Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ 87b2cd45

History | View | Annotate | Download (368 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")
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
      else:
4252
        raise errors.ParameterError(field)
4253
      values.append(entry)
4254
    return values
4255

    
4256

    
4257
class LUActivateInstanceDisks(NoHooksLU):
4258
  """Bring up an instance's disks.
4259

4260
  """
4261
  _OP_PARAMS = [
4262
    _PInstanceName,
4263
    ("ignore_size", False, _TBool),
4264
    ]
4265
  REQ_BGL = False
4266

    
4267
  def ExpandNames(self):
4268
    self._ExpandAndLockInstance()
4269
    self.needed_locks[locking.LEVEL_NODE] = []
4270
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4271

    
4272
  def DeclareLocks(self, level):
4273
    if level == locking.LEVEL_NODE:
4274
      self._LockInstancesNodes()
4275

    
4276
  def CheckPrereq(self):
4277
    """Check prerequisites.
4278

4279
    This checks that the instance is in the cluster.
4280

4281
    """
4282
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4283
    assert self.instance is not None, \
4284
      "Cannot retrieve locked instance %s" % self.op.instance_name
4285
    _CheckNodeOnline(self, self.instance.primary_node)
4286

    
4287
  def Exec(self, feedback_fn):
4288
    """Activate the disks.
4289

4290
    """
4291
    disks_ok, disks_info = \
4292
              _AssembleInstanceDisks(self, self.instance,
4293
                                     ignore_size=self.op.ignore_size)
4294
    if not disks_ok:
4295
      raise errors.OpExecError("Cannot activate block devices")
4296

    
4297
    return disks_info
4298

    
4299

    
4300
def _AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False,
4301
                           ignore_size=False):
4302
  """Prepare the block devices for an instance.
4303

4304
  This sets up the block devices on all nodes.
4305

4306
  @type lu: L{LogicalUnit}
4307
  @param lu: the logical unit on whose behalf we execute
4308
  @type instance: L{objects.Instance}
4309
  @param instance: the instance for whose disks we assemble
4310
  @type disks: list of L{objects.Disk} or None
4311
  @param disks: which disks to assemble (or all, if None)
4312
  @type ignore_secondaries: boolean
4313
  @param ignore_secondaries: if true, errors on secondary nodes
4314
      won't result in an error return from the function
4315
  @type ignore_size: boolean
4316
  @param ignore_size: if true, the current known size of the disk
4317
      will not be used during the disk activation, useful for cases
4318
      when the size is wrong
4319
  @return: False if the operation failed, otherwise a list of
4320
      (host, instance_visible_name, node_visible_name)
4321
      with the mapping from node devices to instance devices
4322

4323
  """
4324
  device_info = []
4325
  disks_ok = True
4326
  iname = instance.name
4327
  disks = _ExpandCheckDisks(instance, disks)
4328

    
4329
  # With the two passes mechanism we try to reduce the window of
4330
  # opportunity for the race condition of switching DRBD to primary
4331
  # before handshaking occured, but we do not eliminate it
4332

    
4333
  # The proper fix would be to wait (with some limits) until the
4334
  # connection has been made and drbd transitions from WFConnection
4335
  # into any other network-connected state (Connected, SyncTarget,
4336
  # SyncSource, etc.)
4337

    
4338
  # 1st pass, assemble on all nodes in secondary mode
4339
  for inst_disk in disks:
4340
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
4341
      if ignore_size:
4342
        node_disk = node_disk.Copy()
4343
        node_disk.UnsetSize()
4344
      lu.cfg.SetDiskID(node_disk, node)
4345
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
4346
      msg = result.fail_msg
4347
      if msg:
4348
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
4349
                           " (is_primary=False, pass=1): %s",
4350
                           inst_disk.iv_name, node, msg)
4351
        if not ignore_secondaries:
4352
          disks_ok = False
4353

    
4354
  # FIXME: race condition on drbd migration to primary
4355

    
4356
  # 2nd pass, do only the primary node
4357
  for inst_disk in disks:
4358
    dev_path = None
4359

    
4360
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
4361
      if node != instance.primary_node:
4362
        continue
4363
      if ignore_size:
4364
        node_disk = node_disk.Copy()
4365
        node_disk.UnsetSize()
4366
      lu.cfg.SetDiskID(node_disk, node)
4367
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
4368
      msg = result.fail_msg
4369
      if msg:
4370
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
4371
                           " (is_primary=True, pass=2): %s",
4372
                           inst_disk.iv_name, node, msg)
4373
        disks_ok = False
4374
      else:
4375
        dev_path = result.payload
4376

    
4377
    device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
4378

    
4379
  # leave the disks configured for the primary node
4380
  # this is a workaround that would be fixed better by
4381
  # improving the logical/physical id handling
4382
  for disk in disks:
4383
    lu.cfg.SetDiskID(disk, instance.primary_node)
4384

    
4385
  return disks_ok, device_info
4386

    
4387

    
4388
def _StartInstanceDisks(lu, instance, force):
4389
  """Start the disks of an instance.
4390

4391
  """
4392
  disks_ok, _ = _AssembleInstanceDisks(lu, instance,
4393
                                           ignore_secondaries=force)
4394
  if not disks_ok:
4395
    _ShutdownInstanceDisks(lu, instance)
4396
    if force is not None and not force:
4397
      lu.proc.LogWarning("", hint="If the message above refers to a"
4398
                         " secondary node,"
4399
                         " you can retry the operation using '--force'.")
4400
    raise errors.OpExecError("Disk consistency error")
4401

    
4402

    
4403
class LUDeactivateInstanceDisks(NoHooksLU):
4404
  """Shutdown an instance's disks.
4405

4406
  """
4407
  _OP_PARAMS = [
4408
    _PInstanceName,
4409
    ]
4410
  REQ_BGL = False
4411

    
4412
  def ExpandNames(self):
4413
    self._ExpandAndLockInstance()
4414
    self.needed_locks[locking.LEVEL_NODE] = []
4415
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4416

    
4417
  def DeclareLocks(self, level):
4418
    if level == locking.LEVEL_NODE:
4419
      self._LockInstancesNodes()
4420

    
4421
  def CheckPrereq(self):
4422
    """Check prerequisites.
4423

4424
    This checks that the instance is in the cluster.
4425

4426
    """
4427
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4428
    assert self.instance is not None, \
4429
      "Cannot retrieve locked instance %s" % self.op.instance_name
4430

    
4431
  def Exec(self, feedback_fn):
4432
    """Deactivate the disks
4433

4434
    """
4435
    instance = self.instance
4436
    _SafeShutdownInstanceDisks(self, instance)
4437

    
4438

    
4439
def _SafeShutdownInstanceDisks(lu, instance, disks=None):
4440
  """Shutdown block devices of an instance.
4441

4442
  This function checks if an instance is running, before calling
4443
  _ShutdownInstanceDisks.
4444

4445
  """
4446
  _CheckInstanceDown(lu, instance, "cannot shutdown disks")
4447
  _ShutdownInstanceDisks(lu, instance, disks=disks)
4448

    
4449

    
4450
def _ExpandCheckDisks(instance, disks):
4451
  """Return the instance disks selected by the disks list
4452

4453
  @type disks: list of L{objects.Disk} or None
4454
  @param disks: selected disks
4455
  @rtype: list of L{objects.Disk}
4456
  @return: selected instance disks to act on
4457

4458
  """
4459
  if disks is None:
4460
    return instance.disks
4461
  else:
4462
    if not set(disks).issubset(instance.disks):
4463
      raise errors.ProgrammerError("Can only act on disks belonging to the"
4464
                                   " target instance")
4465
    return disks
4466

    
4467

    
4468
def _ShutdownInstanceDisks(lu, instance, disks=None, ignore_primary=False):
4469
  """Shutdown block devices of an instance.
4470

4471
  This does the shutdown on all nodes of the instance.
4472

4473
  If the ignore_primary is false, errors on the primary node are
4474
  ignored.
4475

4476
  """
4477
  all_result = True
4478
  disks = _ExpandCheckDisks(instance, disks)
4479

    
4480
  for disk in disks:
4481
    for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
4482
      lu.cfg.SetDiskID(top_disk, node)
4483
      result = lu.rpc.call_blockdev_shutdown(node, top_disk)
4484
      msg = result.fail_msg
4485
      if msg:
4486
        lu.LogWarning("Could not shutdown block device %s on node %s: %s",
4487
                      disk.iv_name, node, msg)
4488
        if not ignore_primary or node != instance.primary_node:
4489
          all_result = False
4490
  return all_result
4491

    
4492

    
4493
def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
4494
  """Checks if a node has enough free memory.
4495

4496
  This function check if a given node has the needed amount of free
4497
  memory. In case the node has less memory or we cannot get the
4498
  information from the node, this function raise an OpPrereqError
4499
  exception.
4500

4501
  @type lu: C{LogicalUnit}
4502
  @param lu: a logical unit from which we get configuration data
4503
  @type node: C{str}
4504
  @param node: the node to check
4505
  @type reason: C{str}
4506
  @param reason: string to use in the error message
4507
  @type requested: C{int}
4508
  @param requested: the amount of memory in MiB to check for
4509
  @type hypervisor_name: C{str}
4510
  @param hypervisor_name: the hypervisor to ask for memory stats
4511
  @raise errors.OpPrereqError: if the node doesn't have enough memory, or
4512
      we cannot check the node
4513

4514
  """
4515
  nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
4516
  nodeinfo[node].Raise("Can't get data from node %s" % node,
4517
                       prereq=True, ecode=errors.ECODE_ENVIRON)
4518
  free_mem = nodeinfo[node].payload.get('memory_free', None)
4519
  if not isinstance(free_mem, int):
4520
    raise errors.OpPrereqError("Can't compute free memory on node %s, result"
4521
                               " was '%s'" % (node, free_mem),
4522
                               errors.ECODE_ENVIRON)
4523
  if requested > free_mem:
4524
    raise errors.OpPrereqError("Not enough memory on node %s for %s:"
4525
                               " needed %s MiB, available %s MiB" %
4526
                               (node, reason, requested, free_mem),
4527
                               errors.ECODE_NORES)
4528

    
4529

    
4530
def _CheckNodesFreeDisk(lu, nodenames, requested):
4531
  """Checks if nodes have enough free disk space in the default VG.
4532

4533
  This function check if all given nodes have the needed amount of
4534
  free disk. In case any node has less disk or we cannot get the
4535
  information from the node, this function raise an OpPrereqError
4536
  exception.
4537

4538
  @type lu: C{LogicalUnit}
4539
  @param lu: a logical unit from which we get configuration data
4540
  @type nodenames: C{list}
4541
  @param nodenames: the list of node names to check
4542
  @type requested: C{int}
4543
  @param requested: the amount of disk in MiB to check for
4544
  @raise errors.OpPrereqError: if the node doesn't have enough disk, or
4545
      we cannot check the node
4546

4547
  """
4548
  nodeinfo = lu.rpc.call_node_info(nodenames, lu.cfg.GetVGName(),
4549
                                   lu.cfg.GetHypervisorType())
4550
  for node in nodenames:
4551
    info = nodeinfo[node]
4552
    info.Raise("Cannot get current information from node %s" % node,
4553
               prereq=True, ecode=errors.ECODE_ENVIRON)
4554
    vg_free = info.payload.get("vg_free", None)
4555
    if not isinstance(vg_free, int):
4556
      raise errors.OpPrereqError("Can't compute free disk space on node %s,"
4557
                                 " result was '%s'" % (node, vg_free),
4558
                                 errors.ECODE_ENVIRON)
4559
    if requested > vg_free:
4560
      raise errors.OpPrereqError("Not enough disk space on target node %s:"
4561
                                 " required %d MiB, available %d MiB" %
4562
                                 (node, requested, vg_free),
4563
                                 errors.ECODE_NORES)
4564

    
4565

    
4566
class LUStartupInstance(LogicalUnit):
4567
  """Starts an instance.
4568

4569
  """
4570
  HPATH = "instance-start"
4571
  HTYPE = constants.HTYPE_INSTANCE
4572
  _OP_PARAMS = [
4573
    _PInstanceName,
4574
    _PForce,
4575
    ("hvparams", _EmptyDict, _TDict),
4576
    ("beparams", _EmptyDict, _TDict),
4577
    ]
4578
  REQ_BGL = False
4579

    
4580
  def CheckArguments(self):
4581
    # extra beparams
4582
    if self.op.beparams:
4583
      # fill the beparams dict
4584
      utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
4585

    
4586
  def ExpandNames(self):
4587
    self._ExpandAndLockInstance()
4588

    
4589
  def BuildHooksEnv(self):
4590
    """Build hooks env.
4591

4592
    This runs on master, primary and secondary nodes of the instance.
4593

4594
    """
4595
    env = {
4596
      "FORCE": self.op.force,
4597
      }
4598
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4599
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4600
    return env, nl, nl
4601

    
4602
  def CheckPrereq(self):
4603
    """Check prerequisites.
4604

4605
    This checks that the instance is in the cluster.
4606

4607
    """
4608
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4609
    assert self.instance is not None, \
4610
      "Cannot retrieve locked instance %s" % self.op.instance_name
4611

    
4612
    # extra hvparams
4613
    if self.op.hvparams:
4614
      # check hypervisor parameter syntax (locally)
4615
      cluster = self.cfg.GetClusterInfo()
4616
      utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
4617
      filled_hvp = cluster.FillHV(instance)
4618
      filled_hvp.update(self.op.hvparams)
4619
      hv_type = hypervisor.GetHypervisor(instance.hypervisor)
4620
      hv_type.CheckParameterSyntax(filled_hvp)
4621
      _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
4622

    
4623
    _CheckNodeOnline(self, instance.primary_node)
4624

    
4625
    bep = self.cfg.GetClusterInfo().FillBE(instance)
4626
    # check bridges existence
4627
    _CheckInstanceBridgesExist(self, instance)
4628

    
4629
    remote_info = self.rpc.call_instance_info(instance.primary_node,
4630
                                              instance.name,
4631
                                              instance.hypervisor)
4632
    remote_info.Raise("Error checking node %s" % instance.primary_node,
4633
                      prereq=True, ecode=errors.ECODE_ENVIRON)
4634
    if not remote_info.payload: # not running already
4635
      _CheckNodeFreeMemory(self, instance.primary_node,
4636
                           "starting instance %s" % instance.name,
4637
                           bep[constants.BE_MEMORY], instance.hypervisor)
4638

    
4639
  def Exec(self, feedback_fn):
4640
    """Start the instance.
4641

4642
    """
4643
    instance = self.instance
4644
    force = self.op.force
4645

    
4646
    self.cfg.MarkInstanceUp(instance.name)
4647

    
4648
    node_current = instance.primary_node
4649

    
4650
    _StartInstanceDisks(self, instance, force)
4651

    
4652
    result = self.rpc.call_instance_start(node_current, instance,
4653
                                          self.op.hvparams, self.op.beparams)
4654
    msg = result.fail_msg
4655
    if msg:
4656
      _ShutdownInstanceDisks(self, instance)
4657
      raise errors.OpExecError("Could not start instance: %s" % msg)
4658

    
4659

    
4660
class LURebootInstance(LogicalUnit):
4661
  """Reboot an instance.
4662

4663
  """
4664
  HPATH = "instance-reboot"
4665
  HTYPE = constants.HTYPE_INSTANCE
4666
  _OP_PARAMS = [
4667
    _PInstanceName,
4668
    ("ignore_secondaries", False, _TBool),
4669
    ("reboot_type", _NoDefault, _TElemOf(constants.REBOOT_TYPES)),
4670
    _PShutdownTimeout,
4671
    ]
4672
  REQ_BGL = False
4673

    
4674
  def ExpandNames(self):
4675
    self._ExpandAndLockInstance()
4676

    
4677
  def BuildHooksEnv(self):
4678
    """Build hooks env.
4679

4680
    This runs on master, primary and secondary nodes of the instance.
4681

4682
    """
4683
    env = {
4684
      "IGNORE_SECONDARIES": self.op.ignore_secondaries,
4685
      "REBOOT_TYPE": self.op.reboot_type,
4686
      "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
4687
      }
4688
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4689
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4690
    return env, nl, nl
4691

    
4692
  def CheckPrereq(self):
4693
    """Check prerequisites.
4694

4695
    This checks that the instance is in the cluster.
4696

4697
    """
4698
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4699
    assert self.instance is not None, \
4700
      "Cannot retrieve locked instance %s" % self.op.instance_name
4701

    
4702
    _CheckNodeOnline(self, instance.primary_node)
4703

    
4704
    # check bridges existence
4705
    _CheckInstanceBridgesExist(self, instance)
4706

    
4707
  def Exec(self, feedback_fn):
4708
    """Reboot the instance.
4709

4710
    """
4711
    instance = self.instance
4712
    ignore_secondaries = self.op.ignore_secondaries
4713
    reboot_type = self.op.reboot_type
4714

    
4715
    node_current = instance.primary_node
4716

    
4717
    if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
4718
                       constants.INSTANCE_REBOOT_HARD]:
4719
      for disk in instance.disks:
4720
        self.cfg.SetDiskID(disk, node_current)
4721
      result = self.rpc.call_instance_reboot(node_current, instance,
4722
                                             reboot_type,
4723
                                             self.op.shutdown_timeout)
4724
      result.Raise("Could not reboot instance")
4725
    else:
4726
      result = self.rpc.call_instance_shutdown(node_current, instance,
4727
                                               self.op.shutdown_timeout)
4728
      result.Raise("Could not shutdown instance for full reboot")
4729
      _ShutdownInstanceDisks(self, instance)
4730
      _StartInstanceDisks(self, instance, ignore_secondaries)
4731
      result = self.rpc.call_instance_start(node_current, instance, None, None)
4732
      msg = result.fail_msg
4733
      if msg:
4734
        _ShutdownInstanceDisks(self, instance)
4735
        raise errors.OpExecError("Could not start instance for"
4736
                                 " full reboot: %s" % msg)
4737

    
4738
    self.cfg.MarkInstanceUp(instance.name)
4739

    
4740

    
4741
class LUShutdownInstance(LogicalUnit):
4742
  """Shutdown an instance.
4743

4744
  """
4745
  HPATH = "instance-stop"
4746
  HTYPE = constants.HTYPE_INSTANCE
4747
  _OP_PARAMS = [
4748
    _PInstanceName,
4749
    ("timeout", constants.DEFAULT_SHUTDOWN_TIMEOUT, _TPositiveInt),
4750
    ]
4751
  REQ_BGL = False
4752

    
4753
  def ExpandNames(self):
4754
    self._ExpandAndLockInstance()
4755

    
4756
  def BuildHooksEnv(self):
4757
    """Build hooks env.
4758

4759
    This runs on master, primary and secondary nodes of the instance.
4760

4761
    """
4762
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4763
    env["TIMEOUT"] = self.op.timeout
4764
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4765
    return env, nl, nl
4766

    
4767
  def CheckPrereq(self):
4768
    """Check prerequisites.
4769

4770
    This checks that the instance is in the cluster.
4771

4772
    """
4773
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4774
    assert self.instance is not None, \
4775
      "Cannot retrieve locked instance %s" % self.op.instance_name
4776
    _CheckNodeOnline(self, self.instance.primary_node)
4777

    
4778
  def Exec(self, feedback_fn):
4779
    """Shutdown the instance.
4780

4781
    """
4782
    instance = self.instance
4783
    node_current = instance.primary_node
4784
    timeout = self.op.timeout
4785
    self.cfg.MarkInstanceDown(instance.name)
4786
    result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
4787
    msg = result.fail_msg
4788
    if msg:
4789
      self.proc.LogWarning("Could not shutdown instance: %s" % msg)
4790

    
4791
    _ShutdownInstanceDisks(self, instance)
4792

    
4793

    
4794
class LUReinstallInstance(LogicalUnit):
4795
  """Reinstall an instance.
4796

4797
  """
4798
  HPATH = "instance-reinstall"
4799
  HTYPE = constants.HTYPE_INSTANCE
4800
  _OP_PARAMS = [
4801
    _PInstanceName,
4802
    ("os_type", None, _TMaybeString),
4803
    ("force_variant", False, _TBool),
4804
    ]
4805
  REQ_BGL = False
4806

    
4807
  def ExpandNames(self):
4808
    self._ExpandAndLockInstance()
4809

    
4810
  def BuildHooksEnv(self):
4811
    """Build hooks env.
4812

4813
    This runs on master, primary and secondary nodes of the instance.
4814

4815
    """
4816
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4817
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4818
    return env, nl, nl
4819

    
4820
  def CheckPrereq(self):
4821
    """Check prerequisites.
4822

4823
    This checks that the instance is in the cluster and is not running.
4824

4825
    """
4826
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4827
    assert instance is not None, \
4828
      "Cannot retrieve locked instance %s" % self.op.instance_name
4829
    _CheckNodeOnline(self, instance.primary_node)
4830

    
4831
    if instance.disk_template == constants.DT_DISKLESS:
4832
      raise errors.OpPrereqError("Instance '%s' has no disks" %
4833
                                 self.op.instance_name,
4834
                                 errors.ECODE_INVAL)
4835
    _CheckInstanceDown(self, instance, "cannot reinstall")
4836

    
4837
    if self.op.os_type is not None:
4838
      # OS verification
4839
      pnode = _ExpandNodeName(self.cfg, instance.primary_node)
4840
      _CheckNodeHasOS(self, pnode, self.op.os_type, self.op.force_variant)
4841

    
4842
    self.instance = instance
4843

    
4844
  def Exec(self, feedback_fn):
4845
    """Reinstall the instance.
4846

4847
    """
4848
    inst = self.instance
4849

    
4850
    if self.op.os_type is not None:
4851
      feedback_fn("Changing OS to '%s'..." % self.op.os_type)
4852
      inst.os = self.op.os_type
4853
      self.cfg.Update(inst, feedback_fn)
4854

    
4855
    _StartInstanceDisks(self, inst, None)
4856
    try:
4857
      feedback_fn("Running the instance OS create scripts...")
4858
      # FIXME: pass debug option from opcode to backend
4859
      result = self.rpc.call_instance_os_add(inst.primary_node, inst, True,
4860
                                             self.op.debug_level)
4861
      result.Raise("Could not install OS for instance %s on node %s" %
4862
                   (inst.name, inst.primary_node))
4863
    finally:
4864
      _ShutdownInstanceDisks(self, inst)
4865

    
4866

    
4867
class LURecreateInstanceDisks(LogicalUnit):
4868
  """Recreate an instance's missing disks.
4869

4870
  """
4871
  HPATH = "instance-recreate-disks"
4872
  HTYPE = constants.HTYPE_INSTANCE
4873
  _OP_PARAMS = [
4874
    _PInstanceName,
4875
    ("disks", _EmptyList, _TListOf(_TPositiveInt)),
4876
    ]
4877
  REQ_BGL = False
4878

    
4879
  def ExpandNames(self):
4880
    self._ExpandAndLockInstance()
4881

    
4882
  def BuildHooksEnv(self):
4883
    """Build hooks env.
4884

4885
    This runs on master, primary and secondary nodes of the instance.
4886

4887
    """
4888
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4889
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4890
    return env, nl, nl
4891

    
4892
  def CheckPrereq(self):
4893
    """Check prerequisites.
4894

4895
    This checks that the instance is in the cluster and is not running.
4896

4897
    """
4898
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4899
    assert instance is not None, \
4900
      "Cannot retrieve locked instance %s" % self.op.instance_name
4901
    _CheckNodeOnline(self, instance.primary_node)
4902

    
4903
    if instance.disk_template == constants.DT_DISKLESS:
4904
      raise errors.OpPrereqError("Instance '%s' has no disks" %
4905
                                 self.op.instance_name, errors.ECODE_INVAL)
4906
    _CheckInstanceDown(self, instance, "cannot recreate disks")
4907

    
4908
    if not self.op.disks:
4909
      self.op.disks = range(len(instance.disks))
4910
    else:
4911
      for idx in self.op.disks:
4912
        if idx >= len(instance.disks):
4913
          raise errors.OpPrereqError("Invalid disk index passed '%s'" % idx,
4914
                                     errors.ECODE_INVAL)
4915

    
4916
    self.instance = instance
4917

    
4918
  def Exec(self, feedback_fn):
4919
    """Recreate the disks.
4920

4921
    """
4922
    to_skip = []
4923
    for idx, _ in enumerate(self.instance.disks):
4924
      if idx not in self.op.disks: # disk idx has not been passed in
4925
        to_skip.append(idx)
4926
        continue
4927

    
4928
    _CreateDisks(self, self.instance, to_skip=to_skip)
4929

    
4930

    
4931
class LURenameInstance(LogicalUnit):
4932
  """Rename an instance.
4933

4934
  """
4935
  HPATH = "instance-rename"
4936
  HTYPE = constants.HTYPE_INSTANCE
4937
  _OP_PARAMS = [
4938
    _PInstanceName,
4939
    ("new_name", _NoDefault, _TNonEmptyString),
4940
    ("ip_check", False, _TBool),
4941
    ("name_check", True, _TBool),
4942
    ]
4943

    
4944
  def CheckArguments(self):
4945
    """Check arguments.
4946

4947
    """
4948
    if self.op.ip_check and not self.op.name_check:
4949
      # TODO: make the ip check more flexible and not depend on the name check
4950
      raise errors.OpPrereqError("Cannot do ip check without a name check",
4951
                                 errors.ECODE_INVAL)
4952

    
4953
  def BuildHooksEnv(self):
4954
    """Build hooks env.
4955

4956
    This runs on master, primary and secondary nodes of the instance.
4957

4958
    """
4959
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4960
    env["INSTANCE_NEW_NAME"] = self.op.new_name
4961
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4962
    return env, nl, nl
4963

    
4964
  def CheckPrereq(self):
4965
    """Check prerequisites.
4966

4967
    This checks that the instance is in the cluster and is not running.
4968

4969
    """
4970
    self.op.instance_name = _ExpandInstanceName(self.cfg,
4971
                                                self.op.instance_name)
4972
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4973
    assert instance is not None
4974
    _CheckNodeOnline(self, instance.primary_node)
4975
    _CheckInstanceDown(self, instance, "cannot rename")
4976
    self.instance = instance
4977

    
4978
    new_name = self.op.new_name
4979
    if self.op.name_check:
4980
      hostinfo = netutils.HostInfo(netutils.HostInfo.NormalizeName(new_name))
4981
      new_name = self.op.new_name = hostinfo.name
4982
      if (self.op.ip_check and
4983
          netutils.TcpPing(hostinfo.ip, constants.DEFAULT_NODED_PORT)):
4984
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
4985
                                   (hostinfo.ip, new_name),
4986
                                   errors.ECODE_NOTUNIQUE)
4987

    
4988
    instance_list = self.cfg.GetInstanceList()
4989
    if new_name in instance_list:
4990
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
4991
                                 new_name, errors.ECODE_EXISTS)
4992

    
4993

    
4994
  def Exec(self, feedback_fn):
4995
    """Reinstall the instance.
4996

4997
    """
4998
    inst = self.instance
4999
    old_name = inst.name
5000

    
5001
    if inst.disk_template == constants.DT_FILE:
5002
      old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
5003

    
5004
    self.cfg.RenameInstance(inst.name, self.op.new_name)
5005
    # Change the instance lock. This is definitely safe while we hold the BGL
5006
    self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
5007
    self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
5008

    
5009
    # re-read the instance from the configuration after rename
5010
    inst = self.cfg.GetInstanceInfo(self.op.new_name)
5011

    
5012
    if inst.disk_template == constants.DT_FILE:
5013
      new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
5014
      result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
5015
                                                     old_file_storage_dir,
5016
                                                     new_file_storage_dir)
5017
      result.Raise("Could not rename on node %s directory '%s' to '%s'"
5018
                   " (but the instance has been renamed in Ganeti)" %
5019
                   (inst.primary_node, old_file_storage_dir,
5020
                    new_file_storage_dir))
5021

    
5022
    _StartInstanceDisks(self, inst, None)
5023
    try:
5024
      result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
5025
                                                 old_name, self.op.debug_level)
5026
      msg = result.fail_msg
5027
      if msg:
5028
        msg = ("Could not run OS rename script for instance %s on node %s"
5029
               " (but the instance has been renamed in Ganeti): %s" %
5030
               (inst.name, inst.primary_node, msg))
5031
        self.proc.LogWarning(msg)
5032
    finally:
5033
      _ShutdownInstanceDisks(self, inst)
5034

    
5035
    return inst.name
5036

    
5037

    
5038
class LURemoveInstance(LogicalUnit):
5039
  """Remove an instance.
5040

5041
  """
5042
  HPATH = "instance-remove"
5043
  HTYPE = constants.HTYPE_INSTANCE
5044
  _OP_PARAMS = [
5045
    _PInstanceName,
5046
    ("ignore_failures", False, _TBool),
5047
    _PShutdownTimeout,
5048
    ]
5049
  REQ_BGL = False
5050

    
5051
  def ExpandNames(self):
5052
    self._ExpandAndLockInstance()
5053
    self.needed_locks[locking.LEVEL_NODE] = []
5054
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5055

    
5056
  def DeclareLocks(self, level):
5057
    if level == locking.LEVEL_NODE:
5058
      self._LockInstancesNodes()
5059

    
5060
  def BuildHooksEnv(self):
5061
    """Build hooks env.
5062

5063
    This runs on master, primary and secondary nodes of the instance.
5064

5065
    """
5066
    env = _BuildInstanceHookEnvByObject(self, self.instance)
5067
    env["SHUTDOWN_TIMEOUT"] = self.op.shutdown_timeout
5068
    nl = [self.cfg.GetMasterNode()]
5069
    nl_post = list(self.instance.all_nodes) + nl
5070
    return env, nl, nl_post
5071

    
5072
  def CheckPrereq(self):
5073
    """Check prerequisites.
5074

5075
    This checks that the instance is in the cluster.
5076

5077
    """
5078
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5079
    assert self.instance is not None, \
5080
      "Cannot retrieve locked instance %s" % self.op.instance_name
5081

    
5082
  def Exec(self, feedback_fn):
5083
    """Remove the instance.
5084

5085
    """
5086
    instance = self.instance
5087
    logging.info("Shutting down instance %s on node %s",
5088
                 instance.name, instance.primary_node)
5089

    
5090
    result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
5091
                                             self.op.shutdown_timeout)
5092
    msg = result.fail_msg
5093
    if msg:
5094
      if self.op.ignore_failures:
5095
        feedback_fn("Warning: can't shutdown instance: %s" % msg)
5096
      else:
5097
        raise errors.OpExecError("Could not shutdown instance %s on"
5098
                                 " node %s: %s" %
5099
                                 (instance.name, instance.primary_node, msg))
5100

    
5101
    _RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures)
5102

    
5103

    
5104
def _RemoveInstance(lu, feedback_fn, instance, ignore_failures):
5105
  """Utility function to remove an instance.
5106

5107
  """
5108
  logging.info("Removing block devices for instance %s", instance.name)
5109

    
5110
  if not _RemoveDisks(lu, instance):
5111
    if not ignore_failures:
5112
      raise errors.OpExecError("Can't remove instance's disks")
5113
    feedback_fn("Warning: can't remove instance's disks")
5114

    
5115
  logging.info("Removing instance %s out of cluster config", instance.name)
5116

    
5117
  lu.cfg.RemoveInstance(instance.name)
5118

    
5119
  assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \
5120
    "Instance lock removal conflict"
5121

    
5122
  # Remove lock for the instance
5123
  lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name
5124

    
5125

    
5126
class LUQueryInstances(NoHooksLU):
5127
  """Logical unit for querying instances.
5128

5129
  """
5130
  # pylint: disable-msg=W0142
5131
  _OP_PARAMS = [
5132
    ("output_fields", _NoDefault, _TListOf(_TNonEmptyString)),
5133
    ("names", _EmptyList, _TListOf(_TNonEmptyString)),
5134
    ("use_locking", False, _TBool),
5135
    ]
5136
  REQ_BGL = False
5137
  _SIMPLE_FIELDS = ["name", "os", "network_port", "hypervisor",
5138
                    "serial_no", "ctime", "mtime", "uuid"]
5139
  _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
5140
                                    "admin_state",
5141
                                    "disk_template", "ip", "mac", "bridge",
5142
                                    "nic_mode", "nic_link",
5143
                                    "sda_size", "sdb_size", "vcpus", "tags",
5144
                                    "network_port", "beparams",
5145
                                    r"(disk)\.(size)/([0-9]+)",
5146
                                    r"(disk)\.(sizes)", "disk_usage",
5147
                                    r"(nic)\.(mac|ip|mode|link)/([0-9]+)",
5148
                                    r"(nic)\.(bridge)/([0-9]+)",
5149
                                    r"(nic)\.(macs|ips|modes|links|bridges)",
5150
                                    r"(disk|nic)\.(count)",
5151
                                    "hvparams",
5152
                                    ] + _SIMPLE_FIELDS +
5153
                                  ["hv/%s" % name
5154
                                   for name in constants.HVS_PARAMETERS
5155
                                   if name not in constants.HVC_GLOBALS] +
5156
                                  ["be/%s" % name
5157
                                   for name in constants.BES_PARAMETERS])
5158
  _FIELDS_DYNAMIC = utils.FieldSet("oper_state",
5159
                                   "oper_ram",
5160
                                   "oper_vcpus",
5161
                                   "status")
5162

    
5163

    
5164
  def CheckArguments(self):
5165
    _CheckOutputFields(static=self._FIELDS_STATIC,
5166
                       dynamic=self._FIELDS_DYNAMIC,
5167
                       selected=self.op.output_fields)
5168

    
5169
  def ExpandNames(self):
5170
    self.needed_locks = {}
5171
    self.share_locks[locking.LEVEL_INSTANCE] = 1
5172
    self.share_locks[locking.LEVEL_NODE] = 1
5173

    
5174
    if self.op.names:
5175
      self.wanted = _GetWantedInstances(self, self.op.names)
5176
    else:
5177
      self.wanted = locking.ALL_SET
5178

    
5179
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
5180
    self.do_locking = self.do_node_query and self.op.use_locking
5181
    if self.do_locking:
5182
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
5183
      self.needed_locks[locking.LEVEL_NODE] = []
5184
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5185

    
5186
  def DeclareLocks(self, level):
5187
    if level == locking.LEVEL_NODE and self.do_locking:
5188
      self._LockInstancesNodes()
5189

    
5190
  def Exec(self, feedback_fn):
5191
    """Computes the list of nodes and their attributes.
5192

5193
    """
5194
    # pylint: disable-msg=R0912
5195
    # way too many branches here
5196
    all_info = self.cfg.GetAllInstancesInfo()
5197
    if self.wanted == locking.ALL_SET:
5198
      # caller didn't specify instance names, so ordering is not important
5199
      if self.do_locking:
5200
        instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
5201
      else:
5202
        instance_names = all_info.keys()
5203
      instance_names = utils.NiceSort(instance_names)
5204
    else:
5205
      # caller did specify names, so we must keep the ordering
5206
      if self.do_locking:
5207
        tgt_set = self.acquired_locks[locking.LEVEL_INSTANCE]
5208
      else:
5209
        tgt_set = all_info.keys()
5210
      missing = set(self.wanted).difference(tgt_set)
5211
      if missing:
5212
        raise errors.OpExecError("Some instances were removed before"
5213
                                 " retrieving their data: %s" % missing)
5214
      instance_names = self.wanted
5215

    
5216
    instance_list = [all_info[iname] for iname in instance_names]
5217

    
5218
    # begin data gathering
5219

    
5220
    nodes = frozenset([inst.primary_node for inst in instance_list])
5221
    hv_list = list(set([inst.hypervisor for inst in instance_list]))
5222

    
5223
    bad_nodes = []
5224
    off_nodes = []
5225
    if self.do_node_query:
5226
      live_data = {}
5227
      node_data = self.rpc.call_all_instances_info(nodes, hv_list)
5228
      for name in nodes:
5229
        result = node_data[name]
5230
        if result.offline:
5231
          # offline nodes will be in both lists
5232
          off_nodes.append(name)
5233
        if result.fail_msg:
5234
          bad_nodes.append(name)
5235
        else:
5236
          if result.payload:
5237
            live_data.update(result.payload)
5238
          # else no instance is alive
5239
    else:
5240
      live_data = dict([(name, {}) for name in instance_names])
5241

    
5242
    # end data gathering
5243

    
5244
    HVPREFIX = "hv/"
5245
    BEPREFIX = "be/"
5246
    output = []
5247
    cluster = self.cfg.GetClusterInfo()
5248
    for instance in instance_list:
5249
      iout = []
5250
      i_hv = cluster.FillHV(instance, skip_globals=True)
5251
      i_be = cluster.FillBE(instance)
5252
      i_nicp = [cluster.SimpleFillNIC(nic.nicparams) for nic in instance.nics]
5253
      for field in self.op.output_fields:
5254
        st_match = self._FIELDS_STATIC.Matches(field)
5255
        if field in self._SIMPLE_FIELDS:
5256
          val = getattr(instance, field)
5257
        elif field == "pnode":
5258
          val = instance.primary_node
5259
        elif field == "snodes":
5260
          val = list(instance.secondary_nodes)
5261
        elif field == "admin_state":
5262
          val = instance.admin_up
5263
        elif field == "oper_state":
5264
          if instance.primary_node in bad_nodes:
5265
            val = None
5266
          else:
5267
            val = bool(live_data.get(instance.name))
5268
        elif field == "status":
5269
          if instance.primary_node in off_nodes:
5270
            val = "ERROR_nodeoffline"
5271
          elif instance.primary_node in bad_nodes:
5272
            val = "ERROR_nodedown"
5273
          else:
5274
            running = bool(live_data.get(instance.name))
5275
            if running:
5276
              if instance.admin_up:
5277
                val = "running"
5278
              else:
5279
                val = "ERROR_up"
5280
            else:
5281
              if instance.admin_up:
5282
                val = "ERROR_down"
5283
              else:
5284
                val = "ADMIN_down"
5285
        elif field == "oper_ram":
5286
          if instance.primary_node in bad_nodes:
5287
            val = None
5288
          elif instance.name in live_data:
5289
            val = live_data[instance.name].get("memory", "?")
5290
          else:
5291
            val = "-"
5292
        elif field == "oper_vcpus":
5293
          if instance.primary_node in bad_nodes:
5294
            val = None
5295
          elif instance.name in live_data:
5296
            val = live_data[instance.name].get("vcpus", "?")
5297
          else:
5298
            val = "-"
5299
        elif field == "vcpus":
5300
          val = i_be[constants.BE_VCPUS]
5301
        elif field == "disk_template":
5302
          val = instance.disk_template
5303
        elif field == "ip":
5304
          if instance.nics:
5305
            val = instance.nics[0].ip
5306
          else:
5307
            val = None
5308
        elif field == "nic_mode":
5309
          if instance.nics:
5310
            val = i_nicp[0][constants.NIC_MODE]
5311
          else:
5312
            val = None
5313
        elif field == "nic_link":
5314
          if instance.nics:
5315
            val = i_nicp[0][constants.NIC_LINK]
5316
          else:
5317
            val = None
5318
        elif field == "bridge":
5319
          if (instance.nics and
5320
              i_nicp[0][constants.NIC_MODE] == constants.NIC_MODE_BRIDGED):
5321
            val = i_nicp[0][constants.NIC_LINK]
5322
          else:
5323
            val = None
5324
        elif field == "mac":
5325
          if instance.nics:
5326
            val = instance.nics[0].mac
5327
          else:
5328
            val = None
5329
        elif field == "sda_size" or field == "sdb_size":
5330
          idx = ord(field[2]) - ord('a')
5331
          try:
5332
            val = instance.FindDisk(idx).size
5333
          except errors.OpPrereqError:
5334
            val = None
5335
        elif field == "disk_usage": # total disk usage per node
5336
          disk_sizes = [{'size': disk.size} for disk in instance.disks]
5337
          val = _ComputeDiskSize(instance.disk_template, disk_sizes)
5338
        elif field == "tags":
5339
          val = list(instance.GetTags())
5340
        elif field == "hvparams":
5341
          val = i_hv
5342
        elif (field.startswith(HVPREFIX) and
5343
              field[len(HVPREFIX):] in constants.HVS_PARAMETERS and
5344
              field[len(HVPREFIX):] not in constants.HVC_GLOBALS):
5345
          val = i_hv.get(field[len(HVPREFIX):], None)
5346
        elif field == "beparams":
5347
          val = i_be
5348
        elif (field.startswith(BEPREFIX) and
5349
              field[len(BEPREFIX):] in constants.BES_PARAMETERS):
5350
          val = i_be.get(field[len(BEPREFIX):], None)
5351
        elif st_match and st_match.groups():
5352
          # matches a variable list
5353
          st_groups = st_match.groups()
5354
          if st_groups and st_groups[0] == "disk":
5355
            if st_groups[1] == "count":
5356
              val = len(instance.disks)
5357
            elif st_groups[1] == "sizes":
5358
              val = [disk.size for disk in instance.disks]
5359
            elif st_groups[1] == "size":
5360
              try:
5361
                val = instance.FindDisk(st_groups[2]).size
5362
              except errors.OpPrereqError:
5363
                val = None
5364
            else:
5365
              assert False, "Unhandled disk parameter"
5366
          elif st_groups[0] == "nic":
5367
            if st_groups[1] == "count":
5368
              val = len(instance.nics)
5369
            elif st_groups[1] == "macs":
5370
              val = [nic.mac for nic in instance.nics]
5371
            elif st_groups[1] == "ips":
5372
              val = [nic.ip for nic in instance.nics]
5373
            elif st_groups[1] == "modes":
5374
              val = [nicp[constants.NIC_MODE] for nicp in i_nicp]
5375
            elif st_groups[1] == "links":
5376
              val = [nicp[constants.NIC_LINK] for nicp in i_nicp]
5377
            elif st_groups[1] == "bridges":
5378
              val = []
5379
              for nicp in i_nicp:
5380
                if nicp[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
5381
                  val.append(nicp[constants.NIC_LINK])
5382
                else:
5383
                  val.append(None)
5384
            else:
5385
              # index-based item
5386
              nic_idx = int(st_groups[2])
5387
              if nic_idx >= len(instance.nics):
5388
                val = None
5389
              else:
5390
                if st_groups[1] == "mac":
5391
                  val = instance.nics[nic_idx].mac
5392
                elif st_groups[1] == "ip":
5393
                  val = instance.nics[nic_idx].ip
5394
                elif st_groups[1] == "mode":
5395
                  val = i_nicp[nic_idx][constants.NIC_MODE]
5396
                elif st_groups[1] == "link":
5397
                  val = i_nicp[nic_idx][constants.NIC_LINK]
5398
                elif st_groups[1] == "bridge":
5399
                  nic_mode = i_nicp[nic_idx][constants.NIC_MODE]
5400
                  if nic_mode == constants.NIC_MODE_BRIDGED:
5401
                    val = i_nicp[nic_idx][constants.NIC_LINK]
5402
                  else:
5403
                    val = None
5404
                else:
5405
                  assert False, "Unhandled NIC parameter"
5406
          else:
5407
            assert False, ("Declared but unhandled variable parameter '%s'" %
5408
                           field)
5409
        else:
5410
          assert False, "Declared but unhandled parameter '%s'" % field
5411
        iout.append(val)
5412
      output.append(iout)
5413

    
5414
    return output
5415

    
5416

    
5417
class LUFailoverInstance(LogicalUnit):
5418
  """Failover an instance.
5419

5420
  """
5421
  HPATH = "instance-failover"
5422
  HTYPE = constants.HTYPE_INSTANCE
5423
  _OP_PARAMS = [
5424
    _PInstanceName,
5425
    ("ignore_consistency", False, _TBool),
5426
    _PShutdownTimeout,
5427
    ]
5428
  REQ_BGL = False
5429

    
5430
  def ExpandNames(self):
5431
    self._ExpandAndLockInstance()
5432
    self.needed_locks[locking.LEVEL_NODE] = []
5433
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5434

    
5435
  def DeclareLocks(self, level):
5436
    if level == locking.LEVEL_NODE:
5437
      self._LockInstancesNodes()
5438

    
5439
  def BuildHooksEnv(self):
5440
    """Build hooks env.
5441

5442
    This runs on master, primary and secondary nodes of the instance.
5443

5444
    """
5445
    instance = self.instance
5446
    source_node = instance.primary_node
5447
    target_node = instance.secondary_nodes[0]
5448
    env = {
5449
      "IGNORE_CONSISTENCY": self.op.ignore_consistency,
5450
      "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5451
      "OLD_PRIMARY": source_node,
5452
      "OLD_SECONDARY": target_node,
5453
      "NEW_PRIMARY": target_node,
5454
      "NEW_SECONDARY": source_node,
5455
      }
5456
    env.update(_BuildInstanceHookEnvByObject(self, instance))
5457
    nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5458
    nl_post = list(nl)
5459
    nl_post.append(source_node)
5460
    return env, nl, nl_post
5461

    
5462
  def CheckPrereq(self):
5463
    """Check prerequisites.
5464

5465
    This checks that the instance is in the cluster.
5466

5467
    """
5468
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5469
    assert self.instance is not None, \
5470
      "Cannot retrieve locked instance %s" % self.op.instance_name
5471

    
5472
    bep = self.cfg.GetClusterInfo().FillBE(instance)
5473
    if instance.disk_template not in constants.DTS_NET_MIRROR:
5474
      raise errors.OpPrereqError("Instance's disk layout is not"
5475
                                 " network mirrored, cannot failover.",
5476
                                 errors.ECODE_STATE)
5477

    
5478
    secondary_nodes = instance.secondary_nodes
5479
    if not secondary_nodes:
5480
      raise errors.ProgrammerError("no secondary node but using "
5481
                                   "a mirrored disk template")
5482

    
5483
    target_node = secondary_nodes[0]
5484
    _CheckNodeOnline(self, target_node)
5485
    _CheckNodeNotDrained(self, target_node)
5486
    if instance.admin_up:
5487
      # check memory requirements on the secondary node
5488
      _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5489
                           instance.name, bep[constants.BE_MEMORY],
5490
                           instance.hypervisor)
5491
    else:
5492
      self.LogInfo("Not checking memory on the secondary node as"
5493
                   " instance will not be started")
5494

    
5495
    # check bridge existance
5496
    _CheckInstanceBridgesExist(self, instance, node=target_node)
5497

    
5498
  def Exec(self, feedback_fn):
5499
    """Failover an instance.
5500

5501
    The failover is done by shutting it down on its present node and
5502
    starting it on the secondary.
5503

5504
    """
5505
    instance = self.instance
5506

    
5507
    source_node = instance.primary_node
5508
    target_node = instance.secondary_nodes[0]
5509

    
5510
    if instance.admin_up:
5511
      feedback_fn("* checking disk consistency between source and target")
5512
      for dev in instance.disks:
5513
        # for drbd, these are drbd over lvm
5514
        if not _CheckDiskConsistency(self, dev, target_node, False):
5515
          if not self.op.ignore_consistency:
5516
            raise errors.OpExecError("Disk %s is degraded on target node,"
5517
                                     " aborting failover." % dev.iv_name)
5518
    else:
5519
      feedback_fn("* not checking disk consistency as instance is not running")
5520

    
5521
    feedback_fn("* shutting down instance on source node")
5522
    logging.info("Shutting down instance %s on node %s",
5523
                 instance.name, source_node)
5524

    
5525
    result = self.rpc.call_instance_shutdown(source_node, instance,
5526
                                             self.op.shutdown_timeout)
5527
    msg = result.fail_msg
5528
    if msg:
5529
      if self.op.ignore_consistency:
5530
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
5531
                             " Proceeding anyway. Please make sure node"
5532
                             " %s is down. Error details: %s",
5533
                             instance.name, source_node, source_node, msg)
5534
      else:
5535
        raise errors.OpExecError("Could not shutdown instance %s on"
5536
                                 " node %s: %s" %
5537
                                 (instance.name, source_node, msg))
5538

    
5539
    feedback_fn("* deactivating the instance's disks on source node")
5540
    if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
5541
      raise errors.OpExecError("Can't shut down the instance's disks.")
5542

    
5543
    instance.primary_node = target_node
5544
    # distribute new instance config to the other nodes
5545
    self.cfg.Update(instance, feedback_fn)
5546

    
5547
    # Only start the instance if it's marked as up
5548
    if instance.admin_up:
5549
      feedback_fn("* activating the instance's disks on target node")
5550
      logging.info("Starting instance %s on node %s",
5551
                   instance.name, target_node)
5552

    
5553
      disks_ok, _ = _AssembleInstanceDisks(self, instance,
5554
                                           ignore_secondaries=True)
5555
      if not disks_ok:
5556
        _ShutdownInstanceDisks(self, instance)
5557
        raise errors.OpExecError("Can't activate the instance's disks")
5558

    
5559
      feedback_fn("* starting the instance on the target node")
5560
      result = self.rpc.call_instance_start(target_node, instance, None, None)
5561
      msg = result.fail_msg
5562
      if msg:
5563
        _ShutdownInstanceDisks(self, instance)
5564
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
5565
                                 (instance.name, target_node, msg))
5566

    
5567

    
5568
class LUMigrateInstance(LogicalUnit):
5569
  """Migrate an instance.
5570

5571
  This is migration without shutting down, compared to the failover,
5572
  which is done with shutdown.
5573

5574
  """
5575
  HPATH = "instance-migrate"
5576
  HTYPE = constants.HTYPE_INSTANCE
5577
  _OP_PARAMS = [
5578
    _PInstanceName,
5579
    _PMigrationMode,
5580
    _PMigrationLive,
5581
    ("cleanup", False, _TBool),
5582
    ]
5583

    
5584
  REQ_BGL = False
5585

    
5586
  def ExpandNames(self):
5587
    self._ExpandAndLockInstance()
5588

    
5589
    self.needed_locks[locking.LEVEL_NODE] = []
5590
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5591

    
5592
    self._migrater = TLMigrateInstance(self, self.op.instance_name,
5593
                                       self.op.cleanup)
5594
    self.tasklets = [self._migrater]
5595

    
5596
  def DeclareLocks(self, level):
5597
    if level == locking.LEVEL_NODE:
5598
      self._LockInstancesNodes()
5599

    
5600
  def BuildHooksEnv(self):
5601
    """Build hooks env.
5602

5603
    This runs on master, primary and secondary nodes of the instance.
5604

5605
    """
5606
    instance = self._migrater.instance
5607
    source_node = instance.primary_node
5608
    target_node = instance.secondary_nodes[0]
5609
    env = _BuildInstanceHookEnvByObject(self, instance)
5610
    env["MIGRATE_LIVE"] = self._migrater.live
5611
    env["MIGRATE_CLEANUP"] = self.op.cleanup
5612
    env.update({
5613
        "OLD_PRIMARY": source_node,
5614
        "OLD_SECONDARY": target_node,
5615
        "NEW_PRIMARY": target_node,
5616
        "NEW_SECONDARY": source_node,
5617
        })
5618
    nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5619
    nl_post = list(nl)
5620
    nl_post.append(source_node)
5621
    return env, nl, nl_post
5622

    
5623

    
5624
class LUMoveInstance(LogicalUnit):
5625
  """Move an instance by data-copying.
5626

5627
  """
5628
  HPATH = "instance-move"
5629
  HTYPE = constants.HTYPE_INSTANCE
5630
  _OP_PARAMS = [
5631
    _PInstanceName,
5632
    ("target_node", _NoDefault, _TNonEmptyString),
5633
    _PShutdownTimeout,
5634
    ]
5635
  REQ_BGL = False
5636

    
5637
  def ExpandNames(self):
5638
    self._ExpandAndLockInstance()
5639
    target_node = _ExpandNodeName(self.cfg, self.op.target_node)
5640
    self.op.target_node = target_node
5641
    self.needed_locks[locking.LEVEL_NODE] = [target_node]
5642
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5643

    
5644
  def DeclareLocks(self, level):
5645
    if level == locking.LEVEL_NODE:
5646
      self._LockInstancesNodes(primary_only=True)
5647

    
5648
  def BuildHooksEnv(self):
5649
    """Build hooks env.
5650

5651
    This runs on master, primary and secondary nodes of the instance.
5652

5653
    """
5654
    env = {
5655
      "TARGET_NODE": self.op.target_node,
5656
      "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
5657
      }
5658
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5659
    nl = [self.cfg.GetMasterNode()] + [self.instance.primary_node,
5660
                                       self.op.target_node]
5661
    return env, nl, nl
5662

    
5663
  def CheckPrereq(self):
5664
    """Check prerequisites.
5665

5666
    This checks that the instance is in the cluster.
5667

5668
    """
5669
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5670
    assert self.instance is not None, \
5671
      "Cannot retrieve locked instance %s" % self.op.instance_name
5672

    
5673
    node = self.cfg.GetNodeInfo(self.op.target_node)
5674
    assert node is not None, \
5675
      "Cannot retrieve locked node %s" % self.op.target_node
5676

    
5677
    self.target_node = target_node = node.name
5678

    
5679
    if target_node == instance.primary_node:
5680
      raise errors.OpPrereqError("Instance %s is already on the node %s" %
5681
                                 (instance.name, target_node),
5682
                                 errors.ECODE_STATE)
5683

    
5684
    bep = self.cfg.GetClusterInfo().FillBE(instance)
5685

    
5686
    for idx, dsk in enumerate(instance.disks):
5687
      if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
5688
        raise errors.OpPrereqError("Instance disk %d has a complex layout,"
5689
                                   " cannot copy" % idx, errors.ECODE_STATE)
5690

    
5691
    _CheckNodeOnline(self, target_node)
5692
    _CheckNodeNotDrained(self, target_node)
5693

    
5694
    if instance.admin_up:
5695
      # check memory requirements on the secondary node
5696
      _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5697
                           instance.name, bep[constants.BE_MEMORY],
5698
                           instance.hypervisor)
5699
    else:
5700
      self.LogInfo("Not checking memory on the secondary node as"
5701
                   " instance will not be started")
5702

    
5703
    # check bridge existance
5704
    _CheckInstanceBridgesExist(self, instance, node=target_node)
5705

    
5706
  def Exec(self, feedback_fn):
5707
    """Move an instance.
5708

5709
    The move is done by shutting it down on its present node, copying
5710
    the data over (slow) and starting it on the new node.
5711

5712
    """
5713
    instance = self.instance
5714

    
5715
    source_node = instance.primary_node
5716
    target_node = self.target_node
5717

    
5718
    self.LogInfo("Shutting down instance %s on source node %s",
5719
                 instance.name, source_node)
5720

    
5721
    result = self.rpc.call_instance_shutdown(source_node, instance,
5722
                                             self.op.shutdown_timeout)
5723
    msg = result.fail_msg
5724
    if msg:
5725
      if self.op.ignore_consistency:
5726
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
5727
                             " Proceeding anyway. Please make sure node"
5728
                             " %s is down. Error details: %s",
5729
                             instance.name, source_node, source_node, msg)
5730
      else:
5731
        raise errors.OpExecError("Could not shutdown instance %s on"
5732
                                 " node %s: %s" %
5733
                                 (instance.name, source_node, msg))
5734

    
5735
    # create the target disks
5736
    try:
5737
      _CreateDisks(self, instance, target_node=target_node)
5738
    except errors.OpExecError:
5739
      self.LogWarning("Device creation failed, reverting...")
5740
      try:
5741
        _RemoveDisks(self, instance, target_node=target_node)
5742
      finally:
5743
        self.cfg.ReleaseDRBDMinors(instance.name)
5744
        raise
5745

    
5746
    cluster_name = self.cfg.GetClusterInfo().cluster_name
5747

    
5748
    errs = []
5749
    # activate, get path, copy the data over
5750
    for idx, disk in enumerate(instance.disks):
5751
      self.LogInfo("Copying data for disk %d", idx)
5752
      result = self.rpc.call_blockdev_assemble(target_node, disk,
5753
                                               instance.name, True)
5754
      if result.fail_msg:
5755
        self.LogWarning("Can't assemble newly created disk %d: %s",
5756
                        idx, result.fail_msg)
5757
        errs.append(result.fail_msg)
5758
        break
5759
      dev_path = result.payload
5760
      result = self.rpc.call_blockdev_export(source_node, disk,
5761
                                             target_node, dev_path,
5762
                                             cluster_name)
5763
      if result.fail_msg:
5764
        self.LogWarning("Can't copy data over for disk %d: %s",
5765
                        idx, result.fail_msg)
5766
        errs.append(result.fail_msg)
5767
        break
5768

    
5769
    if errs:
5770
      self.LogWarning("Some disks failed to copy, aborting")
5771
      try:
5772
        _RemoveDisks(self, instance, target_node=target_node)
5773
      finally:
5774
        self.cfg.ReleaseDRBDMinors(instance.name)
5775
        raise errors.OpExecError("Errors during disk copy: %s" %
5776
                                 (",".join(errs),))
5777

    
5778
    instance.primary_node = target_node
5779
    self.cfg.Update(instance, feedback_fn)
5780

    
5781
    self.LogInfo("Removing the disks on the original node")
5782
    _RemoveDisks(self, instance, target_node=source_node)
5783

    
5784
    # Only start the instance if it's marked as up
5785
    if instance.admin_up:
5786
      self.LogInfo("Starting instance %s on node %s",
5787
                   instance.name, target_node)
5788

    
5789
      disks_ok, _ = _AssembleInstanceDisks(self, instance,
5790
                                           ignore_secondaries=True)
5791
      if not disks_ok:
5792
        _ShutdownInstanceDisks(self, instance)
5793
        raise errors.OpExecError("Can't activate the instance's disks")
5794

    
5795
      result = self.rpc.call_instance_start(target_node, instance, None, None)
5796
      msg = result.fail_msg
5797
      if msg:
5798
        _ShutdownInstanceDisks(self, instance)
5799
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
5800
                                 (instance.name, target_node, msg))
5801

    
5802

    
5803
class LUMigrateNode(LogicalUnit):
5804
  """Migrate all instances from a node.
5805

5806
  """
5807
  HPATH = "node-migrate"
5808
  HTYPE = constants.HTYPE_NODE
5809
  _OP_PARAMS = [
5810
    _PNodeName,
5811
    _PMigrationMode,
5812
    _PMigrationLive,
5813
    ]
5814
  REQ_BGL = False
5815

    
5816
  def ExpandNames(self):
5817
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
5818

    
5819
    self.needed_locks = {
5820
      locking.LEVEL_NODE: [self.op.node_name],
5821
      }
5822

    
5823
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5824

    
5825
    # Create tasklets for migrating instances for all instances on this node
5826
    names = []
5827
    tasklets = []
5828

    
5829
    for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name):
5830
      logging.debug("Migrating instance %s", inst.name)
5831
      names.append(inst.name)
5832

    
5833
      tasklets.append(TLMigrateInstance(self, inst.name, False))
5834

    
5835
    self.tasklets = tasklets
5836

    
5837
    # Declare instance locks
5838
    self.needed_locks[locking.LEVEL_INSTANCE] = names
5839

    
5840
  def DeclareLocks(self, level):
5841
    if level == locking.LEVEL_NODE:
5842
      self._LockInstancesNodes()
5843

    
5844
  def BuildHooksEnv(self):
5845
    """Build hooks env.
5846

5847
    This runs on the master, the primary and all the secondaries.
5848

5849
    """
5850
    env = {
5851
      "NODE_NAME": self.op.node_name,
5852
      }
5853

    
5854
    nl = [self.cfg.GetMasterNode()]
5855

    
5856
    return (env, nl, nl)
5857

    
5858

    
5859
class TLMigrateInstance(Tasklet):
5860
  """Tasklet class for instance migration.
5861

5862
  @type live: boolean
5863
  @ivar live: whether the migration will be done live or non-live;
5864
      this variable is initalized only after CheckPrereq has run
5865

5866
  """
5867
  def __init__(self, lu, instance_name, cleanup):
5868
    """Initializes this class.
5869

5870
    """
5871
    Tasklet.__init__(self, lu)
5872

    
5873
    # Parameters
5874
    self.instance_name = instance_name
5875
    self.cleanup = cleanup
5876
    self.live = False # will be overridden later
5877

    
5878
  def CheckPrereq(self):
5879
    """Check prerequisites.
5880

5881
    This checks that the instance is in the cluster.
5882

5883
    """
5884
    instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name)
5885
    instance = self.cfg.GetInstanceInfo(instance_name)
5886
    assert instance is not None
5887

    
5888
    if instance.disk_template != constants.DT_DRBD8:
5889
      raise errors.OpPrereqError("Instance's disk layout is not"
5890
                                 " drbd8, cannot migrate.", errors.ECODE_STATE)
5891

    
5892
    secondary_nodes = instance.secondary_nodes
5893
    if not secondary_nodes:
5894
      raise errors.ConfigurationError("No secondary node but using"
5895
                                      " drbd8 disk template")
5896

    
5897
    i_be = self.cfg.GetClusterInfo().FillBE(instance)
5898

    
5899
    target_node = secondary_nodes[0]
5900
    # check memory requirements on the secondary node
5901
    _CheckNodeFreeMemory(self.lu, target_node, "migrating instance %s" %
5902
                         instance.name, i_be[constants.BE_MEMORY],
5903
                         instance.hypervisor)
5904

    
5905
    # check bridge existance
5906
    _CheckInstanceBridgesExist(self.lu, instance, node=target_node)
5907

    
5908
    if not self.cleanup:
5909
      _CheckNodeNotDrained(self.lu, target_node)
5910
      result = self.rpc.call_instance_migratable(instance.primary_node,
5911
                                                 instance)
5912
      result.Raise("Can't migrate, please use failover",
5913
                   prereq=True, ecode=errors.ECODE_STATE)
5914

    
5915
    self.instance = instance
5916

    
5917
    if self.lu.op.live is not None and self.lu.op.mode is not None:
5918
      raise errors.OpPrereqError("Only one of the 'live' and 'mode'"
5919
                                 " parameters are accepted",
5920
                                 errors.ECODE_INVAL)
5921
    if self.lu.op.live is not None:
5922
      if self.lu.op.live:
5923
        self.lu.op.mode = constants.HT_MIGRATION_LIVE
5924
      else:
5925
        self.lu.op.mode = constants.HT_MIGRATION_NONLIVE
5926
      # reset the 'live' parameter to None so that repeated
5927
      # invocations of CheckPrereq do not raise an exception
5928
      self.lu.op.live = None
5929
    elif self.lu.op.mode is None:
5930
      # read the default value from the hypervisor
5931
      i_hv = self.cfg.GetClusterInfo().FillHV(instance, skip_globals=False)
5932
      self.lu.op.mode = i_hv[constants.HV_MIGRATION_MODE]
5933

    
5934
    self.live = self.lu.op.mode == constants.HT_MIGRATION_LIVE
5935

    
5936
  def _WaitUntilSync(self):
5937
    """Poll with custom rpc for disk sync.
5938

5939
    This uses our own step-based rpc call.
5940

5941
    """
5942
    self.feedback_fn("* wait until resync is done")
5943
    all_done = False
5944
    while not all_done:
5945
      all_done = True
5946
      result = self.rpc.call_drbd_wait_sync(self.all_nodes,
5947
                                            self.nodes_ip,
5948
                                            self.instance.disks)
5949
      min_percent = 100
5950
      for node, nres in result.items():
5951
        nres.Raise("Cannot resync disks on node %s" % node)
5952
        node_done, node_percent = nres.payload
5953
        all_done = all_done and node_done
5954
        if node_percent is not None:
5955
          min_percent = min(min_percent, node_percent)
5956
      if not all_done:
5957
        if min_percent < 100:
5958
          self.feedback_fn("   - progress: %.1f%%" % min_percent)
5959
        time.sleep(2)
5960

    
5961
  def _EnsureSecondary(self, node):
5962
    """Demote a node to secondary.
5963

5964
    """
5965
    self.feedback_fn("* switching node %s to secondary mode" % node)
5966

    
5967
    for dev in self.instance.disks:
5968
      self.cfg.SetDiskID(dev, node)
5969

    
5970
    result = self.rpc.call_blockdev_close(node, self.instance.name,
5971
                                          self.instance.disks)
5972
    result.Raise("Cannot change disk to secondary on node %s" % node)
5973

    
5974
  def _GoStandalone(self):
5975
    """Disconnect from the network.
5976

5977
    """
5978
    self.feedback_fn("* changing into standalone mode")
5979
    result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
5980
                                               self.instance.disks)
5981
    for node, nres in result.items():
5982
      nres.Raise("Cannot disconnect disks node %s" % node)
5983

    
5984
  def _GoReconnect(self, multimaster):
5985
    """Reconnect to the network.
5986

5987
    """
5988
    if multimaster:
5989
      msg = "dual-master"
5990
    else:
5991
      msg = "single-master"
5992
    self.feedback_fn("* changing disks into %s mode" % msg)
5993
    result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
5994
                                           self.instance.disks,
5995
                                           self.instance.name, multimaster)
5996
    for node, nres in result.items():
5997
      nres.Raise("Cannot change disks config on node %s" % node)
5998

    
5999
  def _ExecCleanup(self):
6000
    """Try to cleanup after a failed migration.
6001

6002
    The cleanup is done by:
6003
      - check that the instance is running only on one node
6004
        (and update the config if needed)
6005
      - change disks on its secondary node to secondary
6006
      - wait until disks are fully synchronized
6007
      - disconnect from the network
6008
      - change disks into single-master mode
6009
      - wait again until disks are fully synchronized
6010

6011
    """
6012
    instance = self.instance
6013
    target_node = self.target_node
6014
    source_node = self.source_node
6015

    
6016
    # check running on only one node
6017
    self.feedback_fn("* checking where the instance actually runs"
6018
                     " (if this hangs, the hypervisor might be in"
6019
                     " a bad state)")
6020
    ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
6021
    for node, result in ins_l.items():
6022
      result.Raise("Can't contact node %s" % node)
6023

    
6024
    runningon_source = instance.name in ins_l[source_node].payload
6025
    runningon_target = instance.name in ins_l[target_node].payload
6026

    
6027
    if runningon_source and runningon_target:
6028
      raise errors.OpExecError("Instance seems to be running on two nodes,"
6029
                               " or the hypervisor is confused. You will have"
6030
                               " to ensure manually that it runs only on one"
6031
                               " and restart this operation.")
6032

    
6033
    if not (runningon_source or runningon_target):
6034
      raise errors.OpExecError("Instance does not seem to be running at all."
6035
                               " In this case, it's safer to repair by"
6036
                               " running 'gnt-instance stop' to ensure disk"
6037
                               " shutdown, and then restarting it.")
6038

    
6039
    if runningon_target:
6040
      # the migration has actually succeeded, we need to update the config
6041
      self.feedback_fn("* instance running on secondary node (%s),"
6042
                       " updating config" % target_node)
6043
      instance.primary_node = target_node
6044
      self.cfg.Update(instance, self.feedback_fn)
6045
      demoted_node = source_node
6046
    else:
6047
      self.feedback_fn("* instance confirmed to be running on its"
6048
                       " primary node (%s)" % source_node)
6049
      demoted_node = target_node
6050

    
6051
    self._EnsureSecondary(demoted_node)
6052
    try:
6053
      self._WaitUntilSync()
6054
    except errors.OpExecError:
6055
      # we ignore here errors, since if the device is standalone, it
6056
      # won't be able to sync
6057
      pass
6058
    self._GoStandalone()
6059
    self._GoReconnect(False)
6060
    self._WaitUntilSync()
6061

    
6062
    self.feedback_fn("* done")
6063

    
6064
  def _RevertDiskStatus(self):
6065
    """Try to revert the disk status after a failed migration.
6066

6067
    """
6068
    target_node = self.target_node
6069
    try:
6070
      self._EnsureSecondary(target_node)
6071
      self._GoStandalone()
6072
      self._GoReconnect(False)
6073
      self._WaitUntilSync()
6074
    except errors.OpExecError, err:
6075
      self.lu.LogWarning("Migration failed and I can't reconnect the"
6076
                         " drives: error '%s'\n"
6077
                         "Please look and recover the instance status" %
6078
                         str(err))
6079

    
6080
  def _AbortMigration(self):
6081
    """Call the hypervisor code to abort a started migration.
6082

6083
    """
6084
    instance = self.instance
6085
    target_node = self.target_node
6086
    migration_info = self.migration_info
6087

    
6088
    abort_result = self.rpc.call_finalize_migration(target_node,
6089
                                                    instance,
6090
                                                    migration_info,
6091
                                                    False)
6092
    abort_msg = abort_result.fail_msg
6093
    if abort_msg:
6094
      logging.error("Aborting migration failed on target node %s: %s",
6095
                    target_node, abort_msg)
6096
      # Don't raise an exception here, as we stil have to try to revert the
6097
      # disk status, even if this step failed.
6098

    
6099
  def _ExecMigration(self):
6100
    """Migrate an instance.
6101

6102
    The migrate is done by:
6103
      - change the disks into dual-master mode
6104
      - wait until disks are fully synchronized again
6105
      - migrate the instance
6106
      - change disks on the new secondary node (the old primary) to secondary
6107
      - wait until disks are fully synchronized
6108
      - change disks into single-master mode
6109

6110
    """
6111
    instance = self.instance
6112
    target_node = self.target_node
6113
    source_node = self.source_node
6114

    
6115
    self.feedback_fn("* checking disk consistency between source and target")
6116
    for dev in instance.disks:
6117
      if not _CheckDiskConsistency(self.lu, dev, target_node, False):
6118
        raise errors.OpExecError("Disk %s is degraded or not fully"
6119
                                 " synchronized on target node,"
6120
                                 " aborting migrate." % dev.iv_name)
6121

    
6122
    # First get the migration information from the remote node
6123
    result = self.rpc.call_migration_info(source_node, instance)
6124
    msg = result.fail_msg
6125
    if msg:
6126
      log_err = ("Failed fetching source migration information from %s: %s" %
6127
                 (source_node, msg))
6128
      logging.error(log_err)
6129
      raise errors.OpExecError(log_err)
6130

    
6131
    self.migration_info = migration_info = result.payload
6132

    
6133
    # Then switch the disks to master/master mode
6134
    self._EnsureSecondary(target_node)
6135
    self._GoStandalone()
6136
    self._GoReconnect(True)
6137
    self._WaitUntilSync()
6138

    
6139
    self.feedback_fn("* preparing %s to accept the instance" % target_node)
6140
    result = self.rpc.call_accept_instance(target_node,
6141
                                           instance,
6142
                                           migration_info,
6143
                                           self.nodes_ip[target_node])
6144

    
6145
    msg = result.fail_msg
6146
    if msg:
6147
      logging.error("Instance pre-migration failed, trying to revert"
6148
                    " disk status: %s", msg)
6149
      self.feedback_fn("Pre-migration failed, aborting")
6150
      self._AbortMigration()
6151
      self._RevertDiskStatus()
6152
      raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
6153
                               (instance.name, msg))
6154

    
6155
    self.feedback_fn("* migrating instance to %s" % target_node)
6156
    time.sleep(10)
6157
    result = self.rpc.call_instance_migrate(source_node, instance,
6158
                                            self.nodes_ip[target_node],
6159
                                            self.live)
6160
    msg = result.fail_msg
6161
    if msg:
6162
      logging.error("Instance migration failed, trying to revert"
6163
                    " disk status: %s", msg)
6164
      self.feedback_fn("Migration failed, aborting")
6165
      self._AbortMigration()
6166
      self._RevertDiskStatus()
6167
      raise errors.OpExecError("Could not migrate instance %s: %s" %
6168
                               (instance.name, msg))
6169
    time.sleep(10)
6170

    
6171
    instance.primary_node = target_node
6172
    # distribute new instance config to the other nodes
6173
    self.cfg.Update(instance, self.feedback_fn)
6174

    
6175
    result = self.rpc.call_finalize_migration(target_node,
6176
                                              instance,
6177
                                              migration_info,
6178
                                              True)
6179
    msg = result.fail_msg
6180
    if msg:
6181
      logging.error("Instance migration succeeded, but finalization failed:"
6182
                    " %s", msg)
6183
      raise errors.OpExecError("Could not finalize instance migration: %s" %
6184
                               msg)
6185

    
6186
    self._EnsureSecondary(source_node)
6187
    self._WaitUntilSync()
6188
    self._GoStandalone()
6189
    self._GoReconnect(False)
6190
    self._WaitUntilSync()
6191

    
6192
    self.feedback_fn("* done")
6193

    
6194
  def Exec(self, feedback_fn):
6195
    """Perform the migration.
6196

6197
    """
6198
    feedback_fn("Migrating instance %s" % self.instance.name)
6199

    
6200
    self.feedback_fn = feedback_fn
6201

    
6202
    self.source_node = self.instance.primary_node
6203
    self.target_node = self.instance.secondary_nodes[0]
6204
    self.all_nodes = [self.source_node, self.target_node]
6205
    self.nodes_ip = {
6206
      self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
6207
      self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
6208
      }
6209

    
6210
    if self.cleanup:
6211
      return self._ExecCleanup()
6212
    else:
6213
      return self._ExecMigration()
6214

    
6215

    
6216
def _CreateBlockDev(lu, node, instance, device, force_create,
6217
                    info, force_open):
6218
  """Create a tree of block devices on a given node.
6219

6220
  If this device type has to be created on secondaries, create it and
6221
  all its children.
6222

6223
  If not, just recurse to children keeping the same 'force' value.
6224

6225
  @param lu: the lu on whose behalf we execute
6226
  @param node: the node on which to create the device
6227
  @type instance: L{objects.Instance}
6228
  @param instance: the instance which owns the device
6229
  @type device: L{objects.Disk}
6230
  @param device: the device to create
6231
  @type force_create: boolean
6232
  @param force_create: whether to force creation of this device; this
6233
      will be change to True whenever we find a device which has
6234
      CreateOnSecondary() attribute
6235
  @param info: the extra 'metadata' we should attach to the device
6236
      (this will be represented as a LVM tag)
6237
  @type force_open: boolean
6238
  @param force_open: this parameter will be passes to the
6239
      L{backend.BlockdevCreate} function where it specifies
6240
      whether we run on primary or not, and it affects both
6241
      the child assembly and the device own Open() execution
6242

6243
  """
6244
  if device.CreateOnSecondary():
6245
    force_create = True
6246

    
6247
  if device.children:
6248
    for child in device.children:
6249
      _CreateBlockDev(lu, node, instance, child, force_create,
6250
                      info, force_open)
6251

    
6252
  if not force_create:
6253
    return
6254

    
6255
  _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
6256

    
6257

    
6258
def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
6259
  """Create a single block device on a given node.
6260

6261
  This will not recurse over children of the device, so they must be
6262
  created in advance.
6263

6264
  @param lu: the lu on whose behalf we execute
6265
  @param node: the node on which to create the device
6266
  @type instance: L{objects.Instance}
6267
  @param instance: the instance which owns the device
6268
  @type device: L{objects.Disk}
6269
  @param device: the device to create
6270
  @param info: the extra 'metadata' we should attach to the device
6271
      (this will be represented as a LVM tag)
6272
  @type force_open: boolean
6273
  @param force_open: this parameter will be passes to the
6274
      L{backend.BlockdevCreate} function where it specifies
6275
      whether we run on primary or not, and it affects both
6276
      the child assembly and the device own Open() execution
6277

6278
  """
6279
  lu.cfg.SetDiskID(device, node)
6280
  result = lu.rpc.call_blockdev_create(node, device, device.size,
6281
                                       instance.name, force_open, info)
6282
  result.Raise("Can't create block device %s on"
6283
               " node %s for instance %s" % (device, node, instance.name))
6284
  if device.physical_id is None:
6285
    device.physical_id = result.payload
6286

    
6287

    
6288
def _GenerateUniqueNames(lu, exts):
6289
  """Generate a suitable LV name.
6290

6291
  This will generate a logical volume name for the given instance.
6292

6293
  """
6294
  results = []
6295
  for val in exts:
6296
    new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
6297
    results.append("%s%s" % (new_id, val))
6298
  return results
6299

    
6300

    
6301
def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
6302
                         p_minor, s_minor):
6303
  """Generate a drbd8 device complete with its children.
6304

6305
  """
6306
  port = lu.cfg.AllocatePort()
6307
  vgname = lu.cfg.GetVGName()
6308
  shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
6309
  dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
6310
                          logical_id=(vgname, names[0]))
6311
  dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
6312
                          logical_id=(vgname, names[1]))
6313
  drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
6314
                          logical_id=(primary, secondary, port,
6315
                                      p_minor, s_minor,
6316
                                      shared_secret),
6317
                          children=[dev_data, dev_meta],
6318
                          iv_name=iv_name)
6319
  return drbd_dev
6320

    
6321

    
6322
def _GenerateDiskTemplate(lu, template_name,
6323
                          instance_name, primary_node,
6324
                          secondary_nodes, disk_info,
6325
                          file_storage_dir, file_driver,
6326
                          base_index):
6327
  """Generate the entire disk layout for a given template type.
6328

6329
  """
6330
  #TODO: compute space requirements
6331

    
6332
  vgname = lu.cfg.GetVGName()
6333
  disk_count = len(disk_info)
6334
  disks = []
6335
  if template_name == constants.DT_DISKLESS:
6336
    pass
6337
  elif template_name == constants.DT_PLAIN:
6338
    if len(secondary_nodes) != 0:
6339
      raise errors.ProgrammerError("Wrong template configuration")
6340

    
6341
    names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
6342
                                      for i in range(disk_count)])
6343
    for idx, disk in enumerate(disk_info):
6344
      disk_index = idx + base_index
6345
      disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
6346
                              logical_id=(vgname, names[idx]),
6347
                              iv_name="disk/%d" % disk_index,
6348
                              mode=disk["mode"])
6349
      disks.append(disk_dev)
6350
  elif template_name == constants.DT_DRBD8:
6351
    if len(secondary_nodes) != 1:
6352
      raise errors.ProgrammerError("Wrong template configuration")
6353
    remote_node = secondary_nodes[0]
6354
    minors = lu.cfg.AllocateDRBDMinor(
6355
      [primary_node, remote_node] * len(disk_info), instance_name)
6356

    
6357
    names = []
6358
    for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
6359
                                               for i in range(disk_count)]):
6360
      names.append(lv_prefix + "_data")
6361
      names.append(lv_prefix + "_meta")
6362
    for idx, disk in enumerate(disk_info):
6363
      disk_index = idx + base_index
6364
      disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
6365
                                      disk["size"], names[idx*2:idx*2+2],
6366
                                      "disk/%d" % disk_index,
6367
                                      minors[idx*2], minors[idx*2+1])
6368
      disk_dev.mode = disk["mode"]
6369
      disks.append(disk_dev)
6370
  elif template_name == constants.DT_FILE:
6371
    if len(secondary_nodes) != 0:
6372
      raise errors.ProgrammerError("Wrong template configuration")
6373

    
6374
    _RequireFileStorage()
6375

    
6376
    for idx, disk in enumerate(disk_info):
6377
      disk_index = idx + base_index
6378
      disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
6379
                              iv_name="disk/%d" % disk_index,
6380
                              logical_id=(file_driver,
6381
                                          "%s/disk%d" % (file_storage_dir,
6382
                                                         disk_index)),
6383
                              mode=disk["mode"])
6384
      disks.append(disk_dev)
6385
  else:
6386
    raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
6387
  return disks
6388

    
6389

    
6390
def _GetInstanceInfoText(instance):
6391
  """Compute that text that should be added to the disk's metadata.
6392

6393
  """
6394
  return "originstname+%s" % instance.name
6395

    
6396

    
6397
def _CreateDisks(lu, instance, to_skip=None, target_node=None):
6398
  """Create all disks for an instance.
6399

6400
  This abstracts away some work from AddInstance.
6401

6402
  @type lu: L{LogicalUnit}
6403
  @param lu: the logical unit on whose behalf we execute
6404
  @type instance: L{objects.Instance}
6405
  @param instance: the instance whose disks we should create
6406
  @type to_skip: list
6407
  @param to_skip: list of indices to skip
6408
  @type target_node: string
6409
  @param target_node: if passed, overrides the target node for creation
6410
  @rtype: boolean
6411
  @return: the success of the creation
6412

6413
  """
6414
  info = _GetInstanceInfoText(instance)
6415
  if target_node is None:
6416
    pnode = instance.primary_node
6417
    all_nodes = instance.all_nodes
6418
  else:
6419
    pnode = target_node
6420
    all_nodes = [pnode]
6421

    
6422
  if instance.disk_template == constants.DT_FILE:
6423
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
6424
    result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
6425

    
6426
    result.Raise("Failed to create directory '%s' on"
6427
                 " node %s" % (file_storage_dir, pnode))
6428

    
6429
  # Note: this needs to be kept in sync with adding of disks in
6430
  # LUSetInstanceParams
6431
  for idx, device in enumerate(instance.disks):
6432
    if to_skip and idx in to_skip:
6433
      continue
6434
    logging.info("Creating volume %s for instance %s",
6435
                 device.iv_name, instance.name)
6436
    #HARDCODE
6437
    for node in all_nodes:
6438
      f_create = node == pnode
6439
      _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
6440

    
6441

    
6442
def _RemoveDisks(lu, instance, target_node=None):
6443
  """Remove all disks for an instance.
6444

6445
  This abstracts away some work from `AddInstance()` and
6446
  `RemoveInstance()`. Note that in case some of the devices couldn't
6447
  be removed, the removal will continue with the other ones (compare
6448
  with `_CreateDisks()`).
6449

6450
  @type lu: L{LogicalUnit}
6451
  @param lu: the logical unit on whose behalf we execute
6452
  @type instance: L{objects.Instance}
6453
  @param instance: the instance whose disks we should remove
6454
  @type target_node: string
6455
  @param target_node: used to override the node on which to remove the disks
6456
  @rtype: boolean
6457
  @return: the success of the removal
6458

6459
  """
6460
  logging.info("Removing block devices for instance %s", instance.name)
6461

    
6462
  all_result = True
6463
  for device in instance.disks:
6464
    if target_node:
6465
      edata = [(target_node, device)]
6466
    else:
6467
      edata = device.ComputeNodeTree(instance.primary_node)
6468
    for node, disk in edata:
6469
      lu.cfg.SetDiskID(disk, node)
6470
      msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
6471
      if msg:
6472
        lu.LogWarning("Could not remove block device %s on node %s,"
6473
                      " continuing anyway: %s", device.iv_name, node, msg)
6474
        all_result = False
6475

    
6476
  if instance.disk_template == constants.DT_FILE:
6477
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
6478
    if target_node:
6479
      tgt = target_node
6480
    else:
6481
      tgt = instance.primary_node
6482
    result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
6483
    if result.fail_msg:
6484
      lu.LogWarning("Could not remove directory '%s' on node %s: %s",
6485
                    file_storage_dir, instance.primary_node, result.fail_msg)
6486
      all_result = False
6487

    
6488
  return all_result
6489

    
6490

    
6491
def _ComputeDiskSize(disk_template, disks):
6492
  """Compute disk size requirements in the volume group
6493

6494
  """
6495
  # Required free disk space as a function of disk and swap space
6496
  req_size_dict = {
6497
    constants.DT_DISKLESS: None,
6498
    constants.DT_PLAIN: sum(d["size"] for d in disks),
6499
    # 128 MB are added for drbd metadata for each disk
6500
    constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
6501
    constants.DT_FILE: None,
6502
  }
6503

    
6504
  if disk_template not in req_size_dict:
6505
    raise errors.ProgrammerError("Disk template '%s' size requirement"
6506
                                 " is unknown" %  disk_template)
6507

    
6508
  return req_size_dict[disk_template]
6509

    
6510

    
6511
def _CheckHVParams(lu, nodenames, hvname, hvparams):
6512
  """Hypervisor parameter validation.
6513

6514
  This function abstract the hypervisor parameter validation to be
6515
  used in both instance create and instance modify.
6516

6517
  @type lu: L{LogicalUnit}
6518
  @param lu: the logical unit for which we check
6519
  @type nodenames: list
6520
  @param nodenames: the list of nodes on which we should check
6521
  @type hvname: string
6522
  @param hvname: the name of the hypervisor we should use
6523
  @type hvparams: dict
6524
  @param hvparams: the parameters which we need to check
6525
  @raise errors.OpPrereqError: if the parameters are not valid
6526

6527
  """
6528
  hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
6529
                                                  hvname,
6530
                                                  hvparams)
6531
  for node in nodenames:
6532
    info = hvinfo[node]
6533
    if info.offline:
6534
      continue
6535
    info.Raise("Hypervisor parameter validation failed on node %s" % node)
6536

    
6537

    
6538
def _CheckOSParams(lu, required, nodenames, osname, osparams):
6539
  """OS parameters validation.
6540

6541
  @type lu: L{LogicalUnit}
6542
  @param lu: the logical unit for which we check
6543
  @type required: boolean
6544
  @param required: whether the validation should fail if the OS is not
6545
      found
6546
  @type nodenames: list
6547
  @param nodenames: the list of nodes on which we should check
6548
  @type osname: string
6549
  @param osname: the name of the hypervisor we should use
6550
  @type osparams: dict
6551
  @param osparams: the parameters which we need to check
6552
  @raise errors.OpPrereqError: if the parameters are not valid
6553

6554
  """
6555
  result = lu.rpc.call_os_validate(required, nodenames, osname,
6556
                                   [constants.OS_VALIDATE_PARAMETERS],
6557
                                   osparams)
6558
  for node, nres in result.items():
6559
    # we don't check for offline cases since this should be run only
6560
    # against the master node and/or an instance's nodes
6561
    nres.Raise("OS Parameters validation failed on node %s" % node)
6562
    if not nres.payload:
6563
      lu.LogInfo("OS %s not found on node %s, validation skipped",
6564
                 osname, node)
6565

    
6566

    
6567
class LUCreateInstance(LogicalUnit):
6568
  """Create an instance.
6569

6570
  """
6571
  HPATH = "instance-add"
6572
  HTYPE = constants.HTYPE_INSTANCE
6573
  _OP_PARAMS = [
6574
    _PInstanceName,
6575
    ("mode", _NoDefault, _TElemOf(constants.INSTANCE_CREATE_MODES)),
6576
    ("start", True, _TBool),
6577
    ("wait_for_sync", True, _TBool),
6578
    ("ip_check", True, _TBool),
6579
    ("name_check", True, _TBool),
6580
    ("disks", _NoDefault, _TListOf(_TDict)),
6581
    ("nics", _NoDefault, _TListOf(_TDict)),
6582
    ("hvparams", _EmptyDict, _TDict),
6583
    ("beparams", _EmptyDict, _TDict),
6584
    ("osparams", _EmptyDict, _TDict),
6585
    ("no_install", None, _TMaybeBool),
6586
    ("os_type", None, _TMaybeString),
6587
    ("force_variant", False, _TBool),
6588
    ("source_handshake", None, _TOr(_TList, _TNone)),
6589
    ("source_x509_ca", None, _TMaybeString),
6590
    ("source_instance_name", None, _TMaybeString),
6591
    ("src_node", None, _TMaybeString),
6592
    ("src_path", None, _TMaybeString),
6593
    ("pnode", None, _TMaybeString),
6594
    ("snode", None, _TMaybeString),
6595
    ("iallocator", None, _TMaybeString),
6596
    ("hypervisor", None, _TMaybeString),
6597
    ("disk_template", _NoDefault, _CheckDiskTemplate),
6598
    ("identify_defaults", False, _TBool),
6599
    ("file_driver", None, _TOr(_TNone, _TElemOf(constants.FILE_DRIVER))),
6600
    ("file_storage_dir", None, _TMaybeString),
6601
    ]
6602
  REQ_BGL = False
6603

    
6604
  def CheckArguments(self):
6605
    """Check arguments.
6606

6607
    """
6608
    # do not require name_check to ease forward/backward compatibility
6609
    # for tools
6610
    if self.op.no_install and self.op.start:
6611
      self.LogInfo("No-installation mode selected, disabling startup")
6612
      self.op.start = False
6613
    # validate/normalize the instance name
6614
    self.op.instance_name = \
6615
      netutils.HostInfo.NormalizeName(self.op.instance_name)
6616

    
6617
    if self.op.ip_check and not self.op.name_check:
6618
      # TODO: make the ip check more flexible and not depend on the name check
6619
      raise errors.OpPrereqError("Cannot do ip check without a name check",
6620
                                 errors.ECODE_INVAL)
6621

    
6622
    # check nics' parameter names
6623
    for nic in self.op.nics:
6624
      utils.ForceDictType(nic, constants.INIC_PARAMS_TYPES)
6625

    
6626
    # check disks. parameter names and consistent adopt/no-adopt strategy
6627
    has_adopt = has_no_adopt = False
6628
    for disk in self.op.disks:
6629
      utils.ForceDictType(disk, constants.IDISK_PARAMS_TYPES)
6630
      if "adopt" in disk:
6631
        has_adopt = True
6632
      else:
6633
        has_no_adopt = True
6634
    if has_adopt and has_no_adopt:
6635
      raise errors.OpPrereqError("Either all disks are adopted or none is",
6636
                                 errors.ECODE_INVAL)
6637
    if has_adopt:
6638
      if self.op.disk_template not in constants.DTS_MAY_ADOPT:
6639
        raise errors.OpPrereqError("Disk adoption is not supported for the"
6640
                                   " '%s' disk template" %
6641
                                   self.op.disk_template,
6642
                                   errors.ECODE_INVAL)
6643
      if self.op.iallocator is not None:
6644
        raise errors.OpPrereqError("Disk adoption not allowed with an"
6645
                                   " iallocator script", errors.ECODE_INVAL)
6646
      if self.op.mode == constants.INSTANCE_IMPORT:
6647
        raise errors.OpPrereqError("Disk adoption not allowed for"
6648
                                   " instance import", errors.ECODE_INVAL)
6649

    
6650
    self.adopt_disks = has_adopt
6651

    
6652
    # instance name verification
6653
    if self.op.name_check:
6654
      self.hostname1 = netutils.GetHostInfo(self.op.instance_name)
6655
      self.op.instance_name = self.hostname1.name
6656
      # used in CheckPrereq for ip ping check
6657
      self.check_ip = self.hostname1.ip
6658
    elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
6659
      raise errors.OpPrereqError("Remote imports require names to be checked" %
6660
                                 errors.ECODE_INVAL)
6661
    else:
6662
      self.check_ip = None
6663

    
6664
    # file storage checks
6665
    if (self.op.file_driver and
6666
        not self.op.file_driver in constants.FILE_DRIVER):
6667
      raise errors.OpPrereqError("Invalid file driver name '%s'" %
6668
                                 self.op.file_driver, errors.ECODE_INVAL)
6669

    
6670
    if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
6671
      raise errors.OpPrereqError("File storage directory path not absolute",
6672
                                 errors.ECODE_INVAL)
6673

    
6674
    ### Node/iallocator related checks
6675
    _CheckIAllocatorOrNode(self, "iallocator", "pnode")
6676

    
6677
    if self.op.pnode is not None:
6678
      if self.op.disk_template in constants.DTS_NET_MIRROR:
6679
        if self.op.snode is None:
6680
          raise errors.OpPrereqError("The networked disk templates need"
6681
                                     " a mirror node", errors.ECODE_INVAL)
6682
      elif self.op.snode:
6683
        self.LogWarning("Secondary node will be ignored on non-mirrored disk"
6684
                        " template")
6685
        self.op.snode = None
6686

    
6687
    self._cds = _GetClusterDomainSecret()
6688

    
6689
    if self.op.mode == constants.INSTANCE_IMPORT:
6690
      # On import force_variant must be True, because if we forced it at
6691
      # initial install, our only chance when importing it back is that it
6692
      # works again!
6693
      self.op.force_variant = True
6694

    
6695
      if self.op.no_install:
6696
        self.LogInfo("No-installation mode has no effect during import")
6697

    
6698
    elif self.op.mode == constants.INSTANCE_CREATE:
6699
      if self.op.os_type is None:
6700
        raise errors.OpPrereqError("No guest OS specified",
6701
                                   errors.ECODE_INVAL)
6702
      if self.op.os_type in self.cfg.GetClusterInfo().blacklisted_os:
6703
        raise errors.OpPrereqError("Guest OS '%s' is not allowed for"
6704
                                   " installation" % self.op.os_type,
6705
                                   errors.ECODE_STATE)
6706
      if self.op.disk_template is None:
6707
        raise errors.OpPrereqError("No disk template specified",
6708
                                   errors.ECODE_INVAL)
6709

    
6710
    elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
6711
      # Check handshake to ensure both clusters have the same domain secret
6712
      src_handshake = self.op.source_handshake
6713
      if not src_handshake:
6714
        raise errors.OpPrereqError("Missing source handshake",
6715
                                   errors.ECODE_INVAL)
6716

    
6717
      errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds,
6718
                                                           src_handshake)
6719
      if errmsg:
6720
        raise errors.OpPrereqError("Invalid handshake: %s" % errmsg,
6721
                                   errors.ECODE_INVAL)
6722

    
6723
      # Load and check source CA
6724
      self.source_x509_ca_pem = self.op.source_x509_ca
6725
      if not self.source_x509_ca_pem:
6726
        raise errors.OpPrereqError("Missing source X509 CA",
6727
                                   errors.ECODE_INVAL)
6728

    
6729
      try:
6730
        (cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem,
6731
                                                    self._cds)
6732
      except OpenSSL.crypto.Error, err:
6733
        raise errors.OpPrereqError("Unable to load source X509 CA (%s)" %
6734
                                   (err, ), errors.ECODE_INVAL)
6735

    
6736
      (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
6737
      if errcode is not None:
6738
        raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ),
6739
                                   errors.ECODE_INVAL)
6740

    
6741
      self.source_x509_ca = cert
6742

    
6743
      src_instance_name = self.op.source_instance_name
6744
      if not src_instance_name:
6745
        raise errors.OpPrereqError("Missing source instance name",
6746
                                   errors.ECODE_INVAL)
6747

    
6748
      norm_name = netutils.HostInfo.NormalizeName(src_instance_name)
6749
      self.source_instance_name = netutils.GetHostInfo(norm_name).name
6750

    
6751
    else:
6752
      raise errors.OpPrereqError("Invalid instance creation mode %r" %
6753
                                 self.op.mode, errors.ECODE_INVAL)
6754

    
6755
  def ExpandNames(self):
6756
    """ExpandNames for CreateInstance.
6757

6758
    Figure out the right locks for instance creation.
6759

6760
    """
6761
    self.needed_locks = {}
6762

    
6763
    instance_name = self.op.instance_name
6764
    # this is just a preventive check, but someone might still add this
6765
    # instance in the meantime, and creation will fail at lock-add time
6766
    if instance_name in self.cfg.GetInstanceList():
6767
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
6768
                                 instance_name, errors.ECODE_EXISTS)
6769

    
6770
    self.add_locks[locking.LEVEL_INSTANCE] = instance_name
6771

    
6772
    if self.op.iallocator:
6773
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6774
    else:
6775
      self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode)
6776
      nodelist = [self.op.pnode]
6777
      if self.op.snode is not None:
6778
        self.op.snode = _ExpandNodeName(self.cfg, self.op.snode)
6779
        nodelist.append(self.op.snode)
6780
      self.needed_locks[locking.LEVEL_NODE] = nodelist
6781

    
6782
    # in case of import lock the source node too
6783
    if self.op.mode == constants.INSTANCE_IMPORT:
6784
      src_node = self.op.src_node
6785
      src_path = self.op.src_path
6786

    
6787
      if src_path is None:
6788
        self.op.src_path = src_path = self.op.instance_name
6789

    
6790
      if src_node is None:
6791
        self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6792
        self.op.src_node = None
6793
        if os.path.isabs(src_path):
6794
          raise errors.OpPrereqError("Importing an instance from an absolute"
6795
                                     " path requires a source node option.",
6796
                                     errors.ECODE_INVAL)
6797
      else:
6798
        self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node)
6799
        if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
6800
          self.needed_locks[locking.LEVEL_NODE].append(src_node)
6801
        if not os.path.isabs(src_path):
6802
          self.op.src_path = src_path = \
6803
            utils.PathJoin(constants.EXPORT_DIR, src_path)
6804

    
6805
  def _RunAllocator(self):
6806
    """Run the allocator based on input opcode.
6807

6808
    """
6809
    nics = [n.ToDict() for n in self.nics]
6810
    ial = IAllocator(self.cfg, self.rpc,
6811
                     mode=constants.IALLOCATOR_MODE_ALLOC,
6812
                     name=self.op.instance_name,
6813
                     disk_template=self.op.disk_template,
6814
                     tags=[],
6815
                     os=self.op.os_type,
6816
                     vcpus=self.be_full[constants.BE_VCPUS],
6817
                     mem_size=self.be_full[constants.BE_MEMORY],
6818
                     disks=self.disks,
6819
                     nics=nics,
6820
                     hypervisor=self.op.hypervisor,
6821
                     )
6822

    
6823
    ial.Run(self.op.iallocator)
6824

    
6825
    if not ial.success:
6826
      raise errors.OpPrereqError("Can't compute nodes using"
6827
                                 " iallocator '%s': %s" %
6828
                                 (self.op.iallocator, ial.info),
6829
                                 errors.ECODE_NORES)
6830
    if len(ial.result) != ial.required_nodes:
6831
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
6832
                                 " of nodes (%s), required %s" %
6833
                                 (self.op.iallocator, len(ial.result),
6834
                                  ial.required_nodes), errors.ECODE_FAULT)
6835
    self.op.pnode = ial.result[0]
6836
    self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
6837
                 self.op.instance_name, self.op.iallocator,
6838
                 utils.CommaJoin(ial.result))
6839
    if ial.required_nodes == 2:
6840
      self.op.snode = ial.result[1]
6841

    
6842
  def BuildHooksEnv(self):
6843
    """Build hooks env.
6844

6845
    This runs on master, primary and secondary nodes of the instance.
6846

6847
    """
6848
    env = {
6849
      "ADD_MODE": self.op.mode,
6850
      }
6851
    if self.op.mode == constants.INSTANCE_IMPORT:
6852
      env["SRC_NODE"] = self.op.src_node
6853
      env["SRC_PATH"] = self.op.src_path
6854
      env["SRC_IMAGES"] = self.src_images
6855

    
6856
    env.update(_BuildInstanceHookEnv(
6857
      name=self.op.instance_name,
6858
      primary_node=self.op.pnode,
6859
      secondary_nodes=self.secondaries,
6860
      status=self.op.start,
6861
      os_type=self.op.os_type,
6862
      memory=self.be_full[constants.BE_MEMORY],
6863
      vcpus=self.be_full[constants.BE_VCPUS],
6864
      nics=_NICListToTuple(self, self.nics),
6865
      disk_template=self.op.disk_template,
6866
      disks=[(d["size"], d["mode"]) for d in self.disks],
6867
      bep=self.be_full,
6868
      hvp=self.hv_full,
6869
      hypervisor_name=self.op.hypervisor,
6870
    ))
6871

    
6872
    nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
6873
          self.secondaries)
6874
    return env, nl, nl
6875

    
6876
  def _ReadExportInfo(self):
6877
    """Reads the export information from disk.
6878

6879
    It will override the opcode source node and path with the actual
6880
    information, if these two were not specified before.
6881

6882
    @return: the export information
6883

6884
    """
6885
    assert self.op.mode == constants.INSTANCE_IMPORT
6886

    
6887
    src_node = self.op.src_node
6888
    src_path = self.op.src_path
6889

    
6890
    if src_node is None:
6891
      locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
6892
      exp_list = self.rpc.call_export_list(locked_nodes)
6893
      found = False
6894
      for node in exp_list:
6895
        if exp_list[node].fail_msg:
6896
          continue
6897
        if src_path in exp_list[node].payload:
6898
          found = True
6899
          self.op.src_node = src_node = node
6900
          self.op.src_path = src_path = utils.PathJoin(constants.EXPORT_DIR,
6901
                                                       src_path)
6902
          break
6903
      if not found:
6904
        raise errors.OpPrereqError("No export found for relative path %s" %
6905
                                    src_path, errors.ECODE_INVAL)
6906

    
6907
    _CheckNodeOnline(self, src_node)
6908
    result = self.rpc.call_export_info(src_node, src_path)
6909
    result.Raise("No export or invalid export found in dir %s" % src_path)
6910

    
6911
    export_info = objects.SerializableConfigParser.Loads(str(result.payload))
6912
    if not export_info.has_section(constants.INISECT_EXP):
6913
      raise errors.ProgrammerError("Corrupted export config",
6914
                                   errors.ECODE_ENVIRON)
6915

    
6916
    ei_version = export_info.get(constants.INISECT_EXP, "version")
6917
    if (int(ei_version) != constants.EXPORT_VERSION):
6918
      raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
6919
                                 (ei_version, constants.EXPORT_VERSION),
6920
                                 errors.ECODE_ENVIRON)
6921
    return export_info
6922

    
6923
  def _ReadExportParams(self, einfo):
6924
    """Use export parameters as defaults.
6925

6926
    In case the opcode doesn't specify (as in override) some instance
6927
    parameters, then try to use them from the export information, if
6928
    that declares them.
6929

6930
    """
6931
    self.op.os_type = einfo.get(constants.INISECT_EXP, "os")
6932

    
6933
    if self.op.disk_template is None:
6934
      if einfo.has_option(constants.INISECT_INS, "disk_template"):
6935
        self.op.disk_template = einfo.get(constants.INISECT_INS,
6936
                                          "disk_template")
6937
      else:
6938
        raise errors.OpPrereqError("No disk template specified and the export"
6939
                                   " is missing the disk_template information",
6940
                                   errors.ECODE_INVAL)
6941

    
6942
    if not self.op.disks:
6943
      if einfo.has_option(constants.INISECT_INS, "disk_count"):
6944
        disks = []
6945
        # TODO: import the disk iv_name too
6946
        for idx in range(einfo.getint(constants.INISECT_INS, "disk_count")):
6947
          disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
6948
          disks.append({"size": disk_sz})
6949
        self.op.disks = disks
6950
      else:
6951
        raise errors.OpPrereqError("No disk info specified and the export"
6952
                                   " is missing the disk information",
6953
                                   errors.ECODE_INVAL)
6954

    
6955
    if (not self.op.nics and
6956
        einfo.has_option(constants.INISECT_INS, "nic_count")):
6957
      nics = []
6958
      for idx in range(einfo.getint(constants.INISECT_INS, "nic_count")):
6959
        ndict = {}
6960
        for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]:
6961
          v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
6962
          ndict[name] = v
6963
        nics.append(ndict)
6964
      self.op.nics = nics
6965

    
6966
    if (self.op.hypervisor is None and
6967
        einfo.has_option(constants.INISECT_INS, "hypervisor")):
6968
      self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor")
6969
    if einfo.has_section(constants.INISECT_HYP):
6970
      # use the export parameters but do not override the ones
6971
      # specified by the user
6972
      for name, value in einfo.items(constants.INISECT_HYP):
6973
        if name not in self.op.hvparams:
6974
          self.op.hvparams[name] = value
6975

    
6976
    if einfo.has_section(constants.INISECT_BEP):
6977
      # use the parameters, without overriding
6978
      for name, value in einfo.items(constants.INISECT_BEP):
6979
        if name not in self.op.beparams:
6980
          self.op.beparams[name] = value
6981
    else:
6982
      # try to read the parameters old style, from the main section
6983
      for name in constants.BES_PARAMETERS:
6984
        if (name not in self.op.beparams and
6985
            einfo.has_option(constants.INISECT_INS, name)):
6986
          self.op.beparams[name] = einfo.get(constants.INISECT_INS, name)
6987

    
6988
    if einfo.has_section(constants.INISECT_OSP):
6989
      # use the parameters, without overriding
6990
      for name, value in einfo.items(constants.INISECT_OSP):
6991
        if name not in self.op.osparams:
6992
          self.op.osparams[name] = value
6993

    
6994
  def _RevertToDefaults(self, cluster):
6995
    """Revert the instance parameters to the default values.
6996

6997
    """
6998
    # hvparams
6999
    hv_defs = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type, {})
7000
    for name in self.op.hvparams.keys():
7001
      if name in hv_defs and hv_defs[name] == self.op.hvparams[name]:
7002
        del self.op.hvparams[name]
7003
    # beparams
7004
    be_defs = cluster.SimpleFillBE({})
7005
    for name in self.op.beparams.keys():
7006
      if name in be_defs and be_defs[name] == self.op.beparams[name]:
7007
        del self.op.beparams[name]
7008
    # nic params
7009
    nic_defs = cluster.SimpleFillNIC({})
7010
    for nic in self.op.nics:
7011
      for name in constants.NICS_PARAMETERS:
7012
        if name in nic and name in nic_defs and nic[name] == nic_defs[name]:
7013
          del nic[name]
7014
    # osparams
7015
    os_defs = cluster.SimpleFillOS(self.op.os_type, {})
7016
    for name in self.op.osparams.keys():
7017
      if name in os_defs and os_defs[name] == self.op.osparams[name]:
7018
        del self.op.osparams[name]
7019

    
7020
  def CheckPrereq(self):
7021
    """Check prerequisites.
7022

7023
    """
7024
    if self.op.mode == constants.INSTANCE_IMPORT:
7025
      export_info = self._ReadExportInfo()
7026
      self._ReadExportParams(export_info)
7027

    
7028
    _CheckDiskTemplate(self.op.disk_template)
7029

    
7030
    if (not self.cfg.GetVGName() and
7031
        self.op.disk_template not in constants.DTS_NOT_LVM):
7032
      raise errors.OpPrereqError("Cluster does not support lvm-based"
7033
                                 " instances", errors.ECODE_STATE)
7034

    
7035
    if self.op.hypervisor is None:
7036
      self.op.hypervisor = self.cfg.GetHypervisorType()
7037

    
7038
    cluster = self.cfg.GetClusterInfo()
7039
    enabled_hvs = cluster.enabled_hypervisors
7040
    if self.op.hypervisor not in enabled_hvs:
7041
      raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
7042
                                 " cluster (%s)" % (self.op.hypervisor,
7043
                                  ",".join(enabled_hvs)),
7044
                                 errors.ECODE_STATE)
7045

    
7046
    # check hypervisor parameter syntax (locally)
7047
    utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
7048
    filled_hvp = cluster.SimpleFillHV(self.op.hypervisor, self.op.os_type,
7049
                                      self.op.hvparams)
7050
    hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
7051
    hv_type.CheckParameterSyntax(filled_hvp)
7052
    self.hv_full = filled_hvp
7053
    # check that we don't specify global parameters on an instance
7054
    _CheckGlobalHvParams(self.op.hvparams)
7055

    
7056
    # fill and remember the beparams dict
7057
    utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
7058
    self.be_full = cluster.SimpleFillBE(self.op.beparams)
7059

    
7060
    # build os parameters
7061
    self.os_full = cluster.SimpleFillOS(self.op.os_type, self.op.osparams)
7062

    
7063
    # now that hvp/bep are in final format, let's reset to defaults,
7064
    # if told to do so
7065
    if self.op.identify_defaults:
7066
      self._RevertToDefaults(cluster)
7067

    
7068
    # NIC buildup
7069
    self.nics = []
7070
    for idx, nic in enumerate(self.op.nics):
7071
      nic_mode_req = nic.get("mode", None)
7072
      nic_mode = nic_mode_req
7073
      if nic_mode is None:
7074
        nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
7075

    
7076
      # in routed mode, for the first nic, the default ip is 'auto'
7077
      if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
7078
        default_ip_mode = constants.VALUE_AUTO
7079
      else:
7080
        default_ip_mode = constants.VALUE_NONE
7081

    
7082
      # ip validity checks
7083
      ip = nic.get("ip", default_ip_mode)
7084
      if ip is None or ip.lower() == constants.VALUE_NONE:
7085
        nic_ip = None
7086
      elif ip.lower() == constants.VALUE_AUTO:
7087
        if not self.op.name_check:
7088
          raise errors.OpPrereqError("IP address set to auto but name checks"
7089
                                     " have been skipped. Aborting.",
7090
                                     errors.ECODE_INVAL)
7091
        nic_ip = self.hostname1.ip
7092
      else:
7093
        if not netutils.IsValidIP4(ip):
7094
          raise errors.OpPrereqError("Given IP address '%s' doesn't look"
7095
                                     " like a valid IP" % ip,
7096
                                     errors.ECODE_INVAL)
7097
        nic_ip = ip
7098

    
7099
      # TODO: check the ip address for uniqueness
7100
      if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
7101
        raise errors.OpPrereqError("Routed nic mode requires an ip address",
7102
                                   errors.ECODE_INVAL)
7103

    
7104
      # MAC address verification
7105
      mac = nic.get("mac", constants.VALUE_AUTO)
7106
      if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7107
        mac = utils.NormalizeAndValidateMac(mac)
7108

    
7109
        try:
7110
          self.cfg.ReserveMAC(mac, self.proc.GetECId())
7111
        except errors.ReservationError:
7112
          raise errors.OpPrereqError("MAC address %s already in use"
7113
                                     " in cluster" % mac,
7114
                                     errors.ECODE_NOTUNIQUE)
7115

    
7116
      # bridge verification
7117
      bridge = nic.get("bridge", None)
7118
      link = nic.get("link", None)
7119
      if bridge and link:
7120
        raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
7121
                                   " at the same time", errors.ECODE_INVAL)
7122
      elif bridge and nic_mode == constants.NIC_MODE_ROUTED:
7123
        raise errors.OpPrereqError("Cannot pass 'bridge' on a routed nic",
7124
                                   errors.ECODE_INVAL)
7125
      elif bridge:
7126
        link = bridge
7127

    
7128
      nicparams = {}
7129
      if nic_mode_req:
7130
        nicparams[constants.NIC_MODE] = nic_mode_req
7131
      if link:
7132
        nicparams[constants.NIC_LINK] = link
7133

    
7134
      check_params = cluster.SimpleFillNIC(nicparams)
7135
      objects.NIC.CheckParameterSyntax(check_params)
7136
      self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
7137

    
7138
    # disk checks/pre-build
7139
    self.disks = []
7140
    for disk in self.op.disks:
7141
      mode = disk.get("mode", constants.DISK_RDWR)
7142
      if mode not in constants.DISK_ACCESS_SET:
7143
        raise errors.OpPrereqError("Invalid disk access mode '%s'" %
7144
                                   mode, errors.ECODE_INVAL)
7145
      size = disk.get("size", None)
7146
      if size is None:
7147
        raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
7148
      try:
7149
        size = int(size)
7150
      except (TypeError, ValueError):
7151
        raise errors.OpPrereqError("Invalid disk size '%s'" % size,
7152
                                   errors.ECODE_INVAL)
7153
      new_disk = {"size": size, "mode": mode}
7154
      if "adopt" in disk:
7155
        new_disk["adopt"] = disk["adopt"]
7156
      self.disks.append(new_disk)
7157

    
7158
    if self.op.mode == constants.INSTANCE_IMPORT:
7159

    
7160
      # Check that the new instance doesn't have less disks than the export
7161
      instance_disks = len(self.disks)
7162
      export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
7163
      if instance_disks < export_disks:
7164
        raise errors.OpPrereqError("Not enough disks to import."
7165
                                   " (instance: %d, export: %d)" %
7166
                                   (instance_disks, export_disks),
7167
                                   errors.ECODE_INVAL)
7168

    
7169
      disk_images = []
7170
      for idx in range(export_disks):
7171
        option = 'disk%d_dump' % idx
7172
        if export_info.has_option(constants.INISECT_INS, option):
7173
          # FIXME: are the old os-es, disk sizes, etc. useful?
7174
          export_name = export_info.get(constants.INISECT_INS, option)
7175
          image = utils.PathJoin(self.op.src_path, export_name)
7176
          disk_images.append(image)
7177
        else:
7178
          disk_images.append(False)
7179

    
7180
      self.src_images = disk_images
7181

    
7182
      old_name = export_info.get(constants.INISECT_INS, 'name')
7183
      try:
7184
        exp_nic_count = export_info.getint(constants.INISECT_INS, 'nic_count')
7185
      except (TypeError, ValueError), err:
7186
        raise errors.OpPrereqError("Invalid export file, nic_count is not"
7187
                                   " an integer: %s" % str(err),
7188
                                   errors.ECODE_STATE)
7189
      if self.op.instance_name == old_name:
7190
        for idx, nic in enumerate(self.nics):
7191
          if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
7192
            nic_mac_ini = 'nic%d_mac' % idx
7193
            nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
7194

    
7195
    # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
7196

    
7197
    # ip ping checks (we use the same ip that was resolved in ExpandNames)
7198
    if self.op.ip_check:
7199
      if netutils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
7200
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
7201
                                   (self.check_ip, self.op.instance_name),
7202
                                   errors.ECODE_NOTUNIQUE)
7203

    
7204
    #### mac address generation
7205
    # By generating here the mac address both the allocator and the hooks get
7206
    # the real final mac address rather than the 'auto' or 'generate' value.
7207
    # There is a race condition between the generation and the instance object
7208
    # creation, which means that we know the mac is valid now, but we're not
7209
    # sure it will be when we actually add the instance. If things go bad
7210
    # adding the instance will abort because of a duplicate mac, and the
7211
    # creation job will fail.
7212
    for nic in self.nics:
7213
      if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7214
        nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
7215

    
7216
    #### allocator run
7217

    
7218
    if self.op.iallocator is not None:
7219
      self._RunAllocator()
7220

    
7221
    #### node related checks
7222

    
7223
    # check primary node
7224
    self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
7225
    assert self.pnode is not None, \
7226
      "Cannot retrieve locked node %s" % self.op.pnode
7227
    if pnode.offline:
7228
      raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
7229
                                 pnode.name, errors.ECODE_STATE)
7230
    if pnode.drained:
7231
      raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
7232
                                 pnode.name, errors.ECODE_STATE)
7233

    
7234
    self.secondaries = []
7235

    
7236
    # mirror node verification
7237
    if self.op.disk_template in constants.DTS_NET_MIRROR:
7238
      if self.op.snode == pnode.name:
7239
        raise errors.OpPrereqError("The secondary node cannot be the"
7240
                                   " primary node.", errors.ECODE_INVAL)
7241
      _CheckNodeOnline(self, self.op.snode)
7242
      _CheckNodeNotDrained(self, self.op.snode)
7243
      self.secondaries.append(self.op.snode)
7244

    
7245
    nodenames = [pnode.name] + self.secondaries
7246

    
7247
    req_size = _ComputeDiskSize(self.op.disk_template,
7248
                                self.disks)
7249

    
7250
    # Check lv size requirements, if not adopting
7251
    if req_size is not None and not self.adopt_disks:
7252
      _CheckNodesFreeDisk(self, nodenames, req_size)
7253

    
7254
    if self.adopt_disks: # instead, we must check the adoption data
7255
      all_lvs = set([i["adopt"] for i in self.disks])
7256
      if len(all_lvs) != len(self.disks):
7257
        raise errors.OpPrereqError("Duplicate volume names given for adoption",
7258
                                   errors.ECODE_INVAL)
7259
      for lv_name in all_lvs:
7260
        try:
7261
          self.cfg.ReserveLV(lv_name, self.proc.GetECId())
7262
        except errors.ReservationError:
7263
          raise errors.OpPrereqError("LV named %s used by another instance" %
7264
                                     lv_name, errors.ECODE_NOTUNIQUE)
7265

    
7266
      node_lvs = self.rpc.call_lv_list([pnode.name],
7267
                                       self.cfg.GetVGName())[pnode.name]
7268
      node_lvs.Raise("Cannot get LV information from node %s" % pnode.name)
7269
      node_lvs = node_lvs.payload
7270
      delta = all_lvs.difference(node_lvs.keys())
7271
      if delta:
7272
        raise errors.OpPrereqError("Missing logical volume(s): %s" %
7273
                                   utils.CommaJoin(delta),
7274
                                   errors.ECODE_INVAL)
7275
      online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]]
7276
      if online_lvs:
7277
        raise errors.OpPrereqError("Online logical volumes found, cannot"
7278
                                   " adopt: %s" % utils.CommaJoin(online_lvs),
7279
                                   errors.ECODE_STATE)
7280
      # update the size of disk based on what is found
7281
      for dsk in self.disks:
7282
        dsk["size"] = int(float(node_lvs[dsk["adopt"]][0]))
7283

    
7284
    _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
7285

    
7286
    _CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant)
7287
    # check OS parameters (remotely)
7288
    _CheckOSParams(self, True, nodenames, self.op.os_type, self.os_full)
7289

    
7290
    _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
7291

    
7292
    # memory check on primary node
7293
    if self.op.start:
7294
      _CheckNodeFreeMemory(self, self.pnode.name,
7295
                           "creating instance %s" % self.op.instance_name,
7296
                           self.be_full[constants.BE_MEMORY],
7297
                           self.op.hypervisor)
7298

    
7299
    self.dry_run_result = list(nodenames)
7300

    
7301
  def Exec(self, feedback_fn):
7302
    """Create and add the instance to the cluster.
7303

7304
    """
7305
    instance = self.op.instance_name
7306
    pnode_name = self.pnode.name
7307

    
7308
    ht_kind = self.op.hypervisor
7309
    if ht_kind in constants.HTS_REQ_PORT:
7310
      network_port = self.cfg.AllocatePort()
7311
    else:
7312
      network_port = None
7313

    
7314
    if constants.ENABLE_FILE_STORAGE:
7315
      # this is needed because os.path.join does not accept None arguments
7316
      if self.op.file_storage_dir is None:
7317
        string_file_storage_dir = ""
7318
      else:
7319
        string_file_storage_dir = self.op.file_storage_dir
7320

    
7321
      # build the full file storage dir path
7322
      file_storage_dir = utils.PathJoin(self.cfg.GetFileStorageDir(),
7323
                                        string_file_storage_dir, instance)
7324
    else:
7325
      file_storage_dir = ""
7326

    
7327
    disks = _GenerateDiskTemplate(self,
7328
                                  self.op.disk_template,
7329
                                  instance, pnode_name,
7330
                                  self.secondaries,
7331
                                  self.disks,
7332
                                  file_storage_dir,
7333
                                  self.op.file_driver,
7334
                                  0)
7335

    
7336
    iobj = objects.Instance(name=instance, os=self.op.os_type,
7337
                            primary_node=pnode_name,
7338
                            nics=self.nics, disks=disks,
7339
                            disk_template=self.op.disk_template,
7340
                            admin_up=False,
7341
                            network_port=network_port,
7342
                            beparams=self.op.beparams,
7343
                            hvparams=self.op.hvparams,
7344
                            hypervisor=self.op.hypervisor,
7345
                            osparams=self.op.osparams,
7346
                            )
7347

    
7348
    if self.adopt_disks:
7349
      # rename LVs to the newly-generated names; we need to construct
7350
      # 'fake' LV disks with the old data, plus the new unique_id
7351
      tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks]
7352
      rename_to = []
7353
      for t_dsk, a_dsk in zip (tmp_disks, self.disks):
7354
        rename_to.append(t_dsk.logical_id)
7355
        t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk["adopt"])
7356
        self.cfg.SetDiskID(t_dsk, pnode_name)
7357
      result = self.rpc.call_blockdev_rename(pnode_name,
7358
                                             zip(tmp_disks, rename_to))
7359
      result.Raise("Failed to rename adoped LVs")
7360
    else:
7361
      feedback_fn("* creating instance disks...")
7362
      try:
7363
        _CreateDisks(self, iobj)
7364
      except errors.OpExecError:
7365
        self.LogWarning("Device creation failed, reverting...")
7366
        try:
7367
          _RemoveDisks(self, iobj)
7368
        finally:
7369
          self.cfg.ReleaseDRBDMinors(instance)
7370
          raise
7371

    
7372
    feedback_fn("adding instance %s to cluster config" % instance)
7373

    
7374
    self.cfg.AddInstance(iobj, self.proc.GetECId())
7375

    
7376
    # Declare that we don't want to remove the instance lock anymore, as we've
7377
    # added the instance to the config
7378
    del self.remove_locks[locking.LEVEL_INSTANCE]
7379
    # Unlock all the nodes
7380
    if self.op.mode == constants.INSTANCE_IMPORT:
7381
      nodes_keep = [self.op.src_node]
7382
      nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
7383
                       if node != self.op.src_node]
7384
      self.context.glm.release(locking.LEVEL_NODE, nodes_release)
7385
      self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
7386
    else:
7387
      self.context.glm.release(locking.LEVEL_NODE)
7388
      del self.acquired_locks[locking.LEVEL_NODE]
7389

    
7390
    if self.op.wait_for_sync:
7391
      disk_abort = not _WaitForSync(self, iobj)
7392
    elif iobj.disk_template in constants.DTS_NET_MIRROR:
7393
      # make sure the disks are not degraded (still sync-ing is ok)
7394
      time.sleep(15)
7395
      feedback_fn("* checking mirrors status")
7396
      disk_abort = not _WaitForSync(self, iobj, oneshot=True)
7397
    else:
7398
      disk_abort = False
7399

    
7400
    if disk_abort:
7401
      _RemoveDisks(self, iobj)
7402
      self.cfg.RemoveInstance(iobj.name)
7403
      # Make sure the instance lock gets removed
7404
      self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
7405
      raise errors.OpExecError("There are some degraded disks for"
7406
                               " this instance")
7407

    
7408
    if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks:
7409
      if self.op.mode == constants.INSTANCE_CREATE:
7410
        if not self.op.no_install:
7411
          feedback_fn("* running the instance OS create scripts...")
7412
          # FIXME: pass debug option from opcode to backend
7413
          result = self.rpc.call_instance_os_add(pnode_name, iobj, False,
7414
                                                 self.op.debug_level)
7415
          result.Raise("Could not add os for instance %s"
7416
                       " on node %s" % (instance, pnode_name))
7417

    
7418
      elif self.op.mode == constants.INSTANCE_IMPORT:
7419
        feedback_fn("* running the instance OS import scripts...")
7420

    
7421
        transfers = []
7422

    
7423
        for idx, image in enumerate(self.src_images):
7424
          if not image:
7425
            continue
7426

    
7427
          # FIXME: pass debug option from opcode to backend
7428
          dt = masterd.instance.DiskTransfer("disk/%s" % idx,
7429
                                             constants.IEIO_FILE, (image, ),
7430
                                             constants.IEIO_SCRIPT,
7431
                                             (iobj.disks[idx], idx),
7432
                                             None)
7433
          transfers.append(dt)
7434

    
7435
        import_result = \
7436
          masterd.instance.TransferInstanceData(self, feedback_fn,
7437
                                                self.op.src_node, pnode_name,
7438
                                                self.pnode.secondary_ip,
7439
                                                iobj, transfers)
7440
        if not compat.all(import_result):
7441
          self.LogWarning("Some disks for instance %s on node %s were not"
7442
                          " imported successfully" % (instance, pnode_name))
7443

    
7444
      elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
7445
        feedback_fn("* preparing remote import...")
7446
        connect_timeout = constants.RIE_CONNECT_TIMEOUT
7447
        timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
7448

    
7449
        disk_results = masterd.instance.RemoteImport(self, feedback_fn, iobj,
7450
                                                     self.source_x509_ca,
7451
                                                     self._cds, timeouts)
7452
        if not compat.all(disk_results):
7453
          # TODO: Should the instance still be started, even if some disks
7454
          # failed to import (valid for local imports, too)?
7455
          self.LogWarning("Some disks for instance %s on node %s were not"
7456
                          " imported successfully" % (instance, pnode_name))
7457

    
7458
        # Run rename script on newly imported instance
7459
        assert iobj.name == instance
7460
        feedback_fn("Running rename script for %s" % instance)
7461
        result = self.rpc.call_instance_run_rename(pnode_name, iobj,
7462
                                                   self.source_instance_name,
7463
                                                   self.op.debug_level)
7464
        if result.fail_msg:
7465
          self.LogWarning("Failed to run rename script for %s on node"
7466
                          " %s: %s" % (instance, pnode_name, result.fail_msg))
7467

    
7468
      else:
7469
        # also checked in the prereq part
7470
        raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
7471
                                     % self.op.mode)
7472

    
7473
    if self.op.start:
7474
      iobj.admin_up = True
7475
      self.cfg.Update(iobj, feedback_fn)
7476
      logging.info("Starting instance %s on node %s", instance, pnode_name)
7477
      feedback_fn("* starting instance...")
7478
      result = self.rpc.call_instance_start(pnode_name, iobj, None, None)
7479
      result.Raise("Could not start instance")
7480

    
7481
    return list(iobj.all_nodes)
7482

    
7483

    
7484
class LUConnectConsole(NoHooksLU):
7485
  """Connect to an instance's console.
7486

7487
  This is somewhat special in that it returns the command line that
7488
  you need to run on the master node in order to connect to the
7489
  console.
7490

7491
  """
7492
  _OP_PARAMS = [
7493
    _PInstanceName
7494
    ]
7495
  REQ_BGL = False
7496

    
7497
  def ExpandNames(self):
7498
    self._ExpandAndLockInstance()
7499

    
7500
  def CheckPrereq(self):
7501
    """Check prerequisites.
7502

7503
    This checks that the instance is in the cluster.
7504

7505
    """
7506
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7507
    assert self.instance is not None, \
7508
      "Cannot retrieve locked instance %s" % self.op.instance_name
7509
    _CheckNodeOnline(self, self.instance.primary_node)
7510

    
7511
  def Exec(self, feedback_fn):
7512
    """Connect to the console of an instance
7513

7514
    """
7515
    instance = self.instance
7516
    node = instance.primary_node
7517

    
7518
    node_insts = self.rpc.call_instance_list([node],
7519
                                             [instance.hypervisor])[node]
7520
    node_insts.Raise("Can't get node information from %s" % node)
7521

    
7522
    if instance.name not in node_insts.payload:
7523
      raise errors.OpExecError("Instance %s is not running." % instance.name)
7524

    
7525
    logging.debug("Connecting to console of %s on %s", instance.name, node)
7526

    
7527
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
7528
    cluster = self.cfg.GetClusterInfo()
7529
    # beparams and hvparams are passed separately, to avoid editing the
7530
    # instance and then saving the defaults in the instance itself.
7531
    hvparams = cluster.FillHV(instance)
7532
    beparams = cluster.FillBE(instance)
7533
    console_cmd = hyper.GetShellCommandForConsole(instance, hvparams, beparams)
7534

    
7535
    # build ssh cmdline
7536
    return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
7537

    
7538

    
7539
class LUReplaceDisks(LogicalUnit):
7540
  """Replace the disks of an instance.
7541

7542
  """
7543
  HPATH = "mirrors-replace"
7544
  HTYPE = constants.HTYPE_INSTANCE
7545
  _OP_PARAMS = [
7546
    _PInstanceName,
7547
    ("mode", _NoDefault, _TElemOf(constants.REPLACE_MODES)),
7548
    ("disks", _EmptyList, _TListOf(_TPositiveInt)),
7549
    ("remote_node", None, _TMaybeString),
7550
    ("iallocator", None, _TMaybeString),
7551
    ("early_release", False, _TBool),
7552
    ]
7553
  REQ_BGL = False
7554

    
7555
  def CheckArguments(self):
7556
    TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
7557
                                  self.op.iallocator)
7558

    
7559
  def ExpandNames(self):
7560
    self._ExpandAndLockInstance()
7561

    
7562
    if self.op.iallocator is not None:
7563
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7564

    
7565
    elif self.op.remote_node is not None:
7566
      remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
7567
      self.op.remote_node = remote_node
7568

    
7569
      # Warning: do not remove the locking of the new secondary here
7570
      # unless DRBD8.AddChildren is changed to work in parallel;
7571
      # currently it doesn't since parallel invocations of
7572
      # FindUnusedMinor will conflict
7573
      self.needed_locks[locking.LEVEL_NODE] = [remote_node]
7574
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
7575

    
7576
    else:
7577
      self.needed_locks[locking.LEVEL_NODE] = []
7578
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7579

    
7580
    self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
7581
                                   self.op.iallocator, self.op.remote_node,
7582
                                   self.op.disks, False, self.op.early_release)
7583

    
7584
    self.tasklets = [self.replacer]
7585

    
7586
  def DeclareLocks(self, level):
7587
    # If we're not already locking all nodes in the set we have to declare the
7588
    # instance's primary/secondary nodes.
7589
    if (level == locking.LEVEL_NODE and
7590
        self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
7591
      self._LockInstancesNodes()
7592

    
7593
  def BuildHooksEnv(self):
7594
    """Build hooks env.
7595

7596
    This runs on the master, the primary and all the secondaries.
7597

7598
    """
7599
    instance = self.replacer.instance
7600
    env = {
7601
      "MODE": self.op.mode,
7602
      "NEW_SECONDARY": self.op.remote_node,
7603
      "OLD_SECONDARY": instance.secondary_nodes[0],
7604
      }
7605
    env.update(_BuildInstanceHookEnvByObject(self, instance))
7606
    nl = [
7607
      self.cfg.GetMasterNode(),
7608
      instance.primary_node,
7609
      ]
7610
    if self.op.remote_node is not None:
7611
      nl.append(self.op.remote_node)
7612
    return env, nl, nl
7613

    
7614

    
7615
class TLReplaceDisks(Tasklet):
7616
  """Replaces disks for an instance.
7617

7618
  Note: Locking is not within the scope of this class.
7619

7620
  """
7621
  def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
7622
               disks, delay_iallocator, early_release):
7623
    """Initializes this class.
7624

7625
    """
7626
    Tasklet.__init__(self, lu)
7627

    
7628
    # Parameters
7629
    self.instance_name = instance_name
7630
    self.mode = mode
7631
    self.iallocator_name = iallocator_name
7632
    self.remote_node = remote_node
7633
    self.disks = disks
7634
    self.delay_iallocator = delay_iallocator
7635
    self.early_release = early_release
7636

    
7637
    # Runtime data
7638
    self.instance = None
7639
    self.new_node = None
7640
    self.target_node = None
7641
    self.other_node = None
7642
    self.remote_node_info = None
7643
    self.node_secondary_ip = None
7644

    
7645
  @staticmethod
7646
  def CheckArguments(mode, remote_node, iallocator):
7647
    """Helper function for users of this class.
7648

7649
    """
7650
    # check for valid parameter combination
7651
    if mode == constants.REPLACE_DISK_CHG:
7652
      if remote_node is None and iallocator is None:
7653
        raise errors.OpPrereqError("When changing the secondary either an"
7654
                                   " iallocator script must be used or the"
7655
                                   " new node given", errors.ECODE_INVAL)
7656

    
7657
      if remote_node is not None and iallocator is not None:
7658
        raise errors.OpPrereqError("Give either the iallocator or the new"
7659
                                   " secondary, not both", errors.ECODE_INVAL)
7660

    
7661
    elif remote_node is not None or iallocator is not None:
7662
      # Not replacing the secondary
7663
      raise errors.OpPrereqError("The iallocator and new node options can"
7664
                                 " only be used when changing the"
7665
                                 " secondary node", errors.ECODE_INVAL)
7666

    
7667
  @staticmethod
7668
  def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
7669
    """Compute a new secondary node using an IAllocator.
7670

7671
    """
7672
    ial = IAllocator(lu.cfg, lu.rpc,
7673
                     mode=constants.IALLOCATOR_MODE_RELOC,
7674
                     name=instance_name,
7675
                     relocate_from=relocate_from)
7676

    
7677
    ial.Run(iallocator_name)
7678

    
7679
    if not ial.success:
7680
      raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
7681
                                 " %s" % (iallocator_name, ial.info),
7682
                                 errors.ECODE_NORES)
7683

    
7684
    if len(ial.result) != ial.required_nodes:
7685
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
7686
                                 " of nodes (%s), required %s" %
7687
                                 (iallocator_name,
7688
                                  len(ial.result), ial.required_nodes),
7689
                                 errors.ECODE_FAULT)
7690

    
7691
    remote_node_name = ial.result[0]
7692

    
7693
    lu.LogInfo("Selected new secondary for instance '%s': %s",
7694
               instance_name, remote_node_name)
7695

    
7696
    return remote_node_name
7697

    
7698
  def _FindFaultyDisks(self, node_name):
7699
    return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
7700
                                    node_name, True)
7701

    
7702
  def CheckPrereq(self):
7703
    """Check prerequisites.
7704

7705
    This checks that the instance is in the cluster.
7706

7707
    """
7708
    self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
7709
    assert instance is not None, \
7710
      "Cannot retrieve locked instance %s" % self.instance_name
7711

    
7712
    if instance.disk_template != constants.DT_DRBD8:
7713
      raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
7714
                                 " instances", errors.ECODE_INVAL)
7715

    
7716
    if len(instance.secondary_nodes) != 1:
7717
      raise errors.OpPrereqError("The instance has a strange layout,"
7718
                                 " expected one secondary but found %d" %
7719
                                 len(instance.secondary_nodes),
7720
                                 errors.ECODE_FAULT)
7721

    
7722
    if not self.delay_iallocator:
7723
      self._CheckPrereq2()
7724

    
7725
  def _CheckPrereq2(self):
7726
    """Check prerequisites, second part.
7727

7728
    This function should always be part of CheckPrereq. It was separated and is
7729
    now called from Exec because during node evacuation iallocator was only
7730
    called with an unmodified cluster model, not taking planned changes into
7731
    account.
7732

7733
    """
7734
    instance = self.instance
7735
    secondary_node = instance.secondary_nodes[0]
7736

    
7737
    if self.iallocator_name is None:
7738
      remote_node = self.remote_node
7739
    else:
7740
      remote_node = self._RunAllocator(self.lu, self.iallocator_name,
7741
                                       instance.name, instance.secondary_nodes)
7742

    
7743
    if remote_node is not None:
7744
      self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
7745
      assert self.remote_node_info is not None, \
7746
        "Cannot retrieve locked node %s" % remote_node
7747
    else:
7748
      self.remote_node_info = None
7749

    
7750
    if remote_node == self.instance.primary_node:
7751
      raise errors.OpPrereqError("The specified node is the primary node of"
7752
                                 " the instance.", errors.ECODE_INVAL)
7753

    
7754
    if remote_node == secondary_node:
7755
      raise errors.OpPrereqError("The specified node is already the"
7756
                                 " secondary node of the instance.",
7757
                                 errors.ECODE_INVAL)
7758

    
7759
    if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
7760
                                    constants.REPLACE_DISK_CHG):
7761
      raise errors.OpPrereqError("Cannot specify disks to be replaced",
7762
                                 errors.ECODE_INVAL)
7763

    
7764
    if self.mode == constants.REPLACE_DISK_AUTO:
7765
      faulty_primary = self._FindFaultyDisks(instance.primary_node)
7766
      faulty_secondary = self._FindFaultyDisks(secondary_node)
7767

    
7768
      if faulty_primary and faulty_secondary:
7769
        raise errors.OpPrereqError("Instance %s has faulty disks on more than"
7770
                                   " one node and can not be repaired"
7771
                                   " automatically" % self.instance_name,
7772
                                   errors.ECODE_STATE)
7773

    
7774
      if faulty_primary:
7775
        self.disks = faulty_primary
7776
        self.target_node = instance.primary_node
7777
        self.other_node = secondary_node
7778
        check_nodes = [self.target_node, self.other_node]
7779
      elif faulty_secondary:
7780
        self.disks = faulty_secondary
7781
        self.target_node = secondary_node
7782
        self.other_node = instance.primary_node
7783
        check_nodes = [self.target_node, self.other_node]
7784
      else:
7785
        self.disks = []
7786
        check_nodes = []
7787

    
7788
    else:
7789
      # Non-automatic modes
7790
      if self.mode == constants.REPLACE_DISK_PRI:
7791
        self.target_node = instance.primary_node
7792
        self.other_node = secondary_node
7793
        check_nodes = [self.target_node, self.other_node]
7794

    
7795
      elif self.mode == constants.REPLACE_DISK_SEC:
7796
        self.target_node = secondary_node
7797
        self.other_node = instance.primary_node
7798
        check_nodes = [self.target_node, self.other_node]
7799

    
7800
      elif self.mode == constants.REPLACE_DISK_CHG:
7801
        self.new_node = remote_node
7802
        self.other_node = instance.primary_node
7803
        self.target_node = secondary_node
7804
        check_nodes = [self.new_node, self.other_node]
7805

    
7806
        _CheckNodeNotDrained(self.lu, remote_node)
7807

    
7808
        old_node_info = self.cfg.GetNodeInfo(secondary_node)
7809
        assert old_node_info is not None
7810
        if old_node_info.offline and not self.early_release:
7811
          # doesn't make sense to delay the release
7812
          self.early_release = True
7813
          self.lu.LogInfo("Old secondary %s is offline, automatically enabling"
7814
                          " early-release mode", secondary_node)
7815

    
7816
      else:
7817
        raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
7818
                                     self.mode)
7819

    
7820
      # If not specified all disks should be replaced
7821
      if not self.disks:
7822
        self.disks = range(len(self.instance.disks))
7823

    
7824
    for node in check_nodes:
7825
      _CheckNodeOnline(self.lu, node)
7826

    
7827
    # Check whether disks are valid
7828
    for disk_idx in self.disks:
7829
      instance.FindDisk(disk_idx)
7830

    
7831
    # Get secondary node IP addresses
7832
    node_2nd_ip = {}
7833

    
7834
    for node_name in [self.target_node, self.other_node, self.new_node]:
7835
      if node_name is not None:
7836
        node_2nd_ip[node_name] = self.cfg.GetNodeInfo(node_name).secondary_ip
7837

    
7838
    self.node_secondary_ip = node_2nd_ip
7839

    
7840
  def Exec(self, feedback_fn):
7841
    """Execute disk replacement.
7842

7843
    This dispatches the disk replacement to the appropriate handler.
7844

7845
    """
7846
    if self.delay_iallocator:
7847
      self._CheckPrereq2()
7848

    
7849
    if not self.disks:
7850
      feedback_fn("No disks need replacement")
7851
      return
7852

    
7853
    feedback_fn("Replacing disk(s) %s for %s" %
7854
                (utils.CommaJoin(self.disks), self.instance.name))
7855

    
7856
    activate_disks = (not self.instance.admin_up)
7857

    
7858
    # Activate the instance disks if we're replacing them on a down instance
7859
    if activate_disks:
7860
      _StartInstanceDisks(self.lu, self.instance, True)
7861

    
7862
    try:
7863
      # Should we replace the secondary node?
7864
      if self.new_node is not None:
7865
        fn = self._ExecDrbd8Secondary
7866
      else:
7867
        fn = self._ExecDrbd8DiskOnly
7868

    
7869
      return fn(feedback_fn)
7870

    
7871
    finally:
7872
      # Deactivate the instance disks if we're replacing them on a
7873
      # down instance
7874
      if activate_disks:
7875
        _SafeShutdownInstanceDisks(self.lu, self.instance)
7876

    
7877
  def _CheckVolumeGroup(self, nodes):
7878
    self.lu.LogInfo("Checking volume groups")
7879

    
7880
    vgname = self.cfg.GetVGName()
7881

    
7882
    # Make sure volume group exists on all involved nodes
7883
    results = self.rpc.call_vg_list(nodes)
7884
    if not results:
7885
      raise errors.OpExecError("Can't list volume groups on the nodes")
7886

    
7887
    for node in nodes:
7888
      res = results[node]
7889
      res.Raise("Error checking node %s" % node)
7890
      if vgname not in res.payload:
7891
        raise errors.OpExecError("Volume group '%s' not found on node %s" %
7892
                                 (vgname, node))
7893

    
7894
  def _CheckDisksExistence(self, nodes):
7895
    # Check disk existence
7896
    for idx, dev in enumerate(self.instance.disks):
7897
      if idx not in self.disks:
7898
        continue
7899

    
7900
      for node in nodes:
7901
        self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
7902
        self.cfg.SetDiskID(dev, node)
7903

    
7904
        result = self.rpc.call_blockdev_find(node, dev)
7905

    
7906
        msg = result.fail_msg
7907
        if msg or not result.payload:
7908
          if not msg:
7909
            msg = "disk not found"
7910
          raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
7911
                                   (idx, node, msg))
7912

    
7913
  def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
7914
    for idx, dev in enumerate(self.instance.disks):
7915
      if idx not in self.disks:
7916
        continue
7917

    
7918
      self.lu.LogInfo("Checking disk/%d consistency on node %s" %
7919
                      (idx, node_name))
7920

    
7921
      if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
7922
                                   ldisk=ldisk):
7923
        raise errors.OpExecError("Node %s has degraded storage, unsafe to"
7924
                                 " replace disks for instance %s" %
7925
                                 (node_name, self.instance.name))
7926

    
7927
  def _CreateNewStorage(self, node_name):
7928
    vgname = self.cfg.GetVGName()
7929
    iv_names = {}
7930

    
7931
    for idx, dev in enumerate(self.instance.disks):
7932
      if idx not in self.disks:
7933
        continue
7934

    
7935
      self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
7936

    
7937
      self.cfg.SetDiskID(dev, node_name)
7938

    
7939
      lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
7940
      names = _GenerateUniqueNames(self.lu, lv_names)
7941

    
7942
      lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
7943
                             logical_id=(vgname, names[0]))
7944
      lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
7945
                             logical_id=(vgname, names[1]))
7946

    
7947
      new_lvs = [lv_data, lv_meta]
7948
      old_lvs = dev.children
7949
      iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
7950

    
7951
      # we pass force_create=True to force the LVM creation
7952
      for new_lv in new_lvs:
7953
        _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
7954
                        _GetInstanceInfoText(self.instance), False)
7955

    
7956
    return iv_names
7957

    
7958
  def _CheckDevices(self, node_name, iv_names):
7959
    for name, (dev, _, _) in iv_names.iteritems():
7960
      self.cfg.SetDiskID(dev, node_name)
7961

    
7962
      result = self.rpc.call_blockdev_find(node_name, dev)
7963

    
7964
      msg = result.fail_msg
7965
      if msg or not result.payload:
7966
        if not msg:
7967
          msg = "disk not found"
7968
        raise errors.OpExecError("Can't find DRBD device %s: %s" %
7969
                                 (name, msg))
7970

    
7971
      if result.payload.is_degraded:
7972
        raise errors.OpExecError("DRBD device %s is degraded!" % name)
7973

    
7974
  def _RemoveOldStorage(self, node_name, iv_names):
7975
    for name, (_, old_lvs, _) in iv_names.iteritems():
7976
      self.lu.LogInfo("Remove logical volumes for %s" % name)
7977

    
7978
      for lv in old_lvs:
7979
        self.cfg.SetDiskID(lv, node_name)
7980

    
7981
        msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
7982
        if msg:
7983
          self.lu.LogWarning("Can't remove old LV: %s" % msg,
7984
                             hint="remove unused LVs manually")
7985

    
7986
  def _ReleaseNodeLock(self, node_name):
7987
    """Releases the lock for a given node."""
7988
    self.lu.context.glm.release(locking.LEVEL_NODE, node_name)
7989

    
7990
  def _ExecDrbd8DiskOnly(self, feedback_fn):
7991
    """Replace a disk on the primary or secondary for DRBD 8.
7992

7993
    The algorithm for replace is quite complicated:
7994

7995
      1. for each disk to be replaced:
7996

7997
        1. create new LVs on the target node with unique names
7998
        1. detach old LVs from the drbd device
7999
        1. rename old LVs to name_replaced.<time_t>
8000
        1. rename new LVs to old LVs
8001
        1. attach the new LVs (with the old names now) to the drbd device
8002

8003
      1. wait for sync across all devices
8004

8005
      1. for each modified disk:
8006

8007
        1. remove old LVs (which have the name name_replaces.<time_t>)
8008

8009
    Failures are not very well handled.
8010

8011
    """
8012
    steps_total = 6
8013

    
8014
    # Step: check device activation
8015
    self.lu.LogStep(1, steps_total, "Check device existence")
8016
    self._CheckDisksExistence([self.other_node, self.target_node])
8017
    self._CheckVolumeGroup([self.target_node, self.other_node])
8018

    
8019
    # Step: check other node consistency
8020
    self.lu.LogStep(2, steps_total, "Check peer consistency")
8021
    self._CheckDisksConsistency(self.other_node,
8022
                                self.other_node == self.instance.primary_node,
8023
                                False)
8024

    
8025
    # Step: create new storage
8026
    self.lu.LogStep(3, steps_total, "Allocate new storage")
8027
    iv_names = self._CreateNewStorage(self.target_node)
8028

    
8029
    # Step: for each lv, detach+rename*2+attach
8030
    self.lu.LogStep(4, steps_total, "Changing drbd configuration")
8031
    for dev, old_lvs, new_lvs in iv_names.itervalues():
8032
      self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
8033

    
8034
      result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
8035
                                                     old_lvs)
8036
      result.Raise("Can't detach drbd from local storage on node"
8037
                   " %s for device %s" % (self.target_node, dev.iv_name))
8038
      #dev.children = []
8039
      #cfg.Update(instance)
8040

    
8041
      # ok, we created the new LVs, so now we know we have the needed
8042
      # storage; as such, we proceed on the target node to rename
8043
      # old_lv to _old, and new_lv to old_lv; note that we rename LVs
8044
      # using the assumption that logical_id == physical_id (which in
8045
      # turn is the unique_id on that node)
8046

    
8047
      # FIXME(iustin): use a better name for the replaced LVs
8048
      temp_suffix = int(time.time())
8049
      ren_fn = lambda d, suff: (d.physical_id[0],
8050
                                d.physical_id[1] + "_replaced-%s" % suff)
8051

    
8052
      # Build the rename list based on what LVs exist on the node
8053
      rename_old_to_new = []
8054
      for to_ren in old_lvs:
8055
        result = self.rpc.call_blockdev_find(self.target_node, to_ren)
8056
        if not result.fail_msg and result.payload:
8057
          # device exists
8058
          rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
8059

    
8060
      self.lu.LogInfo("Renaming the old LVs on the target node")
8061
      result = self.rpc.call_blockdev_rename(self.target_node,
8062
                                             rename_old_to_new)
8063
      result.Raise("Can't rename old LVs on node %s" % self.target_node)
8064

    
8065
      # Now we rename the new LVs to the old LVs
8066
      self.lu.LogInfo("Renaming the new LVs on the target node")
8067
      rename_new_to_old = [(new, old.physical_id)
8068
                           for old, new in zip(old_lvs, new_lvs)]
8069
      result = self.rpc.call_blockdev_rename(self.target_node,
8070
                                             rename_new_to_old)
8071
      result.Raise("Can't rename new LVs on node %s" % self.target_node)
8072

    
8073
      for old, new in zip(old_lvs, new_lvs):
8074
        new.logical_id = old.logical_id
8075
        self.cfg.SetDiskID(new, self.target_node)
8076

    
8077
      for disk in old_lvs:
8078
        disk.logical_id = ren_fn(disk, temp_suffix)
8079
        self.cfg.SetDiskID(disk, self.target_node)
8080

    
8081
      # Now that the new lvs have the old name, we can add them to the device
8082
      self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
8083
      result = self.rpc.call_blockdev_addchildren(self.target_node, dev,
8084
                                                  new_lvs)
8085
      msg = result.fail_msg
8086
      if msg:
8087
        for new_lv in new_lvs:
8088
          msg2 = self.rpc.call_blockdev_remove(self.target_node,
8089
                                               new_lv).fail_msg
8090
          if msg2:
8091
            self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
8092
                               hint=("cleanup manually the unused logical"
8093
                                     "volumes"))
8094
        raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
8095

    
8096
      dev.children = new_lvs
8097

    
8098
      self.cfg.Update(self.instance, feedback_fn)
8099

    
8100
    cstep = 5
8101
    if self.early_release:
8102
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
8103
      cstep += 1
8104
      self._RemoveOldStorage(self.target_node, iv_names)
8105
      # WARNING: we release both node locks here, do not do other RPCs
8106
      # than WaitForSync to the primary node
8107
      self._ReleaseNodeLock([self.target_node, self.other_node])
8108

    
8109
    # Wait for sync
8110
    # This can fail as the old devices are degraded and _WaitForSync
8111
    # does a combined result over all disks, so we don't check its return value
8112
    self.lu.LogStep(cstep, steps_total, "Sync devices")
8113
    cstep += 1
8114
    _WaitForSync(self.lu, self.instance)
8115

    
8116
    # Check all devices manually
8117
    self._CheckDevices(self.instance.primary_node, iv_names)
8118

    
8119
    # Step: remove old storage
8120
    if not self.early_release:
8121
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
8122
      cstep += 1
8123
      self._RemoveOldStorage(self.target_node, iv_names)
8124

    
8125
  def _ExecDrbd8Secondary(self, feedback_fn):
8126
    """Replace the secondary node for DRBD 8.
8127

8128
    The algorithm for replace is quite complicated:
8129
      - for all disks of the instance:
8130
        - create new LVs on the new node with same names
8131
        - shutdown the drbd device on the old secondary
8132
        - disconnect the drbd network on the primary
8133
        - create the drbd device on the new secondary
8134
        - network attach the drbd on the primary, using an artifice:
8135
          the drbd code for Attach() will connect to the network if it
8136
          finds a device which is connected to the good local disks but
8137
          not network enabled
8138
      - wait for sync across all devices
8139
      - remove all disks from the old secondary
8140

8141
    Failures are not very well handled.
8142

8143
    """
8144
    steps_total = 6
8145

    
8146
    # Step: check device activation
8147
    self.lu.LogStep(1, steps_total, "Check device existence")
8148
    self._CheckDisksExistence([self.instance.primary_node])
8149
    self._CheckVolumeGroup([self.instance.primary_node])
8150

    
8151
    # Step: check other node consistency
8152
    self.lu.LogStep(2, steps_total, "Check peer consistency")
8153
    self._CheckDisksConsistency(self.instance.primary_node, True, True)
8154

    
8155
    # Step: create new storage
8156
    self.lu.LogStep(3, steps_total, "Allocate new storage")
8157
    for idx, dev in enumerate(self.instance.disks):
8158
      self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
8159
                      (self.new_node, idx))
8160
      # we pass force_create=True to force LVM creation
8161
      for new_lv in dev.children:
8162
        _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
8163
                        _GetInstanceInfoText(self.instance), False)
8164

    
8165
    # Step 4: dbrd minors and drbd setups changes
8166
    # after this, we must manually remove the drbd minors on both the
8167
    # error and the success paths
8168
    self.lu.LogStep(4, steps_total, "Changing drbd configuration")
8169
    minors = self.cfg.AllocateDRBDMinor([self.new_node
8170
                                         for dev in self.instance.disks],
8171
                                        self.instance.name)
8172
    logging.debug("Allocated minors %r", minors)
8173

    
8174
    iv_names = {}
8175
    for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
8176
      self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
8177
                      (self.new_node, idx))
8178
      # create new devices on new_node; note that we create two IDs:
8179
      # one without port, so the drbd will be activated without
8180
      # networking information on the new node at this stage, and one
8181
      # with network, for the latter activation in step 4
8182
      (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
8183
      if self.instance.primary_node == o_node1:
8184
        p_minor = o_minor1
8185
      else:
8186
        assert self.instance.primary_node == o_node2, "Three-node instance?"
8187
        p_minor = o_minor2
8188

    
8189
      new_alone_id = (self.instance.primary_node, self.new_node, None,
8190
                      p_minor, new_minor, o_secret)
8191
      new_net_id = (self.instance.primary_node, self.new_node, o_port,
8192
                    p_minor, new_minor, o_secret)
8193

    
8194
      iv_names[idx] = (dev, dev.children, new_net_id)
8195
      logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
8196
                    new_net_id)
8197
      new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
8198
                              logical_id=new_alone_id,
8199
                              children=dev.children,
8200
                              size=dev.size)
8201
      try:
8202
        _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
8203
                              _GetInstanceInfoText(self.instance), False)
8204
      except errors.GenericError:
8205
        self.cfg.ReleaseDRBDMinors(self.instance.name)
8206
        raise
8207

    
8208
    # We have new devices, shutdown the drbd on the old secondary
8209
    for idx, dev in enumerate(self.instance.disks):
8210
      self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
8211
      self.cfg.SetDiskID(dev, self.target_node)
8212
      msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
8213
      if msg:
8214
        self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
8215
                           "node: %s" % (idx, msg),
8216
                           hint=("Please cleanup this device manually as"
8217
                                 " soon as possible"))
8218

    
8219
    self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
8220
    result = self.rpc.call_drbd_disconnect_net([self.instance.primary_node],
8221
                                               self.node_secondary_ip,
8222
                                               self.instance.disks)\
8223
                                              [self.instance.primary_node]
8224

    
8225
    msg = result.fail_msg
8226
    if msg:
8227
      # detaches didn't succeed (unlikely)
8228
      self.cfg.ReleaseDRBDMinors(self.instance.name)
8229
      raise errors.OpExecError("Can't detach the disks from the network on"
8230
                               " old node: %s" % (msg,))
8231

    
8232
    # if we managed to detach at least one, we update all the disks of
8233
    # the instance to point to the new secondary
8234
    self.lu.LogInfo("Updating instance configuration")
8235
    for dev, _, new_logical_id in iv_names.itervalues():
8236
      dev.logical_id = new_logical_id
8237
      self.cfg.SetDiskID(dev, self.instance.primary_node)
8238

    
8239
    self.cfg.Update(self.instance, feedback_fn)
8240

    
8241
    # and now perform the drbd attach
8242
    self.lu.LogInfo("Attaching primary drbds to new secondary"
8243
                    " (standalone => connected)")
8244
    result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
8245
                                            self.new_node],
8246
                                           self.node_secondary_ip,
8247
                                           self.instance.disks,
8248
                                           self.instance.name,
8249
                                           False)
8250
    for to_node, to_result in result.items():
8251
      msg = to_result.fail_msg
8252
      if msg:
8253
        self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
8254
                           to_node, msg,
8255
                           hint=("please do a gnt-instance info to see the"
8256
                                 " status of disks"))
8257
    cstep = 5
8258
    if self.early_release:
8259
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
8260
      cstep += 1
8261
      self._RemoveOldStorage(self.target_node, iv_names)
8262
      # WARNING: we release all node locks here, do not do other RPCs
8263
      # than WaitForSync to the primary node
8264
      self._ReleaseNodeLock([self.instance.primary_node,
8265
                             self.target_node,
8266
                             self.new_node])
8267

    
8268
    # Wait for sync
8269
    # This can fail as the old devices are degraded and _WaitForSync
8270
    # does a combined result over all disks, so we don't check its return value
8271
    self.lu.LogStep(cstep, steps_total, "Sync devices")
8272
    cstep += 1
8273
    _WaitForSync(self.lu, self.instance)
8274

    
8275
    # Check all devices manually
8276
    self._CheckDevices(self.instance.primary_node, iv_names)
8277

    
8278
    # Step: remove old storage
8279
    if not self.early_release:
8280
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
8281
      self._RemoveOldStorage(self.target_node, iv_names)
8282

    
8283

    
8284
class LURepairNodeStorage(NoHooksLU):
8285
  """Repairs the volume group on a node.
8286

8287
  """
8288
  _OP_PARAMS = [
8289
    _PNodeName,
8290
    ("storage_type", _NoDefault, _CheckStorageType),
8291
    ("name", _NoDefault, _TNonEmptyString),
8292
    ("ignore_consistency", False, _TBool),
8293
    ]
8294
  REQ_BGL = False
8295

    
8296
  def CheckArguments(self):
8297
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
8298

    
8299
    storage_type = self.op.storage_type
8300

    
8301
    if (constants.SO_FIX_CONSISTENCY not in
8302
        constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
8303
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
8304
                                 " repaired" % storage_type,
8305
                                 errors.ECODE_INVAL)
8306

    
8307
  def ExpandNames(self):
8308
    self.needed_locks = {
8309
      locking.LEVEL_NODE: [self.op.node_name],
8310
      }
8311

    
8312
  def _CheckFaultyDisks(self, instance, node_name):
8313
    """Ensure faulty disks abort the opcode or at least warn."""
8314
    try:
8315
      if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
8316
                                  node_name, True):
8317
        raise errors.OpPrereqError("Instance '%s' has faulty disks on"
8318
                                   " node '%s'" % (instance.name, node_name),
8319
                                   errors.ECODE_STATE)
8320
    except errors.OpPrereqError, err:
8321
      if self.op.ignore_consistency:
8322
        self.proc.LogWarning(str(err.args[0]))
8323
      else:
8324
        raise
8325

    
8326
  def CheckPrereq(self):
8327
    """Check prerequisites.
8328

8329
    """
8330
    # Check whether any instance on this node has faulty disks
8331
    for inst in _GetNodeInstances(self.cfg, self.op.node_name):
8332
      if not inst.admin_up:
8333
        continue
8334
      check_nodes = set(inst.all_nodes)
8335
      check_nodes.discard(self.op.node_name)
8336
      for inst_node_name in check_nodes:
8337
        self._CheckFaultyDisks(inst, inst_node_name)
8338

    
8339
  def Exec(self, feedback_fn):
8340
    feedback_fn("Repairing storage unit '%s' on %s ..." %
8341
                (self.op.name, self.op.node_name))
8342

    
8343
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
8344
    result = self.rpc.call_storage_execute(self.op.node_name,
8345
                                           self.op.storage_type, st_args,
8346
                                           self.op.name,
8347
                                           constants.SO_FIX_CONSISTENCY)
8348
    result.Raise("Failed to repair storage unit '%s' on %s" %
8349
                 (self.op.name, self.op.node_name))
8350

    
8351

    
8352
class LUNodeEvacuationStrategy(NoHooksLU):
8353
  """Computes the node evacuation strategy.
8354

8355
  """
8356
  _OP_PARAMS = [
8357
    ("nodes", _NoDefault, _TListOf(_TNonEmptyString)),
8358
    ("remote_node", None, _TMaybeString),
8359
    ("iallocator", None, _TMaybeString),
8360
    ]
8361
  REQ_BGL = False
8362

    
8363
  def CheckArguments(self):
8364
    _CheckIAllocatorOrNode(self, "iallocator", "remote_node")
8365

    
8366
  def ExpandNames(self):
8367
    self.op.nodes = _GetWantedNodes(self, self.op.nodes)
8368
    self.needed_locks = locks = {}
8369
    if self.op.remote_node is None:
8370
      locks[locking.LEVEL_NODE] = locking.ALL_SET
8371
    else:
8372
      self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8373
      locks[locking.LEVEL_NODE] = self.op.nodes + [self.op.remote_node]
8374

    
8375
  def Exec(self, feedback_fn):
8376
    if self.op.remote_node is not None:
8377
      instances = []
8378
      for node in self.op.nodes:
8379
        instances.extend(_GetNodeSecondaryInstances(self.cfg, node))
8380
      result = []
8381
      for i in instances:
8382
        if i.primary_node == self.op.remote_node:
8383
          raise errors.OpPrereqError("Node %s is the primary node of"
8384
                                     " instance %s, cannot use it as"
8385
                                     " secondary" %
8386
                                     (self.op.remote_node, i.name),
8387
                                     errors.ECODE_INVAL)
8388
        result.append([i.name, self.op.remote_node])
8389
    else:
8390
      ial = IAllocator(self.cfg, self.rpc,
8391
                       mode=constants.IALLOCATOR_MODE_MEVAC,
8392
                       evac_nodes=self.op.nodes)
8393
      ial.Run(self.op.iallocator, validate=True)
8394
      if not ial.success:
8395
        raise errors.OpExecError("No valid evacuation solution: %s" % ial.info,
8396
                                 errors.ECODE_NORES)
8397
      result = ial.result
8398
    return result
8399

    
8400

    
8401
class LUGrowDisk(LogicalUnit):
8402
  """Grow a disk of an instance.
8403

8404
  """
8405
  HPATH = "disk-grow"
8406
  HTYPE = constants.HTYPE_INSTANCE
8407
  _OP_PARAMS = [
8408
    _PInstanceName,
8409
    ("disk", _NoDefault, _TInt),
8410
    ("amount", _NoDefault, _TInt),
8411
    ("wait_for_sync", True, _TBool),
8412
    ]
8413
  REQ_BGL = False
8414

    
8415
  def ExpandNames(self):
8416
    self._ExpandAndLockInstance()
8417
    self.needed_locks[locking.LEVEL_NODE] = []
8418
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8419

    
8420
  def DeclareLocks(self, level):
8421
    if level == locking.LEVEL_NODE:
8422
      self._LockInstancesNodes()
8423

    
8424
  def BuildHooksEnv(self):
8425
    """Build hooks env.
8426

8427
    This runs on the master, the primary and all the secondaries.
8428

8429
    """
8430
    env = {
8431
      "DISK": self.op.disk,
8432
      "AMOUNT": self.op.amount,
8433
      }
8434
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
8435
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
8436
    return env, nl, nl
8437

    
8438
  def CheckPrereq(self):
8439
    """Check prerequisites.
8440

8441
    This checks that the instance is in the cluster.
8442

8443
    """
8444
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
8445
    assert instance is not None, \
8446
      "Cannot retrieve locked instance %s" % self.op.instance_name
8447
    nodenames = list(instance.all_nodes)
8448
    for node in nodenames:
8449
      _CheckNodeOnline(self, node)
8450

    
8451
    self.instance = instance
8452

    
8453
    if instance.disk_template not in constants.DTS_GROWABLE:
8454
      raise errors.OpPrereqError("Instance's disk layout does not support"
8455
                                 " growing.", errors.ECODE_INVAL)
8456

    
8457
    self.disk = instance.FindDisk(self.op.disk)
8458

    
8459
    if instance.disk_template != constants.DT_FILE:
8460
      # TODO: check the free disk space for file, when that feature will be
8461
      # supported
8462
      _CheckNodesFreeDisk(self, nodenames, self.op.amount)
8463

    
8464
  def Exec(self, feedback_fn):
8465
    """Execute disk grow.
8466

8467
    """
8468
    instance = self.instance
8469
    disk = self.disk
8470

    
8471
    disks_ok, _ = _AssembleInstanceDisks(self, self.instance, disks=[disk])
8472
    if not disks_ok:
8473
      raise errors.OpExecError("Cannot activate block device to grow")
8474

    
8475
    for node in instance.all_nodes:
8476
      self.cfg.SetDiskID(disk, node)
8477
      result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
8478
      result.Raise("Grow request failed to node %s" % node)
8479

    
8480
      # TODO: Rewrite code to work properly
8481
      # DRBD goes into sync mode for a short amount of time after executing the
8482
      # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
8483
      # calling "resize" in sync mode fails. Sleeping for a short amount of
8484
      # time is a work-around.
8485
      time.sleep(5)
8486

    
8487
    disk.RecordGrow(self.op.amount)
8488
    self.cfg.Update(instance, feedback_fn)
8489
    if self.op.wait_for_sync:
8490
      disk_abort = not _WaitForSync(self, instance, disks=[disk])
8491
      if disk_abort:
8492
        self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
8493
                             " status.\nPlease check the instance.")
8494
      if not instance.admin_up:
8495
        _SafeShutdownInstanceDisks(self, instance, disks=[disk])
8496
    elif not instance.admin_up:
8497
      self.proc.LogWarning("Not shutting down the disk even if the instance is"
8498
                           " not supposed to be running because no wait for"
8499
                           " sync mode was requested.")
8500

    
8501

    
8502
class LUQueryInstanceData(NoHooksLU):
8503
  """Query runtime instance data.
8504

8505
  """
8506
  _OP_PARAMS = [
8507
    ("instances", _EmptyList, _TListOf(_TNonEmptyString)),
8508
    ("static", False, _TBool),
8509
    ]
8510
  REQ_BGL = False
8511

    
8512
  def ExpandNames(self):
8513
    self.needed_locks = {}
8514
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
8515

    
8516
    if self.op.instances:
8517
      self.wanted_names = []
8518
      for name in self.op.instances:
8519
        full_name = _ExpandInstanceName(self.cfg, name)
8520
        self.wanted_names.append(full_name)
8521
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
8522
    else:
8523
      self.wanted_names = None
8524
      self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
8525

    
8526
    self.needed_locks[locking.LEVEL_NODE] = []
8527
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8528

    
8529
  def DeclareLocks(self, level):
8530
    if level == locking.LEVEL_NODE:
8531
      self._LockInstancesNodes()
8532

    
8533
  def CheckPrereq(self):
8534
    """Check prerequisites.
8535

8536
    This only checks the optional instance list against the existing names.
8537

8538
    """
8539
    if self.wanted_names is None:
8540
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
8541

    
8542
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
8543
                             in self.wanted_names]
8544

    
8545
  def _ComputeBlockdevStatus(self, node, instance_name, dev):
8546
    """Returns the status of a block device
8547

8548
    """
8549
    if self.op.static or not node:
8550
      return None
8551

    
8552
    self.cfg.SetDiskID(dev, node)
8553

    
8554
    result = self.rpc.call_blockdev_find(node, dev)
8555
    if result.offline:
8556
      return None
8557

    
8558
    result.Raise("Can't compute disk status for %s" % instance_name)
8559

    
8560
    status = result.payload
8561
    if status is None:
8562
      return None
8563

    
8564
    return (status.dev_path, status.major, status.minor,
8565
            status.sync_percent, status.estimated_time,
8566
            status.is_degraded, status.ldisk_status)
8567

    
8568
  def _ComputeDiskStatus(self, instance, snode, dev):
8569
    """Compute block device status.
8570

8571
    """
8572
    if dev.dev_type in constants.LDS_DRBD:
8573
      # we change the snode then (otherwise we use the one passed in)
8574
      if dev.logical_id[0] == instance.primary_node:
8575
        snode = dev.logical_id[1]
8576
      else:
8577
        snode = dev.logical_id[0]
8578

    
8579
    dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
8580
                                              instance.name, dev)
8581
    dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
8582

    
8583
    if dev.children:
8584
      dev_children = [self._ComputeDiskStatus(instance, snode, child)
8585
                      for child in dev.children]
8586
    else:
8587
      dev_children = []
8588

    
8589
    data = {
8590
      "iv_name": dev.iv_name,
8591
      "dev_type": dev.dev_type,
8592
      "logical_id": dev.logical_id,
8593
      "physical_id": dev.physical_id,
8594
      "pstatus": dev_pstatus,
8595
      "sstatus": dev_sstatus,
8596
      "children": dev_children,
8597
      "mode": dev.mode,
8598
      "size": dev.size,
8599
      }
8600

    
8601
    return data
8602

    
8603
  def Exec(self, feedback_fn):
8604
    """Gather and return data"""
8605
    result = {}
8606

    
8607
    cluster = self.cfg.GetClusterInfo()
8608

    
8609
    for instance in self.wanted_instances:
8610
      if not self.op.static:
8611
        remote_info = self.rpc.call_instance_info(instance.primary_node,
8612
                                                  instance.name,
8613
                                                  instance.hypervisor)
8614
        remote_info.Raise("Error checking node %s" % instance.primary_node)
8615
        remote_info = remote_info.payload
8616
        if remote_info and "state" in remote_info:
8617
          remote_state = "up"
8618
        else:
8619
          remote_state = "down"
8620
      else:
8621
        remote_state = None
8622
      if instance.admin_up:
8623
        config_state = "up"
8624
      else:
8625
        config_state = "down"
8626

    
8627
      disks = [self._ComputeDiskStatus(instance, None, device)
8628
               for device in instance.disks]
8629

    
8630
      idict = {
8631
        "name": instance.name,
8632
        "config_state": config_state,
8633
        "run_state": remote_state,
8634
        "pnode": instance.primary_node,
8635
        "snodes": instance.secondary_nodes,
8636
        "os": instance.os,
8637
        # this happens to be the same format used for hooks
8638
        "nics": _NICListToTuple(self, instance.nics),
8639
        "disk_template": instance.disk_template,
8640
        "disks": disks,
8641
        "hypervisor": instance.hypervisor,
8642
        "network_port": instance.network_port,
8643
        "hv_instance": instance.hvparams,
8644
        "hv_actual": cluster.FillHV(instance, skip_globals=True),
8645
        "be_instance": instance.beparams,
8646
        "be_actual": cluster.FillBE(instance),
8647
        "os_instance": instance.osparams,
8648
        "os_actual": cluster.SimpleFillOS(instance.os, instance.osparams),
8649
        "serial_no": instance.serial_no,
8650
        "mtime": instance.mtime,
8651
        "ctime": instance.ctime,
8652
        "uuid": instance.uuid,
8653
        }
8654

    
8655
      result[instance.name] = idict
8656

    
8657
    return result
8658

    
8659

    
8660
class LUSetInstanceParams(LogicalUnit):
8661
  """Modifies an instances's parameters.
8662

8663
  """
8664
  HPATH = "instance-modify"
8665
  HTYPE = constants.HTYPE_INSTANCE
8666
  _OP_PARAMS = [
8667
    _PInstanceName,
8668
    ("nics", _EmptyList, _TList),
8669
    ("disks", _EmptyList, _TList),
8670
    ("beparams", _EmptyDict, _TDict),
8671
    ("hvparams", _EmptyDict, _TDict),
8672
    ("disk_template", None, _TMaybeString),
8673
    ("remote_node", None, _TMaybeString),
8674
    ("os_name", None, _TMaybeString),
8675
    ("force_variant", False, _TBool),
8676
    ("osparams", None, _TOr(_TDict, _TNone)),
8677
    _PForce,
8678
    ]
8679
  REQ_BGL = False
8680

    
8681
  def CheckArguments(self):
8682
    if not (self.op.nics or self.op.disks or self.op.disk_template or
8683
            self.op.hvparams or self.op.beparams or self.op.os_name):
8684
      raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
8685

    
8686
    if self.op.hvparams:
8687
      _CheckGlobalHvParams(self.op.hvparams)
8688

    
8689
    # Disk validation
8690
    disk_addremove = 0
8691
    for disk_op, disk_dict in self.op.disks:
8692
      utils.ForceDictType(disk_dict, constants.IDISK_PARAMS_TYPES)
8693
      if disk_op == constants.DDM_REMOVE:
8694
        disk_addremove += 1
8695
        continue
8696
      elif disk_op == constants.DDM_ADD:
8697
        disk_addremove += 1
8698
      else:
8699
        if not isinstance(disk_op, int):
8700
          raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL)
8701
        if not isinstance(disk_dict, dict):
8702
          msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
8703
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
8704

    
8705
      if disk_op == constants.DDM_ADD:
8706
        mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
8707
        if mode not in constants.DISK_ACCESS_SET:
8708
          raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
8709
                                     errors.ECODE_INVAL)
8710
        size = disk_dict.get('size', None)
8711
        if size is None:
8712
          raise errors.OpPrereqError("Required disk parameter size missing",
8713
                                     errors.ECODE_INVAL)
8714
        try:
8715
          size = int(size)
8716
        except (TypeError, ValueError), err:
8717
          raise errors.OpPrereqError("Invalid disk size parameter: %s" %
8718
                                     str(err), errors.ECODE_INVAL)
8719
        disk_dict['size'] = size
8720
      else:
8721
        # modification of disk
8722
        if 'size' in disk_dict:
8723
          raise errors.OpPrereqError("Disk size change not possible, use"
8724
                                     " grow-disk", errors.ECODE_INVAL)
8725

    
8726
    if disk_addremove > 1:
8727
      raise errors.OpPrereqError("Only one disk add or remove operation"
8728
                                 " supported at a time", errors.ECODE_INVAL)
8729

    
8730
    if self.op.disks and self.op.disk_template is not None:
8731
      raise errors.OpPrereqError("Disk template conversion and other disk"
8732
                                 " changes not supported at the same time",
8733
                                 errors.ECODE_INVAL)
8734

    
8735
    if self.op.disk_template:
8736
      _CheckDiskTemplate(self.op.disk_template)
8737
      if (self.op.disk_template in constants.DTS_NET_MIRROR and
8738
          self.op.remote_node is None):
8739
        raise errors.OpPrereqError("Changing the disk template to a mirrored"
8740
                                   " one requires specifying a secondary node",
8741
                                   errors.ECODE_INVAL)
8742

    
8743
    # NIC validation
8744
    nic_addremove = 0
8745
    for nic_op, nic_dict in self.op.nics:
8746
      utils.ForceDictType(nic_dict, constants.INIC_PARAMS_TYPES)
8747
      if nic_op == constants.DDM_REMOVE:
8748
        nic_addremove += 1
8749
        continue
8750
      elif nic_op == constants.DDM_ADD:
8751
        nic_addremove += 1
8752
      else:
8753
        if not isinstance(nic_op, int):
8754
          raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL)
8755
        if not isinstance(nic_dict, dict):
8756
          msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
8757
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
8758

    
8759
      # nic_dict should be a dict
8760
      nic_ip = nic_dict.get('ip', None)
8761
      if nic_ip is not None:
8762
        if nic_ip.lower() == constants.VALUE_NONE:
8763
          nic_dict['ip'] = None
8764
        else:
8765
          if not netutils.IsValidIP4(nic_ip):
8766
            raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip,
8767
                                       errors.ECODE_INVAL)
8768

    
8769
      nic_bridge = nic_dict.get('bridge', None)
8770
      nic_link = nic_dict.get('link', None)
8771
      if nic_bridge and nic_link:
8772
        raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
8773
                                   " at the same time", errors.ECODE_INVAL)
8774
      elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
8775
        nic_dict['bridge'] = None
8776
      elif nic_link and nic_link.lower() == constants.VALUE_NONE:
8777
        nic_dict['link'] = None
8778

    
8779
      if nic_op == constants.DDM_ADD:
8780
        nic_mac = nic_dict.get('mac', None)
8781
        if nic_mac is None:
8782
          nic_dict['mac'] = constants.VALUE_AUTO
8783

    
8784
      if 'mac' in nic_dict:
8785
        nic_mac = nic_dict['mac']
8786
        if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
8787
          nic_mac = utils.NormalizeAndValidateMac(nic_mac)
8788

    
8789
        if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
8790
          raise errors.OpPrereqError("'auto' is not a valid MAC address when"
8791
                                     " modifying an existing nic",
8792
                                     errors.ECODE_INVAL)
8793

    
8794
    if nic_addremove > 1:
8795
      raise errors.OpPrereqError("Only one NIC add or remove operation"
8796
                                 " supported at a time", errors.ECODE_INVAL)
8797

    
8798
  def ExpandNames(self):
8799
    self._ExpandAndLockInstance()
8800
    self.needed_locks[locking.LEVEL_NODE] = []
8801
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8802

    
8803
  def DeclareLocks(self, level):
8804
    if level == locking.LEVEL_NODE:
8805
      self._LockInstancesNodes()
8806
      if self.op.disk_template and self.op.remote_node:
8807
        self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8808
        self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)
8809

    
8810
  def BuildHooksEnv(self):
8811
    """Build hooks env.
8812

8813
    This runs on the master, primary and secondaries.
8814

8815
    """
8816
    args = dict()
8817
    if constants.BE_MEMORY in self.be_new:
8818
      args['memory'] = self.be_new[constants.BE_MEMORY]
8819
    if constants.BE_VCPUS in self.be_new:
8820
      args['vcpus'] = self.be_new[constants.BE_VCPUS]
8821
    # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
8822
    # information at all.
8823
    if self.op.nics:
8824
      args['nics'] = []
8825
      nic_override = dict(self.op.nics)
8826
      for idx, nic in enumerate(self.instance.nics):
8827
        if idx in nic_override:
8828
          this_nic_override = nic_override[idx]
8829
        else:
8830
          this_nic_override = {}
8831
        if 'ip' in this_nic_override:
8832
          ip = this_nic_override['ip']
8833
        else:
8834
          ip = nic.ip
8835
        if 'mac' in this_nic_override:
8836
          mac = this_nic_override['mac']
8837
        else:
8838
          mac = nic.mac
8839
        if idx in self.nic_pnew:
8840
          nicparams = self.nic_pnew[idx]
8841
        else:
8842
          nicparams = self.cluster.SimpleFillNIC(nic.nicparams)
8843
        mode = nicparams[constants.NIC_MODE]
8844
        link = nicparams[constants.NIC_LINK]
8845
        args['nics'].append((ip, mac, mode, link))
8846
      if constants.DDM_ADD in nic_override:
8847
        ip = nic_override[constants.DDM_ADD].get('ip', None)
8848
        mac = nic_override[constants.DDM_ADD]['mac']
8849
        nicparams = self.nic_pnew[constants.DDM_ADD]
8850
        mode = nicparams[constants.NIC_MODE]
8851
        link = nicparams[constants.NIC_LINK]
8852
        args['nics'].append((ip, mac, mode, link))
8853
      elif constants.DDM_REMOVE in nic_override:
8854
        del args['nics'][-1]
8855

    
8856
    env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
8857
    if self.op.disk_template:
8858
      env["NEW_DISK_TEMPLATE"] = self.op.disk_template
8859
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
8860
    return env, nl, nl
8861

    
8862
  def CheckPrereq(self):
8863
    """Check prerequisites.
8864

8865
    This only checks the instance list against the existing names.
8866

8867
    """
8868
    # checking the new params on the primary/secondary nodes
8869

    
8870
    instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
8871
    cluster = self.cluster = self.cfg.GetClusterInfo()
8872
    assert self.instance is not None, \
8873
      "Cannot retrieve locked instance %s" % self.op.instance_name
8874
    pnode = instance.primary_node
8875
    nodelist = list(instance.all_nodes)
8876

    
8877
    # OS change
8878
    if self.op.os_name and not self.op.force:
8879
      _CheckNodeHasOS(self, instance.primary_node, self.op.os_name,
8880
                      self.op.force_variant)
8881
      instance_os = self.op.os_name
8882
    else:
8883
      instance_os = instance.os
8884

    
8885
    if self.op.disk_template:
8886
      if instance.disk_template == self.op.disk_template:
8887
        raise errors.OpPrereqError("Instance already has disk template %s" %
8888
                                   instance.disk_template, errors.ECODE_INVAL)
8889

    
8890
      if (instance.disk_template,
8891
          self.op.disk_template) not in self._DISK_CONVERSIONS:
8892
        raise errors.OpPrereqError("Unsupported disk template conversion from"
8893
                                   " %s to %s" % (instance.disk_template,
8894
                                                  self.op.disk_template),
8895
                                   errors.ECODE_INVAL)
8896
      _CheckInstanceDown(self, instance, "cannot change disk template")
8897
      if self.op.disk_template in constants.DTS_NET_MIRROR:
8898
        if self.op.remote_node == pnode:
8899
          raise errors.OpPrereqError("Given new secondary node %s is the same"
8900
                                     " as the primary node of the instance" %
8901
                                     self.op.remote_node, errors.ECODE_STATE)
8902
        _CheckNodeOnline(self, self.op.remote_node)
8903
        _CheckNodeNotDrained(self, self.op.remote_node)
8904
        disks = [{"size": d.size} for d in instance.disks]
8905
        required = _ComputeDiskSize(self.op.disk_template, disks)
8906
        _CheckNodesFreeDisk(self, [self.op.remote_node], required)
8907

    
8908
    # hvparams processing
8909
    if self.op.hvparams:
8910
      hv_type = instance.hypervisor
8911
      i_hvdict = _GetUpdatedParams(instance.hvparams, self.op.hvparams)
8912
      utils.ForceDictType(i_hvdict, constants.HVS_PARAMETER_TYPES)
8913
      hv_new = cluster.SimpleFillHV(hv_type, instance.os, i_hvdict)
8914

    
8915
      # local check
8916
      hypervisor.GetHypervisor(hv_type).CheckParameterSyntax(hv_new)
8917
      _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
8918
      self.hv_new = hv_new # the new actual values
8919
      self.hv_inst = i_hvdict # the new dict (without defaults)
8920
    else:
8921
      self.hv_new = self.hv_inst = {}
8922

    
8923
    # beparams processing
8924
    if self.op.beparams:
8925
      i_bedict = _GetUpdatedParams(instance.beparams, self.op.beparams,
8926
                                   use_none=True)
8927
      utils.ForceDictType(i_bedict, constants.BES_PARAMETER_TYPES)
8928
      be_new = cluster.SimpleFillBE(i_bedict)
8929
      self.be_new = be_new # the new actual values
8930
      self.be_inst = i_bedict # the new dict (without defaults)
8931
    else:
8932
      self.be_new = self.be_inst = {}
8933

    
8934
    # osparams processing
8935
    if self.op.osparams:
8936
      i_osdict = _GetUpdatedParams(instance.osparams, self.op.osparams)
8937
      _CheckOSParams(self, True, nodelist, instance_os, i_osdict)
8938
      self.os_new = cluster.SimpleFillOS(instance_os, i_osdict)
8939
      self.os_inst = i_osdict # the new dict (without defaults)
8940
    else:
8941
      self.os_new = self.os_inst = {}
8942

    
8943
    self.warn = []
8944

    
8945
    if constants.BE_MEMORY in self.op.beparams and not self.op.force:
8946
      mem_check_list = [pnode]
8947
      if be_new[constants.BE_AUTO_BALANCE]:
8948
        # either we changed auto_balance to yes or it was from before
8949
        mem_check_list.extend(instance.secondary_nodes)
8950
      instance_info = self.rpc.call_instance_info(pnode, instance.name,
8951
                                                  instance.hypervisor)
8952
      nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
8953
                                         instance.hypervisor)
8954
      pninfo = nodeinfo[pnode]
8955
      msg = pninfo.fail_msg
8956
      if msg:
8957
        # Assume the primary node is unreachable and go ahead
8958
        self.warn.append("Can't get info from primary node %s: %s" %
8959
                         (pnode,  msg))
8960
      elif not isinstance(pninfo.payload.get('memory_free', None), int):
8961
        self.warn.append("Node data from primary node %s doesn't contain"
8962
                         " free memory information" % pnode)
8963
      elif instance_info.fail_msg:
8964
        self.warn.append("Can't get instance runtime information: %s" %
8965
                        instance_info.fail_msg)
8966
      else:
8967
        if instance_info.payload:
8968
          current_mem = int(instance_info.payload['memory'])
8969
        else:
8970
          # Assume instance not running
8971
          # (there is a slight race condition here, but it's not very probable,
8972
          # and we have no other way to check)
8973
          current_mem = 0
8974
        miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
8975
                    pninfo.payload['memory_free'])
8976
        if miss_mem > 0:
8977
          raise errors.OpPrereqError("This change will prevent the instance"
8978
                                     " from starting, due to %d MB of memory"
8979
                                     " missing on its primary node" % miss_mem,
8980
                                     errors.ECODE_NORES)
8981

    
8982
      if be_new[constants.BE_AUTO_BALANCE]:
8983
        for node, nres in nodeinfo.items():
8984
          if node not in instance.secondary_nodes:
8985
            continue
8986
          msg = nres.fail_msg
8987
          if msg:
8988
            self.warn.append("Can't get info from secondary node %s: %s" %
8989
                             (node, msg))
8990
          elif not isinstance(nres.payload.get('memory_free', None), int):
8991
            self.warn.append("Secondary node %s didn't return free"
8992
                             " memory information" % node)
8993
          elif be_new[constants.BE_MEMORY] > nres.payload['memory_free']:
8994
            self.warn.append("Not enough memory to failover instance to"
8995
                             " secondary node %s" % node)
8996

    
8997
    # NIC processing
8998
    self.nic_pnew = {}
8999
    self.nic_pinst = {}
9000
    for nic_op, nic_dict in self.op.nics:
9001
      if nic_op == constants.DDM_REMOVE:
9002
        if not instance.nics:
9003
          raise errors.OpPrereqError("Instance has no NICs, cannot remove",
9004
                                     errors.ECODE_INVAL)
9005
        continue
9006
      if nic_op != constants.DDM_ADD:
9007
        # an existing nic
9008
        if not instance.nics:
9009
          raise errors.OpPrereqError("Invalid NIC index %s, instance has"
9010
                                     " no NICs" % nic_op,
9011
                                     errors.ECODE_INVAL)
9012
        if nic_op < 0 or nic_op >= len(instance.nics):
9013
          raise errors.OpPrereqError("Invalid NIC index %s, valid values"
9014
                                     " are 0 to %d" %
9015
                                     (nic_op, len(instance.nics) - 1),
9016
                                     errors.ECODE_INVAL)
9017
        old_nic_params = instance.nics[nic_op].nicparams
9018
        old_nic_ip = instance.nics[nic_op].ip
9019
      else:
9020
        old_nic_params = {}
9021
        old_nic_ip = None
9022

    
9023
      update_params_dict = dict([(key, nic_dict[key])
9024
                                 for key in constants.NICS_PARAMETERS
9025
                                 if key in nic_dict])
9026

    
9027
      if 'bridge' in nic_dict:
9028
        update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
9029

    
9030
      new_nic_params = _GetUpdatedParams(old_nic_params,
9031
                                         update_params_dict)
9032
      utils.ForceDictType(new_nic_params, constants.NICS_PARAMETER_TYPES)
9033
      new_filled_nic_params = cluster.SimpleFillNIC(new_nic_params)
9034
      objects.NIC.CheckParameterSyntax(new_filled_nic_params)
9035
      self.nic_pinst[nic_op] = new_nic_params
9036
      self.nic_pnew[nic_op] = new_filled_nic_params
9037
      new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
9038

    
9039
      if new_nic_mode == constants.NIC_MODE_BRIDGED:
9040
        nic_bridge = new_filled_nic_params[constants.NIC_LINK]
9041
        msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
9042
        if msg:
9043
          msg = "Error checking bridges on node %s: %s" % (pnode, msg)
9044
          if self.op.force:
9045
            self.warn.append(msg)
9046
          else:
9047
            raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
9048
      if new_nic_mode == constants.NIC_MODE_ROUTED:
9049
        if 'ip' in nic_dict:
9050
          nic_ip = nic_dict['ip']
9051
        else:
9052
          nic_ip = old_nic_ip
9053
        if nic_ip is None:
9054
          raise errors.OpPrereqError('Cannot set the nic ip to None'
9055
                                     ' on a routed nic', errors.ECODE_INVAL)
9056
      if 'mac' in nic_dict:
9057
        nic_mac = nic_dict['mac']
9058
        if nic_mac is None:
9059
          raise errors.OpPrereqError('Cannot set the nic mac to None',
9060
                                     errors.ECODE_INVAL)
9061
        elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
9062
          # otherwise generate the mac
9063
          nic_dict['mac'] = self.cfg.GenerateMAC(self.proc.GetECId())
9064
        else:
9065
          # or validate/reserve the current one
9066
          try:
9067
            self.cfg.ReserveMAC(nic_mac, self.proc.GetECId())
9068
          except errors.ReservationError:
9069
            raise errors.OpPrereqError("MAC address %s already in use"
9070
                                       " in cluster" % nic_mac,
9071
                                       errors.ECODE_NOTUNIQUE)
9072

    
9073
    # DISK processing
9074
    if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
9075
      raise errors.OpPrereqError("Disk operations not supported for"
9076
                                 " diskless instances",
9077
                                 errors.ECODE_INVAL)
9078
    for disk_op, _ in self.op.disks:
9079
      if disk_op == constants.DDM_REMOVE:
9080
        if len(instance.disks) == 1:
9081
          raise errors.OpPrereqError("Cannot remove the last disk of"
9082
                                     " an instance", errors.ECODE_INVAL)
9083
        _CheckInstanceDown(self, instance, "cannot remove disks")
9084

    
9085
      if (disk_op == constants.DDM_ADD and
9086
          len(instance.nics) >= constants.MAX_DISKS):
9087
        raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
9088
                                   " add more" % constants.MAX_DISKS,
9089
                                   errors.ECODE_STATE)
9090
      if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
9091
        # an existing disk
9092
        if disk_op < 0 or disk_op >= len(instance.disks):
9093
          raise errors.OpPrereqError("Invalid disk index %s, valid values"
9094
                                     " are 0 to %d" %
9095
                                     (disk_op, len(instance.disks)),
9096
                                     errors.ECODE_INVAL)
9097

    
9098
    return
9099

    
9100
  def _ConvertPlainToDrbd(self, feedback_fn):
9101
    """Converts an instance from plain to drbd.
9102

9103
    """
9104
    feedback_fn("Converting template to drbd")
9105
    instance = self.instance
9106
    pnode = instance.primary_node
9107
    snode = self.op.remote_node
9108

    
9109
    # create a fake disk info for _GenerateDiskTemplate
9110
    disk_info = [{"size": d.size, "mode": d.mode} for d in instance.disks]
9111
    new_disks = _GenerateDiskTemplate(self, self.op.disk_template,
9112
                                      instance.name, pnode, [snode],
9113
                                      disk_info, None, None, 0)
9114
    info = _GetInstanceInfoText(instance)
9115
    feedback_fn("Creating aditional volumes...")
9116
    # first, create the missing data and meta devices
9117
    for disk in new_disks:
9118
      # unfortunately this is... not too nice
9119
      _CreateSingleBlockDev(self, pnode, instance, disk.children[1],
9120
                            info, True)
9121
      for child in disk.children:
9122
        _CreateSingleBlockDev(self, snode, instance, child, info, True)
9123
    # at this stage, all new LVs have been created, we can rename the
9124
    # old ones
9125
    feedback_fn("Renaming original volumes...")
9126
    rename_list = [(o, n.children[0].logical_id)
9127
                   for (o, n) in zip(instance.disks, new_disks)]
9128
    result = self.rpc.call_blockdev_rename(pnode, rename_list)
9129
    result.Raise("Failed to rename original LVs")
9130

    
9131
    feedback_fn("Initializing DRBD devices...")
9132
    # all child devices are in place, we can now create the DRBD devices
9133
    for disk in new_disks:
9134
      for node in [pnode, snode]:
9135
        f_create = node == pnode
9136
        _CreateSingleBlockDev(self, node, instance, disk, info, f_create)
9137

    
9138
    # at this point, the instance has been modified
9139
    instance.disk_template = constants.DT_DRBD8
9140
    instance.disks = new_disks
9141
    self.cfg.Update(instance, feedback_fn)
9142

    
9143
    # disks are created, waiting for sync
9144
    disk_abort = not _WaitForSync(self, instance)
9145
    if disk_abort:
9146
      raise errors.OpExecError("There are some degraded disks for"
9147
                               " this instance, please cleanup manually")
9148

    
9149
  def _ConvertDrbdToPlain(self, feedback_fn):
9150
    """Converts an instance from drbd to plain.
9151

9152
    """
9153
    instance = self.instance
9154
    assert len(instance.secondary_nodes) == 1
9155
    pnode = instance.primary_node
9156
    snode = instance.secondary_nodes[0]
9157
    feedback_fn("Converting template to plain")
9158

    
9159
    old_disks = instance.disks
9160
    new_disks = [d.children[0] for d in old_disks]
9161

    
9162
    # copy over size and mode
9163
    for parent, child in zip(old_disks, new_disks):
9164
      child.size = parent.size
9165
      child.mode = parent.mode
9166

    
9167
    # update instance structure
9168
    instance.disks = new_disks
9169
    instance.disk_template = constants.DT_PLAIN
9170
    self.cfg.Update(instance, feedback_fn)
9171

    
9172
    feedback_fn("Removing volumes on the secondary node...")
9173
    for disk in old_disks:
9174
      self.cfg.SetDiskID(disk, snode)
9175
      msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg
9176
      if msg:
9177
        self.LogWarning("Could not remove block device %s on node %s,"
9178
                        " continuing anyway: %s", disk.iv_name, snode, msg)
9179

    
9180
    feedback_fn("Removing unneeded volumes on the primary node...")
9181
    for idx, disk in enumerate(old_disks):
9182
      meta = disk.children[1]
9183
      self.cfg.SetDiskID(meta, pnode)
9184
      msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg
9185
      if msg:
9186
        self.LogWarning("Could not remove metadata for disk %d on node %s,"
9187
                        " continuing anyway: %s", idx, pnode, msg)
9188

    
9189

    
9190
  def Exec(self, feedback_fn):
9191
    """Modifies an instance.
9192

9193
    All parameters take effect only at the next restart of the instance.
9194

9195
    """
9196
    # Process here the warnings from CheckPrereq, as we don't have a
9197
    # feedback_fn there.
9198
    for warn in self.warn:
9199
      feedback_fn("WARNING: %s" % warn)
9200

    
9201
    result = []
9202
    instance = self.instance
9203
    # disk changes
9204
    for disk_op, disk_dict in self.op.disks:
9205
      if disk_op == constants.DDM_REMOVE:
9206
        # remove the last disk
9207
        device = instance.disks.pop()
9208
        device_idx = len(instance.disks)
9209
        for node, disk in device.ComputeNodeTree(instance.primary_node):
9210
          self.cfg.SetDiskID(disk, node)
9211
          msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
9212
          if msg:
9213
            self.LogWarning("Could not remove disk/%d on node %s: %s,"
9214
                            " continuing anyway", device_idx, node, msg)
9215
        result.append(("disk/%d" % device_idx, "remove"))
9216
      elif disk_op == constants.DDM_ADD:
9217
        # add a new disk
9218
        if instance.disk_template == constants.DT_FILE:
9219
          file_driver, file_path = instance.disks[0].logical_id
9220
          file_path = os.path.dirname(file_path)
9221
        else:
9222
          file_driver = file_path = None
9223
        disk_idx_base = len(instance.disks)
9224
        new_disk = _GenerateDiskTemplate(self,
9225
                                         instance.disk_template,
9226
                                         instance.name, instance.primary_node,
9227
                                         instance.secondary_nodes,
9228
                                         [disk_dict],
9229
                                         file_path,
9230
                                         file_driver,
9231
                                         disk_idx_base)[0]
9232
        instance.disks.append(new_disk)
9233
        info = _GetInstanceInfoText(instance)
9234

    
9235
        logging.info("Creating volume %s for instance %s",
9236
                     new_disk.iv_name, instance.name)
9237
        # Note: this needs to be kept in sync with _CreateDisks
9238
        #HARDCODE
9239
        for node in instance.all_nodes:
9240
          f_create = node == instance.primary_node
9241
          try:
9242
            _CreateBlockDev(self, node, instance, new_disk,
9243
                            f_create, info, f_create)
9244
          except errors.OpExecError, err:
9245
            self.LogWarning("Failed to create volume %s (%s) on"
9246
                            " node %s: %s",
9247
                            new_disk.iv_name, new_disk, node, err)
9248
        result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
9249
                       (new_disk.size, new_disk.mode)))
9250
      else:
9251
        # change a given disk
9252
        instance.disks[disk_op].mode = disk_dict['mode']
9253
        result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
9254

    
9255
    if self.op.disk_template:
9256
      r_shut = _ShutdownInstanceDisks(self, instance)
9257
      if not r_shut:
9258
        raise errors.OpExecError("Cannot shutdow instance disks, unable to"
9259
                                 " proceed with disk template conversion")
9260
      mode = (instance.disk_template, self.op.disk_template)
9261
      try:
9262
        self._DISK_CONVERSIONS[mode](self, feedback_fn)
9263
      except:
9264
        self.cfg.ReleaseDRBDMinors(instance.name)
9265
        raise
9266
      result.append(("disk_template", self.op.disk_template))
9267

    
9268
    # NIC changes
9269
    for nic_op, nic_dict in self.op.nics:
9270
      if nic_op == constants.DDM_REMOVE:
9271
        # remove the last nic
9272
        del instance.nics[-1]
9273
        result.append(("nic.%d" % len(instance.nics), "remove"))
9274
      elif nic_op == constants.DDM_ADD:
9275
        # mac and bridge should be set, by now
9276
        mac = nic_dict['mac']
9277
        ip = nic_dict.get('ip', None)
9278
        nicparams = self.nic_pinst[constants.DDM_ADD]
9279
        new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
9280
        instance.nics.append(new_nic)
9281
        result.append(("nic.%d" % (len(instance.nics) - 1),
9282
                       "add:mac=%s,ip=%s,mode=%s,link=%s" %
9283
                       (new_nic.mac, new_nic.ip,
9284
                        self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
9285
                        self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
9286
                       )))
9287
      else:
9288
        for key in 'mac', 'ip':
9289
          if key in nic_dict:
9290
            setattr(instance.nics[nic_op], key, nic_dict[key])
9291
        if nic_op in self.nic_pinst:
9292
          instance.nics[nic_op].nicparams = self.nic_pinst[nic_op]
9293
        for key, val in nic_dict.iteritems():
9294
          result.append(("nic.%s/%d" % (key, nic_op), val))
9295

    
9296
    # hvparams changes
9297
    if self.op.hvparams:
9298
      instance.hvparams = self.hv_inst
9299
      for key, val in self.op.hvparams.iteritems():
9300
        result.append(("hv/%s" % key, val))
9301

    
9302
    # beparams changes
9303
    if self.op.beparams:
9304
      instance.beparams = self.be_inst
9305
      for key, val in self.op.beparams.iteritems():
9306
        result.append(("be/%s" % key, val))
9307

    
9308
    # OS change
9309
    if self.op.os_name:
9310
      instance.os = self.op.os_name
9311

    
9312
    # osparams changes
9313
    if self.op.osparams:
9314
      instance.osparams = self.os_inst
9315
      for key, val in self.op.osparams.iteritems():
9316
        result.append(("os/%s" % key, val))
9317

    
9318
    self.cfg.Update(instance, feedback_fn)
9319

    
9320
    return result
9321

    
9322
  _DISK_CONVERSIONS = {
9323
    (constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd,
9324
    (constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain,
9325
    }
9326

    
9327

    
9328
class LUQueryExports(NoHooksLU):
9329
  """Query the exports list
9330

9331
  """
9332
  _OP_PARAMS = [
9333
    ("nodes", _EmptyList, _TListOf(_TNonEmptyString)),
9334
    ("use_locking", False, _TBool),
9335
    ]
9336
  REQ_BGL = False
9337

    
9338
  def ExpandNames(self):
9339
    self.needed_locks = {}
9340
    self.share_locks[locking.LEVEL_NODE] = 1
9341
    if not self.op.nodes:
9342
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9343
    else:
9344
      self.needed_locks[locking.LEVEL_NODE] = \
9345
        _GetWantedNodes(self, self.op.nodes)
9346

    
9347
  def Exec(self, feedback_fn):
9348
    """Compute the list of all the exported system images.
9349

9350
    @rtype: dict
9351
    @return: a dictionary with the structure node->(export-list)
9352
        where export-list is a list of the instances exported on
9353
        that node.
9354

9355
    """
9356
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
9357
    rpcresult = self.rpc.call_export_list(self.nodes)
9358
    result = {}
9359
    for node in rpcresult:
9360
      if rpcresult[node].fail_msg:
9361
        result[node] = False
9362
      else:
9363
        result[node] = rpcresult[node].payload
9364

    
9365
    return result
9366

    
9367

    
9368
class LUPrepareExport(NoHooksLU):
9369
  """Prepares an instance for an export and returns useful information.
9370

9371
  """
9372
  _OP_PARAMS = [
9373
    _PInstanceName,
9374
    ("mode", _NoDefault, _TElemOf(constants.EXPORT_MODES)),
9375
    ]
9376
  REQ_BGL = False
9377

    
9378
  def ExpandNames(self):
9379
    self._ExpandAndLockInstance()
9380

    
9381
  def CheckPrereq(self):
9382
    """Check prerequisites.
9383

9384
    """
9385
    instance_name = self.op.instance_name
9386

    
9387
    self.instance = self.cfg.GetInstanceInfo(instance_name)
9388
    assert self.instance is not None, \
9389
          "Cannot retrieve locked instance %s" % self.op.instance_name
9390
    _CheckNodeOnline(self, self.instance.primary_node)
9391

    
9392
    self._cds = _GetClusterDomainSecret()
9393

    
9394
  def Exec(self, feedback_fn):
9395
    """Prepares an instance for an export.
9396

9397
    """
9398
    instance = self.instance
9399

    
9400
    if self.op.mode == constants.EXPORT_MODE_REMOTE:
9401
      salt = utils.GenerateSecret(8)
9402

    
9403
      feedback_fn("Generating X509 certificate on %s" % instance.primary_node)
9404
      result = self.rpc.call_x509_cert_create(instance.primary_node,
9405
                                              constants.RIE_CERT_VALIDITY)
9406
      result.Raise("Can't create X509 key and certificate on %s" % result.node)
9407

    
9408
      (name, cert_pem) = result.payload
9409

    
9410
      cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
9411
                                             cert_pem)
9412

    
9413
      return {
9414
        "handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds),
9415
        "x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt),
9416
                          salt),
9417
        "x509_ca": utils.SignX509Certificate(cert, self._cds, salt),
9418
        }
9419

    
9420
    return None
9421

    
9422

    
9423
class LUExportInstance(LogicalUnit):
9424
  """Export an instance to an image in the cluster.
9425

9426
  """
9427
  HPATH = "instance-export"
9428
  HTYPE = constants.HTYPE_INSTANCE
9429
  _OP_PARAMS = [
9430
    _PInstanceName,
9431
    ("target_node", _NoDefault, _TOr(_TNonEmptyString, _TList)),
9432
    ("shutdown", True, _TBool),
9433
    _PShutdownTimeout,
9434
    ("remove_instance", False, _TBool),
9435
    ("ignore_remove_failures", False, _TBool),
9436
    ("mode", constants.EXPORT_MODE_LOCAL, _TElemOf(constants.EXPORT_MODES)),
9437
    ("x509_key_name", None, _TOr(_TList, _TNone)),
9438
    ("destination_x509_ca", None, _TMaybeString),
9439
    ]
9440
  REQ_BGL = False
9441

    
9442
  def CheckArguments(self):
9443
    """Check the arguments.
9444

9445
    """
9446
    self.x509_key_name = self.op.x509_key_name
9447
    self.dest_x509_ca_pem = self.op.destination_x509_ca
9448

    
9449
    if self.op.remove_instance and not self.op.shutdown:
9450
      raise errors.OpPrereqError("Can not remove instance without shutting it"
9451
                                 " down before")
9452

    
9453
    if self.op.mode == constants.EXPORT_MODE_REMOTE:
9454
      if not self.x509_key_name:
9455
        raise errors.OpPrereqError("Missing X509 key name for encryption",
9456
                                   errors.ECODE_INVAL)
9457

    
9458
      if not self.dest_x509_ca_pem:
9459
        raise errors.OpPrereqError("Missing destination X509 CA",
9460
                                   errors.ECODE_INVAL)
9461

    
9462
  def ExpandNames(self):
9463
    self._ExpandAndLockInstance()
9464

    
9465
    # Lock all nodes for local exports
9466
    if self.op.mode == constants.EXPORT_MODE_LOCAL:
9467
      # FIXME: lock only instance primary and destination node
9468
      #
9469
      # Sad but true, for now we have do lock all nodes, as we don't know where
9470
      # the previous export might be, and in this LU we search for it and
9471
      # remove it from its current node. In the future we could fix this by:
9472
      #  - making a tasklet to search (share-lock all), then create the
9473
      #    new one, then one to remove, after
9474
      #  - removing the removal operation altogether
9475
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9476

    
9477
  def DeclareLocks(self, level):
9478
    """Last minute lock declaration."""
9479
    # All nodes are locked anyway, so nothing to do here.
9480

    
9481
  def BuildHooksEnv(self):
9482
    """Build hooks env.
9483

9484
    This will run on the master, primary node and target node.
9485

9486
    """
9487
    env = {
9488
      "EXPORT_MODE": self.op.mode,
9489
      "EXPORT_NODE": self.op.target_node,
9490
      "EXPORT_DO_SHUTDOWN": self.op.shutdown,
9491
      "SHUTDOWN_TIMEOUT": self.op.shutdown_timeout,
9492
      # TODO: Generic function for boolean env variables
9493
      "REMOVE_INSTANCE": str(bool(self.op.remove_instance)),
9494
      }
9495

    
9496
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
9497

    
9498
    nl = [self.cfg.GetMasterNode(), self.instance.primary_node]
9499

    
9500
    if self.op.mode == constants.EXPORT_MODE_LOCAL:
9501
      nl.append(self.op.target_node)
9502

    
9503
    return env, nl, nl
9504

    
9505
  def CheckPrereq(self):
9506
    """Check prerequisites.
9507

9508
    This checks that the instance and node names are valid.
9509

9510
    """
9511
    instance_name = self.op.instance_name
9512

    
9513
    self.instance = self.cfg.GetInstanceInfo(instance_name)
9514
    assert self.instance is not None, \
9515
          "Cannot retrieve locked instance %s" % self.op.instance_name
9516
    _CheckNodeOnline(self, self.instance.primary_node)
9517

    
9518
    if self.op.mode == constants.EXPORT_MODE_LOCAL:
9519
      self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
9520
      self.dst_node = self.cfg.GetNodeInfo(self.op.target_node)
9521
      assert self.dst_node is not None
9522

    
9523
      _CheckNodeOnline(self, self.dst_node.name)
9524
      _CheckNodeNotDrained(self, self.dst_node.name)
9525

    
9526
      self._cds = None
9527
      self.dest_disk_info = None
9528
      self.dest_x509_ca = None
9529

    
9530
    elif self.op.mode == constants.EXPORT_MODE_REMOTE:
9531
      self.dst_node = None
9532

    
9533
      if len(self.op.target_node) != len(self.instance.disks):
9534
        raise errors.OpPrereqError(("Received destination information for %s"
9535
                                    " disks, but instance %s has %s disks") %
9536
                                   (len(self.op.target_node), instance_name,
9537
                                    len(self.instance.disks)),
9538
                                   errors.ECODE_INVAL)
9539

    
9540
      cds = _GetClusterDomainSecret()
9541

    
9542
      # Check X509 key name
9543
      try:
9544
        (key_name, hmac_digest, hmac_salt) = self.x509_key_name
9545
      except (TypeError, ValueError), err:
9546
        raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err)
9547

    
9548
      if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt):
9549
        raise errors.OpPrereqError("HMAC for X509 key name is wrong",
9550
                                   errors.ECODE_INVAL)
9551

    
9552
      # Load and verify CA
9553
      try:
9554
        (cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds)
9555
      except OpenSSL.crypto.Error, err:
9556
        raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" %
9557
                                   (err, ), errors.ECODE_INVAL)
9558

    
9559
      (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
9560
      if errcode is not None:
9561
        raise errors.OpPrereqError("Invalid destination X509 CA (%s)" %
9562
                                   (msg, ), errors.ECODE_INVAL)
9563

    
9564
      self.dest_x509_ca = cert
9565

    
9566
      # Verify target information
9567
      disk_info = []
9568
      for idx, disk_data in enumerate(self.op.target_node):
9569
        try:
9570
          (host, port, magic) = \
9571
            masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data)
9572
        except errors.GenericError, err:
9573
          raise errors.OpPrereqError("Target info for disk %s: %s" %
9574
                                     (idx, err), errors.ECODE_INVAL)
9575

    
9576
        disk_info.append((host, port, magic))
9577

    
9578
      assert len(disk_info) == len(self.op.target_node)
9579
      self.dest_disk_info = disk_info
9580

    
9581
    else:
9582
      raise errors.ProgrammerError("Unhandled export mode %r" %
9583
                                   self.op.mode)
9584

    
9585
    # instance disk type verification
9586
    # TODO: Implement export support for file-based disks
9587
    for disk in self.instance.disks:
9588
      if disk.dev_type == constants.LD_FILE:
9589
        raise errors.OpPrereqError("Export not supported for instances with"
9590
                                   " file-based disks", errors.ECODE_INVAL)
9591

    
9592
  def _CleanupExports(self, feedback_fn):
9593
    """Removes exports of current instance from all other nodes.
9594

9595
    If an instance in a cluster with nodes A..D was exported to node C, its
9596
    exports will be removed from the nodes A, B and D.
9597

9598
    """
9599
    assert self.op.mode != constants.EXPORT_MODE_REMOTE
9600

    
9601
    nodelist = self.cfg.GetNodeList()
9602
    nodelist.remove(self.dst_node.name)
9603

    
9604
    # on one-node clusters nodelist will be empty after the removal
9605
    # if we proceed the backup would be removed because OpQueryExports
9606
    # substitutes an empty list with the full cluster node list.
9607
    iname = self.instance.name
9608
    if nodelist:
9609
      feedback_fn("Removing old exports for instance %s" % iname)
9610
      exportlist = self.rpc.call_export_list(nodelist)
9611
      for node in exportlist:
9612
        if exportlist[node].fail_msg:
9613
          continue
9614
        if iname in exportlist[node].payload:
9615
          msg = self.rpc.call_export_remove(node, iname).fail_msg
9616
          if msg:
9617
            self.LogWarning("Could not remove older export for instance %s"
9618
                            " on node %s: %s", iname, node, msg)
9619

    
9620
  def Exec(self, feedback_fn):
9621
    """Export an instance to an image in the cluster.
9622

9623
    """
9624
    assert self.op.mode in constants.EXPORT_MODES
9625

    
9626
    instance = self.instance
9627
    src_node = instance.primary_node
9628

    
9629
    if self.op.shutdown:
9630
      # shutdown the instance, but not the disks
9631
      feedback_fn("Shutting down instance %s" % instance.name)
9632
      result = self.rpc.call_instance_shutdown(src_node, instance,
9633
                                               self.op.shutdown_timeout)
9634
      # TODO: Maybe ignore failures if ignore_remove_failures is set
9635
      result.Raise("Could not shutdown instance %s on"
9636
                   " node %s" % (instance.name, src_node))
9637

    
9638
    # set the disks ID correctly since call_instance_start needs the
9639
    # correct drbd minor to create the symlinks
9640
    for disk in instance.disks:
9641
      self.cfg.SetDiskID(disk, src_node)
9642

    
9643
    activate_disks = (not instance.admin_up)
9644

    
9645
    if activate_disks:
9646
      # Activate the instance disks if we'exporting a stopped instance
9647
      feedback_fn("Activating disks for %s" % instance.name)
9648
      _StartInstanceDisks(self, instance, None)
9649

    
9650
    try:
9651
      helper = masterd.instance.ExportInstanceHelper(self, feedback_fn,
9652
                                                     instance)
9653

    
9654
      helper.CreateSnapshots()
9655
      try:
9656
        if (self.op.shutdown and instance.admin_up and
9657
            not self.op.remove_instance):
9658
          assert not activate_disks
9659
          feedback_fn("Starting instance %s" % instance.name)
9660
          result = self.rpc.call_instance_start(src_node, instance, None, None)
9661
          msg = result.fail_msg
9662
          if msg:
9663
            feedback_fn("Failed to start instance: %s" % msg)
9664
            _ShutdownInstanceDisks(self, instance)
9665
            raise errors.OpExecError("Could not start instance: %s" % msg)
9666

    
9667
        if self.op.mode == constants.EXPORT_MODE_LOCAL:
9668
          (fin_resu, dresults) = helper.LocalExport(self.dst_node)
9669
        elif self.op.mode == constants.EXPORT_MODE_REMOTE:
9670
          connect_timeout = constants.RIE_CONNECT_TIMEOUT
9671
          timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
9672

    
9673
          (key_name, _, _) = self.x509_key_name
9674

    
9675
          dest_ca_pem = \
9676
            OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM,
9677
                                            self.dest_x509_ca)
9678

    
9679
          (fin_resu, dresults) = helper.RemoteExport(self.dest_disk_info,
9680
                                                     key_name, dest_ca_pem,
9681
                                                     timeouts)
9682
      finally:
9683
        helper.Cleanup()
9684

    
9685
      # Check for backwards compatibility
9686
      assert len(dresults) == len(instance.disks)
9687
      assert compat.all(isinstance(i, bool) for i in dresults), \
9688
             "Not all results are boolean: %r" % dresults
9689

    
9690
    finally:
9691
      if activate_disks:
9692
        feedback_fn("Deactivating disks for %s" % instance.name)
9693
        _ShutdownInstanceDisks(self, instance)
9694

    
9695
    if not (compat.all(dresults) and fin_resu):
9696
      failures = []
9697
      if not fin_resu:
9698
        failures.append("export finalization")
9699
      if not compat.all(dresults):
9700
        fdsk = utils.CommaJoin(idx for (idx, dsk) in enumerate(dresults)
9701
                               if not dsk)
9702
        failures.append("disk export: disk(s) %s" % fdsk)
9703

    
9704
      raise errors.OpExecError("Export failed, errors in %s" %
9705
                               utils.CommaJoin(failures))
9706

    
9707
    # At this point, the export was successful, we can cleanup/finish
9708

    
9709
    # Remove instance if requested
9710
    if self.op.remove_instance:
9711
      feedback_fn("Removing instance %s" % instance.name)
9712
      _RemoveInstance(self, feedback_fn, instance,
9713
                      self.op.ignore_remove_failures)
9714

    
9715
    if self.op.mode == constants.EXPORT_MODE_LOCAL:
9716
      self._CleanupExports(feedback_fn)
9717

    
9718
    return fin_resu, dresults
9719

    
9720

    
9721
class LURemoveExport(NoHooksLU):
9722
  """Remove exports related to the named instance.
9723

9724
  """
9725
  _OP_PARAMS = [
9726
    _PInstanceName,
9727
    ]
9728
  REQ_BGL = False
9729

    
9730
  def ExpandNames(self):
9731
    self.needed_locks = {}
9732
    # We need all nodes to be locked in order for RemoveExport to work, but we
9733
    # don't need to lock the instance itself, as nothing will happen to it (and
9734
    # we can remove exports also for a removed instance)
9735
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9736

    
9737
  def Exec(self, feedback_fn):
9738
    """Remove any export.
9739

9740
    """
9741
    instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
9742
    # If the instance was not found we'll try with the name that was passed in.
9743
    # This will only work if it was an FQDN, though.
9744
    fqdn_warn = False
9745
    if not instance_name:
9746
      fqdn_warn = True
9747
      instance_name = self.op.instance_name
9748

    
9749
    locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
9750
    exportlist = self.rpc.call_export_list(locked_nodes)
9751
    found = False
9752
    for node in exportlist:
9753
      msg = exportlist[node].fail_msg
9754
      if msg:
9755
        self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
9756
        continue
9757
      if instance_name in exportlist[node].payload:
9758
        found = True
9759
        result = self.rpc.call_export_remove(node, instance_name)
9760
        msg = result.fail_msg
9761
        if msg:
9762
          logging.error("Could not remove export for instance %s"
9763
                        " on node %s: %s", instance_name, node, msg)
9764

    
9765
    if fqdn_warn and not found:
9766
      feedback_fn("Export not found. If trying to remove an export belonging"
9767
                  " to a deleted instance please use its Fully Qualified"
9768
                  " Domain Name.")
9769

    
9770

    
9771
class TagsLU(NoHooksLU): # pylint: disable-msg=W0223
9772
  """Generic tags LU.
9773

9774
  This is an abstract class which is the parent of all the other tags LUs.
9775

9776
  """
9777

    
9778
  def ExpandNames(self):
9779
    self.needed_locks = {}
9780
    if self.op.kind == constants.TAG_NODE:
9781
      self.op.name = _ExpandNodeName(self.cfg, self.op.name)
9782
      self.needed_locks[locking.LEVEL_NODE] = self.op.name
9783
    elif self.op.kind == constants.TAG_INSTANCE:
9784
      self.op.name = _ExpandInstanceName(self.cfg, self.op.name)
9785
      self.needed_locks[locking.LEVEL_INSTANCE] = self.op.name
9786

    
9787
  def CheckPrereq(self):
9788
    """Check prerequisites.
9789

9790
    """
9791
    if self.op.kind == constants.TAG_CLUSTER:
9792
      self.target = self.cfg.GetClusterInfo()
9793
    elif self.op.kind == constants.TAG_NODE:
9794
      self.target = self.cfg.GetNodeInfo(self.op.name)
9795
    elif self.op.kind == constants.TAG_INSTANCE:
9796
      self.target = self.cfg.GetInstanceInfo(self.op.name)
9797
    else:
9798
      raise errors.OpPrereqError("Wrong tag type requested (%s)" %
9799
                                 str(self.op.kind), errors.ECODE_INVAL)
9800

    
9801

    
9802
class LUGetTags(TagsLU):
9803
  """Returns the tags of a given object.
9804

9805
  """
9806
  _OP_PARAMS = [
9807
    ("kind", _NoDefault, _TElemOf(constants.VALID_TAG_TYPES)),
9808
    # Name is only meaningful for nodes and instances
9809
    ("name", _NoDefault, _TMaybeString),
9810
    ]
9811
  REQ_BGL = False
9812

    
9813
  def Exec(self, feedback_fn):
9814
    """Returns the tag list.
9815

9816
    """
9817
    return list(self.target.GetTags())
9818

    
9819

    
9820
class LUSearchTags(NoHooksLU):
9821
  """Searches the tags for a given pattern.
9822

9823
  """
9824
  _OP_PARAMS = [
9825
    ("pattern", _NoDefault, _TNonEmptyString),
9826
    ]
9827
  REQ_BGL = False
9828

    
9829
  def ExpandNames(self):
9830
    self.needed_locks = {}
9831

    
9832
  def CheckPrereq(self):
9833
    """Check prerequisites.
9834

9835
    This checks the pattern passed for validity by compiling it.
9836

9837
    """
9838
    try:
9839
      self.re = re.compile(self.op.pattern)
9840
    except re.error, err:
9841
      raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
9842
                                 (self.op.pattern, err), errors.ECODE_INVAL)
9843

    
9844
  def Exec(self, feedback_fn):
9845
    """Returns the tag list.
9846

9847
    """
9848
    cfg = self.cfg
9849
    tgts = [("/cluster", cfg.GetClusterInfo())]
9850
    ilist = cfg.GetAllInstancesInfo().values()
9851
    tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
9852
    nlist = cfg.GetAllNodesInfo().values()
9853
    tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
9854
    results = []
9855
    for path, target in tgts:
9856
      for tag in target.GetTags():
9857
        if self.re.search(tag):
9858
          results.append((path, tag))
9859
    return results
9860

    
9861

    
9862
class LUAddTags(TagsLU):
9863
  """Sets a tag on a given object.
9864

9865
  """
9866
  _OP_PARAMS = [
9867
    ("kind", _NoDefault, _TElemOf(constants.VALID_TAG_TYPES)),
9868
    # Name is only meaningful for nodes and instances
9869
    ("name", _NoDefault, _TMaybeString),
9870
    ("tags", _NoDefault, _TListOf(_TNonEmptyString)),
9871
    ]
9872
  REQ_BGL = False
9873

    
9874
  def CheckPrereq(self):
9875
    """Check prerequisites.
9876

9877
    This checks the type and length of the tag name and value.
9878

9879
    """
9880
    TagsLU.CheckPrereq(self)
9881
    for tag in self.op.tags:
9882
      objects.TaggableObject.ValidateTag(tag)
9883

    
9884
  def Exec(self, feedback_fn):
9885
    """Sets the tag.
9886

9887
    """
9888
    try:
9889
      for tag in self.op.tags:
9890
        self.target.AddTag(tag)
9891
    except errors.TagError, err:
9892
      raise errors.OpExecError("Error while setting tag: %s" % str(err))
9893
    self.cfg.Update(self.target, feedback_fn)
9894

    
9895

    
9896
class LUDelTags(TagsLU):
9897
  """Delete a list of tags from a given object.
9898

9899
  """
9900
  _OP_PARAMS = [
9901
    ("kind", _NoDefault, _TElemOf(constants.VALID_TAG_TYPES)),
9902
    # Name is only meaningful for nodes and instances
9903
    ("name", _NoDefault, _TMaybeString),
9904
    ("tags", _NoDefault, _TListOf(_TNonEmptyString)),
9905
    ]
9906
  REQ_BGL = False
9907

    
9908
  def CheckPrereq(self):
9909
    """Check prerequisites.
9910

9911
    This checks that we have the given tag.
9912

9913
    """
9914
    TagsLU.CheckPrereq(self)
9915
    for tag in self.op.tags:
9916
      objects.TaggableObject.ValidateTag(tag)
9917
    del_tags = frozenset(self.op.tags)
9918
    cur_tags = self.target.GetTags()
9919
    if not del_tags <= cur_tags:
9920
      diff_tags = del_tags - cur_tags
9921
      diff_names = ["'%s'" % tag for tag in diff_tags]
9922
      diff_names.sort()
9923
      raise errors.OpPrereqError("Tag(s) %s not found" %
9924
                                 (",".join(diff_names)), errors.ECODE_NOENT)
9925

    
9926
  def Exec(self, feedback_fn):
9927
    """Remove the tag from the object.
9928

9929
    """
9930
    for tag in self.op.tags:
9931
      self.target.RemoveTag(tag)
9932
    self.cfg.Update(self.target, feedback_fn)
9933

    
9934

    
9935
class LUTestDelay(NoHooksLU):
9936
  """Sleep for a specified amount of time.
9937

9938
  This LU sleeps on the master and/or nodes for a specified amount of
9939
  time.
9940

9941
  """
9942
  _OP_PARAMS = [
9943
    ("duration", _NoDefault, _TFloat),
9944
    ("on_master", True, _TBool),
9945
    ("on_nodes", _EmptyList, _TListOf(_TNonEmptyString)),
9946
    ("repeat", 0, _TPositiveInt)
9947
    ]
9948
  REQ_BGL = False
9949

    
9950
  def ExpandNames(self):
9951
    """Expand names and set required locks.
9952

9953
    This expands the node list, if any.
9954

9955
    """
9956
    self.needed_locks = {}
9957
    if self.op.on_nodes:
9958
      # _GetWantedNodes can be used here, but is not always appropriate to use
9959
      # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
9960
      # more information.
9961
      self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
9962
      self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
9963

    
9964
  def _TestDelay(self):
9965
    """Do the actual sleep.
9966

9967
    """
9968
    if self.op.on_master:
9969
      if not utils.TestDelay(self.op.duration):
9970
        raise errors.OpExecError("Error during master delay test")
9971
    if self.op.on_nodes:
9972
      result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
9973
      for node, node_result in result.items():
9974
        node_result.Raise("Failure during rpc call to node %s" % node)
9975

    
9976
  def Exec(self, feedback_fn):
9977
    """Execute the test delay opcode, with the wanted repetitions.
9978

9979
    """
9980
    if self.op.repeat == 0:
9981
      self._TestDelay()
9982
    else:
9983
      top_value = self.op.repeat - 1
9984
      for i in range(self.op.repeat):
9985
        self.LogInfo("Test delay iteration %d/%d" % (i, top_value))
9986
        self._TestDelay()
9987

    
9988

    
9989
class LUTestJobqueue(NoHooksLU):
9990
  """Utility LU to test some aspects of the job queue.
9991

9992
  """
9993
  _OP_PARAMS = [
9994
    ("notify_waitlock", False, _TBool),
9995
    ("notify_exec", False, _TBool),
9996
    ("log_messages", _EmptyList, _TListOf(_TString)),
9997
    ("fail", False, _TBool),
9998
    ]
9999
  REQ_BGL = False
10000

    
10001
  # Must be lower than default timeout for WaitForJobChange to see whether it
10002
  # notices changed jobs
10003
  _CLIENT_CONNECT_TIMEOUT = 20.0
10004
  _CLIENT_CONFIRM_TIMEOUT = 60.0
10005

    
10006
  @classmethod
10007
  def _NotifyUsingSocket(cls, cb, errcls):
10008
    """Opens a Unix socket and waits for another program to connect.
10009

10010
    @type cb: callable
10011
    @param cb: Callback to send socket name to client
10012
    @type errcls: class
10013
    @param errcls: Exception class to use for errors
10014

10015
    """
10016
    # Using a temporary directory as there's no easy way to create temporary
10017
    # sockets without writing a custom loop around tempfile.mktemp and
10018
    # socket.bind
10019
    tmpdir = tempfile.mkdtemp()
10020
    try:
10021
      tmpsock = utils.PathJoin(tmpdir, "sock")
10022

    
10023
      logging.debug("Creating temporary socket at %s", tmpsock)
10024
      sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
10025
      try:
10026
        sock.bind(tmpsock)
10027
        sock.listen(1)
10028

    
10029
        # Send details to client
10030
        cb(tmpsock)
10031

    
10032
        # Wait for client to connect before continuing
10033
        sock.settimeout(cls._CLIENT_CONNECT_TIMEOUT)
10034
        try:
10035
          (conn, _) = sock.accept()
10036
        except socket.error, err:
10037
          raise errcls("Client didn't connect in time (%s)" % err)
10038
      finally:
10039
        sock.close()
10040
    finally:
10041
      # Remove as soon as client is connected
10042
      shutil.rmtree(tmpdir)
10043

    
10044
    # Wait for client to close
10045
    try:
10046
      try:
10047
        # pylint: disable-msg=E1101
10048
        # Instance of '_socketobject' has no ... member
10049
        conn.settimeout(cls._CLIENT_CONFIRM_TIMEOUT)
10050
        conn.recv(1)
10051
      except socket.error, err:
10052
        raise errcls("Client failed to confirm notification (%s)" % err)
10053
    finally:
10054
      conn.close()
10055

    
10056
  def _SendNotification(self, test, arg, sockname):
10057
    """Sends a notification to the client.
10058

10059
    @type test: string
10060
    @param test: Test name
10061
    @param arg: Test argument (depends on test)
10062
    @type sockname: string
10063
    @param sockname: Socket path
10064

10065
    """
10066
    self.Log(constants.ELOG_JQUEUE_TEST, (sockname, test, arg))
10067

    
10068
  def _Notify(self, prereq, test, arg):
10069
    """Notifies the client of a test.
10070

10071
    @type prereq: bool
10072
    @param prereq: Whether this is a prereq-phase test
10073
    @type test: string
10074
    @param test: Test name
10075
    @param arg: Test argument (depends on test)
10076

10077
    """
10078
    if prereq:
10079
      errcls = errors.OpPrereqError
10080
    else:
10081
      errcls = errors.OpExecError
10082

    
10083
    return self._NotifyUsingSocket(compat.partial(self._SendNotification,
10084
                                                  test, arg),
10085
                                   errcls)
10086

    
10087
  def CheckArguments(self):
10088
    self.checkargs_calls = getattr(self, "checkargs_calls", 0) + 1
10089
    self.expandnames_calls = 0
10090

    
10091
  def ExpandNames(self):
10092
    checkargs_calls = getattr(self, "checkargs_calls", 0)
10093
    if checkargs_calls < 1:
10094
      raise errors.ProgrammerError("CheckArguments was not called")
10095

    
10096
    self.expandnames_calls += 1
10097

    
10098
    if self.op.notify_waitlock:
10099
      self._Notify(True, constants.JQT_EXPANDNAMES, None)
10100

    
10101
    self.LogInfo("Expanding names")
10102

    
10103
    # Get lock on master node (just to get a lock, not for a particular reason)
10104
    self.needed_locks = {
10105
      locking.LEVEL_NODE: self.cfg.GetMasterNode(),
10106
      }
10107

    
10108
  def Exec(self, feedback_fn):
10109
    if self.expandnames_calls < 1:
10110
      raise errors.ProgrammerError("ExpandNames was not called")
10111

    
10112
    if self.op.notify_exec:
10113
      self._Notify(False, constants.JQT_EXEC, None)
10114

    
10115
    self.LogInfo("Executing")
10116

    
10117
    if self.op.log_messages:
10118
      self._Notify(False, constants.JQT_STARTMSG, len(self.op.log_messages))
10119
      for idx, msg in enumerate(self.op.log_messages):
10120
        self.LogInfo("Sending log message %s", idx + 1)
10121
        feedback_fn(constants.JQT_MSGPREFIX + msg)
10122
        # Report how many test messages have been sent
10123
        self._Notify(False, constants.JQT_LOGMSG, idx + 1)
10124

    
10125
    if self.op.fail:
10126
      raise errors.OpExecError("Opcode failure was requested")
10127

    
10128
    return True
10129

    
10130

    
10131
class IAllocator(object):
10132
  """IAllocator framework.
10133

10134
  An IAllocator instance has three sets of attributes:
10135
    - cfg that is needed to query the cluster
10136
    - input data (all members of the _KEYS class attribute are required)
10137
    - four buffer attributes (in|out_data|text), that represent the
10138
      input (to the external script) in text and data structure format,
10139
      and the output from it, again in two formats
10140
    - the result variables from the script (success, info, nodes) for
10141
      easy usage
10142

10143
  """
10144
  # pylint: disable-msg=R0902
10145
  # lots of instance attributes
10146
  _ALLO_KEYS = [
10147
    "name", "mem_size", "disks", "disk_template",
10148
    "os", "tags", "nics", "vcpus", "hypervisor",
10149
    ]
10150
  _RELO_KEYS = [
10151
    "name", "relocate_from",
10152
    ]
10153
  _EVAC_KEYS = [
10154
    "evac_nodes",
10155
    ]
10156

    
10157
  def __init__(self, cfg, rpc, mode, **kwargs):
10158
    self.cfg = cfg
10159
    self.rpc = rpc
10160
    # init buffer variables
10161
    self.in_text = self.out_text = self.in_data = self.out_data = None
10162
    # init all input fields so that pylint is happy
10163
    self.mode = mode
10164
    self.mem_size = self.disks = self.disk_template = None
10165
    self.os = self.tags = self.nics = self.vcpus = None
10166
    self.hypervisor = None
10167
    self.relocate_from = None
10168
    self.name = None
10169
    self.evac_nodes = None
10170
    # computed fields
10171
    self.required_nodes = None
10172
    # init result fields
10173
    self.success = self.info = self.result = None
10174
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
10175
      keyset = self._ALLO_KEYS
10176
      fn = self._AddNewInstance
10177
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
10178
      keyset = self._RELO_KEYS
10179
      fn = self._AddRelocateInstance
10180
    elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
10181
      keyset = self._EVAC_KEYS
10182
      fn = self._AddEvacuateNodes
10183
    else:
10184
      raise errors.ProgrammerError("Unknown mode '%s' passed to the"
10185
                                   " IAllocator" % self.mode)
10186
    for key in kwargs:
10187
      if key not in keyset:
10188
        raise errors.ProgrammerError("Invalid input parameter '%s' to"
10189
                                     " IAllocator" % key)
10190
      setattr(self, key, kwargs[key])
10191

    
10192
    for key in keyset:
10193
      if key not in kwargs:
10194
        raise errors.ProgrammerError("Missing input parameter '%s' to"
10195
                                     " IAllocator" % key)
10196
    self._BuildInputData(fn)
10197

    
10198
  def _ComputeClusterData(self):
10199
    """Compute the generic allocator input data.
10200

10201
    This is the data that is independent of the actual operation.
10202

10203
    """
10204
    cfg = self.cfg
10205
    cluster_info = cfg.GetClusterInfo()
10206
    # cluster data
10207
    data = {
10208
      "version": constants.IALLOCATOR_VERSION,
10209
      "cluster_name": cfg.GetClusterName(),
10210
      "cluster_tags": list(cluster_info.GetTags()),
10211
      "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
10212
      # we don't have job IDs
10213
      }
10214
    iinfo = cfg.GetAllInstancesInfo().values()
10215
    i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
10216

    
10217
    # node data
10218
    node_results = {}
10219
    node_list = cfg.GetNodeList()
10220

    
10221
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
10222
      hypervisor_name = self.hypervisor
10223
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
10224
      hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
10225
    elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
10226
      hypervisor_name = cluster_info.enabled_hypervisors[0]
10227

    
10228
    node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(),
10229
                                        hypervisor_name)
10230
    node_iinfo = \
10231
      self.rpc.call_all_instances_info(node_list,
10232
                                       cluster_info.enabled_hypervisors)
10233
    for nname, nresult in node_data.items():
10234
      # first fill in static (config-based) values
10235
      ninfo = cfg.GetNodeInfo(nname)
10236
      pnr = {
10237
        "tags": list(ninfo.GetTags()),
10238
        "primary_ip": ninfo.primary_ip,
10239
        "secondary_ip": ninfo.secondary_ip,
10240
        "offline": ninfo.offline,
10241
        "drained": ninfo.drained,
10242
        "master_candidate": ninfo.master_candidate,
10243
        }
10244

    
10245
      if not (ninfo.offline or ninfo.drained):
10246
        nresult.Raise("Can't get data for node %s" % nname)
10247
        node_iinfo[nname].Raise("Can't get node instance info from node %s" %
10248
                                nname)
10249
        remote_info = nresult.payload
10250

    
10251
        for attr in ['memory_total', 'memory_free', 'memory_dom0',
10252
                     'vg_size', 'vg_free', 'cpu_total']:
10253
          if attr not in remote_info:
10254
            raise errors.OpExecError("Node '%s' didn't return attribute"
10255
                                     " '%s'" % (nname, attr))
10256
          if not isinstance(remote_info[attr], int):
10257
            raise errors.OpExecError("Node '%s' returned invalid value"
10258
                                     " for '%s': %s" %
10259
                                     (nname, attr, remote_info[attr]))
10260
        # compute memory used by primary instances
10261
        i_p_mem = i_p_up_mem = 0
10262
        for iinfo, beinfo in i_list:
10263
          if iinfo.primary_node == nname:
10264
            i_p_mem += beinfo[constants.BE_MEMORY]
10265
            if iinfo.name not in node_iinfo[nname].payload:
10266
              i_used_mem = 0
10267
            else:
10268
              i_used_mem = int(node_iinfo[nname].payload[iinfo.name]['memory'])
10269
            i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
10270
            remote_info['memory_free'] -= max(0, i_mem_diff)
10271

    
10272
            if iinfo.admin_up:
10273
              i_p_up_mem += beinfo[constants.BE_MEMORY]
10274

    
10275
        # compute memory used by instances
10276
        pnr_dyn = {
10277
          "total_memory": remote_info['memory_total'],
10278
          "reserved_memory": remote_info['memory_dom0'],
10279
          "free_memory": remote_info['memory_free'],
10280
          "total_disk": remote_info['vg_size'],
10281
          "free_disk": remote_info['vg_free'],
10282
          "total_cpus": remote_info['cpu_total'],
10283
          "i_pri_memory": i_p_mem,
10284
          "i_pri_up_memory": i_p_up_mem,
10285
          }
10286
        pnr.update(pnr_dyn)
10287

    
10288
      node_results[nname] = pnr
10289
    data["nodes"] = node_results
10290

    
10291
    # instance data
10292
    instance_data = {}
10293
    for iinfo, beinfo in i_list:
10294
      nic_data = []
10295
      for nic in iinfo.nics:
10296
        filled_params = cluster_info.SimpleFillNIC(nic.nicparams)
10297
        nic_dict = {"mac": nic.mac,
10298
                    "ip": nic.ip,
10299
                    "mode": filled_params[constants.NIC_MODE],
10300
                    "link": filled_params[constants.NIC_LINK],
10301
                   }
10302
        if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
10303
          nic_dict["bridge"] = filled_params[constants.NIC_LINK]
10304
        nic_data.append(nic_dict)
10305
      pir = {
10306
        "tags": list(iinfo.GetTags()),
10307
        "admin_up": iinfo.admin_up,
10308
        "vcpus": beinfo[constants.BE_VCPUS],
10309
        "memory": beinfo[constants.BE_MEMORY],
10310
        "os": iinfo.os,
10311
        "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
10312
        "nics": nic_data,
10313
        "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
10314
        "disk_template": iinfo.disk_template,
10315
        "hypervisor": iinfo.hypervisor,
10316
        }
10317
      pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
10318
                                                 pir["disks"])
10319
      instance_data[iinfo.name] = pir
10320

    
10321
    data["instances"] = instance_data
10322

    
10323
    self.in_data = data
10324

    
10325
  def _AddNewInstance(self):
10326
    """Add new instance data to allocator structure.
10327

10328
    This in combination with _AllocatorGetClusterData will create the
10329
    correct structure needed as input for the allocator.
10330

10331
    The checks for the completeness of the opcode must have already been
10332
    done.
10333

10334
    """
10335
    disk_space = _ComputeDiskSize(self.disk_template, self.disks)
10336

    
10337
    if self.disk_template in constants.DTS_NET_MIRROR:
10338
      self.required_nodes = 2
10339
    else:
10340
      self.required_nodes = 1
10341
    request = {
10342
      "name": self.name,
10343
      "disk_template": self.disk_template,
10344
      "tags": self.tags,
10345
      "os": self.os,
10346
      "vcpus": self.vcpus,
10347
      "memory": self.mem_size,
10348
      "disks": self.disks,
10349
      "disk_space_total": disk_space,
10350
      "nics": self.nics,
10351
      "required_nodes": self.required_nodes,
10352
      }
10353
    return request
10354

    
10355
  def _AddRelocateInstance(self):
10356
    """Add relocate instance data to allocator structure.
10357

10358
    This in combination with _IAllocatorGetClusterData will create the
10359
    correct structure needed as input for the allocator.
10360

10361
    The checks for the completeness of the opcode must have already been
10362
    done.
10363

10364
    """
10365
    instance = self.cfg.GetInstanceInfo(self.name)
10366
    if instance is None:
10367
      raise errors.ProgrammerError("Unknown instance '%s' passed to"
10368
                                   " IAllocator" % self.name)
10369

    
10370
    if instance.disk_template not in constants.DTS_NET_MIRROR:
10371
      raise errors.OpPrereqError("Can't relocate non-mirrored instances",
10372
                                 errors.ECODE_INVAL)
10373

    
10374
    if len(instance.secondary_nodes) != 1:
10375
      raise errors.OpPrereqError("Instance has not exactly one secondary node",
10376
                                 errors.ECODE_STATE)
10377

    
10378
    self.required_nodes = 1
10379
    disk_sizes = [{'size': disk.size} for disk in instance.disks]
10380
    disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
10381

    
10382
    request = {
10383
      "name": self.name,
10384
      "disk_space_total": disk_space,
10385
      "required_nodes": self.required_nodes,
10386
      "relocate_from": self.relocate_from,
10387
      }
10388
    return request
10389

    
10390
  def _AddEvacuateNodes(self):
10391
    """Add evacuate nodes data to allocator structure.
10392

10393
    """
10394
    request = {
10395
      "evac_nodes": self.evac_nodes
10396
      }
10397
    return request
10398

    
10399
  def _BuildInputData(self, fn):
10400
    """Build input data structures.
10401

10402
    """
10403
    self._ComputeClusterData()
10404

    
10405
    request = fn()
10406
    request["type"] = self.mode
10407
    self.in_data["request"] = request
10408

    
10409
    self.in_text = serializer.Dump(self.in_data)
10410

    
10411
  def Run(self, name, validate=True, call_fn=None):
10412
    """Run an instance allocator and return the results.
10413

10414
    """
10415
    if call_fn is None:
10416
      call_fn = self.rpc.call_iallocator_runner
10417

    
10418
    result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
10419
    result.Raise("Failure while running the iallocator script")
10420

    
10421
    self.out_text = result.payload
10422
    if validate:
10423
      self._ValidateResult()
10424

    
10425
  def _ValidateResult(self):
10426
    """Process the allocator results.
10427

10428
    This will process and if successful save the result in
10429
    self.out_data and the other parameters.
10430

10431
    """
10432
    try:
10433
      rdict = serializer.Load(self.out_text)
10434
    except Exception, err:
10435
      raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
10436

    
10437
    if not isinstance(rdict, dict):
10438
      raise errors.OpExecError("Can't parse iallocator results: not a dict")
10439

    
10440
    # TODO: remove backwards compatiblity in later versions
10441
    if "nodes" in rdict and "result" not in rdict:
10442
      rdict["result"] = rdict["nodes"]
10443
      del rdict["nodes"]
10444

    
10445
    for key in "success", "info", "result":
10446
      if key not in rdict:
10447
        raise errors.OpExecError("Can't parse iallocator results:"
10448
                                 " missing key '%s'" % key)
10449
      setattr(self, key, rdict[key])
10450

    
10451
    if not isinstance(rdict["result"], list):
10452
      raise errors.OpExecError("Can't parse iallocator results: 'result' key"
10453
                               " is not a list")
10454
    self.out_data = rdict
10455

    
10456

    
10457
class LUTestAllocator(NoHooksLU):
10458
  """Run allocator tests.
10459

10460
  This LU runs the allocator tests
10461

10462
  """
10463
  _OP_PARAMS = [
10464
    ("direction", _NoDefault, _TElemOf(constants.VALID_IALLOCATOR_DIRECTIONS)),
10465
    ("mode", _NoDefault, _TElemOf(constants.VALID_IALLOCATOR_MODES)),
10466
    ("name", _NoDefault, _TNonEmptyString),
10467
    ("nics", _NoDefault, _TOr(_TNone, _TListOf(
10468
      _TDictOf(_TElemOf(["mac", "ip", "bridge"]),
10469
               _TOr(_TNone, _TNonEmptyString))))),
10470
    ("disks", _NoDefault, _TOr(_TNone, _TList)),
10471
    ("hypervisor", None, _TMaybeString),
10472
    ("allocator", None, _TMaybeString),
10473
    ("tags", _EmptyList, _TListOf(_TNonEmptyString)),
10474
    ("mem_size", None, _TOr(_TNone, _TPositiveInt)),
10475
    ("vcpus", None, _TOr(_TNone, _TPositiveInt)),
10476
    ("os", None, _TMaybeString),
10477
    ("disk_template", None, _TMaybeString),
10478
    ("evac_nodes", None, _TOr(_TNone, _TListOf(_TNonEmptyString))),
10479
    ]
10480

    
10481
  def CheckPrereq(self):
10482
    """Check prerequisites.
10483

10484
    This checks the opcode parameters depending on the director and mode test.
10485

10486
    """
10487
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
10488
      for attr in ["mem_size", "disks", "disk_template",
10489
                   "os", "tags", "nics", "vcpus"]:
10490
        if not hasattr(self.op, attr):
10491
          raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
10492
                                     attr, errors.ECODE_INVAL)
10493
      iname = self.cfg.ExpandInstanceName(self.op.name)
10494
      if iname is not None:
10495
        raise errors.OpPrereqError("Instance '%s' already in the cluster" %
10496
                                   iname, errors.ECODE_EXISTS)
10497
      if not isinstance(self.op.nics, list):
10498
        raise errors.OpPrereqError("Invalid parameter 'nics'",
10499
                                   errors.ECODE_INVAL)
10500
      if not isinstance(self.op.disks, list):
10501
        raise errors.OpPrereqError("Invalid parameter 'disks'",
10502
                                   errors.ECODE_INVAL)
10503
      for row in self.op.disks:
10504
        if (not isinstance(row, dict) or
10505
            "size" not in row or
10506
            not isinstance(row["size"], int) or
10507
            "mode" not in row or
10508
            row["mode"] not in ['r', 'w']):
10509
          raise errors.OpPrereqError("Invalid contents of the 'disks'"
10510
                                     " parameter", errors.ECODE_INVAL)
10511
      if self.op.hypervisor is None:
10512
        self.op.hypervisor = self.cfg.GetHypervisorType()
10513
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
10514
      fname = _ExpandInstanceName(self.cfg, self.op.name)
10515
      self.op.name = fname
10516
      self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
10517
    elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
10518
      if not hasattr(self.op, "evac_nodes"):
10519
        raise errors.OpPrereqError("Missing attribute 'evac_nodes' on"
10520
                                   " opcode input", errors.ECODE_INVAL)
10521
    else:
10522
      raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
10523
                                 self.op.mode, errors.ECODE_INVAL)
10524

    
10525
    if self.op.direction == constants.IALLOCATOR_DIR_OUT:
10526
      if self.op.allocator is None:
10527
        raise errors.OpPrereqError("Missing allocator name",
10528
                                   errors.ECODE_INVAL)
10529
    elif self.op.direction != constants.IALLOCATOR_DIR_IN:
10530
      raise errors.OpPrereqError("Wrong allocator test '%s'" %
10531
                                 self.op.direction, errors.ECODE_INVAL)
10532

    
10533
  def Exec(self, feedback_fn):
10534
    """Run the allocator test.
10535

10536
    """
10537
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
10538
      ial = IAllocator(self.cfg, self.rpc,
10539
                       mode=self.op.mode,
10540
                       name=self.op.name,
10541
                       mem_size=self.op.mem_size,
10542
                       disks=self.op.disks,
10543
                       disk_template=self.op.disk_template,
10544
                       os=self.op.os,
10545
                       tags=self.op.tags,
10546
                       nics=self.op.nics,
10547
                       vcpus=self.op.vcpus,
10548
                       hypervisor=self.op.hypervisor,
10549
                       )
10550
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
10551
      ial = IAllocator(self.cfg, self.rpc,
10552
                       mode=self.op.mode,
10553
                       name=self.op.name,
10554
                       relocate_from=list(self.relocate_from),
10555
                       )
10556
    elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
10557
      ial = IAllocator(self.cfg, self.rpc,
10558
                       mode=self.op.mode,
10559
                       evac_nodes=self.op.evac_nodes)
10560
    else:
10561
      raise errors.ProgrammerError("Uncatched mode %s in"
10562
                                   " LUTestAllocator.Exec", self.op.mode)
10563

    
10564
    if self.op.direction == constants.IALLOCATOR_DIR_IN:
10565
      result = ial.in_text
10566
    else:
10567
      ial.Run(self.op.allocator, validate=False)
10568
      result = ial.out_text
10569
    return result