),
3431
            "dtotal": fn(int, nodeinfo.get('vg_size', None)),
3432
            "dfree": fn(int, nodeinfo.get('vg_free', None)),
3433
            "ctotal": fn(int, nodeinfo.get('cpu_total', None)),
3434
            "bootid": nodeinfo.get('bootid', None),
3435
            "cnodes": fn(int, nodeinfo.get('cpu_nodes', None)),
3436
            "csockets": fn(int, nodeinfo.get('cpu_sockets', None)),
3437
            }
3438
        else:
3439
          live_data[name] = {}
3440
    else:
3441
      live_data = dict.fromkeys(nodenames, {})
3442

    
3443
    node_to_primary = dict([(name, set()) for name in nodenames])
3444
    node_to_secondary = dict([(name, set()) for name in nodenames])
3445

    
3446
    inst_fields = frozenset(("pinst_cnt", "pinst_list",
3447
                             "sinst_cnt", "sinst_list"))
3448
    if inst_fields & frozenset(self.op.output_fields):
3449
      inst_data = self.cfg.GetAllInstancesInfo()
3450

    
3451
      for inst in inst_data.values():
3452
        if inst.primary_node in node_to_primary:
3453
          node_to_primary[inst.primary_node].add(inst.name)
3454
        for secnode in inst.secondary_nodes:
3455
          if secnode in node_to_secondary:
3456
            node_to_secondary[secnode].add(inst.name)
3457

    
3458
    master_node = self.cfg.GetMasterNode()
3459

    
3460
    # end data gathering
3461

    
3462
    output = []
3463
    for node in nodelist:
3464
      node_output = []
3465
      for field in self.op.output_fields:
3466
        if field in self._SIMPLE_FIELDS:
3467
          val = getattr(node, field)
3468
        elif field == "pinst_list":
3469
          val = list(node_to_primary[node.name])
3470
        elif field == "sinst_list":
3471
          val = list(node_to_secondary[node.name])
3472
        elif field == "pinst_cnt":
3473
          val = len(node_to_primary[node.name])
3474
        elif field == "sinst_cnt":
3475
          val = len(node_to_secondary[node.name])
3476
        elif field == "pip":
3477
          val = node.primary_ip
3478
        elif field == "sip":
3479
          val = node.secondary_ip
3480
        elif field == "tags":
3481
          val = list(node.GetTags())
3482
        elif field == "master":
3483
          val = node.name == master_node
3484
        elif self._FIELDS_DYNAMIC.Matches(field):
3485
          val = live_data[node.name].get(field, None)
3486
        elif field == "role":
3487
          if node.name == master_node:
3488
            val = "M"
3489
          elif node.master_candidate:
3490
            val = "C"
3491
          elif node.drained:
3492
            val = "D"
3493
          elif node.offline:
3494
            val = "O"
3495
          else:
3496
            val = "R"
3497
        else:
3498
          raise errors.ParameterError(field)
3499
        node_output.append(val)
3500
      output.append(node_output)
3501

    
3502
    return output
3503

    
3504

    
3505
class LUQueryNodeVolumes(NoHooksLU):
3506
  """Logical unit for getting volumes on node(s).
3507

3508
  """
3509
  _OP_PARAMS = [
3510
    ("nodes", _EmptyList, _TListOf(_TNonEmptyString)),
3511
    ("output_fields", _NoDefault, _TListOf(_TNonEmptyString)),
3512
    ]
3513
  REQ_BGL = False
3514
  _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
3515
  _FIELDS_STATIC = utils.FieldSet("node")
3516

    
3517
  def CheckArguments(self):
3518
    _CheckOutputFields(static=self._FIELDS_STATIC,
3519
                       dynamic=self._FIELDS_DYNAMIC,
3520
                       selected=self.op.output_fields)
3521

    
3522
  def ExpandNames(self):
3523
    self.needed_locks = {}
3524
    self.share_locks[locking.LEVEL_NODE] = 1
3525
    if not self.op.nodes:
3526
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3527
    else:
3528
      self.needed_locks[locking.LEVEL_NODE] = \
3529
        _GetWantedNodes(self, self.op.nodes)
3530

    
3531
  def Exec(self, feedback_fn):
3532
    """Computes the list of nodes and their attributes.
3533

3534
    """
3535
    nodenames = self.acquired_locks[locking.LEVEL_NODE]
3536
    volumes = self.rpc.call_node_volumes(nodenames)
3537

    
3538
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
3539
             in self.cfg.GetInstanceList()]
3540

    
3541
    lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
3542

    
3543
    output = []
3544
    for node in nodenames:
3545
      nresult = volumes[node]
3546
      if nresult.offline:
3547
        continue
3548
      msg = nresult.fail_msg
3549
      if msg:
3550
        self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
3551
        continue
3552

    
3553
      node_vols = nresult.payload[:]
3554
      node_vols.sort(key=lambda vol: vol['dev'])
3555

    
3556
      for vol in node_vols:
3557
        node_output = []
3558
        for field in self.op.output_fields:
3559
          if field == "node":
3560
            val = node
3561
          elif field == "phys":
3562
            val = vol['dev']
3563
          elif field == "vg":
3564
            val = vol['vg']
3565
          elif field == "name":
3566
            val = vol['name']
3567
          elif field == "size":
3568
            val = int(float(vol['size']))
3569
          elif field == "instance":
3570
            for inst in ilist:
3571
              if node not in lv_by_node[inst]:
3572
                continue
3573
              if vol['name'] in lv_by_node[inst][node]:
3574
                val = inst.name
3575
                break
3576
            else:
3577
              val = '-'
3578
          else:
3579
            raise errors.ParameterError(field)
3580
          node_output.append(str(val))
3581

    
3582
        output.append(node_output)
3583

    
3584
    return output
3585

    
3586

    
3587
class LUQueryNodeStorage(NoHooksLU):
3588
  """Logical unit for getting information on storage units on node(s).
3589

3590
  """
3591
  _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
3592
  _OP_PARAMS = [
3593
    ("nodes", _EmptyList, _TListOf(_TNonEmptyString)),
3594
    ("storage_type", _NoDefault, _CheckStorageType),
3595
    ("output_fields", _NoDefault, _TListOf(_TNonEmptyString)),
3596
    ("name", None, _TMaybeString),
3597
    ]
3598
  REQ_BGL = False
3599

    
3600
  def CheckArguments(self):
3601
    _CheckOutputFields(static=self._FIELDS_STATIC,
3602
                       dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
3603
                       selected=self.op.output_fields)
3604

    
3605
  def ExpandNames(self):
3606
    self.needed_locks = {}
3607
    self.share_locks[locking.LEVEL_NODE] = 1
3608

    
3609
    if self.op.nodes:
3610
      self.needed_locks[locking.LEVEL_NODE] = \
3611
        _GetWantedNodes(self, self.op.nodes)
3612
    else:
3613
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3614

    
3615
  def Exec(self, feedback_fn):
3616
    """Computes the list of nodes and their attributes.
3617

3618
    """
3619
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
3620

    
3621
    # Always get name to sort by
3622
    if constants.SF_NAME in self.op.output_fields:
3623
      fields = self.op.output_fields[:]
3624
    else:
3625
      fields = [constants.SF_NAME] + self.op.output_fields
3626

    
3627
    # Never ask for node or type as it's only known to the LU
3628
    for extra in [constants.SF_NODE, constants.SF_TYPE]:
3629
      while extra in fields:
3630
        fields.remove(extra)
3631

    
3632
    field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
3633
    name_idx = field_idx[constants.SF_NAME]
3634

    
3635
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3636
    data = self.rpc.call_storage_list(self.nodes,
3637
                                      self.op.storage_type, st_args,
3638
                                      self.op.name, fields)
3639

    
3640
    result = []
3641

    
3642
    for node in utils.NiceSort(self.nodes):
3643
      nresult = data[node]
3644
      if nresult.offline:
3645
        continue
3646

    
3647
      msg = nresult.fail_msg
3648
      if msg:
3649
        self.LogWarning("Can't get storage data from node %s: %s", node, msg)
3650
        continue
3651

    
3652
      rows = dict([(row[name_idx], row) for row in nresult.payload])
3653

    
3654
      for name in utils.NiceSort(rows.keys()):
3655
        row = rows[name]
3656

    
3657
        out = []
3658

    
3659
        for field in self.op.output_fields:
3660
          if field == constants.SF_NODE:
3661
            val = node
3662
          elif field == constants.SF_TYPE:
3663
            val = self.op.storage_type
3664
          elif field in field_idx:
3665
            val = row[field_idx[field]]
3666
          else:
3667
            raise errors.ParameterError(field)
3668

    
3669
          out.append(val)
3670

    
3671
        result.append(out)
3672

    
3673
    return result
3674

    
3675

    
3676
class LUModifyNodeStorage(NoHooksLU):
3677
  """Logical unit for modifying a storage volume on a node.
3678

3679
  """
3680
  _OP_PARAMS = [
3681
    _PNodeName,
3682
    ("storage_type", _NoDefault, _CheckStorageType),
3683
    ("name", _NoDefault, _TNonEmptyString),
3684
    ("changes", _NoDefault, _TDict),
3685
    ]
3686
  REQ_BGL = False
3687

    
3688
  def CheckArguments(self):
3689
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3690

    
3691
    storage_type = self.op.storage_type
3692

    
3693
    try:
3694
      modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
3695
    except KeyError:
3696
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
3697
                                 " modified" % storage_type,
3698
                                 errors.ECODE_INVAL)
3699

    
3700
    diff = set(self.op.changes.keys()) - modifiable
3701
    if diff:
3702
      raise errors.OpPrereqError("The following fields can not be modified for"
3703
                                 " storage units of type '%s': %r" %
3704
                                 (storage_type, list(diff)),
3705
                                 errors.ECODE_INVAL)
3706

    
3707
  def ExpandNames(self):
3708
    self.needed_locks = {
3709
      locking.LEVEL_NODE: self.op.node_name,
3710
      }
3711

    
3712
  def Exec(self, feedback_fn):
3713
    """Computes the list of nodes and their attributes.
3714

3715
    """
3716
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3717
    result = self.rpc.call_storage_modify(self.op.node_name,
3718
                                          self.op.storage_type, st_args,
3719
                                          self.op.name, self.op.changes)
3720
    result.Raise("Failed to modify storage unit '%s' on %s" %
3721
                 (self.op.name, self.op.node_name))
3722

    
3723

    
3724
class LUAddNode(LogicalUnit):
3725
  """Logical unit for adding node to the cluster.
3726

3727
  """
3728
  HPATH = "node-add"
3729
  HTYPE = constants.HTYPE_NODE
3730
  _OP_PARAMS = [
3731
    _PNodeName,
3732
    ("primary_ip", None, _NoType),
3733
    ("secondary_ip", None, _TMaybeString),
3734
    ("readd", False, _TBool),
3735
    ]
3736

    
3737
  def CheckArguments(self):
3738
    # validate/normalize the node name
3739
    self.op.node_name = netutils.HostInfo.NormalizeName(self.op.node_name)
3740

    
3741
  def BuildHooksEnv(self):
3742
    """Build hooks env.
3743

3744
    This will run on all nodes before, and on all nodes + the new node after.
3745

3746
    """
3747
    env = {
3748
      "OP_TARGET": self.op.node_name,
3749
      "NODE_NAME": self.op.node_name,
3750
      "NODE_PIP": self.op.primary_ip,
3751
      "NODE_SIP": self.op.secondary_ip,
3752
      }
3753
    nodes_0 = self.cfg.GetNodeList()
3754
    nodes_1 = nodes_0 + [self.op.node_name, ]
3755
    return env, nodes_0, nodes_1
3756

    
3757
  def CheckPrereq(self):
3758
    """Check prerequisites.
3759

3760
    This checks:
3761
     - the new node is not already in the config
3762
     - it is resolvable
3763
     - its parameters (single/dual homed) matches the cluster
3764

3765
    Any errors are signaled by raising errors.OpPrereqError.
3766

3767
    """
3768
    node_name = self.op.node_name
3769
    cfg = self.cfg
3770

    
3771
    dns_data = netutils.GetHostInfo(node_name)
3772

    
3773
    node = dns_data.name
3774
    primary_ip = self.op.primary_ip = dns_data.ip
3775
    if self.op.secondary_ip is None:
3776
      self.op.secondary_ip = primary_ip
3777
    if not netutils.IsValidIP4(self.op.secondary_ip):
3778
      raise errors.OpPrereqError("Invalid secondary IP given",
3779
                                 errors.ECODE_INVAL)
3780
    secondary_ip = self.op.secondary_ip
3781

    
3782
    node_list = cfg.GetNodeList()
3783
    if not self.op.readd and node in node_list:
3784
      raise errors.OpPrereqError("Node %s is already in the configuration" %
3785
                                 node, errors.ECODE_EXISTS)
3786
    elif self.op.readd and node not in node_list:
3787
      raise errors.OpPrereqError("Node %s is not in the configuration" % node,
3788
                                 errors.ECODE_NOENT)
3789

    
3790
    self.changed_primary_ip = False
3791

    
3792
    for existing_node_name in node_list:
3793
      existing_node = cfg.GetNodeInfo(existing_node_name)
3794

    
3795
      if self.op.readd and node == existing_node_name:
3796
        if existing_node.secondary_ip != secondary_ip:
3797
          raise errors.OpPrereqError("Readded node doesn't have the same IP"
3798
                                     " address configuration as before",
3799
                                     errors.ECODE_INVAL)
3800
        if existing_node.primary_ip != primary_ip:
3801
          self.changed_primary_ip = True
3802

    
3803
        continue
3804

    
3805
      if (existing_node.primary_ip == primary_ip or
3806
          existing_node.secondary_ip == primary_ip or
3807
          existing_node.primary_ip == secondary_ip or
3808
          existing_node.secondary_ip == secondary_ip):
3809
        raise errors.OpPrereqError("New node ip address(es) conflict with"
3810
                                   " existing node %s" % existing_node.name,
3811
                                   errors.ECODE_NOTUNIQUE)
3812

    
3813
    # check that the type of the node (single versus dual homed) is the
3814
    # same as for the master
3815
    myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
3816
    master_singlehomed = myself.secondary_ip == myself.primary_ip
3817
    newbie_singlehomed = secondary_ip == primary_ip
3818
    if master_singlehomed != newbie_singlehomed:
3819
      if master_singlehomed:
3820
        raise errors.OpPrereqError("The master has no private ip but the"
3821
                                   " new node has one",
3822
                                   errors.ECODE_INVAL)
3823
      else:
3824
        raise errors.OpPrereqError("The master has a private ip but the"
3825
                                   " new node doesn't have one",
3826
                                   errors.ECODE_INVAL)
3827

    
3828
    # checks reachability
3829
    if not netutils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
3830
      raise errors.OpPrereqError("Node not reachable by ping",
3831
                                 errors.ECODE_ENVIRON)
3832

    
3833
    if not newbie_singlehomed:
3834
      # check reachability from my secondary ip to newbie's secondary ip
3835
      if not netutils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
3836
                           source=myself.secondary_ip):
3837
        raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
3838
                                   " based ping to noded port",
3839
                                   errors.ECODE_ENVIRON)
3840

    
3841
    if self.op.readd:
3842
      exceptions = [node]
3843
    else:
3844
      exceptions = []
3845

    
3846
    self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
3847

    
3848
    if self.op.readd:
3849
      self.new_node = self.cfg.GetNodeInfo(node)
3850
      assert self.new_node is not None, "Can't retrieve locked node %s" % node
3851
    else:
3852
      self.new_node = objects.Node(name=node,
3853
                                   primary_ip=primary_ip,
3854
                                   secondary_ip=secondary_ip,
3855
                                   master_candidate=self.master_candidate,
3856
                                   offline=False, drained=False)
3857

    
3858
  def Exec(self, feedback_fn):
3859
    """Adds the new node to the cluster.
3860

3861
    """
3862
    new_node = self.new_node
3863
    node = new_node.name
3864

    
3865
    # for re-adds, reset the offline/drained/master-candidate flags;
3866
    # we need to reset here, otherwise offline would prevent RPC calls
3867
    # later in the procedure; this also means that if the re-add
3868
    # fails, we are left with a non-offlined, broken node
3869
    if self.op.readd:
3870
      new_node.drained = new_node.offline = False # pylint: disable-msg=W0201
3871
      self.LogInfo("Readding a node, the offline/drained flags were reset")
3872
      # if we demote the node, we do cleanup later in the procedure
3873
      new_node.master_candidate = self.master_candidate
3874
      if self.changed_primary_ip:
3875
        new_node.primary_ip = self.op.primary_ip
3876

    
3877
    # notify the user about any possible mc promotion
3878
    if new_node.master_candidate:
3879
      self.LogInfo("Node will be a master candidate")
3880

    
3881
    # check connectivity
3882
    result = self.rpc.call_version([node])[node]
3883
    result.Raise("Can't get version information from node %s" % node)
3884
    if constants.PROTOCOL_VERSION == result.payload:
3885
      logging.info("Communication to node %s fine, sw version %s match",
3886
                   node, result.payload)
3887
    else:
3888
      raise errors.OpExecError("Version mismatch master version %s,"
3889
                               " node version %s" %
3890
                               (constants.PROTOCOL_VERSION, result.payload))
3891

    
3892
    # setup ssh on node
3893
    if self.cfg.GetClusterInfo().modify_ssh_setup:
3894
      logging.info("Copy ssh key to node %s", node)
3895
      priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
3896
      keyarray = []
3897
      keyfiles = [constants.SSH_HOST_DSA_PRIV, constants.SSH_HOST_DSA_PUB,
3898
                  constants.SSH_HOST_RSA_PRIV, constants.SSH_HOST_RSA_PUB,
3899
                  priv_key, pub_key]
3900

    
3901
      for i in keyfiles:
3902
        keyarray.append(utils.ReadFile(i))
3903

    
3904
      result = self.rpc.call_node_add(node, keyarray[0], keyarray[1],
3905
                                      keyarray[2], keyarray[3], keyarray[4],
3906
                                      keyarray[5])
3907
      result.Raise("Cannot transfer ssh keys to the new node")
3908

    
3909
    # Add node to our /etc/hosts, and add key to known_hosts
3910
    if self.cfg.GetClusterInfo().modify_etc_hosts:
3911
      # FIXME: this should be done via an rpc call to node daemon
3912
      utils.AddHostToEtcHosts(new_node.name)
3913

    
3914
    if new_node.secondary_ip != new_node.primary_ip:
3915
      result = self.rpc.call_node_has_ip_address(new_node.name,
3916
                                                 new_node.secondary_ip)
3917
      result.Raise("Failure checking secondary ip on node %s" % new_node.name,
3918
                   prereq=True, ecode=errors.ECODE_ENVIRON)
3919
      if not result.payload:
3920
        raise errors.OpExecError("Node claims it doesn't have the secondary ip"
3921
                                 " you gave (%s). Please fix and re-run this"
3922
                                 " command." % new_node.secondary_ip)
3923

    
3924
    node_verify_list = [self.cfg.GetMasterNode()]
3925
    node_verify_param = {
3926
      constants.NV_NODELIST: [node],
3927
      # TODO: do a node-net-test as well?
3928
    }
3929

    
3930
    result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
3931
                                       self.cfg.GetClusterName())
3932
    for verifier in node_verify_list:
3933
      result[verifier].Raise("Cannot communicate with node %s" % verifier)
3934
      nl_payload = result[verifier].payload[constants.NV_NODELIST]
3935
      if nl_payload:
3936
        for failed in nl_payload:
3937
          feedback_fn("ssh/hostname verification failed"
3938
                      " (checking from %s): %s" %
3939
                      (verifier, nl_payload[failed]))
3940
        raise errors.OpExecError("ssh/hostname verification failed.")
3941

    
3942
    if self.op.readd:
3943
      _RedistributeAncillaryFiles(self)
3944
      self.context.ReaddNode(new_node)
3945
      # make sure we redistribute the config
3946
      self.cfg.Update(new_node, feedback_fn)
3947
      # and make sure the new node will not have old files around
3948
      if not new_node.master_candidate:
3949
        result = self.rpc.call_node_demote_from_mc(new_node.name)
3950
        msg = result.fail_msg
3951
        if msg:
3952
          self.LogWarning("Node failed to demote itself from master"
3953
                          " candidate status: %s" % msg)
3954
    else:
3955
      _RedistributeAncillaryFiles(self, additional_nodes=[node])
3956
      self.context.AddNode(new_node, self.proc.GetECId())
3957

    
3958

    
3959
class LUSetNodeParams(LogicalUnit):
3960
  """Modifies the parameters of a node.
3961

3962
  """
3963
  HPATH = "node-modify"
3964
  HTYPE = constants.HTYPE_NODE
3965
  _OP_PARAMS = [
3966
    _PNodeName,
3967
    ("master_candidate", None, _TMaybeBool),
3968
    ("offline", None, _TMaybeBool),
3969
    ("drained", None, _TMaybeBool),
3970
    ("auto_promote", False, _TBool),
3971
    _PForce,
3972
    ]
3973
  REQ_BGL = False
3974

    
3975
  def CheckArguments(self):
3976
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3977
    all_mods = [self.op.offline, self.op.master_candidate, self.op.drained]
3978
    if all_mods.count(None) == 3:
3979
      raise errors.OpPrereqError("Please pass at least one modification",
3980
                                 errors.ECODE_INVAL)
3981
    if all_mods.count(True) > 1:
3982
      raise errors.OpPrereqError("Can't set the node into more than one"
3983
                                 " state at the same time",
3984
                                 errors.ECODE_INVAL)
3985

    
3986
    # Boolean value that tells us whether we're offlining or draining the node
3987
    self.offline_or_drain = (self.op.offline == True or
3988
                             self.op.drained == True)
3989
    self.deoffline_or_drain = (self.op.offline == False or
3990
                               self.op.drained == False)
3991
    self.might_demote = (self.op.master_candidate == False or
3992
                         self.offline_or_drain)
3993

    
3994
    self.lock_all = self.op.auto_promote and self.might_demote
3995

    
3996

    
3997
  def ExpandNames(self):
3998
    if self.lock_all:
3999
      self.needed_locks = {locking.LEVEL_NODE: locking.ALL_SET}
4000
    else:
4001
      self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
4002

    
4003
  def BuildHooksEnv(self):
4004
    """Build hooks env.
4005

4006
    This runs on the master node.
4007

4008
    """
4009
    env = {
4010
      "OP_TARGET": self.op.node_name,
4011
      "MASTER_CANDIDATE": str(self.op.master_candidate),
4012
      "OFFLINE": str(self.op.offline),
4013
      "DRAINED": str(self.op.drained),
4014
      }
4015
    nl = [self.cfg.GetMasterNode(),
4016
          self.op.node_name]
4017
    return env, nl, nl
4018

    
4019
  def CheckPrereq(self):
4020
    """Check prerequisites.
4021

4022
    This only checks the instance list against the existing names.
4023

4024
    """
4025
    node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
4026

    
4027
    if (self.op.master_candidate is not None or
4028
        self.op.drained is not None or
4029
        self.op.offline is not None):
4030
      # we can't change the master's node flags
4031
      if self.op.node_name == self.cfg.GetMasterNode():
4032
        raise errors.OpPrereqError("The master role can be changed"
4033
                                   " only via master-failover",
4034
                                   errors.ECODE_INVAL)
4035

    
4036

    
4037
    if node.master_candidate and self.might_demote and not self.lock_all:
4038
      assert not self.op.auto_promote, "auto-promote set but lock_all not"
4039
      # check if after removing the current node, we're missing master
4040
      # candidates
4041
      (mc_remaining, mc_should, _) = \
4042
          self.cfg.GetMasterCandidateStats(exceptions=[node.name])
4043
      if mc_remaining < mc_should:
4044
        raise errors.OpPrereqError("Not enough master candidates, please"
4045
                                   " pass auto_promote to allow promotion",
4046
                                   errors.ECODE_INVAL)
4047

    
4048
    if (self.op.master_candidate == True and
4049
        ((node.offline and not self.op.offline == False) or
4050
         (node.drained and not self.op.drained == False))):
4051
      raise errors.OpPrereqError("Node '%s' is offline or drained, can't set"
4052
                                 " to master_candidate" % node.name,
4053
                                 errors.ECODE_INVAL)
4054

    
4055
    # If we're being deofflined/drained, we'll MC ourself if needed
4056
    if (self.deoffline_or_drain and not self.offline_or_drain and not
4057
        self.op.master_candidate == True and not node.master_candidate):
4058
      self.op.master_candidate = _DecideSelfPromotion(self)
4059
      if self.op.master_candidate:
4060
        self.LogInfo("Autopromoting node to master candidate")
4061

    
4062
    return
4063

    
4064
  def Exec(self, feedback_fn):
4065
    """Modifies a node.
4066

4067
    """
4068
    node = self.node
4069

    
4070
    result = []
4071
    changed_mc = False
4072

    
4073
    if self.op.offline is not None:
4074
      node.offline = self.op.offline
4075
      result.append(("offline", str(self.op.offline)))
4076
      if self.op.offline == True:
4077
        if node.master_candidate:
4078
          node.master_candidate = False
4079
          changed_mc = True
4080
          result.append(("master_candidate", "auto-demotion due to offline"))
4081
        if node.drained:
4082
          node.drained = False
4083
          result.append(("drained", "clear drained status due to offline"))
4084

    
4085
    if self.op.master_candidate is not None:
4086
      node.master_candidate = self.op.master_candidate
4087
      changed_mc = True
4088
      result.append(("master_candidate", str(self.op.master_candidate)))
4089
      if self.op.master_candidate == False:
4090
        rrc = self.rpc.call_node_demote_from_mc(node.name)
4091
        msg = rrc.fail_msg
4092
        if msg:
4093
          self.LogWarning("Node failed to demote itself: %s" % msg)
4094

    
4095
    if self.op.drained is not None:
4096
      node.drained = self.op.drained
4097
      result.append(("drained", str(self.op.drained)))
4098
      if self.op.drained == True:
4099
        if node.master_candidate:
4100
          node.master_candidate = False
4101
          changed_mc = True
4102
          result.append(("master_candidate", "auto-demotion due to drain"))
4103
          rrc = self.rpc.call_node_demote_from_mc(node.name)
4104
          msg = rrc.fail_msg
4105
          if msg:
4106
            self.LogWarning("Node failed to demote itself: %s" % msg)
4107
        if node.offline:
4108
          node.offline = False
4109
          result.append(("offline", "clear offline status due to drain"))
4110

    
4111
    # we locked all nodes, we adjust the CP before updating this node
4112
    if self.lock_all:
4113
      _AdjustCandidatePool(self, [node.name])
4114

    
4115
    # this will trigger configuration file update, if needed
4116
    self.cfg.Update(node, feedback_fn)
4117

    
4118
    # this will trigger job queue propagation or cleanup
4119
    if changed_mc:
4120
      self.context.ReaddNode(node)
4121

    
4122
    return result
4123

    
4124

    
4125
class LUPowercycleNode(NoHooksLU):
4126
  """Powercycles a node.
4127

4128
  """
4129
  _OP_PARAMS = [
4130
    _PNodeName,
4131
    _PForce,
4132
    ]
4133
  REQ_BGL = False
4134

    
4135
  def CheckArguments(self):
4136
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
4137
    if self.op.node_name == self.cfg.GetMasterNode() and not self.op.force:
4138
      raise errors.OpPrereqError("The node is the master and the force"
4139
                                 " parameter was not set",
4140
                                 errors.ECODE_INVAL)
4141

    
4142
  def ExpandNames(self):
4143
    """Locking for PowercycleNode.
4144

4145
    This is a last-resort option and shouldn't block on other
4146
    jobs. Therefore, we grab no locks.
4147

4148
    """
4149
    self.needed_locks = {}
4150

    
4151
  def Exec(self, feedback_fn):
4152
    """Reboots a node.
4153

4154
    """
4155
    result = self.rpc.call_node_powercycle(self.op.node_name,
4156
                                           self.cfg.GetHypervisorType())
4157
    result.Raise("Failed to schedule the reboot")
4158
    return result.payload
4159

    
4160

    
4161
class LUQueryClusterInfo(NoHooksLU):
4162
  """Query cluster configuration.
4163

4164
  """
4165
  REQ_BGL = False
4166

    
4167
  def ExpandNames(self):
4168
    self.needed_locks = {}
4169

    
4170
  def Exec(self, feedback_fn):
4171
    """Return cluster config.
4172

4173
    """
4174
    cluster = self.cfg.GetClusterInfo()
4175
    os_hvp = {}
4176

    
4177
    # Filter just for enabled hypervisors
4178
    for os_name, hv_dict in cluster.os_hvp.items():
4179
      os_hvp[os_name] = {}
4180
      for hv_name, hv_params in hv_dict.items():
4181
        if hv_name in cluster.enabled_hypervisors:
4182
          os_hvp[os_name][hv_name] = hv_params
4183

    
4184
    result = {
4185
      "software_version": constants.RELEASE_VERSION,
4186
      "protocol_version": constants.PROTOCOL_VERSION,
4187
      "config_version": constants.CONFIG_VERSION,
4188
      "os_api_version": max(constants.OS_API_VERSIONS),
4189
      "export_version": constants.EXPORT_VERSION,
4190
      "architecture": (platform.architecture()[0], platform.machine()),
4191
      "name": cluster.cluster_name,
4192
      "master": cluster.master_node,
4193
      "default_hypervisor": cluster.enabled_hypervisors[0],
4194
      "enabled_hypervisors": cluster.enabled_hypervisors,
4195
      "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
4196
                        for hypervisor_name in cluster.enabled_hypervisors]),
4197
      "os_hvp": os_hvp,
4198
      "beparams": cluster.beparams,
4199
      "osparams": cluster.osparams,
4200
      "nicparams": cluster.nicparams,
4201
      "candidate_pool_size": cluster.candidate_pool_size,
4202
      "master_netdev": cluster.master_netdev,
4203
      "volume_group_name": cluster.volume_group_name,
4204
      "drbd_usermode_helper": cluster.drbd_usermode_helper,
4205
      "file_storage_dir": cluster.file_storage_dir,
4206
      "maintain_node_health": cluster.maintain_node_health,
4207
      "ctime": cluster.ctime,
4208
      "mtime": cluster.mtime,
4209
      "uuid": cluster.uuid,
4210
      "tags": list(cluster.GetTags()),
4211
      "uid_pool": cluster.uid_pool,
4212
      "default_iallocator": cluster.default_iallocator,
4213
      "reserved_lvs": cluster.reserved_lvs,
4214
      }
4215

    
4216
    return result
4217

    
4218

    
4219
class LUQueryConfigValues(NoHooksLU):
4220
  """Return configuration values.
4221

4222
  """
4223
  _OP_PARAMS = [_POutputFields]
4224
  REQ_BGL = False
4225
  _FIELDS_DYNAMIC = utils.FieldSet()
4226
  _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag",
4227
                                  "watcher_pause", "volume_group_name")
4228

    
4229
  def CheckArguments(self):
4230
    _CheckOutputFields(static=self._FIELDS_STATIC,
4231
                       dynamic=self._FIELDS_DYNAMIC,
4232
                       selected=self.op.output_fields)
4233

    
4234
  def ExpandNames(self):
4235
    self.needed_locks = {}
4236

    
4237
  def Exec(self, feedback_fn):
4238
    """Dump a representation of the cluster config to the standard output.
4239

4240
    """
4241
    values = []
4242
    for field in self.op.output_fields:
4243
      if field == "cluster_name":
4244
        entry = self.cfg.GetClusterName()
4245
      elif field == "master_node":
4246
        entry = self.cfg.GetMasterNode()
4247
      elif field == "drain_flag":
4248
        entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
4249
      elif field == "watcher_pause":
4250
        entry = utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
4251
      elif field == "volume_group_name":
4252
        entry = self.cfg.GetVGName()
4253
      else:
4254
        raise errors.ParameterError(field)
4255
      values.append(entry)
4256
    return values
4257

    
4258

    
4259
class LUActivateInstanceDisks(NoHooksLU):
4260
  """Bring up an instance's disks.
4261

4262
  """
4263
  _OP_PARAMS = [
4264
    _PInstanceName,
4265
    ("ignore_size", False, _TBool),
4266
    ]
4267
  REQ_BGL = False
4268

    
4269
  def ExpandNames(self):
4270
    self._ExpandAndLockInstance()
4271
    self.needed_locks[locking.LEVEL_NODE] = []
4272
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4273

    
4274
  def DeclareLocks(self, level):
4275
    if level == locking.LEVEL_NODE:
4276
      self._LockInstancesNodes()
4277

    
4278
  def CheckPrereq(self):
4279
    """Check prerequisites.
4280

4281
    This checks that the instance is in the cluster.
4282

4283
    """
4284
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4285
    assert self.instance is not None, \
4286
      "Cannot retrieve locked instance %s" % self.op.instance_name
4287
    _CheckNodeOnline(self, self.instance.primary_node)
4288

    
4289
  def Exec(self, feedback_fn):
4290
    """Activate the disks.
4291

4292
    """
4293
    disks_ok, disks_info = \
4294
              _AssembleInstanceDisks(self, self.instance,
4295
                                     ignore_size=self.op.ignore_size)
4296
    if not disks_ok:
4297
      raise errors.OpExecError("Cannot activate block devices")
4298

    
4299
    return disks_info
4300

    
4301

    
4302
def _AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False,
4303
                           ignore_size=False):
4304
  """Prepare the block devices for an instance.
4305

4306
  This sets up the block devices on all nodes.
4307

4308
  @type lu: L{LogicalUnit}
4309
  @param lu: the logical unit on whose behalf we execute
4310
  @type instance: L{objects.Instance}
4311
  @param instance: the instance for whose disks we assemble
4312
  @type disks: list of L{objects.Disk} or None
4313
  @param disks: which disks to assemble (or all, if None)
4314
  @type ignore_secondaries: boolean
4315
  @param ignore_secondaries: if true, errors on secondary nodes
4316
      won't result in an error return from the function
4317
  @type ignore_size: boolean
4318
  @param ignore_size: if true, the current known size of the disk
4319
      will not be used during the disk activation, useful for cases
4320
      when the size is wrong
4321
  @return: False if the operation failed, otherwise a list of
4322
      (host, instance_visible_name, node_visible_name)
4323
      with the mapping from node devices to instance devices
4324

4325
  """
4326
  device_info = []
4327
  disks_ok = True
4328
  iname = instance.name
4329
  disks = _ExpandCheckDisks(instance, disks)
4330

    
4331
  # With the two passes mechanism we try to reduce the window of
4332
  # opportunity for the race condition of switching DRBD to primary
4333
  # before handshaking occured, but we do not eliminate it
4334

    
4335
  # The proper fix would be to wait (with some limits) until the
4336
  # connection has been made and drbd transitions from WFConnection
4337
  # into any other network-connected state (Connected, SyncTarget,
4338
  # SyncSource, etc.)
4339

    
4340
  # 1st pass, assemble on all nodes in secondary mode
4341
  for inst_disk in disks:
4342
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
4343
      if ignore_size:
4344
        node_disk = node_disk.Copy()
4345
        node_disk.UnsetSize()
4346
      lu.cfg.SetDiskID(node_disk, node)
4347
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
4348
      msg = result.fail_msg
4349
      if msg:
4350
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
4351
                           " (is_primary=False, pass=1): %s",
4352
                           inst_disk.iv_name, node, msg)
4353
        if not ignore_secondaries:
4354
          disks_ok = False
4355

    
4356
  # FIXME: race condition on drbd migration to primary
4357

    
4358
  # 2nd pass, do only the primary node
4359
  for inst_disk in disks:
4360
    dev_path = None
4361

    
4362
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
4363
      if node != instance.primary_node:
4364
        continue
4365
      if ignore_size:
4366
        node_disk = node_disk.Copy()
4367
        node_disk.UnsetSize()
4368
      lu.cfg.SetDiskID(node_disk, node)
4369
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
4370
      msg = result.fail_msg
4371
      if msg:
4372
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
4373
                           " (is_primary=True, pass=2): %s",
4374
                           inst_disk.iv_name, node, msg)
4375
        disks_ok = False
4376
      else:
4377
        dev_path = result.payload
4378

    
4379
    device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
4380

    
4381
  # leave the disks configured for the primary node
4382
  # this is a workaround that would be fixed better by
4383
  # improving the logical/physical id handling
4384
  for disk in disks:
4385
    lu.cfg.SetDiskID(disk, instance.primary_node)
4386

    
4387
  return disks_ok, device_info
4388

    
4389

    
4390
def _StartInstanceDisks(lu, instance, force):
4391
  """Start the disks of an instance.
4392

4393
  """
4394
  disks_ok, _ = _AssembleInstanceDisks(lu, instance,
4395
                                           ignore_secondaries=force)
4396
  if not disks_ok:
4397
    _ShutdownInstanceDisks(lu, instance)
4398
    if force is not None and not force:
4399
      lu.proc.LogWarning("", hint="If the message above refers to a"
4400
                         " secondary node,"
4401
                         " you can retry the operation using '--force'.")
4402
    raise errors.OpExecError("Disk consistency error")
4403

    
4404

    
4405
class LUDeactivateInstanceDisks(NoHooksLU):
4406
  """Shutdown an instance's disks.
4407

4408
  """
4409
  _OP_PARAMS = [
4410
    _PInstanceName,
4411
    ]
4412
  REQ_BGL = False
4413

    
4414
  def ExpandNames(self):
4415
    self._ExpandAndLockInstance()
4416
    self.needed_locks[locking.LEVEL_NODE] = []
4417
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4418

    
4419
  def DeclareLocks(self, level):
4420
    if level == locking.LEVEL_NODE:
4421
      self._LockInstancesNodes()
4422

    
4423
  def CheckPrereq(self):
4424
    """Check prerequisites.
4425

4426
    This checks that the instance is in the cluster.
4427

4428
    """
4429
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4430
    assert self.instance is not None, \
4431
      "Cannot retrieve locked instance %s" % self.op.instance_name
4432

    
4433
  def Exec(self, feedback_fn):
4434
    """Deactivate the disks
4435

4436
    """
4437
    instance = self.instance
4438
    _SafeShutdownInstanceDisks(self, instance)
4439

    
4440

    
4441
def _SafeShutdownInstanceDisks(lu, instance, disks=None):
4442
  """Shutdown block devices of an instance.
4443

4444
  This function checks if an instance is running, before calling
4445
  _ShutdownInstanceDisks.
4446

4447
  """
4448
  _CheckInstanceDown(lu, instance, "cannot shutdown disks")
4449
  _ShutdownInstanceDisks(lu, instance, disks=disks)
4450

    
4451

    
4452
def _ExpandCheckDisks(instance, disks):
4453
  """Return the instance disks selected by the disks list
4454

4455
  @type disks: list of L{objects.Disk} or None
4456
  @param disks: selected disks
4457
  @rtype: list of L{objects.Disk}
4458
  @return: selected instance disks to act on
4459

4460
  """
4461
  if disks is None:
4462
    return instance.disks
4463
  else:
4464
    if not set(disks).issubset(instance.disks):
4465
      raise errors.ProgrammerError("Can only act on disks belonging to the"
4466
                                   " target instance")
4467
    return disks
4468

    
4469

    
4470
def _ShutdownInstanceDisks(lu, instance, disks=None, ignore_primary=False):
4471
  """Shutdown block devices of an instance.
4472

4473
  This does the shutdown on all nodes of the instance.
4474

4475
  If the ignore_primary is false, errors on the primary node are
4476
  ignored.
4477

4478
  """
4479
  all_result = True
4480
  disks = _ExpandCheckDisks(instance, disks)
4481

    
4482
  for disk in disks:
4483
    for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
4484
      lu.cfg.SetDiskID(top_disk, node)
4485
      result = lu.rpc.call_blockdev_shutdown(node, top_disk)
4486
      msg = result.fail_msg
4487
      if msg:
4488
        lu.LogWarning("Could not shutdown block device %s on node %s: %s",
4489
                      disk.iv_name, node, msg)
4490
        if not ignore_primary or node != instance.primary_node:
4491
          all_result = False
4492
  return all_result
4493

    
4494

    
4495
def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
4496
  """Checks if a node has enough free memory.
4497

4498
  This function check if a given node has the needed amount of free
4499
  memory. In case the node has less memory or we cannot get the
4500
  information from the node, this function raise an OpPrereqError
4501
  exception.
4502

4503
  @type lu: C{LogicalUnit}
4504
  @param lu: a logical unit from which we get configuration data
4505
  @type node: C{str}
4506
  @param node: the node to check
4507
  @type reason: C{str}
4508
  @param reason: string to use in the error message
4509
  @type requested: C{int}
4510
  @param requested: the amount of memory in MiB to check for
4511
  @type hypervisor_name: C{str}
4512
  @param hypervisor_name: the hypervisor to ask for memory stats
4513
  @raise errors.OpPrereqError: if the node doesn't have enough memory, or
4514
      we cannot check the node
4515

4516
  """
4517
  nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
4518
  nodeinfo[node].Raise("Can't get data from node %s" % node,
4519
                       prereq=True, ecode=errors.ECODE_ENVIRON)
4520
  free_mem = nodeinfo[node].payload.get('memory_free', None)
4521
  if not isinstance(free_mem, int):
4522
    raise errors.OpPrereqError("Can't compute free memory on node %s, result"
4523
                               " was '%s'" % (node, free_mem),
4524
                               errors.ECODE_ENVIRON)
4525
  if requested > free_mem:
4526
    raise errors.OpPrereqError("Not enough memory on node %s for %s:"
4527
                               " needed %s MiB, available %s MiB" %
4528
                               (node, reason, requested, free_mem),
4529
                               errors.ECODE_NORES)
4530

    
4531

    
4532
def _CheckNodesFreeDisk(lu, nodenames, requested):
4533
  """Checks if nodes have enough free disk space in the default VG.
4534

4535
  This function check if all given nodes have the needed amount of
4536
  free disk. In case any node has less disk or we cannot get the
4537
  information from the node, this function raise an OpPrereqError
4538
  exception.
4539

4540
  @type lu: C{LogicalUnit}
4541
  @param lu: a logical unit from which we get configuration data
4542
  @type nodenames: C{list}
4543
  @param nodenames: the list of node names to check
4544
  @type requested: C{int}
4545
  @param requested: the amount of disk in MiB to check for
4546
  @raise errors.OpPrereqError: if the node doesn't have enough disk, or
4547
      we cannot check the node
4548

4549
  """
4550
  nodeinfo = lu.rpc.call_node_info(nodenames, lu.cfg.GetVGName(),
4551
                                   lu.cfg.GetHypervisorType())
4552
  for node in nodenames:
4553
    info = nodeinfo[node]
4554
    info.Raise("Cannot get current information from node %s" % node,
4555
               prereq=True, ecode=errors.ECODE_ENVIRON)
4556
    vg_free = info.payload.get("vg_free", None)
4557
    if not isinstance(vg_free, int):
4558
      raise errors.OpPrereqError("Can't compute free disk space on node %s,"
4559
                                 " result was '%s'" % (node, vg_free),
4560
                                 errors.ECODE_ENVIRON)
4561
    if requested > vg_free:
4562
      raise errors.OpPrereqError("Not enough disk space on target node %s:"
4563
                                 " required %d MiB, available %d MiB" %
4564
                                 (node, requested, vg_free),
4565
                                 errors.ECODE_NORES)
4566

    
4567

    
4568
class LUStartupInstance(LogicalUnit):
4569
  """Starts an instance.
4570

4571
  """
4572
  HPATH = "instance-start"
4573
  HTYPE = constants.HTYPE_INSTANCE
4574
  _OP_PARAMS = [
4575
    _PInstanceName,
4576
    _PForce,
4577
    ("hvparams", _EmptyDict, _TDict),
4578
    ("beparams", _EmptyDict, _TDict),
4579
    ]
4580
  REQ_BGL = False
4581

    
4582
  def CheckArguments(self):
4583
    # extra beparams
4584
    if self.op.beparams:
4585
      # fill the beparams dict
4586
      utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
4587

    
4588
  def ExpandNames(self):
4589
    self._ExpandAndLockInstance()
4590

    
4591
  def BuildHooksEnv(self):
4592
    """Build hooks env.
4593

4594
    This runs on master, primary and secondary nodes of the instance.
4595

4596
    """
4597
    env = {
4598
      "FORCE": self.op.force,
4599
      }
4600
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4601
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4602
    return env, nl, nl
4603

    
4604
  def CheckPrereq(self):
4605
    """Check prerequisites.
4606

4607
    This checks that the instance is in the cluster.
4608

4609
    """
4610
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4611
    assert self.instance is not None, \
4612
      "Cannot retrieve locked instance %s" % self.op.instance_name
4613

    
4614
    # extra hvparams
4615
    if self.op.hvparams:
4616
      # check hypervisor parameter syntax (locally)
4617
      cluster = self.cfg.GetClusterInfo()
4618
      utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
4619
      filled_hvp = cluster.FillHV(instance)
4620
      filled_hvp.update(self.op.hvparams)
4621
      hv_type = hypervisor.GetHypervisor(instance.hypervisor)
4622
      hv_type.CheckParameterSyntax(filled_hvp)
4623
      _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
4624

    
4625
    _CheckNodeOnline(self, instance.primary_node)
4626

    
4627
    bep = self.cfg.GetClusterInfo().FillBE(instance)
4628
    # check bridges existence
4629
    _CheckInstanceBridgesExist(self, instance)
4630

    
4631
    remote_info = self.rpc.call_instance_info(instance.primary_node,
4632
                                              instance.name,
4633
                                              instance.hypervisor)
4634
    remote_info.Raise("Error checking node %s" % instance.primary_node,
4635
                      prereq=True, ecode=errors.ECODE_ENVIRON)
4636
    if not remote_info.payload: # not running already
4637
      _CheckNodeFreeMemory(self, instance.primary_node,
4638
                           "starting instance %s" % instance.name,
4639
                           bep[constants.BE_MEMORY], instance.hypervisor)
4640

    
4641
  def Exec(self, feedback_fn):
4642
    """Start the instance.
4643

4644
    """
4645
    instance = self.instance
4646
    force = self.op.force
4647

    
4648
    self.cfg.MarkInstanceUp(instance.name)
4649

    
4650
    node_current = instance.primary_node
4651

    
4652
    _StartInstanceDisks(self, instance, force)
4653

    
4654
    result = self.rpc.call_instance_start(node_current, instance,
4655
                                          self.op.hvparams, self.op.beparams)
4656
    msg = result.fail_msg
4657
    if msg:
4658
      _ShutdownInstanceDisks(self, instance)
4659
      raise errors.OpExecError("Could not start instance: %s" % msg)
4660

    
4661

    
4662
class LURebootInstance(LogicalUnit):
4663
  """Reboot an instance.
4664

4665
  """
4666
  HPATH = "instance-reboot"
4667
  HTYPE = constants.HTYPE_INSTANCE
4668
  _OP_PARAMS = [
4669
    _PInstanceName,
4670
    ("ignore_secondaries", False, _TBool),
4671
    ("reboot_type", _NoDefault, _TElemOf(constants.REBOOT_TYPES)),
4672
    _PShutdownTimeout,
4673
    ]
4674
  REQ_BGL = False
4675

    
4676
  def ExpandNames(self):
4677
    self._ExpandAndLockInstance()
4678

    
4679
  def BuildHooksEnv(self):
4680
    """Build hooks env.
4681

4682
    This runs on master, primary and secondary nodes of the instance.
4683

4684
    """
4685
    env = {
4686
      "IGNORE_SECONDARIES": self.op.ignore_secondaries,
4687
      "REBOOT_TYPE": self.op.reboot_type,
4688
      "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
4689
      }
4690
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4691
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4692
    return env, nl, nl
4693

    
4694
  def CheckPrereq(self):
4695
    """Check prerequisites.
4696

4697
    This checks that the instance is in the cluster.
4698

4699
    """
4700
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4701
    assert self.instance is not None, \
4702
      "Cannot retrieve locked instance %s" % self.op.instance_name
4703

    
4704
    _CheckNodeOnline(self, instance.primary_node)
4705

    
4706
    # check bridges existence
4707
    _CheckInstanceBridgesExist(self, instance)
4708

    
4709
  def Exec(self, feedback_fn):
4710
    """Reboot the instance.
4711

4712
    """
4713
    instance = self.instance
4714
    ignore_secondaries = self.op.ignore_secondaries
4715
    reboot_type = self.op.reboot_type
4716

    
4717
    node_current = instance.primary_node
4718

    
4719
    if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
4720
                       constants.INSTANCE_REBOOT_HARD]:
4721
      for disk in instance.disks:
4722
        self.cfg.SetDiskID(disk, node_current)
4723
      result = self.rpc.call_instance_reboot(node_current, instance,
4724
                                             reboot_type,
4725
                                             self.op.shutdown_timeout)
4726
      result.Raise("Could not reboot instance")
4727
    else:
4728
      result = self.rpc.call_instance_shutdown(node_current, instance,
4729
                                               self.op.shutdown_timeout)
4730
      result.Raise("Could not shutdown instance for full reboot")
4731
      _ShutdownInstanceDisks(self, instance)
4732
      _StartInstanceDisks(self, instance, ignore_secondaries)
4733
      result = self.rpc.call_instance_start(node_current, instance, None, None)
4734
      msg = result.fail_msg
4735
      if msg:
4736
        _ShutdownInstanceDisks(self, instance)
4737
        raise errors.OpExecError("Could not start instance for"
4738
                                 " full reboot: %s" % msg)
4739

    
4740
    self.cfg.MarkInstanceUp(instance.name)
4741

    
4742

    
4743
class LUShutdownInstance(LogicalUnit):
4744
  """Shutdown an instance.
4745

4746
  """
4747
  HPATH = "instance-stop"
4748
  HTYPE = constants.HTYPE_INSTANCE
4749
  _OP_PARAMS = [
4750
    _PInstanceName,
4751
    ("timeout", constants.DEFAULT_SHUTDOWN_TIMEOUT, _TPositiveInt),
4752
    ]
4753
  REQ_BGL = False
4754

    
4755
  def ExpandNames(self):
4756
    self._ExpandAndLockInstance()
4757

    
4758
  def BuildHooksEnv(self):
4759
    """Build hooks env.
4760

4761
    This runs on master, primary and secondary nodes of the instance.
4762

4763
    """
4764
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4765
    env["TIMEOUT"] = self.op.timeout
4766
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4767
    return env, nl, nl
4768

    
4769
  def CheckPrereq(self):
4770
    """Check prerequisites.
4771

4772
    This checks that the instance is in the cluster.
4773

4774
    """
4775
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4776
    assert self.instance is not None, \
4777
      "Cannot retrieve locked instance %s" % self.op.instance_name
4778
    _CheckNodeOnline(self, self.instance.primary_node)
4779

    
4780
  def Exec(self, feedback_fn):
4781
    """Shutdown the instance.
4782

4783
    """
4784
    instance = self.instance
4785
    node_current = instance.primary_node
4786
    timeout = self.op.timeout
4787
    self.cfg.MarkInstanceDown(instance.name)
4788
    result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
4789
    msg = result.fail_msg
4790
    if msg:
4791
      self.proc.LogWarning("Could not shutdown instance: %s" % msg)
4792

    
4793
    _ShutdownInstanceDisks(self, instance)
4794

    
4795

    
4796
class LUReinstallInstance(LogicalUnit):
4797
  """Reinstall an instance.
4798

4799
  """
4800
  HPATH = "instance-reinstall"
4801
  HTYPE = constants.HTYPE_INSTANCE
4802
  _OP_PARAMS = [
4803
    _PInstanceName,
4804
    ("os_type", None, _TMaybeString),
4805
    ("force_variant", False, _TBool),
4806
    ]
4807
  REQ_BGL = False
4808

    
4809
  def ExpandNames(self):
4810
    self._ExpandAndLockInstance()
4811

    
4812
  def BuildHooksEnv(self):
4813
    """Build hooks env.
4814

4815
    This runs on master, primary and secondary nodes of the instance.
4816

4817
    """
4818
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4819
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4820
    return env, nl, nl
4821

    
4822
  def CheckPrereq(self):
4823
    """Check prerequisites.
4824

4825
    This checks that the instance is in the cluster and is not running.
4826

4827
    """
4828
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4829
    assert instance is not None, \
4830
      "Cannot retrieve locked instance %s" % self.op.instance_name
4831
    _CheckNodeOnline(self, instance.primary_node)
4832

    
4833
    if instance.disk_template == constants.DT_DISKLESS:
4834
      raise errors.OpPrereqError("Instance '%s' has no disks" %
4835
                                 self.op.instance_name,
4836
                                 errors.ECODE_INVAL)
4837
    _CheckInstanceDown(self, instance, "cannot reinstall")
4838

    
4839
    if self.op.os_type is not None:
4840
      # OS verification
4841
      pnode = _ExpandNodeName(self.cfg, instance.primary_node)
4842
      _CheckNodeHasOS(self, pnode, self.op.os_type, self.op.force_variant)
4843

    
4844
    self.instance = instance
4845

    
4846
  def Exec(self, feedback_fn):
4847
    """Reinstall the instance.
4848

4849
    """
4850
    inst = self.instance
4851

    
4852
    if self.op.os_type is not None:
4853
      feedback_fn("Changing OS to '%s'..." % self.op.os_type)
4854
      inst.os = self.op.os_type
4855
      self.cfg.Update(inst, feedback_fn)
4856

    
4857
    _StartInstanceDisks(self, inst, None)
4858
    try:
4859
      feedback_fn("Running the instance OS create scripts...")
4860
      # FIXME: pass debug option from opcode to backend
4861
      result = self.rpc.call_instance_os_add(inst.primary_node, inst, True,
4862
                                             self.op.debug_level)
4863
      result.Raise("Could not install OS for instance %s on node %s" %
4864
                   (inst.name, inst.primary_node))
4865
    finally:
4866
      _ShutdownInstanceDisks(self, inst)
4867

    
4868

    
4869
class LURecreateInstanceDisks(LogicalUnit):
4870
  """Recreate an instance's missing disks.
4871

4872
  """
4873
  HPATH = "instance-recreate-disks"
4874
  HTYPE = constants.HTYPE_INSTANCE
4875
  _OP_PARAMS = [
4876
    _PInstanceName,
4877
    ("disks", _EmptyList, _TListOf(_TPositiveInt)),
4878
    ]
4879
  REQ_BGL = False
4880

    
4881
  def ExpandNames(self):
4882
    self._ExpandAndLockInstance()
4883

    
4884
  def BuildHooksEnv(self):
4885
    """Build hooks env.
4886

4887
    This runs on master, primary and secondary nodes of the instance.
4888

4889
    """
4890
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4891
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4892
    return env, nl, nl
4893

    
4894
  def CheckPrereq(self):
4895
    """Check prerequisites.
4896

4897
    This checks that the instance is in the cluster and is not running.
4898

4899
    """
4900
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4901
    assert instance is not None, \
4902
      "Cannot retrieve locked instance %s" % self.op.instance_name
4903
    _CheckNodeOnline(self, instance.primary_node)
4904

    
4905
    if instance.disk_template == constants.DT_DISKLESS:
4906
      raise errors.OpPrereqError("Instance '%s' has no disks" %
4907
                                 self.op.instance_name, errors.ECODE_INVAL)
4908
    _CheckInstanceDown(self, instance, "cannot recreate disks")
4909

    
4910
    if not self.op.disks:
4911
      self.op.disks = range(len(instance.disks))
4912
    else:
4913
      for idx in self.op.disks:
4914
        if idx >= len(instance.disks):
4915
          raise errors.OpPrereqError("Invalid disk index passed '%s'" % idx,
4916
                                     errors.ECODE_INVAL)
4917

    
4918
    self.instance = instance
4919

    
4920
  def Exec(self, feedback_fn):
4921
    """Recreate the disks.
4922

4923
    """
4924
    to_skip = []
4925
    for idx, _ in enumerate(self.instance.disks):
4926
      if idx not in self.op.disks: # disk idx has not been passed in
4927
        to_skip.append(idx)
4928
        continue
4929

    
4930
    _CreateDisks(self, self.instance, to_skip=to_skip)
4931

    
4932

    
4933
class LURenameInstance(LogicalUnit):
4934
  """Rename an instance.
4935

4936
  """
4937
  HPATH = "instance-rename"
4938
  HTYPE = constants.HTYPE_INSTANCE
4939
  _OP_PARAMS = [
4940
    _PInstanceName,
4941
    ("new_name", _NoDefault, _TNonEmptyString),
4942
    ("ip_check", False, _TBool),
4943
    ("name_check", True, _TBool),
4944
    ]
4945

    
4946
  def CheckArguments(self):
4947
    """Check arguments.
4948

4949
    """
4950
    if self.op.ip_check and not self.op.name_check:
4951
      # TODO: make the ip check more flexible and not depend on the name check
4952
      raise errors.OpPrereqError("Cannot do ip check without a name check",
4953
                                 errors.ECODE_INVAL)
4954

    
4955
  def BuildHooksEnv(self):
4956
    """Build hooks env.
4957

4958
    This runs on master, primary and secondary nodes of the instance.
4959

4960
    """
4961
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4962
    env["INSTANCE_NEW_NAME"] = self.op.new_name
4963
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4964
    return env, nl, nl
4965

    
4966
  def CheckPrereq(self):
4967
    """Check prerequisites.
4968

4969
    This checks that the instance is in the cluster and is not running.
4970

4971
    """
4972
    self.op.instance_name = _ExpandInstanceName(self.cfg,
4973
                                                self.op.instance_name)
4974
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4975
    assert instance is not None
4976
    _CheckNodeOnline(self, instance.primary_node)
4977
    _CheckInstanceDown(self, instance, "cannot rename")
4978
    self.instance = instance
4979

    
4980
    new_name = self.op.new_name
4981
    if self.op.name_check:
4982
      hostinfo = netutils.HostInfo(netutils.HostInfo.NormalizeName(new_name))
4983
      new_name = self.op.new_name = hostinfo.name
4984
      if (self.op.ip_check and
4985
          netutils.TcpPing(hostinfo.ip, constants.DEFAULT_NODED_PORT)):
4986
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
4987
                                   (hostinfo.ip, new_name),
4988
                                   errors.ECODE_NOTUNIQUE)
4989

    
4990
    instance_list = self.cfg.GetInstanceList()
4991
    if new_name in instance_list:
4992
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
4993
                                 new_name, errors.ECODE_EXISTS)
4994

    
4995

    
4996
  def Exec(self, feedback_fn):
4997
    """Reinstall the instance.
4998

4999
    """
5000
    inst = self.instance
5001
    old_name = inst.name
5002

    
5003
    if inst.disk_template == constants.DT_FILE:
5004
      old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
5005

    
5006
    self.cfg.RenameInstance(inst.name, self.op.new_name)
5007
    # Change the instance lock. This is definitely safe while we hold the BGL
5008
    self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
5009
    self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
5010

    
5011
    # re-read the instance from the configuration after rename
5012
    inst = self.cfg.GetInstanceInfo(self.op.new_name)
5013

    
5014
    if inst.disk_template == constants.DT_FILE:
5015
      new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
5016
      result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
5017
                                                     old_file_storage_dir,
5018
                                                     new_file_storage_dir)
5019
      result.Raise("Could not rename on node %s directory '%s' to '%s'"
5020
                   " (but the instance has been renamed in Ganeti)" %
5021
                   (inst.primary_node, old_file_storage_dir,
5022
                    new_file_storage_dir))
5023

    
5024
    _StartInstanceDisks(self, inst, None)
5025
    try:
5026
      result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
5027
                                                 old_name, self.op.debug_level)
5028
      msg = result.fail_msg
5029
      if msg:
5030
        msg = ("Could not run OS rename script for instance %s on node %s"
5031
               " (but the instance has been renamed in Ganeti): %s" %
5032
               (inst.name, inst.primary_node, msg))
5033
        self.proc.LogWarning(msg)
5034
    finally:
5035
      _ShutdownInstanceDisks(self, inst)
5036

    
5037
    return inst.name
5038

    
5039

    
5040
class LURemoveInstance(LogicalUnit):
5041
  """Remove an instance.
5042

5043
  """
5044
  HPATH = "instance-remove"
5045
  HTYPE = constants.HTYPE_INSTANCE
5046
  _OP_PARAMS = [
5047
    _PInstanceName,
5048
    ("ignore_failures", False, _TBool),
5049
    _PShutdownTimeout,
5050
    ]
5051
  REQ_BGL = False
5052

    
5053
  def ExpandNames(self):
5054
    self._ExpandAndLockInstance()
5055
    self.needed_locks[locking.LEVEL_NODE] = []
5056
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5057

    
5058
  def DeclareLocks(self, level):
5059
    if level == locking.LEVEL_NODE:
5060
      self._LockInstancesNodes()
5061

    
5062
  def BuildHooksEnv(self):
5063
    """Build hooks env.
5064

5065
    This runs on master, primary and secondary nodes of the instance.
5066

5067
    """
5068
    env = _BuildInstanceHookEnvByObject(self, self.instance)
5069
    env["SHUTDOWN_TIMEOUT"] = self.op.shutdown_timeout
5070
    nl = [self.cfg.GetMasterNode()]
5071
    nl_post = list(self.instance.all_nodes) + nl
5072
    return env, nl, nl_post
5073

    
5074
  def CheckPrereq(self):
5075
    """Check prerequisites.
5076

5077
    This checks that the instance is in the cluster.
5078

5079
    """
5080
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5081
    assert self.instance is not None, \
5082
      "Cannot retrieve locked instance %s" % self.op.instance_name
5083

    
5084
  def Exec(self, feedback_fn):
5085
    """Remove the instance.
5086

5087
    """
5088
    instance = self.instance
5089
    logging.info("Shutting down instance %s on node %s",
5090
                 instance.name, instance.primary_node)
5091

    
5092
    result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
5093
                                             self.op.shutdown_timeout)
5094
    msg = result.fail_msg
5095
    if msg:
5096
      if self.op.ignore_failures:
5097
        feedback_fn("Warning: can't shutdown instance: %s" % msg)
5098
      else:
5099
        raise errors.OpExecError("Could not shutdown instance %s on"
5100
                                 " node %s: %s" %
5101
                                 (instance.name, instance.primary_node, msg))
5102

    
5103
    _RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures)
5104

    
5105

    
5106
def _RemoveInstance(lu, feedback_fn, instance, ignore_failures):
5107
  """Utility function to remove an instance.
5108

5109
  """
5110
  logging.info("Removing block devices for instance %s", instance.name)
5111

    
5112
  if not _RemoveDisks(lu, instance):
5113
    if not ignore_failures:
5114
      raise errors.OpExecError("Can't remove instance's disks")
5115
    feedback_fn("Warning: can't remove instance's disks")
5116

    
5117
  logging.info("Removing instance %s out of cluster config", instance.name)
5118

    
5119
  lu.cfg.RemoveInstance(instance.name)
5120

    
5121
  assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \
5122
    "Instance lock removal conflict"
5123

    
5124
  # Remove lock for the instance
5125
  lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name
5126

    
5127

    
5128
class LUQueryInstances(NoHooksLU):
5129
  """Logical unit for querying instances.
5130

5131
  """
5132
  # pylint: disable-msg=W0142
5133
  _OP_PARAMS = [
5134
    ("output_fields", _NoDefault, _TListOf(_TNonEmptyString)),
5135
    ("names", _EmptyList, _TListOf(_TNonEmptyString)),
5136
    ("use_locking", False, _TBool),
5137
    ]
5138
  REQ_BGL = False
5139
  _SIMPLE_FIELDS = ["name", "os", "network_port", "hypervisor",
5140
                    "serial_no", "ctime", "mtime", "uuid"]
5141
  _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
5142
                                    "admin_state",
5143
                                    "disk_template", "ip", "mac", "bridge",
5144
                                    "nic_mode", "nic_link",
5145
                                    "sda_size", "sdb_size", "vcpus", "tags",
5146
                                    "network_port", "beparams",
5147
                                    r"(disk)\.(size)/([0-9]+)",
5148
                                    r"(disk)\.(sizes)", "disk_usage",
5149
                                    r"(nic)\.(mac|ip|mode|link)/([0-9]+)",
5150
                                    r"(nic)\.(bridge)/([0-9]+)",
5151
                                    r"(nic)\.(macs|ips|modes|links|bridges)",
5152
                                    r"(disk|nic)\.(count)",
5153
                                    "hvparams",
5154
                                    ] + _SIMPLE_FIELDS +
5155
                                  ["hv/%s" % name
5156
                                   for name in constants.HVS_PARAMETERS
5157
                                   if name not in constants.HVC_GLOBALS] +
5158
                                  ["be/%s" % name
5159
                                   for name in constants.BES_PARAMETERS])
5160
  _FIELDS_DYNAMIC = utils.FieldSet("oper_state",
5161
                                   "oper_ram",
5162
                                   "oper_vcpus",
5163
                                   "status")
5164

    
5165

    
5166
  def CheckArguments(self):
5167
    _CheckOutputFields(static=self._FIELDS_STATIC,
5168
                       dynamic=self._FIELDS_DYNAMIC,
5169
                       selected=self.op.output_fields)
5170

    
5171
  def ExpandNames(self):
5172
    self.needed_locks = {}
5173
    self.share_locks[locking.LEVEL_INSTANCE] = 1
5174
    self.share_locks[locking.LEVEL_NODE] = 1
5175

    
5176
    if self.op.names:
5177
      self.wanted = _GetWantedInstances(self, self.op.names)
5178
    else:
5179
      self.wanted = locking.ALL_SET
5180

    
5181
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
5182
    self.do_locking = self.do_node_query and self.op.use_locking
5183
    if self.do_locking:
5184
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
5185
      self.needed_locks[locking.LEVEL_NODE] = []
5186
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5187

    
5188
  def DeclareLocks(self, level):
5189
    if level == locking.LEVEL_NODE and self.do_locking:
5190
      self._LockInstancesNodes()
5191

    
5192
  def Exec(self, feedback_fn):
5193
    """Computes the list of nodes and their attributes.
5194

5195
    """
5196
    # pylint: disable-msg=R0912
5197
    # way too many branches here
5198
    all_info = self.cfg.GetAllInstancesInfo()
5199
    if self.wanted == locking.ALL_SET:
5200
      # caller didn't specify instance names, so ordering is not important
5201
      if self.do_locking:
5202
        instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
5203
      else:
5204
        instance_names = all_info.keys()
5205
      instance_names = utils.NiceSort(instance_names)
5206
    else:
5207
      # caller did specify names, so we must keep the ordering
5208
      if self.do_locking:
5209
        tgt_set = self.acquired_locks[locking.LEVEL_INSTANCE]
5210
      else:
5211
        tgt_set = all_info.keys()
5212
      missing = set(self.wanted).difference(tgt_set)
5213
      if missing:
5214
        raise errors.OpExecError("Some instances were removed before"
5215
                                 " retrieving their data: %s" % missing)
5216
      instance_names = self.wanted
5217

    
5218
    instance_list = [all_info[iname] for iname in instance_names]
5219

    
5220
    # begin data gathering
5221

    
5222
    nodes = frozenset([inst.primary_node for inst in instance_list])
5223
    hv_list = list(set([inst.hypervisor for inst in instance_list]))
5224

    
5225
    bad_nodes = []
5226
    off_nodes = []
5227
    if self.do_node_query:
5228
      live_data = {}
5229
      node_data = self.rpc.call_all_instances_info(nodes, hv_list)
5230
      for name in nodes:
5231
        result = node_data[name]
5232
        if result.offline:
5233
          # offline nodes will be in both lists
5234
          off_nodes.append(name)
5235
        if result.fail_msg:
5236
          bad_nodes.append(name)
5237
        else:
5238
          if result.payload:
5239
            live_data.update(result.payload)
5240
          # else no instance is alive
5241
    else:
5242
      live_data = dict([(name, {}) for name in instance_names])
5243

    
5244
    # end data gathering
5245

    
5246
    HVPREFIX = "hv/"
5247
    BEPREFIX = "be/"
5248
    output = []
5249
    cluster = self.cfg.GetClusterInfo()
5250
    for instance in instance_list:
5251
      iout = []
5252
      i_hv = cluster.FillHV(instance, skip_globals=True)
5253
      i_be = cluster.FillBE(instance)
5254
      i_nicp = [cluster.SimpleFillNIC(nic.nicparams) for nic in instance.nics]
5255
      for field in self.op.output_fields:
5256
        st_match = self._FIELDS_STATIC.Matches(field)
5257
        if field in self._SIMPLE_FIELDS:
5258
          val = getattr(instance, field)
5259
        elif field == "pnode":
5260
          val = instance.primary_node
5261
        elif field == "snodes":
5262
          val = list(instance.secondary_nodes)
5263
        elif field == "admin_state":
5264
          val = instance.admin_up
5265
        elif field == "oper_state":
5266
          if instance.primary_node in bad_nodes:
5267
            val = None
5268
          else:
5269
            val = bool(live_data.get(instance.name))
5270
        elif field == "status":
5271
          if instance.primary_node in off_nodes:
5272
            val = "ERROR_nodeoffline"
5273
          elif instance.primary_node in bad_nodes:
5274
            val = "ERROR_nodedown"
5275
          else:
5276
            running = bool(live_data.get(instance.name))
5277
            if running:
5278
              if instance.admin_up:
5279
                val = "running"
5280
              else:
5281
                val = "ERROR_up"
5282
            else:
5283
              if instance.admin_up:
5284
                val = "ERROR_down"
5285
              else:
5286
                val = "ADMIN_down"
5287
        elif field == "oper_ram":
5288
          if instance.primary_node in bad_nodes:
5289
            val = None
5290
          elif instance.name in live_data:
5291
            val = live_data[instance.name].get("memory", "?")
5292
          else:
5293
            val = "-"
5294
        elif field == "oper_vcpus":
5295
          if instance.primary_node in bad_nodes:
5296
            val = None
5297
          elif instance.name in live_data:
5298
            val = live_data[instance.name].get("vcpus", "?")
5299
          else:
5300
            val = "-"
5301
        elif field == "vcpus":
5302
          val = i_be[constants.BE_VCPUS]
5303
        elif field == "disk_template":
5304
          val = instance.disk_template
5305
        elif field == "ip":
5306
          if instance.nics:
5307
            val = instance.nics[0].ip
5308
          else:
5309
            val = None
5310
        elif field == "nic_mode":
5311
          if instance.nics:
5312
            val = i_nicp[0][constants.NIC_MODE]
5313
          else:
5314
            val = None
5315
        elif field == "nic_link":
5316
          if instance.nics:
5317
            val = i_nicp[0][constants.NIC_LINK]
5318
          else:
5319
            val = None
5320
        elif field == "bridge":
5321
          if (instance.nics and
5322
              i_nicp[0][constants.NIC_MODE] == constants.NIC_MODE_BRIDGED):
5323
            val = i_nicp[0][constants.NIC_LINK]
5324
          else:
5325
            val = None
5326
        elif field == "mac":
5327
          if instance.nics:
5328
            val = instance.nics[0].mac
5329
          else:
5330
            val = None
5331
        elif field == "sda_size" or field == "sdb_size":
5332
          idx = ord(field[2]) - ord('a')
5333
          try:
5334
            val = instance.FindDisk(idx).size
5335
          except errors.OpPrereqError:
5336
            val = None
5337
        elif field == "disk_usage": # total disk usage per node
5338
          disk_sizes = [{'size': disk.size} for disk in instance.disks]
5339
          val = _ComputeDiskSize(instance.disk_template, disk_sizes)
5340
        elif field == "tags":
5341
          val = list(instance.GetTags())
5342
        elif field == "hvparams":
5343
          val = i_hv
5344
        elif (field.startswith(HVPREFIX) and
5345
              field[len(HVPREFIX):] in constants.HVS_PARAMETERS and
5346
              field[len(HVPREFIX):] not in constants.HVC_GLOBALS):
5347
          val = i_hv.get(field[len(HVPREFIX):], None)
5348
        elif field == "beparams":
5349
          val = i_be
5350
        elif (field.startswith(BEPREFIX) and
5351
              field[len(BEPREFIX):] in constants.BES_PARAMETERS):
5352
          val = i_be.get(field[len(BEPREFIX):], None)
5353
        elif st_match and st_match.groups():
5354
          # matches a variable list
5355
          st_groups = st_match.groups()
5356
          if st_groups and st_groups[0] == "disk":
5357
            if st_groups[1] == "count":
5358
              val = len(instance.disks)
5359
            elif st_groups[1] == "sizes":
5360
              val = [disk.size for disk in instance.disks]
5361
            elif st_groups[1] == "size":
5362
              try:
5363
                val = instance.FindDisk(st_groups[2]).size
5364
              except errors.OpPrereqError:
5365
                val = None
5366
            else:
5367
              assert False, "Unhandled disk parameter"
5368
          elif st_groups[0] == "nic":
5369
            if st_groups[1] == "count":
5370
              val = len(instance.nics)
5371
            elif st_groups[1] == "macs":
5372
              val = [nic.mac for nic in instance.nics]
5373
            elif st_groups[1] == "ips":
5374
              val = [nic.ip for nic in instance.nics]
5375
            elif st_groups[1] == "modes":
5376
              val = [nicp[constants.NIC_MODE] for nicp in i_nicp]
5377
            elif st_groups[1] == "links":
5378
              val = [nicp[constants.NIC_LINK] for nicp in i_nicp]
5379
            elif st_groups[1] == "bridges":
5380
              val = []
5381
              for nicp in i_nicp:
5382
                if nicp[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
5383
                  val.append(nicp[constants.NIC_LINK])
5384
                else:
5385
                  val.append(None)
5386
            else:
5387
              # index-based item
5388
              nic_idx = int(st_groups[2])
5389
              if nic_idx >= len(instance.nics):
5390
                val = None
5391
              else:
5392
                if st_groups[1] == "mac":
5393
                  val = instance.nics[nic_idx].mac
5394
                elif st_groups[1] == "ip":
5395
                  val = instance.nics[nic_idx].ip
5396
                elif st_groups[1] == "mode":
5397
                  val = i_nicp[nic_idx][constants.NIC_MODE]
5398
                elif st_groups[1] == "link":
5399
                  val = i_nicp[nic_idx][constants.NIC_LINK]
5400
                elif st_groups[1] == "bridge":
5401
                  nic_mode = i_nicp[nic_idx][constants.NIC_MODE]
5402
                  if nic_mode == constants.NIC_MODE_BRIDGED:
5403
                    val = i_nicp[nic_idx][constants.NIC_LINK]
5404
                  else:
5405
                    val = None
5406
                else:
5407
                  assert False, "Unhandled NIC parameter"
5408
          else:
5409
            assert False, ("Declared but unhandled variable parameter '%s'" %
5410
                           field)
5411
        else:
5412
          assert False, "Declared but unhandled parameter '%s'" % field
5413
        iout.append(val)
5414
      output.append(iout)
5415

    
5416
    return output
5417

    
5418

    
5419
class LUFailoverInstance(LogicalUnit):
5420
  """Failover an instance.
5421

5422
  """
5423
  HPATH = "instance-failover"
5424
  HTYPE = constants.HTYPE_INSTANCE
5425
  _OP_PARAMS = [
5426
    _PInstanceName,
5427
    ("ignore_consistency", False, _TBool),
5428
    _PShutdownTimeout,
5429
    ]
5430
  REQ_BGL = False
5431

    
5432
  def ExpandNames(self):
5433
    self._ExpandAndLockInstance()
5434
    self.needed_locks[locking.LEVEL_NODE] = []
5435
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5436

    
5437
  def DeclareLocks(self, level):
5438
    if level == locking.LEVEL_NODE:
5439
      self._LockInstancesNodes()
5440

    
5441
  def BuildHooksEnv(self):
5442
    """Build hooks env.
5443

5444
    This runs on master, primary and secondary nodes of the instance.
5445

5446
    """
5447
    instance = self.instance
5448
    source_node = instance.primary_node
5449
    target_node = instance.secondary_nodes[0]
5450
    env = {
5451
      "IGNORE_CONSISTENCY": self.op.ignore_consistency,
5452
      "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5453
      "OLD_PRIMARY": source_node,
5454
      "OLD_SECONDARY": target_node,
5455
      "NEW_PRIMARY": target_node,
5456
      "NEW_SECONDARY": source_node,
5457
      }
5458
    env.update(_BuildInstanceHookEnvByObject(self, instance))
5459
    nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5460
    nl_post = list(nl)
5461
    nl_post.append(source_node)
5462
    return env, nl, nl_post
5463

    
5464
  def CheckPrereq(self):
5465
    """Check prerequisites.
5466

5467
    This checks that the instance is in the cluster.
5468

5469
    """
5470
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5471
    assert self.instance is not None, \
5472
      "Cannot retrieve locked instance %s" % self.op.instance_name
5473

    
5474
    bep = self.cfg.GetClusterInfo().FillBE(instance)
5475
    if instance.disk_template not in constants.DTS_NET_MIRROR:
5476
      raise errors.OpPrereqError("Instance's disk layout is not"
5477
                                 " network mirrored, cannot failover.",
5478
                                 errors.ECODE_STATE)
5479

    
5480
    secondary_nodes = instance.secondary_nodes
5481
    if not secondary_nodes:
5482
      raise errors.ProgrammerError("no secondary node but using "
5483
                                   "a mirrored disk template")
5484

    
5485
    target_node = secondary_nodes[0]
5486
    _CheckNodeOnline(self, target_node)
5487
    _CheckNodeNotDrained(self, target_node)
5488
    if instance.admin_up:
5489
      # check memory requirements on the secondary node
5490
      _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5491
                           instance.name, bep[constants.BE_MEMORY],
5492
                           instance.hypervisor)
5493
    else:
5494
      self.LogInfo("Not checking memory on the secondary node as"
5495
                   " instance will not be started")
5496

    
5497
    # check bridge existance
5498
    _CheckInstanceBridgesExist(self, instance, node=target_node)
5499

    
5500
  def Exec(self, feedback_fn):
5501
    """Failover an instance.
5502

5503
    The failover is done by shutting it down on its present node and
5504
    starting it on the secondary.
5505

5506
    """
5507
    instance = self.instance
5508

    
5509
    source_node = instance.primary_node
5510
    target_node = instance.secondary_nodes[0]
5511

    
5512
    if instance.admin_up:
5513
      feedback_fn("* checking disk consistency between source and target")
5514
      for dev in instance.disks:
5515
        # for drbd, these are drbd over lvm
5516
        if not _CheckDiskConsistency(self, dev, target_node, False):
5517
          if not self.op.ignore_consistency:
5518
            raise errors.OpExecError("Disk %s is degraded on target node,"
5519
                                     " aborting failover." % dev.iv_name)
5520
    else:
5521
      feedback_fn("* not checking disk consistency as instance is not running")
5522

    
5523
    feedback_fn("* shutting down instance on source node")
5524
    logging.info("Shutting down instance %s on node %s",
5525
                 instance.name, source_node)
5526

    
5527
    result = self.rpc.call_instance_shutdown(source_node, instance,
5528
                                             self.op.shutdown_timeout)
5529
    msg = result.fail_msg
5530
    if msg:
5531
      if self.op.ignore_consistency:
5532
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
5533
                             " Proceeding anyway. Please make sure node"
5534
                             " %s is down. Error details: %s",
5535
                             instance.name, source_node, source_node, msg)
5536
      else:
5537
        raise errors.OpExecError("Could not shutdown instance %s on"
5538
                                 " node %s: %s" %
5539
                                 (instance.name, source_node, msg))
5540

    
5541
    feedback_fn("* deactivating the instance's disks on source node")
5542
    if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
5543
      raise errors.OpExecError("Can't shut down the instance's disks.")
5544

    
5545
    instance.primary_node = target_node
5546
    # distribute new instance config to the other nodes
5547
    self.cfg.Update(instance, feedback_fn)
5548

    
5549
    # Only start the instance if it's marked as up
5550
    if instance.admin_up:
5551
      feedback_fn("* activating the instance's disks on target node")
5552
      logging.info("Starting instance %s on node %s",
5553
                   instance.name, target_node)
5554

    
5555
      disks_ok, _ = _AssembleInstanceDisks(self, instance,
5556
                                           ignore_secondaries=True)
5557
      if not disks_ok:
5558
        _ShutdownInstanceDisks(self, instance)
5559
        raise errors.OpExecError("Can't activate the instance's disks")
5560

    
5561
      feedback_fn("* starting the instance on the target node")
5562
      result = self.rpc.call_instance_start(target_node, instance, None, None)
5563
      msg = result.fail_msg
5564
      if msg:
5565
        _ShutdownInstanceDisks(self, instance)
5566
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
5567
                                 (instance.name, target_node, msg))
5568

    
5569

    
5570
class LUMigrateInstance(LogicalUnit):
5571
  """Migrate an instance.
5572

5573
  This is migration without shutting down, compared to the failover,
5574
  which is done with shutdown.
5575

5576
  """
5577
  HPATH = "instance-migrate"
5578
  HTYPE = constants.HTYPE_INSTANCE
5579
  _OP_PARAMS = [
5580
    _PInstanceName,
5581
    _PMigrationMode,
5582
    _PMigrationLive,
5583
    ("cleanup", False, _TBool),
5584
    ]
5585

    
5586
  REQ_BGL = False
5587

    
5588
  def ExpandNames(self):
5589
    self._ExpandAndLockInstance()
5590

    
5591
    self.needed_locks[locking.LEVEL_NODE] = []
5592
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5593

    
5594
    self._migrater = TLMigrateInstance(self, self.op.instance_name,
5595
                                       self.op.cleanup)
5596
    self.tasklets = [self._migrater]
5597

    
5598
  def DeclareLocks(self, level):
5599
    if level == locking.LEVEL_NODE:
5600
      self._LockInstancesNodes()
5601

    
5602
  def BuildHooksEnv(self):
5603
    """Build hooks env.
5604

5605
    This runs on master, primary and secondary nodes of the instance.
5606

5607
    """
5608
    instance = self._migrater.instance
5609
    source_node = instance.primary_node
5610
    target_node = instance.secondary_nodes[0]
5611
    env = _BuildInstanceHookEnvByObject(self, instance)
5612
    env["MIGRATE_LIVE"] = self._migrater.live
5613
    env["MIGRATE_CLEANUP"] = self.op.cleanup
5614
    env.update({
5615
        "OLD_PRIMARY": source_node,
5616
        "OLD_SECONDARY": target_node,
5617
        "NEW_PRIMARY": target_node,
5618
        "NEW_SECONDARY": source_node,
5619
        })
5620
    nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5621
    nl_post = list(nl)
5622
    nl_post.append(source_node)
5623
    return env, nl, nl_post
5624

    
5625

    
5626
class LUMoveInstance(LogicalUnit):
5627
  """Move an instance by data-copying.
5628

5629
  """
5630
  HPATH = "instance-move"
5631
  HTYPE = constants.HTYPE_INSTANCE
5632
  _OP_PARAMS = [
5633
    _PInstanceName,
5634
    ("target_node", _NoDefault, _TNonEmptyString),
5635
    _PShutdownTimeout,
5636
    ]
5637
  REQ_BGL = False
5638

    
5639
  def ExpandNames(self):
5640
    self._ExpandAndLockInstance()
5641
    target_node = _ExpandNodeName(self.cfg, self.op.target_node)
5642
    self.op.target_node = target_node
5643
    self.needed_locks[locking.LEVEL_NODE] = [target_node]
5644
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5645

    
5646
  def DeclareLocks(self, level):
5647
    if level == locking.LEVEL_NODE:
5648
      self._LockInstancesNodes(primary_only=True)
5649

    
5650
  def BuildHooksEnv(self):
5651
    """Build hooks env.
5652

5653
    This runs on master, primary and secondary nodes of the instance.
5654

5655
    """
5656
    env = {
5657
      "TARGET_NODE": self.op.target_node,
5658
      "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5659
      }
5660
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5661
    nl = [self.cfg.GetMasterNode()] + [self.instance.primary_node,
5662
                                       self.op.target_node]
5663
    return env, nl, nl
5664

    
5665
  def CheckPrereq(self):
5666
    """Check prerequisites.
5667

5668
    This checks that the instance is in the cluster.
5669

5670
    """
5671
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5672
    assert self.instance is not None, \
5673
      "Cannot retrieve locked instance %s" % self.op.instance_name
5674

    
5675
    node = self.cfg.GetNodeInfo(self.op.target_node)
5676
    assert node is not None, \
5677
      "Cannot retrieve locked node %s" % self.op.target_node
5678

    
5679
    self.target_node = target_node = node.name
5680

    
5681
    if target_node == instance.primary_node:
5682
      raise errors.OpPrereqError("Instance %s is already on the node %s" %
5683
                                 (instance.name, target_node),
5684
                                 errors.ECODE_STATE)
5685

    
5686
    bep = self.cfg.GetClusterInfo().FillBE(instance)
5687

    
5688
    for idx, dsk in enumerate(instance.disks):
5689
      if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
5690
        raise errors.OpPrereqError("Instance disk %d has a complex layout,"
5691
                                   " cannot copy" % idx, errors.ECODE_STATE)
5692

    
5693
    _CheckNodeOnline(self, target_node)
5694
    _CheckNodeNotDrained(self, target_node)
5695

    
5696
    if instance.admin_up:
5697
      # check memory requirements on the secondary node
5698
      _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5699
                           instance.name, bep[constants.BE_MEMORY],
5700
                           instance.hypervisor)
5701
    else:
5702
      self.LogInfo("Not checking memory on the secondary node as"
5703
                   " instance will not be started")
5704

    
5705
    # check bridge existance
5706
    _CheckInstanceBridgesExist(self, instance, node=target_node)
5707

    
5708
  def Exec(self, feedback_fn):
5709
    """Move an instance.
5710

5711
    The move is done by shutting it down on its present node, copying
5712
    the data over (slow) and starting it on the new node.
5713

5714
    """
5715
    instance = self.instance
5716

    
5717
    source_node = instance.primary_node
5718
    target_node = self.target_node
5719

    
5720
    self.LogInfo("Shutting down instance %s on source node %s",
5721
                 instance.name, source_node)
5722

    
5723
    result = self.rpc.call_instance_shutdown(source_node, instance,
5724
                                             self.op.shutdown_timeout)
5725
    msg = result.fail_msg
5726
    if msg:
5727
      if self.op.ignore_consistency:
5728
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
5729
                             " Proceeding anyway. Please make sure node"
5730
                             " %s is down. Error details: %s",
5731
                             instance.name, source_node, source_node, msg)
5732
      else:
5733
        raise errors.OpExecError("Could not shutdown instance %s on"
5734
                                 " node %s: %s" %
5735
                                 (instance.name, source_node, msg))
5736

    
5737
    # create the target disks
5738
    try:
5739
      _CreateDisks(self, instance, target_node=target_node)
5740
    except errors.OpExecError:
5741
      self.LogWarning("Device creation failed, reverting...")
5742
      try:
5743
        _RemoveDisks(self, instance, target_node=target_node)
5744
      finally:
5745
        self.cfg.ReleaseDRBDMinors(instance.name)
5746
        raise
5747

    
5748
    cluster_name = self.cfg.GetClusterInfo().cluster_name
5749

    
5750
    errs = []
5751
    # activate, get path, copy the data over
5752
    for idx, disk in enumerate(instance.disks):
5753
      self.LogInfo("Copying data for disk %d", idx)
5754
      result = self.rpc.call_blockdev_assemble(target_node, disk,
5755
                                               instance.name, True)
5756
      if result.fail_msg:
5757
        self.LogWarning("Can't assemble newly created disk %d: %s",
5758
                        idx, result.fail_msg)
5759
        errs.append(result.fail_msg)
5760
        break
5761
      dev_path = result.payload
5762
      result = self.rpc.call_blockdev_export(source_node, disk,
5763
                                             target_node, dev_path,
5764
                                             cluster_name)
5765
      if result.fail_msg:
5766
        self.LogWarning("Can't copy data over for disk %d: %s",
5767
                        idx, result.fail_msg)
5768
        errs.append(result.fail_msg)
5769
        break
5770

    
5771
    if errs:
5772
      self.LogWarning("Some disks failed to copy, aborting")
5773
      try:
5774
        _RemoveDisks(self, instance, target_node=target_node)
5775
      finally:
5776
        self.cfg.ReleaseDRBDMinors(instance.name)
5777
        raise errors.OpExecError("Errors during disk copy: %s" %
5778
                                 (",".join(errs),))
5779

    
5780
    instance.primary_node = target_node
5781
    self.cfg.Update(instance, feedback_fn)
5782

    
5783
    self.LogInfo("Removing the disks on the original node")
5784
    _RemoveDisks(self, instance, target_node=source_node)
5785

    
5786
    # Only start the instance if it's marked as up
5787
    if instance.admin_up:
5788
      self.LogInfo("Starting instance %s on node %s",
5789
                   instance.name, target_node)
5790

    
5791
      disks_ok, _ = _AssembleInstanceDisks(self, instance,
5792
                                           ignore_secondaries=True)
5793
      if not disks_ok:
5794
        _ShutdownInstanceDisks(self, instance)
5795
        raise errors.OpExecError("Can't activate the instance's disks")
5796

    
5797
      result = self.rpc.call_instance_start(target_node, instance, None, None)
5798
      msg = result.fail_msg
5799
      if msg:
5800
        _ShutdownInstanceDisks(self, instance)
5801
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
5802
                                 (instance.name, target_node, msg))
5803

    
5804

    
5805
class LUMigrateNode(LogicalUnit):
5806
  """Migrate all instances from a node.
5807

5808
  """
5809
  HPATH = "node-migrate"
5810
  HTYPE = constants.HTYPE_NODE
5811
  _OP_PARAMS = [
5812
    _PNodeName,
5813
    _PMigrationMode,
5814
    _PMigrationLive,
5815
    ]
5816
  REQ_BGL = False
5817

    
5818
  def ExpandNames(self):
5819
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
5820

    
5821
    self.needed_locks = {
5822
      locking.LEVEL_NODE: [self.op.node_name],
5823
      }
5824

    
5825
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5826

    
5827
    # Create tasklets for migrating instances for all instances on this node
5828
    names = []
5829
    tasklets = []
5830

    
5831
    for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name):
5832
      logging.debug("Migrating instance %s", inst.name)
5833
      names.append(inst.name)
5834

    
5835
      tasklets.append(TLMigrateInstance(self, inst.name, False))
5836

    
5837
    self.tasklets = tasklets
5838

    
5839
    # Declare instance locks
5840
    self.needed_locks[locking.LEVEL_INSTANCE] = names
5841

    
5842
  def DeclareLocks(self, level):
5843
    if level == locking.LEVEL_NODE:
5844
      self._LockInstancesNodes()
5845

    
5846
  def BuildHooksEnv(self):
5847
    """Build hooks env.
5848

5849
    This runs on the master, the primary and all the secondaries.
5850

5851
    """
5852
    env = {
5853
      "NODE_NAME": self.op.node_name,
5854
      }
5855

    
5856
    nl = [self.cfg.GetMasterNode()]
5857

    
5858
    return (env, nl, nl)
5859

    
5860

    
5861
class TLMigrateInstance(Tasklet):
5862
  """Tasklet class for instance migration.
5863

5864
  @type live: boolean
5865
  @ivar live: whether the migration will be done live or non-live;
5866
      this variable is initalized only after CheckPrereq has run
5867

5868
  """
5869
  def __init__(self, lu, instance_name, cleanup):
5870
    """Initializes this class.
5871

5872
    """
5873
    Tasklet.__init__(self, lu)
5874

    
5875
    # Parameters
5876
    self.instance_name = instance_name
5877
    self.cleanup = cleanup
5878
    self.live = False # will be overridden later
5879

    
5880
  def CheckPrereq(self):
5881
    """Check prerequisites.
5882

5883
    This checks that the instance is in the cluster.
5884

5885
    """
5886
    instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name)
5887
    instance = self.cfg.GetInstanceInfo(instance_name)
5888
    assert instance is not None
5889

    
5890
    if instance.disk_template != constants.DT_DRBD8:
5891
      raise errors.OpPrereqError("Instance's disk layout is not"
5892
                                 " drbd8, cannot migrate.", errors.ECODE_STATE)
5893

    
5894
    secondary_nodes = instance.secondary_nodes
5895
    if not secondary_nodes:
5896
      raise errors.ConfigurationError("No secondary node but using"
5897
                                      " drbd8 disk template")
5898

    
5899
    i_be = self.cfg.GetClusterInfo().FillBE(instance)
5900

    
5901
    target_node = secondary_nodes[0]
5902
    # check memory requirements on the secondary node
5903
    _CheckNodeFreeMemory(self.lu, target_node, "migrating instance %s" %
5904
                         instance.name, i_be[constants.BE_MEMORY],
5905
                         instance.hypervisor)
5906

    
5907
    # check bridge existance
5908
    _CheckInstanceBridgesExist(self.lu, instance, node=target_node)
5909

    
5910
    if not self.cleanup:
5911
      _CheckNodeNotDrained(self.lu, target_node)
5912
      result = self.rpc.call_instance_migratable(instance.primary_node,
5913
                                                 instance)
5914
      result.Raise("Can't migrate, please use failover",
5915
                   prereq=True, ecode=errors.ECODE_STATE)
5916

    
5917
    self.instance = instance
5918

    
5919
    if self.lu.op.live is not None and self.lu.op.mode is not None:
5920
      raise errors.OpPrereqError("Only one of the 'live' and 'mode'"
5921
                                 " parameters are accepted",
5922
                                 errors.ECODE_INVAL)
5923
    if self.lu.op.live is not None:
5924
      if self.lu.op.live:
5925
        self.lu.op.mode = constants.HT_MIGRATION_LIVE
5926
      else:
5927
        self.lu.op.mode = constants.HT_MIGRATION_NONLIVE
5928
      # reset the 'live' parameter to None so that repeated
5929
      # invocations of CheckPrereq do not raise an exception
5930
      self.lu.op.live = None
5931
    elif self.lu.op.mode is None:
5932
      # read the default value from the hypervisor
5933
      i_hv = self.cfg.GetClusterInfo().FillHV(instance, skip_globals=False)
5934
      self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE]
5935

    
5936
    self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE
5937

    
5938
  def _WaitUntilSync(self):
5939
    """Poll with custom rpc for disk sync.
5940

5941
    This uses our own step-based rpc call.
5942

5943
    """
5944
    self.feedback_fn("* wait until resync is done")
5945
    all_done = False
5946
    while not all_done:
5947
      all_done = True
5948
      result = self.rpc.call_drbd_wait_sync(self.all_nodes,
5949
                                            self.nodes_ip,
5950
                                            self.instance.disks)
5951
      min_percent = 100
5952
      for node, nres in result.items():
5953
        nres.Raise("Cannot resync disks on node %s" % node)
5954
        node_done, node_percent = nres.payload
5955
        all_done = all_done and node_done
5956
        if node_percent is not None:
5957
          min_percent = min(min_percent, node_percent)
5958
      if not all_done:
5959
        if min_percent < 100:
5960
          self.feedback_fn("   - progress: %.1f%%" % min_percent)
5961
        time.sleep(2)
5962

    
5963
  def _EnsureSecondary(self, node):
5964
    """Demote a node to secondary.
5965

5966
    """
5967
    self.feedback_fn("* switching node %s to secondary mode" % node)
5968

    
5969
    for dev in self.instance.disks:
5970
      self.cfg.SetDiskID(dev, node)
5971

    
5972
    result = self.rpc.call_blockdev_close(node, self.instance.name,
5973
                                          self.instance.disks)
5974
    result.Raise("Cannot change disk to secondary on node %s" % node)
5975

    
5976
  def _GoStandalone(self):
5977
    """Disconnect from the network.
5978

5979
    """
5980
    self.feedback_fn("* changing into standalone mode")
5981
    result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
5982
                                               self.instance.disks)
5983
    for node, nres in result.items():
5984
      nres.Raise("Cannot disconnect disks node %s" % node)
5985

    
5986
  def _GoReconnect(self, multimaster):
5987
    """Reconnect to the network.
5988

5989
    """
5990
    if multimaster:
5991
      msg = "dual-master"
5992
    else:
5993
      msg = "single-master"
5994
    self.feedback_fn("* changing disks into %s mode" % msg)
5995
    result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
5996
                                           self.instance.disks,
5997
                                           self.instance.name, multimaster)
5998
    for node, nres in result.items():
5999
      nres.Raise("Cannot change disks config on node %s" % node)
6000

    
6001
  def _ExecCleanup(self):
6002
    """Try to cleanup after a failed migration.
6003

6004
    The cleanup is done by:
6005
      - check that the instance is running only on one node
6006
        (and update the config if needed)
6007
      - change disks on its secondary node to secondary
6008
      - wait until disks are fully synchronized
6009
      - disconnect from the network
6010
      - change disks into single-master mode
6011
      - wait again until disks are fully synchronized
6012

6013
    """
6014
    instance = self.instance
6015
    target_node = self.target_node
6016
    source_node = self.source_node
6017

    
6018
    # check running on only one node
6019
    self.feedback_fn("* checking where the instance actually runs"
6020
                     " (if this hangs, the hypervisor might be in"
6021
                     " a bad state)")
6022
    ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
6023
    for node, result in ins_l.items():
6024
      result.Raise("Can't contact node %s" % node)
6025

    
6026
    runningon_source = instance.name in ins_l[source_node].payload
6027
    runningon_target = instance.name in ins_l[target_node].payload
6028

    
6029
    if runningon_source and runningon_target:
6030
      raise errors.OpExecError("Instance seems to be running on two nodes,"
6031
                               " or the hypervisor is confused. You will have"
6032
                               " to ensure manually that it runs only on one"
6033
                               " and restart this operation.")
6034

    
6035
    if not (runningon_source or runningon_target):
6036
      raise errors.OpExecError("Instance does not seem to be running at all."
6037
                               " In this case, it's safer to repair by"
6038
                               " running 'gnt-instance stop' to ensure disk"
6039
                               " shutdown, and then restarting it.")
6040

    
6041
    if runningon_target:
6042
      # the migration has actually succeeded, we need to update the config
6043
      self.feedback_fn("* instance running on secondary node (%s),"
6044
                       " updating config" % target_node)
6045
      instance.primary_node = target_node
6046
      self.cfg.Update(instance, self.feedback_fn)
6047
      demoted_node = source_node
6048
    else:
6049
      self.feedback_fn("* instance confirmed to be running on its"
6050
                       " primary node (%s)" % source_node)
6051
      demoted_node = target_node
6052

    
6053
    self._EnsureSecondary(demoted_node)
6054
    try:
6055
      self._WaitUntilSync()
6056
    except errors.OpExecError:
6057
      # we ignore here errors, since if the device is standalone, it
6058
      # won't be able to sync
6059
      pass
6060
    self._GoStandalone()
6061
    self._GoReconnect(False)
6062
    self._WaitUntilSync()
6063

    
6064
    self.feedback_fn("* done")
6065

    
6066
  def _RevertDiskStatus(self):
6067
    """Try to revert the disk status after a failed migration.
6068

6069
    """
6070
    target_node = self.target_node
6071
    try:
6072
      self._EnsureSecondary(target_node)
6073
      self._GoStandalone()
6074
      self._GoReconnect(False)
6075
      self._WaitUntilSync()
6076
    except errors.OpExecError, err:
6077
      self.lu.LogWarning("Migration failed and I can't reconnect the"
6078
                         " drives: error '%s'\n"
6079
                         "Please look and recover the instance status" %
6080
                         str(err))
6081

    
6082
  def _AbortMigration(self):
6083
    """Call the hypervisor code to abort a started migration.
6084

6085
    """
6086
    instance = self.instance
6087
    target_node = self.target_node
6088
    migration_info = self.migration_info
6089

    
6090
    abort_result = self.rpc.call_finalize_migration(target_node,
6091
                                                    instance,
6092
                                                    migration_info,
6093
                                                    False)
6094
    abort_msg = abort_result.fail_msg
6095
    if abort_msg:
6096
      logging.error("Aborting migration failed on target node %s: %s",
6097
                    target_node, abort_msg)
6098
      # Don't raise an exception here, as we stil have to try to revert the
6099
      # disk status, even if this step failed.
6100

    
6101
  def _ExecMigration(self):
6102
    """Migrate an instance.
6103

6104
    The migrate is done by:
6105
      - change the disks into dual-master mode
6106
      - wait until disks are fully synchronized again
6107
      - migrate the instance
6108
      - change disks on the new secondary node (the old primary) to secondary
6109
      - wait until disks are fully synchronized
6110
      - change disks into single-master mode
6111

6112
    """
6113
    instance = self.instance
6114
    target_node = self.target_node
6115
    source_node = self.source_node
6116

    
6117
    self.feedback_fn("* checking disk consistency between source and target")
6118
    for dev in instance.disks:
6119
      if not _CheckDiskConsistency(self.lu, dev, target_node, False):
6120
        raise errors.OpExecError("Disk %s is degraded or not fully"
6121
                                 " synchronized on target node,"
6122
                                 " aborting migrate." % dev.iv_name)
6123

    
6124
    # First get the migration information from the remote node
6125
    result = self.rpc.call_migration_info(source_node, instance)
6126
    msg = result.fail_msg
6127
    if msg:
6128
      log_err = ("Failed fetching source migration information from %s: %s" %
6129
                 (source_node, msg))
6130
      logging.error(log_err)
6131
      raise errors.OpExecError(log_err)
6132

    
6133
    self.migration_info = migration_info = result.payload
6134

    
6135
    # Then switch the disks to master/master mode
6136
    self._EnsureSecondary(target_node)
6137
    self._GoStandalone()
6138
    self._GoReconnect(True)
6139
    self._WaitUntilSync()
6140

    
6141
    self.feedback_fn("* preparing %s to accept the instance" % target_node)
6142
    result = self.rpc.call_accept_instance(target_node,
6143
                                           instance,
6144
                                           migration_info,
6145
                                           self.nodes_ip[target_node])
6146

    
6147
    msg = result.fail_msg
6148
    if msg:
6149
      logging.error("Instance pre-migration failed, trying to revert"
6150
                    " disk status: %s", msg)
6151
      self.feedback_fn("Pre-migration failed, aborting")
6152
      self._AbortMigration()
6153
      self._RevertDiskStatus()
6154
      raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
6155
                               (instance.name, msg))
6156

    
6157
    self.feedback_fn("* migrating instance to %s" % target_node)
6158
    time.sleep(10)
6159
    result = self.rpc.call_instance_migrate(source_node, instance,
6160
                                            self.nodes_ip[target_node],
6161
                                            self.live)
6162
    msg = result.fail_msg
6163
    if msg:
6164
      logging.error("Instance migration failed, trying to revert"
6165
                    " disk status: %s", msg)
6166
      self.feedback_fn("Migration failed, aborting")
6167
      self._AbortMigration()
6168
      self._RevertDiskStatus()
6169
      raise errors.OpExecError("Could not migrate instance %s: %s" %
6170
                               (instance.name, msg))
6171
    time.sleep(10)
6172

    
6173
    instance.primary_node = target_node
6174
    # distribute new instance config to the other nodes
6175
    self.cfg.Update(instance, self.feedback_fn)
6176

    
6177
    result = self.rpc.call_finalize_migration(target_node,
6178
                                              instance,
6179
                                              migration_info,
6180
                                              True)
6181
    msg = result.fail_msg
6182
    if msg:
6183
      logging.error("Instance migration succeeded, but finalization failed:"
6184
                    " %s", msg)
6185
      raise errors.OpExecError("Could not finalize instance migration: %s" %
6186
                               msg)
6187

    
6188
    self._EnsureSecondary(source_node)
6189
    self._WaitUntilSync()
6190
    self._GoStandalone()
6191
    self._GoReconnect(False)
6192
    self._WaitUntilSync()
6193

    
6194
    self.feedback_fn("* done")
6195

    
6196
  def Exec(self, feedback_fn):
6197
    """Perform the migration.
6198

6199
    """
6200
    feedback_fn("Migrating instance %s" % self.instance.name)
6201

    
6202
    self.feedback_fn = feedback_fn
6203

    
6204
    self.source_node = self.instance.primary_node
6205
    self.target_node = self.instance.secondary_nodes[0]
6206
    self.all_nodes = [self.source_node, self.target_node]
6207
    self.nodes_ip = {
6208
      self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
6209
      self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
6210
      }
6211

    
6212
    if self.cleanup:
6213
      return self._ExecCleanup()
6214
    else:
6215
      return self._ExecMigration()
6216

    
6217

    
6218
def _CreateBlockDev(lu, node, instance, device, force_create,
6219
                    info, force_open):
6220
  """Create a tree of block devices on a given node.
6221

6222
  If this device type has to be created on secondaries, create it and
6223
  all its children.
6224

6225
  If not, just recurse to children keeping the same 'force' value.
6226

6227
  @param lu: the lu on whose behalf we execute
6228
  @param node: the node on which to create the device
6229
  @type instance: L{objects.Instance}
6230
  @param instance: the instance which owns the device
6231
  @type device: L{objects.Disk}
6232
  @param device: the device to create
6233
  @type force_create: boolean
6234
  @param force_create: whether to force creation of this device; this
6235
      will be change to True whenever we find a device which has
6236
      CreateOnSecondary() attribute
6237
  @param info: the extra 'metadata' we should attach to the device
6238
      (this will be represented as a LVM tag)
6239
  @type force_open: boolean
6240
  @param force_open: this parameter will be passes to the
6241
      L{backend.BlockdevCreate} function where it specifies
6242
      whether we run on primary or not, and it affects both
6243
      the child assembly and the device own Open() execution
6244

6245
  """
6246
  if device.CreateOnSecondary():
6247
    force_create = True
6248

    
6249
  if device.children:
6250
    for child in device.children:
6251
      _CreateBlockDev(lu, node, instance, child, force_create,
6252
                      info, force_open)
6253

    
6254
  if not force_create:
6255
    return
6256

    
6257
  _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
6258

    
6259

    
6260
def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
6261
  """Create a single block device on a given node.
6262

6263
  This will not recurse over children of the device, so they must be
6264
  created in advance.
6265

6266
  @param lu: the lu on whose behalf we execute
6267
  @param node: the node on which to create the device
6268
  @type instance: L{objects.Instance}
6269
  @param instance: the instance which owns the device
6270
  @type device: L{objects.Disk}
6271
  @param device: the device to create
6272
  @param info: the extra 'metadata' we should attach to the device
6273
      (this will be represented as a LVM tag)
6274
  @type force_open: boolean
6275
  @param force_open: this parameter will be passes to the
6276
      L{backend.BlockdevCreate} function where it specifies
6277
      whether we run on primary or not, and it affects both
6278
      the child assembly and the device own Open() execution
6279

6280
  """
6281
  lu.cfg.SetDiskID(device, node)
6282
  result = lu.rpc.call_blockdev_create(node, device, device.size,
6283
                                       instance.name, force_open, info)
6284
  result.Raise("Can't create block device %s on"
6285
               " node %s for instance %s" % (device, node, instance.name))
6286
  if device.physical_id is None:
6287
    device.physical_id = result.payload
6288

    
6289

    
6290
def _GenerateUniqueNames(lu, exts):
6291
  """Generate a suitable LV name.
6292

6293
  This will generate a logical volume name for the given instance.
6294

6295
  """
6296
  results = []
6297
  for val in exts:
6298
    new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
6299
    results.append("%s%s" % (new_id, val))
6300
  return results
6301

    
6302

    
6303
def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
6304
                         p_minor, s_minor):
6305
  """Generate a drbd8 device complete with its children.
6306

6307
  """
6308
  port = lu.cfg.AllocatePort()
6309
  vgname = lu.cfg.GetVGName()
6310
  shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
6311
  dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
6312
                          logical_id=(vgname, names[0]))
6313
  dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
6314
                          logical_id=(vgname, names[1]))
6315
  drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
6316
                          logical_id=(primary, secondary, port,
6317
                                      p_minor, s_minor,
6318
                                      shared_secret),
6319
                          children=[dev_data, dev_meta],
6320
                          iv_name=iv_name)
6321
  return drbd_dev
6322

    
6323

    
6324
def _GenerateDiskTemplate(lu, template_name,
6325
                          instance_name, primary_node,
6326
                          secondary_nodes, disk_info,
6327
                          file_storage_dir, file_driver,
6328
                          base_index):
6329
  """Generate the entire disk layout for a given template type.
6330

6331
  """
6332
  #TODO: compute space requirements
6333

    
6334
  vgname = lu.cfg.GetVGName()
6335
  disk_count = len(disk_info)
6336
  disks = []
6337
  if template_name == constants.DT_DISKLESS:
6338
    pass
6339
  elif template_name == constants.DT_PLAIN:
6340
    if len(secondary_nodes) != 0:
6341
      raise errors.ProgrammerError("Wrong template configuration")
6342

    
6343
    names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
6344
                                      for i in range(disk_count)])
6345
    for idx, disk in enumerate(disk_info):
6346
      disk_index = idx + base_index
6347
      disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
6348
                              logical_id=(vgname, names[idx]),
6349
                              iv_name="disk/%d" % disk_index,
6350
                              mode=disk["mode"])
6351
      disks.append(disk_dev)
6352
  elif template_name == constants.DT_DRBD8:
6353
    if len(secondary_nodes) != 1:
6354
      raise errors.ProgrammerError("Wrong template configuration")
6355
    remote_node = secondary_nodes[0]
6356
    minors = lu.cfg.AllocateDRBDMinor(
6357
      [primary_node, remote_node] * len(disk_info), instance_name)
6358

    
6359
    names = []
6360
    for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
6361
                                               for i in range(disk_count)]):
6362
      names.append(lv_prefix + "_data")
6363
      names.append(lv_prefix + "_meta")
6364
    for idx, disk in enumerate(disk_info):
6365
      disk_index = idx + base_index
6366
      disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
6367
                                      disk["size"], names[idx*2:idx*2+2],
6368
                                      "disk/%d" % disk_index,
6369
                                      minors[idx*2], minors[idx*2+1])
6370
      disk_dev.mode = disk["mode"]
6371
      disks.append(disk_dev)
6372
  elif template_name == constants.DT_FILE:
6373
    if len(secondary_nodes) != 0:
6374
      raise errors.ProgrammerError("Wrong template configuration")
6375

    
6376
    _RequireFileStorage()
6377

    
6378
    for idx, disk in enumerate(disk_info):
6379
      disk_index = idx + base_index
6380
      disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
6381
                              iv_name="disk/%d" % disk_index,
6382
                              logical_id=(file_driver,
6383
                                          "%s/disk%d" % (file_storage_dir,
6384
                                                         disk_index)),
6385
                              mode=disk["mode"])
6386
      disks.append(disk_dev)
6387
  else:
6388
    raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
6389
  return disks
6390

    
6391

    
6392
def _GetInstanceInfoText(instance):
6393
  """Compute that text that should be added to the disk's metadata.
6394

6395
  """
6396
  return "originstname+%s" % instance.name
6397

    
6398

    
6399
def _CreateDisks(lu, instance, to_skip=None, target_node=None):
6400
  """Create all disks for an instance.
6401

6402
  This abstracts away some work from AddInstance.
6403

6404
  @type lu: L{LogicalUnit}
6405
  @param lu: the logical unit on whose behalf we execute
6406
  @type instance: L{objects.Instance}
6407
  @param instance: the instance whose disks we should create
6408
  @type to_skip: list
6409
  @param to_skip: list of indices to skip
6410
  @type target_node: string
6411
  @param target_node: if passed, overrides the target node for creation
6412
  @rtype: boolean
6413
  @return: the success of the creation
6414

6415
  """
6416
  info = _GetInstanceInfoText(instance)
6417
  if target_node is None:
6418
    pnode = instance.primary_node
6419
    all_nodes = instance.all_nodes
6420
  else:
6421
    pnode = target_node
6422
    all_nodes = [pnode]
6423

    
6424
  if instance.disk_template == constants.DT_FILE:
6425
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
6426
    result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
6427

    
6428
    result.Raise("Failed to create directory '%s' on"
6429
                 " node %s" % (file_storage_dir, pnode))
6430

    
6431
  # Note: this needs to be kept in sync with adding of disks in
6432
  # LUSetInstanceParams
6433
  for idx, device in enumerate(instance.disks):
6434
    if to_skip and idx in to_skip:
6435
      continue
6436
    logging.info("Creating volume %s for instance %s",
6437
                 device.iv_name, instance.name)
6438
    #HARDCODE
6439
    for node in all_nodes:
6440
      f_create = node == pnode
6441
      _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
6442

    
6443

    
6444
def _RemoveDisks(lu, instance, target_node=None):
6445
  """Remove all disks for an instance.
6446

6447
  This abstracts away some work from `AddInstance()` and
6448
  `RemoveInstance()`. Note that in case some of the devices couldn't
6449
  be removed, the removal will continue with the other ones (compare
6450
  with `_CreateDisks()`).
6451

6452
  @type lu: L{LogicalUnit}
6453
  @param lu: the logical unit on whose behalf we execute
6454
  @type instance: L{objects.Instance}
6455
  @param instance: the instance whose disks we should remove
6456
  @type target_node: string
6457
  @param target_node: used to override the node on which to remove the disks
6458
  @rtype: boolean
6459
  @return: the success of the removal
6460

6461
  """
6462
  logging.info("Removing block devices for instance %s", instance.name)
6463

    
6464
  all_result = True
6465
  for device in instance.disks:
6466
    if target_node:
6467
      edata = [(target_node, device)]
6468
    else:
6469
      edata = device.ComputeNodeTree(instance.primary_node)
6470
    for node, disk in edata:
6471
      lu.cfg.SetDiskID(disk, node)
6472
      msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
6473
      if msg:
6474
        lu.LogWarning("Could not remove block device %s on node %s,"
6475
                      " continuing anyway: %s", device.iv_name, node, msg)
6476
        all_result = False
6477

    
6478
  if instance.disk_template == constants.DT_FILE:
6479
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
6480
    if target_node:
6481
      tgt = target_node
6482
    else:
6483
      tgt = instance.primary_node
6484
    result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
6485
    if result.fail_msg:
6486
      lu.LogWarning("Could not remove directory '%s' on node %s: %s",
6487
                    file_storage_dir, instance.primary_node, result.fail_msg)
6488
      all_result = False
6489

    
6490
  return all_result
6491

    
6492

    
6493
def _ComputeDiskSize(disk_template, disks):
6494
  """Compute disk size requirements in the volume group
6495

6496
  """
6497
  # Required free disk space as a function of disk and swap space
6498
  req_size_dict = {
6499
    constants.DT_DISKLESS: None,
6500
    constants.DT_PLAIN: sum(d["size"] for d in disks),
6501
    # 128 MB are added for drbd metadata for each disk
6502
    constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
6503
    constants.DT_FILE: None,
6504
  }
6505

    
6506
  if disk_template not in req_size_dict:
6507
    raise errors.ProgrammerError("Disk template '%s' size requirement"
6508
                                 " is unknown" %  disk_template)
6509

    
6510
  return req_size_dict[disk_template]
6511

    
6512

    
6513
def _CheckHVParams(lu, nodenames, hvname, hvparams):
6514
  """Hypervisor parameter validation.
6515

6516
  This function abstract the hypervisor parameter validation to be
6517
  used in both instance create and instance modify.
6518

6519
  @type lu: L{LogicalUnit}
6520
  @param lu: the logical unit for which we check
6521
  @type nodenames: list
6522
  @param nodenames: the list of nodes on which we should check
6523
  @type hvname: string
6524
  @param hvname: the name of the hypervisor we should use
6525
  @type hvparams: dict
6526
  @param hvparams: the parameters which we need to check
6527
  @raise errors.OpPrereqError: if the parameters are not valid
6528

6529
  """
6530
  hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
6531
                                                  hvname,
6532
                                                  hvparams)
6533
  for node in nodenames:
6534
    info = hvinfo[node]
6535
    if info.offline:
6536
      continue
6537
    info.Raise("Hypervisor parameter validation failed on node %s" % node)
6538

    
6539

    
6540
def _CheckOSParams(lu, required, nodenames, osname, osparams):
6541
  """OS parameters validation.
6542

6543
  @type lu: L{LogicalUnit}
6544
  @param lu: the logical unit for which we check
6545
  @type required: boolean
6546
  @param required: whether the validation should fail if the OS is not
6547
      found
6548
  @type nodenames: list
6549
  @param nodenames: the list of nodes on which we should check
6550
  @type osname: string
6551
  @param osname: the name of the hypervisor we should use
6552
  @type osparams: dict
6553
  @param osparams: the parameters which we need to check
6554
  @raise errors.OpPrereqError: if the parameters are not valid
6555

6556
  """
6557
  result = lu.rpc.call_os_validate(required, nodenames, osname,
6558
                                   [constants.OS_VALIDATE_PARAMETERS],
6559
                                   osparams)
6560
  for node, nres in result.items():
6561
    # we don't check for offline cases since this should be run only
6562
    # against the master node and/or an instance's nodes
6563
    nres.Raise("OS Parameters validation failed on node %s" % node)
6564
    if not nres.payload:
6565
      lu.LogInfo("OS %s not found on node %s, validation skipped",
6566
                 osname, node)
6567

    
6568

    
6569
class LUCreateInstance(LogicalUnit):
6570
  """Create an instance.
6571

6572
  """
6573
  HPATH = "instance-add"
6574
  HTYPE = constants.HTYPE_INSTANCE
6575
  _OP_PARAMS = [
6576
    _PInstanceName,
6577
    ("mode", _NoDefault, _TElemOf(constants.INSTANCE_CREATE_MODES)),
6578
    ("start", True, _TBool),
6579
    ("wait_for_sync", True, _TBool),
6580
    ("ip_check", True, _TBool),
6581
    ("name_check", True, _TBool),
6582
    ("disks", _NoDefault, _TListOf(_TDict)),
6583
    ("nics", _NoDefault, _TListOf(_TDict)),
6584
    ("hvparams", _EmptyDict, _TDict),
6585
    ("beparams", _EmptyDict, _TDict),
6586
    ("osparams", _EmptyDict, _TDict),
6587
    ("no_install", None, _TMaybeBool),
6588
    ("os_type", None, _TMaybeString),
6589
    ("force_variant", False, _TBool),
6590
    ("source_handshake", None, _TOr(_TList, _TNone)),
6591
    ("source_x509_ca", None, _TMaybeString),
6592
    ("source_instance_name", None, _TMaybeString),
6593
    ("src_node", None, _TMaybeString),
6594
    ("src_path", None, _TMaybeString),
6595
    ("pnode", None, _TMaybeString),
6596
    ("snode", None, _TMaybeString),
6597
    ("iallocator", None, _TMaybeString),
6598
    ("hypervisor", None, _TMaybeString),
6599
    ("disk_template", _NoDefault, _CheckDiskTemplate),
6600
    ("identify_defaults", False, _TBool),
6601
    ("file_driver", None, _TOr(_TNone, _TElemOf(constants.FILE_DRIVER))),
6602
    ("file_storage_dir", None, _TMaybeString),
6603
    ]
6604
  REQ_BGL = False
6605

    
6606
  def CheckArguments(self):
6607
    """Check arguments.
6608

6609
    """
6610
    # do not require name_check to ease forward/backward compatibility
6611
    # for tools
6612
    if self.op.no_install and self.op.start:
6613
      self.LogInfo("No-installation mode selected, disabling startup")
6614
      self.op.start = False
6615
    # validate/normalize the instance name
6616
    self.op.instance_name = \
6617
      netutils.HostInfo.NormalizeName(self.op.instance_name)
6618

    
6619
    if self.op.ip_check and not self.op.name_check:
6620
      # TODO: make the ip check more flexible and not depend on the name check
6621
      raise errors.OpPrereqError("Cannot do ip check without a name check",
6622
                                 errors.ECODE_INVAL)
6623

    
6624
    # check nics' parameter names
6625
    for nic in self.op.nics:
6626
      utils.ForceDictType(nic, constants.INIC_PARAMS_TYPES)
6627

    
6628
    # check disks. parameter names and consistent adopt/no-adopt strategy
6629
    has_adopt = has_no_adopt = False
6630
    for disk in self.op.disks:
6631
      utils.ForceDictType(disk, constants.IDISK_PARAMS_TYPES)
6632
      if "adopt" in disk:
6633
        has_adopt = True
6634
      else:
6635
        has_no_adopt = True
6636
    if has_adopt and has_no_adopt:
6637
      raise errors.OpPrereqError("Either all disks are adopted or none is",
6638
                                 errors.ECODE_INVAL)
6639
    if has_adopt:
6640
      if self.op.disk_template not in constants.DTS_MAY_ADOPT:
6641
        raise errors.OpPrereqError("Disk adoption is not supported for the"
6642
                                   " '%s' disk template" %
6643
                                   self.op.disk_template,
6644
                                   errors.ECODE_INVAL)
6645
      if self.op.iallocator is not None:
6646
        raise errors.OpPrereqError("Disk adoption not allowed with an"
6647
                                   " iallocator script", errors.ECODE_INVAL)
6648
      if self.op.mode == constants.INSTANCE_IMPORT:
6649
        raise errors.OpPrereqError("Disk adoption not allowed for"
6650
                                   " instance import", errors.ECODE_INVAL)
6651

    
6652
    self.adopt_disks = has_adopt
6653

    
6654
    # instance name verification
6655
    if self.op.name_check:
6656
      self.hostname1 = netutils.GetHostInfo(self.op.instance_name)
6657
      self.op.instance_name = self.hostname1.name
6658
      # used in CheckPrereq for ip ping check
6659
      self.check_ip = self.hostname1.ip
6660
    else:
6661
      self.check_ip = None
6662

    
6663
    # file storage checks
6664
    if (self.op.file_driver and
6665
        not self.op.file_driver in constants.FILE_DRIVER):
6666
      raise errors.OpPrereqError("Invalid file driver name '%s'" %
6667
                                 self.op.file_driver, errors.ECODE_INVAL)
6668

    
6669
    if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
6670
      raise errors.OpPrereqError("File storage directory path not absolute",
6671
                                 errors.ECODE_INVAL)
6672

    
6673
    ### Node/iallocator related checks
6674
    _CheckIAllocatorOrNode(self, "iallocator", "pnode")
6675

    
6676
    if self.op.pnode is not None:
6677
      if self.op.disk_template in constants.DTS_NET_MIRROR:
6678
        if self.op.snode is None:
6679
          raise errors.OpPrereqError("The networked disk templates need"
6680
                                     " a mirror node", errors.ECODE_INVAL)
6681
      elif self.op.snode:
6682
        self.LogWarning("Secondary node will be ignored on non-mirrored disk"
6683
                        " template")
6684
        self.op.snode = None
6685

    
6686
    self._cds = _GetClusterDomainSecret()
6687

    
6688
    if self.op.mode == constants.INSTANCE_IMPORT:
6689
      # On import force_variant must be True, because if we forced it at
6690
      # initial install, our only chance when importing it back is that it
6691
      # works again!
6692
      self.op.force_variant = True
6693

    
6694
      if self.op.no_install:
6695
        self.LogInfo("No-installation mode has no effect during import")
6696

    
6697
    elif self.op.mode == constants.INSTANCE_CREATE:
6698
      if self.op.os_type is None:
6699
        raise errors.OpPrereqError("No guest OS specified",
6700
                                   errors.ECODE_INVAL)
6701
      if self.op.os_type in self.cfg.GetClusterInfo().blacklisted_os:
6702
        raise errors.OpPrereqError("Guest OS '%s' is not allowed for"
6703
                                   " installation" % self.op.os_type,
6704
                                   errors.ECODE_STATE)
6705
      if self.op.disk_template is None:
6706
        raise errors.OpPrereqError("No disk template specified",
6707
                                   errors.ECODE_INVAL)
6708

    
6709
    elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
6710
      # Check handshake to ensure both clusters have the same domain secret
6711
      src_handshake = self.op.source_handshake
6712
      if not src_handshake:
6713
        raise errors.OpPrereqError("Missing source handshake",
6714
                                   errors.ECODE_INVAL)
6715

    
6716
      errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds,
6717
                                                           src_handshake)
6718
      if errmsg:
6719
        raise errors.OpPrereqError("Invalid handshake: %s" % errmsg,
6720
                                   errors.ECODE_INVAL)
6721

    
6722
      # Load and check source CA
6723
      self.source_x509_ca_pem = self.op.source_x509_ca
6724
      if not self.source_x509_ca_pem:
6725
        raise errors.OpPrereqError("Missing source X509 CA",
6726
                                   errors.ECODE_INVAL)
6727

    
6728
      try:
6729
        (cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem,
6730
                                                    self._cds)
6731
      except OpenSSL.crypto.Error, err:
6732
        raise errors.OpPrereqError("Unable to load source X509 CA (%s)" %
6733
                                   (err, ), errors.ECODE_INVAL)
6734

    
6735
      (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
6736
      if errcode is not None:
6737
        raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ),
6738
                                   errors.ECODE_INVAL)
6739

    
6740
      self.source_x509_ca = cert
6741

    
6742
      src_instance_name = self.op.source_instance_name
6743
      if not src_instance_name:
6744
        raise errors.OpPrereqError("Missing source instance name",
6745
                                   errors.ECODE_INVAL)
6746

    
6747
      norm_name = netutils.HostInfo.NormalizeName(src_instance_name)
6748
      self.source_instance_name = netutils.GetHostInfo(norm_name).name
6749

    
6750
    else:
6751
      raise errors.OpPrereqError("Invalid instance creation mode %r" %
6752
                                 self.op.mode, errors.ECODE_INVAL)
6753

    
6754
  def ExpandNames(self):
6755
    """ExpandNames for CreateInstance.
6756

6757
    Figure out the right locks for instance creation.
6758

6759
    """
6760
    self.needed_locks = {}
6761

    
6762
    instance_name = self.op.instance_name
6763
    # this is just a preventive check, but someone might still add this
6764
    # instance in the meantime, and creation will fail at lock-add time
6765
    if instance_name in self.cfg.GetInstanceList():
6766
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
6767
                                 instance_name, errors.ECODE_EXISTS)
6768

    
6769
    self.add_locks[locking.LEVEL_INSTANCE] = instance_name
6770

    
6771
    if self.op.iallocator:
6772
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6773
    else:
6774
      self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode)
6775
      nodelist = [self.op.pnode]
6776
      if self.op.snode is not None:
6777
        self.op.snode = _ExpandNodeName(self.cfg, self.op.snode)
6778
        nodelist.append(self.op.snode)
6779
      self.needed_locks[locking.LEVEL_NODE] = nodelist
6780

    
6781
    # in case of import lock the source node too
6782
    if self.op.mode == constants.INSTANCE_IMPORT:
6783
      src_node = self.op.src_node
6784
      src_path = self.op.src_path
6785

    
6786
      if src_path is None:
6787
        self.op.src_path = src_path = self.op.instance_name
6788

    
6789
      if src_node is None:
6790
        self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6791
        self.op.src_node = None
6792
        if os.path.isabs(src_path):
6793
          raise errors.OpPrereqError("Importing an instance from an absolute"
6794
                                     " path requires a source node option.",
6795
                                     errors.ECODE_INVAL)
6796
      else:
6797
        self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node)
6798
        if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
6799
          self.needed_locks[locking.LEVEL_NODE].append(src_node)
6800
        if not os.path.isabs(src_path):
6801
          self.op.src_path = src_path = \
6802
            utils.PathJoin(constants.EXPORT_DIR, src_path)
6803

    
6804
  def _RunAllocator(self):
6805
    """Run the allocator based on input opcode.
6806

6807
    """
6808
    nics = [n.ToDict() for n in self.nics]
6809
    ial = IAllocator(self.cfg, self.rpc,
6810
                     mode=constants.IALLOCATOR_MODE_ALLOC,
6811
                     name=self.op.instance_name,
6812
                     disk_template=self.op.disk_template,
6813
                     tags=[],
6814
                     os=self.op.os_type,
6815
                     vcpus=self.be_full[constants.BE_VCPUS],
6816
                     mem_size=self.be_full[constants.BE_MEMORY],
6817
                     disks=self.disks,
6818
                     nics=nics,
6819
                     hypervisor=self.op.hypervisor,
6820
                     )
6821

    
6822
    ial.Run(self.op.iallocator)
6823

    
6824
    if not ial.success:
6825
      raise errors.OpPrereqError("Can't compute nodes using"
6826
                                 " iallocator '%s': %s" %
6827
                                 (self.op.iallocator, ial.info),
6828
                                 errors.ECODE_NORES)
6829
    if len(ial.result) != ial.required_nodes:
6830
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
6831
                                 " of nodes (%s), required %s" %
6832
                                 (self.op.iallocator, len(ial.result),
6833
                                  ial.required_nodes), errors.ECODE_FAULT)
6834
    self.op.pnode = ial.result[0]
6835
    self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
6836
                 self.op.instance_name, self.op.iallocator,
6837
                 utils.CommaJoin(ial.result))
6838
    if ial.required_nodes == 2:
6839
      self.op.snode = ial.result[1]
6840

    
6841
  def BuildHooksEnv(self):
6842
    """Build hooks env.
6843

6844
    This runs on master, primary and secondary nodes of the instance.
6845

6846
    """
6847
    env = {
6848
      "ADD_MODE": self.op.mode,
6849
      }
6850
    if self.op.mode == constants.INSTANCE_IMPORT:
6851
      env["SRC_NODE"] = self.op.src_node
6852
      env["SRC_PATH"] = self.op.src_path
6853
      env["SRC_IMAGES"] = self.src_images
6854

    
6855
    env.update(_BuildInstanceHookEnv(
6856
      name=self.op.instance_name,
6857
      primary_node=self.op.pnode,
6858
      secondary_nodes=self.secondaries,
6859
      status=self.op.start,
6860
      os_type=self.op.os_type,
6861
      memory=self.be_full[constants.BE_MEMORY],
6862
      vcpus=self.be_full[constants.BE_VCPUS],
6863
      nics=_NICListToTuple(self, self.nics),
6864
      disk_template=self.op.disk_template,
6865
      disks=[(d["size"], d["mode"]) for d in self.disks],
6866
      bep=self.be_full,
6867
      hvp=self.hv_full,
6868
      hypervisor_name=self.op.hypervisor,
6869
    ))
6870

    
6871
    nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
6872
          self.secondaries)
6873
    return env, nl, nl
6874

    
6875
  def _ReadExportInfo(self):
6876
    """Reads the export information from disk.
6877

6878
    It will override the opcode source node and path with the actual
6879
    information, if these two were not specified before.
6880

6881
    @return: the export information
6882

6883
    """
6884
    assert self.op.mode == constants.INSTANCE_IMPORT
6885

    
6886
    src_node = self.op.src_node
6887
    src_path = self.op.src_path
6888

    
6889
    if src_node is None:
6890
      locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
6891
      exp_list = self.rpc.call_export_list(locked_nodes)
6892
      found = False
6893
      for node in exp_list:
6894
        if exp_list[node].fail_msg:
6895
          continue
6896
        if src_path in exp_list[node].payload:
6897
          found = True
6898
          self.op.src_node = src_node = node
6899
          self.op.src_path = src_path = utils.PathJoin(constants.EXPORT_DIR,
6900
                                                       src_path)
6901
          break
6902
      if not found:
6903
        raise errors.OpPrereqError("No export found for relative path %s" %
6904
                                    src_path, errors.ECODE_INVAL)
6905

    
6906
    _CheckNodeOnline(self, src_node)
6907
    result = self.rpc.call_export_info(src_node, src_path)
6908
    result.Raise("No export or invalid export found in dir %s" % src_path)
6909

    
6910
    export_info = objects.SerializableConfigParser.Loads(str(result.payload))
6911
    if not export_info.has_section(constants.INISECT_EXP):
6912
      raise errors.ProgrammerError("Corrupted export config",
6913
                                   errors.ECODE_ENVIRON)
6914

    
6915
    ei_version = export_info.get(constants.INISECT_EXP, "version")
6916
    if (int(ei_version) != constants.EXPORT_VERSION):
6917
      raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
6918
                                 (ei_version, constants.EXPORT_VERSION),
6919
                                 errors.ECODE_ENVIRON)
6920
    return export_info
6921

    
6922
  def _ReadExportParams(self, einfo):
6923
    """Use export parameters as defaults.
6924

6925
    In case the opcode doesn't specify (as in override) some instance
6926
    parameters, then try to use them from the export information, if
6927
    that declares them.
6928

6929
    """
6930
    self.op.os_type = einfo.get(constants.INISECT_EXP, "os")
6931

    
6932
    if self.op.disk_template is None:
6933
      if einfo.has_option(constants.INISECT_INS, "disk_template"):
6934
        self.op.disk_template = einfo.get(constants.INISECT_INS,
6935
                                          "disk_template")
6936
      else:
6937
        raise errors.OpPrereqError("No disk template specified and the export"
6938
                                   " is missing the disk_template information",
6939
                                   errors.ECODE_INVAL)
6940

    
6941
    if not self.op.disks:
6942
      if einfo.has_option(constants.INISECT_INS, "disk_count"):
6943
        disks = []
6944
        # TODO: import the disk iv_name too
6945
        for idx in range(einfo.getint(constants.INISECT_INS, "disk_count")):
6946
          disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
6947
          disks.append({"size": disk_sz})
6948
        self.op.disks = disks
6949
      else:
6950
        raise errors.OpPrereqError("No disk info specified and the export"
6951
                                   " is missing the disk information",
6952
                                   errors.ECODE_INVAL)
6953

    
6954
    if (not self.op.nics and
6955
        einfo.has_option(constants.INISECT_INS, "nic_count")):
6956
      nics = []
6957
      for idx in range(einfo.getint(constants.INISECT_INS, "nic_count")):
6958
        ndict = {}
6959
        for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]:
6960
          v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
6961
          ndict[name] = v
6962
        nics.append(ndict)
6963
      self.op.nics = nics
6964

    
6965
    if (self.op.hypervisor is None and
6966
        einfo.has_option(constants.INISECT_INS, "hypervisor")):
6967
      self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor")
6968
    if einfo.has_section(constants.INISECT_HYP):
6969
      # use the export parameters but do not override the ones
6970
      # specified by the user
6971
      for name, value in einfo.items(constants.INISECT_HYP):
6972
        if name not in self.op.hvparams:
6973
          self.op.hvparams[name] = value
6974

    
6975
    if einfo.has_section(constants.INISECT_BEP):
6976
      # use the parameters, without overriding
6977
      for name, value in einfo.items(constants.INISECT_BEP):
6978
        if name not in self.op.beparams:
6979
          self.op.beparams[name] = value
6980
    else:
6981
      # try to read the parameters old style, from the main section
6982
      for name in constants.BES_PARAMETERS:
6983
        if (name not in self.op.beparams and
6984
            einfo.has_option(constants.INISECT_INS, name)):
6985
          self.op.beparams[name] = einfo.get(constants.INISECT_INS, name)
6986

    
6987
    if einfo.has_section(constants.INISECT_OSP):
6988
      # use the parameters, without overriding
6989
      for name, value in einfo.items(constants.INISECT_OSP):
6990
        if name not in self.op.osparams:
6991
          self.op.osparams[name] = value
6992

    
6993
  def _RevertToDefaults(self, cluster):
6994
    """Revert the instance parameters to the default values.
6995

6996
    """
6997
    # hvparams
6998
    hv_defs = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, {})
6999
    for name in self.op.hvparams.keys():
7000
      if name in hv_defs and hv_defs[name] == self.op.hvparams[name]:
7001
        del self.op.hvparams[name]
7002
    # beparams
7003
    be_defs = cluster.SimpleFillBE({})
7004
    for name in self.op.beparams.keys():
7005
      if name in be_defs and be_defs[name] == self.op.beparams[name]:
7006
        del self.op.beparams[name]
7007
    # nic params
7008
    nic_defs = cluster.SimpleFillNIC({})
7009
    for nic in self.op.nics:
7010
      for name in constants.NICS_PARAMETERS:
7011
        if name in nic and name in nic_defs and nic[name] == nic_defs[name]:
7012
          del nic[name]
7013
    # osparams
7014
    os_defs = cluster.SimpleFillOS(self.op.os_type, {})
7015
    for name in self.op.osparams.keys():
7016
      if name in os_defs and os_defs[name] == self.op.osparams[name]:
7017
        del self.op.osparams[name]
7018

    
7019
  def CheckPrereq(self):
7020
    """Check prerequisites.
7021

7022
    """
7023
    if self.op.mode == constants.INSTANCE_IMPORT:
7024
      export_info = self._ReadExportInfo()
7025
      self._ReadExportParams(export_info)
7026

    
7027
    _CheckDiskTemplate(self.op.disk_template)
7028

    
7029
    if (not self.cfg.GetVGName() and
7030
        self.op.disk_template not in constants.DTS_NOT_LVM):
7031
      raise errors.OpPrereqError("Cluster does not support lvm-based"
7032
                                 " instances", errors.ECODE_STATE)
7033

    
7034
    if self.op.hypervisor is None:
7035
      self.op.hypervisor = self.cfg.GetHypervisorType()
7036

    
7037
    cluster = self.cfg.GetClusterInfo()
7038
    enabled_hvs = cluster.enabled_hypervisors
7039
    if self.op.hypervisor not in enabled_hvs:
7040
      raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
7041
                                 " cluster (%s)" % (self.op.hypervisor,
7042
                                  ",".join(enabled_hvs)),
7043
                                 errors.ECODE_STATE)
7044

    
7045
    # check hypervisor parameter syntax (locally)
7046
    utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
7047
    filled_hvp = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type,
7048
                                      self.op.hvparams)
7049
    hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
7050
    hv_type.CheckParameterSyntax(filled_hvp)
7051
    self.hv_full = filled_hvp
7052
    # check that we don't specify global parameters on an instance
7053
    _CheckGlobalHvParams(self.op.hvparams)
7054

    
7055
    # fill and remember the beparams dict
7056
    utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
7057
    self.be_full = cluster.SimpleFillBE(self.op.beparams)
7058

    
7059
    # build os parameters
7060
    self.os_full = cluster.SimpleFillOS(self.op.os_type, self.op.osparams)
7061

    
7062
    # now that hvp/bep are in final format, let's reset to defaults,
7063
    # if told to do so
7064
    if self.op.identify_defaults:
7065
      self._RevertToDefaults(cluster)
7066

    
7067
    # NIC buildup
7068
    self.nics = []
7069
    for idx, nic in enumerate(self.op.nics):
7070
      nic_mode_req = nic.get("mode", None)
7071
      nic_mode = nic_mode_req
7072
      if nic_mode is None:
7073
        nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
7074

    
7075
      # in routed mode, for the first nic, the default ip is 'auto'
7076
      if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
7077
        default_ip_mode = constants.VALUE_AUTO
7078
      else:
7079
        default_ip_mode = constants.VALUE_NONE
7080

    
7081
      # ip validity checks
7082
      ip = nic.get("ip", default_ip_mode)
7083
      if ip is None or ip.lower() == constants.VALUE_NONE:
7084
        nic_ip = None
7085
      elif ip.lower() == constants.VALUE_AUTO:
7086
        if not self.op.name_check:
7087
          raise errors.OpPrereqError("IP address set to auto but name checks"
7088
                                     " have been skipped. Aborting.",
7089
                                     errors.ECODE_INVAL)
7090
        nic_ip = self.hostname1.ip
7091
      else:
7092
        if not netutils.IsValidIP4(ip):
7093
          raise errors.OpPrereqError("Given IP address '%s' doesn't look"
7094
                                     " like a valid IP" % ip,
7095
                                     errors.ECODE_INVAL)
7096
        nic_ip = ip
7097

    
7098
      # TODO: check the ip address for uniqueness
7099
      if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
7100
        raise errors.OpPrereqError("Routed nic mode requires an ip address",
7101
                                   errors.ECODE_INVAL)
7102

    
7103
      # MAC address verification
7104
      mac = nic.get("mac", constants.VALUE_AUTO)
7105
      if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7106
        mac = utils.NormalizeAndValidateMac(mac)
7107

    
7108
        try:
7109
          self.cfg.ReserveMAC(mac, self.proc.GetECId())
7110
        except errors.ReservationError:
7111
          raise errors.OpPrereqError("MAC address %s already in use"
7112
                                     " in cluster" % mac,
7113
                                     errors.ECODE_NOTUNIQUE)
7114

    
7115
      # bridge verification
7116
      bridge = nic.get("bridge", None)
7117
      link = nic.get("link", None)
7118
      if bridge and link:
7119
        raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
7120
                                   " at the same time", errors.ECODE_INVAL)
7121
      elif bridge and nic_mode == constants.NIC_MODE_ROUTED:
7122
        raise errors.OpPrereqError("Cannot pass 'bridge' on a routed nic",
7123
                                   errors.ECODE_INVAL)
7124
      elif bridge:
7125
        link = bridge
7126

    
7127
      nicparams = {}
7128
      if nic_mode_req:
7129
        nicparams[constants.NIC_MODE] = nic_mode_req
7130
      if link:
7131
        nicparams[constants.NIC_LINK] = link
7132

    
7133
      check_params = cluster.SimpleFillNIC(nicparams)
7134
      objects.NIC.CheckParameterSyntax(check_params)
7135
      self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
7136

    
7137
    # disk checks/pre-build
7138
    self.disks = []
7139
    for disk in self.op.disks:
7140
      mode = disk.get("mode", constants.DISK_RDWR)
7141
      if mode not in constants.DISK_ACCESS_SET:
7142
        raise errors.OpPrereqError("Invalid disk access mode '%s'" %
7143
                                   mode, errors.ECODE_INVAL)
7144
      size = disk.get("size", None)
7145
      if size is None:
7146
        raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
7147
      try:
7148
        size = int(size)
7149
      except (TypeError, ValueError):
7150
        raise errors.OpPrereqError("Invalid disk size '%s'" % size,
7151
                                   errors.ECODE_INVAL)
7152
      new_disk = {"size": size, "mode": mode}
7153
      if "adopt" in disk:
7154
        new_disk["adopt"] = disk["adopt"]
7155
      self.disks.append(new_disk)
7156

    
7157
    if self.op.mode == constants.INSTANCE_IMPORT:
7158

    
7159
      # Check that the new instance doesn't have less disks than the export
7160
      instance_disks = len(self.disks)
7161
      export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
7162
      if instance_disks < export_disks:
7163
        raise errors.OpPrereqError("Not enough disks to import."
7164
                                   " (instance: %d, export: %d)" %
7165
                                   (instance_disks, export_disks),
7166
                                   errors.ECODE_INVAL)
7167

    
7168
      disk_images = []
7169
      for idx in range(export_disks):
7170
        option = 'disk%d_dump' % idx
7171
        if export_info.has_option(constants.INISECT_INS, option):
7172
          # FIXME: are the old os-es, disk sizes, etc. useful?
7173
          export_name = export_info.get(constants.INISECT_INS, option)
7174
          image = utils.PathJoin(self.op.src_path, export_name)
7175
          disk_images.append(image)
7176
        else:
7177
          disk_images.append(False)
7178

    
7179
      self.src_images = disk_images
7180

    
7181
      old_name = export_info.get(constants.INISECT_INS, 'name')
7182
      try:
7183
        exp_nic_count = export_info.getint(constants.INISECT_INS, 'nic_count')
7184
      except (TypeError, ValueError), err:
7185
        raise errors.OpPrereqError("Invalid export file, nic_count is not"
7186
                                   " an integer: %s" % str(err),
7187
                                   errors.ECODE_STATE)
7188
      if self.op.instance_name == old_name:
7189
        for idx, nic in enumerate(self.nics):
7190
          if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
7191
            nic_mac_ini = 'nic%d_mac' % idx
7192
            nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
7193

    
7194
    # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
7195

    
7196
    # ip ping checks (we use the same ip that was resolved in ExpandNames)
7197
    if self.op.ip_check:
7198
      if netutils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
7199
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
7200
                                   (self.check_ip, self.op.instance_name),
7201
                                   errors.ECODE_NOTUNIQUE)
7202

    
7203
    #### mac address generation
7204
    # By generating here the mac address both the allocator and the hooks get
7205
    # the real final mac address rather than the 'auto' or 'generate' value.
7206
    # There is a race condition between the generation and the instance object
7207
    # creation, which means that we know the mac is valid now, but we're not
7208
    # sure it will be when we actually add the instance. If things go bad
7209
    # adding the instance will abort because of a duplicate mac, and the
7210
    # creation job will fail.
7211
    for nic in self.nics:
7212
      if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7213
        nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
7214

    
7215
    #### allocator run
7216

    
7217
    if self.op.iallocator is not None:
7218
      self._RunAllocator()
7219

    
7220
    #### node related checks
7221

    
7222
    # check primary node
7223
    self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
7224
    assert self.pnode is not None, \
7225
      "Cannot retrieve locked node %s" % self.op.pnode
7226
    if pnode.offline:
7227
      raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
7228
                                 pnode.name, errors.ECODE_STATE)
7229
    if pnode.drained:
7230
      raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
7231
                                 pnode.name, errors.ECODE_STATE)
7232

    
7233
    self.secondaries = []
7234

    
7235
    # mirror node verification
7236
    if self.op.disk_template in constants.DTS_NET_MIRROR:
7237
      if self.op.snode == pnode.name:
7238
        raise errors.OpPrereqError("The secondary node cannot be the"
7239
                                   " primary node.", errors.ECODE_INVAL)
7240
      _CheckNodeOnline(self, self.op.snode)
7241
      _CheckNodeNotDrained(self, self.op.snode)
7242
      self.secondaries.append(self.op.snode)
7243

    
7244
    nodenames = [pnode.name] + self.secondaries
7245

    
7246
    req_size = _ComputeDiskSize(self.op.disk_template,
7247
                                self.disks)
7248

    
7249
    # Check lv size requirements, if not adopting
7250
    if req_size is not None and not self.adopt_disks:
7251
      _CheckNodesFreeDisk(self, nodenames, req_size)
7252

    
7253
    if self.adopt_disks: # instead, we must check the adoption data
7254
      all_lvs = set([i["adopt"] for i in self.disks])
7255
      if len(all_lvs) != len(self.disks):
7256
        raise errors.OpPrereqError("Duplicate volume names given for adoption",
7257
                                   errors.ECODE_INVAL)
7258
      for lv_name in all_lvs:
7259
        try:
7260
          self.cfg.ReserveLV(lv_name, self.proc.GetECId())
7261
        except errors.ReservationError:
7262
          raise errors.OpPrereqError("LV named %s used by another instance" %
7263
                                     lv_name, errors.ECODE_NOTUNIQUE)
7264

    
7265
      node_lvs = self.rpc.call_lv_list([pnode.name],
7266
                                       self.cfg.GetVGName())[pnode.name]
7267
      node_lvs.Raise("Cannot get LV information from node %s" % pnode.name)
7268
      node_lvs = node_lvs.payload
7269
      delta = all_lvs.difference(node_lvs.keys())
7270
      if delta:
7271
        raise errors.OpPrereqError("Missing logical volume(s): %s" %
7272
                                   utils.CommaJoin(delta),
7273
                                   errors.ECODE_INVAL)
7274
      online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]]
7275
      if online_lvs:
7276
        raise errors.OpPrereqError("Online logical volumes found, cannot"
7277
                                   " adopt: %s" % utils.CommaJoin(online_lvs),
7278
                                   errors.ECODE_STATE)
7279
      # update the size of disk based on what is found
7280
      for dsk in self.disks:
7281
        dsk["size"] = int(float(node_lvs[dsk["adopt"]][0]))
7282

    
7283
    _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
7284

    
7285
    _CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant)
7286
    # check OS parameters (remotely)
7287
    _CheckOSParams(self, True, nodenames, self.op.os_type, self.os_full)
7288

    
7289
    _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
7290

    
7291
    # memory check on primary node
7292
    if self.op.start:
7293
      _CheckNodeFreeMemory(self, self.pnode.name,
7294
                           "creating instance %s" % self.op.instance_name,
7295
                           self.be_full[constants.BE_MEMORY],
7296
                           self.op.hypervisor)
7297

    
7298
    self.dry_run_result = list(nodenames)
7299

    
7300
  def Exec(self, feedback_fn):
7301
    """Create and add the instance to the cluster.
7302

7303
    """
7304
    instance = self.op.instance_name
7305
    pnode_name = self.pnode.name
7306

    
7307
    ht_kind = self.op.hypervisor
7308
    if ht_kind in constants.HTS_REQ_PORT:
7309
      network_port = self.cfg.AllocatePort()
7310
    else:
7311
      network_port = None
7312

    
7313
    if constants.ENABLE_FILE_STORAGE:
7314
      # this is needed because os.path.join does not accept None arguments
7315
      if self.op.file_storage_dir is None:
7316
        string_file_storage_dir = ""
7317
      else:
7318
        string_file_storage_dir = self.op.file_storage_dir
7319

    
7320
      # build the full file storage dir path
7321
      file_storage_dir = utils.PathJoin(self.cfg.GetFileStorageDir(),
7322
                                        string_file_storage_dir, instance)
7323
    else:
7324
      file_storage_dir = ""
7325

    
7326
    disks = _GenerateDiskTemplate(self,
7327
                                  self.op.disk_template,
7328
                                  instance, pnode_name,
7329
                                  self.secondaries,
7330
                                  self.disks,
7331
                                  file_storage_dir,
7332
                                  self.op.file_driver,
7333
                                  0)
7334

    
7335
    iobj = objects.Instance(name=instance, os=self.op.os_type,
7336
                            primary_node=pnode_name,
7337
                            nics=self.nics, disks=disks,
7338
                            disk_template=self.op.disk_template,
7339
                            admin_up=False,
7340
                            network_port=network_port,
7341
                            beparams=self.op.beparams,
7342
                            hvparams=self.op.hvparams,
7343
                            hypervisor=self.op.hypervisor,
7344
                            osparams=self.op.osparams,
7345
                            )
7346

    
7347
    if self.adopt_disks:
7348
      # rename LVs to the newly-generated names; we need to construct
7349
      # 'fake' LV disks with the old data, plus the new unique_id
7350
      tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks]
7351
      rename_to = []
7352
      for t_dsk, a_dsk in zip (tmp_disks, self.disks):
7353
        rename_to.append(t_dsk.logical_id)
7354
        t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk["adopt"])
7355
        self.cfg.SetDiskID(t_dsk, pnode_name)
7356
      result = self.rpc.call_blockdev_rename(pnode_name,
7357
                                             zip(tmp_disks, rename_to))
7358
      result.Raise("Failed to rename adoped LVs")
7359
    else:
7360
      feedback_fn("* creating instance disks...")
7361
      try:
7362
        _CreateDisks(self, iobj)
7363
      except errors.OpExecError:
7364
        self.LogWarning("Device creation failed, reverting...")
7365
        try:
7366
          _RemoveDisks(self, iobj)
7367
        finally:
7368
          self.cfg.ReleaseDRBDMinors(instance)
7369
          raise
7370

    
7371
    feedback_fn("adding instance %s to cluster config" % instance)
7372

    
7373
    self.cfg.AddInstance(iobj, self.proc.GetECId())
7374

    
7375
    # Declare that we don't want to remove the instance lock anymore, as we've
7376
    # added the instance to the config
7377
    del self.remove_locks[locking.LEVEL_INSTANCE]
7378
    # Unlock all the nodes
7379
    if self.op.mode == constants.INSTANCE_IMPORT:
7380
      nodes_keep = [self.op.src_node]
7381
      nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
7382
                       if node != self.op.src_node]
7383
      self.context.glm.release(locking.LEVEL_NODE, nodes_release)
7384
      self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
7385
    else:
7386
      self.context.glm.release(locking.LEVEL_NODE)
7387
      del self.acquired_locks[locking.LEVEL_NODE]
7388

    
7389
    if self.op.wait_for_sync:
7390
      disk_abort = not _WaitForSync(self, iobj)
7391
    elif iobj.disk_template in constants.DTS_NET_MIRROR:
7392
      # make sure the disks are not degraded (still sync-ing is ok)
7393
      time.sleep(15)
7394
      feedback_fn("* checking mirrors status")
7395
      disk_abort = not _WaitForSync(self, iobj, oneshot=True)
7396
    else:
7397
      disk_abort = False
7398

    
7399
    if disk_abort:
7400
      _RemoveDisks(self, iobj)
7401
      self.cfg.RemoveInstance(iobj.name)
7402
      # Make sure the instance lock gets removed
7403
      self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
7404
      raise errors.OpExecError("There are some degraded disks for"
7405
                               " this instance")
7406

    
7407
    if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks:
7408
      if self.op.mode == constants.INSTANCE_CREATE:
7409
        if not self.op.no_install:
7410
          feedback_fn("* running the instance OS create scripts...")
7411
          # FIXME: pass debug option from opcode to backend
7412
          result = self.rpc.call_instance_os_add(pnode_name, iobj, False,
7413
                                                 self.op.debug_level)
7414
          result.Raise("Could not add os for instance %s"
7415
                       " on node %s" % (instance, pnode_name))
7416

    
7417
      elif self.op.mode == constants.INSTANCE_IMPORT:
7418
        feedback_fn("* running the instance OS import scripts...")
7419

    
7420
        transfers = []
7421

    
7422
        for idx, image in enumerate(self.src_images):
7423
          if not image:
7424
            continue
7425

    
7426
          # FIXME: pass debug option from opcode to backend
7427
          dt = masterd.instance.DiskTransfer("disk/%s" % idx,
7428
                                             constants.IEIO_FILE, (image, ),
7429
                                             constants.IEIO_SCRIPT,
7430
                                             (iobj.disks[idx], idx),
7431
                                             None)
7432
          transfers.append(dt)
7433

    
7434
        import_result = \
7435
          masterd.instance.TransferInstanceData(self, feedback_fn,
7436
                                                self.op.src_node, pnode_name,
7437
                                                self.pnode.secondary_ip,
7438
                                                iobj, transfers)
7439
        if not compat.all(import_result):
7440
          self.LogWarning("Some disks for instance %s on node %s were not"
7441
                          " imported successfully" % (instance, pnode_name))
7442

    
7443
      elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
7444
        feedback_fn("* preparing remote import...")
7445
        connect_timeout = constants.RIE_CONNECT_TIMEOUT
7446
        timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
7447

    
7448
        disk_results = masterd.instance.RemoteImport(self, feedback_fn, iobj,
7449
                                                     self.source_x509_ca,
7450
                                                     self._cds, timeouts)
7451
        if not compat.all(disk_results):
7452
          # TODO: Should the instance still be started, even if some disks
7453
          # failed to import (valid for local imports, too)?
7454
          self.LogWarning("Some disks for instance %s on node %s were not"
7455
                          " imported successfully" % (instance, pnode_name))
7456

    
7457
        # Run rename script on newly imported instance
7458
        assert iobj.name == instance
7459
        feedback_fn("Running rename script for %s" % instance)
7460
        result = self.rpc.call_instance_run_rename(pnode_name, iobj,
7461
                                                   self.source_instance_name,
7462
                                                   self.op.debug_level)
7463
        if result.fail_msg:
7464
          self.LogWarning("Failed to run rename script for %s on node"
7465
                          " %s: %s" % (instance, pnode_name, result.fail_msg))
7466

    
7467
      else:
7468
        # also checked in the prereq part
7469
        raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
7470
                                     % self.op.mode)
7471

    
7472
    if self.op.start:
7473
      iobj.admin_up = True
7474
      self.cfg.Update(iobj, feedback_fn)
7475
      logging.info("Starting instance %s on node %s", instance, pnode_name)
7476
      feedback_fn("* starting instance...")
7477
      result = self.rpc.call_instance_start(pnode_name, iobj, None, None)
7478
      result.Raise("Could not start instance")
7479

    
7480
    return list(iobj.all_nodes)
7481

    
7482

    
7483
class LUConnectConsole(NoHooksLU):
7484
  """Connect to an instance's console.
7485

7486
  This is somewhat special in that it returns the command line that
7487
  you need to run on the master node in order to connect to the
7488
  console.
7489

7490
  """
7491
  _OP_PARAMS = [
7492
    _PInstanceName
7493
    ]
7494
  REQ_BGL = False
7495

    
7496
  def ExpandNames(self):
7497
    self._ExpandAndLockInstance()
7498

    
7499
  def CheckPrereq(self):
7500
    """Check prerequisites.
7501

7502
    This checks that the instance is in the cluster.
7503

7504
    """
7505
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7506
    assert self.instance is not None, \
7507
      "Cannot retrieve locked instance %s" % self.op.instance_name
7508
    _CheckNodeOnline(self, self.instance.primary_node)
7509

    
7510
  def Exec(self, feedback_fn):
7511
    """Connect to the console of an instance
7512

7513
    """
7514
    instance = self.instance
7515
    node = instance.primary_node
7516

    
7517
    node_insts = self.rpc.call_instance_list([node],
7518
                                             [instance.hypervisor])[node]
7519
    node_insts.Raise("Can't get node information from %s" % node)
7520

    
7521
    if instance.name not in node_insts.payload:
7522
      raise errors.OpExecError("Instance %s is not running." % instance.name)
7523

    
7524
    logging.debug("Connecting to console of %s on %s", instance.name, node)
7525

    
7526
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
7527
    cluster = self.cfg.GetClusterInfo()
7528
    # beparams and hvparams are passed separately, to avoid editing the
7529
    # instance and then saving the defaults in the instance itself.
7530
    hvparams = cluster.FillHV(instance)
7531
    beparams = cluster.FillBE(instance)
7532
    console_cmd = hyper.GetShellCommandForConsole(instance, hvparams, beparams)
7533

    
7534
    # build ssh cmdline
7535
    return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
7536

    
7537

    
7538
class LUReplaceDisks(LogicalUnit):
7539
  """Replace the disks of an instance.
7540

7541
  """
7542
  HPATH = "mirrors-replace"
7543
  HTYPE = constants.HTYPE_INSTANCE
7544
  _OP_PARAMS = [
7545
    _PInstanceName,
7546
    ("mode", _NoDefault, _TElemOf(constants.REPLACE_MODES)),
7547
    ("disks", _EmptyList, _TListOf(_TPositiveInt)),
7548
    ("remote_node", None, _TMaybeString),
7549
    ("iallocator", None, _TMaybeString),
7550
    ("early_release", False, _TBool),
7551
    ]
7552
  REQ_BGL = False
7553

    
7554
  def CheckArguments(self):
7555
    TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
7556
                                  self.op.iallocator)
7557

    
7558
  def ExpandNames(self):
7559
    self._ExpandAndLockInstance()
7560

    
7561
    if self.op.iallocator is not None:
7562
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7563

    
7564
    elif self.op.remote_node is not None:
7565
      remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
7566
      self.op.remote_node = remote_node
7567

    
7568
      # Warning: do not remove the locking of the new secondary here
7569
      # unless DRBD8.AddChildren is changed to work in parallel;
7570
      # currently it doesn't since parallel invocations of
7571
      # FindUnusedMinor will conflict
7572
      self.needed_locks[locking.LEVEL_NODE] = [remote_node]
7573
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
7574

    
7575
    else:
7576
      self.needed_locks[locking.LEVEL_NODE] = []
7577
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7578

    
7579
    self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
7580
                                   self.op.iallocator, self.op.remote_node,
7581
                                   self.op.disks, False, self.op.early_release)
7582

    
7583
    self.tasklets = [self.replacer]
7584

    
7585
  def DeclareLocks(self, level):
7586
    # If we're not already locking all nodes in the set we have to declare the
7587
    # instance's primary/secondary nodes.
7588
    if (level == locking.LEVEL_NODE and
7589
        self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
7590
      self._LockInstancesNodes()
7591

    
7592
  def BuildHooksEnv(self):
7593
    """Build hooks env.
7594

7595
    This runs on the master, the primary and all the secondaries.
7596

7597
    """
7598
    instance = self.replacer.instance
7599
    env = {
7600
      "MODE": self.op.mode,
7601
      "NEW_SECONDARY": self.op.remote_node,
7602
      "OLD_SECONDARY": instance.secondary_nodes[0],
7603
      }
7604
    env.update(_BuildInstanceHookEnvByObject(self, instance))
7605
    nl = [
7606
      self.cfg.GetMasterNode(),
7607
      instance.primary_node,
7608
      ]
7609
    if self.op.remote_node is not None:
7610
      nl.append(self.op.remote_node)
7611
    return env, nl, nl
7612

    
7613

    
7614
class TLReplaceDisks(Tasklet):
7615
  """Replaces disks for an instance.
7616

7617
  Note: Locking is not within the scope of this class.
7618

7619
  """
7620
  def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
7621
               disks, delay_iallocator, early_release):
7622
    """Initializes this class.
7623

7624
    """
7625
    Tasklet.__init__(self, lu)
7626

    
7627
    # Parameters
7628
    self.instance_name = instance_name
7629
    self.mode = mode
7630
    self.iallocator_name = iallocator_name
7631
    self.remote_node = remote_node
7632
    self.disks = disks
7633
    self.delay_iallocator = delay_iallocator
7634
    self.early_release = early_release
7635

    
7636
    # Runtime data
7637
    self.instance = None
7638
    self.new_node = None
7639
    self.target_node = None
7640
    self.other_node = None
7641
    self.remote_node_info = None
7642
    self.node_secondary_ip = None
7643

    
7644
  @staticmethod
7645
  def CheckArguments(mode, remote_node, iallocator):
7646
    """Helper function for users of this class.
7647

7648
    """
7649
    # check for valid parameter combination
7650
    if mode == constants.REPLACE_DISK_CHG:
7651
      if remote_node is None and iallocator is None:
7652
        raise errors.OpPrereqError("When changing the secondary either an"
7653
                                   " iallocator script must be used or the"
7654
                                   " new node given", errors.ECODE_INVAL)
7655

    
7656
      if remote_node is not None and iallocator is not None:
7657
        raise errors.OpPrereqError("Give either the iallocator or the new"
7658
                                   " secondary, not both", errors.ECODE_INVAL)
7659

    
7660
    elif remote_node is not None or iallocator is not None:
7661
      # Not replacing the secondary
7662
      raise errors.OpPrereqError("The iallocator and new node options can"
7663
                                 " only be used when changing the"
7664
                                 " secondary node", errors.ECODE_INVAL)
7665

    
7666
  @staticmethod
7667
  def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
7668
    """Compute a new secondary node using an IAllocator.
7669

7670
    """
7671
    ial = IAllocator(lu.cfg, lu.rpc,
7672
                     mode=constants.IALLOCATOR_MODE_RELOC,
7673
                     name=instance_name,
7674
                     relocate_from=relocate_from)
7675

    
7676
    ial.Run(iallocator_name)
7677

    
7678
    if not ial.success:
7679
      raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
7680
                                 " %s" % (iallocator_name, ial.info),
7681
                                 errors.ECODE_NORES)
7682

    
7683
    if len(ial.result) != ial.required_nodes:
7684
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
7685
                                 " of nodes (%s), required %s" %
7686
                                 (iallocator_name,
7687
                                  len(ial.result), ial.required_nodes),
7688
                                 errors.ECODE_FAULT)
7689

    
7690
    remote_node_name = ial.result[0]
7691

    
7692
    lu.LogInfo("Selected new secondary for instance '%s': %s",
7693
               instance_name, remote_node_name)
7694

    
7695
    return remote_node_name
7696

    
7697
  def _FindFaultyDisks(self, node_name):
7698
    return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
7699
                                    node_name, True)
7700

    
7701
  def CheckPrereq(self):
7702
    """Check prerequisites.
7703

7704
    This checks that the instance is in the cluster.
7705

7706
    """
7707
    self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
7708
    assert instance is not None, \
7709
      "Cannot retrieve locked instance %s" % self.instance_name
7710

    
7711
    if instance.disk_template != constants.DT_DRBD8:
7712
      raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
7713
                                 " instances", errors.ECODE_INVAL)
7714

    
7715
    if len(instance.secondary_nodes) != 1:
7716
      raise errors.OpPrereqError("The instance has a strange layout,"
7717
                                 " expected one secondary but found %d" %
7718
                                 len(instance.secondary_nodes),
7719
                                 errors.ECODE_FAULT)
7720

    
7721
    if not self.delay_iallocator:
7722
      self._CheckPrereq2()
7723

    
7724
  def _CheckPrereq2(self):
7725
    """Check prerequisites, second part.
7726

7727
    This function should always be part of CheckPrereq. It was separated and is
7728
    now called from Exec because during node evacuation iallocator was only
7729
    called with an unmodified cluster model, not taking planned changes into
7730
    account.
7731

7732
    """
7733
    instance = self.instance
7734
    secondary_node = instance.secondary_nodes[0]
7735

    
7736
    if self.iallocator_name is None:
7737
      remote_node = self.remote_node
7738
    else:
7739
      remote_node = self._RunAllocator(self.lu, self.iallocator_name,
7740
                                       instance.name, instance.secondary_nodes)
7741

    
7742
    if remote_node is not None:
7743
      self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
7744
      assert self.remote_node_info is not None, \
7745
        "Cannot retrieve locked node %s" % remote_node
7746
    else:
7747
      self.remote_node_info = None
7748

    
7749
    if remote_node == self.instance.primary_node:
7750
      raise errors.OpPrereqError("The specified node is the primary node of"
7751
                                 " the instance.", errors.ECODE_INVAL)
7752

    
7753
    if remote_node == secondary_node:
7754
      raise errors.OpPrereqError("The specified node is already the"
7755
                                 " secondary node of the instance.",
7756
                                 errors.ECODE_INVAL)
7757

    
7758
    if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
7759
                                    constants.REPLACE_DISK_CHG):
7760
      raise errors.OpPrereqError("Cannot specify disks to be replaced",
7761
                                 errors.ECODE_INVAL)
7762

    
7763
    if self.mode == constants.REPLACE_DISK_AUTO:
7764
      faulty_primary = self._FindFaultyDisks(instance.primary_node)
7765
      faulty_secondary = self._FindFaultyDisks(secondary_node)
7766

    
7767
      if faulty_primary and faulty_secondary:
7768
        raise errors.OpPrereqError("Instance %s has faulty disks on more than"
7769
                                   " one node and can not be repaired"
7770
                                   " automatically" % self.instance_name,
7771
                                   errors.ECODE_STATE)
7772

    
7773
      if faulty_primary:
7774
        self.disks = faulty_primary
7775
        self.target_node = instance.primary_node
7776
        self.other_node = secondary_node
7777
        check_nodes = [self.target_node, self.other_node]
7778
      elif faulty_secondary:
7779
        self.disks = faulty_secondary
7780
        self.target_node = secondary_node
7781
        self.other_node = instance.primary_node
7782
        check_nodes = [self.target_node, self.other_node]
7783
      else:
7784
        self.disks = []
7785
        check_nodes = []
7786

    
7787
    else:
7788
      # Non-automatic modes
7789
      if self.mode == constants.REPLACE_DISK_PRI:
7790
        self.target_node = instance.primary_node
7791
        self.other_node = secondary_node
7792
        check_nodes = [self.target_node, self.other_node]
7793

    
7794
      elif self.mode == constants.REPLACE_DISK_SEC:
7795
        self.target_node = secondary_node
7796
        self.other_node = instance.primary_node
7797
        check_nodes = [self.target_node, self.other_node]
7798

    
7799
      elif self.mode == constants.REPLACE_DISK_CHG:
7800
        self.new_node = remote_node
7801
        self.other_node = instance.primary_node
7802
        self.target_node = secondary_node
7803
        check_nodes = [self.new_node, self.other_node]
7804

    
7805
        _CheckNodeNotDrained(self.lu, remote_node)
7806

    
7807
        old_node_info = self.cfg.GetNodeInfo(secondary_node)
7808
        assert old_node_info is not None
7809
        if old_node_info.offline and not self.early_release:
7810
          # doesn't make sense to delay the release
7811
          self.early_release = True
7812
          self.lu.LogInfo("Old secondary %s is offline, automatically enabling"
7813
                          " early-release mode", secondary_node)
7814

    
7815
      else:
7816
        raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
7817
                                     self.mode)
7818

    
7819
      # If not specified all disks should be replaced
7820
      if not self.disks:
7821
        self.disks = range(len(self.instance.disks))
7822

    
7823
    for node in check_nodes:
7824
      _CheckNodeOnline(self.lu, node)
7825

    
7826
    # Check whether disks are valid
7827
    for disk_idx in self.disks:
7828
      instance.FindDisk(disk_idx)
7829

    
7830
    # Get secondary node IP addresses
7831
    node_2nd_ip = {}
7832

    
7833
    for node_name in [self.target_node, self.other_node, self.new_node]:
7834
      if node_name is not None:
7835
        node_2nd_ip[node_name] = self.cfg.GetNodeInfo(node_name).secondary_ip
7836

    
7837
    self.node_secondary_ip = node_2nd_ip
7838

    
7839
  def Exec(self, feedback_fn):
7840
    """Execute disk replacement.
7841

7842
    This dispatches the disk replacement to the appropriate handler.
7843

7844
    """
7845
    if self.delay_iallocator:
7846
      self._CheckPrereq2()
7847

    
7848
    if not self.disks:
7849
      feedback_fn("No disks need replacement")
7850
      return
7851

    
7852
    feedback_fn("Replacing disk(s) %s for %s" %
7853
                (utils.CommaJoin(self.disks), self.instance.name))
7854

    
7855
    activate_disks = (not self.instance.admin_up)
7856

    
7857
    # Activate the instance disks if we're replacing them on a down instance
7858
    if activate_disks:
7859
      _StartInstanceDisks(self.lu, self.instance, True)
7860

    
7861
    try:
7862
      # Should we replace the secondary node?
7863
      if self.new_node is not None:
7864
        fn = self._ExecDrbd8Secondary
7865
      else:
7866
        fn = self._ExecDrbd8DiskOnly
7867

    
7868
      return fn(feedback_fn)
7869

    
7870
    finally:
7871
      # Deactivate the instance disks if we're replacing them on a
7872
      # down instance
7873
      if activate_disks:
7874
        _SafeShutdownInstanceDisks(self.lu, self.instance)
7875

    
7876
  def _CheckVolumeGroup(self, nodes):
7877
    self.lu.LogInfo("Checking volume groups")
7878

    
7879
    vgname = self.cfg.GetVGName()
7880

    
7881
    # Make sure volume group exists on all involved nodes
7882
    results = self.rpc.call_vg_list(nodes)
7883
    if not results:
7884
      raise errors.OpExecError("Can't list volume groups on the nodes")
7885

    
7886
    for node in nodes:
7887
      res = results[node]
7888
      res.Raise("Error checking node %s" % node)
7889
      if vgname not in res.payload:
7890
        raise errors.OpExecError("Volume group '%s' not found on node %s" %
7891
                                 (vgname, node))
7892

    
7893
  def _CheckDisksExistence(self, nodes):
7894
    # Check disk existence
7895
    for idx, dev in enumerate(self.instance.disks):
7896
      if idx not in self.disks:
7897
        continue
7898

    
7899
      for node in nodes:
7900
        self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
7901
        self.cfg.SetDiskID(dev, node)
7902

    
7903
        result = self.rpc.call_blockdev_find(node, dev)
7904

    
7905
        msg = result.fail_msg
7906
        if msg or not result.payload:
7907
          if not msg:
7908
            msg = "disk not found"
7909
          raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
7910
                                   (idx, node, msg))
7911

    
7912
  def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
7913
    for idx, dev in enumerate(self.instance.disks):
7914
      if idx not in self.disks:
7915
        continue
7916

    
7917
      self.lu.LogInfo("Checking disk/%d consistency on node %s" %
7918
                      (idx, node_name))
7919

    
7920
      if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
7921
                                   ldisk=ldisk):
7922
        raise errors.OpExecError("Node %s has degraded storage, unsafe to"
7923
                                 " replace disks for instance %s" %
7924
                                 (node_name, self.instance.name))
7925

    
7926
  def _CreateNewStorage(self, node_name):
7927
    vgname = self.cfg.GetVGName()
7928
    iv_names = {}
7929

    
7930
    for idx, dev in enumerate(self.instance.disks):
7931
      if idx not in self.disks:
7932
        continue
7933

    
7934
      self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
7935

    
7936
      self.cfg.SetDiskID(dev, node_name)
7937

    
7938
      lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
7939
      names = _GenerateUniqueNames(self.lu, lv_names)
7940

    
7941
      lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
7942
                             logical_id=(vgname, names[0]))
7943
      lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
7944
                             logical_id=(vgname, names[1]))
7945

    
7946
      new_lvs = [lv_data, lv_meta]
7947
      old_lvs = dev.children
7948
      iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
7949

    
7950
      # we pass force_create=True to force the LVM creation
7951
      for new_lv in new_lvs:
7952
        _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
7953
                        _GetInstanceInfoText(self.instance), False)
7954

    
7955
    return iv_names
7956

    
7957
  def _CheckDevices(self, node_name, iv_names):
7958
    for name, (dev, _, _) in iv_names.iteritems():
7959
      self.cfg.SetDiskID(dev, node_name)
7960

    
7961
      result = self.rpc.call_blockdev_find(node_name, dev)
7962

    
7963
      msg = result.fail_msg
7964
      if msg or not result.payload:
7965
        if not msg:
7966
          msg = "disk not found"
7967
        raise errors.OpExecError("Can't find DRBD device %s: %s" %
7968
                                 (name, msg))
7969

    
7970
      if result.payload.is_degraded:
7971
        raise errors.OpExecError("DRBD device %s is degraded!" % name)
7972

    
7973
  def _RemoveOldStorage(self, node_name, iv_names):
7974
    for name, (_, old_lvs, _) in iv_names.iteritems():
7975
      self.lu.LogInfo("Remove logical volumes for %s" % name)
7976

    
7977
      for lv in old_lvs:
7978
        self.cfg.SetDiskID(lv, node_name)
7979

    
7980
        msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
7981
        if msg:
7982
          self.lu.LogWarning("Can't remove old LV: %s" % msg,
7983
                             hint="remove unused LVs manually")
7984

    
7985
  def _ReleaseNodeLock(self, node_name):
7986
    """Releases the lock for a given node."""
7987
    self.lu.context.glm.release(locking.LEVEL_NODE, node_name)
7988

    
7989
  def _ExecDrbd8DiskOnly(self, feedback_fn):
7990
    """Replace a disk on the primary or secondary for DRBD 8.
7991

7992
    The algorithm for replace is quite complicated:
7993

7994
      1. for each disk to be replaced:
7995

7996
        1. create new LVs on the target node with unique names
7997
        1. detach old LVs from the drbd device
7998
        1. rename old LVs to name_replaced.<time_t>
7999
        1. rename new LVs to old LVs
8000
        1. attach the new LVs (with the old names now) to the drbd device
8001

8002
      1. wait for sync across all devices
8003

8004
      1. for each modified disk:
8005

8006
        1. remove old LVs (which have the name name_replaces.<time_t>)
8007

8008
    Failures are not very well handled.
8009

8010
    """
8011
    steps_total = 6
8012

    
8013
    # Step: check device activation
8014
    self.lu.LogStep(1, steps_total, "Check device existence")
8015
    self._CheckDisksExistence([self.other_node, self.target_node])
8016
    self._CheckVolumeGroup([self.target_node, self.other_node])
8017

    
8018
    # Step: check other node consistency
8019
    self.lu.LogStep(2, steps_total, "Check peer consistency")
8020
    self._CheckDisksConsistency(self.other_node,
8021
                                self.other_node == self.instance.primary_node,
8022
                                False)
8023

    
8024
    # Step: create new storage
8025
    self.lu.LogStep(3, steps_total, "Allocate new storage")
8026
    iv_names = self._CreateNewStorage(self.target_node)
8027

    
8028
    # Step: for each lv, detach+rename*2+attach
8029
    self.lu.LogStep(4, steps_total, "Changing drbd configuration")
8030
    for dev, old_lvs, new_lvs in iv_names.itervalues():
8031
      self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
8032

    
8033
      result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
8034
                                                     old_lvs)
8035
      result.Raise("Can't detach drbd from local storage on node"
8036
                   " %s for device %s" % (self.target_node, dev.iv_name))
8037
      #dev.children = []
8038
      #cfg.Update(instance)
8039

    
8040
      # ok, we created the new LVs, so now we know we have the needed
8041
      # storage; as such, we proceed on the target node to rename
8042
      # old_lv to _old, and new_lv to old_lv; note that we rename LVs
8043
      # using the assumption that logical_id == physical_id (which in
8044
      # turn is the unique_id on that node)
8045

    
8046
      # FIXME(iustin): use a better name for the replaced LVs
8047
      temp_suffix = int(time.time())
8048
      ren_fn = lambda d, suff: (d.physical_id[0],
8049
                                d.physical_id[1] + "_replaced-%s" % suff)
8050

    
8051
      # Build the rename list based on what LVs exist on the node
8052
      rename_old_to_new = []
8053
      for to_ren in old_lvs:
8054
        result = self.rpc.call_blockdev_find(self.target_node, to_ren)
8055
        if not result.fail_msg and result.payload:
8056
          # device exists
8057
          rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
8058

    
8059
      self.lu.LogInfo("Renaming the old LVs on the target node")
8060
      result = self.rpc.call_blockdev_rename(self.target_node,
8061
                                             rename_old_to_new)
8062
      result.Raise("Can't rename old LVs on node %s" % self.target_node)
8063

    
8064
      # Now we rename the new LVs to the old LVs
8065
      self.lu.LogInfo("Renaming the new LVs on the target node")
8066
      rename_new_to_old = [(new, old.physical_id)
8067
                           for old, new in zip(old_lvs, new_lvs)]
8068
      result = self.rpc.call_blockdev_rename(self.target_node,
8069
                                             rename_new_to_old)
8070
      result.Raise("Can't rename new LVs on node %s" % self.target_node)
8071

    
8072
      for old, new in zip(old_lvs, new_lvs):
8073
        new.logical_id = old.logical_id
8074
        self.cfg.SetDiskID(new, self.target_node)
8075

    
8076
      for disk in old_lvs:
8077
        disk.logical_id = ren_fn(disk, temp_suffix)
8078
        self.cfg.SetDiskID(disk, self.target_node)
8079

    
8080
      # Now that the new lvs have the old name, we can add them to the device
8081
      self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
8082
      result = self.rpc.call_blockdev_addchildren(self.target_node, dev,
8083
                                                  new_lvs)
8084
      msg = result.fail_msg
8085
      if msg:
8086
        for new_lv in new_lvs:
8087
          msg2 = self.rpc.call_blockdev_remove(self.target_node,
8088
                                               new_lv).fail_msg
8089
          if msg2:
8090
            self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
8091
                               hint=("cleanup manually the unused logical"
8092
                                     "volumes"))
8093
        raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
8094

    
8095
      dev.children = new_lvs
8096

    
8097
      self.cfg.Update(self.instance, feedback_fn)
8098

    
8099
    cstep = 5
8100
    if self.early_release:
8101
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
8102
      cstep += 1
8103
      self._RemoveOldStorage(self.target_node, iv_names)
8104
      # WARNING: we release both node locks here, do not do other RPCs
8105
      # than WaitForSync to the primary node
8106
      self._ReleaseNodeLock([self.target_node, self.other_node])
8107

    
8108
    # Wait for sync
8109
    # This can fail as the old devices are degraded and _WaitForSync
8110
    # does a combined result over all disks, so we don't check its return value
8111
    self.lu.LogStep(cstep, steps_total, "Sync devices")
8112
    cstep += 1
8113
    _WaitForSync(self.lu, self.instance)
8114

    
8115
    # Check all devices manually
8116
    self._CheckDevices(self.instance.primary_node, iv_names)
8117

    
8118
    # Step: remove old storage
8119
    if not self.early_release:
8120
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
8121
      cstep += 1
8122
      self._RemoveOldStorage(self.target_node, iv_names)
8123

    
8124
  def _ExecDrbd8Secondary(self, feedback_fn):
8125
    """Replace the secondary node for DRBD 8.
8126

8127
    The algorithm for replace is quite complicated:
8128
      - for all disks of the instance:
8129
        - create new LVs on the new node with same names
8130
        - shutdown the drbd device on the old secondary
8131
        - disconnect the drbd network on the primary
8132
        - create the drbd device on the new secondary
8133
        - network attach the drbd on the primary, using an artifice:
8134
          the drbd code for Attach() will connect to the network if it
8135
          finds a device which is connected to the good local disks but
8136
          not network enabled
8137
      - wait for sync across all devices
8138
      - remove all disks from the old secondary
8139

8140
    Failures are not very well handled.
8141

8142
    """
8143
    steps_total = 6
8144

    
8145
    # Step: check device activation
8146
    self.lu.LogStep(1, steps_total, "Check device existence")
8147
    self._CheckDisksExistence([self.instance.primary_node])
8148
    self._CheckVolumeGroup([self.instance.primary_node])
8149

    
8150
    # Step: check other node consistency
8151
    self.lu.LogStep(2, steps_total, "Check peer consistency")
8152
    self._CheckDisksConsistency(self.instance.primary_node, True, True)
8153

    
8154
    # Step: create new storage
8155
    self.lu.LogStep(3, steps_total, "Allocate new storage")
8156
    for idx, dev in enumerate(self.instance.disks):
8157
      self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
8158
                      (self.new_node, idx))
8159
      # we pass force_create=True to force LVM creation
8160
      for new_lv in dev.children:
8161
        _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
8162
                        _GetInstanceInfoText(self.instance), False)
8163

    
8164
    # Step 4: dbrd minors and drbd setups changes
8165
    # after this, we must manually remove the drbd minors on both the
8166
    # error and the success paths
8167
    self.lu.LogStep(4, steps_total, "Changing drbd configuration")
8168
    minors = self.cfg.AllocateDRBDMinor([self.new_node
8169
                                         for dev in self.instance.disks],
8170
                                        self.instance.name)
8171
    logging.debug("Allocated minors %r", minors)
8172

    
8173
    iv_names = {}
8174
    for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
8175
      self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
8176
                      (self.new_node, idx))
8177
      # create new devices on new_node; note that we create two IDs:
8178
      # one without port, so the drbd will be activated without
8179
      # networking information on the new node at this stage, and one
8180
      # with network, for the latter activation in step 4
8181
      (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
8182
      if self.instance.primary_node == o_node1:
8183
        p_minor = o_minor1
8184
      else:
8185
        assert self.instance.primary_node == o_node2, "Three-node instance?"
8186
        p_minor = o_minor2
8187

    
8188
      new_alone_id = (self.instance.primary_node, self.new_node, None,
8189
                      p_minor, new_minor, o_secret)
8190
      new_net_id = (self.instance.primary_node, self.new_node, o_port,
8191
                    p_minor, new_minor, o_secret)
8192

    
8193
      iv_names[idx] = (dev, dev.children, new_net_id)
8194
      logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
8195
                    new_net_id)
8196
      new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
8197
                              logical_id=new_alone_id,
8198
                              children=dev.children,
8199
                              size=dev.size)
8200
      try:
8201
        _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
8202
                              _GetInstanceInfoText(self.instance), False)
8203
      except errors.GenericError:
8204
        self.cfg.ReleaseDRBDMinors(self.instance.name)
8205
        raise
8206

    
8207
    # We have new devices, shutdown the drbd on the old secondary
8208
    for idx, dev in enumerate(self.instance.disks):
8209
      self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
8210
      self.cfg.SetDiskID(dev, self.target_node)
8211
      msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
8212
      if msg:
8213
        self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
8214
                           "node: %s" % (idx, msg),
8215
                           hint=("Please cleanup this device manually as"
8216
                                 " soon as possible"))
8217

    
8218
    self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
8219
    result = self.rpc.call_drbd_disconnect_net([self.instance.primary_node],
8220
                                               self.node_secondary_ip,
8221
                                               self.instance.disks)\
8222
                                              [self.instance.primary_node]
8223

    
8224
    msg = result.fail_msg
8225
    if msg:
8226
      # detaches didn't succeed (unlikely)
8227
      self.cfg.ReleaseDRBDMinors(self.instance.name)
8228
      raise errors.OpExecError("Can't detach the disks from the network on"
8229
                               " old node: %s" % (msg,))
8230

    
8231
    # if we managed to detach at least one, we update all the disks of
8232
    # the instance to point to the new secondary
8233
    self.lu.LogInfo("Updating instance configuration")
8234
    for dev, _, new_logical_id in iv_names.itervalues():
8235
      dev.logical_id = new_logical_id
8236
      self.cfg.SetDiskID(dev, self.instance.primary_node)
8237

    
8238
    self.cfg.Update(self.instance, feedback_fn)
8239

    
8240
    # and now perform the drbd attach
8241
    self.lu.LogInfo("Attaching primary drbds to new secondary"
8242
                    " (standalone => connected)")
8243
    result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
8244
                                            self.new_node],
8245
                                           self.node_secondary_ip,
8246
                                           self.instance.disks,
8247
                                           self.instance.name,
8248
                                           False)
8249
    for to_node, to_result in result.items():
8250
      msg = to_result.fail_msg
8251
      if msg:
8252
        self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
8253
                           to_node, msg,
8254
                           hint=("please do a gnt-instance info to see the"
8255
                                 " status of disks"))
8256
    cstep = 5
8257
    if self.early_release:
8258
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
8259
      cstep += 1
8260
      self._RemoveOldStorage(self.target_node, iv_names)
8261
      # WARNING: we release all node locks here, do not do other RPCs
8262
      # than WaitForSync to the primary node
8263
      self._ReleaseNodeLock([self.instance.primary_node,
8264
                             self.target_node,
8265
                             self.new_node])
8266

    
8267
    # Wait for sync
8268
    # This can fail as the old devices are degraded and _WaitForSync
8269
    # does a combined result over all disks, so we don't check its return value
8270
    self.lu.LogStep(cstep, steps_total, "Sync devices")
8271
    cstep += 1
8272
    _WaitForSync(self.lu, self.instance)
8273

    
8274
    # Check all devices manually
8275
    self._CheckDevices(self.instance.primary_node, iv_names)
8276

    
8277
    # Step: remove old storage
8278
    if not self.early_release:
8279
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
8280
      self._RemoveOldStorage(self.target_node, iv_names)
8281

    
8282

    
8283
class LURepairNodeStorage(NoHooksLU):
8284
  """Repairs the volume group on a node.
8285

8286
  """
8287
  _OP_PARAMS = [
8288
    _PNodeName,
8289
    ("storage_type", _NoDefault, _CheckStorageType),
8290
    ("name", _NoDefault, _TNonEmptyString),
8291
    ("ignore_consistency", False, _TBool),
8292
    ]
8293
  REQ_BGL = False
8294

    
8295
  def CheckArguments(self):
8296
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
8297

    
8298
    storage_type = self.op.storage_type
8299

    
8300
    if (constants.SO_FIX_CONSISTENCY not in
8301
        constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
8302
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
8303
                                 " repaired" % storage_type,
8304
                                 errors.ECODE_INVAL)
8305

    
8306
  def ExpandNames(self):
8307
    self.needed_locks = {
8308
      locking.LEVEL_NODE: [self.op.node_name],
8309
      }
8310

    
8311
  def _CheckFaultyDisks(self, instance, node_name):
8312
    """Ensure faulty disks abort the opcode or at least warn."""
8313
    try:
8314
      if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
8315
                                  node_name, True):
8316
        raise errors.OpPrereqError("Instance '%s' has faulty disks on"
8317
                                   " node '%s'" % (instance.name, node_name),
8318
                                   errors.ECODE_STATE)
8319
    except errors.OpPrereqError, err:
8320
      if self.op.ignore_consistency:
8321
        self.proc.LogWarning(str(err.args[0]))
8322
      else:
8323
        raise
8324

    
8325
  def CheckPrereq(self):
8326
    """Check prerequisites.
8327

8328
    """
8329
    # Check whether any instance on this node has faulty disks
8330
    for inst in _GetNodeInstances(self.cfg, self.op.node_name):
8331
      if not inst.admin_up:
8332
        continue
8333
      check_nodes = set(inst.all_nodes)
8334
      check_nodes.discard(self.op.node_name)
8335
      for inst_node_name in check_nodes:
8336
        self._CheckFaultyDisks(inst, inst_node_name)
8337

    
8338
  def Exec(self, feedback_fn):
8339
    feedback_fn("Repairing storage unit '%s' on %s ..." %
8340
                (self.op.name, self.op.node_name))
8341

    
8342
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
8343
    result = self.rpc.call_storage_execute(self.op.node_name,
8344
                                           self.op.storage_type, st_args,
8345
                                           self.op.name,
8346
                                           constants.SO_FIX_CONSISTENCY)
8347
    result.Raise("Failed to repair storage unit '%s' on %s" %
8348
                 (self.op.name, self.op.node_name))
8349

    
8350

    
8351
class LUNodeEvacuationStrategy(NoHooksLU):
8352
  """Computes the node evacuation strategy.
8353

8354
  """
8355
  _OP_PARAMS = [
8356
    ("nodes", _NoDefault, _TListOf(_TNonEmptyString)),
8357
    ("remote_node", None, _TMaybeString),
8358
    ("iallocator", None, _TMaybeString),
8359
    ]
8360
  REQ_BGL = False
8361

    
8362
  def CheckArguments(self):
8363
    _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
8364

    
8365
  def ExpandNames(self):
8366
    self.op.nodes = _GetWantedNodes(self, self.op.nodes)
8367
    self.needed_locks = locks = {}
8368
    if self.op.remote_node is None:
8369
      locks[locking.LEVEL_NODE] = locking.ALL_SET
8370
    else:
8371
      self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8372
      locks[locking.LEVEL_NODE] = self.op.nodes + [self.op.remote_node]
8373

    
8374
  def Exec(self, feedback_fn):
8375
    if self.op.remote_node is not None:
8376
      instances = []
8377
      for node in self.op.nodes:
8378
        instances.extend(_GetNodeSecondaryInstances(self.cfg, node))
8379
      result = []
8380
      for i in instances:
8381
        if i.primary_node == self.op.remote_node:
8382
          raise errors.OpPrereqError("Node %s is the primary node of"
8383
                                     " instance %s, cannot use it as"
8384
                                     " secondary" %
8385
                                     (self.op.remote_node, i.name),
8386
                                     errors.ECODE_INVAL)
8387
        result.append([i.name, self.op.remote_node])
8388
    else:
8389
      ial = IAllocator(self.cfg, self.rpc,
8390
                       mode=constants.IALLOCATOR_MODE_MEVAC,
8391
                       evac_nodes=self.op.nodes)
8392
      ial.Run(self.op.iallocator, validate=True)
8393
      if not ial.success:
8394
        raise errors.OpExecError("No valid evacuation solution: %s" % ial.info,
8395
                                 errors.ECODE_NORES)
8396
      result = ial.result
8397
    return result
8398

    
8399

    
8400
class LUGrowDisk(LogicalUnit):
8401
  """Grow a disk of an instance.
8402

8403
  """
8404
  HPATH = "disk-grow"
8405
  HTYPE = constants.HTYPE_INSTANCE
8406
  _OP_PARAMS = [
8407
    _PInstanceName,
8408
    ("disk", _NoDefault, _TInt),
8409
    ("amount", _NoDefault, _TInt),
8410
    ("wait_for_sync", True, _TBool),
8411
    ]
8412
  REQ_BGL = False
8413

    
8414
  def ExpandNames(self):
8415
    self._ExpandAndLockInstance()
8416
    self.needed_locks[locking.LEVEL_NODE] = []
8417
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8418

    
8419
  def DeclareLocks(self, level):
8420
    if level == locking.LEVEL_NODE:
8421
      self._LockInstancesNodes()
8422

    
8423
  def BuildHooksEnv(self):
8424
    """Build hooks env.
8425

8426
    This runs on the master, the primary and all the secondaries.
8427

8428
    """
8429
    env = {
8430
      "DISK": self.op.disk,
8431
      "AMOUNT": self.op.amount,
8432
      }
8433
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
8434
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
8435
    return env, nl, nl
8436

    
8437
  def CheckPrereq(self):
8438
    """Check prerequisites.
8439

8440
    This checks that the instance is in the cluster.
8441

8442
    """
8443
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
8444
    assert instance is not None, \
8445
      "Cannot retrieve locked instance %s" % self.op.instance_name
8446
    nodenames = list(instance.all_nodes)
8447
    for node in nodenames:
8448
      _CheckNodeOnline(self, node)
8449

    
8450
    self.instance = instance
8451

    
8452
    if instance.disk_template not in constants.DTS_GROWABLE:
8453
      raise errors.OpPrereqError("Instance's disk layout does not support"
8454
                                 " growing.", errors.ECODE_INVAL)
8455

    
8456
    self.disk = instance.FindDisk(self.op.disk)
8457

    
8458
    if instance.disk_template != constants.DT_FILE:
8459
      # TODO: check the free disk space for file, when that feature will be
8460
      # supported
8461
      _CheckNodesFreeDisk(self, nodenames, self.op.amount)
8462

    
8463
  def Exec(self, feedback_fn):
8464
    """Execute disk grow.
8465

8466
    """
8467
    instance = self.instance
8468
    disk = self.disk
8469

    
8470
    disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk])
8471
    if not disks_ok:
8472
      raise errors.OpExecError("Cannot activate block device to grow")
8473

    
8474
    for node in instance.all_nodes:
8475
      self.cfg.SetDiskID(disk, node)
8476
      result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
8477
      result.Raise("Grow request failed to node %s" % node)
8478

    
8479
      # TODO: Rewrite code to work properly
8480
      # DRBD goes into sync mode for a short amount of time after executing the
8481
      # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
8482
      # calling "resize" in sync mode fails. Sleeping for a short amount of
8483
      # time is a work-around.
8484
      time.sleep(5)
8485

    
8486
    disk.RecordGrow(self.op.amount)
8487
    self.cfg.Update(instance, feedback_fn)
8488
    if self.op.wait_for_sync:
8489
      disk_abort = not _WaitForSync(self, instance, disks=[disk])
8490
      if disk_abort:
8491
        self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
8492
                             " status.\nPlease check the instance.")
8493
      if not instance.admin_up:
8494
        _SafeShutdownInstanceDisks(self, instance, disks=[disk])
8495
    elif not instance.admin_up:
8496
      self.proc.LogWarning("Not shutting down the disk even if the instance is"
8497
                           " not supposed to be running because no wait for"
8498
                           " sync mode was requested.")
8499

    
8500

    
8501
class LUQueryInstanceData(NoHooksLU):
8502
  """Query runtime instance data.
8503

8504
  """
8505
  _OP_PARAMS = [
8506
    ("instances", _EmptyList, _TListOf(_TNonEmptyString)),
8507
    ("static", False, _TBool),
8508
    ]
8509
  REQ_BGL = False
8510

    
8511
  def ExpandNames(self):
8512
    self.needed_locks = {}
8513
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
8514

    
8515
    if self.op.instances:
8516
      self.wanted_names = []
8517
      for name in self.op.instances:
8518
        full_name = _ExpandInstanceName(self.cfg, name)
8519
        self.wanted_names.append(full_name)
8520
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
8521
    else:
8522
      self.wanted_names = None
8523
      self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
8524

    
8525
    self.needed_locks[locking.LEVEL_NODE] = []
8526
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8527

    
8528
  def DeclareLocks(self, level):
8529
    if level == locking.LEVEL_NODE:
8530
      self._LockInstancesNodes()
8531

    
8532
  def CheckPrereq(self):
8533
    """Check prerequisites.
8534

8535
    This only checks the optional instance list against the existing names.
8536

8537
    """
8538
    if self.wanted_names is None:
8539
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
8540

    
8541
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
8542
                             in self.wanted_names]
8543

    
8544
  def _ComputeBlockdevStatus(self, node, instance_name, dev):
8545
    """Returns the status of a block device
8546

8547
    """
8548
    if self.op.static or not node:
8549
      return None
8550

    
8551
    self.cfg.SetDiskID(dev, node)
8552

    
8553
    result = self.rpc.call_blockdev_find(node, dev)
8554
    if result.offline:
8555
      return None
8556

    
8557
    result.Raise("Can't compute disk status for %s" % instance_name)
8558

    
8559
    status = result.payload
8560
    if status is None:
8561
      return None
8562

    
8563
    return (status.dev_path, status.major, status.minor,
8564
            status.sync_percent, status.estimated_time,
8565
            status.is_degraded, status.ldisk_status)
8566

    
8567
  def _ComputeDiskStatus(self, instance, snode, dev):
8568
    """Compute block device status.
8569

8570
    """
8571
    if dev.dev_type in constants.LDS_DRBD:
8572
      # we change the snode then (otherwise we use the one passed in)
8573
      if dev.logical_id[0] == instance.primary_node:
8574
        snode = dev.logical_id[1]
8575
      else:
8576
        snode = dev.logical_id[0]
8577

    
8578
    dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
8579
                                              instance.name, dev)
8580
    dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
8581

    
8582
    if dev.children:
8583
      dev_children = [self._ComputeDiskStatus(instance, snode, child)
8584
                      for child in dev.children]
8585
    else:
8586
      dev_children = []
8587

    
8588
    data = {
8589
      "iv_name": dev.iv_name,
8590
      "dev_type": dev.dev_type,
8591
      "logical_id": dev.logical_id,
8592
      "physical_id": dev.physical_id,
8593
      "pstatus": dev_pstatus,
8594
      "sstatus": dev_sstatus,
8595
      "children": dev_children,
8596
      "mode": dev.mode,
8597
      "size": dev.size,
8598
      }
8599

    
8600
    return data
8601

    
8602
  def Exec(self, feedback_fn):
8603
    """Gather and return data"""
8604
    result = {}
8605

    
8606
    cluster = self.cfg.GetClusterInfo()
8607

    
8608
    for instance in self.wanted_instances:
8609
      if not self.op.static:
8610
        remote_info = self.rpc.call_instance_info(instance.primary_node,
8611
                                                  instance.name,
8612
                                                  instance.hypervisor)
8613
        remote_info.Raise("Error checking node %s" % instance.primary_node)
8614
        remote_info = remote_info.payload
8615
        if remote_info and "state" in remote_info:
8616
          remote_state = "up"
8617
        else:
8618
          remote_state = "down"
8619
      else:
8620
        remote_state = None
8621
      if instance.admin_up:
8622
        config_state = "up"
8623
      else:
8624
        config_state = "down"
8625

    
8626
      disks = [self._ComputeDiskStatus(instance, None, device)
8627
               for device in instance.disks]
8628

    
8629
      idict = {
8630
        "name": instance.name,
8631
        "config_state": config_state,
8632
        "run_state": remote_state,
8633
        "pnode": instance.primary_node,
8634
        "snodes": instance.secondary_nodes,
8635
        "os": instance.os,
8636
        # this happens to be the same format used for hooks
8637
        "nics": _NICListToTuple(self, instance.nics),
8638
        "disk_template": instance.disk_template,
8639
        "disks": disks,
8640
        "hypervisor": instance.hypervisor,
8641
        "network_port": instance.network_port,
8642
        "hv_instance": instance.hvparams,
8643
        "hv_actual": cluster.FillHV(instance, skip_globals=True),
8644
        "be_instance": instance.beparams,
8645
        "be_actual": cluster.FillBE(instance),
8646
        "os_instance": instance.osparams,
8647
        "os_actual": cluster.SimpleFillOS(instance.os, instance.osparams),
8648
        "serial_no": instance.serial_no,
8649
        "mtime": instance.mtime,
8650
        "ctime": instance.ctime,
8651
        "uuid": instance.uuid,
8652
        }
8653

    
8654
      result[instance.name] = idict
8655

    
8656
    return result
8657

    
8658

    
8659
class LUSetInstanceParams(LogicalUnit):
8660
  """Modifies an instances's parameters.
8661

8662
  """
8663
  HPATH = "instance-modify"
8664
  HTYPE = constants.HTYPE_INSTANCE
8665
  _OP_PARAMS = [
8666
    _PInstanceName,
8667
    ("nics", _EmptyList, _TList),
8668
    ("disks", _EmptyList, _TList),
8669
    ("beparams", _EmptyDict, _TDict),
8670
    ("hvparams", _EmptyDict, _TDict),
8671
    ("disk_template", None, _TMaybeString),
8672
    ("remote_node", None, _TMaybeString),
8673
    ("os_name", None, _TMaybeString),
8674
    ("force_variant", False, _TBool),
8675
    ("osparams", None, _TOr(_TDict, _TNone)),
8676
    _PForce,
8677
    ]
8678
  REQ_BGL = False
8679

    
8680
  def CheckArguments(self):
8681
    if not (self.op.nics or self.op.disks or self.op.disk_template or
8682
            self.op.hvparams or self.op.beparams or self.op.os_name):
8683
      raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
8684

    
8685
    if self.op.hvparams:
8686
      _CheckGlobalHvParams(self.op.hvparams)
8687

    
8688
    # Disk validation
8689
    disk_addremove = 0
8690
    for disk_op, disk_dict in self.op.disks:
8691
      utils.ForceDictType(disk_dict, constants.IDISK_PARAMS_TYPES)
8692
      if disk_op == constants.DDM_REMOVE:
8693
        disk_addremove += 1
8694
        continue
8695
      elif disk_op == constants.DDM_ADD:
8696
        disk_addremove += 1
8697
      else:
8698
        if not isinstance(disk_op, int):
8699
          raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL)
8700
        if not isinstance(disk_dict, dict):
8701
          msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
8702
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
8703

    
8704
      if disk_op == constants.DDM_ADD:
8705
        mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
8706
        if mode not in constants.DISK_ACCESS_SET:
8707
          raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
8708
                                     errors.ECODE_INVAL)
8709
        size = disk_dict.get('size', None)
8710
        if size is None:
8711
          raise errors.OpPrereqError("Required disk parameter size missing",
8712
                                     errors.ECODE_INVAL)
8713
        try:
8714
          size = int(size)
8715
        except (TypeError, ValueError), err:
8716
          raise errors.OpPrereqError("Invalid disk size parameter: %s" %
8717
                                     str(err), errors.ECODE_INVAL)
8718
        disk_dict['size'] = size
8719
      else:
8720
        # modification of disk
8721
        if 'size' in disk_dict:
8722
          raise errors.OpPrereqError("Disk size change not possible, use"
8723
                                     " grow-disk", errors.ECODE_INVAL)
8724

    
8725
    if disk_addremove > 1:
8726
      raise errors.OpPrereqError("Only one disk add or remove operation"
8727
                                 " supported at a time", errors.ECODE_INVAL)
8728

    
8729
    if self.op.disks and self.op.disk_template is not None:
8730
      raise errors.OpPrereqError("Disk template conversion and other disk"
8731
                                 " changes not supported at the same time",
8732
                                 errors.ECODE_INVAL)
8733

    
8734
    if self.op.disk_template:
8735
      _CheckDiskTemplate(self.op.disk_template)
8736
      if (self.op.disk_template in constants.DTS_NET_MIRROR and
8737
          self.op.remote_node is None):
8738
        raise errors.OpPrereqError("Changing the disk template to a mirrored"
8739
                                   " one requires specifying a secondary node",
8740
                                   errors.ECODE_INVAL)
8741

    
8742
    # NIC validation
8743
    nic_addremove = 0
8744
    for nic_op, nic_dict in self.op.nics:
8745
      utils.ForceDictType(nic_dict, constants.INIC_PARAMS_TYPES)
8746
      if nic_op == constants.DDM_REMOVE:
8747
        nic_addremove += 1
8748
        continue
8749
      elif nic_op == constants.DDM_ADD:
8750
        nic_addremove += 1
8751
      else:
8752
        if not isinstance(nic_op, int):
8753
          raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL)
8754
        if not isinstance(nic_dict, dict):
8755
          msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
8756
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
8757

    
8758
      # nic_dict should be a dict
8759
      nic_ip = nic_dict.get('ip', None)
8760
      if nic_ip is not None:
8761
        if nic_ip.lower() == constants.VALUE_NONE:
8762
          nic_dict['ip'] = None
8763
        else:
8764
          if not netutils.IsValidIP4(nic_ip):
8765
            raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip,
8766
                                       errors.ECODE_INVAL)
8767

    
8768
      nic_bridge = nic_dict.get('bridge', None)
8769
      nic_link = nic_dict.get('link', None)
8770
      if nic_bridge and nic_link:
8771
        raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
8772
                                   " at the same time", errors.ECODE_INVAL)
8773
      elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
8774
        nic_dict['bridge'] = None
8775
      elif nic_link and nic_link.lower() == constants.VALUE_NONE:
8776
        nic_dict['link'] = None
8777

    
8778
      if nic_op == constants.DDM_ADD:
8779
        nic_mac = nic_dict.get('mac', None)
8780
        if nic_mac is None:
8781
          nic_dict['mac'] = constants.VALUE_AUTO
8782

    
8783
      if 'mac' in nic_dict:
8784
        nic_mac = nic_dict['mac']
8785
        if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
8786
          nic_mac = utils.NormalizeAndValidateMac(nic_mac)
8787

    
8788
        if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
8789
          raise errors.OpPrereqError("'auto' is not a valid MAC address when"
8790
                                     " modifying an existing nic",
8791
                                     errors.ECODE_INVAL)
8792

    
8793
    if nic_addremove > 1:
8794
      raise errors.OpPrereqError("Only one NIC add or remove operation"
8795
                                 " supported at a time", errors.ECODE_INVAL)
8796

    
8797
  def ExpandNames(self):
8798
    self._ExpandAndLockInstance()
8799
    self.needed_locks[locking.LEVEL_NODE] = []
8800
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8801

    
8802
  def DeclareLocks(self, level):
8803
    if level == locking.LEVEL_NODE:
8804
      self._LockInstancesNodes()
8805
      if self.op.disk_template and self.op.remote_node:
8806
        self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8807
        self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)
8808

    
8809
  def BuildHooksEnv(self):
8810
    """Build hooks env.
8811

8812
    This runs on the master, primary and secondaries.
8813

8814
    """
8815
    args = dict()
8816
    if constants.BE_MEMORY in self.be_new:
8817
      args['memory'] = self.be_new[constants.BE_MEMORY]
8818
    if constants.BE_VCPUS in self.be_new:
8819
      args['vcpus'] = self.be_new[constants.BE_VCPUS]
8820
    # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
8821
    # information at all.
8822
    if self.op.nics:
8823
      args['nics'] = []
8824
      nic_override = dict(self.op.nics)
8825
      for idx, nic in enumerate(self.instance.nics):
8826
        if idx in nic_override:
8827
          this_nic_override = nic_override[idx]
8828
        else:
8829
          this_nic_override = {}
8830
        if 'ip' in this_nic_override:
8831
          ip = this_nic_override['ip']
8832
        else:
8833
          ip = nic.ip
8834
        if 'mac' in this_nic_override:
8835
          mac = this_nic_override['mac']
8836
        else:
8837
          mac = nic.mac
8838
        if idx in self.nic_pnew:
8839
          nicparams = self.nic_pnew[idx]
8840
        else:
8841
          nicparams = self.cluster.SimpleFillNIC(nic.nicparams)
8842
        mode = nicparams[constants.NIC_MODE]
8843
        link = nicparams[constants.NIC_LINK]
8844
        args['nics'].append((ip, mac, mode, link))
8845
      if constants.DDM_ADD in nic_override:
8846
        ip = nic_override[constants.DDM_ADD].get('ip', None)
8847
        mac = nic_override[constants.DDM_ADD]['mac']
8848
        nicparams = self.nic_pnew[constants.DDM_ADD]
8849
        mode = nicparams[constants.NIC_MODE]
8850
        link = nicparams[constants.NIC_LINK]
8851
        args['nics'].append((ip, mac, mode, link))
8852
      elif constants.DDM_REMOVE in nic_override:
8853
        del args['nics'][-1]
8854

    
8855
    env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
8856
    if self.op.disk_template:
8857
      env["NEW_DISK_TEMPLATE"] = self.op.disk_template
8858
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
8859
    return env, nl, nl
8860

    
8861
  def CheckPrereq(self):
8862
    """Check prerequisites.
8863

8864
    This only checks the instance list against the existing names.
8865

8866
    """
8867
    # checking the new params on the primary/secondary nodes
8868

    
8869
    instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
8870
    cluster = self.cluster = self.cfg.GetClusterInfo()
8871
    assert self.instance is not None, \
8872
      "Cannot retrieve locked instance %s" % self.op.instance_name
8873
    pnode = instance.primary_node
8874
    nodelist = list(instance.all_nodes)
8875

    
8876
    # OS change
8877
    if self.op.os_name and not self.op.force:
8878
      _CheckNodeHasOS(self, instance.primary_node, self.op.os_name,
8879
                      self.op.force_variant)
8880
      instance_os = self.op.os_name
8881
    else:
8882
      instance_os = instance.os
8883

    
8884
    if self.op.disk_template:
8885
      if instance.disk_template == self.op.disk_template:
8886
        raise errors.OpPrereqError("Instance already has disk template %s" %
8887
                                   instance.disk_template, errors.ECODE_INVAL)
8888

    
8889
      if (instance.disk_template,
8890
          self.op.disk_template) not in self._DISK_CONVERSIONS:
8891
        raise errors.OpPrereqError("Unsupported disk template conversion from"
8892
                                   " %s to %s" % (instance.disk_template,
8893
                                                  self.op.disk_template),
8894
                                   errors.ECODE_INVAL)
8895
      _CheckInstanceDown(self, instance, "cannot change disk template")
8896
      if self.op.disk_template in constants.DTS_NET_MIRROR:
8897
        if self.op.remote_node == pnode:
8898
          raise errors.OpPrereqError("Given new secondary node %s is the same"
8899
                                     " as the primary node of the instance" %
8900
                                     self.op.remote_node, errors.ECODE_STATE)
8901
        _CheckNodeOnline(self, self.op.remote_node)
8902
        _CheckNodeNotDrained(self, self.op.remote_node)
8903
        disks = [{"size": d.size} for d in instance.disks]
8904
        required = _ComputeDiskSize(self.op.disk_template, disks)
8905
        _CheckNodesFreeDisk(self, [self.op.remote_node], required)
8906

    
8907
    # hvparams processing
8908
    if self.op.hvparams:
8909
      hv_type = instance.hypervisor
8910
      i_hvdict = _GetUpdatedParams(instance.hvparams, self.op.hvparams)
8911
      utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES)
8912
      hv_new = cluster.SimpleFillHV(hv_type, instance.os, i_hvdict)
8913

    
8914
      # local check
8915
      hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new)
8916
      _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
8917
      self.hv_new = hv_new # the new actual values
8918
      self.hv_inst = i_hvdict # the new dict (without defaults)
8919
    else:
8920
      self.hv_new = self.hv_inst = {}
8921

    
8922
    # beparams processing
8923
    if self.op.beparams:
8924
      i_bedict = _GetUpdatedParams(instance.beparams, self.op.beparams,
8925
                                   use_none=True)
8926
      utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
8927
      be_new = cluster.SimpleFillBE(i_bedict)
8928
      self.be_new = be_new # the new actual values
8929
      self.be_inst = i_bedict # the new dict (without defaults)
8930
    else:
8931
      self.be_new = self.be_inst = {}
8932

    
8933
    # osparams processing
8934
    if self.op.osparams:
8935
      i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
8936
      _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
8937
      self.os_new = cluster.SimpleFillOS(instance_os, i_osdict)
8938
      self.os_inst = i_osdict # the new dict (without defaults)
8939
    else:
8940
      self.os_new = self.os_inst = {}
8941

    
8942
    self.warn = []
8943

    
8944
    if constants.BE_MEMORY in self.op.beparams and not self.op.force:
8945
      mem_check_list = [pnode]
8946
      if be_new[constants.BE_AUTO_BALANCE]:
8947
        # either we changed auto_balance to yes or it was from before
8948
        mem_check_list.extend(instance.secondary_nodes)
8949
      instance_info = self.rpc.call_instance_info(pnode, instance.name,
8950
                                                  instance.hypervisor)
8951
      nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
8952
                                         instance.hypervisor)
8953
      pninfo = nodeinfo[pnode]
8954
      msg = pninfo.fail_msg
8955
      if msg:
8956
        # Assume the primary node is unreachable and go ahead
8957
        self.warn.append("Can't get info from primary node %s: %s" %
8958
                         (pnode,  msg))
8959
      elif not isinstance(pninfo.payload.get('memory_free', None), int):
8960
        self.warn.append("Node data from primary node %s doesn't contain"
8961
                         " free memory information" % pnode)
8962
      elif instance_info.fail_msg:
8963
        self.warn.append("Can't get instance runtime information: %s" %
8964
                        instance_info.fail_msg)
8965
      else:
8966
        if instance_info.payload:
8967
          current_mem = int(instance_info.payload['memory'])
8968
        else:
8969
          # Assume instance not running
8970
          # (there is a slight race condition here, but it's not very probable,
8971
          # and we have no other way to check)
8972
          current_mem = 0
8973
        miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
8974
                    pninfo.payload['memory_free'])
8975
        if miss_mem > 0:
8976
          raise errors.OpPrereqError("This change will prevent the instance"
8977
                                     " from starting, due to %d MB of memory"
8978
                                     " missing on its primary node" % miss_mem,
8979
                                     errors.ECODE_NORES)
8980

    
8981
      if be_new[constants.BE_AUTO_BALANCE]:
8982
        for node, nres in nodeinfo.items():
8983
          if node not in instance.secondary_nodes:
8984
            continue
8985
          msg = nres.fail_msg
8986
          if msg:
8987
            self.warn.append("Can't get info from secondary node %s: %s" %
8988
                             (node, msg))
8989
          elif not isinstance(nres.payload.get('memory_free', None), int):
8990
            self.warn.append("Secondary node %s didn't return free"
8991
                             " memory information" % node)
8992
          elif be_new[constants.BE_MEMORY] > nres.payload['memory_free']:
8993
            self.warn.append("Not enough memory to failover instance to"
8994
                             " secondary node %s" % node)
8995

    
8996
    # NIC processing
8997
    self.nic_pnew = {}
8998
    self.nic_pinst = {}
8999
    for nic_op, nic_dict in self.op.nics:
9000
      if nic_op == constants.DDM_REMOVE:
9001
        if not instance.nics:
9002
          raise errors.OpPrereqError("Instance has no NICs, cannot remove",
9003
                                     errors.ECODE_INVAL)
9004
        continue
9005
      if nic_op != constants.DDM_ADD:
9006
        # an existing nic
9007
        if not instance.nics:
9008
          raise errors.OpPrereqError("Invalid NIC index %s, instance has"
9009
                                     " no NICs" % nic_op,
9010
                                     errors.ECODE_INVAL)
9011
        if nic_op < 0 or nic_op >= len(instance.nics):
9012
          raise errors.OpPrereqError("Invalid NIC index %s, valid values"
9013
                                     " are 0 to %d" %
9014
                                     (nic_op, len(instance.nics) - 1),
9015
                                     errors.ECODE_INVAL)
9016
        old_nic_params = instance.nics[nic_op].nicparams
9017
        old_nic_ip = instance.nics[nic_op].ip
9018
      else:
9019
        old_nic_params = {}
9020
        old_nic_ip = None
9021

    
9022
      update_params_dict = dict([(key, nic_dict[key])
9023
                                 for key in constants.NICS_PARAMETERS
9024
                                 if key in nic_dict])
9025

    
9026
      if 'bridge' in nic_dict:
9027
        update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
9028

    
9029
      new_nic_params = _GetUpdatedParams(old_nic_params,
9030
                                         update_params_dict)
9031
      utils.ForceDictType(new_nic_params, constants.NICS_PARAMETER_TYPES)
9032
      new_filled_nic_params = cluster.SimpleFillNIC(new_nic_params)
9033
      objects.NIC.CheckParameterSyntax(new_filled_nic_params)
9034
      self.nic_pinst[nic_op] = new_nic_params
9035
      self.nic_pnew[nic_op] = new_filled_nic_params
9036
      new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
9037

    
9038
      if new_nic_mode == constants.NIC_MODE_BRIDGED:
9039
        nic_bridge = new_filled_nic_params[constants.NIC_LINK]
9040
        msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
9041
        if msg:
9042
          msg = "Error checking bridges on node %s: %s" % (pnode, msg)
9043
          if self.op.force:
9044
            self.warn.append(msg)
9045
          else:
9046
            raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
9047
      if new_nic_mode == constants.NIC_MODE_ROUTED:
9048
        if 'ip' in nic_dict:
9049
          nic_ip = nic_dict['ip']
9050
        else:
9051
          nic_ip = old_nic_ip
9052
        if nic_ip is None:
9053
          raise errors.OpPrereqError('Cannot set the nic ip to None'
9054
                                     ' on a routed nic', errors.ECODE_INVAL)
9055
      if 'mac' in nic_dict:
9056
        nic_mac = nic_dict['mac']
9057
        if nic_mac is None:
9058
          raise errors.OpPrereqError('Cannot set the nic mac to None',
9059
                                     errors.ECODE_INVAL)
9060
        elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
9061
          # otherwise generate the mac
9062
          nic_dict['mac'] = self.cfg.GenerateMAC(self.proc.GetECId())
9063
        else:
9064
          # or validate/reserve the current one
9065
          try:
9066
            self.cfg.ReserveMAC(nic_mac, self.proc.GetECId())
9067
          except errors.ReservationError:
9068
            raise errors.OpPrereqError("MAC address %s already in use"
9069
                                       " in cluster" % nic_mac,
9070
                                       errors.ECODE_NOTUNIQUE)
9071

    
9072
    # DISK processing
9073
    if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
9074
      raise errors.OpPrereqError("Disk operations not supported for"
9075
                                 " diskless instances",
9076
                                 errors.ECODE_INVAL)
9077
    for disk_op, _ in self.op.disks:
9078
      if disk_op == constants.DDM_REMOVE:
9079
        if len(instance.disks) == 1:
9080
          raise errors.OpPrereqError("Cannot remove the last disk of"
9081
                                     " an instance", errors.ECODE_INVAL)
9082
        _CheckInstanceDown(self, instance, "cannot remove disks")
9083

    
9084
      if (disk_op == constants.DDM_ADD and
9085
          len(instance.nics) >= constants.MAX_DISKS):
9086
        raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
9087
                                   " add more" % constants.MAX_DISKS,
9088
                                   errors.ECODE_STATE)
9089
      if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
9090
        # an existing disk
9091
        if disk_op < 0 or disk_op >= len(instance.disks):
9092
          raise errors.OpPrereqError("Invalid disk index %s, valid values"
9093
                                     " are 0 to %d" %
9094
                                     (disk_op, len(instance.disks)),
9095
                                     errors.ECODE_INVAL)
9096

    
9097
    return
9098

    
9099
  def _ConvertPlainToDrbd(self, feedback_fn):
9100
    """Converts an instance from plain to drbd.
9101

9102
    """
9103
    feedback_fn("Converting template to drbd")
9104
    instance = self.instance
9105
    pnode = instance.primary_node
9106
    snode = self.op.remote_node
9107

    
9108
    # create a fake disk info for _GenerateDiskTemplate
9109
    disk_info = [{"size": d.size, "mode": d.mode} for d in instance.disks]
9110
    new_disks = _GenerateDiskTemplate(self, self.op.disk_template,
9111
                                      instance.name, pnode, [snode],
9112
                                      disk_info, None, None, 0)
9113
    info = _GetInstanceInfoText(instance)
9114
    feedback_fn("Creating aditional volumes...")
9115
    # first, create the missing data and meta devices
9116
    for disk in new_disks:
9117
      # unfortunately this is... not too nice
9118
      _CreateSingleBlockDev(self, pnode, instance, disk.children[1],
9119
                            info, True)
9120
      for child in disk.children:
9121
        _CreateSingleBlockDev(self, snode, instance, child, info, True)
9122
    # at this stage, all new LVs have been created, we can rename the
9123
    # old ones
9124
    feedback_fn("Renaming original volumes...")
9125
    rename_list = [(o, n.children[0].logical_id)
9126
                   for (o, n) in zip(instance.disks, new_disks)]
9127
    result = self.rpc.call_blockdev_rename(pnode, rename_list)
9128
    result.Raise("Failed to rename original LVs")
9129

    
9130
    feedback_fn("Initializing DRBD devices...")
9131
    # all child devices are in place, we can now create the DRBD devices
9132
    for disk in new_disks:
9133
      for node in [pnode, snode]:
9134
        f_create = node == pnode
9135
        _CreateSingleBlockDev(self, node, instance, disk, info, f_create)
9136

    
9137
    # at this point, the instance has been modified
9138
    instance.disk_template = constants.DT_DRBD8
9139
    instance.disks = new_disks
9140
    self.cfg.Update(instance, feedback_fn)
9141

    
9142
    # disks are created, waiting for sync
9143
    disk_abort = not _WaitForSync(self, instance)
9144
    if disk_abort:
9145
      raise errors.OpExecError("There are some degraded disks for"
9146
                               " this instance, please cleanup manually")
9147

    
9148
  def _ConvertDrbdToPlain(self, feedback_fn):
9149
    """Converts an instance from drbd to plain.
9150

9151
    """
9152
    instance = self.instance
9153
    assert len(instance.secondary_nodes) == 1
9154
    pnode = instance.primary_node
9155
    snode = instance.secondary_nodes[0]
9156
    feedback_fn("Converting template to plain")
9157

    
9158
    old_disks = instance.disks
9159
    new_disks = [d.children[0] for d in old_disks]
9160

    
9161
    # copy over size and mode
9162
    for parent, child in zip(old_disks, new_disks):
9163
      child.size = parent.size
9164
      child.mode = parent.mode
9165

    
9166
    # update instance structure
9167
    instance.disks = new_disks
9168
    instance.disk_template = constants.DT_PLAIN
9169
    self.cfg.Update(instance, feedback_fn)
9170

    
9171
    feedback_fn("Removing volumes on the secondary node...")
9172
    for disk in old_disks:
9173
      self.cfg.SetDiskID(disk, snode)
9174
      msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg
9175
      if msg:
9176
        self.LogWarning("Could not remove block device %s on node %s,"
9177
                        " continuing anyway: %s", disk.iv_name, snode, msg)
9178

    
9179
    feedback_fn("Removing unneeded volumes on the primary node...")
9180
    for idx, disk in enumerate(old_disks):
9181
      meta = disk.children[1]
9182
      self.cfg.SetDiskID(meta, pnode)
9183
      msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg
9184
      if msg:
9185
        self.LogWarning("Could not remove metadata for disk %d on node %s,"
9186
                        " continuing anyway: %s", idx, pnode, msg)
9187

    
9188

    
9189
  def Exec(self, feedback_fn):
9190
    """Modifies an instance.
9191

9192
    All parameters take effect only at the next restart of the instance.
9193

9194
    """
9195
    # Process here the warnings from CheckPrereq, as we don't have a
9196
    # feedback_fn there.
9197
    for warn in self.warn:
9198
      feedback_fn("WARNING: %s" % warn)
9199

    
9200
    result = []
9201
    instance = self.instance
9202
    # disk changes
9203
    for disk_op, disk_dict in self.op.disks:
9204
      if disk_op == constants.DDM_REMOVE:
9205
        # remove the last disk
9206
        device = instance.disks.pop()
9207
        device_idx = len(instance.disks)
9208
        for node, disk in device.ComputeNodeTree(instance.primary_node):
9209
          self.cfg.SetDiskID(disk, node)
9210
          msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
9211
          if msg:
9212
            self.LogWarning("Could not remove disk/%d on node %s: %s,"
9213
                            " continuing anyway", device_idx, node, msg)
9214
        result.append(("disk/%d" % device_idx, "remove"))
9215
      elif disk_op == constants.DDM_ADD:
9216
        # add a new disk
9217
        if instance.disk_template == constants.DT_FILE:
9218
          file_driver, file_path = instance.disks[0].logical_id
9219
          file_path = os.path.dirname(file_path)
9220
        else:
9221
          file_driver = file_path = None
9222
        disk_idx_base = len(instance.disks)
9223
        new_disk = _GenerateDiskTemplate(self,
9224
                                         instance.disk_template,
9225
                                         instance.name, instance.primary_node,
9226
                                         instance.secondary_nodes,
9227
                                         [disk_dict],
9228
                                         file_path,
9229
                                         file_driver,
9230
                                         disk_idx_base)[0]
9231
        instance.disks.append(new_disk)
9232
        info = _GetInstanceInfoText(instance)
9233

    
9234
        logging.info("Creating volume %s for instance %s",
9235
                     new_disk.iv_name, instance.name)
9236
        # Note: this needs to be kept in sync with _CreateDisks
9237
        #HARDCODE
9238
        for node in instance.all_nodes:
9239
          f_create = node == instance.primary_node
9240
          try:
9241
            _CreateBlockDev(self, node, instance, new_disk,
9242
                            f_create, info, f_create)
9243
          except errors.OpExecError, err:
9244
            self.LogWarning("Failed to create volume %s (%s) on"
9245
                            " node %s: %s",
9246
                            new_disk.iv_name, new_disk, node, err)
9247
        result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
9248
                       (new_disk.size, new_disk.mode)))
9249
      else:
9250
        # change a given disk
9251
        instance.disks[disk_op].mode = disk_dict['mode']
9252
        result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
9253

    
9254
    if self.op.disk_template:
9255
      r_shut = _ShutdownInstanceDisks(self, instance)
9256
      if not r_shut:
9257
        raise errors.OpExecError("Cannot shutdow instance disks, unable to"
9258
                                 " proceed with disk template conversion")
9259
      mode = (instance.disk_template, self.op.disk_template)
9260
      try:
9261
        self._DISK_CONVERSIONS[mode](self, feedback_fn)
9262
      except:
9263
        self.cfg.ReleaseDRBDMinors(instance.name)
9264
        raise
9265
      result.append(("disk_template", self.op.disk_template))
9266

    
9267
    # NIC changes
9268
    for nic_op, nic_dict in self.op.nics:
9269
      if nic_op == constants.DDM_REMOVE:
9270
        # remove the last nic
9271
        del instance.nics[-1]
9272
        result.append(("nic.%d" % len(instance.nics), "remove"))
9273
      elif nic_op == constants.DDM_ADD:
9274
        # mac and bridge should be set, by now
9275
        mac = nic_dict['mac']
9276
        ip = nic_dict.get('ip', None)
9277
        nicparams = self.nic_pinst[constants.DDM_ADD]
9278
        new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
9279
        instance.nics.append(new_nic)
9280
        result.append(("nic.%d" % (len(instance.nics) - 1),
9281
                       "add:mac=%s,ip=%s,mode=%s,link=%s" %
9282
                       (new_nic.mac, new_nic.ip,
9283
                        self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
9284
                        self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
9285
                       )))
9286
      else:
9287
        for key in 'mac', 'ip':
9288
          if key in nic_dict:
9289
            setattr(instance.nics[nic_op], key, nic_dict[key])
9290
        if nic_op in self.nic_pinst:
9291
          instance.nics[nic_op].nicparams = self.nic_pinst[nic_op]
9292
        for key, val in nic_dict.iteritems():
9293
          result.append(("nic.%s/%d" % (key, nic_op), val))
9294

    
9295
    # hvparams changes
9296
    if self.op.hvparams:
9297
      instance.hvparams = self.hv_inst
9298
      for key, val in self.op.hvparams.iteritems():
9299
        result.append(("hv/%s" % key, val))
9300

    
9301
    # beparams changes
9302
    if self.op.beparams:
9303
      instance.beparams = self.be_inst
9304
      for key, val in self.op.beparams.iteritems():
9305
        result.append(("be/%s" % key, val))
9306

    
9307
    # OS change
9308
    if self.op.os_name:
9309
      instance.os = self.op.os_name
9310

    
9311
    # osparams changes
9312
    if self.op.osparams:
9313
      instance.osparams = self.os_inst
9314
      for key, val in self.op.osparams.iteritems():
9315
        result.append(("os/%s" % key, val))
9316

    
9317
    self.cfg.Update(instance, feedback_fn)
9318

    
9319
    return result
9320

    
9321
  _DISK_CONVERSIONS = {
9322
    (constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd,
9323
    (constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain,
9324
    }
9325

    
9326

    
9327
class LUQueryExports(NoHooksLU):
9328
  """Query the exports list
9329

9330
  """
9331
  _OP_PARAMS = [
9332
    ("nodes", _EmptyList, _TListOf(_TNonEmptyString)),
9333
    ("use_locking", False, _TBool),
9334
    ]
9335
  REQ_BGL = False
9336

    
9337
  def ExpandNames(self):
9338
    self.needed_locks = {}
9339
    self.share_locks[locking.LEVEL_NODE] = 1
9340
    if not self.op.nodes:
9341
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9342
    else:
9343
      self.needed_locks[locking.LEVEL_NODE] = \
9344
        _GetWantedNodes(self, self.op.nodes)
9345

    
9346
  def Exec(self, feedback_fn):
9347
    """Compute the list of all the exported system images.
9348

9349
    @rtype: dict
9350
    @return: a dictionary with the structure node->(export-list)
9351
        where export-list is a list of the instances exported on
9352
        that node.
9353

9354
    """
9355
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
9356
    rpcresult = self.rpc.call_export_list(self.nodes)
9357
    result = {}
9358
    for node in rpcresult:
9359
      if rpcresult[node].fail_msg:
9360
        result[node] = False
9361
      else:
9362
        result[node] = rpcresult[node].payload
9363

    
9364
    return result
9365

    
9366

    
9367
class LUPrepareExport(NoHooksLU):
9368
  """Prepares an instance for an export and returns useful information.
9369

9370
  """
9371
  _OP_PARAMS = [
9372
    _PInstanceName,
9373
    ("mode", _NoDefault, _TElemOf(constants.EXPORT_MODES)),
9374
    ]
9375
  REQ_BGL = False
9376

    
9377
  def ExpandNames(self):
9378
    self._ExpandAndLockInstance()
9379

    
9380
  def CheckPrereq(self):
9381
    """Check prerequisites.
9382

9383
    """
9384
    instance_name = self.op.instance_name
9385

    
9386
    self.instance = self.cfg.GetInstanceInfo(instance_name)
9387
    assert self.instance is not None, \
9388
          "Cannot retrieve locked instance %s" % self.op.instance_name
9389
    _CheckNodeOnline(self, self.instance.primary_node)
9390

    
9391
    self._cds = _GetClusterDomainSecret()
9392

    
9393
  def Exec(self, feedback_fn):
9394
    """Prepares an instance for an export.
9395

9396
    """
9397
    instance = self.instance
9398

    
9399
    if self.op.mode == constants.EXPORT_MODE_REMOTE:
9400
      salt = utils.GenerateSecret(8)
9401

    
9402
      feedback_fn("Generating X509 certificate on %s" % instance.primary_node)
9403
      result = self.rpc.call_x509_cert_create(instance.primary_node,
9404
                                              constants.RIE_CERT_VALIDITY)
9405
      result.Raise("Can't create X509 key and certificate on %s" % result.node)
9406

    
9407
      (name, cert_pem) = result.payload
9408

    
9409
      cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
9410
                                             cert_pem)
9411

    
9412
      return {
9413
        "handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds),
9414
        "x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt),
9415
                          salt),
9416
        "x509_ca": utils.SignX509Certificate(cert, self._cds, salt),
9417
        }
9418

    
9419
    return None
9420

    
9421

    
9422
class LUExportInstance(LogicalUnit):
9423
  """Export an instance to an image in the cluster.
9424

9425
  """
9426
  HPATH = "instance-export"
9427
  HTYPE = constants.HTYPE_INSTANCE
9428
  _OP_PARAMS = [
9429
    _PInstanceName,
9430
    ("target_node", _NoDefault, _TOr(_TNonEmptyString, _TList)),
9431
    ("shutdown", True, _TBool),
9432
    _PShutdownTimeout,
9433
    ("remove_instance", False, _TBool),
9434
    ("ignore_remove_failures", False, _TBool),
9435
    ("mode", constants.EXPORT_MODE_LOCAL, _TElemOf(constants.EXPORT_MODES)),
9436
    ("x509_key_name", None, _TOr(_TList, _TNone)),
9437
    ("destination_x509_ca", None, _TMaybeString),
9438
    ]
9439
  REQ_BGL = False
9440

    
9441
  def CheckArguments(self):
9442
    """Check the arguments.
9443

9444
    """
9445
    self.x509_key_name = self.op.x509_key_name
9446
    self.dest_x509_ca_pem = self.op.destination_x509_ca
9447

    
9448
    if self.op.remove_instance and not self.op.shutdown:
9449
      raise errors.OpPrereqError("Can not remove instance without shutting it"
9450
                                 " down before")
9451

    
9452
    if self.op.mode == constants.EXPORT_MODE_REMOTE:
9453
      if not self.x509_key_name:
9454
        raise errors.OpPrereqError("Missing X509 key name for encryption",
9455
                                   errors.ECODE_INVAL)
9456

    
9457
      if not self.dest_x509_ca_pem:
9458
        raise errors.OpPrereqError("Missing destination X509 CA",
9459
                                   errors.ECODE_INVAL)
9460

    
9461
  def ExpandNames(self):
9462
    self._ExpandAndLockInstance()
9463

    
9464
    # Lock all nodes for local exports
9465
    if self.op.mode == constants.EXPORT_MODE_LOCAL:
9466
      # FIXME: lock only instance primary and destination node
9467
      #
9468
      # Sad but true, for now we have do lock all nodes, as we don't know where
9469
      # the previous export might be, and in this LU we search for it and
9470
      # remove it from its current node. In the future we could fix this by:
9471
      #  - making a tasklet to search (share-lock all), then create the
9472
      #    new one, then one to remove, after
9473
      #  - removing the removal operation altogether
9474
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9475

    
9476
  def DeclareLocks(self, level):
9477
    """Last minute lock declaration."""
9478
    # All nodes are locked anyway, so nothing to do here.
9479

    
9480
  def BuildHooksEnv(self):
9481
    """Build hooks env.
9482

9483
    This will run on the master, primary node and target node.
9484

9485
    """
9486
    env = {
9487
      "EXPORT_MODE": self.op.mode,
9488
      "EXPORT_NODE": self.op.target_node,
9489
      "EXPORT_DO_SHUTDOWN": self.op.shutdown,
9490
      "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
9491
      # TODO: Generic function for boolean env variables
9492
      "REMOVE_INSTANCE": str(bool(self.op.remove_instance)),
9493
      }
9494

    
9495
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
9496

    
9497
    nl = [self.cfg.GetMasterNode(), self.instance.primary_node]
9498

    
9499
    if self.op.mode == constants.EXPORT_MODE_LOCAL:
9500
      nl.append(self.op.target_node)
9501

    
9502
    return env, nl, nl
9503

    
9504
  def CheckPrereq(self):
9505
    """Check prerequisites.
9506

9507
    This checks that the instance and node names are valid.
9508

9509
    """
9510
    instance_name = self.op.instance_name
9511

    
9512
    self.instance = self.cfg.GetInstanceInfo(instance_name)
9513
    assert self.instance is not None, \
9514
          "Cannot retrieve locked instance %s" % self.op.instance_name
9515
    _CheckNodeOnline(self, self.instance.primary_node)
9516

    
9517
    if self.op.mode == constants.EXPORT_MODE_LOCAL:
9518
      self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
9519
      self.dst_node = self.cfg.GetNodeInfo(self.op.target_node)
9520
      assert self.dst_node is not None
9521

    
9522
      _CheckNodeOnline(self, self.dst_node.name)
9523
      _CheckNodeNotDrained(self, self.dst_node.name)
9524

    
9525
      self._cds = None
9526
      self.dest_disk_info = None
9527
      self.dest_x509_ca = None
9528

    
9529
    elif self.op.mode == constants.EXPORT_MODE_REMOTE:
9530
      self.dst_node = None
9531

    
9532
      if len(self.op.target_node) != len(self.instance.disks):
9533
        raise errors.OpPrereqError(("Received destination information for %s"
9534
                                    " disks, but instance %s has %s disks") %
9535
                                   (len(self.op.target_node), instance_name,
9536
                                    len(self.instance.disks)),
9537
                                   errors.ECODE_INVAL)
9538

    
9539
      cds = _GetClusterDomainSecret()
9540

    
9541
      # Check X509 key name
9542
      try:
9543
        (key_name, hmac_digest, hmac_salt) = self.x509_key_name
9544
      except (TypeError, ValueError), err:
9545
        raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err)
9546

    
9547
      if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt):
9548
        raise errors.OpPrereqError("HMAC for X509 key name is wrong",
9549
                                   errors.ECODE_INVAL)
9550

    
9551
      # Load and verify CA
9552
      try:
9553
        (cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds)
9554
      except OpenSSL.crypto.Error, err:
9555
        raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" %
9556
                                   (err, ), errors.ECODE_INVAL)
9557

    
9558
      (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
9559
      if errcode is not None:
9560
        raise errors.OpPrereqError("Invalid destination X509 CA (%s)" %
9561
                                   (msg, ), errors.ECODE_INVAL)
9562

    
9563
      self.dest_x509_ca = cert
9564

    
9565
      # Verify target information
9566
      disk_info = []
9567
      for idx, disk_data in enumerate(self.op.target_node):
9568
        try:
9569
          (host, port, magic) = \
9570
            masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data)
9571
        except errors.GenericError, err:
9572
          raise errors.OpPrereqError("Target info for disk %s: %s" %
9573
                                     (idx, err), errors.ECODE_INVAL)
9574

    
9575
        disk_info.append((host, port, magic))
9576

    
9577
      assert len(disk_info) == len(self.op.target_node)
9578
      self.dest_disk_info = disk_info
9579

    
9580
    else:
9581
      raise errors.ProgrammerError("Unhandled export mode %r" %
9582
                                   self.op.mode)
9583

    
9584
    # instance disk type verification
9585
    # TODO: Implement export support for file-based disks
9586
    for disk in self.instance.disks:
9587
      if disk.dev_type == constants.LD_FILE:
9588
        raise errors.OpPrereqError("Export not supported for instances with"
9589
                                   " file-based disks", errors.ECODE_INVAL)
9590

    
9591
  def _CleanupExports(self, feedback_fn):
9592
    """Removes exports of current instance from all other nodes.
9593

9594
    If an instance in a cluster with nodes A..D was exported to node C, its
9595
    exports will be removed from the nodes A, B and D.
9596

9597
    """
9598
    assert self.op.mode != constants.EXPORT_MODE_REMOTE
9599

    
9600
    nodelist = self.cfg.GetNodeList()
9601
    nodelist.remove(self.dst_node.name)
9602

    
9603
    # on one-node clusters nodelist will be empty after the removal
9604
    # if we proceed the backup would be removed because OpQueryExports
9605
    # substitutes an empty list with the full cluster node list.
9606
    iname = self.instance.name
9607
    if nodelist:
9608
      feedback_fn("Removing old exports for instance %s" % iname)
9609
      exportlist = self.rpc.call_export_list(nodelist)
9610
      for node in exportlist:
9611
        if exportlist[node].fail_msg:
9612
          continue
9613
        if iname in exportlist[node].payload:
9614
          msg = self.rpc.call_export_remove(node, iname).fail_msg
9615
          if msg:
9616
            self.LogWarning("Could not remove older export for instance %s"
9617
                            " on node %s: %s", iname, node, msg)
9618

    
9619
  def Exec(self, feedback_fn):
9620
    """Export an instance to an image in the cluster.
9621

9622
    """
9623
    assert self.op.mode in constants.EXPORT_MODES
9624

    
9625
    instance = self.instance
9626
    src_node = instance.primary_node
9627

    
9628
    if self.op.shutdown:
9629
      # shutdown the instance, but not the disks
9630
      feedback_fn("Shutting down instance %s" % instance.name)
9631
      result = self.rpc.call_instance_shutdown(src_node, instance,
9632
                                               self.op.shutdown_timeout)
9633
      # TODO: Maybe ignore failures if ignore_remove_failures is set
9634
      result.Raise("Could not shutdown instance %s on"
9635
                   " node %s" % (instance.name, src_node))
9636

    
9637
    # set the disks ID correctly since call_instance_start needs the
9638
    # correct drbd minor to create the symlinks
9639
    for disk in instance.disks:
9640
      self.cfg.SetDiskID(disk, src_node)
9641

    
9642
    activate_disks = (not instance.admin_up)
9643

    
9644
    if activate_disks:
9645
      # Activate the instance disks if we'exporting a stopped instance
9646
      feedback_fn("Activating disks for %s" % instance.name)
9647
      _StartInstanceDisks(self, instance, None)
9648

    
9649
    try:
9650
      helper = masterd.instance.ExportInstanceHelper(self, feedback_fn,
9651
                                                     instance)
9652

    
9653
      helper.CreateSnapshots()
9654
      try:
9655
        if (self.op.shutdown and instance.admin_up and
9656
            not self.op.remove_instance):
9657
          assert not activate_disks
9658
          feedback_fn("Starting instance %s" % instance.name)
9659
          result = self.rpc.call_instance_start(src_node, instance, None, None)
9660
          msg = result.fail_msg
9661
          if msg:
9662
            feedback_fn("Failed to start instance: %s" % msg)
9663
            _ShutdownInstanceDisks(self, instance)
9664
            raise errors.OpExecError("Could not start instance: %s" % msg)
9665

    
9666
        if self.op.mode == constants.EXPORT_MODE_LOCAL:
9667
          (fin_resu, dresults) = helper.LocalExport(self.dst_node)
9668
        elif self.op.mode == constants.EXPORT_MODE_REMOTE:
9669
          connect_timeout = constants.RIE_CONNECT_TIMEOUT
9670
          timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
9671

    
9672
          (key_name, _, _) = self.x509_key_name
9673

    
9674
          dest_ca_pem = \
9675
            OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM,
9676
                                            self.dest_x509_ca)
9677

    
9678
          (fin_resu, dresults) = helper.RemoteExport(self.dest_disk_info,
9679
                                                     key_name, dest_ca_pem,
9680
                                                     timeouts)
9681
      finally:
9682
        helper.Cleanup()
9683

    
9684
      # Check for backwards compatibility
9685
      assert len(dresults) == len(instance.disks)
9686
      assert compat.all(isinstance(i, bool) for i in dresults), \
9687
             "Not all results are boolean: %r" % dresults
9688

    
9689
    finally:
9690
      if activate_disks:
9691
        feedback_fn("Deactivating disks for %s" % instance.name)
9692
        _ShutdownInstanceDisks(self, instance)
9693

    
9694
    if not (compat.all(dresults) and fin_resu):
9695
      failures = []
9696
      if not fin_resu:
9697
        failures.append("export finalization")
9698
      if not compat.all(dresults):
9699
        fdsk = utils.CommaJoin(idx for (idx, dsk) in enumerate(dresults)
9700
                               if not dsk)
9701
        failures.append("disk export: disk(s) %s" % fdsk)
9702

    
9703
      raise errors.OpExecError("Export failed, errors in %s" %
9704
                               utils.CommaJoin(failures))
9705

    
9706
    # At this point, the export was successful, we can cleanup/finish
9707

    
9708
    # Remove instance if requested
9709
    if self.op.remove_instance:
9710
      feedback_fn("Removing instance %s" % instance.name)
9711
      _RemoveInstance(self, feedback_fn, instance,
9712
                      self.op.ignore_remove_failures)
9713

    
9714
    if self.op.mode == constants.EXPORT_MODE_LOCAL:
9715
      self._CleanupExports(feedback_fn)
9716

    
9717
    return fin_resu, dresults
9718

    
9719

    
9720
class LURemoveExport(NoHooksLU):
9721
  """Remove exports related to the named instance.
9722

9723
  """
9724
  _OP_PARAMS = [
9725
    _PInstanceName,
9726
    ]
9727
  REQ_BGL = False
9728

    
9729
  def ExpandNames(self):
9730
    self.needed_locks = {}
9731
    # We need all nodes to be locked in order for RemoveExport to work, but we
9732
    # don't need to lock the instance itself, as nothing will happen to it (and
9733
    # we can remove exports also for a removed instance)
9734
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9735

    
9736
  def Exec(self, feedback_fn):
9737
    """Remove any export.
9738

9739
    """
9740
    instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
9741
    # If the instance was not found we'll try with the name that was passed in.
9742
    # This will only work if it was an FQDN, though.
9743
    fqdn_warn = False
9744
    if not instance_name:
9745
      fqdn_warn = True
9746
      instance_name = self.op.instance_name
9747

    
9748
    locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
9749
    exportlist = self.rpc.call_export_list(locked_nodes)
9750
    found = False
9751
    for node in exportlist:
9752
      msg = exportlist[node].fail_msg
9753
      if msg:
9754
        self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
9755
        continue
9756
      if instance_name in exportlist[node].payload:
9757
        found = True
9758
        result = self.rpc.call_export_remove(node, instance_name)
9759
        msg = result.fail_msg
9760
        if msg:
9761
          logging.error("Could not remove export for instance %s"
9762
                        " on node %s: %s", instance_name, node, msg)
9763

    
9764
    if fqdn_warn and not found:
9765
      feedback_fn("Export not found. If trying to remove an export belonging"
9766
                  " to a deleted instance please use its Fully Qualified"
9767
                  " Domain Name.")
9768

    
9769

    
9770
class TagsLU(NoHooksLU): # pylint: disable-msg=W0223
9771
  """Generic tags LU.
9772

9773
  This is an abstract class which is the parent of all the other tags LUs.
9774

9775
  """
9776

    
9777
  def ExpandNames(self):
9778
    self.needed_locks = {}
9779
    if self.op.kind == constants.TAG_NODE:
9780
      self.op.name = _ExpandNodeName(self.cfg, self.op.name)
9781
      self.needed_locks[locking.LEVEL_NODE] = self.op.name
9782
    elif self.op.kind == constants.TAG_INSTANCE:
9783
      self.op.name = _ExpandInstanceName(self.cfg, self.op.name)
9784
      self.needed_locks[locking.LEVEL_INSTANCE] = self.op.name
9785

    
9786
    # FIXME: Acquire BGL for cluster tag operations (as of this writing it's
9787
    # not possible to acquire the BGL based on opcode parameters)
9788

    
9789
  def CheckPrereq(self):
9790
    """Check prerequisites.
9791

9792
    """
9793
    if self.op.kind == constants.TAG_CLUSTER:
9794
      self.target = self.cfg.GetClusterInfo()
9795
    elif self.op.kind == constants.TAG_NODE:
9796
      self.target = self.cfg.GetNodeInfo(self.op.name)
9797
    elif self.op.kind == constants.TAG_INSTANCE:
9798
      self.target = self.cfg.GetInstanceInfo(self.op.name)
9799
    else:
9800
      raise errors.OpPrereqError("Wrong tag type requested (%s)" %
9801
                                 str(self.op.kind), errors.ECODE_INVAL)
9802

    
9803

    
9804
class LUGetTags(TagsLU):
9805
  """Returns the tags of a given object.
9806

9807
  """
9808
  _OP_PARAMS = [
9809
    ("kind", _NoDefault, _TElemOf(constants.VALID_TAG_TYPES)),
9810
    # Name is only meaningful for nodes and instances
9811
    ("name", _NoDefault, _TMaybeString),
9812
    ]
9813
  REQ_BGL = False
9814

    
9815
  def ExpandNames(self):
9816
    TagsLU.ExpandNames(self)
9817

    
9818
    # Share locks as this is only a read operation
9819
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
9820

    
9821
  def Exec(self, feedback_fn):
9822
    """Returns the tag list.
9823

9824
    """
9825
    return list(self.target.GetTags())
9826

    
9827

    
9828
class LUSearchTags(NoHooksLU):
9829
  """Searches the tags for a given pattern.
9830

9831
  """
9832
  _OP_PARAMS = [
9833
    ("pattern", _NoDefault, _TNonEmptyString),
9834
    ]
9835
  REQ_BGL = False
9836

    
9837
  def ExpandNames(self):
9838
    self.needed_locks = {}
9839

    
9840
  def CheckPrereq(self):
9841
    """Check prerequisites.
9842

9843
    This checks the pattern passed for validity by compiling it.
9844

9845
    """
9846
    try:
9847
      self.re = re.compile(self.op.pattern)
9848
    except re.error, err:
9849
      raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
9850
                                 (self.op.pattern, err), errors.ECODE_INVAL)
9851

    
9852
  def Exec(self, feedback_fn):
9853
    """Returns the tag list.
9854

9855
    """
9856
    cfg = self.cfg
9857
    tgts = [("/cluster", cfg.GetClusterInfo())]
9858
    ilist = cfg.GetAllInstancesInfo().values()
9859
    tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
9860
    nlist = cfg.GetAllNodesInfo().values()
9861
    tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
9862
    results = []
9863
    for path, target in tgts:
9864
      for tag in target.GetTags():
9865
        if self.re.search(tag):
9866
          results.append((path, tag))
9867
    return results
9868

    
9869

    
9870
class LUAddTags(TagsLU):
9871
  """Sets a tag on a given object.
9872

9873
  """
9874
  _OP_PARAMS = [
9875
    ("kind", _NoDefault, _TElemOf(constants.VALID_TAG_TYPES)),
9876
    # Name is only meaningful for nodes and instances
9877
    ("name", _NoDefault, _TMaybeString),
9878
    ("tags", _NoDefault, _TListOf(_TNonEmptyString)),
9879
    ]
9880
  REQ_BGL = False
9881

    
9882
  def CheckPrereq(self):
9883
    """Check prerequisites.
9884

9885
    This checks the type and length of the tag name and value.
9886

9887
    """
9888
    TagsLU.CheckPrereq(self)
9889
    for tag in self.op.tags:
9890
      objects.TaggableObject.ValidateTag(tag)
9891

    
9892
  def Exec(self, feedback_fn):
9893
    """Sets the tag.
9894

9895
    """
9896
    try:
9897
      for tag in self.op.tags:
9898
        self.target.AddTag(tag)
9899
    except errors.TagError, err:
9900
      raise errors.OpExecError("Error while setting tag: %s" % str(err))
9901
    self.cfg.Update(self.target, feedback_fn)
9902

    
9903

    
9904
class LUDelTags(TagsLU):
9905
  """Delete a list of tags from a given object.
9906

9907
  """
9908
  _OP_PARAMS = [
9909
    ("kind", _NoDefault, _TElemOf(constants.VALID_TAG_TYPES)),
9910
    # Name is only meaningful for nodes and instances
9911
    ("name", _NoDefault, _TMaybeString),
9912
    ("tags", _NoDefault, _TListOf(_TNonEmptyString)),
9913
    ]
9914
  REQ_BGL = False
9915

    
9916
  def CheckPrereq(self):
9917
    """Check prerequisites.
9918

9919
    This checks that we have the given tag.
9920

9921
    """
9922
    TagsLU.CheckPrereq(self)
9923
    for tag in self.op.tags:
9924
      objects.TaggableObject.ValidateTag(tag)
9925
    del_tags = frozenset(self.op.tags)
9926
    cur_tags = self.target.GetTags()
9927

    
9928
    diff_tags = del_tags - cur_tags
9929
    if diff_tags:
9930
      diff_names = ("'%s'" % i for i in sorted(diff_tags))
9931
      raise errors.OpPrereqError("Tag(s) %s not found" %
9932
                                 (utils.CommaJoin(diff_names), ),
9933
                                 errors.ECODE_NOENT)
9934

    
9935
  def Exec(self, feedback_fn):
9936
    """Remove the tag from the object.
9937

9938
    """
9939
    for tag in self.op.tags:
9940
      self.target.RemoveTag(tag)
9941
    self.cfg.Update(self.target, feedback_fn)
9942

    
9943

    
9944
class LUTestDelay(NoHooksLU):
9945
  """Sleep for a specified amount of time.
9946

9947
  This LU sleeps on the master and/or nodes for a specified amount of
9948
  time.
9949

9950
  """
9951
  _OP_PARAMS = [
9952
    ("duration", _NoDefault, _TFloat),
9953
    ("on_master", True, _TBool),
9954
    ("on_nodes", _EmptyList, _TListOf(_TNonEmptyString)),
9955
    ("repeat", 0, _TPositiveInt)
9956
    ]
9957
  REQ_BGL = False
9958

    
9959
  def ExpandNames(self):
9960
    """Expand names and set required locks.
9961

9962
    This expands the node list, if any.
9963

9964
    """
9965
    self.needed_locks = {}
9966
    if self.op.on_nodes:
9967
      # _GetWantedNodes can be used here, but is not always appropriate to use
9968
      # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
9969
      # more information.
9970
      self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
9971
      self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
9972

    
9973
  def _TestDelay(self):
9974
    """Do the actual sleep.
9975

9976
    """
9977
    if self.op.on_master:
9978
      if not utils.TestDelay(self.op.duration):
9979
        raise errors.OpExecError("Error during master delay test")
9980
    if self.op.on_nodes:
9981
      result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
9982
      for node, node_result in result.items():
9983
        node_result.Raise("Failure during rpc call to node %s" % node)
9984

    
9985
  def Exec(self, feedback_fn):
9986
    """Execute the test delay opcode, with the wanted repetitions.
9987

9988
    """
9989
    if self.op.repeat == 0:
9990
      self._TestDelay()
9991
    else:
9992
      top_value = self.op.repeat - 1
9993
      for i in range(self.op.repeat):
9994
        self.LogInfo("Test delay iteration %d/%d" % (i, top_value))
9995
        self._TestDelay()
9996

    
9997

    
9998
class LUTestJobqueue(NoHooksLU):
9999
  """Utility LU to test some aspects of the job queue.
10000

10001
  """
10002
  _OP_PARAMS = [
10003
    ("notify_waitlock", False, _TBool),
10004
    ("notify_exec", False, _TBool),
10005
    ("log_messages", _EmptyList, _TListOf(_TString)),
10006
    ("fail", False, _TBool),
10007
    ]
10008
  REQ_BGL = False
10009

    
10010
  # Must be lower than default timeout for WaitForJobChange to see whether it
10011
  # notices changed jobs
10012
  _CLIENT_CONNECT_TIMEOUT = 20.0
10013
  _CLIENT_CONFIRM_TIMEOUT = 60.0
10014

    
10015
  @classmethod
10016
  def _NotifyUsingSocket(cls, cb, errcls):
10017
    """Opens a Unix socket and waits for another program to connect.
10018

10019
    @type cb: callable
10020
    @param cb: Callback to send socket name to client
10021
    @type errcls: class
10022
    @param errcls: Exception class to use for errors
10023

10024
    """
10025
    # Using a temporary directory as there's no easy way to create temporary
10026
    # sockets without writing a custom loop around tempfile.mktemp and
10027
    # socket.bind
10028
    tmpdir = tempfile.mkdtemp()
10029
    try:
10030
      tmpsock = utils.PathJoin(tmpdir, "sock")
10031

    
10032
      logging.debug("Creating temporary socket at %s", tmpsock)
10033
      sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
10034
      try:
10035
        sock.bind(tmpsock)
10036
        sock.listen(1)
10037

    
10038
        # Send details to client
10039
        cb(tmpsock)
10040

    
10041
        # Wait for client to connect before continuing
10042
        sock.settimeout(cls._CLIENT_CONNECT_TIMEOUT)
10043
        try:
10044
          (conn, _) = sock.accept()
10045
        except socket.error, err:
10046
          raise errcls("Client didn't connect in time (%s)" % err)
10047
      finally:
10048
        sock.close()
10049
    finally:
10050
      # Remove as soon as client is connected
10051
      shutil.rmtree(tmpdir)
10052

    
10053
    # Wait for client to close
10054
    try:
10055
      try:
10056
        # pylint: disable-msg=E1101
10057
        # Instance of '_socketobject' has no ... member
10058
        conn.settimeout(cls._CLIENT_CONFIRM_TIMEOUT)
10059
        conn.recv(1)
10060
      except socket.error, err:
10061
        raise errcls("Client failed to confirm notification (%s)" % err)
10062
    finally:
10063
      conn.close()
10064

    
10065
  def _SendNotification(self, test, arg, sockname):
10066
    """Sends a notification to the client.
10067

10068
    @type test: string
10069
    @param test: Test name
10070
    @param arg: Test argument (depends on test)
10071
    @type sockname: string
10072
    @param sockname: Socket path
10073

10074
    """
10075
    self.Log(constants.ELOG_JQUEUE_TEST, (sockname, test, arg))
10076

    
10077
  def _Notify(self, prereq, test, arg):
10078
    """Notifies the client of a test.
10079

10080
    @type prereq: bool
10081
    @param prereq: Whether this is a prereq-phase test
10082
    @type test: string
10083
    @param test: Test name
10084
    @param arg: Test argument (depends on test)
10085

10086
    """
10087
    if prereq:
10088
      errcls = errors.OpPrereqError
10089
    else:
10090
      errcls = errors.OpExecError
10091

    
10092
    return self._NotifyUsingSocket(compat.partial(self._SendNotification,
10093
                                                  test, arg),
10094
                                   errcls)
10095

    
10096
  def CheckArguments(self):
10097
    self.checkargs_calls = getattr(self, "checkargs_calls", 0) + 1
10098
    self.expandnames_calls = 0
10099

    
10100
  def ExpandNames(self):
10101
    checkargs_calls = getattr(self, "checkargs_calls", 0)
10102
    if checkargs_calls < 1:
10103
      raise errors.ProgrammerError("CheckArguments was not called")
10104

    
10105
    self.expandnames_calls += 1
10106

    
10107
    if self.op.notify_waitlock:
10108
      self._Notify(True, constants.JQT_EXPANDNAMES, None)
10109

    
10110
    self.LogInfo("Expanding names")
10111

    
10112
    # Get lock on master node (just to get a lock, not for a particular reason)
10113
    self.needed_locks = {
10114
      locking.LEVEL_NODE: self.cfg.GetMasterNode(),
10115
      }
10116

    
10117
  def Exec(self, feedback_fn):
10118
    if self.expandnames_calls < 1:
10119
      raise errors.ProgrammerError("ExpandNames was not called")
10120

    
10121
    if self.op.notify_exec:
10122
      self._Notify(False, constants.JQT_EXEC, None)
10123

    
10124
    self.LogInfo("Executing")
10125

    
10126
    if self.op.log_messages:
10127
      self._Notify(False, constants.JQT_STARTMSG, len(self.op.log_messages))
10128
      for idx, msg in enumerate(self.op.log_messages):
10129
        self.LogInfo("Sending log message %s", idx + 1)
10130
        feedback_fn(constants.JQT_MSGPREFIX + msg)
10131
        # Report how many test messages have been sent
10132
        self._Notify(False, constants.JQT_LOGMSG, idx + 1)
10133

    
10134
    if self.op.fail:
10135
      raise errors.OpExecError("Opcode failure was requested")
10136

    
10137
    return True
10138

    
10139

    
10140
class IAllocator(object):
10141
  """IAllocator framework.
10142

10143
  An IAllocator instance has three sets of attributes:
10144
    - cfg that is needed to query the cluster
10145
    - input data (all members of the _KEYS class attribute are required)
10146
    - four buffer attributes (in|out_data|text), that represent the
10147
      input (to the external script) in text and data structure format,
10148
      and the output from it, again in two formats
10149
    - the result variables from the script (success, info, nodes) for
10150
      easy usage
10151

10152
  """
10153
  # pylint: disable-msg=R0902
10154
  # lots of instance attributes
10155
  _ALLO_KEYS = [
10156
    "name", "mem_size", "disks", "disk_template",
10157
    "os", "tags", "nics", "vcpus", "hypervisor",
10158
    ]
10159
  _RELO_KEYS = [
10160
    "name", "relocate_from",
10161
    ]
10162
  _EVAC_KEYS = [
10163
    "evac_nodes",
10164
    ]
10165

    
10166
  def __init__(self, cfg, rpc, mode, **kwargs):
10167
    self.cfg = cfg
10168
    self.rpc = rpc
10169
    # init buffer variables
10170
    self.in_text = self.out_text = self.in_data = self.out_data = None
10171
    # init all input fields so that pylint is happy
10172
    self.mode = mode
10173
    self.mem_size = self.disks = self.disk_template = None
10174
    self.os = self.tags = self.nics = self.vcpus = None
10175
    self.hypervisor = None
10176
    self.relocate_from = None
10177
    self.name = None
10178
    self.evac_nodes = None
10179
    # computed fields
10180
    self.required_nodes = None
10181
    # init result fields
10182
    self.success = self.info = self.result = None
10183
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
10184
      keyset = self._ALLO_KEYS
10185
      fn = self._AddNewInstance
10186
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
10187
      keyset = self._RELO_KEYS
10188
      fn = self._AddRelocateInstance
10189
    elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
10190
      keyset = self._EVAC_KEYS
10191
      fn = self._AddEvacuateNodes
10192
    else:
10193
      raise errors.ProgrammerError("Unknown mode '%s' passed to the"
10194
                                   " IAllocator" % self.mode)
10195
    for key in kwargs:
10196
      if key not in keyset:
10197
        raise errors.ProgrammerError("Invalid input parameter '%s' to"
10198
                                     " IAllocator" % key)
10199
      setattr(self, key, kwargs[key])
10200

    
10201
    for key in keyset:
10202
      if key not in kwargs:
10203
        raise errors.ProgrammerError("Missing input parameter '%s' to"
10204
                                     " IAllocator" % key)
10205
    self._BuildInputData(fn)
10206

    
10207
  def _ComputeClusterData(self):
10208
    """Compute the generic allocator input data.
10209

10210
    This is the data that is independent of the actual operation.
10211

10212
    """
10213
    cfg = self.cfg
10214
    cluster_info = cfg.GetClusterInfo()
10215
    # cluster data
10216
    data = {
10217
      "version": constants.IALLOCATOR_VERSION,
10218
      "cluster_name": cfg.GetClusterName(),
10219
      "cluster_tags": list(cluster_info.GetTags()),
10220
      "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
10221
      # we don't have job IDs
10222
      }
10223
    iinfo = cfg.GetAllInstancesInfo().values()
10224
    i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
10225

    
10226
    # node data
10227
    node_results = {}
10228
    node_list = cfg.GetNodeList()
10229

    
10230
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
10231
      hypervisor_name = self.hypervisor
10232
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
10233
      hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
10234
    elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
10235
      hypervisor_name = cluster_info.enabled_hypervisors[0]
10236

    
10237
    node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(),
10238
                                        hypervisor_name)
10239
    node_iinfo = \
10240
      self.rpc.call_all_instances_info(node_list,
10241
                                       cluster_info.enabled_hypervisors)
10242
    for nname, nresult in node_data.items():
10243
      # first fill in static (config-based) values
10244
      ninfo = cfg.GetNodeInfo(nname)
10245
      pnr = {
10246
        "tags": list(ninfo.GetTags()),
10247
        "primary_ip": ninfo.primary_ip,
10248
        "secondary_ip": ninfo.secondary_ip,
10249
        "offline": ninfo.offline,
10250
        "drained": ninfo.drained,
10251
        "master_candidate": ninfo.master_candidate,
10252
        }
10253

    
10254
      if not (ninfo.offline or ninfo.drained):
10255
        nresult.Raise("Can't get data for node %s" % nname)
10256
        node_iinfo[nname].Raise("Can't get node instance info from node %s" %
10257
                                nname)
10258
        remote_info = nresult.payload
10259

    
10260
        for attr in ['memory_total', 'memory_free', 'memory_dom0',
10261
                     'vg_size', 'vg_free', 'cpu_total']:
10262
          if attr not in remote_info:
10263
            raise errors.OpExecError("Node '%s' didn't return attribute"
10264
                                     " '%s'" % (nname, attr))
10265
          if not isinstance(remote_info[attr], int):
10266
            raise errors.OpExecError("Node '%s' returned invalid value"
10267
                                     " for '%s': %s" %
10268
                                     (nname, attr, remote_info[attr]))
10269
        # compute memory used by primary instances
10270
        i_p_mem = i_p_up_mem = 0
10271
        for iinfo, beinfo in i_list:
10272
          if iinfo.primary_node == nname:
10273
            i_p_mem += beinfo[constants.BE_MEMORY]
10274
            if iinfo.name not in node_iinfo[nname].payload:
10275
              i_used_mem = 0
10276
            else:
10277
              i_used_mem = int(node_iinfo[nname].payload[iinfo.name]['memory'])
10278
            i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
10279
            remote_info['memory_free'] -= max(0, i_mem_diff)
10280

    
10281
            if iinfo.admin_up:
10282
              i_p_up_mem += beinfo[constants.BE_MEMORY]
10283

    
10284
        # compute memory used by instances
10285
        pnr_dyn = {
10286
          "total_memory": remote_info['memory_total'],
10287
          "reserved_memory": remote_info['memory_dom0'],
10288
          "free_memory": remote_info['memory_free'],
10289
          "total_disk": remote_info['vg_size'],
10290
          "free_disk": remote_info['vg_free'],
10291
          "total_cpus": remote_info['cpu_total'],
10292
          "i_pri_memory": i_p_mem,
10293
          "i_pri_up_memory": i_p_up_mem,
10294
          }
10295
        pnr.update(pnr_dyn)
10296

    
10297
      node_results[nname] = pnr
10298
    data["nodes"] = node_results
10299

    
10300
    # instance data
10301
    instance_data = {}
10302
    for iinfo, beinfo in i_list:
10303
      nic_data = []
10304
      for nic in iinfo.nics:
10305
        filled_params = cluster_info.SimpleFillNIC(nic.nicparams)
10306
        nic_dict = {"mac": nic.mac,
10307
                    "ip": nic.ip,
10308
                    "mode": filled_params[constants.NIC_MODE],
10309
                    "link": filled_params[constants.NIC_LINK],
10310
                   }
10311
        if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
10312
          nic_dict["bridge"] = filled_params[constants.NIC_LINK]
10313
        nic_data.append(nic_dict)
10314
      pir = {
10315
        "tags": list(iinfo.GetTags()),
10316
        "admin_up": iinfo.admin_up,
10317
        "vcpus": beinfo[constants.BE_VCPUS],
10318
        "memory": beinfo[constants.BE_MEMORY],
10319
        "os": iinfo.os,
10320
        "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
10321
        "nics": nic_data,
10322
        "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
10323
        "disk_template": iinfo.disk_template,
10324
        "hypervisor": iinfo.hypervisor,
10325
        }
10326
      pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
10327
                                                 pir["disks"])
10328
      instance_data[iinfo.name] = pir
10329

    
10330
    data["instances"] = instance_data
10331

    
10332
    self.in_data = data
10333

    
10334
  def _AddNewInstance(self):
10335
    """Add new instance data to allocator structure.
10336

10337
    This in combination with _AllocatorGetClusterData will create the
10338
    correct structure needed as input for the allocator.
10339

10340
    The checks for the completeness of the opcode must have already been
10341
    done.
10342

10343
    """
10344
    disk_space = _ComputeDiskSize(self.disk_template, self.disks)
10345

    
10346
    if self.disk_template in constants.DTS_NET_MIRROR:
10347
      self.required_nodes = 2
10348
    else:
10349
      self.required_nodes = 1
10350
    request = {
10351
      "name": self.name,
10352
      "disk_template": self.disk_template,
10353
      "tags": self.tags,
10354
      "os": self.os,
10355
      "vcpus": self.vcpus,
10356
      "memory": self.mem_size,
10357
      "disks": self.disks,
10358
      "disk_space_total": disk_space,
10359
      "nics": self.nics,
10360
      "required_nodes": self.required_nodes,
10361
      }
10362
    return request
10363

    
10364
  def _AddRelocateInstance(self):
10365
    """Add relocate instance data to allocator structure.
10366

10367
    This in combination with _IAllocatorGetClusterData will create the
10368
    correct structure needed as input for the allocator.
10369

10370
    The checks for the completeness of the opcode must have already been
10371
    done.
10372

10373
    """
10374
    instance = self.cfg.GetInstanceInfo(self.name)
10375
    if instance is None:
10376
      raise errors.ProgrammerError("Unknown instance '%s' passed to"
10377
                                   " IAllocator" % self.name)
10378

    
10379
    if instance.disk_template not in constants.DTS_NET_MIRROR:
10380
      raise errors.OpPrereqError("Can't relocate non-mirrored instances",
10381
                                 errors.ECODE_INVAL)
10382

    
10383
    if len(instance.secondary_nodes) != 1:
10384
      raise errors.OpPrereqError("Instance has not exactly one secondary node",
10385
                                 errors.ECODE_STATE)
10386

    
10387
    self.required_nodes = 1
10388
    disk_sizes = [{'size': disk.size} for disk in instance.disks]
10389
    disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
10390

    
10391
    request = {
10392
      "name": self.name,
10393
      "disk_space_total": disk_space,
10394
      "required_nodes": self.required_nodes,
10395
      "relocate_from": self.relocate_from,
10396
      }
10397
    return request
10398

    
10399
  def _AddEvacuateNodes(self):
10400
    """Add evacuate nodes data to allocator structure.
10401

10402
    """
10403
    request = {
10404
      "evac_nodes": self.evac_nodes
10405
      }
10406
    return request
10407

    
10408
  def _BuildInputData(self, fn):
10409
    """Build input data structures.
10410

10411
    """
10412
    self._ComputeClusterData()
10413

    
10414
    request = fn()
10415
    request["type"] = self.mode
10416
    self.in_data["request"] = request
10417

    
10418
    self.in_text = serializer.Dump(self.in_data)
10419

    
10420
  def Run(self, name, validate=True, call_fn=None):
10421
    """Run an instance allocator and return the results.
10422

10423
    """
10424
    if call_fn is None:
10425
      call_fn = self.rpc.call_iallocator_runner
10426

    
10427
    result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
10428
    result.Raise("Failure while running the iallocator script")
10429

    
10430
    self.out_text = result.payload
10431
    if validate:
10432
      self._ValidateResult()
10433

    
10434
  def _ValidateResult(self):
10435
    """Process the allocator results.
10436

10437
    This will process and if successful save the result in
10438
    self.out_data and the other parameters.
10439

10440
    """
10441
    try:
10442
      rdict = serializer.Load(self.out_text)
10443
    except Exception, err:
10444
      raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
10445

    
10446
    if not isinstance(rdict, dict):
10447
      raise errors.OpExecError("Can't parse iallocator results: not a dict")
10448

    
10449
    # TODO: remove backwards compatiblity in later versions
10450
    if "nodes" in rdict and "result" not in rdict:
10451
      rdict["result"] = rdict["nodes"]
10452
      del rdict["nodes"]
10453

    
10454
    for key in "success", "info", "result":
10455
      if key not in rdict:
10456
        raise errors.OpExecError("Can't parse iallocator results:"
10457
                                 " missing key '%s'" % key)
10458
      setattr(self, key, rdict[key])
10459

    
10460
    if not isinstance(rdict["result"], list):
10461
      raise errors.OpExecError("Can't parse iallocator results: 'result' key"
10462
                               " is not a list")
10463
    self.out_data = rdict
10464

    
10465

    
10466
class LUTestAllocator(NoHooksLU):
10467
  """Run allocator tests.
10468

10469
  This LU runs the allocator tests
10470

10471
  """
10472
  _OP_PARAMS = [
10473
    ("direction", _NoDefault, _TElemOf(constants.VALID_IALLOCATOR_DIRECTIONS)),
10474
    ("mode", _NoDefault, _TElemOf(constants.VALID_IALLOCATOR_MODES)),
10475
    ("name", _NoDefault, _TNonEmptyString),
10476
    ("nics", _NoDefault, _TOr(_TNone, _TListOf(
10477
      _TDictOf(_TElemOf(["mac", "ip", "bridge"]),
10478
               _TOr(_TNone, _TNonEmptyString))))),
10479
    ("disks", _NoDefault, _TOr(_TNone, _TList)),
10480
    ("hypervisor", None, _TMaybeString),
10481
    ("allocator", None, _TMaybeString),
10482
    ("tags", _EmptyList, _TListOf(_TNonEmptyString)),
10483
    ("mem_size", None, _TOr(_TNone, _TPositiveInt)),
10484
    ("vcpus", None, _TOr(_TNone, _TPositiveInt)),
10485
    ("os", None, _TMaybeString),
10486
    ("disk_template", None, _TMaybeString),
10487
    ("evac_nodes", None, _TOr(_TNone, _TListOf(_TNonEmptyString))),
10488
    ]
10489

    
10490
  def CheckPrereq(self):
10491
    """Check prerequisites.
10492

10493
    This checks the opcode parameters depending on the director and mode test.
10494

10495
    """
10496
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
10497
      for attr in ["mem_size", "disks", "disk_template",
10498
                   "os", "tags", "nics", "vcpus"]:
10499
        if not hasattr(self.op, attr):
10500
          raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
10501
                                     attr, errors.ECODE_INVAL)
10502
      iname = self.cfg.ExpandInstanceName(self.op.name)
10503
      if iname is not None:
10504
        raise errors.OpPrereqError("Instance '%s' already in the cluster" %
10505
                                   iname, errors.ECODE_EXISTS)
10506
      if not isinstance(self.op.nics, list):
10507
        raise errors.OpPrereqError("Invalid parameter 'nics'",
10508
                                   errors.ECODE_INVAL)
10509
      if not isinstance(self.op.disks, list):
10510
        raise errors.OpPrereqError("Invalid parameter 'disks'",
10511
                                   errors.ECODE_INVAL)
10512
      for row in self.op.disks:
10513
        if (not isinstance(row, dict) or
10514
            "size" not in row or
10515
            not isinstance(row["size"], int) or
10516
            "mode" not in row or
10517
            row["mode"] not in ['r', 'w']):
10518
          raise errors.OpPrereqError("Invalid contents of the 'disks'"
10519
                                     " parameter", errors.ECODE_INVAL)
10520
      if self.op.hypervisor is None:
10521
        self.op.hypervisor = self.cfg.GetHypervisorType()
10522
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
10523
      fname = _ExpandInstanceName(self.cfg, self.op.name)
10524
      self.op.name = fname
10525
      self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
10526
    elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
10527
      if not hasattr(self.op, "evac_nodes"):
10528
        raise errors.OpPrereqError("Missing attribute 'evac_nodes' on"
10529
                                   " opcode input", errors.ECODE_INVAL)
10530
    else:
10531
      raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
10532
                                 self.op.mode, errors.ECODE_INVAL)
10533

    
10534
    if self.op.direction == constants.IALLOCATOR_DIR_OUT:
10535
      if self.op.allocator is None:
10536
        raise errors.OpPrereqError("Missing allocator name",
10537
                                   errors.ECODE_INVAL)
10538
    elif self.op.direction != constants.IALLOCATOR_DIR_IN:
10539
      raise errors.OpPrereqError("Wrong allocator test '%s'" %
10540
                                 self.op.direction, errors.ECODE_INVAL)
10541

    
10542
  def Exec(self, feedback_fn):
10543
    """Run the allocator test.
10544

10545
    """
10546
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
10547
      ial = IAllocator(self.cfg, self.rpc,
10548
                       mode=self.op.mode,
10549
                       name=self.op.name,
10550
                       mem_size=self.op.mem_size,
10551
                       disks=self.op.disks,
10552
                       disk_template=self.op.disk_template,
10553
                       os=self.op.os,
10554
                       tags=self.op.tags,
10555
                       nics=self.op.nics,
10556
                       vcpus=self.op.vcpus,
10557
                       hypervisor=self.op.hypervisor,
10558
                       )
10559
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
10560
      ial = IAllocator(self.cfg, self.rpc,
10561
                       mode=self.op.mode,
10562
                       name=self.op.name,
10563
                       relocate_from=list(self.relocate_from),
10564
                       )
10565
    elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
10566
      ial = IAllocator(self.cfg, self.rpc,
10567
                       mode=self.op.mode,
10568
                       evac_nodes=self.op.evac_nodes)
10569
    else:
10570
      raise errors.ProgrammerError("Uncatched mode %s in"
10571
                                   " LUTestAllocator.Exec", self.op.mode)
10572

    
10573
    if self.op.direction == constants.IALLOCATOR_DIR_IN:
10574
      result = ial.in_text
10575
    else:
10576
      ial.Run(self.op.allocator, validate=False)
10577
      result = ial.out_text
10578
    return result