Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ ff89a747

History | View | Annotate | Download (347.7 kB)

1
#
2
#
3

    
4
# Copyright (C) 2006, 2007, 2008 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
25

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

    
29
import os
30
import os.path
31
import time
32
import re
33
import platform
34
import logging
35
import copy
36
import OpenSSL
37

    
38
from ganeti import ssh
39
from ganeti import utils
40
from ganeti import errors
41
from ganeti import hypervisor
42
from ganeti import locking
43
from ganeti import constants
44
from ganeti import objects
45
from ganeti import serializer
46
from ganeti import ssconf
47
from ganeti import uidpool
48
from ganeti import compat
49
from ganeti import masterd
50

    
51
import ganeti.masterd.instance # pylint: disable-msg=W0611
52

    
53

    
54
class LogicalUnit(object):
55
  """Logical Unit base class.
56

57
  Subclasses must follow these rules:
58
    - implement ExpandNames
59
    - implement CheckPrereq (except when tasklets are used)
60
    - implement Exec (except when tasklets are used)
61
    - implement BuildHooksEnv
62
    - redefine HPATH and HTYPE
63
    - optionally redefine their run requirements:
64
        REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
65

66
  Note that all commands require root permissions.
67

68
  @ivar dry_run_result: the value (if any) that will be returned to the caller
69
      in dry-run mode (signalled by opcode dry_run parameter)
70

71
  """
72
  HPATH = None
73
  HTYPE = None
74
  _OP_REQP = []
75
  REQ_BGL = True
76

    
77
  def __init__(self, processor, op, context, rpc):
78
    """Constructor for LogicalUnit.
79

80
    This needs to be overridden in derived classes in order to check op
81
    validity.
82

83
    """
84
    self.proc = processor
85
    self.op = op
86
    self.cfg = context.cfg
87
    self.context = context
88
    self.rpc = rpc
89
    # Dicts used to declare locking needs to mcpu
90
    self.needed_locks = None
91
    self.acquired_locks = {}
92
    self.share_locks = dict.fromkeys(locking.LEVELS, 0)
93
    self.add_locks = {}
94
    self.remove_locks = {}
95
    # Used to force good behavior when calling helper functions
96
    self.recalculate_locks = {}
97
    self.__ssh = None
98
    # logging
99
    self.LogWarning = processor.LogWarning # pylint: disable-msg=C0103
100
    self.LogInfo = processor.LogInfo # pylint: disable-msg=C0103
101
    self.LogStep = processor.LogStep # pylint: disable-msg=C0103
102
    # support for dry-run
103
    self.dry_run_result = None
104
    # support for generic debug attribute
105
    if (not hasattr(self.op, "debug_level") or
106
        not isinstance(self.op.debug_level, int)):
107
      self.op.debug_level = 0
108

    
109
    # Tasklets
110
    self.tasklets = None
111

    
112
    for attr_name in self._OP_REQP:
113
      attr_val = getattr(op, attr_name, None)
114
      if attr_val is None:
115
        raise errors.OpPrereqError("Required parameter '%s' missing" %
116
                                   attr_name, errors.ECODE_INVAL)
117

    
118
    self.CheckArguments()
119

    
120
  def __GetSSH(self):
121
    """Returns the SshRunner object
122

123
    """
124
    if not self.__ssh:
125
      self.__ssh = ssh.SshRunner(self.cfg.GetClusterName())
126
    return self.__ssh
127

    
128
  ssh = property(fget=__GetSSH)
129

    
130
  def CheckArguments(self):
131
    """Check syntactic validity for the opcode arguments.
132

133
    This method is for doing a simple syntactic check and ensure
134
    validity of opcode parameters, without any cluster-related
135
    checks. While the same can be accomplished in ExpandNames and/or
136
    CheckPrereq, doing these separate is better because:
137

138
      - ExpandNames is left as as purely a lock-related function
139
      - CheckPrereq is run after we have acquired locks (and possible
140
        waited for them)
141

142
    The function is allowed to change the self.op attribute so that
143
    later methods can no longer worry about missing parameters.
144

145
    """
146
    pass
147

    
148
  def ExpandNames(self):
149
    """Expand names for this LU.
150

151
    This method is called before starting to execute the opcode, and it should
152
    update all the parameters of the opcode to their canonical form (e.g. a
153
    short node name must be fully expanded after this method has successfully
154
    completed). This way locking, hooks, logging, ecc. can work correctly.
155

156
    LUs which implement this method must also populate the self.needed_locks
157
    member, as a dict with lock levels as keys, and a list of needed lock names
158
    as values. Rules:
159

160
      - use an empty dict if you don't need any lock
161
      - if you don't need any lock at a particular level omit that level
162
      - don't put anything for the BGL level
163
      - if you want all locks at a level use locking.ALL_SET as a value
164

165
    If you need to share locks (rather than acquire them exclusively) at one
166
    level you can modify self.share_locks, setting a true value (usually 1) for
167
    that level. By default locks are not shared.
168

169
    This function can also define a list of tasklets, which then will be
170
    executed in order instead of the usual LU-level CheckPrereq and Exec
171
    functions, if those are not defined by the LU.
172

173
    Examples::
174

175
      # Acquire all nodes and one instance
176
      self.needed_locks = {
177
        locking.LEVEL_NODE: locking.ALL_SET,
178
        locking.LEVEL_INSTANCE: ['instance1.example.tld'],
179
      }
180
      # Acquire just two nodes
181
      self.needed_locks = {
182
        locking.LEVEL_NODE: ['node1.example.tld', 'node2.example.tld'],
183
      }
184
      # Acquire no locks
185
      self.needed_locks = {} # No, you can't leave it to the default value None
186

187
    """
188
    # The implementation of this method is mandatory only if the new LU is
189
    # concurrent, so that old LUs don't need to be changed all at the same
190
    # time.
191
    if self.REQ_BGL:
192
      self.needed_locks = {} # Exclusive LUs don't need locks.
193
    else:
194
      raise NotImplementedError
195

    
196
  def DeclareLocks(self, level):
197
    """Declare LU locking needs for a level
198

199
    While most LUs can just declare their locking needs at ExpandNames time,
200
    sometimes there's the need to calculate some locks after having acquired
201
    the ones before. This function is called just before acquiring locks at a
202
    particular level, but after acquiring the ones at lower levels, and permits
203
    such calculations. It can be used to modify self.needed_locks, and by
204
    default it does nothing.
205

206
    This function is only called if you have something already set in
207
    self.needed_locks for the level.
208

209
    @param level: Locking level which is going to be locked
210
    @type level: member of ganeti.locking.LEVELS
211

212
    """
213

    
214
  def CheckPrereq(self):
215
    """Check prerequisites for this LU.
216

217
    This method should check that the prerequisites for the execution
218
    of this LU are fulfilled. It can do internode communication, but
219
    it should be idempotent - no cluster or system changes are
220
    allowed.
221

222
    The method should raise errors.OpPrereqError in case something is
223
    not fulfilled. Its return value is ignored.
224

225
    This method should also update all the parameters of the opcode to
226
    their canonical form if it hasn't been done by ExpandNames before.
227

228
    """
229
    if self.tasklets is not None:
230
      for (idx, tl) in enumerate(self.tasklets):
231
        logging.debug("Checking prerequisites for tasklet %s/%s",
232
                      idx + 1, len(self.tasklets))
233
        tl.CheckPrereq()
234
    else:
235
      raise NotImplementedError
236

    
237
  def Exec(self, feedback_fn):
238
    """Execute the LU.
239

240
    This method should implement the actual work. It should raise
241
    errors.OpExecError for failures that are somewhat dealt with in
242
    code, or expected.
243

244
    """
245
    if self.tasklets is not None:
246
      for (idx, tl) in enumerate(self.tasklets):
247
        logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets))
248
        tl.Exec(feedback_fn)
249
    else:
250
      raise NotImplementedError
251

    
252
  def BuildHooksEnv(self):
253
    """Build hooks environment for this LU.
254

255
    This method should return a three-node tuple consisting of: a dict
256
    containing the environment that will be used for running the
257
    specific hook for this LU, a list of node names on which the hook
258
    should run before the execution, and a list of node names on which
259
    the hook should run after the execution.
260

261
    The keys of the dict must not have 'GANETI_' prefixed as this will
262
    be handled in the hooks runner. Also note additional keys will be
263
    added by the hooks runner. If the LU doesn't define any
264
    environment, an empty dict (and not None) should be returned.
265

266
    No nodes should be returned as an empty list (and not None).
267

268
    Note that if the HPATH for a LU class is None, this function will
269
    not be called.
270

271
    """
272
    raise NotImplementedError
273

    
274
  def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
275
    """Notify the LU about the results of its hooks.
276

277
    This method is called every time a hooks phase is executed, and notifies
278
    the Logical Unit about the hooks' result. The LU can then use it to alter
279
    its result based on the hooks.  By default the method does nothing and the
280
    previous result is passed back unchanged but any LU can define it if it
281
    wants to use the local cluster hook-scripts somehow.
282

283
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
284
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
285
    @param hook_results: the results of the multi-node hooks rpc call
286
    @param feedback_fn: function used send feedback back to the caller
287
    @param lu_result: the previous Exec result this LU had, or None
288
        in the PRE phase
289
    @return: the new Exec result, based on the previous result
290
        and hook results
291

292
    """
293
    # API must be kept, thus we ignore the unused argument and could
294
    # be a function warnings
295
    # pylint: disable-msg=W0613,R0201
296
    return lu_result
297

    
298
  def _ExpandAndLockInstance(self):
299
    """Helper function to expand and lock an instance.
300

301
    Many LUs that work on an instance take its name in self.op.instance_name
302
    and need to expand it and then declare the expanded name for locking. This
303
    function does it, and then updates self.op.instance_name to the expanded
304
    name. It also initializes needed_locks as a dict, if this hasn't been done
305
    before.
306

307
    """
308
    if self.needed_locks is None:
309
      self.needed_locks = {}
310
    else:
311
      assert locking.LEVEL_INSTANCE not in self.needed_locks, \
312
        "_ExpandAndLockInstance called with instance-level locks set"
313
    self.op.instance_name = _ExpandInstanceName(self.cfg,
314
                                                self.op.instance_name)
315
    self.needed_locks[locking.LEVEL_INSTANCE] = self.op.instance_name
316

    
317
  def _LockInstancesNodes(self, primary_only=False):
318
    """Helper function to declare instances' nodes for locking.
319

320
    This function should be called after locking one or more instances to lock
321
    their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
322
    with all primary or secondary nodes for instances already locked and
323
    present in self.needed_locks[locking.LEVEL_INSTANCE].
324

325
    It should be called from DeclareLocks, and for safety only works if
326
    self.recalculate_locks[locking.LEVEL_NODE] is set.
327

328
    In the future it may grow parameters to just lock some instance's nodes, or
329
    to just lock primaries or secondary nodes, if needed.
330

331
    If should be called in DeclareLocks in a way similar to::
332

333
      if level == locking.LEVEL_NODE:
334
        self._LockInstancesNodes()
335

336
    @type primary_only: boolean
337
    @param primary_only: only lock primary nodes of locked instances
338

339
    """
340
    assert locking.LEVEL_NODE in self.recalculate_locks, \
341
      "_LockInstancesNodes helper function called with no nodes to recalculate"
342

    
343
    # TODO: check if we're really been called with the instance locks held
344

    
345
    # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
346
    # future we might want to have different behaviors depending on the value
347
    # of self.recalculate_locks[locking.LEVEL_NODE]
348
    wanted_nodes = []
349
    for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
350
      instance = self.context.cfg.GetInstanceInfo(instance_name)
351
      wanted_nodes.append(instance.primary_node)
352
      if not primary_only:
353
        wanted_nodes.extend(instance.secondary_nodes)
354

    
355
    if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE:
356
      self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
357
    elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND:
358
      self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
359

    
360
    del self.recalculate_locks[locking.LEVEL_NODE]
361

    
362

    
363
class NoHooksLU(LogicalUnit): # pylint: disable-msg=W0223
364
  """Simple LU which runs no hooks.
365

366
  This LU is intended as a parent for other LogicalUnits which will
367
  run no hooks, in order to reduce duplicate code.
368

369
  """
370
  HPATH = None
371
  HTYPE = None
372

    
373
  def BuildHooksEnv(self):
374
    """Empty BuildHooksEnv for NoHooksLu.
375

376
    This just raises an error.
377

378
    """
379
    assert False, "BuildHooksEnv called for NoHooksLUs"
380

    
381

    
382
class Tasklet:
383
  """Tasklet base class.
384

385
  Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
386
  they can mix legacy code with tasklets. Locking needs to be done in the LU,
387
  tasklets know nothing about locks.
388

389
  Subclasses must follow these rules:
390
    - Implement CheckPrereq
391
    - Implement Exec
392

393
  """
394
  def __init__(self, lu):
395
    self.lu = lu
396

    
397
    # Shortcuts
398
    self.cfg = lu.cfg
399
    self.rpc = lu.rpc
400

    
401
  def CheckPrereq(self):
402
    """Check prerequisites for this tasklets.
403

404
    This method should check whether the prerequisites for the execution of
405
    this tasklet are fulfilled. It can do internode communication, but it
406
    should be idempotent - no cluster or system changes are allowed.
407

408
    The method should raise errors.OpPrereqError in case something is not
409
    fulfilled. Its return value is ignored.
410

411
    This method should also update all parameters to their canonical form if it
412
    hasn't been done before.
413

414
    """
415
    raise NotImplementedError
416

    
417
  def Exec(self, feedback_fn):
418
    """Execute the tasklet.
419

420
    This method should implement the actual work. It should raise
421
    errors.OpExecError for failures that are somewhat dealt with in code, or
422
    expected.
423

424
    """
425
    raise NotImplementedError
426

    
427

    
428
def _GetWantedNodes(lu, nodes):
429
  """Returns list of checked and expanded node names.
430

431
  @type lu: L{LogicalUnit}
432
  @param lu: the logical unit on whose behalf we execute
433
  @type nodes: list
434
  @param nodes: list of node names or None for all nodes
435
  @rtype: list
436
  @return: the list of nodes, sorted
437
  @raise errors.ProgrammerError: if the nodes parameter is wrong type
438

439
  """
440
  if not isinstance(nodes, list):
441
    raise errors.OpPrereqError("Invalid argument type 'nodes'",
442
                               errors.ECODE_INVAL)
443

    
444
  if not nodes:
445
    raise errors.ProgrammerError("_GetWantedNodes should only be called with a"
446
      " non-empty list of nodes whose name is to be expanded.")
447

    
448
  wanted = [_ExpandNodeName(lu.cfg, name) for name in nodes]
449
  return utils.NiceSort(wanted)
450

    
451

    
452
def _GetWantedInstances(lu, instances):
453
  """Returns list of checked and expanded instance names.
454

455
  @type lu: L{LogicalUnit}
456
  @param lu: the logical unit on whose behalf we execute
457
  @type instances: list
458
  @param instances: list of instance names or None for all instances
459
  @rtype: list
460
  @return: the list of instances, sorted
461
  @raise errors.OpPrereqError: if the instances parameter is wrong type
462
  @raise errors.OpPrereqError: if any of the passed instances is not found
463

464
  """
465
  if not isinstance(instances, list):
466
    raise errors.OpPrereqError("Invalid argument type 'instances'",
467
                               errors.ECODE_INVAL)
468

    
469
  if instances:
470
    wanted = [_ExpandInstanceName(lu.cfg, name) for name in instances]
471
  else:
472
    wanted = utils.NiceSort(lu.cfg.GetInstanceList())
473
  return wanted
474

    
475

    
476
def _CheckOutputFields(static, dynamic, selected):
477
  """Checks whether all selected fields are valid.
478

479
  @type static: L{utils.FieldSet}
480
  @param static: static fields set
481
  @type dynamic: L{utils.FieldSet}
482
  @param dynamic: dynamic fields set
483

484
  """
485
  f = utils.FieldSet()
486
  f.Extend(static)
487
  f.Extend(dynamic)
488

    
489
  delta = f.NonMatching(selected)
490
  if delta:
491
    raise errors.OpPrereqError("Unknown output fields selected: %s"
492
                               % ",".join(delta), errors.ECODE_INVAL)
493

    
494

    
495
def _CheckBooleanOpField(op, name):
496
  """Validates boolean opcode parameters.
497

498
  This will ensure that an opcode parameter is either a boolean value,
499
  or None (but that it always exists).
500

501
  """
502
  val = getattr(op, name, None)
503
  if not (val is None or isinstance(val, bool)):
504
    raise errors.OpPrereqError("Invalid boolean parameter '%s' (%s)" %
505
                               (name, str(val)), errors.ECODE_INVAL)
506
  setattr(op, name, val)
507

    
508

    
509
def _CheckGlobalHvParams(params):
510
  """Validates that given hypervisor params are not global ones.
511

512
  This will ensure that instances don't get customised versions of
513
  global params.
514

515
  """
516
  used_globals = constants.HVC_GLOBALS.intersection(params)
517
  if used_globals:
518
    msg = ("The following hypervisor parameters are global and cannot"
519
           " be customized at instance level, please modify them at"
520
           " cluster level: %s" % utils.CommaJoin(used_globals))
521
    raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
522

    
523

    
524
def _CheckNodeOnline(lu, node):
525
  """Ensure that a given node is online.
526

527
  @param lu: the LU on behalf of which we make the check
528
  @param node: the node to check
529
  @raise errors.OpPrereqError: if the node is offline
530

531
  """
532
  if lu.cfg.GetNodeInfo(node).offline:
533
    raise errors.OpPrereqError("Can't use offline node %s" % node,
534
                               errors.ECODE_INVAL)
535

    
536

    
537
def _CheckNodeNotDrained(lu, node):
538
  """Ensure that a given node is not drained.
539

540
  @param lu: the LU on behalf of which we make the check
541
  @param node: the node to check
542
  @raise errors.OpPrereqError: if the node is drained
543

544
  """
545
  if lu.cfg.GetNodeInfo(node).drained:
546
    raise errors.OpPrereqError("Can't use drained node %s" % node,
547
                               errors.ECODE_INVAL)
548

    
549

    
550
def _CheckNodeHasOS(lu, node, os_name, force_variant):
551
  """Ensure that a node supports a given OS.
552

553
  @param lu: the LU on behalf of which we make the check
554
  @param node: the node to check
555
  @param os_name: the OS to query about
556
  @param force_variant: whether to ignore variant errors
557
  @raise errors.OpPrereqError: if the node is not supporting the OS
558

559
  """
560
  result = lu.rpc.call_os_get(node, os_name)
561
  result.Raise("OS '%s' not in supported OS list for node %s" %
562
               (os_name, node),
563
               prereq=True, ecode=errors.ECODE_INVAL)
564
  if not force_variant:
565
    _CheckOSVariant(result.payload, os_name)
566

    
567

    
568
def _RequireFileStorage():
569
  """Checks that file storage is enabled.
570

571
  @raise errors.OpPrereqError: when file storage is disabled
572

573
  """
574
  if not constants.ENABLE_FILE_STORAGE:
575
    raise errors.OpPrereqError("File storage disabled at configure time",
576
                               errors.ECODE_INVAL)
577

    
578

    
579
def _CheckDiskTemplate(template):
580
  """Ensure a given disk template is valid.
581

582
  """
583
  if template not in constants.DISK_TEMPLATES:
584
    msg = ("Invalid disk template name '%s', valid templates are: %s" %
585
           (template, utils.CommaJoin(constants.DISK_TEMPLATES)))
586
    raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
587
  if template == constants.DT_FILE:
588
    _RequireFileStorage()
589

    
590

    
591
def _CheckStorageType(storage_type):
592
  """Ensure a given storage type is valid.
593

594
  """
595
  if storage_type not in constants.VALID_STORAGE_TYPES:
596
    raise errors.OpPrereqError("Unknown storage type: %s" % storage_type,
597
                               errors.ECODE_INVAL)
598
  if storage_type == constants.ST_FILE:
599
    _RequireFileStorage()
600

    
601

    
602
def _GetClusterDomainSecret():
603
  """Reads the cluster domain secret.
604

605
  """
606
  return utils.ReadOneLineFile(constants.CLUSTER_DOMAIN_SECRET_FILE,
607
                               strict=True)
608

    
609

    
610
def _CheckInstanceDown(lu, instance, reason):
611
  """Ensure that an instance is not running."""
612
  if instance.admin_up:
613
    raise errors.OpPrereqError("Instance %s is marked to be up, %s" %
614
                               (instance.name, reason), errors.ECODE_STATE)
615

    
616
  pnode = instance.primary_node
617
  ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
618
  ins_l.Raise("Can't contact node %s for instance information" % pnode,
619
              prereq=True, ecode=errors.ECODE_ENVIRON)
620

    
621
  if instance.name in ins_l.payload:
622
    raise errors.OpPrereqError("Instance %s is running, %s" %
623
                               (instance.name, reason), errors.ECODE_STATE)
624

    
625

    
626
def _ExpandItemName(fn, name, kind):
627
  """Expand an item name.
628

629
  @param fn: the function to use for expansion
630
  @param name: requested item name
631
  @param kind: text description ('Node' or 'Instance')
632
  @return: the resolved (full) name
633
  @raise errors.OpPrereqError: if the item is not found
634

635
  """
636
  full_name = fn(name)
637
  if full_name is None:
638
    raise errors.OpPrereqError("%s '%s' not known" % (kind, name),
639
                               errors.ECODE_NOENT)
640
  return full_name
641

    
642

    
643
def _ExpandNodeName(cfg, name):
644
  """Wrapper over L{_ExpandItemName} for nodes."""
645
  return _ExpandItemName(cfg.ExpandNodeName, name, "Node")
646

    
647

    
648
def _ExpandInstanceName(cfg, name):
649
  """Wrapper over L{_ExpandItemName} for instance."""
650
  return _ExpandItemName(cfg.ExpandInstanceName, name, "Instance")
651

    
652

    
653
def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
654
                          memory, vcpus, nics, disk_template, disks,
655
                          bep, hvp, hypervisor_name):
656
  """Builds instance related env variables for hooks
657

658
  This builds the hook environment from individual variables.
659

660
  @type name: string
661
  @param name: the name of the instance
662
  @type primary_node: string
663
  @param primary_node: the name of the instance's primary node
664
  @type secondary_nodes: list
665
  @param secondary_nodes: list of secondary nodes as strings
666
  @type os_type: string
667
  @param os_type: the name of the instance's OS
668
  @type status: boolean
669
  @param status: the should_run status of the instance
670
  @type memory: string
671
  @param memory: the memory size of the instance
672
  @type vcpus: string
673
  @param vcpus: the count of VCPUs the instance has
674
  @type nics: list
675
  @param nics: list of tuples (ip, mac, mode, link) representing
676
      the NICs the instance has
677
  @type disk_template: string
678
  @param disk_template: the disk template of the instance
679
  @type disks: list
680
  @param disks: the list of (size, mode) pairs
681
  @type bep: dict
682
  @param bep: the backend parameters for the instance
683
  @type hvp: dict
684
  @param hvp: the hypervisor parameters for the instance
685
  @type hypervisor_name: string
686
  @param hypervisor_name: the hypervisor for the instance
687
  @rtype: dict
688
  @return: the hook environment for this instance
689

690
  """
691
  if status:
692
    str_status = "up"
693
  else:
694
    str_status = "down"
695
  env = {
696
    "OP_TARGET": name,
697
    "INSTANCE_NAME": name,
698
    "INSTANCE_PRIMARY": primary_node,
699
    "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
700
    "INSTANCE_OS_TYPE": os_type,
701
    "INSTANCE_STATUS": str_status,
702
    "INSTANCE_MEMORY": memory,
703
    "INSTANCE_VCPUS": vcpus,
704
    "INSTANCE_DISK_TEMPLATE": disk_template,
705
    "INSTANCE_HYPERVISOR": hypervisor_name,
706
  }
707

    
708
  if nics:
709
    nic_count = len(nics)
710
    for idx, (ip, mac, mode, link) in enumerate(nics):
711
      if ip is None:
712
        ip = ""
713
      env["INSTANCE_NIC%d_IP" % idx] = ip
714
      env["INSTANCE_NIC%d_MAC" % idx] = mac
715
      env["INSTANCE_NIC%d_MODE" % idx] = mode
716
      env["INSTANCE_NIC%d_LINK" % idx] = link
717
      if mode == constants.NIC_MODE_BRIDGED:
718
        env["INSTANCE_NIC%d_BRIDGE" % idx] = link
719
  else:
720
    nic_count = 0
721

    
722
  env["INSTANCE_NIC_COUNT"] = nic_count
723

    
724
  if disks:
725
    disk_count = len(disks)
726
    for idx, (size, mode) in enumerate(disks):
727
      env["INSTANCE_DISK%d_SIZE" % idx] = size
728
      env["INSTANCE_DISK%d_MODE" % idx] = mode
729
  else:
730
    disk_count = 0
731

    
732
  env["INSTANCE_DISK_COUNT"] = disk_count
733

    
734
  for source, kind in [(bep, "BE"), (hvp, "HV")]:
735
    for key, value in source.items():
736
      env["INSTANCE_%s_%s" % (kind, key)] = value
737

    
738
  return env
739

    
740

    
741
def _NICListToTuple(lu, nics):
742
  """Build a list of nic information tuples.
743

744
  This list is suitable to be passed to _BuildInstanceHookEnv or as a return
745
  value in LUQueryInstanceData.
746

747
  @type lu:  L{LogicalUnit}
748
  @param lu: the logical unit on whose behalf we execute
749
  @type nics: list of L{objects.NIC}
750
  @param nics: list of nics to convert to hooks tuples
751

752
  """
753
  hooks_nics = []
754
  c_nicparams = lu.cfg.GetClusterInfo().nicparams[constants.PP_DEFAULT]
755
  for nic in nics:
756
    ip = nic.ip
757
    mac = nic.mac
758
    filled_params = objects.FillDict(c_nicparams, nic.nicparams)
759
    mode = filled_params[constants.NIC_MODE]
760
    link = filled_params[constants.NIC_LINK]
761
    hooks_nics.append((ip, mac, mode, link))
762
  return hooks_nics
763

    
764

    
765
def _BuildInstanceHookEnvByObject(lu, instance, override=None):
766
  """Builds instance related env variables for hooks from an object.
767

768
  @type lu: L{LogicalUnit}
769
  @param lu: the logical unit on whose behalf we execute
770
  @type instance: L{objects.Instance}
771
  @param instance: the instance for which we should build the
772
      environment
773
  @type override: dict
774
  @param override: dictionary with key/values that will override
775
      our values
776
  @rtype: dict
777
  @return: the hook environment dictionary
778

779
  """
780
  cluster = lu.cfg.GetClusterInfo()
781
  bep = cluster.FillBE(instance)
782
  hvp = cluster.FillHV(instance)
783
  args = {
784
    'name': instance.name,
785
    'primary_node': instance.primary_node,
786
    'secondary_nodes': instance.secondary_nodes,
787
    'os_type': instance.os,
788
    'status': instance.admin_up,
789
    'memory': bep[constants.BE_MEMORY],
790
    'vcpus': bep[constants.BE_VCPUS],
791
    'nics': _NICListToTuple(lu, instance.nics),
792
    'disk_template': instance.disk_template,
793
    'disks': [(disk.size, disk.mode) for disk in instance.disks],
794
    'bep': bep,
795
    'hvp': hvp,
796
    'hypervisor_name': instance.hypervisor,
797
  }
798
  if override:
799
    args.update(override)
800
  return _BuildInstanceHookEnv(**args) # pylint: disable-msg=W0142
801

    
802

    
803
def _AdjustCandidatePool(lu, exceptions):
804
  """Adjust the candidate pool after node operations.
805

806
  """
807
  mod_list = lu.cfg.MaintainCandidatePool(exceptions)
808
  if mod_list:
809
    lu.LogInfo("Promoted nodes to master candidate role: %s",
810
               utils.CommaJoin(node.name for node in mod_list))
811
    for name in mod_list:
812
      lu.context.ReaddNode(name)
813
  mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
814
  if mc_now > mc_max:
815
    lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
816
               (mc_now, mc_max))
817

    
818

    
819
def _DecideSelfPromotion(lu, exceptions=None):
820
  """Decide whether I should promote myself as a master candidate.
821

822
  """
823
  cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
824
  mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
825
  # the new node will increase mc_max with one, so:
826
  mc_should = min(mc_should + 1, cp_size)
827
  return mc_now < mc_should
828

    
829

    
830
def _CheckNicsBridgesExist(lu, target_nics, target_node,
831
                               profile=constants.PP_DEFAULT):
832
  """Check that the brigdes needed by a list of nics exist.
833

834
  """
835
  c_nicparams = lu.cfg.GetClusterInfo().nicparams[profile]
836
  paramslist = [objects.FillDict(c_nicparams, nic.nicparams)
837
                for nic in target_nics]
838
  brlist = [params[constants.NIC_LINK] for params in paramslist
839
            if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
840
  if brlist:
841
    result = lu.rpc.call_bridges_exist(target_node, brlist)
842
    result.Raise("Error checking bridges on destination node '%s'" %
843
                 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
844

    
845

    
846
def _CheckInstanceBridgesExist(lu, instance, node=None):
847
  """Check that the brigdes needed by an instance exist.
848

849
  """
850
  if node is None:
851
    node = instance.primary_node
852
  _CheckNicsBridgesExist(lu, instance.nics, node)
853

    
854

    
855
def _CheckOSVariant(os_obj, name):
856
  """Check whether an OS name conforms to the os variants specification.
857

858
  @type os_obj: L{objects.OS}
859
  @param os_obj: OS object to check
860
  @type name: string
861
  @param name: OS name passed by the user, to check for validity
862

863
  """
864
  if not os_obj.supported_variants:
865
    return
866
  try:
867
    variant = name.split("+", 1)[1]
868
  except IndexError:
869
    raise errors.OpPrereqError("OS name must include a variant",
870
                               errors.ECODE_INVAL)
871

    
872
  if variant not in os_obj.supported_variants:
873
    raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
874

    
875

    
876
def _GetNodeInstancesInner(cfg, fn):
877
  return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
878

    
879

    
880
def _GetNodeInstances(cfg, node_name):
881
  """Returns a list of all primary and secondary instances on a node.
882

883
  """
884

    
885
  return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
886

    
887

    
888
def _GetNodePrimaryInstances(cfg, node_name):
889
  """Returns primary instances on a node.
890

891
  """
892
  return _GetNodeInstancesInner(cfg,
893
                                lambda inst: node_name == inst.primary_node)
894

    
895

    
896
def _GetNodeSecondaryInstances(cfg, node_name):
897
  """Returns secondary instances on a node.
898

899
  """
900
  return _GetNodeInstancesInner(cfg,
901
                                lambda inst: node_name in inst.secondary_nodes)
902

    
903

    
904
def _GetStorageTypeArgs(cfg, storage_type):
905
  """Returns the arguments for a storage type.
906

907
  """
908
  # Special case for file storage
909
  if storage_type == constants.ST_FILE:
910
    # storage.FileStorage wants a list of storage directories
911
    return [[cfg.GetFileStorageDir()]]
912

    
913
  return []
914

    
915

    
916
def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq):
917
  faulty = []
918

    
919
  for dev in instance.disks:
920
    cfg.SetDiskID(dev, node_name)
921

    
922
  result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks)
923
  result.Raise("Failed to get disk status from node %s" % node_name,
924
               prereq=prereq, ecode=errors.ECODE_ENVIRON)
925

    
926
  for idx, bdev_status in enumerate(result.payload):
927
    if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
928
      faulty.append(idx)
929

    
930
  return faulty
931

    
932

    
933
class LUPostInitCluster(LogicalUnit):
934
  """Logical unit for running hooks after cluster initialization.
935

936
  """
937
  HPATH = "cluster-init"
938
  HTYPE = constants.HTYPE_CLUSTER
939
  _OP_REQP = []
940

    
941
  def BuildHooksEnv(self):
942
    """Build hooks env.
943

944
    """
945
    env = {"OP_TARGET": self.cfg.GetClusterName()}
946
    mn = self.cfg.GetMasterNode()
947
    return env, [], [mn]
948

    
949
  def CheckPrereq(self):
950
    """No prerequisites to check.
951

952
    """
953
    return True
954

    
955
  def Exec(self, feedback_fn):
956
    """Nothing to do.
957

958
    """
959
    return True
960

    
961

    
962
class LUDestroyCluster(LogicalUnit):
963
  """Logical unit for destroying the cluster.
964

965
  """
966
  HPATH = "cluster-destroy"
967
  HTYPE = constants.HTYPE_CLUSTER
968
  _OP_REQP = []
969

    
970
  def BuildHooksEnv(self):
971
    """Build hooks env.
972

973
    """
974
    env = {"OP_TARGET": self.cfg.GetClusterName()}
975
    return env, [], []
976

    
977
  def CheckPrereq(self):
978
    """Check prerequisites.
979

980
    This checks whether the cluster is empty.
981

982
    Any errors are signaled by raising errors.OpPrereqError.
983

984
    """
985
    master = self.cfg.GetMasterNode()
986

    
987
    nodelist = self.cfg.GetNodeList()
988
    if len(nodelist) != 1 or nodelist[0] != master:
989
      raise errors.OpPrereqError("There are still %d node(s) in"
990
                                 " this cluster." % (len(nodelist) - 1),
991
                                 errors.ECODE_INVAL)
992
    instancelist = self.cfg.GetInstanceList()
993
    if instancelist:
994
      raise errors.OpPrereqError("There are still %d instance(s) in"
995
                                 " this cluster." % len(instancelist),
996
                                 errors.ECODE_INVAL)
997

    
998
  def Exec(self, feedback_fn):
999
    """Destroys the cluster.
1000

1001
    """
1002
    master = self.cfg.GetMasterNode()
1003
    modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
1004

    
1005
    # Run post hooks on master node before it's removed
1006
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
1007
    try:
1008
      hm.RunPhase(constants.HOOKS_PHASE_POST, [master])
1009
    except:
1010
      # pylint: disable-msg=W0702
1011
      self.LogWarning("Errors occurred running hooks on %s" % master)
1012

    
1013
    result = self.rpc.call_node_stop_master(master, False)
1014
    result.Raise("Could not disable the master role")
1015

    
1016
    if modify_ssh_setup:
1017
      priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
1018
      utils.CreateBackup(priv_key)
1019
      utils.CreateBackup(pub_key)
1020

    
1021
    return master
1022

    
1023

    
1024
def _VerifyCertificate(filename):
1025
  """Verifies a certificate for LUVerifyCluster.
1026

1027
  @type filename: string
1028
  @param filename: Path to PEM file
1029

1030
  """
1031
  try:
1032
    cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
1033
                                           utils.ReadFile(filename))
1034
  except Exception, err: # pylint: disable-msg=W0703
1035
    return (LUVerifyCluster.ETYPE_ERROR,
1036
            "Failed to load X509 certificate %s: %s" % (filename, err))
1037

    
1038
  (errcode, msg) = \
1039
    utils.VerifyX509Certificate(cert, constants.SSL_CERT_EXPIRATION_WARN,
1040
                                constants.SSL_CERT_EXPIRATION_ERROR)
1041

    
1042
  if msg:
1043
    fnamemsg = "While verifying %s: %s" % (filename, msg)
1044
  else:
1045
    fnamemsg = None
1046

    
1047
  if errcode is None:
1048
    return (None, fnamemsg)
1049
  elif errcode == utils.CERT_WARNING:
1050
    return (LUVerifyCluster.ETYPE_WARNING, fnamemsg)
1051
  elif errcode == utils.CERT_ERROR:
1052
    return (LUVerifyCluster.ETYPE_ERROR, fnamemsg)
1053

    
1054
  raise errors.ProgrammerError("Unhandled certificate error code %r" % errcode)
1055

    
1056

    
1057
class LUVerifyCluster(LogicalUnit):
1058
  """Verifies the cluster status.
1059

1060
  """
1061
  HPATH = "cluster-verify"
1062
  HTYPE = constants.HTYPE_CLUSTER
1063
  _OP_REQP = ["skip_checks", "verbose", "error_codes", "debug_simulate_errors"]
1064
  REQ_BGL = False
1065

    
1066
  TCLUSTER = "cluster"
1067
  TNODE = "node"
1068
  TINSTANCE = "instance"
1069

    
1070
  ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
1071
  ECLUSTERCERT = (TCLUSTER, "ECLUSTERCERT")
1072
  EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
1073
  EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
1074
  EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
1075
  EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
1076
  EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
1077
  EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE")
1078
  ENODEDRBD = (TNODE, "ENODEDRBD")
1079
  ENODEFILECHECK = (TNODE, "ENODEFILECHECK")
1080
  ENODEHOOKS = (TNODE, "ENODEHOOKS")
1081
  ENODEHV = (TNODE, "ENODEHV")
1082
  ENODELVM = (TNODE, "ENODELVM")
1083
  ENODEN1 = (TNODE, "ENODEN1")
1084
  ENODENET = (TNODE, "ENODENET")
1085
  ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE")
1086
  ENODEORPHANLV = (TNODE, "ENODEORPHANLV")
1087
  ENODERPC = (TNODE, "ENODERPC")
1088
  ENODESSH = (TNODE, "ENODESSH")
1089
  ENODEVERSION = (TNODE, "ENODEVERSION")
1090
  ENODESETUP = (TNODE, "ENODESETUP")
1091
  ENODETIME = (TNODE, "ENODETIME")
1092

    
1093
  ETYPE_FIELD = "code"
1094
  ETYPE_ERROR = "ERROR"
1095
  ETYPE_WARNING = "WARNING"
1096

    
1097
  class NodeImage(object):
1098
    """A class representing the logical and physical status of a node.
1099

1100
    @ivar volumes: a structure as returned from
1101
        L{ganeti.backend.GetVolumeList} (runtime)
1102
    @ivar instances: a list of running instances (runtime)
1103
    @ivar pinst: list of configured primary instances (config)
1104
    @ivar sinst: list of configured secondary instances (config)
1105
    @ivar sbp: diction of {secondary-node: list of instances} of all peers
1106
        of this node (config)
1107
    @ivar mfree: free memory, as reported by hypervisor (runtime)
1108
    @ivar dfree: free disk, as reported by the node (runtime)
1109
    @ivar offline: the offline status (config)
1110
    @type rpc_fail: boolean
1111
    @ivar rpc_fail: whether the RPC verify call was successfull (overall,
1112
        not whether the individual keys were correct) (runtime)
1113
    @type lvm_fail: boolean
1114
    @ivar lvm_fail: whether the RPC call didn't return valid LVM data
1115
    @type hyp_fail: boolean
1116
    @ivar hyp_fail: whether the RPC call didn't return the instance list
1117
    @type ghost: boolean
1118
    @ivar ghost: whether this is a known node or not (config)
1119

1120
    """
1121
    def __init__(self, offline=False):
1122
      self.volumes = {}
1123
      self.instances = []
1124
      self.pinst = []
1125
      self.sinst = []
1126
      self.sbp = {}
1127
      self.mfree = 0
1128
      self.dfree = 0
1129
      self.offline = offline
1130
      self.rpc_fail = False
1131
      self.lvm_fail = False
1132
      self.hyp_fail = False
1133
      self.ghost = False
1134

    
1135
  def ExpandNames(self):
1136
    self.needed_locks = {
1137
      locking.LEVEL_NODE: locking.ALL_SET,
1138
      locking.LEVEL_INSTANCE: locking.ALL_SET,
1139
    }
1140
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
1141

    
1142
  def _Error(self, ecode, item, msg, *args, **kwargs):
1143
    """Format an error message.
1144

1145
    Based on the opcode's error_codes parameter, either format a
1146
    parseable error code, or a simpler error string.
1147

1148
    This must be called only from Exec and functions called from Exec.
1149

1150
    """
1151
    ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
1152
    itype, etxt = ecode
1153
    # first complete the msg
1154
    if args:
1155
      msg = msg % args
1156
    # then format the whole message
1157
    if self.op.error_codes:
1158
      msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
1159
    else:
1160
      if item:
1161
        item = " " + item
1162
      else:
1163
        item = ""
1164
      msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
1165
    # and finally report it via the feedback_fn
1166
    self._feedback_fn("  - %s" % msg)
1167

    
1168
  def _ErrorIf(self, cond, *args, **kwargs):
1169
    """Log an error message if the passed condition is True.
1170

1171
    """
1172
    cond = bool(cond) or self.op.debug_simulate_errors
1173
    if cond:
1174
      self._Error(*args, **kwargs)
1175
    # do not mark the operation as failed for WARN cases only
1176
    if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
1177
      self.bad = self.bad or cond
1178

    
1179
  def _VerifyNode(self, ninfo, nresult):
1180
    """Run multiple tests against a node.
1181

1182
    Test list:
1183

1184
      - compares ganeti version
1185
      - checks vg existence and size > 20G
1186
      - checks config file checksum
1187
      - checks ssh to other nodes
1188

1189
    @type ninfo: L{objects.Node}
1190
    @param ninfo: the node to check
1191
    @param nresult: the results from the node
1192
    @rtype: boolean
1193
    @return: whether overall this call was successful (and we can expect
1194
         reasonable values in the respose)
1195

1196
    """
1197
    node = ninfo.name
1198
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1199

    
1200
    # main result, nresult should be a non-empty dict
1201
    test = not nresult or not isinstance(nresult, dict)
1202
    _ErrorIf(test, self.ENODERPC, node,
1203
                  "unable to verify node: no data returned")
1204
    if test:
1205
      return False
1206

    
1207
    # compares ganeti version
1208
    local_version = constants.PROTOCOL_VERSION
1209
    remote_version = nresult.get("version", None)
1210
    test = not (remote_version and
1211
                isinstance(remote_version, (list, tuple)) and
1212
                len(remote_version) == 2)
1213
    _ErrorIf(test, self.ENODERPC, node,
1214
             "connection to node returned invalid data")
1215
    if test:
1216
      return False
1217

    
1218
    test = local_version != remote_version[0]
1219
    _ErrorIf(test, self.ENODEVERSION, node,
1220
             "incompatible protocol versions: master %s,"
1221
             " node %s", local_version, remote_version[0])
1222
    if test:
1223
      return False
1224

    
1225
    # node seems compatible, we can actually try to look into its results
1226

    
1227
    # full package version
1228
    self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
1229
                  self.ENODEVERSION, node,
1230
                  "software version mismatch: master %s, node %s",
1231
                  constants.RELEASE_VERSION, remote_version[1],
1232
                  code=self.ETYPE_WARNING)
1233

    
1234
    hyp_result = nresult.get(constants.NV_HYPERVISOR, None)
1235
    if isinstance(hyp_result, dict):
1236
      for hv_name, hv_result in hyp_result.iteritems():
1237
        test = hv_result is not None
1238
        _ErrorIf(test, self.ENODEHV, node,
1239
                 "hypervisor %s verify failure: '%s'", hv_name, hv_result)
1240

    
1241

    
1242
    test = nresult.get(constants.NV_NODESETUP,
1243
                           ["Missing NODESETUP results"])
1244
    _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s",
1245
             "; ".join(test))
1246

    
1247
    return True
1248

    
1249
  def _VerifyNodeTime(self, ninfo, nresult,
1250
                      nvinfo_starttime, nvinfo_endtime):
1251
    """Check the node time.
1252

1253
    @type ninfo: L{objects.Node}
1254
    @param ninfo: the node to check
1255
    @param nresult: the remote results for the node
1256
    @param nvinfo_starttime: the start time of the RPC call
1257
    @param nvinfo_endtime: the end time of the RPC call
1258

1259
    """
1260
    node = ninfo.name
1261
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1262

    
1263
    ntime = nresult.get(constants.NV_TIME, None)
1264
    try:
1265
      ntime_merged = utils.MergeTime(ntime)
1266
    except (ValueError, TypeError):
1267
      _ErrorIf(True, self.ENODETIME, node, "Node returned invalid time")
1268
      return
1269

    
1270
    if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
1271
      ntime_diff = "%.01fs" % abs(nvinfo_starttime - ntime_merged)
1272
    elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
1273
      ntime_diff = "%.01fs" % abs(ntime_merged - nvinfo_endtime)
1274
    else:
1275
      ntime_diff = None
1276

    
1277
    _ErrorIf(ntime_diff is not None, self.ENODETIME, node,
1278
             "Node time diverges by at least %s from master node time",
1279
             ntime_diff)
1280

    
1281
  def _VerifyNodeLVM(self, ninfo, nresult, vg_name):
1282
    """Check the node time.
1283

1284
    @type ninfo: L{objects.Node}
1285
    @param ninfo: the node to check
1286
    @param nresult: the remote results for the node
1287
    @param vg_name: the configured VG name
1288

1289
    """
1290
    if vg_name is None:
1291
      return
1292

    
1293
    node = ninfo.name
1294
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1295

    
1296
    # checks vg existence and size > 20G
1297
    vglist = nresult.get(constants.NV_VGLIST, None)
1298
    test = not vglist
1299
    _ErrorIf(test, self.ENODELVM, node, "unable to check volume groups")
1300
    if not test:
1301
      vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
1302
                                            constants.MIN_VG_SIZE)
1303
      _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
1304

    
1305
    # check pv names
1306
    pvlist = nresult.get(constants.NV_PVLIST, None)
1307
    test = pvlist is None
1308
    _ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node")
1309
    if not test:
1310
      # check that ':' is not present in PV names, since it's a
1311
      # special character for lvcreate (denotes the range of PEs to
1312
      # use on the PV)
1313
      for _, pvname, owner_vg in pvlist:
1314
        test = ":" in pvname
1315
        _ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV"
1316
                 " '%s' of VG '%s'", pvname, owner_vg)
1317

    
1318
  def _VerifyNodeNetwork(self, ninfo, nresult):
1319
    """Check the node time.
1320

1321
    @type ninfo: L{objects.Node}
1322
    @param ninfo: the node to check
1323
    @param nresult: the remote results for the node
1324

1325
    """
1326
    node = ninfo.name
1327
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1328

    
1329
    test = constants.NV_NODELIST not in nresult
1330
    _ErrorIf(test, self.ENODESSH, node,
1331
             "node hasn't returned node ssh connectivity data")
1332
    if not test:
1333
      if nresult[constants.NV_NODELIST]:
1334
        for a_node, a_msg in nresult[constants.NV_NODELIST].items():
1335
          _ErrorIf(True, self.ENODESSH, node,
1336
                   "ssh communication with node '%s': %s", a_node, a_msg)
1337

    
1338
    test = constants.NV_NODENETTEST not in nresult
1339
    _ErrorIf(test, self.ENODENET, node,
1340
             "node hasn't returned node tcp connectivity data")
1341
    if not test:
1342
      if nresult[constants.NV_NODENETTEST]:
1343
        nlist = utils.NiceSort(nresult[constants.NV_NODENETTEST].keys())
1344
        for anode in nlist:
1345
          _ErrorIf(True, self.ENODENET, node,
1346
                   "tcp communication with node '%s': %s",
1347
                   anode, nresult[constants.NV_NODENETTEST][anode])
1348

    
1349
    test = constants.NV_MASTERIP not in nresult
1350
    _ErrorIf(test, self.ENODENET, node,
1351
             "node hasn't returned node master IP reachability data")
1352
    if not test:
1353
      if not nresult[constants.NV_MASTERIP]:
1354
        if node == self.master_node:
1355
          msg = "the master node cannot reach the master IP (not configured?)"
1356
        else:
1357
          msg = "cannot reach the master IP"
1358
        _ErrorIf(True, self.ENODENET, node, msg)
1359

    
1360

    
1361
  def _VerifyInstance(self, instance, instanceconfig, node_image):
1362
    """Verify an instance.
1363

1364
    This function checks to see if the required block devices are
1365
    available on the instance's node.
1366

1367
    """
1368
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1369
    node_current = instanceconfig.primary_node
1370

    
1371
    node_vol_should = {}
1372
    instanceconfig.MapLVsByNode(node_vol_should)
1373

    
1374
    for node in node_vol_should:
1375
      n_img = node_image[node]
1376
      if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1377
        # ignore missing volumes on offline or broken nodes
1378
        continue
1379
      for volume in node_vol_should[node]:
1380
        test = volume not in n_img.volumes
1381
        _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
1382
                 "volume %s missing on node %s", volume, node)
1383

    
1384
    if instanceconfig.admin_up:
1385
      pri_img = node_image[node_current]
1386
      test = instance not in pri_img.instances and not pri_img.offline
1387
      _ErrorIf(test, self.EINSTANCEDOWN, instance,
1388
               "instance not running on its primary node %s",
1389
               node_current)
1390

    
1391
    for node, n_img in node_image.items():
1392
      if (not node == node_current):
1393
        test = instance in n_img.instances
1394
        _ErrorIf(test, self.EINSTANCEWRONGNODE, instance,
1395
                 "instance should not run on node %s", node)
1396

    
1397
  def _VerifyOrphanVolumes(self, node_vol_should, node_image):
1398
    """Verify if there are any unknown volumes in the cluster.
1399

1400
    The .os, .swap and backup volumes are ignored. All other volumes are
1401
    reported as unknown.
1402

1403
    """
1404
    for node, n_img in node_image.items():
1405
      if n_img.offline or n_img.rpc_fail or n_img.lvm_fail:
1406
        # skip non-healthy nodes
1407
        continue
1408
      for volume in n_img.volumes:
1409
        test = (node not in node_vol_should or
1410
                volume not in node_vol_should[node])
1411
        self._ErrorIf(test, self.ENODEORPHANLV, node,
1412
                      "volume %s is unknown", volume)
1413

    
1414
  def _VerifyOrphanInstances(self, instancelist, node_image):
1415
    """Verify the list of running instances.
1416

1417
    This checks what instances are running but unknown to the cluster.
1418

1419
    """
1420
    for node, n_img in node_image.items():
1421
      for o_inst in n_img.instances:
1422
        test = o_inst not in instancelist
1423
        self._ErrorIf(test, self.ENODEORPHANINSTANCE, node,
1424
                      "instance %s on node %s should not exist", o_inst, node)
1425

    
1426
  def _VerifyNPlusOneMemory(self, node_image, instance_cfg):
1427
    """Verify N+1 Memory Resilience.
1428

1429
    Check that if one single node dies we can still start all the
1430
    instances it was primary for.
1431

1432
    """
1433
    for node, n_img in node_image.items():
1434
      # This code checks that every node which is now listed as
1435
      # secondary has enough memory to host all instances it is
1436
      # supposed to should a single other node in the cluster fail.
1437
      # FIXME: not ready for failover to an arbitrary node
1438
      # FIXME: does not support file-backed instances
1439
      # WARNING: we currently take into account down instances as well
1440
      # as up ones, considering that even if they're down someone
1441
      # might want to start them even in the event of a node failure.
1442
      for prinode, instances in n_img.sbp.items():
1443
        needed_mem = 0
1444
        for instance in instances:
1445
          bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
1446
          if bep[constants.BE_AUTO_BALANCE]:
1447
            needed_mem += bep[constants.BE_MEMORY]
1448
        test = n_img.mfree < needed_mem
1449
        self._ErrorIf(test, self.ENODEN1, node,
1450
                      "not enough memory on to accommodate"
1451
                      " failovers should peer node %s fail", prinode)
1452

    
1453
  def _VerifyNodeFiles(self, ninfo, nresult, file_list, local_cksum,
1454
                       master_files):
1455
    """Verifies and computes the node required file checksums.
1456

1457
    @type ninfo: L{objects.Node}
1458
    @param ninfo: the node to check
1459
    @param nresult: the remote results for the node
1460
    @param file_list: required list of files
1461
    @param local_cksum: dictionary of local files and their checksums
1462
    @param master_files: list of files that only masters should have
1463

1464
    """
1465
    node = ninfo.name
1466
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1467

    
1468
    remote_cksum = nresult.get(constants.NV_FILELIST, None)
1469
    test = not isinstance(remote_cksum, dict)
1470
    _ErrorIf(test, self.ENODEFILECHECK, node,
1471
             "node hasn't returned file checksum data")
1472
    if test:
1473
      return
1474

    
1475
    for file_name in file_list:
1476
      node_is_mc = ninfo.master_candidate
1477
      must_have = (file_name not in master_files) or node_is_mc
1478
      # missing
1479
      test1 = file_name not in remote_cksum
1480
      # invalid checksum
1481
      test2 = not test1 and remote_cksum[file_name] != local_cksum[file_name]
1482
      # existing and good
1483
      test3 = not test1 and remote_cksum[file_name] == local_cksum[file_name]
1484
      _ErrorIf(test1 and must_have, self.ENODEFILECHECK, node,
1485
               "file '%s' missing", file_name)
1486
      _ErrorIf(test2 and must_have, self.ENODEFILECHECK, node,
1487
               "file '%s' has wrong checksum", file_name)
1488
      # not candidate and this is not a must-have file
1489
      _ErrorIf(test2 and not must_have, self.ENODEFILECHECK, node,
1490
               "file '%s' should not exist on non master"
1491
               " candidates (and the file is outdated)", file_name)
1492
      # all good, except non-master/non-must have combination
1493
      _ErrorIf(test3 and not must_have, self.ENODEFILECHECK, node,
1494
               "file '%s' should not exist"
1495
               " on non master candidates", file_name)
1496

    
1497
  def _VerifyNodeDrbd(self, ninfo, nresult, instanceinfo, drbd_map):
1498
    """Verifies and the node DRBD status.
1499

1500
    @type ninfo: L{objects.Node}
1501
    @param ninfo: the node to check
1502
    @param nresult: the remote results for the node
1503
    @param instanceinfo: the dict of instances
1504
    @param drbd_map: the DRBD map as returned by
1505
        L{ganeti.config.ConfigWriter.ComputeDRBDMap}
1506

1507
    """
1508
    node = ninfo.name
1509
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1510

    
1511
    # compute the DRBD minors
1512
    node_drbd = {}
1513
    for minor, instance in drbd_map[node].items():
1514
      test = instance not in instanceinfo
1515
      _ErrorIf(test, self.ECLUSTERCFG, None,
1516
               "ghost instance '%s' in temporary DRBD map", instance)
1517
        # ghost instance should not be running, but otherwise we
1518
        # don't give double warnings (both ghost instance and
1519
        # unallocated minor in use)
1520
      if test:
1521
        node_drbd[minor] = (instance, False)
1522
      else:
1523
        instance = instanceinfo[instance]
1524
        node_drbd[minor] = (instance.name, instance.admin_up)
1525

    
1526
    # and now check them
1527
    used_minors = nresult.get(constants.NV_DRBDLIST, [])
1528
    test = not isinstance(used_minors, (tuple, list))
1529
    _ErrorIf(test, self.ENODEDRBD, node,
1530
             "cannot parse drbd status file: %s", str(used_minors))
1531
    if test:
1532
      # we cannot check drbd status
1533
      return
1534

    
1535
    for minor, (iname, must_exist) in node_drbd.items():
1536
      test = minor not in used_minors and must_exist
1537
      _ErrorIf(test, self.ENODEDRBD, node,
1538
               "drbd minor %d of instance %s is not active", minor, iname)
1539
    for minor in used_minors:
1540
      test = minor not in node_drbd
1541
      _ErrorIf(test, self.ENODEDRBD, node,
1542
               "unallocated drbd minor %d is in use", minor)
1543

    
1544
  def _UpdateNodeVolumes(self, ninfo, nresult, nimg, vg_name):
1545
    """Verifies and updates the node volume data.
1546

1547
    This function will update a L{NodeImage}'s internal structures
1548
    with data from the remote call.
1549

1550
    @type ninfo: L{objects.Node}
1551
    @param ninfo: the node to check
1552
    @param nresult: the remote results for the node
1553
    @param nimg: the node image object
1554
    @param vg_name: the configured VG name
1555

1556
    """
1557
    node = ninfo.name
1558
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1559

    
1560
    nimg.lvm_fail = True
1561
    lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
1562
    if vg_name is None:
1563
      pass
1564
    elif isinstance(lvdata, basestring):
1565
      _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
1566
               utils.SafeEncode(lvdata))
1567
    elif not isinstance(lvdata, dict):
1568
      _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
1569
    else:
1570
      nimg.volumes = lvdata
1571
      nimg.lvm_fail = False
1572

    
1573
  def _UpdateNodeInstances(self, ninfo, nresult, nimg):
1574
    """Verifies and updates the node instance list.
1575

1576
    If the listing was successful, then updates this node's instance
1577
    list. Otherwise, it marks the RPC call as failed for the instance
1578
    list key.
1579

1580
    @type ninfo: L{objects.Node}
1581
    @param ninfo: the node to check
1582
    @param nresult: the remote results for the node
1583
    @param nimg: the node image object
1584

1585
    """
1586
    idata = nresult.get(constants.NV_INSTANCELIST, None)
1587
    test = not isinstance(idata, list)
1588
    self._ErrorIf(test, self.ENODEHV, ninfo.name, "rpc call to node failed"
1589
                  " (instancelist): %s", utils.SafeEncode(str(idata)))
1590
    if test:
1591
      nimg.hyp_fail = True
1592
    else:
1593
      nimg.instances = idata
1594

    
1595
  def _UpdateNodeInfo(self, ninfo, nresult, nimg, vg_name):
1596
    """Verifies and computes a node information map
1597

1598
    @type ninfo: L{objects.Node}
1599
    @param ninfo: the node to check
1600
    @param nresult: the remote results for the node
1601
    @param nimg: the node image object
1602
    @param vg_name: the configured VG name
1603

1604
    """
1605
    node = ninfo.name
1606
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1607

    
1608
    # try to read free memory (from the hypervisor)
1609
    hv_info = nresult.get(constants.NV_HVINFO, None)
1610
    test = not isinstance(hv_info, dict) or "memory_free" not in hv_info
1611
    _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
1612
    if not test:
1613
      try:
1614
        nimg.mfree = int(hv_info["memory_free"])
1615
      except (ValueError, TypeError):
1616
        _ErrorIf(True, self.ENODERPC, node,
1617
                 "node returned invalid nodeinfo, check hypervisor")
1618

    
1619
    # FIXME: devise a free space model for file based instances as well
1620
    if vg_name is not None:
1621
      test = (constants.NV_VGLIST not in nresult or
1622
              vg_name not in nresult[constants.NV_VGLIST])
1623
      _ErrorIf(test, self.ENODELVM, node,
1624
               "node didn't return data for the volume group '%s'"
1625
               " - it is either missing or broken", vg_name)
1626
      if not test:
1627
        try:
1628
          nimg.dfree = int(nresult[constants.NV_VGLIST][vg_name])
1629
        except (ValueError, TypeError):
1630
          _ErrorIf(True, self.ENODERPC, node,
1631
                   "node returned invalid LVM info, check LVM status")
1632

    
1633
  def CheckPrereq(self):
1634
    """Check prerequisites.
1635

1636
    Transform the list of checks we're going to skip into a set and check that
1637
    all its members are valid.
1638

1639
    """
1640
    self.skip_set = frozenset(self.op.skip_checks)
1641
    if not constants.VERIFY_OPTIONAL_CHECKS.issuperset(self.skip_set):
1642
      raise errors.OpPrereqError("Invalid checks to be skipped specified",
1643
                                 errors.ECODE_INVAL)
1644

    
1645
  def BuildHooksEnv(self):
1646
    """Build hooks env.
1647

1648
    Cluster-Verify hooks just ran in the post phase and their failure makes
1649
    the output be logged in the verify output and the verification to fail.
1650

1651
    """
1652
    all_nodes = self.cfg.GetNodeList()
1653
    env = {
1654
      "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
1655
      }
1656
    for node in self.cfg.GetAllNodesInfo().values():
1657
      env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
1658

    
1659
    return env, [], all_nodes
1660

    
1661
  def Exec(self, feedback_fn):
1662
    """Verify integrity of cluster, performing various test on nodes.
1663

1664
    """
1665
    self.bad = False
1666
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1667
    verbose = self.op.verbose
1668
    self._feedback_fn = feedback_fn
1669
    feedback_fn("* Verifying global settings")
1670
    for msg in self.cfg.VerifyConfig():
1671
      _ErrorIf(True, self.ECLUSTERCFG, None, msg)
1672

    
1673
    # Check the cluster certificates
1674
    for cert_filename in constants.ALL_CERT_FILES:
1675
      (errcode, msg) = _VerifyCertificate(cert_filename)
1676
      _ErrorIf(errcode, self.ECLUSTERCERT, None, msg, code=errcode)
1677

    
1678
    vg_name = self.cfg.GetVGName()
1679
    hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
1680
    cluster = self.cfg.GetClusterInfo()
1681
    nodelist = utils.NiceSort(self.cfg.GetNodeList())
1682
    nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
1683
    instancelist = utils.NiceSort(self.cfg.GetInstanceList())
1684
    instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
1685
                        for iname in instancelist)
1686
    i_non_redundant = [] # Non redundant instances
1687
    i_non_a_balanced = [] # Non auto-balanced instances
1688
    n_offline = 0 # Count of offline nodes
1689
    n_drained = 0 # Count of nodes being drained
1690
    node_vol_should = {}
1691

    
1692
    # FIXME: verify OS list
1693
    # do local checksums
1694
    master_files = [constants.CLUSTER_CONF_FILE]
1695
    master_node = self.master_node = self.cfg.GetMasterNode()
1696
    master_ip = self.cfg.GetMasterIP()
1697

    
1698
    file_names = ssconf.SimpleStore().GetFileList()
1699
    file_names.extend(constants.ALL_CERT_FILES)
1700
    file_names.extend(master_files)
1701
    if cluster.modify_etc_hosts:
1702
      file_names.append(constants.ETC_HOSTS)
1703

    
1704
    local_checksums = utils.FingerprintFiles(file_names)
1705

    
1706
    feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
1707
    node_verify_param = {
1708
      constants.NV_FILELIST: file_names,
1709
      constants.NV_NODELIST: [node.name for node in nodeinfo
1710
                              if not node.offline],
1711
      constants.NV_HYPERVISOR: hypervisors,
1712
      constants.NV_NODENETTEST: [(node.name, node.primary_ip,
1713
                                  node.secondary_ip) for node in nodeinfo
1714
                                 if not node.offline],
1715
      constants.NV_INSTANCELIST: hypervisors,
1716
      constants.NV_VERSION: None,
1717
      constants.NV_HVINFO: self.cfg.GetHypervisorType(),
1718
      constants.NV_NODESETUP: None,
1719
      constants.NV_TIME: None,
1720
      constants.NV_MASTERIP: (master_node, master_ip),
1721
      }
1722

    
1723
    if vg_name is not None:
1724
      node_verify_param[constants.NV_VGLIST] = None
1725
      node_verify_param[constants.NV_LVLIST] = vg_name
1726
      node_verify_param[constants.NV_PVLIST] = [vg_name]
1727
      node_verify_param[constants.NV_DRBDLIST] = None
1728

    
1729
    # Build our expected cluster state
1730
    node_image = dict((node.name, self.NodeImage(offline=node.offline))
1731
                      for node in nodeinfo)
1732

    
1733
    for instance in instancelist:
1734
      inst_config = instanceinfo[instance]
1735

    
1736
      for nname in inst_config.all_nodes:
1737
        if nname not in node_image:
1738
          # ghost node
1739
          gnode = self.NodeImage()
1740
          gnode.ghost = True
1741
          node_image[nname] = gnode
1742

    
1743
      inst_config.MapLVsByNode(node_vol_should)
1744

    
1745
      pnode = inst_config.primary_node
1746
      node_image[pnode].pinst.append(instance)
1747

    
1748
      for snode in inst_config.secondary_nodes:
1749
        nimg = node_image[snode]
1750
        nimg.sinst.append(instance)
1751
        if pnode not in nimg.sbp:
1752
          nimg.sbp[pnode] = []
1753
        nimg.sbp[pnode].append(instance)
1754

    
1755
    # At this point, we have the in-memory data structures complete,
1756
    # except for the runtime information, which we'll gather next
1757

    
1758
    # Due to the way our RPC system works, exact response times cannot be
1759
    # guaranteed (e.g. a broken node could run into a timeout). By keeping the
1760
    # time before and after executing the request, we can at least have a time
1761
    # window.
1762
    nvinfo_starttime = time.time()
1763
    all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
1764
                                           self.cfg.GetClusterName())
1765
    nvinfo_endtime = time.time()
1766

    
1767
    all_drbd_map = self.cfg.ComputeDRBDMap()
1768

    
1769
    feedback_fn("* Verifying node status")
1770
    for node_i in nodeinfo:
1771
      node = node_i.name
1772
      nimg = node_image[node]
1773

    
1774
      if node_i.offline:
1775
        if verbose:
1776
          feedback_fn("* Skipping offline node %s" % (node,))
1777
        n_offline += 1
1778
        continue
1779

    
1780
      if node == master_node:
1781
        ntype = "master"
1782
      elif node_i.master_candidate:
1783
        ntype = "master candidate"
1784
      elif node_i.drained:
1785
        ntype = "drained"
1786
        n_drained += 1
1787
      else:
1788
        ntype = "regular"
1789
      if verbose:
1790
        feedback_fn("* Verifying node %s (%s)" % (node, ntype))
1791

    
1792
      msg = all_nvinfo[node].fail_msg
1793
      _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
1794
      if msg:
1795
        nimg.rpc_fail = True
1796
        continue
1797

    
1798
      nresult = all_nvinfo[node].payload
1799

    
1800
      nimg.call_ok = self._VerifyNode(node_i, nresult)
1801
      self._VerifyNodeNetwork(node_i, nresult)
1802
      self._VerifyNodeLVM(node_i, nresult, vg_name)
1803
      self._VerifyNodeFiles(node_i, nresult, file_names, local_checksums,
1804
                            master_files)
1805
      self._VerifyNodeDrbd(node_i, nresult, instanceinfo, all_drbd_map)
1806
      self._VerifyNodeTime(node_i, nresult, nvinfo_starttime, nvinfo_endtime)
1807

    
1808
      self._UpdateNodeVolumes(node_i, nresult, nimg, vg_name)
1809
      self._UpdateNodeInstances(node_i, nresult, nimg)
1810
      self._UpdateNodeInfo(node_i, nresult, nimg, vg_name)
1811

    
1812
    feedback_fn("* Verifying instance status")
1813
    for instance in instancelist:
1814
      if verbose:
1815
        feedback_fn("* Verifying instance %s" % instance)
1816
      inst_config = instanceinfo[instance]
1817
      self._VerifyInstance(instance, inst_config, node_image)
1818
      inst_nodes_offline = []
1819

    
1820
      pnode = inst_config.primary_node
1821
      pnode_img = node_image[pnode]
1822
      _ErrorIf(pnode_img.rpc_fail and not pnode_img.offline,
1823
               self.ENODERPC, pnode, "instance %s, connection to"
1824
               " primary node failed", instance)
1825

    
1826
      if pnode_img.offline:
1827
        inst_nodes_offline.append(pnode)
1828

    
1829
      # If the instance is non-redundant we cannot survive losing its primary
1830
      # node, so we are not N+1 compliant. On the other hand we have no disk
1831
      # templates with more than one secondary so that situation is not well
1832
      # supported either.
1833
      # FIXME: does not support file-backed instances
1834
      if not inst_config.secondary_nodes:
1835
        i_non_redundant.append(instance)
1836
      _ErrorIf(len(inst_config.secondary_nodes) > 1, self.EINSTANCELAYOUT,
1837
               instance, "instance has multiple secondary nodes: %s",
1838
               utils.CommaJoin(inst_config.secondary_nodes),
1839
               code=self.ETYPE_WARNING)
1840

    
1841
      if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
1842
        i_non_a_balanced.append(instance)
1843

    
1844
      for snode in inst_config.secondary_nodes:
1845
        s_img = node_image[snode]
1846
        _ErrorIf(s_img.rpc_fail and not s_img.offline, self.ENODERPC, snode,
1847
                 "instance %s, connection to secondary node failed", instance)
1848

    
1849
        if s_img.offline:
1850
          inst_nodes_offline.append(snode)
1851

    
1852
      # warn that the instance lives on offline nodes
1853
      _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
1854
               "instance lives on offline node(s) %s",
1855
               utils.CommaJoin(inst_nodes_offline))
1856
      # ... or ghost nodes
1857
      for node in inst_config.all_nodes:
1858
        _ErrorIf(node_image[node].ghost, self.EINSTANCEBADNODE, instance,
1859
                 "instance lives on ghost node %s", node)
1860

    
1861
    feedback_fn("* Verifying orphan volumes")
1862
    self._VerifyOrphanVolumes(node_vol_should, node_image)
1863

    
1864
    feedback_fn("* Verifying orphan instances")
1865
    self._VerifyOrphanInstances(instancelist, node_image)
1866

    
1867
    if constants.VERIFY_NPLUSONE_MEM not in self.skip_set:
1868
      feedback_fn("* Verifying N+1 Memory redundancy")
1869
      self._VerifyNPlusOneMemory(node_image, instanceinfo)
1870

    
1871
    feedback_fn("* Other Notes")
1872
    if i_non_redundant:
1873
      feedback_fn("  - NOTICE: %d non-redundant instance(s) found."
1874
                  % len(i_non_redundant))
1875

    
1876
    if i_non_a_balanced:
1877
      feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
1878
                  % len(i_non_a_balanced))
1879

    
1880
    if n_offline:
1881
      feedback_fn("  - NOTICE: %d offline node(s) found." % n_offline)
1882

    
1883
    if n_drained:
1884
      feedback_fn("  - NOTICE: %d drained node(s) found." % n_drained)
1885

    
1886
    return not self.bad
1887

    
1888
  def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
1889
    """Analyze the post-hooks' result
1890

1891
    This method analyses the hook result, handles it, and sends some
1892
    nicely-formatted feedback back to the user.
1893

1894
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
1895
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
1896
    @param hooks_results: the results of the multi-node hooks rpc call
1897
    @param feedback_fn: function used send feedback back to the caller
1898
    @param lu_result: previous Exec result
1899
    @return: the new Exec result, based on the previous result
1900
        and hook results
1901

1902
    """
1903
    # We only really run POST phase hooks, and are only interested in
1904
    # their results
1905
    if phase == constants.HOOKS_PHASE_POST:
1906
      # Used to change hooks' output to proper indentation
1907
      indent_re = re.compile('^', re.M)
1908
      feedback_fn("* Hooks Results")
1909
      assert hooks_results, "invalid result from hooks"
1910

    
1911
      for node_name in hooks_results:
1912
        res = hooks_results[node_name]
1913
        msg = res.fail_msg
1914
        test = msg and not res.offline
1915
        self._ErrorIf(test, self.ENODEHOOKS, node_name,
1916
                      "Communication failure in hooks execution: %s", msg)
1917
        if res.offline or msg:
1918
          # No need to investigate payload if node is offline or gave an error.
1919
          # override manually lu_result here as _ErrorIf only
1920
          # overrides self.bad
1921
          lu_result = 1
1922
          continue
1923
        for script, hkr, output in res.payload:
1924
          test = hkr == constants.HKR_FAIL
1925
          self._ErrorIf(test, self.ENODEHOOKS, node_name,
1926
                        "Script %s failed, output:", script)
1927
          if test:
1928
            output = indent_re.sub('      ', output)
1929
            feedback_fn("%s" % output)
1930
            lu_result = 0
1931

    
1932
      return lu_result
1933

    
1934

    
1935
class LUVerifyDisks(NoHooksLU):
1936
  """Verifies the cluster disks status.
1937

1938
  """
1939
  _OP_REQP = []
1940
  REQ_BGL = False
1941

    
1942
  def ExpandNames(self):
1943
    self.needed_locks = {
1944
      locking.LEVEL_NODE: locking.ALL_SET,
1945
      locking.LEVEL_INSTANCE: locking.ALL_SET,
1946
    }
1947
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
1948

    
1949
  def CheckPrereq(self):
1950
    """Check prerequisites.
1951

1952
    This has no prerequisites.
1953

1954
    """
1955
    pass
1956

    
1957
  def Exec(self, feedback_fn):
1958
    """Verify integrity of cluster disks.
1959

1960
    @rtype: tuple of three items
1961
    @return: a tuple of (dict of node-to-node_error, list of instances
1962
        which need activate-disks, dict of instance: (node, volume) for
1963
        missing volumes
1964

1965
    """
1966
    result = res_nodes, res_instances, res_missing = {}, [], {}
1967

    
1968
    vg_name = self.cfg.GetVGName()
1969
    nodes = utils.NiceSort(self.cfg.GetNodeList())
1970
    instances = [self.cfg.GetInstanceInfo(name)
1971
                 for name in self.cfg.GetInstanceList()]
1972

    
1973
    nv_dict = {}
1974
    for inst in instances:
1975
      inst_lvs = {}
1976
      if (not inst.admin_up or
1977
          inst.disk_template not in constants.DTS_NET_MIRROR):
1978
        continue
1979
      inst.MapLVsByNode(inst_lvs)
1980
      # transform { iname: {node: [vol,],},} to {(node, vol): iname}
1981
      for node, vol_list in inst_lvs.iteritems():
1982
        for vol in vol_list:
1983
          nv_dict[(node, vol)] = inst
1984

    
1985
    if not nv_dict:
1986
      return result
1987

    
1988
    node_lvs = self.rpc.call_lv_list(nodes, vg_name)
1989

    
1990
    for node in nodes:
1991
      # node_volume
1992
      node_res = node_lvs[node]
1993
      if node_res.offline:
1994
        continue
1995
      msg = node_res.fail_msg
1996
      if msg:
1997
        logging.warning("Error enumerating LVs on node %s: %s", node, msg)
1998
        res_nodes[node] = msg
1999
        continue
2000

    
2001
      lvs = node_res.payload
2002
      for lv_name, (_, _, lv_online) in lvs.items():
2003
        inst = nv_dict.pop((node, lv_name), None)
2004
        if (not lv_online and inst is not None
2005
            and inst.name not in res_instances):
2006
          res_instances.append(inst.name)
2007

    
2008
    # any leftover items in nv_dict are missing LVs, let's arrange the
2009
    # data better
2010
    for key, inst in nv_dict.iteritems():
2011
      if inst.name not in res_missing:
2012
        res_missing[inst.name] = []
2013
      res_missing[inst.name].append(key)
2014

    
2015
    return result
2016

    
2017

    
2018
class LURepairDiskSizes(NoHooksLU):
2019
  """Verifies the cluster disks sizes.
2020

2021
  """
2022
  _OP_REQP = ["instances"]
2023
  REQ_BGL = False
2024

    
2025
  def ExpandNames(self):
2026
    if not isinstance(self.op.instances, list):
2027
      raise errors.OpPrereqError("Invalid argument type 'instances'",
2028
                                 errors.ECODE_INVAL)
2029

    
2030
    if self.op.instances:
2031
      self.wanted_names = []
2032
      for name in self.op.instances:
2033
        full_name = _ExpandInstanceName(self.cfg, name)
2034
        self.wanted_names.append(full_name)
2035
      self.needed_locks = {
2036
        locking.LEVEL_NODE: [],
2037
        locking.LEVEL_INSTANCE: self.wanted_names,
2038
        }
2039
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2040
    else:
2041
      self.wanted_names = None
2042
      self.needed_locks = {
2043
        locking.LEVEL_NODE: locking.ALL_SET,
2044
        locking.LEVEL_INSTANCE: locking.ALL_SET,
2045
        }
2046
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
2047

    
2048
  def DeclareLocks(self, level):
2049
    if level == locking.LEVEL_NODE and self.wanted_names is not None:
2050
      self._LockInstancesNodes(primary_only=True)
2051

    
2052
  def CheckPrereq(self):
2053
    """Check prerequisites.
2054

2055
    This only checks the optional instance list against the existing names.
2056

2057
    """
2058
    if self.wanted_names is None:
2059
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
2060

    
2061
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
2062
                             in self.wanted_names]
2063

    
2064
  def _EnsureChildSizes(self, disk):
2065
    """Ensure children of the disk have the needed disk size.
2066

2067
    This is valid mainly for DRBD8 and fixes an issue where the
2068
    children have smaller disk size.
2069

2070
    @param disk: an L{ganeti.objects.Disk} object
2071

2072
    """
2073
    if disk.dev_type == constants.LD_DRBD8:
2074
      assert disk.children, "Empty children for DRBD8?"
2075
      fchild = disk.children[0]
2076
      mismatch = fchild.size < disk.size
2077
      if mismatch:
2078
        self.LogInfo("Child disk has size %d, parent %d, fixing",
2079
                     fchild.size, disk.size)
2080
        fchild.size = disk.size
2081

    
2082
      # and we recurse on this child only, not on the metadev
2083
      return self._EnsureChildSizes(fchild) or mismatch
2084
    else:
2085
      return False
2086

    
2087
  def Exec(self, feedback_fn):
2088
    """Verify the size of cluster disks.
2089

2090
    """
2091
    # TODO: check child disks too
2092
    # TODO: check differences in size between primary/secondary nodes
2093
    per_node_disks = {}
2094
    for instance in self.wanted_instances:
2095
      pnode = instance.primary_node
2096
      if pnode not in per_node_disks:
2097
        per_node_disks[pnode] = []
2098
      for idx, disk in enumerate(instance.disks):
2099
        per_node_disks[pnode].append((instance, idx, disk))
2100

    
2101
    changed = []
2102
    for node, dskl in per_node_disks.items():
2103
      newl = [v[2].Copy() for v in dskl]
2104
      for dsk in newl:
2105
        self.cfg.SetDiskID(dsk, node)
2106
      result = self.rpc.call_blockdev_getsizes(node, newl)
2107
      if result.fail_msg:
2108
        self.LogWarning("Failure in blockdev_getsizes call to node"
2109
                        " %s, ignoring", node)
2110
        continue
2111
      if len(result.data) != len(dskl):
2112
        self.LogWarning("Invalid result from node %s, ignoring node results",
2113
                        node)
2114
        continue
2115
      for ((instance, idx, disk), size) in zip(dskl, result.data):
2116
        if size is None:
2117
          self.LogWarning("Disk %d of instance %s did not return size"
2118
                          " information, ignoring", idx, instance.name)
2119
          continue
2120
        if not isinstance(size, (int, long)):
2121
          self.LogWarning("Disk %d of instance %s did not return valid"
2122
                          " size information, ignoring", idx, instance.name)
2123
          continue
2124
        size = size >> 20
2125
        if size != disk.size:
2126
          self.LogInfo("Disk %d of instance %s has mismatched size,"
2127
                       " correcting: recorded %d, actual %d", idx,
2128
                       instance.name, disk.size, size)
2129
          disk.size = size
2130
          self.cfg.Update(instance, feedback_fn)
2131
          changed.append((instance.name, idx, size))
2132
        if self._EnsureChildSizes(disk):
2133
          self.cfg.Update(instance, feedback_fn)
2134
          changed.append((instance.name, idx, disk.size))
2135
    return changed
2136

    
2137

    
2138
class LURenameCluster(LogicalUnit):
2139
  """Rename the cluster.
2140

2141
  """
2142
  HPATH = "cluster-rename"
2143
  HTYPE = constants.HTYPE_CLUSTER
2144
  _OP_REQP = ["name"]
2145

    
2146
  def BuildHooksEnv(self):
2147
    """Build hooks env.
2148

2149
    """
2150
    env = {
2151
      "OP_TARGET": self.cfg.GetClusterName(),
2152
      "NEW_NAME": self.op.name,
2153
      }
2154
    mn = self.cfg.GetMasterNode()
2155
    all_nodes = self.cfg.GetNodeList()
2156
    return env, [mn], all_nodes
2157

    
2158
  def CheckPrereq(self):
2159
    """Verify that the passed name is a valid one.
2160

2161
    """
2162
    hostname = utils.GetHostInfo(self.op.name)
2163

    
2164
    new_name = hostname.name
2165
    self.ip = new_ip = hostname.ip
2166
    old_name = self.cfg.GetClusterName()
2167
    old_ip = self.cfg.GetMasterIP()
2168
    if new_name == old_name and new_ip == old_ip:
2169
      raise errors.OpPrereqError("Neither the name nor the IP address of the"
2170
                                 " cluster has changed",
2171
                                 errors.ECODE_INVAL)
2172
    if new_ip != old_ip:
2173
      if utils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
2174
        raise errors.OpPrereqError("The given cluster IP address (%s) is"
2175
                                   " reachable on the network. Aborting." %
2176
                                   new_ip, errors.ECODE_NOTUNIQUE)
2177

    
2178
    self.op.name = new_name
2179

    
2180
  def Exec(self, feedback_fn):
2181
    """Rename the cluster.
2182

2183
    """
2184
    clustername = self.op.name
2185
    ip = self.ip
2186

    
2187
    # shutdown the master IP
2188
    master = self.cfg.GetMasterNode()
2189
    result = self.rpc.call_node_stop_master(master, False)
2190
    result.Raise("Could not disable the master role")
2191

    
2192
    try:
2193
      cluster = self.cfg.GetClusterInfo()
2194
      cluster.cluster_name = clustername
2195
      cluster.master_ip = ip
2196
      self.cfg.Update(cluster, feedback_fn)
2197

    
2198
      # update the known hosts file
2199
      ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
2200
      node_list = self.cfg.GetNodeList()
2201
      try:
2202
        node_list.remove(master)
2203
      except ValueError:
2204
        pass
2205
      result = self.rpc.call_upload_file(node_list,
2206
                                         constants.SSH_KNOWN_HOSTS_FILE)
2207
      for to_node, to_result in result.iteritems():
2208
        msg = to_result.fail_msg
2209
        if msg:
2210
          msg = ("Copy of file %s to node %s failed: %s" %
2211
                 (constants.SSH_KNOWN_HOSTS_FILE, to_node, msg))
2212
          self.proc.LogWarning(msg)
2213

    
2214
    finally:
2215
      result = self.rpc.call_node_start_master(master, False, False)
2216
      msg = result.fail_msg
2217
      if msg:
2218
        self.LogWarning("Could not re-enable the master role on"
2219
                        " the master, please restart manually: %s", msg)
2220

    
2221

    
2222
def _RecursiveCheckIfLVMBased(disk):
2223
  """Check if the given disk or its children are lvm-based.
2224

2225
  @type disk: L{objects.Disk}
2226
  @param disk: the disk to check
2227
  @rtype: boolean
2228
  @return: boolean indicating whether a LD_LV dev_type was found or not
2229

2230
  """
2231
  if disk.children:
2232
    for chdisk in disk.children:
2233
      if _RecursiveCheckIfLVMBased(chdisk):
2234
        return True
2235
  return disk.dev_type == constants.LD_LV
2236

    
2237

    
2238
class LUSetClusterParams(LogicalUnit):
2239
  """Change the parameters of the cluster.
2240

2241
  """
2242
  HPATH = "cluster-modify"
2243
  HTYPE = constants.HTYPE_CLUSTER
2244
  _OP_REQP = []
2245
  REQ_BGL = False
2246

    
2247
  def CheckArguments(self):
2248
    """Check parameters
2249

2250
    """
2251
    for attr in ["candidate_pool_size",
2252
                 "uid_pool", "add_uids", "remove_uids"]:
2253
      if not hasattr(self.op, attr):
2254
        setattr(self.op, attr, None)
2255

    
2256
    if self.op.candidate_pool_size is not None:
2257
      try:
2258
        self.op.candidate_pool_size = int(self.op.candidate_pool_size)
2259
      except (ValueError, TypeError), err:
2260
        raise errors.OpPrereqError("Invalid candidate_pool_size value: %s" %
2261
                                   str(err), errors.ECODE_INVAL)
2262
      if self.op.candidate_pool_size < 1:
2263
        raise errors.OpPrereqError("At least one master candidate needed",
2264
                                   errors.ECODE_INVAL)
2265

    
2266
    _CheckBooleanOpField(self.op, "maintain_node_health")
2267

    
2268
    if self.op.uid_pool:
2269
      uidpool.CheckUidPool(self.op.uid_pool)
2270

    
2271
    if self.op.add_uids:
2272
      uidpool.CheckUidPool(self.op.add_uids)
2273

    
2274
    if self.op.remove_uids:
2275
      uidpool.CheckUidPool(self.op.remove_uids)
2276

    
2277
  def ExpandNames(self):
2278
    # FIXME: in the future maybe other cluster params won't require checking on
2279
    # all nodes to be modified.
2280
    self.needed_locks = {
2281
      locking.LEVEL_NODE: locking.ALL_SET,
2282
    }
2283
    self.share_locks[locking.LEVEL_NODE] = 1
2284

    
2285
  def BuildHooksEnv(self):
2286
    """Build hooks env.
2287

2288
    """
2289
    env = {
2290
      "OP_TARGET": self.cfg.GetClusterName(),
2291
      "NEW_VG_NAME": self.op.vg_name,
2292
      }
2293
    mn = self.cfg.GetMasterNode()
2294
    return env, [mn], [mn]
2295

    
2296
  def CheckPrereq(self):
2297
    """Check prerequisites.
2298

2299
    This checks whether the given params don't conflict and
2300
    if the given volume group is valid.
2301

2302
    """
2303
    if self.op.vg_name is not None and not self.op.vg_name:
2304
      instances = self.cfg.GetAllInstancesInfo().values()
2305
      for inst in instances:
2306
        for disk in inst.disks:
2307
          if _RecursiveCheckIfLVMBased(disk):
2308
            raise errors.OpPrereqError("Cannot disable lvm storage while"
2309
                                       " lvm-based instances exist",
2310
                                       errors.ECODE_INVAL)
2311

    
2312
    node_list = self.acquired_locks[locking.LEVEL_NODE]
2313

    
2314
    # if vg_name not None, checks given volume group on all nodes
2315
    if self.op.vg_name:
2316
      vglist = self.rpc.call_vg_list(node_list)
2317
      for node in node_list:
2318
        msg = vglist[node].fail_msg
2319
        if msg:
2320
          # ignoring down node
2321
          self.LogWarning("Error while gathering data on node %s"
2322
                          " (ignoring node): %s", node, msg)
2323
          continue
2324
        vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
2325
                                              self.op.vg_name,
2326
                                              constants.MIN_VG_SIZE)
2327
        if vgstatus:
2328
          raise errors.OpPrereqError("Error on node '%s': %s" %
2329
                                     (node, vgstatus), errors.ECODE_ENVIRON)
2330

    
2331
    self.cluster = cluster = self.cfg.GetClusterInfo()
2332
    # validate params changes
2333
    if self.op.beparams:
2334
      utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
2335
      self.new_beparams = objects.FillDict(
2336
        cluster.beparams[constants.PP_DEFAULT], self.op.beparams)
2337

    
2338
    if self.op.nicparams:
2339
      utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
2340
      self.new_nicparams = objects.FillDict(
2341
        cluster.nicparams[constants.PP_DEFAULT], self.op.nicparams)
2342
      objects.NIC.CheckParameterSyntax(self.new_nicparams)
2343
      nic_errors = []
2344

    
2345
      # check all instances for consistency
2346
      for instance in self.cfg.GetAllInstancesInfo().values():
2347
        for nic_idx, nic in enumerate(instance.nics):
2348
          params_copy = copy.deepcopy(nic.nicparams)
2349
          params_filled = objects.FillDict(self.new_nicparams, params_copy)
2350

    
2351
          # check parameter syntax
2352
          try:
2353
            objects.NIC.CheckParameterSyntax(params_filled)
2354
          except errors.ConfigurationError, err:
2355
            nic_errors.append("Instance %s, nic/%d: %s" %
2356
                              (instance.name, nic_idx, err))
2357

    
2358
          # if we're moving instances to routed, check that they have an ip
2359
          target_mode = params_filled[constants.NIC_MODE]
2360
          if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
2361
            nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
2362
                              (instance.name, nic_idx))
2363
      if nic_errors:
2364
        raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
2365
                                   "\n".join(nic_errors))
2366

    
2367
    # hypervisor list/parameters
2368
    self.new_hvparams = new_hvp = objects.FillDict(cluster.hvparams, {})
2369
    if self.op.hvparams:
2370
      if not isinstance(self.op.hvparams, dict):
2371
        raise errors.OpPrereqError("Invalid 'hvparams' parameter on input",
2372
                                   errors.ECODE_INVAL)
2373
      for hv_name, hv_dict in self.op.hvparams.items():
2374
        if hv_name not in self.new_hvparams:
2375
          self.new_hvparams[hv_name] = hv_dict
2376
        else:
2377
          self.new_hvparams[hv_name].update(hv_dict)
2378

    
2379
    # os hypervisor parameters
2380
    self.new_os_hvp = objects.FillDict(cluster.os_hvp, {})
2381
    if self.op.os_hvp:
2382
      if not isinstance(self.op.os_hvp, dict):
2383
        raise errors.OpPrereqError("Invalid 'os_hvp' parameter on input",
2384
                                   errors.ECODE_INVAL)
2385
      for os_name, hvs in self.op.os_hvp.items():
2386
        if not isinstance(hvs, dict):
2387
          raise errors.OpPrereqError(("Invalid 'os_hvp' parameter on"
2388
                                      " input"), errors.ECODE_INVAL)
2389
        if os_name not in self.new_os_hvp:
2390
          self.new_os_hvp[os_name] = hvs
2391
        else:
2392
          for hv_name, hv_dict in hvs.items():
2393
            if hv_name not in self.new_os_hvp[os_name]:
2394
              self.new_os_hvp[os_name][hv_name] = hv_dict
2395
            else:
2396
              self.new_os_hvp[os_name][hv_name].update(hv_dict)
2397

    
2398
    # changes to the hypervisor list
2399
    if self.op.enabled_hypervisors is not None:
2400
      self.hv_list = self.op.enabled_hypervisors
2401
      if not self.hv_list:
2402
        raise errors.OpPrereqError("Enabled hypervisors list must contain at"
2403
                                   " least one member",
2404
                                   errors.ECODE_INVAL)
2405
      invalid_hvs = set(self.hv_list) - constants.HYPER_TYPES
2406
      if invalid_hvs:
2407
        raise errors.OpPrereqError("Enabled hypervisors contains invalid"
2408
                                   " entries: %s" %
2409
                                   utils.CommaJoin(invalid_hvs),
2410
                                   errors.ECODE_INVAL)
2411
      for hv in self.hv_list:
2412
        # if the hypervisor doesn't already exist in the cluster
2413
        # hvparams, we initialize it to empty, and then (in both
2414
        # cases) we make sure to fill the defaults, as we might not
2415
        # have a complete defaults list if the hypervisor wasn't
2416
        # enabled before
2417
        if hv not in new_hvp:
2418
          new_hvp[hv] = {}
2419
        new_hvp[hv] = objects.FillDict(constants.HVC_DEFAULTS[hv], new_hvp[hv])
2420
        utils.ForceDictType(new_hvp[hv], constants.HVS_PARAMETER_TYPES)
2421
    else:
2422
      self.hv_list = cluster.enabled_hypervisors
2423

    
2424
    if self.op.hvparams or self.op.enabled_hypervisors is not None:
2425
      # either the enabled list has changed, or the parameters have, validate
2426
      for hv_name, hv_params in self.new_hvparams.items():
2427
        if ((self.op.hvparams and hv_name in self.op.hvparams) or
2428
            (self.op.enabled_hypervisors and
2429
             hv_name in self.op.enabled_hypervisors)):
2430
          # either this is a new hypervisor, or its parameters have changed
2431
          hv_class = hypervisor.GetHypervisor(hv_name)
2432
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2433
          hv_class.CheckParameterSyntax(hv_params)
2434
          _CheckHVParams(self, node_list, hv_name, hv_params)
2435

    
2436
    if self.op.os_hvp:
2437
      # no need to check any newly-enabled hypervisors, since the
2438
      # defaults have already been checked in the above code-block
2439
      for os_name, os_hvp in self.new_os_hvp.items():
2440
        for hv_name, hv_params in os_hvp.items():
2441
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2442
          # we need to fill in the new os_hvp on top of the actual hv_p
2443
          cluster_defaults = self.new_hvparams.get(hv_name, {})
2444
          new_osp = objects.FillDict(cluster_defaults, hv_params)
2445
          hv_class = hypervisor.GetHypervisor(hv_name)
2446
          hv_class.CheckParameterSyntax(new_osp)
2447
          _CheckHVParams(self, node_list, hv_name, new_osp)
2448

    
2449

    
2450
  def Exec(self, feedback_fn):
2451
    """Change the parameters of the cluster.
2452

2453
    """
2454
    if self.op.vg_name is not None:
2455
      new_volume = self.op.vg_name
2456
      if not new_volume:
2457
        new_volume = None
2458
      if new_volume != self.cfg.GetVGName():
2459
        self.cfg.SetVGName(new_volume)
2460
      else:
2461
        feedback_fn("Cluster LVM configuration already in desired"
2462
                    " state, not changing")
2463
    if self.op.hvparams:
2464
      self.cluster.hvparams = self.new_hvparams
2465
    if self.op.os_hvp:
2466
      self.cluster.os_hvp = self.new_os_hvp
2467
    if self.op.enabled_hypervisors is not None:
2468
      self.cluster.hvparams = self.new_hvparams
2469
      self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
2470
    if self.op.beparams:
2471
      self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
2472
    if self.op.nicparams:
2473
      self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
2474

    
2475
    if self.op.candidate_pool_size is not None:
2476
      self.cluster.candidate_pool_size = self.op.candidate_pool_size
2477
      # we need to update the pool size here, otherwise the save will fail
2478
      _AdjustCandidatePool(self, [])
2479

    
2480
    if self.op.maintain_node_health is not None:
2481
      self.cluster.maintain_node_health = self.op.maintain_node_health
2482

    
2483
    if self.op.add_uids is not None:
2484
      uidpool.AddToUidPool(self.cluster.uid_pool, self.op.add_uids)
2485

    
2486
    if self.op.remove_uids is not None:
2487
      uidpool.RemoveFromUidPool(self.cluster.uid_pool, self.op.remove_uids)
2488

    
2489
    if self.op.uid_pool is not None:
2490
      self.cluster.uid_pool = self.op.uid_pool
2491

    
2492
    self.cfg.Update(self.cluster, feedback_fn)
2493

    
2494

    
2495
def _RedistributeAncillaryFiles(lu, additional_nodes=None):
2496
  """Distribute additional files which are part of the cluster configuration.
2497

2498
  ConfigWriter takes care of distributing the config and ssconf files, but
2499
  there are more files which should be distributed to all nodes. This function
2500
  makes sure those are copied.
2501

2502
  @param lu: calling logical unit
2503
  @param additional_nodes: list of nodes not in the config to distribute to
2504

2505
  """
2506
  # 1. Gather target nodes
2507
  myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
2508
  dist_nodes = lu.cfg.GetOnlineNodeList()
2509
  if additional_nodes is not None:
2510
    dist_nodes.extend(additional_nodes)
2511
  if myself.name in dist_nodes:
2512
    dist_nodes.remove(myself.name)
2513

    
2514
  # 2. Gather files to distribute
2515
  dist_files = set([constants.ETC_HOSTS,
2516
                    constants.SSH_KNOWN_HOSTS_FILE,
2517
                    constants.RAPI_CERT_FILE,
2518
                    constants.RAPI_USERS_FILE,
2519
                    constants.CONFD_HMAC_KEY,
2520
                    constants.CLUSTER_DOMAIN_SECRET_FILE,
2521
                   ])
2522

    
2523
  enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
2524
  for hv_name in enabled_hypervisors:
2525
    hv_class = hypervisor.GetHypervisor(hv_name)
2526
    dist_files.update(hv_class.GetAncillaryFiles())
2527

    
2528
  # 3. Perform the files upload
2529
  for fname in dist_files:
2530
    if os.path.exists(fname):
2531
      result = lu.rpc.call_upload_file(dist_nodes, fname)
2532
      for to_node, to_result in result.items():
2533
        msg = to_result.fail_msg
2534
        if msg:
2535
          msg = ("Copy of file %s to node %s failed: %s" %
2536
                 (fname, to_node, msg))
2537
          lu.proc.LogWarning(msg)
2538

    
2539

    
2540
class LURedistributeConfig(NoHooksLU):
2541
  """Force the redistribution of cluster configuration.
2542

2543
  This is a very simple LU.
2544

2545
  """
2546
  _OP_REQP = []
2547
  REQ_BGL = False
2548

    
2549
  def ExpandNames(self):
2550
    self.needed_locks = {
2551
      locking.LEVEL_NODE: locking.ALL_SET,
2552
    }
2553
    self.share_locks[locking.LEVEL_NODE] = 1
2554

    
2555
  def CheckPrereq(self):
2556
    """Check prerequisites.
2557

2558
    """
2559

    
2560
  def Exec(self, feedback_fn):
2561
    """Redistribute the configuration.
2562

2563
    """
2564
    self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
2565
    _RedistributeAncillaryFiles(self)
2566

    
2567

    
2568
def _WaitForSync(lu, instance, oneshot=False):
2569
  """Sleep and poll for an instance's disk to sync.
2570

2571
  """
2572
  if not instance.disks:
2573
    return True
2574

    
2575
  if not oneshot:
2576
    lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
2577

    
2578
  node = instance.primary_node
2579

    
2580
  for dev in instance.disks:
2581
    lu.cfg.SetDiskID(dev, node)
2582

    
2583
  # TODO: Convert to utils.Retry
2584

    
2585
  retries = 0
2586
  degr_retries = 10 # in seconds, as we sleep 1 second each time
2587
  while True:
2588
    max_time = 0
2589
    done = True
2590
    cumul_degraded = False
2591
    rstats = lu.rpc.call_blockdev_getmirrorstatus(node, instance.disks)
2592
    msg = rstats.fail_msg
2593
    if msg:
2594
      lu.LogWarning("Can't get any data from node %s: %s", node, msg)
2595
      retries += 1
2596
      if retries >= 10:
2597
        raise errors.RemoteError("Can't contact node %s for mirror data,"
2598
                                 " aborting." % node)
2599
      time.sleep(6)
2600
      continue
2601
    rstats = rstats.payload
2602
    retries = 0
2603
    for i, mstat in enumerate(rstats):
2604
      if mstat is None:
2605
        lu.LogWarning("Can't compute data for node %s/%s",
2606
                           node, instance.disks[i].iv_name)
2607
        continue
2608

    
2609
      cumul_degraded = (cumul_degraded or
2610
                        (mstat.is_degraded and mstat.sync_percent is None))
2611
      if mstat.sync_percent is not None:
2612
        done = False
2613
        if mstat.estimated_time is not None:
2614
          rem_time = "%d estimated seconds remaining" % mstat.estimated_time
2615
          max_time = mstat.estimated_time
2616
        else:
2617
          rem_time = "no time estimate"
2618
        lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
2619
                        (instance.disks[i].iv_name, mstat.sync_percent,
2620
                         rem_time))
2621

    
2622
    # if we're done but degraded, let's do a few small retries, to
2623
    # make sure we see a stable and not transient situation; therefore
2624
    # we force restart of the loop
2625
    if (done or oneshot) and cumul_degraded and degr_retries > 0:
2626
      logging.info("Degraded disks found, %d retries left", degr_retries)
2627
      degr_retries -= 1
2628
      time.sleep(1)
2629
      continue
2630

    
2631
    if done or oneshot:
2632
      break
2633

    
2634
    time.sleep(min(60, max_time))
2635

    
2636
  if done:
2637
    lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
2638
  return not cumul_degraded
2639

    
2640

    
2641
def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
2642
  """Check that mirrors are not degraded.
2643

2644
  The ldisk parameter, if True, will change the test from the
2645
  is_degraded attribute (which represents overall non-ok status for
2646
  the device(s)) to the ldisk (representing the local storage status).
2647

2648
  """
2649
  lu.cfg.SetDiskID(dev, node)
2650

    
2651
  result = True
2652

    
2653
  if on_primary or dev.AssembleOnSecondary():
2654
    rstats = lu.rpc.call_blockdev_find(node, dev)
2655
    msg = rstats.fail_msg
2656
    if msg:
2657
      lu.LogWarning("Can't find disk on node %s: %s", node, msg)
2658
      result = False
2659
    elif not rstats.payload:
2660
      lu.LogWarning("Can't find disk on node %s", node)
2661
      result = False
2662
    else:
2663
      if ldisk:
2664
        result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
2665
      else:
2666
        result = result and not rstats.payload.is_degraded
2667

    
2668
  if dev.children:
2669
    for child in dev.children:
2670
      result = result and _CheckDiskConsistency(lu, child, node, on_primary)
2671

    
2672
  return result
2673

    
2674

    
2675
class LUDiagnoseOS(NoHooksLU):
2676
  """Logical unit for OS diagnose/query.
2677

2678
  """
2679
  _OP_REQP = ["output_fields", "names"]
2680
  REQ_BGL = False
2681
  _FIELDS_STATIC = utils.FieldSet()
2682
  _FIELDS_DYNAMIC = utils.FieldSet("name", "valid", "node_status", "variants")
2683
  # Fields that need calculation of global os validity
2684
  _FIELDS_NEEDVALID = frozenset(["valid", "variants"])
2685

    
2686
  def ExpandNames(self):
2687
    if self.op.names:
2688
      raise errors.OpPrereqError("Selective OS query not supported",
2689
                                 errors.ECODE_INVAL)
2690

    
2691
    _CheckOutputFields(static=self._FIELDS_STATIC,
2692
                       dynamic=self._FIELDS_DYNAMIC,
2693
                       selected=self.op.output_fields)
2694

    
2695
    # Lock all nodes, in shared mode
2696
    # Temporary removal of locks, should be reverted later
2697
    # TODO: reintroduce locks when they are lighter-weight
2698
    self.needed_locks = {}
2699
    #self.share_locks[locking.LEVEL_NODE] = 1
2700
    #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
2701

    
2702
  def CheckPrereq(self):
2703
    """Check prerequisites.
2704

2705
    """
2706

    
2707
  @staticmethod
2708
  def _DiagnoseByOS(rlist):
2709
    """Remaps a per-node return list into an a per-os per-node dictionary
2710

2711
    @param rlist: a map with node names as keys and OS objects as values
2712

2713
    @rtype: dict
2714
    @return: a dictionary with osnames as keys and as value another map, with
2715
        nodes as keys and tuples of (path, status, diagnose) as values, eg::
2716

2717
          {"debian-etch": {"node1": [(/usr/lib/..., True, ""),
2718
                                     (/srv/..., False, "invalid api")],
2719
                           "node2": [(/srv/..., True, "")]}
2720
          }
2721

2722
    """
2723
    all_os = {}
2724
    # we build here the list of nodes that didn't fail the RPC (at RPC
2725
    # level), so that nodes with a non-responding node daemon don't
2726
    # make all OSes invalid
2727
    good_nodes = [node_name for node_name in rlist
2728
                  if not rlist[node_name].fail_msg]
2729
    for node_name, nr in rlist.items():
2730
      if nr.fail_msg or not nr.payload:
2731
        continue
2732
      for name, path, status, diagnose, variants in nr.payload:
2733
        if name not in all_os:
2734
          # build a list of nodes for this os containing empty lists
2735
          # for each node in node_list
2736
          all_os[name] = {}
2737
          for nname in good_nodes:
2738
            all_os[name][nname] = []
2739
        all_os[name][node_name].append((path, status, diagnose, variants))
2740
    return all_os
2741

    
2742
  def Exec(self, feedback_fn):
2743
    """Compute the list of OSes.
2744

2745
    """
2746
    valid_nodes = [node for node in self.cfg.GetOnlineNodeList()]
2747
    node_data = self.rpc.call_os_diagnose(valid_nodes)
2748
    pol = self._DiagnoseByOS(node_data)
2749
    output = []
2750
    calc_valid = self._FIELDS_NEEDVALID.intersection(self.op.output_fields)
2751
    calc_variants = "variants" in self.op.output_fields
2752

    
2753
    for os_name, os_data in pol.items():
2754
      row = []
2755
      if calc_valid:
2756
        valid = True
2757
        variants = None
2758
        for osl in os_data.values():
2759
          valid = valid and osl and osl[0][1]
2760
          if not valid:
2761
            variants = None
2762
            break
2763
          if calc_variants:
2764
            node_variants = osl[0][3]
2765
            if variants is None:
2766
              variants = node_variants
2767
            else:
2768
              variants = [v for v in variants if v in node_variants]
2769

    
2770
      for field in self.op.output_fields:
2771
        if field == "name":
2772
          val = os_name
2773
        elif field == "valid":
2774
          val = valid
2775
        elif field == "node_status":
2776
          # this is just a copy of the dict
2777
          val = {}
2778
          for node_name, nos_list in os_data.items():
2779
            val[node_name] = nos_list
2780
        elif field == "variants":
2781
          val =  variants
2782
        else:
2783
          raise errors.ParameterError(field)
2784
        row.append(val)
2785
      output.append(row)
2786

    
2787
    return output
2788

    
2789

    
2790
class LURemoveNode(LogicalUnit):
2791
  """Logical unit for removing a node.
2792

2793
  """
2794
  HPATH = "node-remove"
2795
  HTYPE = constants.HTYPE_NODE
2796
  _OP_REQP = ["node_name"]
2797

    
2798
  def BuildHooksEnv(self):
2799
    """Build hooks env.
2800

2801
    This doesn't run on the target node in the pre phase as a failed
2802
    node would then be impossible to remove.
2803

2804
    """
2805
    env = {
2806
      "OP_TARGET": self.op.node_name,
2807
      "NODE_NAME": self.op.node_name,
2808
      }
2809
    all_nodes = self.cfg.GetNodeList()
2810
    try:
2811
      all_nodes.remove(self.op.node_name)
2812
    except ValueError:
2813
      logging.warning("Node %s which is about to be removed not found"
2814
                      " in the all nodes list", self.op.node_name)
2815
    return env, all_nodes, all_nodes
2816

    
2817
  def CheckPrereq(self):
2818
    """Check prerequisites.
2819

2820
    This checks:
2821
     - the node exists in the configuration
2822
     - it does not have primary or secondary instances
2823
     - it's not the master
2824

2825
    Any errors are signaled by raising errors.OpPrereqError.
2826

2827
    """
2828
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
2829
    node = self.cfg.GetNodeInfo(self.op.node_name)
2830
    assert node is not None
2831

    
2832
    instance_list = self.cfg.GetInstanceList()
2833

    
2834
    masternode = self.cfg.GetMasterNode()
2835
    if node.name == masternode:
2836
      raise errors.OpPrereqError("Node is the master node,"
2837
                                 " you need to failover first.",
2838
                                 errors.ECODE_INVAL)
2839

    
2840
    for instance_name in instance_list:
2841
      instance = self.cfg.GetInstanceInfo(instance_name)
2842
      if node.name in instance.all_nodes:
2843
        raise errors.OpPrereqError("Instance %s is still running on the node,"
2844
                                   " please remove first." % instance_name,
2845
                                   errors.ECODE_INVAL)
2846
    self.op.node_name = node.name
2847
    self.node = node
2848

    
2849
  def Exec(self, feedback_fn):
2850
    """Removes the node from the cluster.
2851

2852
    """
2853
    node = self.node
2854
    logging.info("Stopping the node daemon and removing configs from node %s",
2855
                 node.name)
2856

    
2857
    modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
2858

    
2859
    # Promote nodes to master candidate as needed
2860
    _AdjustCandidatePool(self, exceptions=[node.name])
2861
    self.context.RemoveNode(node.name)
2862

    
2863
    # Run post hooks on the node before it's removed
2864
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
2865
    try:
2866
      hm.RunPhase(constants.HOOKS_PHASE_POST, [node.name])
2867
    except:
2868
      # pylint: disable-msg=W0702
2869
      self.LogWarning("Errors occurred running hooks on %s" % node.name)
2870

    
2871
    result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
2872
    msg = result.fail_msg
2873
    if msg:
2874
      self.LogWarning("Errors encountered on the remote node while leaving"
2875
                      " the cluster: %s", msg)
2876

    
2877
    # Remove node from our /etc/hosts
2878
    if self.cfg.GetClusterInfo().modify_etc_hosts:
2879
      # FIXME: this should be done via an rpc call to node daemon
2880
      utils.RemoveHostFromEtcHosts(node.name)
2881
      _RedistributeAncillaryFiles(self)
2882

    
2883

    
2884
class LUQueryNodes(NoHooksLU):
2885
  """Logical unit for querying nodes.
2886

2887
  """
2888
  # pylint: disable-msg=W0142
2889
  _OP_REQP = ["output_fields", "names", "use_locking"]
2890
  REQ_BGL = False
2891

    
2892
  _SIMPLE_FIELDS = ["name", "serial_no", "ctime", "mtime", "uuid",
2893
                    "master_candidate", "offline", "drained"]
2894

    
2895
  _FIELDS_DYNAMIC = utils.FieldSet(
2896
    "dtotal", "dfree",
2897
    "mtotal", "mnode", "mfree",
2898
    "bootid",
2899
    "ctotal", "cnodes", "csockets",
2900
    )
2901

    
2902
  _FIELDS_STATIC = utils.FieldSet(*[
2903
    "pinst_cnt", "sinst_cnt",
2904
    "pinst_list", "sinst_list",
2905
    "pip", "sip", "tags",
2906
    "master",
2907
    "role"] + _SIMPLE_FIELDS
2908
    )
2909

    
2910
  def ExpandNames(self):
2911
    _CheckOutputFields(static=self._FIELDS_STATIC,
2912
                       dynamic=self._FIELDS_DYNAMIC,
2913
                       selected=self.op.output_fields)
2914

    
2915
    self.needed_locks = {}
2916
    self.share_locks[locking.LEVEL_NODE] = 1
2917

    
2918
    if self.op.names:
2919
      self.wanted = _GetWantedNodes(self, self.op.names)
2920
    else:
2921
      self.wanted = locking.ALL_SET
2922

    
2923
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
2924
    self.do_locking = self.do_node_query and self.op.use_locking
2925
    if self.do_locking:
2926
      # if we don't request only static fields, we need to lock the nodes
2927
      self.needed_locks[locking.LEVEL_NODE] = self.wanted
2928

    
2929
  def CheckPrereq(self):
2930
    """Check prerequisites.
2931

2932
    """
2933
    # The validation of the node list is done in the _GetWantedNodes,
2934
    # if non empty, and if empty, there's no validation to do
2935
    pass
2936

    
2937
  def Exec(self, feedback_fn):
2938
    """Computes the list of nodes and their attributes.
2939

2940
    """
2941
    all_info = self.cfg.GetAllNodesInfo()
2942
    if self.do_locking:
2943
      nodenames = self.acquired_locks[locking.LEVEL_NODE]
2944
    elif self.wanted != locking.ALL_SET:
2945
      nodenames = self.wanted
2946
      missing = set(nodenames).difference(all_info.keys())
2947
      if missing:
2948
        raise errors.OpExecError(
2949
          "Some nodes were removed before retrieving their data: %s" % missing)
2950
    else:
2951
      nodenames = all_info.keys()
2952

    
2953
    nodenames = utils.NiceSort(nodenames)
2954
    nodelist = [all_info[name] for name in nodenames]
2955

    
2956
    # begin data gathering
2957

    
2958
    if self.do_node_query:
2959
      live_data = {}
2960
      node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
2961
                                          self.cfg.GetHypervisorType())
2962
      for name in nodenames:
2963
        nodeinfo = node_data[name]
2964
        if not nodeinfo.fail_msg and nodeinfo.payload:
2965
          nodeinfo = nodeinfo.payload
2966
          fn = utils.TryConvert
2967
          live_data[name] = {
2968
            "mtotal": fn(int, nodeinfo.get('memory_total', None)),
2969
            "mnode": fn(int, nodeinfo.get('memory_dom0', None)),
2970
            "mfree": fn(int, nodeinfo.get('memory_free', None)),
2971
            "dtotal": fn(int, nodeinfo.get('vg_size', None)),
2972
            "dfree": fn(int, nodeinfo.get('vg_free', None)),
2973
            "ctotal": fn(int, nodeinfo.get('cpu_total', None)),
2974
            "bootid": nodeinfo.get('bootid', None),
2975
            "cnodes": fn(int, nodeinfo.get('cpu_nodes', None)),
2976
            "csockets": fn(int, nodeinfo.get('cpu_sockets', None)),
2977
            }
2978
        else:
2979
          live_data[name] = {}
2980
    else:
2981
      live_data = dict.fromkeys(nodenames, {})
2982

    
2983
    node_to_primary = dict([(name, set()) for name in nodenames])
2984
    node_to_secondary = dict([(name, set()) for name in nodenames])
2985

    
2986
    inst_fields = frozenset(("pinst_cnt", "pinst_list",
2987
                             "sinst_cnt", "sinst_list"))
2988
    if inst_fields & frozenset(self.op.output_fields):
2989
      inst_data = self.cfg.GetAllInstancesInfo()
2990

    
2991
      for inst in inst_data.values():
2992
        if inst.primary_node in node_to_primary:
2993
          node_to_primary[inst.primary_node].add(inst.name)
2994
        for secnode in inst.secondary_nodes:
2995
          if secnode in node_to_secondary:
2996
            node_to_secondary[secnode].add(inst.name)
2997

    
2998
    master_node = self.cfg.GetMasterNode()
2999

    
3000
    # end data gathering
3001

    
3002
    output = []
3003
    for node in nodelist:
3004
      node_output = []
3005
      for field in self.op.output_fields:
3006
        if field in self._SIMPLE_FIELDS:
3007
          val = getattr(node, field)
3008
        elif field == "pinst_list":
3009
          val = list(node_to_primary[node.name])
3010
        elif field == "sinst_list":
3011
          val = list(node_to_secondary[node.name])
3012
        elif field == "pinst_cnt":
3013
          val = len(node_to_primary[node.name])
3014
        elif field == "sinst_cnt":
3015
          val = len(node_to_secondary[node.name])
3016
        elif field == "pip":
3017
          val = node.primary_ip
3018
        elif field == "sip":
3019
          val = node.secondary_ip
3020
        elif field == "tags":
3021
          val = list(node.GetTags())
3022
        elif field == "master":
3023
          val = node.name == master_node
3024
        elif self._FIELDS_DYNAMIC.Matches(field):
3025
          val = live_data[node.name].get(field, None)
3026
        elif field == "role":
3027
          if node.name == master_node:
3028
            val = "M"
3029
          elif node.master_candidate:
3030
            val = "C"
3031
          elif node.drained:
3032
            val = "D"
3033
          elif node.offline:
3034
            val = "O"
3035
          else:
3036
            val = "R"
3037
        else:
3038
          raise errors.ParameterError(field)
3039
        node_output.append(val)
3040
      output.append(node_output)
3041

    
3042
    return output
3043

    
3044

    
3045
class LUQueryNodeVolumes(NoHooksLU):
3046
  """Logical unit for getting volumes on node(s).
3047

3048
  """
3049
  _OP_REQP = ["nodes", "output_fields"]
3050
  REQ_BGL = False
3051
  _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
3052
  _FIELDS_STATIC = utils.FieldSet("node")
3053

    
3054
  def ExpandNames(self):
3055
    _CheckOutputFields(static=self._FIELDS_STATIC,
3056
                       dynamic=self._FIELDS_DYNAMIC,
3057
                       selected=self.op.output_fields)
3058

    
3059
    self.needed_locks = {}
3060
    self.share_locks[locking.LEVEL_NODE] = 1
3061
    if not self.op.nodes:
3062
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3063
    else:
3064
      self.needed_locks[locking.LEVEL_NODE] = \
3065
        _GetWantedNodes(self, self.op.nodes)
3066

    
3067
  def CheckPrereq(self):
3068
    """Check prerequisites.
3069

3070
    This checks that the fields required are valid output fields.
3071

3072
    """
3073
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
3074

    
3075
  def Exec(self, feedback_fn):
3076
    """Computes the list of nodes and their attributes.
3077

3078
    """
3079
    nodenames = self.nodes
3080
    volumes = self.rpc.call_node_volumes(nodenames)
3081

    
3082
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
3083
             in self.cfg.GetInstanceList()]
3084

    
3085
    lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
3086

    
3087
    output = []
3088
    for node in nodenames:
3089
      nresult = volumes[node]
3090
      if nresult.offline:
3091
        continue
3092
      msg = nresult.fail_msg
3093
      if msg:
3094
        self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
3095
        continue
3096

    
3097
      node_vols = nresult.payload[:]
3098
      node_vols.sort(key=lambda vol: vol['dev'])
3099

    
3100
      for vol in node_vols:
3101
        node_output = []
3102
        for field in self.op.output_fields:
3103
          if field == "node":
3104
            val = node
3105
          elif field == "phys":
3106
            val = vol['dev']
3107
          elif field == "vg":
3108
            val = vol['vg']
3109
          elif field == "name":
3110
            val = vol['name']
3111
          elif field == "size":
3112
            val = int(float(vol['size']))
3113
          elif field == "instance":
3114
            for inst in ilist:
3115
              if node not in lv_by_node[inst]:
3116
                continue
3117
              if vol['name'] in lv_by_node[inst][node]:
3118
                val = inst.name
3119
                break
3120
            else:
3121
              val = '-'
3122
          else:
3123
            raise errors.ParameterError(field)
3124
          node_output.append(str(val))
3125

    
3126
        output.append(node_output)
3127

    
3128
    return output
3129

    
3130

    
3131
class LUQueryNodeStorage(NoHooksLU):
3132
  """Logical unit for getting information on storage units on node(s).
3133

3134
  """
3135
  _OP_REQP = ["nodes", "storage_type", "output_fields"]
3136
  REQ_BGL = False
3137
  _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
3138

    
3139
  def CheckArguments(self):
3140
    _CheckStorageType(self.op.storage_type)
3141

    
3142
    _CheckOutputFields(static=self._FIELDS_STATIC,
3143
                       dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
3144
                       selected=self.op.output_fields)
3145

    
3146
  def ExpandNames(self):
3147
    self.needed_locks = {}
3148
    self.share_locks[locking.LEVEL_NODE] = 1
3149

    
3150
    if self.op.nodes:
3151
      self.needed_locks[locking.LEVEL_NODE] = \
3152
        _GetWantedNodes(self, self.op.nodes)
3153
    else:
3154
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3155

    
3156
  def CheckPrereq(self):
3157
    """Check prerequisites.
3158

3159
    This checks that the fields required are valid output fields.
3160

3161
    """
3162
    self.op.name = getattr(self.op, "name", None)
3163

    
3164
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
3165

    
3166
  def Exec(self, feedback_fn):
3167
    """Computes the list of nodes and their attributes.
3168

3169
    """
3170
    # Always get name to sort by
3171
    if constants.SF_NAME in self.op.output_fields:
3172
      fields = self.op.output_fields[:]
3173
    else:
3174
      fields = [constants.SF_NAME] + self.op.output_fields
3175

    
3176
    # Never ask for node or type as it's only known to the LU
3177
    for extra in [constants.SF_NODE, constants.SF_TYPE]:
3178
      while extra in fields:
3179
        fields.remove(extra)
3180

    
3181
    field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
3182
    name_idx = field_idx[constants.SF_NAME]
3183

    
3184
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3185
    data = self.rpc.call_storage_list(self.nodes,
3186
                                      self.op.storage_type, st_args,
3187
                                      self.op.name, fields)
3188

    
3189
    result = []
3190

    
3191
    for node in utils.NiceSort(self.nodes):
3192
      nresult = data[node]
3193
      if nresult.offline:
3194
        continue
3195

    
3196
      msg = nresult.fail_msg
3197
      if msg:
3198
        self.LogWarning("Can't get storage data from node %s: %s", node, msg)
3199
        continue
3200

    
3201
      rows = dict([(row[name_idx], row) for row in nresult.payload])
3202

    
3203
      for name in utils.NiceSort(rows.keys()):
3204
        row = rows[name]
3205

    
3206
        out = []
3207

    
3208
        for field in self.op.output_fields:
3209
          if field == constants.SF_NODE:
3210
            val = node
3211
          elif field == constants.SF_TYPE:
3212
            val = self.op.storage_type
3213
          elif field in field_idx:
3214
            val = row[field_idx[field]]
3215
          else:
3216
            raise errors.ParameterError(field)
3217

    
3218
          out.append(val)
3219

    
3220
        result.append(out)
3221

    
3222
    return result
3223

    
3224

    
3225
class LUModifyNodeStorage(NoHooksLU):
3226
  """Logical unit for modifying a storage volume on a node.
3227

3228
  """
3229
  _OP_REQP = ["node_name", "storage_type", "name", "changes"]
3230
  REQ_BGL = False
3231

    
3232
  def CheckArguments(self):
3233
    self.opnode_name = _ExpandNodeName(self.cfg, self.op.node_name)
3234

    
3235
    _CheckStorageType(self.op.storage_type)
3236

    
3237
  def ExpandNames(self):
3238
    self.needed_locks = {
3239
      locking.LEVEL_NODE: self.op.node_name,
3240
      }
3241

    
3242
  def CheckPrereq(self):
3243
    """Check prerequisites.
3244

3245
    """
3246
    storage_type = self.op.storage_type
3247

    
3248
    try:
3249
      modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
3250
    except KeyError:
3251
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
3252
                                 " modified" % storage_type,
3253
                                 errors.ECODE_INVAL)
3254

    
3255
    diff = set(self.op.changes.keys()) - modifiable
3256
    if diff:
3257
      raise errors.OpPrereqError("The following fields can not be modified for"
3258
                                 " storage units of type '%s': %r" %
3259
                                 (storage_type, list(diff)),
3260
                                 errors.ECODE_INVAL)
3261

    
3262
  def Exec(self, feedback_fn):
3263
    """Computes the list of nodes and their attributes.
3264

3265
    """
3266
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
3267
    result = self.rpc.call_storage_modify(self.op.node_name,
3268
                                          self.op.storage_type, st_args,
3269
                                          self.op.name, self.op.changes)
3270
    result.Raise("Failed to modify storage unit '%s' on %s" %
3271
                 (self.op.name, self.op.node_name))
3272

    
3273

    
3274
class LUAddNode(LogicalUnit):
3275
  """Logical unit for adding node to the cluster.
3276

3277
  """
3278
  HPATH = "node-add"
3279
  HTYPE = constants.HTYPE_NODE
3280
  _OP_REQP = ["node_name"]
3281

    
3282
  def CheckArguments(self):
3283
    # validate/normalize the node name
3284
    self.op.node_name = utils.HostInfo.NormalizeName(self.op.node_name)
3285

    
3286
  def BuildHooksEnv(self):
3287
    """Build hooks env.
3288

3289
    This will run on all nodes before, and on all nodes + the new node after.
3290

3291
    """
3292
    env = {
3293
      "OP_TARGET": self.op.node_name,
3294
      "NODE_NAME": self.op.node_name,
3295
      "NODE_PIP": self.op.primary_ip,
3296
      "NODE_SIP": self.op.secondary_ip,
3297
      }
3298
    nodes_0 = self.cfg.GetNodeList()
3299
    nodes_1 = nodes_0 + [self.op.node_name, ]
3300
    return env, nodes_0, nodes_1
3301

    
3302
  def CheckPrereq(self):
3303
    """Check prerequisites.
3304

3305
    This checks:
3306
     - the new node is not already in the config
3307
     - it is resolvable
3308
     - its parameters (single/dual homed) matches the cluster
3309

3310
    Any errors are signaled by raising errors.OpPrereqError.
3311

3312
    """
3313
    node_name = self.op.node_name
3314
    cfg = self.cfg
3315

    
3316
    dns_data = utils.GetHostInfo(node_name)
3317

    
3318
    node = dns_data.name
3319
    primary_ip = self.op.primary_ip = dns_data.ip
3320
    secondary_ip = getattr(self.op, "secondary_ip", None)
3321
    if secondary_ip is None:
3322
      secondary_ip = primary_ip
3323
    if not utils.IsValidIP(secondary_ip):
3324
      raise errors.OpPrereqError("Invalid secondary IP given",
3325
                                 errors.ECODE_INVAL)
3326
    self.op.secondary_ip = secondary_ip
3327

    
3328
    node_list = cfg.GetNodeList()
3329
    if not self.op.readd and node in node_list:
3330
      raise errors.OpPrereqError("Node %s is already in the configuration" %
3331
                                 node, errors.ECODE_EXISTS)
3332
    elif self.op.readd and node not in node_list:
3333
      raise errors.OpPrereqError("Node %s is not in the configuration" % node,
3334
                                 errors.ECODE_NOENT)
3335

    
3336
    self.changed_primary_ip = False
3337

    
3338
    for existing_node_name in node_list:
3339
      existing_node = cfg.GetNodeInfo(existing_node_name)
3340

    
3341
      if self.op.readd and node == existing_node_name:
3342
        if existing_node.secondary_ip != secondary_ip:
3343
          raise errors.OpPrereqError("Readded node doesn't have the same IP"
3344
                                     " address configuration as before",
3345
                                     errors.ECODE_INVAL)
3346
        if existing_node.primary_ip != primary_ip:
3347
          self.changed_primary_ip = True
3348

    
3349
        continue
3350

    
3351
      if (existing_node.primary_ip == primary_ip or
3352
          existing_node.secondary_ip == primary_ip or
3353
          existing_node.primary_ip == secondary_ip or
3354
          existing_node.secondary_ip == secondary_ip):
3355
        raise errors.OpPrereqError("New node ip address(es) conflict with"
3356
                                   " existing node %s" % existing_node.name,
3357
                                   errors.ECODE_NOTUNIQUE)
3358

    
3359
    # check that the type of the node (single versus dual homed) is the
3360
    # same as for the master
3361
    myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
3362
    master_singlehomed = myself.secondary_ip == myself.primary_ip
3363
    newbie_singlehomed = secondary_ip == primary_ip
3364
    if master_singlehomed != newbie_singlehomed:
3365
      if master_singlehomed:
3366
        raise errors.OpPrereqError("The master has no private ip but the"
3367
                                   " new node has one",
3368
                                   errors.ECODE_INVAL)
3369
      else:
3370
        raise errors.OpPrereqError("The master has a private ip but the"
3371
                                   " new node doesn't have one",
3372
                                   errors.ECODE_INVAL)
3373

    
3374
    # checks reachability
3375
    if not utils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
3376
      raise errors.OpPrereqError("Node not reachable by ping",
3377
                                 errors.ECODE_ENVIRON)
3378

    
3379
    if not newbie_singlehomed:
3380
      # check reachability from my secondary ip to newbie's secondary ip
3381
      if not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
3382
                           source=myself.secondary_ip):
3383
        raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
3384
                                   " based ping to noded port",
3385
                                   errors.ECODE_ENVIRON)
3386

    
3387
    if self.op.readd:
3388
      exceptions = [node]
3389
    else:
3390
      exceptions = []
3391

    
3392
    self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
3393

    
3394
    if self.op.readd:
3395
      self.new_node = self.cfg.GetNodeInfo(node)
3396
      assert self.new_node is not None, "Can't retrieve locked node %s" % node
3397
    else:
3398
      self.new_node = objects.Node(name=node,
3399
                                   primary_ip=primary_ip,
3400
                                   secondary_ip=secondary_ip,
3401
                                   master_candidate=self.master_candidate,
3402
                                   offline=False, drained=False)
3403

    
3404
  def Exec(self, feedback_fn):
3405
    """Adds the new node to the cluster.
3406

3407
    """
3408
    new_node = self.new_node
3409
    node = new_node.name
3410

    
3411
    # for re-adds, reset the offline/drained/master-candidate flags;
3412
    # we need to reset here, otherwise offline would prevent RPC calls
3413
    # later in the procedure; this also means that if the re-add
3414
    # fails, we are left with a non-offlined, broken node
3415
    if self.op.readd:
3416
      new_node.drained = new_node.offline = False # pylint: disable-msg=W0201
3417
      self.LogInfo("Readding a node, the offline/drained flags were reset")
3418
      # if we demote the node, we do cleanup later in the procedure
3419
      new_node.master_candidate = self.master_candidate
3420
      if self.changed_primary_ip:
3421
        new_node.primary_ip = self.op.primary_ip
3422

    
3423
    # notify the user about any possible mc promotion
3424
    if new_node.master_candidate:
3425
      self.LogInfo("Node will be a master candidate")
3426

    
3427
    # check connectivity
3428
    result = self.rpc.call_version([node])[node]
3429
    result.Raise("Can't get version information from node %s" % node)
3430
    if constants.PROTOCOL_VERSION == result.payload:
3431
      logging.info("Communication to node %s fine, sw version %s match",
3432
                   node, result.payload)
3433
    else:
3434
      raise errors.OpExecError("Version mismatch master version %s,"
3435
                               " node version %s" %
3436
                               (constants.PROTOCOL_VERSION, result.payload))
3437

    
3438
    # setup ssh on node
3439
    if self.cfg.GetClusterInfo().modify_ssh_setup:
3440
      logging.info("Copy ssh key to node %s", node)
3441
      priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
3442
      keyarray = []
3443
      keyfiles = [constants.SSH_HOST_DSA_PRIV, constants.SSH_HOST_DSA_PUB,
3444
                  constants.SSH_HOST_RSA_PRIV, constants.SSH_HOST_RSA_PUB,
3445
                  priv_key, pub_key]
3446

    
3447
      for i in keyfiles:
3448
        keyarray.append(utils.ReadFile(i))
3449

    
3450
      result = self.rpc.call_node_add(node, keyarray[0], keyarray[1],
3451
                                      keyarray[2], keyarray[3], keyarray[4],
3452
                                      keyarray[5])
3453
      result.Raise("Cannot transfer ssh keys to the new node")
3454

    
3455
    # Add node to our /etc/hosts, and add key to known_hosts
3456
    if self.cfg.GetClusterInfo().modify_etc_hosts:
3457
      # FIXME: this should be done via an rpc call to node daemon
3458
      utils.AddHostToEtcHosts(new_node.name)
3459

    
3460
    if new_node.secondary_ip != new_node.primary_ip:
3461
      result = self.rpc.call_node_has_ip_address(new_node.name,
3462
                                                 new_node.secondary_ip)
3463
      result.Raise("Failure checking secondary ip on node %s" % new_node.name,
3464
                   prereq=True, ecode=errors.ECODE_ENVIRON)
3465
      if not result.payload:
3466
        raise errors.OpExecError("Node claims it doesn't have the secondary ip"
3467
                                 " you gave (%s). Please fix and re-run this"
3468
                                 " command." % new_node.secondary_ip)
3469

    
3470
    node_verify_list = [self.cfg.GetMasterNode()]
3471
    node_verify_param = {
3472
      constants.NV_NODELIST: [node],
3473
      # TODO: do a node-net-test as well?
3474
    }
3475

    
3476
    result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
3477
                                       self.cfg.GetClusterName())
3478
    for verifier in node_verify_list:
3479
      result[verifier].Raise("Cannot communicate with node %s" % verifier)
3480
      nl_payload = result[verifier].payload[constants.NV_NODELIST]
3481
      if nl_payload:
3482
        for failed in nl_payload:
3483
          feedback_fn("ssh/hostname verification failed"
3484
                      " (checking from %s): %s" %
3485
                      (verifier, nl_payload[failed]))
3486
        raise errors.OpExecError("ssh/hostname verification failed.")
3487

    
3488
    if self.op.readd:
3489
      _RedistributeAncillaryFiles(self)
3490
      self.context.ReaddNode(new_node)
3491
      # make sure we redistribute the config
3492
      self.cfg.Update(new_node, feedback_fn)
3493
      # and make sure the new node will not have old files around
3494
      if not new_node.master_candidate:
3495
        result = self.rpc.call_node_demote_from_mc(new_node.name)
3496
        msg = result.fail_msg
3497
        if msg:
3498
          self.LogWarning("Node failed to demote itself from master"
3499
                          " candidate status: %s" % msg)
3500
    else:
3501
      _RedistributeAncillaryFiles(self, additional_nodes=[node])
3502
      self.context.AddNode(new_node, self.proc.GetECId())
3503

    
3504

    
3505
class LUSetNodeParams(LogicalUnit):
3506
  """Modifies the parameters of a node.
3507

3508
  """
3509
  HPATH = "node-modify"
3510
  HTYPE = constants.HTYPE_NODE
3511
  _OP_REQP = ["node_name"]
3512
  REQ_BGL = False
3513

    
3514
  def CheckArguments(self):
3515
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3516
    _CheckBooleanOpField(self.op, 'master_candidate')
3517
    _CheckBooleanOpField(self.op, 'offline')
3518
    _CheckBooleanOpField(self.op, 'drained')
3519
    _CheckBooleanOpField(self.op, 'auto_promote')
3520
    all_mods = [self.op.offline, self.op.master_candidate, self.op.drained]
3521
    if all_mods.count(None) == 3:
3522
      raise errors.OpPrereqError("Please pass at least one modification",
3523
                                 errors.ECODE_INVAL)
3524
    if all_mods.count(True) > 1:
3525
      raise errors.OpPrereqError("Can't set the node into more than one"
3526
                                 " state at the same time",
3527
                                 errors.ECODE_INVAL)
3528

    
3529
    # Boolean value that tells us whether we're offlining or draining the node
3530
    self.offline_or_drain = (self.op.offline == True or
3531
                             self.op.drained == True)
3532
    self.deoffline_or_drain = (self.op.offline == False or
3533
                               self.op.drained == False)
3534
    self.might_demote = (self.op.master_candidate == False or
3535
                         self.offline_or_drain)
3536

    
3537
    self.lock_all = self.op.auto_promote and self.might_demote
3538

    
3539

    
3540
  def ExpandNames(self):
3541
    if self.lock_all:
3542
      self.needed_locks = {locking.LEVEL_NODE: locking.ALL_SET}
3543
    else:
3544
      self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
3545

    
3546
  def BuildHooksEnv(self):
3547
    """Build hooks env.
3548

3549
    This runs on the master node.
3550

3551
    """
3552
    env = {
3553
      "OP_TARGET": self.op.node_name,
3554
      "MASTER_CANDIDATE": str(self.op.master_candidate),
3555
      "OFFLINE": str(self.op.offline),
3556
      "DRAINED": str(self.op.drained),
3557
      }
3558
    nl = [self.cfg.GetMasterNode(),
3559
          self.op.node_name]
3560
    return env, nl, nl
3561

    
3562
  def CheckPrereq(self):
3563
    """Check prerequisites.
3564

3565
    This only checks the instance list against the existing names.
3566

3567
    """
3568
    node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
3569

    
3570
    if (self.op.master_candidate is not None or
3571
        self.op.drained is not None or
3572
        self.op.offline is not None):
3573
      # we can't change the master's node flags
3574
      if self.op.node_name == self.cfg.GetMasterNode():
3575
        raise errors.OpPrereqError("The master role can be changed"
3576
                                   " only via masterfailover",
3577
                                   errors.ECODE_INVAL)
3578

    
3579

    
3580
    if node.master_candidate and self.might_demote and not self.lock_all:
3581
      assert not self.op.auto_promote, "auto-promote set but lock_all not"
3582
      # check if after removing the current node, we're missing master
3583
      # candidates
3584
      (mc_remaining, mc_should, _) = \
3585
          self.cfg.GetMasterCandidateStats(exceptions=[node.name])
3586
      if mc_remaining < mc_should:
3587
        raise errors.OpPrereqError("Not enough master candidates, please"
3588
                                   " pass auto_promote to allow promotion",
3589
                                   errors.ECODE_INVAL)
3590

    
3591
    if (self.op.master_candidate == True and
3592
        ((node.offline and not self.op.offline == False) or
3593
         (node.drained and not self.op.drained == False))):
3594
      raise errors.OpPrereqError("Node '%s' is offline or drained, can't set"
3595
                                 " to master_candidate" % node.name,
3596
                                 errors.ECODE_INVAL)
3597

    
3598
    # If we're being deofflined/drained, we'll MC ourself if needed
3599
    if (self.deoffline_or_drain and not self.offline_or_drain and not
3600
        self.op.master_candidate == True and not node.master_candidate):
3601
      self.op.master_candidate = _DecideSelfPromotion(self)
3602
      if self.op.master_candidate:
3603
        self.LogInfo("Autopromoting node to master candidate")
3604

    
3605
    return
3606

    
3607
  def Exec(self, feedback_fn):
3608
    """Modifies a node.
3609

3610
    """
3611
    node = self.node
3612

    
3613
    result = []
3614
    changed_mc = False
3615

    
3616
    if self.op.offline is not None:
3617
      node.offline = self.op.offline
3618
      result.append(("offline", str(self.op.offline)))
3619
      if self.op.offline == True:
3620
        if node.master_candidate:
3621
          node.master_candidate = False
3622
          changed_mc = True
3623
          result.append(("master_candidate", "auto-demotion due to offline"))
3624
        if node.drained:
3625
          node.drained = False
3626
          result.append(("drained", "clear drained status due to offline"))
3627

    
3628
    if self.op.master_candidate is not None:
3629
      node.master_candidate = self.op.master_candidate
3630
      changed_mc = True
3631
      result.append(("master_candidate", str(self.op.master_candidate)))
3632
      if self.op.master_candidate == False:
3633
        rrc = self.rpc.call_node_demote_from_mc(node.name)
3634
        msg = rrc.fail_msg
3635
        if msg:
3636
          self.LogWarning("Node failed to demote itself: %s" % msg)
3637

    
3638
    if self.op.drained is not None:
3639
      node.drained = self.op.drained
3640
      result.append(("drained", str(self.op.drained)))
3641
      if self.op.drained == True:
3642
        if node.master_candidate:
3643
          node.master_candidate = False
3644
          changed_mc = True
3645
          result.append(("master_candidate", "auto-demotion due to drain"))
3646
          rrc = self.rpc.call_node_demote_from_mc(node.name)
3647
          msg = rrc.fail_msg
3648
          if msg:
3649
            self.LogWarning("Node failed to demote itself: %s" % msg)
3650
        if node.offline:
3651
          node.offline = False
3652
          result.append(("offline", "clear offline status due to drain"))
3653

    
3654
    # we locked all nodes, we adjust the CP before updating this node
3655
    if self.lock_all:
3656
      _AdjustCandidatePool(self, [node.name])
3657

    
3658
    # this will trigger configuration file update, if needed
3659
    self.cfg.Update(node, feedback_fn)
3660

    
3661
    # this will trigger job queue propagation or cleanup
3662
    if changed_mc:
3663
      self.context.ReaddNode(node)
3664

    
3665
    return result
3666

    
3667

    
3668
class LUPowercycleNode(NoHooksLU):
3669
  """Powercycles a node.
3670

3671
  """
3672
  _OP_REQP = ["node_name", "force"]
3673
  REQ_BGL = False
3674

    
3675
  def CheckArguments(self):
3676
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
3677
    if self.op.node_name == self.cfg.GetMasterNode() and not self.op.force:
3678
      raise errors.OpPrereqError("The node is the master and the force"
3679
                                 " parameter was not set",
3680
                                 errors.ECODE_INVAL)
3681

    
3682
  def ExpandNames(self):
3683
    """Locking for PowercycleNode.
3684

3685
    This is a last-resort option and shouldn't block on other
3686
    jobs. Therefore, we grab no locks.
3687

3688
    """
3689
    self.needed_locks = {}
3690

    
3691
  def CheckPrereq(self):
3692
    """Check prerequisites.
3693

3694
    This LU has no prereqs.
3695

3696
    """
3697
    pass
3698

    
3699
  def Exec(self, feedback_fn):
3700
    """Reboots a node.
3701

3702
    """
3703
    result = self.rpc.call_node_powercycle(self.op.node_name,
3704
                                           self.cfg.GetHypervisorType())
3705
    result.Raise("Failed to schedule the reboot")
3706
    return result.payload
3707

    
3708

    
3709
class LUQueryClusterInfo(NoHooksLU):
3710
  """Query cluster configuration.
3711

3712
  """
3713
  _OP_REQP = []
3714
  REQ_BGL = False
3715

    
3716
  def ExpandNames(self):
3717
    self.needed_locks = {}
3718

    
3719
  def CheckPrereq(self):
3720
    """No prerequsites needed for this LU.
3721

3722
    """
3723
    pass
3724

    
3725
  def Exec(self, feedback_fn):
3726
    """Return cluster config.
3727

3728
    """
3729
    cluster = self.cfg.GetClusterInfo()
3730
    os_hvp = {}
3731

    
3732
    # Filter just for enabled hypervisors
3733
    for os_name, hv_dict in cluster.os_hvp.items():
3734
      os_hvp[os_name] = {}
3735
      for hv_name, hv_params in hv_dict.items():
3736
        if hv_name in cluster.enabled_hypervisors:
3737
          os_hvp[os_name][hv_name] = hv_params
3738

    
3739
    result = {
3740
      "software_version": constants.RELEASE_VERSION,
3741
      "protocol_version": constants.PROTOCOL_VERSION,
3742
      "config_version": constants.CONFIG_VERSION,
3743
      "os_api_version": max(constants.OS_API_VERSIONS),
3744
      "export_version": constants.EXPORT_VERSION,
3745
      "architecture": (platform.architecture()[0], platform.machine()),
3746
      "name": cluster.cluster_name,
3747
      "master": cluster.master_node,
3748
      "default_hypervisor": cluster.enabled_hypervisors[0],
3749
      "enabled_hypervisors": cluster.enabled_hypervisors,
3750
      "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
3751
                        for hypervisor_name in cluster.enabled_hypervisors]),
3752
      "os_hvp": os_hvp,
3753
      "beparams": cluster.beparams,
3754
      "nicparams": cluster.nicparams,
3755
      "candidate_pool_size": cluster.candidate_pool_size,
3756
      "master_netdev": cluster.master_netdev,
3757
      "volume_group_name": cluster.volume_group_name,
3758
      "file_storage_dir": cluster.file_storage_dir,
3759
      "maintain_node_health": cluster.maintain_node_health,
3760
      "ctime": cluster.ctime,
3761
      "mtime": cluster.mtime,
3762
      "uuid": cluster.uuid,
3763
      "tags": list(cluster.GetTags()),
3764
      "uid_pool": cluster.uid_pool,
3765
      }
3766

    
3767
    return result
3768

    
3769

    
3770
class LUQueryConfigValues(NoHooksLU):
3771
  """Return configuration values.
3772

3773
  """
3774
  _OP_REQP = []
3775
  REQ_BGL = False
3776
  _FIELDS_DYNAMIC = utils.FieldSet()
3777
  _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag",
3778
                                  "watcher_pause")
3779

    
3780
  def ExpandNames(self):
3781
    self.needed_locks = {}
3782

    
3783
    _CheckOutputFields(static=self._FIELDS_STATIC,
3784
                       dynamic=self._FIELDS_DYNAMIC,
3785
                       selected=self.op.output_fields)
3786

    
3787
  def CheckPrereq(self):
3788
    """No prerequisites.
3789

3790
    """
3791
    pass
3792

    
3793
  def Exec(self, feedback_fn):
3794
    """Dump a representation of the cluster config to the standard output.
3795

3796
    """
3797
    values = []
3798
    for field in self.op.output_fields:
3799
      if field == "cluster_name":
3800
        entry = self.cfg.GetClusterName()
3801
      elif field == "master_node":
3802
        entry = self.cfg.GetMasterNode()
3803
      elif field == "drain_flag":
3804
        entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
3805
      elif field == "watcher_pause":
3806
        entry = utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
3807
      else:
3808
        raise errors.ParameterError(field)
3809
      values.append(entry)
3810
    return values
3811

    
3812

    
3813
class LUActivateInstanceDisks(NoHooksLU):
3814
  """Bring up an instance's disks.
3815

3816
  """
3817
  _OP_REQP = ["instance_name"]
3818
  REQ_BGL = False
3819

    
3820
  def ExpandNames(self):
3821
    self._ExpandAndLockInstance()
3822
    self.needed_locks[locking.LEVEL_NODE] = []
3823
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3824

    
3825
  def DeclareLocks(self, level):
3826
    if level == locking.LEVEL_NODE:
3827
      self._LockInstancesNodes()
3828

    
3829
  def CheckPrereq(self):
3830
    """Check prerequisites.
3831

3832
    This checks that the instance is in the cluster.
3833

3834
    """
3835
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3836
    assert self.instance is not None, \
3837
      "Cannot retrieve locked instance %s" % self.op.instance_name
3838
    _CheckNodeOnline(self, self.instance.primary_node)
3839
    if not hasattr(self.op, "ignore_size"):
3840
      self.op.ignore_size = False
3841

    
3842
  def Exec(self, feedback_fn):
3843
    """Activate the disks.
3844

3845
    """
3846
    disks_ok, disks_info = \
3847
              _AssembleInstanceDisks(self, self.instance,
3848
                                     ignore_size=self.op.ignore_size)
3849
    if not disks_ok:
3850
      raise errors.OpExecError("Cannot activate block devices")
3851

    
3852
    return disks_info
3853

    
3854

    
3855
def _AssembleInstanceDisks(lu, instance, ignore_secondaries=False,
3856
                           ignore_size=False):
3857
  """Prepare the block devices for an instance.
3858

3859
  This sets up the block devices on all nodes.
3860

3861
  @type lu: L{LogicalUnit}
3862
  @param lu: the logical unit on whose behalf we execute
3863
  @type instance: L{objects.Instance}
3864
  @param instance: the instance for whose disks we assemble
3865
  @type ignore_secondaries: boolean
3866
  @param ignore_secondaries: if true, errors on secondary nodes
3867
      won't result in an error return from the function
3868
  @type ignore_size: boolean
3869
  @param ignore_size: if true, the current known size of the disk
3870
      will not be used during the disk activation, useful for cases
3871
      when the size is wrong
3872
  @return: False if the operation failed, otherwise a list of
3873
      (host, instance_visible_name, node_visible_name)
3874
      with the mapping from node devices to instance devices
3875

3876
  """
3877
  device_info = []
3878
  disks_ok = True
3879
  iname = instance.name
3880
  # With the two passes mechanism we try to reduce the window of
3881
  # opportunity for the race condition of switching DRBD to primary
3882
  # before handshaking occured, but we do not eliminate it
3883

    
3884
  # The proper fix would be to wait (with some limits) until the
3885
  # connection has been made and drbd transitions from WFConnection
3886
  # into any other network-connected state (Connected, SyncTarget,
3887
  # SyncSource, etc.)
3888

    
3889
  # 1st pass, assemble on all nodes in secondary mode
3890
  for inst_disk in instance.disks:
3891
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
3892
      if ignore_size:
3893
        node_disk = node_disk.Copy()
3894
        node_disk.UnsetSize()
3895
      lu.cfg.SetDiskID(node_disk, node)
3896
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
3897
      msg = result.fail_msg
3898
      if msg:
3899
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
3900
                           " (is_primary=False, pass=1): %s",
3901
                           inst_disk.iv_name, node, msg)
3902
        if not ignore_secondaries:
3903
          disks_ok = False
3904

    
3905
  # FIXME: race condition on drbd migration to primary
3906

    
3907
  # 2nd pass, do only the primary node
3908
  for inst_disk in instance.disks:
3909
    dev_path = None
3910

    
3911
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
3912
      if node != instance.primary_node:
3913
        continue
3914
      if ignore_size:
3915
        node_disk = node_disk.Copy()
3916
        node_disk.UnsetSize()
3917
      lu.cfg.SetDiskID(node_disk, node)
3918
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
3919
      msg = result.fail_msg
3920
      if msg:
3921
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
3922
                           " (is_primary=True, pass=2): %s",
3923
                           inst_disk.iv_name, node, msg)
3924
        disks_ok = False
3925
      else:
3926
        dev_path = result.payload
3927

    
3928
    device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
3929

    
3930
  # leave the disks configured for the primary node
3931
  # this is a workaround that would be fixed better by
3932
  # improving the logical/physical id handling
3933
  for disk in instance.disks:
3934
    lu.cfg.SetDiskID(disk, instance.primary_node)
3935

    
3936
  return disks_ok, device_info
3937

    
3938

    
3939
def _StartInstanceDisks(lu, instance, force):
3940
  """Start the disks of an instance.
3941

3942
  """
3943
  disks_ok, _ = _AssembleInstanceDisks(lu, instance,
3944
                                           ignore_secondaries=force)
3945
  if not disks_ok:
3946
    _ShutdownInstanceDisks(lu, instance)
3947
    if force is not None and not force:
3948
      lu.proc.LogWarning("", hint="If the message above refers to a"
3949
                         " secondary node,"
3950
                         " you can retry the operation using '--force'.")
3951
    raise errors.OpExecError("Disk consistency error")
3952

    
3953

    
3954
class LUDeactivateInstanceDisks(NoHooksLU):
3955
  """Shutdown an instance's disks.
3956

3957
  """
3958
  _OP_REQP = ["instance_name"]
3959
  REQ_BGL = False
3960

    
3961
  def ExpandNames(self):
3962
    self._ExpandAndLockInstance()
3963
    self.needed_locks[locking.LEVEL_NODE] = []
3964
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3965

    
3966
  def DeclareLocks(self, level):
3967
    if level == locking.LEVEL_NODE:
3968
      self._LockInstancesNodes()
3969

    
3970
  def CheckPrereq(self):
3971
    """Check prerequisites.
3972

3973
    This checks that the instance is in the cluster.
3974

3975
    """
3976
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3977
    assert self.instance is not None, \
3978
      "Cannot retrieve locked instance %s" % self.op.instance_name
3979

    
3980
  def Exec(self, feedback_fn):
3981
    """Deactivate the disks
3982

3983
    """
3984
    instance = self.instance
3985
    _SafeShutdownInstanceDisks(self, instance)
3986

    
3987

    
3988
def _SafeShutdownInstanceDisks(lu, instance):
3989
  """Shutdown block devices of an instance.
3990

3991
  This function checks if an instance is running, before calling
3992
  _ShutdownInstanceDisks.
3993

3994
  """
3995
  _CheckInstanceDown(lu, instance, "cannot shutdown disks")
3996
  _ShutdownInstanceDisks(lu, instance)
3997

    
3998

    
3999
def _ShutdownInstanceDisks(lu, instance, ignore_primary=False):
4000
  """Shutdown block devices of an instance.
4001

4002
  This does the shutdown on all nodes of the instance.
4003

4004
  If the ignore_primary is false, errors on the primary node are
4005
  ignored.
4006

4007
  """
4008
  all_result = True
4009
  for disk in instance.disks:
4010
    for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
4011
      lu.cfg.SetDiskID(top_disk, node)
4012
      result = lu.rpc.call_blockdev_shutdown(node, top_disk)
4013
      msg = result.fail_msg
4014
      if msg:
4015
        lu.LogWarning("Could not shutdown block device %s on node %s: %s",
4016
                      disk.iv_name, node, msg)
4017
        if not ignore_primary or node != instance.primary_node:
4018
          all_result = False
4019
  return all_result
4020

    
4021

    
4022
def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
4023
  """Checks if a node has enough free memory.
4024

4025
  This function check if a given node has the needed amount of free
4026
  memory. In case the node has less memory or we cannot get the
4027
  information from the node, this function raise an OpPrereqError
4028
  exception.
4029

4030
  @type lu: C{LogicalUnit}
4031
  @param lu: a logical unit from which we get configuration data
4032
  @type node: C{str}
4033
  @param node: the node to check
4034
  @type reason: C{str}
4035
  @param reason: string to use in the error message
4036
  @type requested: C{int}
4037
  @param requested: the amount of memory in MiB to check for
4038
  @type hypervisor_name: C{str}
4039
  @param hypervisor_name: the hypervisor to ask for memory stats
4040
  @raise errors.OpPrereqError: if the node doesn't have enough memory, or
4041
      we cannot check the node
4042

4043
  """
4044
  nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
4045
  nodeinfo[node].Raise("Can't get data from node %s" % node,
4046
                       prereq=True, ecode=errors.ECODE_ENVIRON)
4047
  free_mem = nodeinfo[node].payload.get('memory_free', None)
4048
  if not isinstance(free_mem, int):
4049
    raise errors.OpPrereqError("Can't compute free memory on node %s, result"
4050
                               " was '%s'" % (node, free_mem),
4051
                               errors.ECODE_ENVIRON)
4052
  if requested > free_mem:
4053
    raise errors.OpPrereqError("Not enough memory on node %s for %s:"
4054
                               " needed %s MiB, available %s MiB" %
4055
                               (node, reason, requested, free_mem),
4056
                               errors.ECODE_NORES)
4057

    
4058

    
4059
def _CheckNodesFreeDisk(lu, nodenames, requested):
4060
  """Checks if nodes have enough free disk space in the default VG.
4061

4062
  This function check if all given nodes have the needed amount of
4063
  free disk. In case any node has less disk or we cannot get the
4064
  information from the node, this function raise an OpPrereqError
4065
  exception.
4066

4067
  @type lu: C{LogicalUnit}
4068
  @param lu: a logical unit from which we get configuration data
4069
  @type nodenames: C{list}
4070
  @param nodenames: the list of node names to check
4071
  @type requested: C{int}
4072
  @param requested: the amount of disk in MiB to check for
4073
  @raise errors.OpPrereqError: if the node doesn't have enough disk, or
4074
      we cannot check the node
4075

4076
  """
4077
  nodeinfo = lu.rpc.call_node_info(nodenames, lu.cfg.GetVGName(),
4078
                                   lu.cfg.GetHypervisorType())
4079
  for node in nodenames:
4080
    info = nodeinfo[node]
4081
    info.Raise("Cannot get current information from node %s" % node,
4082
               prereq=True, ecode=errors.ECODE_ENVIRON)
4083
    vg_free = info.payload.get("vg_free", None)
4084
    if not isinstance(vg_free, int):
4085
      raise errors.OpPrereqError("Can't compute free disk space on node %s,"
4086
                                 " result was '%s'" % (node, vg_free),
4087
                                 errors.ECODE_ENVIRON)
4088
    if requested > vg_free:
4089
      raise errors.OpPrereqError("Not enough disk space on target node %s:"
4090
                                 " required %d MiB, available %d MiB" %
4091
                                 (node, requested, vg_free),
4092
                                 errors.ECODE_NORES)
4093

    
4094

    
4095
class LUStartupInstance(LogicalUnit):
4096
  """Starts an instance.
4097

4098
  """
4099
  HPATH = "instance-start"
4100
  HTYPE = constants.HTYPE_INSTANCE
4101
  _OP_REQP = ["instance_name", "force"]
4102
  REQ_BGL = False
4103

    
4104
  def ExpandNames(self):
4105
    self._ExpandAndLockInstance()
4106

    
4107
  def BuildHooksEnv(self):
4108
    """Build hooks env.
4109

4110
    This runs on master, primary and secondary nodes of the instance.
4111

4112
    """
4113
    env = {
4114
      "FORCE": self.op.force,
4115
      }
4116
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4117
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4118
    return env, nl, nl
4119

    
4120
  def CheckPrereq(self):
4121
    """Check prerequisites.
4122

4123
    This checks that the instance is in the cluster.
4124

4125
    """
4126
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4127
    assert self.instance is not None, \
4128
      "Cannot retrieve locked instance %s" % self.op.instance_name
4129

    
4130
    # extra beparams
4131
    self.beparams = getattr(self.op, "beparams", {})
4132
    if self.beparams:
4133
      if not isinstance(self.beparams, dict):
4134
        raise errors.OpPrereqError("Invalid beparams passed: %s, expected"
4135
                                   " dict" % (type(self.beparams), ),
4136
                                   errors.ECODE_INVAL)
4137
      # fill the beparams dict
4138
      utils.ForceDictType(self.beparams, constants.BES_PARAMETER_TYPES)
4139
      self.op.beparams = self.beparams
4140

    
4141
    # extra hvparams
4142
    self.hvparams = getattr(self.op, "hvparams", {})
4143
    if self.hvparams:
4144
      if not isinstance(self.hvparams, dict):
4145
        raise errors.OpPrereqError("Invalid hvparams passed: %s, expected"
4146
                                   " dict" % (type(self.hvparams), ),
4147
                                   errors.ECODE_INVAL)
4148

    
4149
      # check hypervisor parameter syntax (locally)
4150
      cluster = self.cfg.GetClusterInfo()
4151
      utils.ForceDictType(self.hvparams, constants.HVS_PARAMETER_TYPES)
4152
      filled_hvp = objects.FillDict(cluster.hvparams[instance.hypervisor],
4153
                                    instance.hvparams)
4154
      filled_hvp.update(self.hvparams)
4155
      hv_type = hypervisor.GetHypervisor(instance.hypervisor)
4156
      hv_type.CheckParameterSyntax(filled_hvp)
4157
      _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
4158
      self.op.hvparams = self.hvparams
4159

    
4160
    _CheckNodeOnline(self, instance.primary_node)
4161

    
4162
    bep = self.cfg.GetClusterInfo().FillBE(instance)
4163
    # check bridges existence
4164
    _CheckInstanceBridgesExist(self, instance)
4165

    
4166
    remote_info = self.rpc.call_instance_info(instance.primary_node,
4167
                                              instance.name,
4168
                                              instance.hypervisor)
4169
    remote_info.Raise("Error checking node %s" % instance.primary_node,
4170
                      prereq=True, ecode=errors.ECODE_ENVIRON)
4171
    if not remote_info.payload: # not running already
4172
      _CheckNodeFreeMemory(self, instance.primary_node,
4173
                           "starting instance %s" % instance.name,
4174
                           bep[constants.BE_MEMORY], instance.hypervisor)
4175

    
4176
  def Exec(self, feedback_fn):
4177
    """Start the instance.
4178

4179
    """
4180
    instance = self.instance
4181
    force = self.op.force
4182

    
4183
    self.cfg.MarkInstanceUp(instance.name)
4184

    
4185
    node_current = instance.primary_node
4186

    
4187
    _StartInstanceDisks(self, instance, force)
4188

    
4189
    result = self.rpc.call_instance_start(node_current, instance,
4190
                                          self.hvparams, self.beparams)
4191
    msg = result.fail_msg
4192
    if msg:
4193
      _ShutdownInstanceDisks(self, instance)
4194
      raise errors.OpExecError("Could not start instance: %s" % msg)
4195

    
4196

    
4197
class LURebootInstance(LogicalUnit):
4198
  """Reboot an instance.
4199

4200
  """
4201
  HPATH = "instance-reboot"
4202
  HTYPE = constants.HTYPE_INSTANCE
4203
  _OP_REQP = ["instance_name", "ignore_secondaries", "reboot_type"]
4204
  REQ_BGL = False
4205

    
4206
  def CheckArguments(self):
4207
    """Check the arguments.
4208

4209
    """
4210
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
4211
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
4212

    
4213
  def ExpandNames(self):
4214
    if self.op.reboot_type not in [constants.INSTANCE_REBOOT_SOFT,
4215
                                   constants.INSTANCE_REBOOT_HARD,
4216
                                   constants.INSTANCE_REBOOT_FULL]:
4217
      raise errors.ParameterError("reboot type not in [%s, %s, %s]" %
4218
                                  (constants.INSTANCE_REBOOT_SOFT,
4219
                                   constants.INSTANCE_REBOOT_HARD,
4220
                                   constants.INSTANCE_REBOOT_FULL))
4221
    self._ExpandAndLockInstance()
4222

    
4223
  def BuildHooksEnv(self):
4224
    """Build hooks env.
4225

4226
    This runs on master, primary and secondary nodes of the instance.
4227

4228
    """
4229
    env = {
4230
      "IGNORE_SECONDARIES": self.op.ignore_secondaries,
4231
      "REBOOT_TYPE": self.op.reboot_type,
4232
      "SHUTDOWN_TIMEOUT": self.shutdown_timeout,
4233
      }
4234
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4235
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4236
    return env, nl, nl
4237

    
4238
  def CheckPrereq(self):
4239
    """Check prerequisites.
4240

4241
    This checks that the instance is in the cluster.
4242

4243
    """
4244
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4245
    assert self.instance is not None, \
4246
      "Cannot retrieve locked instance %s" % self.op.instance_name
4247

    
4248
    _CheckNodeOnline(self, instance.primary_node)
4249

    
4250
    # check bridges existence
4251
    _CheckInstanceBridgesExist(self, instance)
4252

    
4253
  def Exec(self, feedback_fn):
4254
    """Reboot the instance.
4255

4256
    """
4257
    instance = self.instance
4258
    ignore_secondaries = self.op.ignore_secondaries
4259
    reboot_type = self.op.reboot_type
4260

    
4261
    node_current = instance.primary_node
4262

    
4263
    if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
4264
                       constants.INSTANCE_REBOOT_HARD]:
4265
      for disk in instance.disks:
4266
        self.cfg.SetDiskID(disk, node_current)
4267
      result = self.rpc.call_instance_reboot(node_current, instance,
4268
                                             reboot_type,
4269
                                             self.shutdown_timeout)
4270
      result.Raise("Could not reboot instance")
4271
    else:
4272
      result = self.rpc.call_instance_shutdown(node_current, instance,
4273
                                               self.shutdown_timeout)
4274
      result.Raise("Could not shutdown instance for full reboot")
4275
      _ShutdownInstanceDisks(self, instance)
4276
      _StartInstanceDisks(self, instance, ignore_secondaries)
4277
      result = self.rpc.call_instance_start(node_current, instance, None, None)
4278
      msg = result.fail_msg
4279
      if msg:
4280
        _ShutdownInstanceDisks(self, instance)
4281
        raise errors.OpExecError("Could not start instance for"
4282
                                 " full reboot: %s" % msg)
4283

    
4284
    self.cfg.MarkInstanceUp(instance.name)
4285

    
4286

    
4287
class LUShutdownInstance(LogicalUnit):
4288
  """Shutdown an instance.
4289

4290
  """
4291
  HPATH = "instance-stop"
4292
  HTYPE = constants.HTYPE_INSTANCE
4293
  _OP_REQP = ["instance_name"]
4294
  REQ_BGL = False
4295

    
4296
  def CheckArguments(self):
4297
    """Check the arguments.
4298

4299
    """
4300
    self.timeout = getattr(self.op, "timeout",
4301
                           constants.DEFAULT_SHUTDOWN_TIMEOUT)
4302

    
4303
  def ExpandNames(self):
4304
    self._ExpandAndLockInstance()
4305

    
4306
  def BuildHooksEnv(self):
4307
    """Build hooks env.
4308

4309
    This runs on master, primary and secondary nodes of the instance.
4310

4311
    """
4312
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4313
    env["TIMEOUT"] = self.timeout
4314
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4315
    return env, nl, nl
4316

    
4317
  def CheckPrereq(self):
4318
    """Check prerequisites.
4319

4320
    This checks that the instance is in the cluster.
4321

4322
    """
4323
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4324
    assert self.instance is not None, \
4325
      "Cannot retrieve locked instance %s" % self.op.instance_name
4326
    _CheckNodeOnline(self, self.instance.primary_node)
4327

    
4328
  def Exec(self, feedback_fn):
4329
    """Shutdown the instance.
4330

4331
    """
4332
    instance = self.instance
4333
    node_current = instance.primary_node
4334
    timeout = self.timeout
4335
    self.cfg.MarkInstanceDown(instance.name)
4336
    result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
4337
    msg = result.fail_msg
4338
    if msg:
4339
      self.proc.LogWarning("Could not shutdown instance: %s" % msg)
4340

    
4341
    _ShutdownInstanceDisks(self, instance)
4342

    
4343

    
4344
class LUReinstallInstance(LogicalUnit):
4345
  """Reinstall an instance.
4346

4347
  """
4348
  HPATH = "instance-reinstall"
4349
  HTYPE = constants.HTYPE_INSTANCE
4350
  _OP_REQP = ["instance_name"]
4351
  REQ_BGL = False
4352

    
4353
  def ExpandNames(self):
4354
    self._ExpandAndLockInstance()
4355

    
4356
  def BuildHooksEnv(self):
4357
    """Build hooks env.
4358

4359
    This runs on master, primary and secondary nodes of the instance.
4360

4361
    """
4362
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4363
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4364
    return env, nl, nl
4365

    
4366
  def CheckPrereq(self):
4367
    """Check prerequisites.
4368

4369
    This checks that the instance is in the cluster and is not running.
4370

4371
    """
4372
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4373
    assert instance is not None, \
4374
      "Cannot retrieve locked instance %s" % self.op.instance_name
4375
    _CheckNodeOnline(self, instance.primary_node)
4376

    
4377
    if instance.disk_template == constants.DT_DISKLESS:
4378
      raise errors.OpPrereqError("Instance '%s' has no disks" %
4379
                                 self.op.instance_name,
4380
                                 errors.ECODE_INVAL)
4381
    _CheckInstanceDown(self, instance, "cannot reinstall")
4382

    
4383
    self.op.os_type = getattr(self.op, "os_type", None)
4384
    self.op.force_variant = getattr(self.op, "force_variant", False)
4385
    if self.op.os_type is not None:
4386
      # OS verification
4387
      pnode = _ExpandNodeName(self.cfg, instance.primary_node)
4388
      _CheckNodeHasOS(self, pnode, self.op.os_type, self.op.force_variant)
4389

    
4390
    self.instance = instance
4391

    
4392
  def Exec(self, feedback_fn):
4393
    """Reinstall the instance.
4394

4395
    """
4396
    inst = self.instance
4397

    
4398
    if self.op.os_type is not None:
4399
      feedback_fn("Changing OS to '%s'..." % self.op.os_type)
4400
      inst.os = self.op.os_type
4401
      self.cfg.Update(inst, feedback_fn)
4402

    
4403
    _StartInstanceDisks(self, inst, None)
4404
    try:
4405
      feedback_fn("Running the instance OS create scripts...")
4406
      # FIXME: pass debug option from opcode to backend
4407
      result = self.rpc.call_instance_os_add(inst.primary_node, inst, True,
4408
                                             self.op.debug_level)
4409
      result.Raise("Could not install OS for instance %s on node %s" %
4410
                   (inst.name, inst.primary_node))
4411
    finally:
4412
      _ShutdownInstanceDisks(self, inst)
4413

    
4414

    
4415
class LURecreateInstanceDisks(LogicalUnit):
4416
  """Recreate an instance's missing disks.
4417

4418
  """
4419
  HPATH = "instance-recreate-disks"
4420
  HTYPE = constants.HTYPE_INSTANCE
4421
  _OP_REQP = ["instance_name", "disks"]
4422
  REQ_BGL = False
4423

    
4424
  def CheckArguments(self):
4425
    """Check the arguments.
4426

4427
    """
4428
    if not isinstance(self.op.disks, list):
4429
      raise errors.OpPrereqError("Invalid disks parameter", errors.ECODE_INVAL)
4430
    for item in self.op.disks:
4431
      if (not isinstance(item, int) or
4432
          item < 0):
4433
        raise errors.OpPrereqError("Invalid disk specification '%s'" %
4434
                                   str(item), errors.ECODE_INVAL)
4435

    
4436
  def ExpandNames(self):
4437
    self._ExpandAndLockInstance()
4438

    
4439
  def BuildHooksEnv(self):
4440
    """Build hooks env.
4441

4442
    This runs on master, primary and secondary nodes of the instance.
4443

4444
    """
4445
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4446
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4447
    return env, nl, nl
4448

    
4449
  def CheckPrereq(self):
4450
    """Check prerequisites.
4451

4452
    This checks that the instance is in the cluster and is not running.
4453

4454
    """
4455
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4456
    assert instance is not None, \
4457
      "Cannot retrieve locked instance %s" % self.op.instance_name
4458
    _CheckNodeOnline(self, instance.primary_node)
4459

    
4460
    if instance.disk_template == constants.DT_DISKLESS:
4461
      raise errors.OpPrereqError("Instance '%s' has no disks" %
4462
                                 self.op.instance_name, errors.ECODE_INVAL)
4463
    _CheckInstanceDown(self, instance, "cannot recreate disks")
4464

    
4465
    if not self.op.disks:
4466
      self.op.disks = range(len(instance.disks))
4467
    else:
4468
      for idx in self.op.disks:
4469
        if idx >= len(instance.disks):
4470
          raise errors.OpPrereqError("Invalid disk index passed '%s'" % idx,
4471
                                     errors.ECODE_INVAL)
4472

    
4473
    self.instance = instance
4474

    
4475
  def Exec(self, feedback_fn):
4476
    """Recreate the disks.
4477

4478
    """
4479
    to_skip = []
4480
    for idx, _ in enumerate(self.instance.disks):
4481
      if idx not in self.op.disks: # disk idx has not been passed in
4482
        to_skip.append(idx)
4483
        continue
4484

    
4485
    _CreateDisks(self, self.instance, to_skip=to_skip)
4486

    
4487

    
4488
class LURenameInstance(LogicalUnit):
4489
  """Rename an instance.
4490

4491
  """
4492
  HPATH = "instance-rename"
4493
  HTYPE = constants.HTYPE_INSTANCE
4494
  _OP_REQP = ["instance_name", "new_name"]
4495

    
4496
  def BuildHooksEnv(self):
4497
    """Build hooks env.
4498

4499
    This runs on master, primary and secondary nodes of the instance.
4500

4501
    """
4502
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4503
    env["INSTANCE_NEW_NAME"] = self.op.new_name
4504
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4505
    return env, nl, nl
4506

    
4507
  def CheckPrereq(self):
4508
    """Check prerequisites.
4509

4510
    This checks that the instance is in the cluster and is not running.
4511

4512
    """
4513
    self.op.instance_name = _ExpandInstanceName(self.cfg,
4514
                                                self.op.instance_name)
4515
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4516
    assert instance is not None
4517
    _CheckNodeOnline(self, instance.primary_node)
4518
    _CheckInstanceDown(self, instance, "cannot rename")
4519
    self.instance = instance
4520

    
4521
    # new name verification
4522
    name_info = utils.GetHostInfo(self.op.new_name)
4523

    
4524
    self.op.new_name = new_name = name_info.name
4525
    instance_list = self.cfg.GetInstanceList()
4526
    if new_name in instance_list:
4527
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
4528
                                 new_name, errors.ECODE_EXISTS)
4529

    
4530
    if not getattr(self.op, "ignore_ip", False):
4531
      if utils.TcpPing(name_info.ip, constants.DEFAULT_NODED_PORT):
4532
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
4533
                                   (name_info.ip, new_name),
4534
                                   errors.ECODE_NOTUNIQUE)
4535

    
4536

    
4537
  def Exec(self, feedback_fn):
4538
    """Reinstall the instance.
4539

4540
    """
4541
    inst = self.instance
4542
    old_name = inst.name
4543

    
4544
    if inst.disk_template == constants.DT_FILE:
4545
      old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
4546

    
4547
    self.cfg.RenameInstance(inst.name, self.op.new_name)
4548
    # Change the instance lock. This is definitely safe while we hold the BGL
4549
    self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
4550
    self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
4551

    
4552
    # re-read the instance from the configuration after rename
4553
    inst = self.cfg.GetInstanceInfo(self.op.new_name)
4554

    
4555
    if inst.disk_template == constants.DT_FILE:
4556
      new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
4557
      result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
4558
                                                     old_file_storage_dir,
4559
                                                     new_file_storage_dir)
4560
      result.Raise("Could not rename on node %s directory '%s' to '%s'"
4561
                   " (but the instance has been renamed in Ganeti)" %
4562
                   (inst.primary_node, old_file_storage_dir,
4563
                    new_file_storage_dir))
4564

    
4565
    _StartInstanceDisks(self, inst, None)
4566
    try:
4567
      result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
4568
                                                 old_name, self.op.debug_level)
4569
      msg = result.fail_msg
4570
      if msg:
4571
        msg = ("Could not run OS rename script for instance %s on node %s"
4572
               " (but the instance has been renamed in Ganeti): %s" %
4573
               (inst.name, inst.primary_node, msg))
4574
        self.proc.LogWarning(msg)
4575
    finally:
4576
      _ShutdownInstanceDisks(self, inst)
4577

    
4578

    
4579
class LURemoveInstance(LogicalUnit):
4580
  """Remove an instance.
4581

4582
  """
4583
  HPATH = "instance-remove"
4584
  HTYPE = constants.HTYPE_INSTANCE
4585
  _OP_REQP = ["instance_name", "ignore_failures"]
4586
  REQ_BGL = False
4587

    
4588
  def CheckArguments(self):
4589
    """Check the arguments.
4590

4591
    """
4592
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
4593
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
4594

    
4595
  def ExpandNames(self):
4596
    self._ExpandAndLockInstance()
4597
    self.needed_locks[locking.LEVEL_NODE] = []
4598
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4599

    
4600
  def DeclareLocks(self, level):
4601
    if level == locking.LEVEL_NODE:
4602
      self._LockInstancesNodes()
4603

    
4604
  def BuildHooksEnv(self):
4605
    """Build hooks env.
4606

4607
    This runs on master, primary and secondary nodes of the instance.
4608

4609
    """
4610
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4611
    env["SHUTDOWN_TIMEOUT"] = self.shutdown_timeout
4612
    nl = [self.cfg.GetMasterNode()]
4613
    nl_post = list(self.instance.all_nodes) + nl
4614
    return env, nl, nl_post
4615

    
4616
  def CheckPrereq(self):
4617
    """Check prerequisites.
4618

4619
    This checks that the instance is in the cluster.
4620

4621
    """
4622
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4623
    assert self.instance is not None, \
4624
      "Cannot retrieve locked instance %s" % self.op.instance_name
4625

    
4626
  def Exec(self, feedback_fn):
4627
    """Remove the instance.
4628

4629
    """
4630
    instance = self.instance
4631
    logging.info("Shutting down instance %s on node %s",
4632
                 instance.name, instance.primary_node)
4633

    
4634
    result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
4635
                                             self.shutdown_timeout)
4636
    msg = result.fail_msg
4637
    if msg:
4638
      if self.op.ignore_failures:
4639
        feedback_fn("Warning: can't shutdown instance: %s" % msg)
4640
      else:
4641
        raise errors.OpExecError("Could not shutdown instance %s on"
4642
                                 " node %s: %s" %
4643
                                 (instance.name, instance.primary_node, msg))
4644

    
4645
    _RemoveInstance(self, feedback_fn, instance, self.op.ignore_failures)
4646

    
4647

    
4648
def _RemoveInstance(lu, feedback_fn, instance, ignore_failures):
4649
  """Utility function to remove an instance.
4650

4651
  """
4652
  logging.info("Removing block devices for instance %s", instance.name)
4653

    
4654
  if not _RemoveDisks(lu, instance):
4655
    if not ignore_failures:
4656
      raise errors.OpExecError("Can't remove instance's disks")
4657
    feedback_fn("Warning: can't remove instance's disks")
4658

    
4659
  logging.info("Removing instance %s out of cluster config", instance.name)
4660

    
4661
  lu.cfg.RemoveInstance(instance.name)
4662

    
4663
  assert not lu.remove_locks.get(locking.LEVEL_INSTANCE), \
4664
    "Instance lock removal conflict"
4665

    
4666
  # Remove lock for the instance
4667
  lu.remove_locks[locking.LEVEL_INSTANCE] = instance.name
4668

    
4669

    
4670
class LUQueryInstances(NoHooksLU):
4671
  """Logical unit for querying instances.
4672

4673
  """
4674
  # pylint: disable-msg=W0142
4675
  _OP_REQP = ["output_fields", "names", "use_locking"]
4676
  REQ_BGL = False
4677
  _SIMPLE_FIELDS = ["name", "os", "network_port", "hypervisor",
4678
                    "serial_no", "ctime", "mtime", "uuid"]
4679
  _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
4680
                                    "admin_state",
4681
                                    "disk_template", "ip", "mac", "bridge",
4682
                                    "nic_mode", "nic_link",
4683
                                    "sda_size", "sdb_size", "vcpus", "tags",
4684
                                    "network_port", "beparams",
4685
                                    r"(disk)\.(size)/([0-9]+)",
4686
                                    r"(disk)\.(sizes)", "disk_usage",
4687
                                    r"(nic)\.(mac|ip|mode|link)/([0-9]+)",
4688
                                    r"(nic)\.(bridge)/([0-9]+)",
4689
                                    r"(nic)\.(macs|ips|modes|links|bridges)",
4690
                                    r"(disk|nic)\.(count)",
4691
                                    "hvparams",
4692
                                    ] + _SIMPLE_FIELDS +
4693
                                  ["hv/%s" % name
4694
                                   for name in constants.HVS_PARAMETERS
4695
                                   if name not in constants.HVC_GLOBALS] +
4696
                                  ["be/%s" % name
4697
                                   for name in constants.BES_PARAMETERS])
4698
  _FIELDS_DYNAMIC = utils.FieldSet("oper_state", "oper_ram", "status")
4699

    
4700

    
4701
  def ExpandNames(self):
4702
    _CheckOutputFields(static=self._FIELDS_STATIC,
4703
                       dynamic=self._FIELDS_DYNAMIC,
4704
                       selected=self.op.output_fields)
4705

    
4706
    self.needed_locks = {}
4707
    self.share_locks[locking.LEVEL_INSTANCE] = 1
4708
    self.share_locks[locking.LEVEL_NODE] = 1
4709

    
4710
    if self.op.names:
4711
      self.wanted = _GetWantedInstances(self, self.op.names)
4712
    else:
4713
      self.wanted = locking.ALL_SET
4714

    
4715
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
4716
    self.do_locking = self.do_node_query and self.op.use_locking
4717
    if self.do_locking:
4718
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
4719
      self.needed_locks[locking.LEVEL_NODE] = []
4720
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4721

    
4722
  def DeclareLocks(self, level):
4723
    if level == locking.LEVEL_NODE and self.do_locking:
4724
      self._LockInstancesNodes()
4725

    
4726
  def CheckPrereq(self):
4727
    """Check prerequisites.
4728

4729
    """
4730
    pass
4731

    
4732
  def Exec(self, feedback_fn):
4733
    """Computes the list of nodes and their attributes.
4734

4735
    """
4736
    # pylint: disable-msg=R0912
4737
    # way too many branches here
4738
    all_info = self.cfg.GetAllInstancesInfo()
4739
    if self.wanted == locking.ALL_SET:
4740
      # caller didn't specify instance names, so ordering is not important
4741
      if self.do_locking:
4742
        instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
4743
      else:
4744
        instance_names = all_info.keys()
4745
      instance_names = utils.NiceSort(instance_names)
4746
    else:
4747
      # caller did specify names, so we must keep the ordering
4748
      if self.do_locking:
4749
        tgt_set = self.acquired_locks[locking.LEVEL_INSTANCE]
4750
      else:
4751
        tgt_set = all_info.keys()
4752
      missing = set(self.wanted).difference(tgt_set)
4753
      if missing:
4754
        raise errors.OpExecError("Some instances were removed before"
4755
                                 " retrieving their data: %s" % missing)
4756
      instance_names = self.wanted
4757

    
4758
    instance_list = [all_info[iname] for iname in instance_names]
4759

    
4760
    # begin data gathering
4761

    
4762
    nodes = frozenset([inst.primary_node for inst in instance_list])
4763
    hv_list = list(set([inst.hypervisor for inst in instance_list]))
4764

    
4765
    bad_nodes = []
4766
    off_nodes = []
4767
    if self.do_node_query:
4768
      live_data = {}
4769
      node_data = self.rpc.call_all_instances_info(nodes, hv_list)
4770
      for name in nodes:
4771
        result = node_data[name]
4772
        if result.offline:
4773
          # offline nodes will be in both lists
4774
          off_nodes.append(name)
4775
        if result.fail_msg:
4776
          bad_nodes.append(name)
4777
        else:
4778
          if result.payload:
4779
            live_data.update(result.payload)
4780
          # else no instance is alive
4781
    else:
4782
      live_data = dict([(name, {}) for name in instance_names])
4783

    
4784
    # end data gathering
4785

    
4786
    HVPREFIX = "hv/"
4787
    BEPREFIX = "be/"
4788
    output = []
4789
    cluster = self.cfg.GetClusterInfo()
4790
    for instance in instance_list:
4791
      iout = []
4792
      i_hv = cluster.FillHV(instance, skip_globals=True)
4793
      i_be = cluster.FillBE(instance)
4794
      i_nicp = [objects.FillDict(cluster.nicparams[constants.PP_DEFAULT],
4795
                                 nic.nicparams) for nic in instance.nics]
4796
      for field in self.op.output_fields:
4797
        st_match = self._FIELDS_STATIC.Matches(field)
4798
        if field in self._SIMPLE_FIELDS:
4799
          val = getattr(instance, field)
4800
        elif field == "pnode":
4801
          val = instance.primary_node
4802
        elif field == "snodes":
4803
          val = list(instance.secondary_nodes)
4804
        elif field == "admin_state":
4805
          val = instance.admin_up
4806
        elif field == "oper_state":
4807
          if instance.primary_node in bad_nodes:
4808
            val = None
4809
          else:
4810
            val = bool(live_data.get(instance.name))
4811
        elif field == "status":
4812
          if instance.primary_node in off_nodes:
4813
            val = "ERROR_nodeoffline"
4814
          elif instance.primary_node in bad_nodes:
4815
            val = "ERROR_nodedown"
4816
          else:
4817
            running = bool(live_data.get(instance.name))
4818
            if running:
4819
              if instance.admin_up:
4820
                val = "running"
4821
              else:
4822
                val = "ERROR_up"
4823
            else:
4824
              if instance.admin_up:
4825
                val = "ERROR_down"
4826
              else:
4827
                val = "ADMIN_down"
4828
        elif field == "oper_ram":
4829
          if instance.primary_node in bad_nodes:
4830
            val = None
4831
          elif instance.name in live_data:
4832
            val = live_data[instance.name].get("memory", "?")
4833
          else:
4834
            val = "-"
4835
        elif field == "vcpus":
4836
          val = i_be[constants.BE_VCPUS]
4837
        elif field == "disk_template":
4838
          val = instance.disk_template
4839
        elif field == "ip":
4840
          if instance.nics:
4841
            val = instance.nics[0].ip
4842
          else:
4843
            val = None
4844
        elif field == "nic_mode":
4845
          if instance.nics:
4846
            val = i_nicp[0][constants.NIC_MODE]
4847
          else:
4848
            val = None
4849
        elif field == "nic_link":
4850
          if instance.nics:
4851
            val = i_nicp[0][constants.NIC_LINK]
4852
          else:
4853
            val = None
4854
        elif field == "bridge":
4855
          if (instance.nics and
4856
              i_nicp[0][constants.NIC_MODE] == constants.NIC_MODE_BRIDGED):
4857
            val = i_nicp[0][constants.NIC_LINK]
4858
          else:
4859
            val = None
4860
        elif field == "mac":
4861
          if instance.nics:
4862
            val = instance.nics[0].mac
4863
          else:
4864
            val = None
4865
        elif field == "sda_size" or field == "sdb_size":
4866
          idx = ord(field[2]) - ord('a')
4867
          try:
4868
            val = instance.FindDisk(idx).size
4869
          except errors.OpPrereqError:
4870
            val = None
4871
        elif field == "disk_usage": # total disk usage per node
4872
          disk_sizes = [{'size': disk.size} for disk in instance.disks]
4873
          val = _ComputeDiskSize(instance.disk_template, disk_sizes)
4874
        elif field == "tags":
4875
          val = list(instance.GetTags())
4876
        elif field == "hvparams":
4877
          val = i_hv
4878
        elif (field.startswith(HVPREFIX) and
4879
              field[len(HVPREFIX):] in constants.HVS_PARAMETERS and
4880
              field[len(HVPREFIX):] not in constants.HVC_GLOBALS):
4881
          val = i_hv.get(field[len(HVPREFIX):], None)
4882
        elif field == "beparams":
4883
          val = i_be
4884
        elif (field.startswith(BEPREFIX) and
4885
              field[len(BEPREFIX):] in constants.BES_PARAMETERS):
4886
          val = i_be.get(field[len(BEPREFIX):], None)
4887
        elif st_match and st_match.groups():
4888
          # matches a variable list
4889
          st_groups = st_match.groups()
4890
          if st_groups and st_groups[0] == "disk":
4891
            if st_groups[1] == "count":
4892
              val = len(instance.disks)
4893
            elif st_groups[1] == "sizes":
4894
              val = [disk.size for disk in instance.disks]
4895
            elif st_groups[1] == "size":
4896
              try:
4897
                val = instance.FindDisk(st_groups[2]).size
4898
              except errors.OpPrereqError:
4899
                val = None
4900
            else:
4901
              assert False, "Unhandled disk parameter"
4902
          elif st_groups[0] == "nic":
4903
            if st_groups[1] == "count":
4904
              val = len(instance.nics)
4905
            elif st_groups[1] == "macs":
4906
              val = [nic.mac for nic in instance.nics]
4907
            elif st_groups[1] == "ips":
4908
              val = [nic.ip for nic in instance.nics]
4909
            elif st_groups[1] == "modes":
4910
              val = [nicp[constants.NIC_MODE] for nicp in i_nicp]
4911
            elif st_groups[1] == "links":
4912
              val = [nicp[constants.NIC_LINK] for nicp in i_nicp]
4913
            elif st_groups[1] == "bridges":
4914
              val = []
4915
              for nicp in i_nicp:
4916
                if nicp[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
4917
                  val.append(nicp[constants.NIC_LINK])
4918
                else:
4919
                  val.append(None)
4920
            else:
4921
              # index-based item
4922
              nic_idx = int(st_groups[2])
4923
              if nic_idx >= len(instance.nics):
4924
                val = None
4925
              else:
4926
                if st_groups[1] == "mac":
4927
                  val = instance.nics[nic_idx].mac
4928
                elif st_groups[1] == "ip":
4929
                  val = instance.nics[nic_idx].ip
4930
                elif st_groups[1] == "mode":
4931
                  val = i_nicp[nic_idx][constants.NIC_MODE]
4932
                elif st_groups[1] == "link":
4933
                  val = i_nicp[nic_idx][constants.NIC_LINK]
4934
                elif st_groups[1] == "bridge":
4935
                  nic_mode = i_nicp[nic_idx][constants.NIC_MODE]
4936
                  if nic_mode == constants.NIC_MODE_BRIDGED:
4937
                    val = i_nicp[nic_idx][constants.NIC_LINK]
4938
                  else:
4939
                    val = None
4940
                else:
4941
                  assert False, "Unhandled NIC parameter"
4942
          else:
4943
            assert False, ("Declared but unhandled variable parameter '%s'" %
4944
                           field)
4945
        else:
4946
          assert False, "Declared but unhandled parameter '%s'" % field
4947
        iout.append(val)
4948
      output.append(iout)
4949

    
4950
    return output
4951

    
4952

    
4953
class LUFailoverInstance(LogicalUnit):
4954
  """Failover an instance.
4955

4956
  """
4957
  HPATH = "instance-failover"
4958
  HTYPE = constants.HTYPE_INSTANCE
4959
  _OP_REQP = ["instance_name", "ignore_consistency"]
4960
  REQ_BGL = False
4961

    
4962
  def CheckArguments(self):
4963
    """Check the arguments.
4964

4965
    """
4966
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
4967
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
4968

    
4969
  def ExpandNames(self):
4970
    self._ExpandAndLockInstance()
4971
    self.needed_locks[locking.LEVEL_NODE] = []
4972
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4973

    
4974
  def DeclareLocks(self, level):
4975
    if level == locking.LEVEL_NODE:
4976
      self._LockInstancesNodes()
4977

    
4978
  def BuildHooksEnv(self):
4979
    """Build hooks env.
4980

4981
    This runs on master, primary and secondary nodes of the instance.
4982

4983
    """
4984
    instance = self.instance
4985
    source_node = instance.primary_node
4986
    target_node = instance.secondary_nodes[0]
4987
    env = {
4988
      "IGNORE_CONSISTENCY": self.op.ignore_consistency,
4989
      "SHUTDOWN_TIMEOUT": self.shutdown_timeout,
4990
      "OLD_PRIMARY": source_node,
4991
      "OLD_SECONDARY": target_node,
4992
      "NEW_PRIMARY": target_node,
4993
      "NEW_SECONDARY": source_node,
4994
      }
4995
    env.update(_BuildInstanceHookEnvByObject(self, instance))
4996
    nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
4997
    nl_post = list(nl)
4998
    nl_post.append(source_node)
4999
    return env, nl, nl_post
5000

    
5001
  def CheckPrereq(self):
5002
    """Check prerequisites.
5003

5004
    This checks that the instance is in the cluster.
5005

5006
    """
5007
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5008
    assert self.instance is not None, \
5009
      "Cannot retrieve locked instance %s" % self.op.instance_name
5010

    
5011
    bep = self.cfg.GetClusterInfo().FillBE(instance)
5012
    if instance.disk_template not in constants.DTS_NET_MIRROR:
5013
      raise errors.OpPrereqError("Instance's disk layout is not"
5014
                                 " network mirrored, cannot failover.",
5015
                                 errors.ECODE_STATE)
5016

    
5017
    secondary_nodes = instance.secondary_nodes
5018
    if not secondary_nodes:
5019
      raise errors.ProgrammerError("no secondary node but using "
5020
                                   "a mirrored disk template")
5021

    
5022
    target_node = secondary_nodes[0]
5023
    _CheckNodeOnline(self, target_node)
5024
    _CheckNodeNotDrained(self, target_node)
5025
    if instance.admin_up:
5026
      # check memory requirements on the secondary node
5027
      _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5028
                           instance.name, bep[constants.BE_MEMORY],
5029
                           instance.hypervisor)
5030
    else:
5031
      self.LogInfo("Not checking memory on the secondary node as"
5032
                   " instance will not be started")
5033

    
5034
    # check bridge existance
5035
    _CheckInstanceBridgesExist(self, instance, node=target_node)
5036

    
5037
  def Exec(self, feedback_fn):
5038
    """Failover an instance.
5039

5040
    The failover is done by shutting it down on its present node and
5041
    starting it on the secondary.
5042

5043
    """
5044
    instance = self.instance
5045

    
5046
    source_node = instance.primary_node
5047
    target_node = instance.secondary_nodes[0]
5048

    
5049
    if instance.admin_up:
5050
      feedback_fn("* checking disk consistency between source and target")
5051
      for dev in instance.disks:
5052
        # for drbd, these are drbd over lvm
5053
        if not _CheckDiskConsistency(self, dev, target_node, False):
5054
          if not self.op.ignore_consistency:
5055
            raise errors.OpExecError("Disk %s is degraded on target node,"
5056
                                     " aborting failover." % dev.iv_name)
5057
    else:
5058
      feedback_fn("* not checking disk consistency as instance is not running")
5059

    
5060
    feedback_fn("* shutting down instance on source node")
5061
    logging.info("Shutting down instance %s on node %s",
5062
                 instance.name, source_node)
5063

    
5064
    result = self.rpc.call_instance_shutdown(source_node, instance,
5065
                                             self.shutdown_timeout)
5066
    msg = result.fail_msg
5067
    if msg:
5068
      if self.op.ignore_consistency:
5069
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
5070
                             " Proceeding anyway. Please make sure node"
5071
                             " %s is down. Error details: %s",
5072
                             instance.name, source_node, source_node, msg)
5073
      else:
5074
        raise errors.OpExecError("Could not shutdown instance %s on"
5075
                                 " node %s: %s" %
5076
                                 (instance.name, source_node, msg))
5077

    
5078
    feedback_fn("* deactivating the instance's disks on source node")
5079
    if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
5080
      raise errors.OpExecError("Can't shut down the instance's disks.")
5081

    
5082
    instance.primary_node = target_node
5083
    # distribute new instance config to the other nodes
5084
    self.cfg.Update(instance, feedback_fn)
5085

    
5086
    # Only start the instance if it's marked as up
5087
    if instance.admin_up:
5088
      feedback_fn("* activating the instance's disks on target node")
5089
      logging.info("Starting instance %s on node %s",
5090
                   instance.name, target_node)
5091

    
5092
      disks_ok, _ = _AssembleInstanceDisks(self, instance,
5093
                                               ignore_secondaries=True)
5094
      if not disks_ok:
5095
        _ShutdownInstanceDisks(self, instance)
5096
        raise errors.OpExecError("Can't activate the instance's disks")
5097

    
5098
      feedback_fn("* starting the instance on the target node")
5099
      result = self.rpc.call_instance_start(target_node, instance, None, None)
5100
      msg = result.fail_msg
5101
      if msg:
5102
        _ShutdownInstanceDisks(self, instance)
5103
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
5104
                                 (instance.name, target_node, msg))
5105

    
5106

    
5107
class LUMigrateInstance(LogicalUnit):
5108
  """Migrate an instance.
5109

5110
  This is migration without shutting down, compared to the failover,
5111
  which is done with shutdown.
5112

5113
  """
5114
  HPATH = "instance-migrate"
5115
  HTYPE = constants.HTYPE_INSTANCE
5116
  _OP_REQP = ["instance_name", "live", "cleanup"]
5117

    
5118
  REQ_BGL = False
5119

    
5120
  def ExpandNames(self):
5121
    self._ExpandAndLockInstance()
5122

    
5123
    self.needed_locks[locking.LEVEL_NODE] = []
5124
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5125

    
5126
    self._migrater = TLMigrateInstance(self, self.op.instance_name,
5127
                                       self.op.live, self.op.cleanup)
5128
    self.tasklets = [self._migrater]
5129

    
5130
  def DeclareLocks(self, level):
5131
    if level == locking.LEVEL_NODE:
5132
      self._LockInstancesNodes()
5133

    
5134
  def BuildHooksEnv(self):
5135
    """Build hooks env.
5136

5137
    This runs on master, primary and secondary nodes of the instance.
5138

5139
    """
5140
    instance = self._migrater.instance
5141
    source_node = instance.primary_node
5142
    target_node = instance.secondary_nodes[0]
5143
    env = _BuildInstanceHookEnvByObject(self, instance)
5144
    env["MIGRATE_LIVE"] = self.op.live
5145
    env["MIGRATE_CLEANUP"] = self.op.cleanup
5146
    env.update({
5147
        "OLD_PRIMARY": source_node,
5148
        "OLD_SECONDARY": target_node,
5149
        "NEW_PRIMARY": target_node,
5150
        "NEW_SECONDARY": source_node,
5151
        })
5152
    nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
5153
    nl_post = list(nl)
5154
    nl_post.append(source_node)
5155
    return env, nl, nl_post
5156

    
5157

    
5158
class LUMoveInstance(LogicalUnit):
5159
  """Move an instance by data-copying.
5160

5161
  """
5162
  HPATH = "instance-move"
5163
  HTYPE = constants.HTYPE_INSTANCE
5164
  _OP_REQP = ["instance_name", "target_node"]
5165
  REQ_BGL = False
5166

    
5167
  def CheckArguments(self):
5168
    """Check the arguments.
5169

5170
    """
5171
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
5172
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
5173

    
5174
  def ExpandNames(self):
5175
    self._ExpandAndLockInstance()
5176
    target_node = _ExpandNodeName(self.cfg, self.op.target_node)
5177
    self.op.target_node = target_node
5178
    self.needed_locks[locking.LEVEL_NODE] = [target_node]
5179
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5180

    
5181
  def DeclareLocks(self, level):
5182
    if level == locking.LEVEL_NODE:
5183
      self._LockInstancesNodes(primary_only=True)
5184

    
5185
  def BuildHooksEnv(self):
5186
    """Build hooks env.
5187

5188
    This runs on master, primary and secondary nodes of the instance.
5189

5190
    """
5191
    env = {
5192
      "TARGET_NODE": self.op.target_node,
5193
      "SHUTDOWN_TIMEOUT": self.shutdown_timeout,
5194
      }
5195
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5196
    nl = [self.cfg.GetMasterNode()] + [self.instance.primary_node,
5197
                                       self.op.target_node]
5198
    return env, nl, nl
5199

    
5200
  def CheckPrereq(self):
5201
    """Check prerequisites.
5202

5203
    This checks that the instance is in the cluster.
5204

5205
    """
5206
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5207
    assert self.instance is not None, \
5208
      "Cannot retrieve locked instance %s" % self.op.instance_name
5209

    
5210
    node = self.cfg.GetNodeInfo(self.op.target_node)
5211
    assert node is not None, \
5212
      "Cannot retrieve locked node %s" % self.op.target_node
5213

    
5214
    self.target_node = target_node = node.name
5215

    
5216
    if target_node == instance.primary_node:
5217
      raise errors.OpPrereqError("Instance %s is already on the node %s" %
5218
                                 (instance.name, target_node),
5219
                                 errors.ECODE_STATE)
5220

    
5221
    bep = self.cfg.GetClusterInfo().FillBE(instance)
5222

    
5223
    for idx, dsk in enumerate(instance.disks):
5224
      if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
5225
        raise errors.OpPrereqError("Instance disk %d has a complex layout,"
5226
                                   " cannot copy" % idx, errors.ECODE_STATE)
5227

    
5228
    _CheckNodeOnline(self, target_node)
5229
    _CheckNodeNotDrained(self, target_node)
5230

    
5231
    if instance.admin_up:
5232
      # check memory requirements on the secondary node
5233
      _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
5234
                           instance.name, bep[constants.BE_MEMORY],
5235
                           instance.hypervisor)
5236
    else:
5237
      self.LogInfo("Not checking memory on the secondary node as"
5238
                   " instance will not be started")
5239

    
5240
    # check bridge existance
5241
    _CheckInstanceBridgesExist(self, instance, node=target_node)
5242

    
5243
  def Exec(self, feedback_fn):
5244
    """Move an instance.
5245

5246
    The move is done by shutting it down on its present node, copying
5247
    the data over (slow) and starting it on the new node.
5248

5249
    """
5250
    instance = self.instance
5251

    
5252
    source_node = instance.primary_node
5253
    target_node = self.target_node
5254

    
5255
    self.LogInfo("Shutting down instance %s on source node %s",
5256
                 instance.name, source_node)
5257

    
5258
    result = self.rpc.call_instance_shutdown(source_node, instance,
5259
                                             self.shutdown_timeout)
5260
    msg = result.fail_msg
5261
    if msg:
5262
      if self.op.ignore_consistency:
5263
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
5264
                             " Proceeding anyway. Please make sure node"
5265
                             " %s is down. Error details: %s",
5266
                             instance.name, source_node, source_node, msg)
5267
      else:
5268
        raise errors.OpExecError("Could not shutdown instance %s on"
5269
                                 " node %s: %s" %
5270
                                 (instance.name, source_node, msg))
5271

    
5272
    # create the target disks
5273
    try:
5274
      _CreateDisks(self, instance, target_node=target_node)
5275
    except errors.OpExecError:
5276
      self.LogWarning("Device creation failed, reverting...")
5277
      try:
5278
        _RemoveDisks(self, instance, target_node=target_node)
5279
      finally:
5280
        self.cfg.ReleaseDRBDMinors(instance.name)
5281
        raise
5282

    
5283
    cluster_name = self.cfg.GetClusterInfo().cluster_name
5284

    
5285
    errs = []
5286
    # activate, get path, copy the data over
5287
    for idx, disk in enumerate(instance.disks):
5288
      self.LogInfo("Copying data for disk %d", idx)
5289
      result = self.rpc.call_blockdev_assemble(target_node, disk,
5290
                                               instance.name, True)
5291
      if result.fail_msg:
5292
        self.LogWarning("Can't assemble newly created disk %d: %s",
5293
                        idx, result.fail_msg)
5294
        errs.append(result.fail_msg)
5295
        break
5296
      dev_path = result.payload
5297
      result = self.rpc.call_blockdev_export(source_node, disk,
5298
                                             target_node, dev_path,
5299
                                             cluster_name)
5300
      if result.fail_msg:
5301
        self.LogWarning("Can't copy data over for disk %d: %s",
5302
                        idx, result.fail_msg)
5303
        errs.append(result.fail_msg)
5304
        break
5305

    
5306
    if errs:
5307
      self.LogWarning("Some disks failed to copy, aborting")
5308
      try:
5309
        _RemoveDisks(self, instance, target_node=target_node)
5310
      finally:
5311
        self.cfg.ReleaseDRBDMinors(instance.name)
5312
        raise errors.OpExecError("Errors during disk copy: %s" %
5313
                                 (",".join(errs),))
5314

    
5315
    instance.primary_node = target_node
5316
    self.cfg.Update(instance, feedback_fn)
5317

    
5318
    self.LogInfo("Removing the disks on the original node")
5319
    _RemoveDisks(self, instance, target_node=source_node)
5320

    
5321
    # Only start the instance if it's marked as up
5322
    if instance.admin_up:
5323
      self.LogInfo("Starting instance %s on node %s",
5324
                   instance.name, target_node)
5325

    
5326
      disks_ok, _ = _AssembleInstanceDisks(self, instance,
5327
                                           ignore_secondaries=True)
5328
      if not disks_ok:
5329
        _ShutdownInstanceDisks(self, instance)
5330
        raise errors.OpExecError("Can't activate the instance's disks")
5331

    
5332
      result = self.rpc.call_instance_start(target_node, instance, None, None)
5333
      msg = result.fail_msg
5334
      if msg:
5335
        _ShutdownInstanceDisks(self, instance)
5336
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
5337
                                 (instance.name, target_node, msg))
5338

    
5339

    
5340
class LUMigrateNode(LogicalUnit):
5341
  """Migrate all instances from a node.
5342

5343
  """
5344
  HPATH = "node-migrate"
5345
  HTYPE = constants.HTYPE_NODE
5346
  _OP_REQP = ["node_name", "live"]
5347
  REQ_BGL = False
5348

    
5349
  def ExpandNames(self):
5350
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
5351

    
5352
    self.needed_locks = {
5353
      locking.LEVEL_NODE: [self.op.node_name],
5354
      }
5355

    
5356
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
5357

    
5358
    # Create tasklets for migrating instances for all instances on this node
5359
    names = []
5360
    tasklets = []
5361

    
5362
    for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name):
5363
      logging.debug("Migrating instance %s", inst.name)
5364
      names.append(inst.name)
5365

    
5366
      tasklets.append(TLMigrateInstance(self, inst.name, self.op.live, False))
5367

    
5368
    self.tasklets = tasklets
5369

    
5370
    # Declare instance locks
5371
    self.needed_locks[locking.LEVEL_INSTANCE] = names
5372

    
5373
  def DeclareLocks(self, level):
5374
    if level == locking.LEVEL_NODE:
5375
      self._LockInstancesNodes()
5376

    
5377
  def BuildHooksEnv(self):
5378
    """Build hooks env.
5379

5380
    This runs on the master, the primary and all the secondaries.
5381

5382
    """
5383
    env = {
5384
      "NODE_NAME": self.op.node_name,
5385
      }
5386

    
5387
    nl = [self.cfg.GetMasterNode()]
5388

    
5389
    return (env, nl, nl)
5390

    
5391

    
5392
class TLMigrateInstance(Tasklet):
5393
  def __init__(self, lu, instance_name, live, cleanup):
5394
    """Initializes this class.
5395

5396
    """
5397
    Tasklet.__init__(self, lu)
5398

    
5399
    # Parameters
5400
    self.instance_name = instance_name
5401
    self.live = live
5402
    self.cleanup = cleanup
5403

    
5404
  def CheckPrereq(self):
5405
    """Check prerequisites.
5406

5407
    This checks that the instance is in the cluster.
5408

5409
    """
5410
    instance_name = _ExpandInstanceName(self.lu.cfg, self.instance_name)
5411
    instance = self.cfg.GetInstanceInfo(instance_name)
5412
    assert instance is not None
5413

    
5414
    if instance.disk_template != constants.DT_DRBD8:
5415
      raise errors.OpPrereqError("Instance's disk layout is not"
5416
                                 " drbd8, cannot migrate.", errors.ECODE_STATE)
5417

    
5418
    secondary_nodes = instance.secondary_nodes
5419
    if not secondary_nodes:
5420
      raise errors.ConfigurationError("No secondary node but using"
5421
                                      " drbd8 disk template")
5422

    
5423
    i_be = self.cfg.GetClusterInfo().FillBE(instance)
5424

    
5425
    target_node = secondary_nodes[0]
5426
    # check memory requirements on the secondary node
5427
    _CheckNodeFreeMemory(self, target_node, "migrating instance %s" %
5428
                         instance.name, i_be[constants.BE_MEMORY],
5429
                         instance.hypervisor)
5430

    
5431
    # check bridge existance
5432
    _CheckInstanceBridgesExist(self, instance, node=target_node)
5433

    
5434
    if not self.cleanup:
5435
      _CheckNodeNotDrained(self, target_node)
5436
      result = self.rpc.call_instance_migratable(instance.primary_node,
5437
                                                 instance)
5438
      result.Raise("Can't migrate, please use failover",
5439
                   prereq=True, ecode=errors.ECODE_STATE)
5440

    
5441
    self.instance = instance
5442

    
5443
  def _WaitUntilSync(self):
5444
    """Poll with custom rpc for disk sync.
5445

5446
    This uses our own step-based rpc call.
5447

5448
    """
5449
    self.feedback_fn("* wait until resync is done")
5450
    all_done = False
5451
    while not all_done:
5452
      all_done = True
5453
      result = self.rpc.call_drbd_wait_sync(self.all_nodes,
5454
                                            self.nodes_ip,
5455
                                            self.instance.disks)
5456
      min_percent = 100
5457
      for node, nres in result.items():
5458
        nres.Raise("Cannot resync disks on node %s" % node)
5459
        node_done, node_percent = nres.payload
5460
        all_done = all_done and node_done
5461
        if node_percent is not None:
5462
          min_percent = min(min_percent, node_percent)
5463
      if not all_done:
5464
        if min_percent < 100:
5465
          self.feedback_fn("   - progress: %.1f%%" % min_percent)
5466
        time.sleep(2)
5467

    
5468
  def _EnsureSecondary(self, node):
5469
    """Demote a node to secondary.
5470

5471
    """
5472
    self.feedback_fn("* switching node %s to secondary mode" % node)
5473

    
5474
    for dev in self.instance.disks:
5475
      self.cfg.SetDiskID(dev, node)
5476

    
5477
    result = self.rpc.call_blockdev_close(node, self.instance.name,
5478
                                          self.instance.disks)
5479
    result.Raise("Cannot change disk to secondary on node %s" % node)
5480

    
5481
  def _GoStandalone(self):
5482
    """Disconnect from the network.
5483

5484
    """
5485
    self.feedback_fn("* changing into standalone mode")
5486
    result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
5487
                                               self.instance.disks)
5488
    for node, nres in result.items():
5489
      nres.Raise("Cannot disconnect disks node %s" % node)
5490

    
5491
  def _GoReconnect(self, multimaster):
5492
    """Reconnect to the network.
5493

5494
    """
5495
    if multimaster:
5496
      msg = "dual-master"
5497
    else:
5498
      msg = "single-master"
5499
    self.feedback_fn("* changing disks into %s mode" % msg)
5500
    result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
5501
                                           self.instance.disks,
5502
                                           self.instance.name, multimaster)
5503
    for node, nres in result.items():
5504
      nres.Raise("Cannot change disks config on node %s" % node)
5505

    
5506
  def _ExecCleanup(self):
5507
    """Try to cleanup after a failed migration.
5508

5509
    The cleanup is done by:
5510
      - check that the instance is running only on one node
5511
        (and update the config if needed)
5512
      - change disks on its secondary node to secondary
5513
      - wait until disks are fully synchronized
5514
      - disconnect from the network
5515
      - change disks into single-master mode
5516
      - wait again until disks are fully synchronized
5517

5518
    """
5519
    instance = self.instance
5520
    target_node = self.target_node
5521
    source_node = self.source_node
5522

    
5523
    # check running on only one node
5524
    self.feedback_fn("* checking where the instance actually runs"
5525
                     " (if this hangs, the hypervisor might be in"
5526
                     " a bad state)")
5527
    ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
5528
    for node, result in ins_l.items():
5529
      result.Raise("Can't contact node %s" % node)
5530

    
5531
    runningon_source = instance.name in ins_l[source_node].payload
5532
    runningon_target = instance.name in ins_l[target_node].payload
5533

    
5534
    if runningon_source and runningon_target:
5535
      raise errors.OpExecError("Instance seems to be running on two nodes,"
5536
                               " or the hypervisor is confused. You will have"
5537
                               " to ensure manually that it runs only on one"
5538
                               " and restart this operation.")
5539

    
5540
    if not (runningon_source or runningon_target):
5541
      raise errors.OpExecError("Instance does not seem to be running at all."
5542
                               " In this case, it's safer to repair by"
5543
                               " running 'gnt-instance stop' to ensure disk"
5544
                               " shutdown, and then restarting it.")
5545

    
5546
    if runningon_target:
5547
      # the migration has actually succeeded, we need to update the config
5548
      self.feedback_fn("* instance running on secondary node (%s),"
5549
                       " updating config" % target_node)
5550
      instance.primary_node = target_node
5551
      self.cfg.Update(instance, self.feedback_fn)
5552
      demoted_node = source_node
5553
    else:
5554
      self.feedback_fn("* instance confirmed to be running on its"
5555
                       " primary node (%s)" % source_node)
5556
      demoted_node = target_node
5557

    
5558
    self._EnsureSecondary(demoted_node)
5559
    try:
5560
      self._WaitUntilSync()
5561
    except errors.OpExecError:
5562
      # we ignore here errors, since if the device is standalone, it
5563
      # won't be able to sync
5564
      pass
5565
    self._GoStandalone()
5566
    self._GoReconnect(False)
5567
    self._WaitUntilSync()
5568

    
5569
    self.feedback_fn("* done")
5570

    
5571
  def _RevertDiskStatus(self):
5572
    """Try to revert the disk status after a failed migration.
5573

5574
    """
5575
    target_node = self.target_node
5576
    try:
5577
      self._EnsureSecondary(target_node)
5578
      self._GoStandalone()
5579
      self._GoReconnect(False)
5580
      self._WaitUntilSync()
5581
    except errors.OpExecError, err:
5582
      self.lu.LogWarning("Migration failed and I can't reconnect the"
5583
                         " drives: error '%s'\n"
5584
                         "Please look and recover the instance status" %
5585
                         str(err))
5586

    
5587
  def _AbortMigration(self):
5588
    """Call the hypervisor code to abort a started migration.
5589

5590
    """
5591
    instance = self.instance
5592
    target_node = self.target_node
5593
    migration_info = self.migration_info
5594

    
5595
    abort_result = self.rpc.call_finalize_migration(target_node,
5596
                                                    instance,
5597
                                                    migration_info,
5598
                                                    False)
5599
    abort_msg = abort_result.fail_msg
5600
    if abort_msg:
5601
      logging.error("Aborting migration failed on target node %s: %s",
5602
                    target_node, abort_msg)
5603
      # Don't raise an exception here, as we stil have to try to revert the
5604
      # disk status, even if this step failed.
5605

    
5606
  def _ExecMigration(self):
5607
    """Migrate an instance.
5608

5609
    The migrate is done by:
5610
      - change the disks into dual-master mode
5611
      - wait until disks are fully synchronized again
5612
      - migrate the instance
5613
      - change disks on the new secondary node (the old primary) to secondary
5614
      - wait until disks are fully synchronized
5615
      - change disks into single-master mode
5616

5617
    """
5618
    instance = self.instance
5619
    target_node = self.target_node
5620
    source_node = self.source_node
5621

    
5622
    self.feedback_fn("* checking disk consistency between source and target")
5623
    for dev in instance.disks:
5624
      if not _CheckDiskConsistency(self, dev, target_node, False):
5625
        raise errors.OpExecError("Disk %s is degraded or not fully"
5626
                                 " synchronized on target node,"
5627
                                 " aborting migrate." % dev.iv_name)
5628

    
5629
    # First get the migration information from the remote node
5630
    result = self.rpc.call_migration_info(source_node, instance)
5631
    msg = result.fail_msg
5632
    if msg:
5633
      log_err = ("Failed fetching source migration information from %s: %s" %
5634
                 (source_node, msg))
5635
      logging.error(log_err)
5636
      raise errors.OpExecError(log_err)
5637

    
5638
    self.migration_info = migration_info = result.payload
5639

    
5640
    # Then switch the disks to master/master mode
5641
    self._EnsureSecondary(target_node)
5642
    self._GoStandalone()
5643
    self._GoReconnect(True)
5644
    self._WaitUntilSync()
5645

    
5646
    self.feedback_fn("* preparing %s to accept the instance" % target_node)
5647
    result = self.rpc.call_accept_instance(target_node,
5648
                                           instance,
5649
                                           migration_info,
5650
                                           self.nodes_ip[target_node])
5651

    
5652
    msg = result.fail_msg
5653
    if msg:
5654
      logging.error("Instance pre-migration failed, trying to revert"
5655
                    " disk status: %s", msg)
5656
      self.feedback_fn("Pre-migration failed, aborting")
5657
      self._AbortMigration()
5658
      self._RevertDiskStatus()
5659
      raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
5660
                               (instance.name, msg))
5661

    
5662
    self.feedback_fn("* migrating instance to %s" % target_node)
5663
    time.sleep(10)
5664
    result = self.rpc.call_instance_migrate(source_node, instance,
5665
                                            self.nodes_ip[target_node],
5666
                                            self.live)
5667
    msg = result.fail_msg
5668
    if msg:
5669
      logging.error("Instance migration failed, trying to revert"
5670
                    " disk status: %s", msg)
5671
      self.feedback_fn("Migration failed, aborting")
5672
      self._AbortMigration()
5673
      self._RevertDiskStatus()
5674
      raise errors.OpExecError("Could not migrate instance %s: %s" %
5675
                               (instance.name, msg))
5676
    time.sleep(10)
5677

    
5678
    instance.primary_node = target_node
5679
    # distribute new instance config to the other nodes
5680
    self.cfg.Update(instance, self.feedback_fn)
5681

    
5682
    result = self.rpc.call_finalize_migration(target_node,
5683
                                              instance,
5684
                                              migration_info,
5685
                                              True)
5686
    msg = result.fail_msg
5687
    if msg:
5688
      logging.error("Instance migration succeeded, but finalization failed:"
5689
                    " %s", msg)
5690
      raise errors.OpExecError("Could not finalize instance migration: %s" %
5691
                               msg)
5692

    
5693
    self._EnsureSecondary(source_node)
5694
    self._WaitUntilSync()
5695
    self._GoStandalone()
5696
    self._GoReconnect(False)
5697
    self._WaitUntilSync()
5698

    
5699
    self.feedback_fn("* done")
5700

    
5701
  def Exec(self, feedback_fn):
5702
    """Perform the migration.
5703

5704
    """
5705
    feedback_fn("Migrating instance %s" % self.instance.name)
5706

    
5707
    self.feedback_fn = feedback_fn
5708

    
5709
    self.source_node = self.instance.primary_node
5710
    self.target_node = self.instance.secondary_nodes[0]
5711
    self.all_nodes = [self.source_node, self.target_node]
5712
    self.nodes_ip = {
5713
      self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
5714
      self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
5715
      }
5716

    
5717
    if self.cleanup:
5718
      return self._ExecCleanup()
5719
    else:
5720
      return self._ExecMigration()
5721

    
5722

    
5723
def _CreateBlockDev(lu, node, instance, device, force_create,
5724
                    info, force_open):
5725
  """Create a tree of block devices on a given node.
5726

5727
  If this device type has to be created on secondaries, create it and
5728
  all its children.
5729

5730
  If not, just recurse to children keeping the same 'force' value.
5731

5732
  @param lu: the lu on whose behalf we execute
5733
  @param node: the node on which to create the device
5734
  @type instance: L{objects.Instance}
5735
  @param instance: the instance which owns the device
5736
  @type device: L{objects.Disk}
5737
  @param device: the device to create
5738
  @type force_create: boolean
5739
  @param force_create: whether to force creation of this device; this
5740
      will be change to True whenever we find a device which has
5741
      CreateOnSecondary() attribute
5742
  @param info: the extra 'metadata' we should attach to the device
5743
      (this will be represented as a LVM tag)
5744
  @type force_open: boolean
5745
  @param force_open: this parameter will be passes to the
5746
      L{backend.BlockdevCreate} function where it specifies
5747
      whether we run on primary or not, and it affects both
5748
      the child assembly and the device own Open() execution
5749

5750
  """
5751
  if device.CreateOnSecondary():
5752
    force_create = True
5753

    
5754
  if device.children:
5755
    for child in device.children:
5756
      _CreateBlockDev(lu, node, instance, child, force_create,
5757
                      info, force_open)
5758

    
5759
  if not force_create:
5760
    return
5761

    
5762
  _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
5763

    
5764

    
5765
def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
5766
  """Create a single block device on a given node.
5767

5768
  This will not recurse over children of the device, so they must be
5769
  created in advance.
5770

5771
  @param lu: the lu on whose behalf we execute
5772
  @param node: the node on which to create the device
5773
  @type instance: L{objects.Instance}
5774
  @param instance: the instance which owns the device
5775
  @type device: L{objects.Disk}
5776
  @param device: the device to create
5777
  @param info: the extra 'metadata' we should attach to the device
5778
      (this will be represented as a LVM tag)
5779
  @type force_open: boolean
5780
  @param force_open: this parameter will be passes to the
5781
      L{backend.BlockdevCreate} function where it specifies
5782
      whether we run on primary or not, and it affects both
5783
      the child assembly and the device own Open() execution
5784

5785
  """
5786
  lu.cfg.SetDiskID(device, node)
5787
  result = lu.rpc.call_blockdev_create(node, device, device.size,
5788
                                       instance.name, force_open, info)
5789
  result.Raise("Can't create block device %s on"
5790
               " node %s for instance %s" % (device, node, instance.name))
5791
  if device.physical_id is None:
5792
    device.physical_id = result.payload
5793

    
5794

    
5795
def _GenerateUniqueNames(lu, exts):
5796
  """Generate a suitable LV name.
5797

5798
  This will generate a logical volume name for the given instance.
5799

5800
  """
5801
  results = []
5802
  for val in exts:
5803
    new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
5804
    results.append("%s%s" % (new_id, val))
5805
  return results
5806

    
5807

    
5808
def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
5809
                         p_minor, s_minor):
5810
  """Generate a drbd8 device complete with its children.
5811

5812
  """
5813
  port = lu.cfg.AllocatePort()
5814
  vgname = lu.cfg.GetVGName()
5815
  shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
5816
  dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
5817
                          logical_id=(vgname, names[0]))
5818
  dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
5819
                          logical_id=(vgname, names[1]))
5820
  drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
5821
                          logical_id=(primary, secondary, port,
5822
                                      p_minor, s_minor,
5823
                                      shared_secret),
5824
                          children=[dev_data, dev_meta],
5825
                          iv_name=iv_name)
5826
  return drbd_dev
5827

    
5828

    
5829
def _GenerateDiskTemplate(lu, template_name,
5830
                          instance_name, primary_node,
5831
                          secondary_nodes, disk_info,
5832
                          file_storage_dir, file_driver,
5833
                          base_index):
5834
  """Generate the entire disk layout for a given template type.
5835

5836
  """
5837
  #TODO: compute space requirements
5838

    
5839
  vgname = lu.cfg.GetVGName()
5840
  disk_count = len(disk_info)
5841
  disks = []
5842
  if template_name == constants.DT_DISKLESS:
5843
    pass
5844
  elif template_name == constants.DT_PLAIN:
5845
    if len(secondary_nodes) != 0:
5846
      raise errors.ProgrammerError("Wrong template configuration")
5847

    
5848
    names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
5849
                                      for i in range(disk_count)])
5850
    for idx, disk in enumerate(disk_info):
5851
      disk_index = idx + base_index
5852
      disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
5853
                              logical_id=(vgname, names[idx]),
5854
                              iv_name="disk/%d" % disk_index,
5855
                              mode=disk["mode"])
5856
      disks.append(disk_dev)
5857
  elif template_name == constants.DT_DRBD8:
5858
    if len(secondary_nodes) != 1:
5859
      raise errors.ProgrammerError("Wrong template configuration")
5860
    remote_node = secondary_nodes[0]
5861
    minors = lu.cfg.AllocateDRBDMinor(
5862
      [primary_node, remote_node] * len(disk_info), instance_name)
5863

    
5864
    names = []
5865
    for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
5866
                                               for i in range(disk_count)]):
5867
      names.append(lv_prefix + "_data")
5868
      names.append(lv_prefix + "_meta")
5869
    for idx, disk in enumerate(disk_info):
5870
      disk_index = idx + base_index
5871
      disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
5872
                                      disk["size"], names[idx*2:idx*2+2],
5873
                                      "disk/%d" % disk_index,
5874
                                      minors[idx*2], minors[idx*2+1])
5875
      disk_dev.mode = disk["mode"]
5876
      disks.append(disk_dev)
5877
  elif template_name == constants.DT_FILE:
5878
    if len(secondary_nodes) != 0:
5879
      raise errors.ProgrammerError("Wrong template configuration")
5880

    
5881
    _RequireFileStorage()
5882

    
5883
    for idx, disk in enumerate(disk_info):
5884
      disk_index = idx + base_index
5885
      disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
5886
                              iv_name="disk/%d" % disk_index,
5887
                              logical_id=(file_driver,
5888
                                          "%s/disk%d" % (file_storage_dir,
5889
                                                         disk_index)),
5890
                              mode=disk["mode"])
5891
      disks.append(disk_dev)
5892
  else:
5893
    raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
5894
  return disks
5895

    
5896

    
5897
def _GetInstanceInfoText(instance):
5898
  """Compute that text that should be added to the disk's metadata.
5899

5900
  """
5901
  return "originstname+%s" % instance.name
5902

    
5903

    
5904
def _CreateDisks(lu, instance, to_skip=None, target_node=None):
5905
  """Create all disks for an instance.
5906

5907
  This abstracts away some work from AddInstance.
5908

5909
  @type lu: L{LogicalUnit}
5910
  @param lu: the logical unit on whose behalf we execute
5911
  @type instance: L{objects.Instance}
5912
  @param instance: the instance whose disks we should create
5913
  @type to_skip: list
5914
  @param to_skip: list of indices to skip
5915
  @type target_node: string
5916
  @param target_node: if passed, overrides the target node for creation
5917
  @rtype: boolean
5918
  @return: the success of the creation
5919

5920
  """
5921
  info = _GetInstanceInfoText(instance)
5922
  if target_node is None:
5923
    pnode = instance.primary_node
5924
    all_nodes = instance.all_nodes
5925
  else:
5926
    pnode = target_node
5927
    all_nodes = [pnode]
5928

    
5929
  if instance.disk_template == constants.DT_FILE:
5930
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
5931
    result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
5932

    
5933
    result.Raise("Failed to create directory '%s' on"
5934
                 " node %s" % (file_storage_dir, pnode))
5935

    
5936
  # Note: this needs to be kept in sync with adding of disks in
5937
  # LUSetInstanceParams
5938
  for idx, device in enumerate(instance.disks):
5939
    if to_skip and idx in to_skip:
5940
      continue
5941
    logging.info("Creating volume %s for instance %s",
5942
                 device.iv_name, instance.name)
5943
    #HARDCODE
5944
    for node in all_nodes:
5945
      f_create = node == pnode
5946
      _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
5947

    
5948

    
5949
def _RemoveDisks(lu, instance, target_node=None):
5950
  """Remove all disks for an instance.
5951

5952
  This abstracts away some work from `AddInstance()` and
5953
  `RemoveInstance()`. Note that in case some of the devices couldn't
5954
  be removed, the removal will continue with the other ones (compare
5955
  with `_CreateDisks()`).
5956

5957
  @type lu: L{LogicalUnit}
5958
  @param lu: the logical unit on whose behalf we execute
5959
  @type instance: L{objects.Instance}
5960
  @param instance: the instance whose disks we should remove
5961
  @type target_node: string
5962
  @param target_node: used to override the node on which to remove the disks
5963
  @rtype: boolean
5964
  @return: the success of the removal
5965

5966
  """
5967
  logging.info("Removing block devices for instance %s", instance.name)
5968

    
5969
  all_result = True
5970
  for device in instance.disks:
5971
    if target_node:
5972
      edata = [(target_node, device)]
5973
    else:
5974
      edata = device.ComputeNodeTree(instance.primary_node)
5975
    for node, disk in edata:
5976
      lu.cfg.SetDiskID(disk, node)
5977
      msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
5978
      if msg:
5979
        lu.LogWarning("Could not remove block device %s on node %s,"
5980
                      " continuing anyway: %s", device.iv_name, node, msg)
5981
        all_result = False
5982

    
5983
  if instance.disk_template == constants.DT_FILE:
5984
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
5985
    if target_node:
5986
      tgt = target_node
5987
    else:
5988
      tgt = instance.primary_node
5989
    result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
5990
    if result.fail_msg:
5991
      lu.LogWarning("Could not remove directory '%s' on node %s: %s",
5992
                    file_storage_dir, instance.primary_node, result.fail_msg)
5993
      all_result = False
5994

    
5995
  return all_result
5996

    
5997

    
5998
def _ComputeDiskSize(disk_template, disks):
5999
  """Compute disk size requirements in the volume group
6000

6001
  """
6002
  # Required free disk space as a function of disk and swap space
6003
  req_size_dict = {
6004
    constants.DT_DISKLESS: None,
6005
    constants.DT_PLAIN: sum(d["size"] for d in disks),
6006
    # 128 MB are added for drbd metadata for each disk
6007
    constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
6008
    constants.DT_FILE: None,
6009
  }
6010

    
6011
  if disk_template not in req_size_dict:
6012
    raise errors.ProgrammerError("Disk template '%s' size requirement"
6013
                                 " is unknown" %  disk_template)
6014

    
6015
  return req_size_dict[disk_template]
6016

    
6017

    
6018
def _CheckHVParams(lu, nodenames, hvname, hvparams):
6019
  """Hypervisor parameter validation.
6020

6021
  This function abstract the hypervisor parameter validation to be
6022
  used in both instance create and instance modify.
6023

6024
  @type lu: L{LogicalUnit}
6025
  @param lu: the logical unit for which we check
6026
  @type nodenames: list
6027
  @param nodenames: the list of nodes on which we should check
6028
  @type hvname: string
6029
  @param hvname: the name of the hypervisor we should use
6030
  @type hvparams: dict
6031
  @param hvparams: the parameters which we need to check
6032
  @raise errors.OpPrereqError: if the parameters are not valid
6033

6034
  """
6035
  hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
6036
                                                  hvname,
6037
                                                  hvparams)
6038
  for node in nodenames:
6039
    info = hvinfo[node]
6040
    if info.offline:
6041
      continue
6042
    info.Raise("Hypervisor parameter validation failed on node %s" % node)
6043

    
6044

    
6045
class LUCreateInstance(LogicalUnit):
6046
  """Create an instance.
6047

6048
  """
6049
  HPATH = "instance-add"
6050
  HTYPE = constants.HTYPE_INSTANCE
6051
  _OP_REQP = ["instance_name", "disks",
6052
              "mode", "start",
6053
              "wait_for_sync", "ip_check", "nics",
6054
              "hvparams", "beparams"]
6055
  REQ_BGL = False
6056

    
6057
  def CheckArguments(self):
6058
    """Check arguments.
6059

6060
    """
6061
    # set optional parameters to none if they don't exist
6062
    for attr in ["pnode", "snode", "iallocator", "hypervisor",
6063
                 "disk_template", "identify_defaults"]:
6064
      if not hasattr(self.op, attr):
6065
        setattr(self.op, attr, None)
6066

    
6067
    # do not require name_check to ease forward/backward compatibility
6068
    # for tools
6069
    if not hasattr(self.op, "name_check"):
6070
      self.op.name_check = True
6071
    if not hasattr(self.op, "no_install"):
6072
      self.op.no_install = False
6073
    if self.op.no_install and self.op.start:
6074
      self.LogInfo("No-installation mode selected, disabling startup")
6075
      self.op.start = False
6076
    # validate/normalize the instance name
6077
    self.op.instance_name = utils.HostInfo.NormalizeName(self.op.instance_name)
6078
    if self.op.ip_check and not self.op.name_check:
6079
      # TODO: make the ip check more flexible and not depend on the name check
6080
      raise errors.OpPrereqError("Cannot do ip checks without a name check",
6081
                                 errors.ECODE_INVAL)
6082
    # check disk information: either all adopt, or no adopt
6083
    has_adopt = has_no_adopt = False
6084
    for disk in self.op.disks:
6085
      if "adopt" in disk:
6086
        has_adopt = True
6087
      else:
6088
        has_no_adopt = True
6089
    if has_adopt and has_no_adopt:
6090
      raise errors.OpPrereqError("Either all disks are adopted or none is",
6091
                                 errors.ECODE_INVAL)
6092
    if has_adopt:
6093
      if self.op.disk_template != constants.DT_PLAIN:
6094
        raise errors.OpPrereqError("Disk adoption is only supported for the"
6095
                                   " 'plain' disk template",
6096
                                   errors.ECODE_INVAL)
6097
      if self.op.iallocator is not None:
6098
        raise errors.OpPrereqError("Disk adoption not allowed with an"
6099
                                   " iallocator script", errors.ECODE_INVAL)
6100
      if self.op.mode == constants.INSTANCE_IMPORT:
6101
        raise errors.OpPrereqError("Disk adoption not allowed for"
6102
                                   " instance import", errors.ECODE_INVAL)
6103

    
6104
    self.adopt_disks = has_adopt
6105

    
6106
    # verify creation mode
6107
    if self.op.mode not in constants.INSTANCE_CREATE_MODES:
6108
      raise errors.OpPrereqError("Invalid instance creation mode '%s'" %
6109
                                 self.op.mode, errors.ECODE_INVAL)
6110

    
6111
    # instance name verification
6112
    if self.op.name_check:
6113
      self.hostname1 = utils.GetHostInfo(self.op.instance_name)
6114
      self.op.instance_name = self.hostname1.name
6115
      # used in CheckPrereq for ip ping check
6116
      self.check_ip = self.hostname1.ip
6117
    elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
6118
      raise errors.OpPrereqError("Remote imports require names to be checked" %
6119
                                 errors.ECODE_INVAL)
6120
    else:
6121
      self.check_ip = None
6122

    
6123
    # file storage checks
6124
    if (self.op.file_driver and
6125
        not self.op.file_driver in constants.FILE_DRIVER):
6126
      raise errors.OpPrereqError("Invalid file driver name '%s'" %
6127
                                 self.op.file_driver, errors.ECODE_INVAL)
6128

    
6129
    if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
6130
      raise errors.OpPrereqError("File storage directory path not absolute",
6131
                                 errors.ECODE_INVAL)
6132

    
6133
    ### Node/iallocator related checks
6134
    if [self.op.iallocator, self.op.pnode].count(None) != 1:
6135
      raise errors.OpPrereqError("One and only one of iallocator and primary"
6136
                                 " node must be given",
6137
                                 errors.ECODE_INVAL)
6138

    
6139
    self._cds = _GetClusterDomainSecret()
6140

    
6141
    if self.op.mode == constants.INSTANCE_IMPORT:
6142
      # On import force_variant must be True, because if we forced it at
6143
      # initial install, our only chance when importing it back is that it
6144
      # works again!
6145
      self.op.force_variant = True
6146

    
6147
      if self.op.no_install:
6148
        self.LogInfo("No-installation mode has no effect during import")
6149

    
6150
    elif self.op.mode == constants.INSTANCE_CREATE:
6151
      if getattr(self.op, "os_type", None) is None:
6152
        raise errors.OpPrereqError("No guest OS specified",
6153
                                   errors.ECODE_INVAL)
6154
      self.op.force_variant = getattr(self.op, "force_variant", False)
6155
      if self.op.disk_template is None:
6156
        raise errors.OpPrereqError("No disk template specified",
6157
                                   errors.ECODE_INVAL)
6158

    
6159
    elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
6160
      # Check handshake to ensure both clusters have the same domain secret
6161
      src_handshake = getattr(self.op, "source_handshake", None)
6162
      if not src_handshake:
6163
        raise errors.OpPrereqError("Missing source handshake",
6164
                                   errors.ECODE_INVAL)
6165

    
6166
      errmsg = masterd.instance.CheckRemoteExportHandshake(self._cds,
6167
                                                           src_handshake)
6168
      if errmsg:
6169
        raise errors.OpPrereqError("Invalid handshake: %s" % errmsg,
6170
                                   errors.ECODE_INVAL)
6171

    
6172
      # Load and check source CA
6173
      self.source_x509_ca_pem = getattr(self.op, "source_x509_ca", None)
6174
      if not self.source_x509_ca_pem:
6175
        raise errors.OpPrereqError("Missing source X509 CA",
6176
                                   errors.ECODE_INVAL)
6177

    
6178
      try:
6179
        (cert, _) = utils.LoadSignedX509Certificate(self.source_x509_ca_pem,
6180
                                                    self._cds)
6181
      except OpenSSL.crypto.Error, err:
6182
        raise errors.OpPrereqError("Unable to load source X509 CA (%s)" %
6183
                                   (err, ), errors.ECODE_INVAL)
6184

    
6185
      (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
6186
      if errcode is not None:
6187
        raise errors.OpPrereqError("Invalid source X509 CA (%s)" % (msg, ),
6188
                                   errors.ECODE_INVAL)
6189

    
6190
      self.source_x509_ca = cert
6191

    
6192
      src_instance_name = getattr(self.op, "source_instance_name", None)
6193
      if not src_instance_name:
6194
        raise errors.OpPrereqError("Missing source instance name",
6195
                                   errors.ECODE_INVAL)
6196

    
6197
      self.source_instance_name = \
6198
        utils.GetHostInfo(utils.HostInfo.NormalizeName(src_instance_name)).name
6199

    
6200
    else:
6201
      raise errors.OpPrereqError("Invalid instance creation mode %r" %
6202
                                 self.op.mode, errors.ECODE_INVAL)
6203

    
6204
  def ExpandNames(self):
6205
    """ExpandNames for CreateInstance.
6206

6207
    Figure out the right locks for instance creation.
6208

6209
    """
6210
    self.needed_locks = {}
6211

    
6212
    instance_name = self.op.instance_name
6213
    # this is just a preventive check, but someone might still add this
6214
    # instance in the meantime, and creation will fail at lock-add time
6215
    if instance_name in self.cfg.GetInstanceList():
6216
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
6217
                                 instance_name, errors.ECODE_EXISTS)
6218

    
6219
    self.add_locks[locking.LEVEL_INSTANCE] = instance_name
6220

    
6221
    if self.op.iallocator:
6222
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6223
    else:
6224
      self.op.pnode = _ExpandNodeName(self.cfg, self.op.pnode)
6225
      nodelist = [self.op.pnode]
6226
      if self.op.snode is not None:
6227
        self.op.snode = _ExpandNodeName(self.cfg, self.op.snode)
6228
        nodelist.append(self.op.snode)
6229
      self.needed_locks[locking.LEVEL_NODE] = nodelist
6230

    
6231
    # in case of import lock the source node too
6232
    if self.op.mode == constants.INSTANCE_IMPORT:
6233
      src_node = getattr(self.op, "src_node", None)
6234
      src_path = getattr(self.op, "src_path", None)
6235

    
6236
      if src_path is None:
6237
        self.op.src_path = src_path = self.op.instance_name
6238

    
6239
      if src_node is None:
6240
        self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6241
        self.op.src_node = None
6242
        if os.path.isabs(src_path):
6243
          raise errors.OpPrereqError("Importing an instance from an absolute"
6244
                                     " path requires a source node option.",
6245
                                     errors.ECODE_INVAL)
6246
      else:
6247
        self.op.src_node = src_node = _ExpandNodeName(self.cfg, src_node)
6248
        if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
6249
          self.needed_locks[locking.LEVEL_NODE].append(src_node)
6250
        if not os.path.isabs(src_path):
6251
          self.op.src_path = src_path = \
6252
            utils.PathJoin(constants.EXPORT_DIR, src_path)
6253

    
6254
  def _RunAllocator(self):
6255
    """Run the allocator based on input opcode.
6256

6257
    """
6258
    nics = [n.ToDict() for n in self.nics]
6259
    ial = IAllocator(self.cfg, self.rpc,
6260
                     mode=constants.IALLOCATOR_MODE_ALLOC,
6261
                     name=self.op.instance_name,
6262
                     disk_template=self.op.disk_template,
6263
                     tags=[],
6264
                     os=self.op.os_type,
6265
                     vcpus=self.be_full[constants.BE_VCPUS],
6266
                     mem_size=self.be_full[constants.BE_MEMORY],
6267
                     disks=self.disks,
6268
                     nics=nics,
6269
                     hypervisor=self.op.hypervisor,
6270
                     )
6271

    
6272
    ial.Run(self.op.iallocator)
6273

    
6274
    if not ial.success:
6275
      raise errors.OpPrereqError("Can't compute nodes using"
6276
                                 " iallocator '%s': %s" %
6277
                                 (self.op.iallocator, ial.info),
6278
                                 errors.ECODE_NORES)
6279
    if len(ial.result) != ial.required_nodes:
6280
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
6281
                                 " of nodes (%s), required %s" %
6282
                                 (self.op.iallocator, len(ial.result),
6283
                                  ial.required_nodes), errors.ECODE_FAULT)
6284
    self.op.pnode = ial.result[0]
6285
    self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
6286
                 self.op.instance_name, self.op.iallocator,
6287
                 utils.CommaJoin(ial.result))
6288
    if ial.required_nodes == 2:
6289
      self.op.snode = ial.result[1]
6290

    
6291
  def BuildHooksEnv(self):
6292
    """Build hooks env.
6293

6294
    This runs on master, primary and secondary nodes of the instance.
6295

6296
    """
6297
    env = {
6298
      "ADD_MODE": self.op.mode,
6299
      }
6300
    if self.op.mode == constants.INSTANCE_IMPORT:
6301
      env["SRC_NODE"] = self.op.src_node
6302
      env["SRC_PATH"] = self.op.src_path
6303
      env["SRC_IMAGES"] = self.src_images
6304

    
6305
    env.update(_BuildInstanceHookEnv(
6306
      name=self.op.instance_name,
6307
      primary_node=self.op.pnode,
6308
      secondary_nodes=self.secondaries,
6309
      status=self.op.start,
6310
      os_type=self.op.os_type,
6311
      memory=self.be_full[constants.BE_MEMORY],
6312
      vcpus=self.be_full[constants.BE_VCPUS],
6313
      nics=_NICListToTuple(self, self.nics),
6314
      disk_template=self.op.disk_template,
6315
      disks=[(d["size"], d["mode"]) for d in self.disks],
6316
      bep=self.be_full,
6317
      hvp=self.hv_full,
6318
      hypervisor_name=self.op.hypervisor,
6319
    ))
6320

    
6321
    nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
6322
          self.secondaries)
6323
    return env, nl, nl
6324

    
6325
  def _ReadExportInfo(self):
6326
    """Reads the export information from disk.
6327

6328
    It will override the opcode source node and path with the actual
6329
    information, if these two were not specified before.
6330

6331
    @return: the export information
6332

6333
    """
6334
    assert self.op.mode == constants.INSTANCE_IMPORT
6335

    
6336
    src_node = self.op.src_node
6337
    src_path = self.op.src_path
6338

    
6339
    if src_node is None:
6340
      locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
6341
      exp_list = self.rpc.call_export_list(locked_nodes)
6342
      found = False
6343
      for node in exp_list:
6344
        if exp_list[node].fail_msg:
6345
          continue
6346
        if src_path in exp_list[node].payload:
6347
          found = True
6348
          self.op.src_node = src_node = node
6349
          self.op.src_path = src_path = utils.PathJoin(constants.EXPORT_DIR,
6350
                                                       src_path)
6351
          break
6352
      if not found:
6353
        raise errors.OpPrereqError("No export found for relative path %s" %
6354
                                    src_path, errors.ECODE_INVAL)
6355

    
6356
    _CheckNodeOnline(self, src_node)
6357
    result = self.rpc.call_export_info(src_node, src_path)
6358
    result.Raise("No export or invalid export found in dir %s" % src_path)
6359

    
6360
    export_info = objects.SerializableConfigParser.Loads(str(result.payload))
6361
    if not export_info.has_section(constants.INISECT_EXP):
6362
      raise errors.ProgrammerError("Corrupted export config",
6363
                                   errors.ECODE_ENVIRON)
6364

    
6365
    ei_version = export_info.get(constants.INISECT_EXP, "version")
6366
    if (int(ei_version) != constants.EXPORT_VERSION):
6367
      raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
6368
                                 (ei_version, constants.EXPORT_VERSION),
6369
                                 errors.ECODE_ENVIRON)
6370
    return export_info
6371

    
6372
  def _ReadExportParams(self, einfo):
6373
    """Use export parameters as defaults.
6374

6375
    In case the opcode doesn't specify (as in override) some instance
6376
    parameters, then try to use them from the export information, if
6377
    that declares them.
6378

6379
    """
6380
    self.op.os_type = einfo.get(constants.INISECT_EXP, "os")
6381

    
6382
    if self.op.disk_template is None:
6383
      if einfo.has_option(constants.INISECT_INS, "disk_template"):
6384
        self.op.disk_template = einfo.get(constants.INISECT_INS,
6385
                                          "disk_template")
6386
      else:
6387
        raise errors.OpPrereqError("No disk template specified and the export"
6388
                                   " is missing the disk_template information",
6389
                                   errors.ECODE_INVAL)
6390

    
6391
    if not self.op.disks:
6392
      if einfo.has_option(constants.INISECT_INS, "disk_count"):
6393
        disks = []
6394
        # TODO: import the disk iv_name too
6395
        for idx in range(einfo.getint(constants.INISECT_INS, "disk_count")):
6396
          disk_sz = einfo.getint(constants.INISECT_INS, "disk%d_size" % idx)
6397
          disks.append({"size": disk_sz})
6398
        self.op.disks = disks
6399
      else:
6400
        raise errors.OpPrereqError("No disk info specified and the export"
6401
                                   " is missing the disk information",
6402
                                   errors.ECODE_INVAL)
6403

    
6404
    if (not self.op.nics and
6405
        einfo.has_option(constants.INISECT_INS, "nic_count")):
6406
      nics = []
6407
      for idx in range(einfo.getint(constants.INISECT_INS, "nic_count")):
6408
        ndict = {}
6409
        for name in list(constants.NICS_PARAMETERS) + ["ip", "mac"]:
6410
          v = einfo.get(constants.INISECT_INS, "nic%d_%s" % (idx, name))
6411
          ndict[name] = v
6412
        nics.append(ndict)
6413
      self.op.nics = nics
6414

    
6415
    if (self.op.hypervisor is None and
6416
        einfo.has_option(constants.INISECT_INS, "hypervisor")):
6417
      self.op.hypervisor = einfo.get(constants.INISECT_INS, "hypervisor")
6418
    if einfo.has_section(constants.INISECT_HYP):
6419
      # use the export parameters but do not override the ones
6420
      # specified by the user
6421
      for name, value in einfo.items(constants.INISECT_HYP):
6422
        if name not in self.op.hvparams:
6423
          self.op.hvparams[name] = value
6424

    
6425
    if einfo.has_section(constants.INISECT_BEP):
6426
      # use the parameters, without overriding
6427
      for name, value in einfo.items(constants.INISECT_BEP):
6428
        if name not in self.op.beparams:
6429
          self.op.beparams[name] = value
6430
    else:
6431
      # try to read the parameters old style, from the main section
6432
      for name in constants.BES_PARAMETERS:
6433
        if (name not in self.op.beparams and
6434
            einfo.has_option(constants.INISECT_INS, name)):
6435
          self.op.beparams[name] = einfo.get(constants.INISECT_INS, name)
6436

    
6437
  def _RevertToDefaults(self, cluster):
6438
    """Revert the instance parameters to the default values.
6439

6440
    """
6441
    # hvparams
6442
    hv_defs = cluster.GetHVDefaults(self.op.hypervisor, self.op.os_type)
6443
    for name in self.op.hvparams.keys():
6444
      if name in hv_defs and hv_defs[name] == self.op.hvparams[name]:
6445
        del self.op.hvparams[name]
6446
    # beparams
6447
    be_defs = cluster.beparams.get(constants.PP_DEFAULT, {})
6448
    for name in self.op.beparams.keys():
6449
      if name in be_defs and be_defs[name] == self.op.beparams[name]:
6450
        del self.op.beparams[name]
6451
    # nic params
6452
    nic_defs = cluster.nicparams.get(constants.PP_DEFAULT, {})
6453
    for nic in self.op.nics:
6454
      for name in constants.NICS_PARAMETERS:
6455
        if name in nic and name in nic_defs and nic[name] == nic_defs[name]:
6456
          del nic[name]
6457

    
6458
  def CheckPrereq(self):
6459
    """Check prerequisites.
6460

6461
    """
6462
    if self.op.mode == constants.INSTANCE_IMPORT:
6463
      export_info = self._ReadExportInfo()
6464
      self._ReadExportParams(export_info)
6465

    
6466
    _CheckDiskTemplate(self.op.disk_template)
6467

    
6468
    if (not self.cfg.GetVGName() and
6469
        self.op.disk_template not in constants.DTS_NOT_LVM):
6470
      raise errors.OpPrereqError("Cluster does not support lvm-based"
6471
                                 " instances", errors.ECODE_STATE)
6472

    
6473
    if self.op.hypervisor is None:
6474
      self.op.hypervisor = self.cfg.GetHypervisorType()
6475

    
6476
    cluster = self.cfg.GetClusterInfo()
6477
    enabled_hvs = cluster.enabled_hypervisors
6478
    if self.op.hypervisor not in enabled_hvs:
6479
      raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
6480
                                 " cluster (%s)" % (self.op.hypervisor,
6481
                                  ",".join(enabled_hvs)),
6482
                                 errors.ECODE_STATE)
6483

    
6484
    # check hypervisor parameter syntax (locally)
6485
    utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
6486
    filled_hvp = objects.FillDict(cluster.GetHVDefaults(self.op.hypervisor,
6487
                                                        self.op.os_type),
6488
                                  self.op.hvparams)
6489
    hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
6490
    hv_type.CheckParameterSyntax(filled_hvp)
6491
    self.hv_full = filled_hvp
6492
    # check that we don't specify global parameters on an instance
6493
    _CheckGlobalHvParams(self.op.hvparams)
6494

    
6495
    # fill and remember the beparams dict
6496
    utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
6497
    self.be_full = objects.FillDict(cluster.beparams[constants.PP_DEFAULT],
6498
                                    self.op.beparams)
6499

    
6500
    # now that hvp/bep are in final format, let's reset to defaults,
6501
    # if told to do so
6502
    if self.op.identify_defaults:
6503
      self._RevertToDefaults(cluster)
6504

    
6505
    # NIC buildup
6506
    self.nics = []
6507
    for idx, nic in enumerate(self.op.nics):
6508
      nic_mode_req = nic.get("mode", None)
6509
      nic_mode = nic_mode_req
6510
      if nic_mode is None:
6511
        nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
6512

    
6513
      # in routed mode, for the first nic, the default ip is 'auto'
6514
      if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
6515
        default_ip_mode = constants.VALUE_AUTO
6516
      else:
6517
        default_ip_mode = constants.VALUE_NONE
6518

    
6519
      # ip validity checks
6520
      ip = nic.get("ip", default_ip_mode)
6521
      if ip is None or ip.lower() == constants.VALUE_NONE:
6522
        nic_ip = None
6523
      elif ip.lower() == constants.VALUE_AUTO:
6524
        if not self.op.name_check:
6525
          raise errors.OpPrereqError("IP address set to auto but name checks"
6526
                                     " have been skipped. Aborting.",
6527
                                     errors.ECODE_INVAL)
6528
        nic_ip = self.hostname1.ip
6529
      else:
6530
        if not utils.IsValidIP(ip):
6531
          raise errors.OpPrereqError("Given IP address '%s' doesn't look"
6532
                                     " like a valid IP" % ip,
6533
                                     errors.ECODE_INVAL)
6534
        nic_ip = ip
6535

    
6536
      # TODO: check the ip address for uniqueness
6537
      if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
6538
        raise errors.OpPrereqError("Routed nic mode requires an ip address",
6539
                                   errors.ECODE_INVAL)
6540

    
6541
      # MAC address verification
6542
      mac = nic.get("mac", constants.VALUE_AUTO)
6543
      if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
6544
        mac = utils.NormalizeAndValidateMac(mac)
6545

    
6546
        try:
6547
          self.cfg.ReserveMAC(mac, self.proc.GetECId())
6548
        except errors.ReservationError:
6549
          raise errors.OpPrereqError("MAC address %s already in use"
6550
                                     " in cluster" % mac,
6551
                                     errors.ECODE_NOTUNIQUE)
6552

    
6553
      # bridge verification
6554
      bridge = nic.get("bridge", None)
6555
      link = nic.get("link", None)
6556
      if bridge and link:
6557
        raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
6558
                                   " at the same time", errors.ECODE_INVAL)
6559
      elif bridge and nic_mode == constants.NIC_MODE_ROUTED:
6560
        raise errors.OpPrereqError("Cannot pass 'bridge' on a routed nic",
6561
                                   errors.ECODE_INVAL)
6562
      elif bridge:
6563
        link = bridge
6564

    
6565
      nicparams = {}
6566
      if nic_mode_req:
6567
        nicparams[constants.NIC_MODE] = nic_mode_req
6568
      if link:
6569
        nicparams[constants.NIC_LINK] = link
6570

    
6571
      check_params = objects.FillDict(cluster.nicparams[constants.PP_DEFAULT],
6572
                                      nicparams)
6573
      objects.NIC.CheckParameterSyntax(check_params)
6574
      self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
6575

    
6576
    # disk checks/pre-build
6577
    self.disks = []
6578
    for disk in self.op.disks:
6579
      mode = disk.get("mode", constants.DISK_RDWR)
6580
      if mode not in constants.DISK_ACCESS_SET:
6581
        raise errors.OpPrereqError("Invalid disk access mode '%s'" %
6582
                                   mode, errors.ECODE_INVAL)
6583
      size = disk.get("size", None)
6584
      if size is None:
6585
        raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
6586
      try:
6587
        size = int(size)
6588
      except (TypeError, ValueError):
6589
        raise errors.OpPrereqError("Invalid disk size '%s'" % size,
6590
                                   errors.ECODE_INVAL)
6591
      new_disk = {"size": size, "mode": mode}
6592
      if "adopt" in disk:
6593
        new_disk["adopt"] = disk["adopt"]
6594
      self.disks.append(new_disk)
6595

    
6596
    if self.op.mode == constants.INSTANCE_IMPORT:
6597

    
6598
      # Check that the new instance doesn't have less disks than the export
6599
      instance_disks = len(self.disks)
6600
      export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
6601
      if instance_disks < export_disks:
6602
        raise errors.OpPrereqError("Not enough disks to import."
6603
                                   " (instance: %d, export: %d)" %
6604
                                   (instance_disks, export_disks),
6605
                                   errors.ECODE_INVAL)
6606

    
6607
      disk_images = []
6608
      for idx in range(export_disks):
6609
        option = 'disk%d_dump' % idx
6610
        if export_info.has_option(constants.INISECT_INS, option):
6611
          # FIXME: are the old os-es, disk sizes, etc. useful?
6612
          export_name = export_info.get(constants.INISECT_INS, option)
6613
          image = utils.PathJoin(self.op.src_path, export_name)
6614
          disk_images.append(image)
6615
        else:
6616
          disk_images.append(False)
6617

    
6618
      self.src_images = disk_images
6619

    
6620
      old_name = export_info.get(constants.INISECT_INS, 'name')
6621
      try:
6622
        exp_nic_count = export_info.getint(constants.INISECT_INS, 'nic_count')
6623
      except (TypeError, ValueError), err:
6624
        raise errors.OpPrereqError("Invalid export file, nic_count is not"
6625
                                   " an integer: %s" % str(err),
6626
                                   errors.ECODE_STATE)
6627
      if self.op.instance_name == old_name:
6628
        for idx, nic in enumerate(self.nics):
6629
          if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
6630
            nic_mac_ini = 'nic%d_mac' % idx
6631
            nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
6632

    
6633
    # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
6634

    
6635
    # ip ping checks (we use the same ip that was resolved in ExpandNames)
6636
    if self.op.ip_check:
6637
      if utils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
6638
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
6639
                                   (self.check_ip, self.op.instance_name),
6640
                                   errors.ECODE_NOTUNIQUE)
6641

    
6642
    #### mac address generation
6643
    # By generating here the mac address both the allocator and the hooks get
6644
    # the real final mac address rather than the 'auto' or 'generate' value.
6645
    # There is a race condition between the generation and the instance object
6646
    # creation, which means that we know the mac is valid now, but we're not
6647
    # sure it will be when we actually add the instance. If things go bad
6648
    # adding the instance will abort because of a duplicate mac, and the
6649
    # creation job will fail.
6650
    for nic in self.nics:
6651
      if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
6652
        nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
6653

    
6654
    #### allocator run
6655

    
6656
    if self.op.iallocator is not None:
6657
      self._RunAllocator()
6658

    
6659
    #### node related checks
6660

    
6661
    # check primary node
6662
    self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
6663
    assert self.pnode is not None, \
6664
      "Cannot retrieve locked node %s" % self.op.pnode
6665
    if pnode.offline:
6666
      raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
6667
                                 pnode.name, errors.ECODE_STATE)
6668
    if pnode.drained:
6669
      raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
6670
                                 pnode.name, errors.ECODE_STATE)
6671

    
6672
    self.secondaries = []
6673

    
6674
    # mirror node verification
6675
    if self.op.disk_template in constants.DTS_NET_MIRROR:
6676
      if self.op.snode is None:
6677
        raise errors.OpPrereqError("The networked disk templates need"
6678
                                   " a mirror node", errors.ECODE_INVAL)
6679
      if self.op.snode == pnode.name:
6680
        raise errors.OpPrereqError("The secondary node cannot be the"
6681
                                   " primary node.", errors.ECODE_INVAL)
6682
      _CheckNodeOnline(self, self.op.snode)
6683
      _CheckNodeNotDrained(self, self.op.snode)
6684
      self.secondaries.append(self.op.snode)
6685

    
6686
    nodenames = [pnode.name] + self.secondaries
6687

    
6688
    req_size = _ComputeDiskSize(self.op.disk_template,
6689
                                self.disks)
6690

    
6691
    # Check lv size requirements, if not adopting
6692
    if req_size is not None and not self.adopt_disks:
6693
      _CheckNodesFreeDisk(self, nodenames, req_size)
6694

    
6695
    if self.adopt_disks: # instead, we must check the adoption data
6696
      all_lvs = set([i["adopt"] for i in self.disks])
6697
      if len(all_lvs) != len(self.disks):
6698
        raise errors.OpPrereqError("Duplicate volume names given for adoption",
6699
                                   errors.ECODE_INVAL)
6700
      for lv_name in all_lvs:
6701
        try:
6702
          self.cfg.ReserveLV(lv_name, self.proc.GetECId())
6703
        except errors.ReservationError:
6704
          raise errors.OpPrereqError("LV named %s used by another instance" %
6705
                                     lv_name, errors.ECODE_NOTUNIQUE)
6706

    
6707
      node_lvs = self.rpc.call_lv_list([pnode.name],
6708
                                       self.cfg.GetVGName())[pnode.name]
6709
      node_lvs.Raise("Cannot get LV information from node %s" % pnode.name)
6710
      node_lvs = node_lvs.payload
6711
      delta = all_lvs.difference(node_lvs.keys())
6712
      if delta:
6713
        raise errors.OpPrereqError("Missing logical volume(s): %s" %
6714
                                   utils.CommaJoin(delta),
6715
                                   errors.ECODE_INVAL)
6716
      online_lvs = [lv for lv in all_lvs if node_lvs[lv][2]]
6717
      if online_lvs:
6718
        raise errors.OpPrereqError("Online logical volumes found, cannot"
6719
                                   " adopt: %s" % utils.CommaJoin(online_lvs),
6720
                                   errors.ECODE_STATE)
6721
      # update the size of disk based on what is found
6722
      for dsk in self.disks:
6723
        dsk["size"] = int(float(node_lvs[dsk["adopt"]][0]))
6724

    
6725
    _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
6726

    
6727
    _CheckNodeHasOS(self, pnode.name, self.op.os_type, self.op.force_variant)
6728

    
6729
    _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
6730

    
6731
    # memory check on primary node
6732
    if self.op.start:
6733
      _CheckNodeFreeMemory(self, self.pnode.name,
6734
                           "creating instance %s" % self.op.instance_name,
6735
                           self.be_full[constants.BE_MEMORY],
6736
                           self.op.hypervisor)
6737

    
6738
    self.dry_run_result = list(nodenames)
6739

    
6740
  def Exec(self, feedback_fn):
6741
    """Create and add the instance to the cluster.
6742

6743
    """
6744
    instance = self.op.instance_name
6745
    pnode_name = self.pnode.name
6746

    
6747
    ht_kind = self.op.hypervisor
6748
    if ht_kind in constants.HTS_REQ_PORT:
6749
      network_port = self.cfg.AllocatePort()
6750
    else:
6751
      network_port = None
6752

    
6753
    if constants.ENABLE_FILE_STORAGE:
6754
      # this is needed because os.path.join does not accept None arguments
6755
      if self.op.file_storage_dir is None:
6756
        string_file_storage_dir = ""
6757
      else:
6758
        string_file_storage_dir = self.op.file_storage_dir
6759

    
6760
      # build the full file storage dir path
6761
      file_storage_dir = utils.PathJoin(self.cfg.GetFileStorageDir(),
6762
                                        string_file_storage_dir, instance)
6763
    else:
6764
      file_storage_dir = ""
6765

    
6766
    disks = _GenerateDiskTemplate(self,
6767
                                  self.op.disk_template,
6768
                                  instance, pnode_name,
6769
                                  self.secondaries,
6770
                                  self.disks,
6771
                                  file_storage_dir,
6772
                                  self.op.file_driver,
6773
                                  0)
6774

    
6775
    iobj = objects.Instance(name=instance, os=self.op.os_type,
6776
                            primary_node=pnode_name,
6777
                            nics=self.nics, disks=disks,
6778
                            disk_template=self.op.disk_template,
6779
                            admin_up=False,
6780
                            network_port=network_port,
6781
                            beparams=self.op.beparams,
6782
                            hvparams=self.op.hvparams,
6783
                            hypervisor=self.op.hypervisor,
6784
                            )
6785

    
6786
    if self.adopt_disks:
6787
      # rename LVs to the newly-generated names; we need to construct
6788
      # 'fake' LV disks with the old data, plus the new unique_id
6789
      tmp_disks = [objects.Disk.FromDict(v.ToDict()) for v in disks]
6790
      rename_to = []
6791
      for t_dsk, a_dsk in zip (tmp_disks, self.disks):
6792
        rename_to.append(t_dsk.logical_id)
6793
        t_dsk.logical_id = (t_dsk.logical_id[0], a_dsk["adopt"])
6794
        self.cfg.SetDiskID(t_dsk, pnode_name)
6795
      result = self.rpc.call_blockdev_rename(pnode_name,
6796
                                             zip(tmp_disks, rename_to))
6797
      result.Raise("Failed to rename adoped LVs")
6798
    else:
6799
      feedback_fn("* creating instance disks...")
6800
      try:
6801
        _CreateDisks(self, iobj)
6802
      except errors.OpExecError:
6803
        self.LogWarning("Device creation failed, reverting...")
6804
        try:
6805
          _RemoveDisks(self, iobj)
6806
        finally:
6807
          self.cfg.ReleaseDRBDMinors(instance)
6808
          raise
6809

    
6810
    feedback_fn("adding instance %s to cluster config" % instance)
6811

    
6812
    self.cfg.AddInstance(iobj, self.proc.GetECId())
6813

    
6814
    # Declare that we don't want to remove the instance lock anymore, as we've
6815
    # added the instance to the config
6816
    del self.remove_locks[locking.LEVEL_INSTANCE]
6817
    # Unlock all the nodes
6818
    if self.op.mode == constants.INSTANCE_IMPORT:
6819
      nodes_keep = [self.op.src_node]
6820
      nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
6821
                       if node != self.op.src_node]
6822
      self.context.glm.release(locking.LEVEL_NODE, nodes_release)
6823
      self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
6824
    else:
6825
      self.context.glm.release(locking.LEVEL_NODE)
6826
      del self.acquired_locks[locking.LEVEL_NODE]
6827

    
6828
    if self.op.wait_for_sync:
6829
      disk_abort = not _WaitForSync(self, iobj)
6830
    elif iobj.disk_template in constants.DTS_NET_MIRROR:
6831
      # make sure the disks are not degraded (still sync-ing is ok)
6832
      time.sleep(15)
6833
      feedback_fn("* checking mirrors status")
6834
      disk_abort = not _WaitForSync(self, iobj, oneshot=True)
6835
    else:
6836
      disk_abort = False
6837

    
6838
    if disk_abort:
6839
      _RemoveDisks(self, iobj)
6840
      self.cfg.RemoveInstance(iobj.name)
6841
      # Make sure the instance lock gets removed
6842
      self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
6843
      raise errors.OpExecError("There are some degraded disks for"
6844
                               " this instance")
6845

    
6846
    if iobj.disk_template != constants.DT_DISKLESS and not self.adopt_disks:
6847
      if self.op.mode == constants.INSTANCE_CREATE:
6848
        if not self.op.no_install:
6849
          feedback_fn("* running the instance OS create scripts...")
6850
          # FIXME: pass debug option from opcode to backend
6851
          result = self.rpc.call_instance_os_add(pnode_name, iobj, False,
6852
                                                 self.op.debug_level)
6853
          result.Raise("Could not add os for instance %s"
6854
                       " on node %s" % (instance, pnode_name))
6855

    
6856
      elif self.op.mode == constants.INSTANCE_IMPORT:
6857
        feedback_fn("* running the instance OS import scripts...")
6858

    
6859
        transfers = []
6860

    
6861
        for idx, image in enumerate(self.src_images):
6862
          if not image:
6863
            continue
6864

    
6865
          # FIXME: pass debug option from opcode to backend
6866
          dt = masterd.instance.DiskTransfer("disk/%s" % idx,
6867
                                             constants.IEIO_FILE, (image, ),
6868
                                             constants.IEIO_SCRIPT,
6869
                                             (iobj.disks[idx], idx),
6870
                                             None)
6871
          transfers.append(dt)
6872

    
6873
        import_result = \
6874
          masterd.instance.TransferInstanceData(self, feedback_fn,
6875
                                                self.op.src_node, pnode_name,
6876
                                                self.pnode.secondary_ip,
6877
                                                iobj, transfers)
6878
        if not compat.all(import_result):
6879
          self.LogWarning("Some disks for instance %s on node %s were not"
6880
                          " imported successfully" % (instance, pnode_name))
6881

    
6882
      elif self.op.mode == constants.INSTANCE_REMOTE_IMPORT:
6883
        feedback_fn("* preparing remote import...")
6884
        connect_timeout = constants.RIE_CONNECT_TIMEOUT
6885
        timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
6886

    
6887
        disk_results = masterd.instance.RemoteImport(self, feedback_fn, iobj,
6888
                                                     self.source_x509_ca,
6889
                                                     self._cds, timeouts)
6890
        if not compat.all(disk_results):
6891
          # TODO: Should the instance still be started, even if some disks
6892
          # failed to import (valid for local imports, too)?
6893
          self.LogWarning("Some disks for instance %s on node %s were not"
6894
                          " imported successfully" % (instance, pnode_name))
6895

    
6896
        # Run rename script on newly imported instance
6897
        assert iobj.name == instance
6898
        feedback_fn("Running rename script for %s" % instance)
6899
        result = self.rpc.call_instance_run_rename(pnode_name, iobj,
6900
                                                   self.source_instance_name,
6901
                                                   self.op.debug_level)
6902
        if result.fail_msg:
6903
          self.LogWarning("Failed to run rename script for %s on node"
6904
                          " %s: %s" % (instance, pnode_name, result.fail_msg))
6905

    
6906
      else:
6907
        # also checked in the prereq part
6908
        raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
6909
                                     % self.op.mode)
6910

    
6911
    if self.op.start:
6912
      iobj.admin_up = True
6913
      self.cfg.Update(iobj, feedback_fn)
6914
      logging.info("Starting instance %s on node %s", instance, pnode_name)
6915
      feedback_fn("* starting instance...")
6916
      result = self.rpc.call_instance_start(pnode_name, iobj, None, None)
6917
      result.Raise("Could not start instance")
6918

    
6919
    return list(iobj.all_nodes)
6920

    
6921

    
6922
class LUConnectConsole(NoHooksLU):
6923
  """Connect to an instance's console.
6924

6925
  This is somewhat special in that it returns the command line that
6926
  you need to run on the master node in order to connect to the
6927
  console.
6928

6929
  """
6930
  _OP_REQP = ["instance_name"]
6931
  REQ_BGL = False
6932

    
6933
  def ExpandNames(self):
6934
    self._ExpandAndLockInstance()
6935

    
6936
  def CheckPrereq(self):
6937
    """Check prerequisites.
6938

6939
    This checks that the instance is in the cluster.
6940

6941
    """
6942
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6943
    assert self.instance is not None, \
6944
      "Cannot retrieve locked instance %s" % self.op.instance_name
6945
    _CheckNodeOnline(self, self.instance.primary_node)
6946

    
6947
  def Exec(self, feedback_fn):
6948
    """Connect to the console of an instance
6949

6950
    """
6951
    instance = self.instance
6952
    node = instance.primary_node
6953

    
6954
    node_insts = self.rpc.call_instance_list([node],
6955
                                             [instance.hypervisor])[node]
6956
    node_insts.Raise("Can't get node information from %s" % node)
6957

    
6958
    if instance.name not in node_insts.payload:
6959
      raise errors.OpExecError("Instance %s is not running." % instance.name)
6960

    
6961
    logging.debug("Connecting to console of %s on %s", instance.name, node)
6962

    
6963
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
6964
    cluster = self.cfg.GetClusterInfo()
6965
    # beparams and hvparams are passed separately, to avoid editing the
6966
    # instance and then saving the defaults in the instance itself.
6967
    hvparams = cluster.FillHV(instance)
6968
    beparams = cluster.FillBE(instance)
6969
    console_cmd = hyper.GetShellCommandForConsole(instance, hvparams, beparams)
6970

    
6971
    # build ssh cmdline
6972
    return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
6973

    
6974

    
6975
class LUReplaceDisks(LogicalUnit):
6976
  """Replace the disks of an instance.
6977

6978
  """
6979
  HPATH = "mirrors-replace"
6980
  HTYPE = constants.HTYPE_INSTANCE
6981
  _OP_REQP = ["instance_name", "mode", "disks"]
6982
  REQ_BGL = False
6983

    
6984
  def CheckArguments(self):
6985
    if not hasattr(self.op, "remote_node"):
6986
      self.op.remote_node = None
6987
    if not hasattr(self.op, "iallocator"):
6988
      self.op.iallocator = None
6989
    if not hasattr(self.op, "early_release"):
6990
      self.op.early_release = False
6991

    
6992
    TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
6993
                                  self.op.iallocator)
6994

    
6995
  def ExpandNames(self):
6996
    self._ExpandAndLockInstance()
6997

    
6998
    if self.op.iallocator is not None:
6999
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7000

    
7001
    elif self.op.remote_node is not None:
7002
      remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
7003
      self.op.remote_node = remote_node
7004

    
7005
      # Warning: do not remove the locking of the new secondary here
7006
      # unless DRBD8.AddChildren is changed to work in parallel;
7007
      # currently it doesn't since parallel invocations of
7008
      # FindUnusedMinor will conflict
7009
      self.needed_locks[locking.LEVEL_NODE] = [remote_node]
7010
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
7011

    
7012
    else:
7013
      self.needed_locks[locking.LEVEL_NODE] = []
7014
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7015

    
7016
    self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
7017
                                   self.op.iallocator, self.op.remote_node,
7018
                                   self.op.disks, False, self.op.early_release)
7019

    
7020
    self.tasklets = [self.replacer]
7021

    
7022
  def DeclareLocks(self, level):
7023
    # If we're not already locking all nodes in the set we have to declare the
7024
    # instance's primary/secondary nodes.
7025
    if (level == locking.LEVEL_NODE and
7026
        self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
7027
      self._LockInstancesNodes()
7028

    
7029
  def BuildHooksEnv(self):
7030
    """Build hooks env.
7031

7032
    This runs on the master, the primary and all the secondaries.
7033

7034
    """
7035
    instance = self.replacer.instance
7036
    env = {
7037
      "MODE": self.op.mode,
7038
      "NEW_SECONDARY": self.op.remote_node,
7039
      "OLD_SECONDARY": instance.secondary_nodes[0],
7040
      }
7041
    env.update(_BuildInstanceHookEnvByObject(self, instance))
7042
    nl = [
7043
      self.cfg.GetMasterNode(),
7044
      instance.primary_node,
7045
      ]
7046
    if self.op.remote_node is not None:
7047
      nl.append(self.op.remote_node)
7048
    return env, nl, nl
7049

    
7050

    
7051
class LUEvacuateNode(LogicalUnit):
7052
  """Relocate the secondary instances from a node.
7053

7054
  """
7055
  HPATH = "node-evacuate"
7056
  HTYPE = constants.HTYPE_NODE
7057
  _OP_REQP = ["node_name"]
7058
  REQ_BGL = False
7059

    
7060
  def CheckArguments(self):
7061
    if not hasattr(self.op, "remote_node"):
7062
      self.op.remote_node = None
7063
    if not hasattr(self.op, "iallocator"):
7064
      self.op.iallocator = None
7065
    if not hasattr(self.op, "early_release"):
7066
      self.op.early_release = False
7067

    
7068
    TLReplaceDisks.CheckArguments(constants.REPLACE_DISK_CHG,
7069
                                  self.op.remote_node,
7070
                                  self.op.iallocator)
7071

    
7072
  def ExpandNames(self):
7073
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
7074

    
7075
    self.needed_locks = {}
7076

    
7077
    # Declare node locks
7078
    if self.op.iallocator is not None:
7079
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7080

    
7081
    elif self.op.remote_node is not None:
7082
      self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
7083

    
7084
      # Warning: do not remove the locking of the new secondary here
7085
      # unless DRBD8.AddChildren is changed to work in parallel;
7086
      # currently it doesn't since parallel invocations of
7087
      # FindUnusedMinor will conflict
7088
      self.needed_locks[locking.LEVEL_NODE] = [self.op.remote_node]
7089
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
7090

    
7091
    else:
7092
      raise errors.OpPrereqError("Invalid parameters", errors.ECODE_INVAL)
7093

    
7094
    # Create tasklets for replacing disks for all secondary instances on this
7095
    # node
7096
    names = []
7097
    tasklets = []
7098

    
7099
    for inst in _GetNodeSecondaryInstances(self.cfg, self.op.node_name):
7100
      logging.debug("Replacing disks for instance %s", inst.name)
7101
      names.append(inst.name)
7102

    
7103
      replacer = TLReplaceDisks(self, inst.name, constants.REPLACE_DISK_CHG,
7104
                                self.op.iallocator, self.op.remote_node, [],
7105
                                True, self.op.early_release)
7106
      tasklets.append(replacer)
7107

    
7108
    self.tasklets = tasklets
7109
    self.instance_names = names
7110

    
7111
    # Declare instance locks
7112
    self.needed_locks[locking.LEVEL_INSTANCE] = self.instance_names
7113

    
7114
  def DeclareLocks(self, level):
7115
    # If we're not already locking all nodes in the set we have to declare the
7116
    # instance's primary/secondary nodes.
7117
    if (level == locking.LEVEL_NODE and
7118
        self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
7119
      self._LockInstancesNodes()
7120

    
7121
  def BuildHooksEnv(self):
7122
    """Build hooks env.
7123

7124
    This runs on the master, the primary and all the secondaries.
7125

7126
    """
7127
    env = {
7128
      "NODE_NAME": self.op.node_name,
7129
      }
7130

    
7131
    nl = [self.cfg.GetMasterNode()]
7132

    
7133
    if self.op.remote_node is not None:
7134
      env["NEW_SECONDARY"] = self.op.remote_node
7135
      nl.append(self.op.remote_node)
7136

    
7137
    return (env, nl, nl)
7138

    
7139

    
7140
class TLReplaceDisks(Tasklet):
7141
  """Replaces disks for an instance.
7142

7143
  Note: Locking is not within the scope of this class.
7144

7145
  """
7146
  def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
7147
               disks, delay_iallocator, early_release):
7148
    """Initializes this class.
7149

7150
    """
7151
    Tasklet.__init__(self, lu)
7152

    
7153
    # Parameters
7154
    self.instance_name = instance_name
7155
    self.mode = mode
7156
    self.iallocator_name = iallocator_name
7157
    self.remote_node = remote_node
7158
    self.disks = disks
7159
    self.delay_iallocator = delay_iallocator
7160
    self.early_release = early_release
7161

    
7162
    # Runtime data
7163
    self.instance = None
7164
    self.new_node = None
7165
    self.target_node = None
7166
    self.other_node = None
7167
    self.remote_node_info = None
7168
    self.node_secondary_ip = None
7169

    
7170
  @staticmethod
7171
  def CheckArguments(mode, remote_node, iallocator):
7172
    """Helper function for users of this class.
7173

7174
    """
7175
    # check for valid parameter combination
7176
    if mode == constants.REPLACE_DISK_CHG:
7177
      if remote_node is None and iallocator is None:
7178
        raise errors.OpPrereqError("When changing the secondary either an"
7179
                                   " iallocator script must be used or the"
7180
                                   " new node given", errors.ECODE_INVAL)
7181

    
7182
      if remote_node is not None and iallocator is not None:
7183
        raise errors.OpPrereqError("Give either the iallocator or the new"
7184
                                   " secondary, not both", errors.ECODE_INVAL)
7185

    
7186
    elif remote_node is not None or iallocator is not None:
7187
      # Not replacing the secondary
7188
      raise errors.OpPrereqError("The iallocator and new node options can"
7189
                                 " only be used when changing the"
7190
                                 " secondary node", errors.ECODE_INVAL)
7191

    
7192
  @staticmethod
7193
  def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
7194
    """Compute a new secondary node using an IAllocator.
7195

7196
    """
7197
    ial = IAllocator(lu.cfg, lu.rpc,
7198
                     mode=constants.IALLOCATOR_MODE_RELOC,
7199
                     name=instance_name,
7200
                     relocate_from=relocate_from)
7201

    
7202
    ial.Run(iallocator_name)
7203

    
7204
    if not ial.success:
7205
      raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
7206
                                 " %s" % (iallocator_name, ial.info),
7207
                                 errors.ECODE_NORES)
7208

    
7209
    if len(ial.result) != ial.required_nodes:
7210
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
7211
                                 " of nodes (%s), required %s" %
7212
                                 (iallocator_name,
7213
                                  len(ial.result), ial.required_nodes),
7214
                                 errors.ECODE_FAULT)
7215

    
7216
    remote_node_name = ial.result[0]
7217

    
7218
    lu.LogInfo("Selected new secondary for instance '%s': %s",
7219
               instance_name, remote_node_name)
7220

    
7221
    return remote_node_name
7222

    
7223
  def _FindFaultyDisks(self, node_name):
7224
    return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
7225
                                    node_name, True)
7226

    
7227
  def CheckPrereq(self):
7228
    """Check prerequisites.
7229

7230
    This checks that the instance is in the cluster.
7231

7232
    """
7233
    self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
7234
    assert instance is not None, \
7235
      "Cannot retrieve locked instance %s" % self.instance_name
7236

    
7237
    if instance.disk_template != constants.DT_DRBD8:
7238
      raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
7239
                                 " instances", errors.ECODE_INVAL)
7240

    
7241
    if len(instance.secondary_nodes) != 1:
7242
      raise errors.OpPrereqError("The instance has a strange layout,"
7243
                                 " expected one secondary but found %d" %
7244
                                 len(instance.secondary_nodes),
7245
                                 errors.ECODE_FAULT)
7246

    
7247
    if not self.delay_iallocator:
7248
      self._CheckPrereq2()
7249

    
7250
  def _CheckPrereq2(self):
7251
    """Check prerequisites, second part.
7252

7253
    This function should always be part of CheckPrereq. It was separated and is
7254
    now called from Exec because during node evacuation iallocator was only
7255
    called with an unmodified cluster model, not taking planned changes into
7256
    account.
7257

7258
    """
7259
    instance = self.instance
7260
    secondary_node = instance.secondary_nodes[0]
7261

    
7262
    if self.iallocator_name is None:
7263
      remote_node = self.remote_node
7264
    else:
7265
      remote_node = self._RunAllocator(self.lu, self.iallocator_name,
7266
                                       instance.name, instance.secondary_nodes)
7267

    
7268
    if remote_node is not None:
7269
      self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
7270
      assert self.remote_node_info is not None, \
7271
        "Cannot retrieve locked node %s" % remote_node
7272
    else:
7273
      self.remote_node_info = None
7274

    
7275
    if remote_node == self.instance.primary_node:
7276
      raise errors.OpPrereqError("The specified node is the primary node of"
7277
                                 " the instance.", errors.ECODE_INVAL)
7278

    
7279
    if remote_node == secondary_node:
7280
      raise errors.OpPrereqError("The specified node is already the"
7281
                                 " secondary node of the instance.",
7282
                                 errors.ECODE_INVAL)
7283

    
7284
    if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
7285
                                    constants.REPLACE_DISK_CHG):
7286
      raise errors.OpPrereqError("Cannot specify disks to be replaced",
7287
                                 errors.ECODE_INVAL)
7288

    
7289
    if self.mode == constants.REPLACE_DISK_AUTO:
7290
      faulty_primary = self._FindFaultyDisks(instance.primary_node)
7291
      faulty_secondary = self._FindFaultyDisks(secondary_node)
7292

    
7293
      if faulty_primary and faulty_secondary:
7294
        raise errors.OpPrereqError("Instance %s has faulty disks on more than"
7295
                                   " one node and can not be repaired"
7296
                                   " automatically" % self.instance_name,
7297
                                   errors.ECODE_STATE)
7298

    
7299
      if faulty_primary:
7300
        self.disks = faulty_primary
7301
        self.target_node = instance.primary_node
7302
        self.other_node = secondary_node
7303
        check_nodes = [self.target_node, self.other_node]
7304
      elif faulty_secondary:
7305
        self.disks = faulty_secondary
7306
        self.target_node = secondary_node
7307
        self.other_node = instance.primary_node
7308
        check_nodes = [self.target_node, self.other_node]
7309
      else:
7310
        self.disks = []
7311
        check_nodes = []
7312

    
7313
    else:
7314
      # Non-automatic modes
7315
      if self.mode == constants.REPLACE_DISK_PRI:
7316
        self.target_node = instance.primary_node
7317
        self.other_node = secondary_node
7318
        check_nodes = [self.target_node, self.other_node]
7319

    
7320
      elif self.mode == constants.REPLACE_DISK_SEC:
7321
        self.target_node = secondary_node
7322
        self.other_node = instance.primary_node
7323
        check_nodes = [self.target_node, self.other_node]
7324

    
7325
      elif self.mode == constants.REPLACE_DISK_CHG:
7326
        self.new_node = remote_node
7327
        self.other_node = instance.primary_node
7328
        self.target_node = secondary_node
7329
        check_nodes = [self.new_node, self.other_node]
7330

    
7331
        _CheckNodeNotDrained(self.lu, remote_node)
7332

    
7333
        old_node_info = self.cfg.GetNodeInfo(secondary_node)
7334
        assert old_node_info is not None
7335
        if old_node_info.offline and not self.early_release:
7336
          # doesn't make sense to delay the release
7337
          self.early_release = True
7338
          self.lu.LogInfo("Old secondary %s is offline, automatically enabling"
7339
                          " early-release mode", secondary_node)
7340

    
7341
      else:
7342
        raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
7343
                                     self.mode)
7344

    
7345
      # If not specified all disks should be replaced
7346
      if not self.disks:
7347
        self.disks = range(len(self.instance.disks))
7348

    
7349
    for node in check_nodes:
7350
      _CheckNodeOnline(self.lu, node)
7351

    
7352
    # Check whether disks are valid
7353
    for disk_idx in self.disks:
7354
      instance.FindDisk(disk_idx)
7355

    
7356
    # Get secondary node IP addresses
7357
    node_2nd_ip = {}
7358

    
7359
    for node_name in [self.target_node, self.other_node, self.new_node]:
7360
      if node_name is not None:
7361
        node_2nd_ip[node_name] = self.cfg.GetNodeInfo(node_name).secondary_ip
7362

    
7363
    self.node_secondary_ip = node_2nd_ip
7364

    
7365
  def Exec(self, feedback_fn):
7366
    """Execute disk replacement.
7367

7368
    This dispatches the disk replacement to the appropriate handler.
7369

7370
    """
7371
    if self.delay_iallocator:
7372
      self._CheckPrereq2()
7373

    
7374
    if not self.disks:
7375
      feedback_fn("No disks need replacement")
7376
      return
7377

    
7378
    feedback_fn("Replacing disk(s) %s for %s" %
7379
                (utils.CommaJoin(self.disks), self.instance.name))
7380

    
7381
    activate_disks = (not self.instance.admin_up)
7382

    
7383
    # Activate the instance disks if we're replacing them on a down instance
7384
    if activate_disks:
7385
      _StartInstanceDisks(self.lu, self.instance, True)
7386

    
7387
    try:
7388
      # Should we replace the secondary node?
7389
      if self.new_node is not None:
7390
        fn = self._ExecDrbd8Secondary
7391
      else:
7392
        fn = self._ExecDrbd8DiskOnly
7393

    
7394
      return fn(feedback_fn)
7395

    
7396
    finally:
7397
      # Deactivate the instance disks if we're replacing them on a
7398
      # down instance
7399
      if activate_disks:
7400
        _SafeShutdownInstanceDisks(self.lu, self.instance)
7401

    
7402
  def _CheckVolumeGroup(self, nodes):
7403
    self.lu.LogInfo("Checking volume groups")
7404

    
7405
    vgname = self.cfg.GetVGName()
7406

    
7407
    # Make sure volume group exists on all involved nodes
7408
    results = self.rpc.call_vg_list(nodes)
7409
    if not results:
7410
      raise errors.OpExecError("Can't list volume groups on the nodes")
7411

    
7412
    for node in nodes:
7413
      res = results[node]
7414
      res.Raise("Error checking node %s" % node)
7415
      if vgname not in res.payload:
7416
        raise errors.OpExecError("Volume group '%s' not found on node %s" %
7417
                                 (vgname, node))
7418

    
7419
  def _CheckDisksExistence(self, nodes):
7420
    # Check disk existence
7421
    for idx, dev in enumerate(self.instance.disks):
7422
      if idx not in self.disks:
7423
        continue
7424

    
7425
      for node in nodes:
7426
        self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
7427
        self.cfg.SetDiskID(dev, node)
7428

    
7429
        result = self.rpc.call_blockdev_find(node, dev)
7430

    
7431
        msg = result.fail_msg
7432
        if msg or not result.payload:
7433
          if not msg:
7434
            msg = "disk not found"
7435
          raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
7436
                                   (idx, node, msg))
7437

    
7438
  def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
7439
    for idx, dev in enumerate(self.instance.disks):
7440
      if idx not in self.disks:
7441
        continue
7442

    
7443
      self.lu.LogInfo("Checking disk/%d consistency on node %s" %
7444
                      (idx, node_name))
7445

    
7446
      if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
7447
                                   ldisk=ldisk):
7448
        raise errors.OpExecError("Node %s has degraded storage, unsafe to"
7449
                                 " replace disks for instance %s" %
7450
                                 (node_name, self.instance.name))
7451

    
7452
  def _CreateNewStorage(self, node_name):
7453
    vgname = self.cfg.GetVGName()
7454
    iv_names = {}
7455

    
7456
    for idx, dev in enumerate(self.instance.disks):
7457
      if idx not in self.disks:
7458
        continue
7459

    
7460
      self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
7461

    
7462
      self.cfg.SetDiskID(dev, node_name)
7463

    
7464
      lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
7465
      names = _GenerateUniqueNames(self.lu, lv_names)
7466

    
7467
      lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
7468
                             logical_id=(vgname, names[0]))
7469
      lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
7470
                             logical_id=(vgname, names[1]))
7471

    
7472
      new_lvs = [lv_data, lv_meta]
7473
      old_lvs = dev.children
7474
      iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
7475

    
7476
      # we pass force_create=True to force the LVM creation
7477
      for new_lv in new_lvs:
7478
        _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
7479
                        _GetInstanceInfoText(self.instance), False)
7480

    
7481
    return iv_names
7482

    
7483
  def _CheckDevices(self, node_name, iv_names):
7484
    for name, (dev, _, _) in iv_names.iteritems():
7485
      self.cfg.SetDiskID(dev, node_name)
7486

    
7487
      result = self.rpc.call_blockdev_find(node_name, dev)
7488

    
7489
      msg = result.fail_msg
7490
      if msg or not result.payload:
7491
        if not msg:
7492
          msg = "disk not found"
7493
        raise errors.OpExecError("Can't find DRBD device %s: %s" %
7494
                                 (name, msg))
7495

    
7496
      if result.payload.is_degraded:
7497
        raise errors.OpExecError("DRBD device %s is degraded!" % name)
7498

    
7499
  def _RemoveOldStorage(self, node_name, iv_names):
7500
    for name, (_, old_lvs, _) in iv_names.iteritems():
7501
      self.lu.LogInfo("Remove logical volumes for %s" % name)
7502

    
7503
      for lv in old_lvs:
7504
        self.cfg.SetDiskID(lv, node_name)
7505

    
7506
        msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
7507
        if msg:
7508
          self.lu.LogWarning("Can't remove old LV: %s" % msg,
7509
                             hint="remove unused LVs manually")
7510

    
7511
  def _ReleaseNodeLock(self, node_name):
7512
    """Releases the lock for a given node."""
7513
    self.lu.context.glm.release(locking.LEVEL_NODE, node_name)
7514

    
7515
  def _ExecDrbd8DiskOnly(self, feedback_fn):
7516
    """Replace a disk on the primary or secondary for DRBD 8.
7517

7518
    The algorithm for replace is quite complicated:
7519

7520
      1. for each disk to be replaced:
7521

7522
        1. create new LVs on the target node with unique names
7523
        1. detach old LVs from the drbd device
7524
        1. rename old LVs to name_replaced.<time_t>
7525
        1. rename new LVs to old LVs
7526
        1. attach the new LVs (with the old names now) to the drbd device
7527

7528
      1. wait for sync across all devices
7529

7530
      1. for each modified disk:
7531

7532
        1. remove old LVs (which have the name name_replaces.<time_t>)
7533

7534
    Failures are not very well handled.
7535

7536
    """
7537
    steps_total = 6
7538

    
7539
    # Step: check device activation
7540
    self.lu.LogStep(1, steps_total, "Check device existence")
7541
    self._CheckDisksExistence([self.other_node, self.target_node])
7542
    self._CheckVolumeGroup([self.target_node, self.other_node])
7543

    
7544
    # Step: check other node consistency
7545
    self.lu.LogStep(2, steps_total, "Check peer consistency")
7546
    self._CheckDisksConsistency(self.other_node,
7547
                                self.other_node == self.instance.primary_node,
7548
                                False)
7549

    
7550
    # Step: create new storage
7551
    self.lu.LogStep(3, steps_total, "Allocate new storage")
7552
    iv_names = self._CreateNewStorage(self.target_node)
7553

    
7554
    # Step: for each lv, detach+rename*2+attach
7555
    self.lu.LogStep(4, steps_total, "Changing drbd configuration")
7556
    for dev, old_lvs, new_lvs in iv_names.itervalues():
7557
      self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
7558

    
7559
      result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
7560
                                                     old_lvs)
7561
      result.Raise("Can't detach drbd from local storage on node"
7562
                   " %s for device %s" % (self.target_node, dev.iv_name))
7563
      #dev.children = []
7564
      #cfg.Update(instance)
7565

    
7566
      # ok, we created the new LVs, so now we know we have the needed
7567
      # storage; as such, we proceed on the target node to rename
7568
      # old_lv to _old, and new_lv to old_lv; note that we rename LVs
7569
      # using the assumption that logical_id == physical_id (which in
7570
      # turn is the unique_id on that node)
7571

    
7572
      # FIXME(iustin): use a better name for the replaced LVs
7573
      temp_suffix = int(time.time())
7574
      ren_fn = lambda d, suff: (d.physical_id[0],
7575
                                d.physical_id[1] + "_replaced-%s" % suff)
7576

    
7577
      # Build the rename list based on what LVs exist on the node
7578
      rename_old_to_new = []
7579
      for to_ren in old_lvs:
7580
        result = self.rpc.call_blockdev_find(self.target_node, to_ren)
7581
        if not result.fail_msg and result.payload:
7582
          # device exists
7583
          rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
7584

    
7585
      self.lu.LogInfo("Renaming the old LVs on the target node")
7586
      result = self.rpc.call_blockdev_rename(self.target_node,
7587
                                             rename_old_to_new)
7588
      result.Raise("Can't rename old LVs on node %s" % self.target_node)
7589

    
7590
      # Now we rename the new LVs to the old LVs
7591
      self.lu.LogInfo("Renaming the new LVs on the target node")
7592
      rename_new_to_old = [(new, old.physical_id)
7593
                           for old, new in zip(old_lvs, new_lvs)]
7594
      result = self.rpc.call_blockdev_rename(self.target_node,
7595
                                             rename_new_to_old)
7596
      result.Raise("Can't rename new LVs on node %s" % self.target_node)
7597

    
7598
      for old, new in zip(old_lvs, new_lvs):
7599
        new.logical_id = old.logical_id
7600
        self.cfg.SetDiskID(new, self.target_node)
7601

    
7602
      for disk in old_lvs:
7603
        disk.logical_id = ren_fn(disk, temp_suffix)
7604
        self.cfg.SetDiskID(disk, self.target_node)
7605

    
7606
      # Now that the new lvs have the old name, we can add them to the device
7607
      self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
7608
      result = self.rpc.call_blockdev_addchildren(self.target_node, dev,
7609
                                                  new_lvs)
7610
      msg = result.fail_msg
7611
      if msg:
7612
        for new_lv in new_lvs:
7613
          msg2 = self.rpc.call_blockdev_remove(self.target_node,
7614
                                               new_lv).fail_msg
7615
          if msg2:
7616
            self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
7617
                               hint=("cleanup manually the unused logical"
7618
                                     "volumes"))
7619
        raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
7620

    
7621
      dev.children = new_lvs
7622

    
7623
      self.cfg.Update(self.instance, feedback_fn)
7624

    
7625
    cstep = 5
7626
    if self.early_release:
7627
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
7628
      cstep += 1
7629
      self._RemoveOldStorage(self.target_node, iv_names)
7630
      # WARNING: we release both node locks here, do not do other RPCs
7631
      # than WaitForSync to the primary node
7632
      self._ReleaseNodeLock([self.target_node, self.other_node])
7633

    
7634
    # Wait for sync
7635
    # This can fail as the old devices are degraded and _WaitForSync
7636
    # does a combined result over all disks, so we don't check its return value
7637
    self.lu.LogStep(cstep, steps_total, "Sync devices")
7638
    cstep += 1
7639
    _WaitForSync(self.lu, self.instance)
7640

    
7641
    # Check all devices manually
7642
    self._CheckDevices(self.instance.primary_node, iv_names)
7643

    
7644
    # Step: remove old storage
7645
    if not self.early_release:
7646
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
7647
      cstep += 1
7648
      self._RemoveOldStorage(self.target_node, iv_names)
7649

    
7650
  def _ExecDrbd8Secondary(self, feedback_fn):
7651
    """Replace the secondary node for DRBD 8.
7652

7653
    The algorithm for replace is quite complicated:
7654
      - for all disks of the instance:
7655
        - create new LVs on the new node with same names
7656
        - shutdown the drbd device on the old secondary
7657
        - disconnect the drbd network on the primary
7658
        - create the drbd device on the new secondary
7659
        - network attach the drbd on the primary, using an artifice:
7660
          the drbd code for Attach() will connect to the network if it
7661
          finds a device which is connected to the good local disks but
7662
          not network enabled
7663
      - wait for sync across all devices
7664
      - remove all disks from the old secondary
7665

7666
    Failures are not very well handled.
7667

7668
    """
7669
    steps_total = 6
7670

    
7671
    # Step: check device activation
7672
    self.lu.LogStep(1, steps_total, "Check device existence")
7673
    self._CheckDisksExistence([self.instance.primary_node])
7674
    self._CheckVolumeGroup([self.instance.primary_node])
7675

    
7676
    # Step: check other node consistency
7677
    self.lu.LogStep(2, steps_total, "Check peer consistency")
7678
    self._CheckDisksConsistency(self.instance.primary_node, True, True)
7679

    
7680
    # Step: create new storage
7681
    self.lu.LogStep(3, steps_total, "Allocate new storage")
7682
    for idx, dev in enumerate(self.instance.disks):
7683
      self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
7684
                      (self.new_node, idx))
7685
      # we pass force_create=True to force LVM creation
7686
      for new_lv in dev.children:
7687
        _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
7688
                        _GetInstanceInfoText(self.instance), False)
7689

    
7690
    # Step 4: dbrd minors and drbd setups changes
7691
    # after this, we must manually remove the drbd minors on both the
7692
    # error and the success paths
7693
    self.lu.LogStep(4, steps_total, "Changing drbd configuration")
7694
    minors = self.cfg.AllocateDRBDMinor([self.new_node
7695
                                         for dev in self.instance.disks],
7696
                                        self.instance.name)
7697
    logging.debug("Allocated minors %r", minors)
7698

    
7699
    iv_names = {}
7700
    for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
7701
      self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
7702
                      (self.new_node, idx))
7703
      # create new devices on new_node; note that we create two IDs:
7704
      # one without port, so the drbd will be activated without
7705
      # networking information on the new node at this stage, and one
7706
      # with network, for the latter activation in step 4
7707
      (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
7708
      if self.instance.primary_node == o_node1:
7709
        p_minor = o_minor1
7710
      else:
7711
        assert self.instance.primary_node == o_node2, "Three-node instance?"
7712
        p_minor = o_minor2
7713

    
7714
      new_alone_id = (self.instance.primary_node, self.new_node, None,
7715
                      p_minor, new_minor, o_secret)
7716
      new_net_id = (self.instance.primary_node, self.new_node, o_port,
7717
                    p_minor, new_minor, o_secret)
7718

    
7719
      iv_names[idx] = (dev, dev.children, new_net_id)
7720
      logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
7721
                    new_net_id)
7722
      new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
7723
                              logical_id=new_alone_id,
7724
                              children=dev.children,
7725
                              size=dev.size)
7726
      try:
7727
        _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
7728
                              _GetInstanceInfoText(self.instance), False)
7729
      except errors.GenericError:
7730
        self.cfg.ReleaseDRBDMinors(self.instance.name)
7731
        raise
7732

    
7733
    # We have new devices, shutdown the drbd on the old secondary
7734
    for idx, dev in enumerate(self.instance.disks):
7735
      self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
7736
      self.cfg.SetDiskID(dev, self.target_node)
7737
      msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
7738
      if msg:
7739
        self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
7740
                           "node: %s" % (idx, msg),
7741
                           hint=("Please cleanup this device manually as"
7742
                                 " soon as possible"))
7743

    
7744
    self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
7745
    result = self.rpc.call_drbd_disconnect_net([self.instance.primary_node],
7746
                                               self.node_secondary_ip,
7747
                                               self.instance.disks)\
7748
                                              [self.instance.primary_node]
7749

    
7750
    msg = result.fail_msg
7751
    if msg:
7752
      # detaches didn't succeed (unlikely)
7753
      self.cfg.ReleaseDRBDMinors(self.instance.name)
7754
      raise errors.OpExecError("Can't detach the disks from the network on"
7755
                               " old node: %s" % (msg,))
7756

    
7757
    # if we managed to detach at least one, we update all the disks of
7758
    # the instance to point to the new secondary
7759
    self.lu.LogInfo("Updating instance configuration")
7760
    for dev, _, new_logical_id in iv_names.itervalues():
7761
      dev.logical_id = new_logical_id
7762
      self.cfg.SetDiskID(dev, self.instance.primary_node)
7763

    
7764
    self.cfg.Update(self.instance, feedback_fn)
7765

    
7766
    # and now perform the drbd attach
7767
    self.lu.LogInfo("Attaching primary drbds to new secondary"
7768
                    " (standalone => connected)")
7769
    result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
7770
                                            self.new_node],
7771
                                           self.node_secondary_ip,
7772
                                           self.instance.disks,
7773
                                           self.instance.name,
7774
                                           False)
7775
    for to_node, to_result in result.items():
7776
      msg = to_result.fail_msg
7777
      if msg:
7778
        self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
7779
                           to_node, msg,
7780
                           hint=("please do a gnt-instance info to see the"
7781
                                 " status of disks"))
7782
    cstep = 5
7783
    if self.early_release:
7784
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
7785
      cstep += 1
7786
      self._RemoveOldStorage(self.target_node, iv_names)
7787
      # WARNING: we release all node locks here, do not do other RPCs
7788
      # than WaitForSync to the primary node
7789
      self._ReleaseNodeLock([self.instance.primary_node,
7790
                             self.target_node,
7791
                             self.new_node])
7792

    
7793
    # Wait for sync
7794
    # This can fail as the old devices are degraded and _WaitForSync
7795
    # does a combined result over all disks, so we don't check its return value
7796
    self.lu.LogStep(cstep, steps_total, "Sync devices")
7797
    cstep += 1
7798
    _WaitForSync(self.lu, self.instance)
7799

    
7800
    # Check all devices manually
7801
    self._CheckDevices(self.instance.primary_node, iv_names)
7802

    
7803
    # Step: remove old storage
7804
    if not self.early_release:
7805
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
7806
      self._RemoveOldStorage(self.target_node, iv_names)
7807

    
7808

    
7809
class LURepairNodeStorage(NoHooksLU):
7810
  """Repairs the volume group on a node.
7811

7812
  """
7813
  _OP_REQP = ["node_name"]
7814
  REQ_BGL = False
7815

    
7816
  def CheckArguments(self):
7817
    self.op.node_name = _ExpandNodeName(self.cfg, self.op.node_name)
7818

    
7819
    _CheckStorageType(self.op.storage_type)
7820

    
7821
  def ExpandNames(self):
7822
    self.needed_locks = {
7823
      locking.LEVEL_NODE: [self.op.node_name],
7824
      }
7825

    
7826
  def _CheckFaultyDisks(self, instance, node_name):
7827
    """Ensure faulty disks abort the opcode or at least warn."""
7828
    try:
7829
      if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
7830
                                  node_name, True):
7831
        raise errors.OpPrereqError("Instance '%s' has faulty disks on"
7832
                                   " node '%s'" % (instance.name, node_name),
7833
                                   errors.ECODE_STATE)
7834
    except errors.OpPrereqError, err:
7835
      if self.op.ignore_consistency:
7836
        self.proc.LogWarning(str(err.args[0]))
7837
      else:
7838
        raise
7839

    
7840
  def CheckPrereq(self):
7841
    """Check prerequisites.
7842

7843
    """
7844
    storage_type = self.op.storage_type
7845

    
7846
    if (constants.SO_FIX_CONSISTENCY not in
7847
        constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
7848
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
7849
                                 " repaired" % storage_type,
7850
                                 errors.ECODE_INVAL)
7851

    
7852
    # Check whether any instance on this node has faulty disks
7853
    for inst in _GetNodeInstances(self.cfg, self.op.node_name):
7854
      if not inst.admin_up:
7855
        continue
7856
      check_nodes = set(inst.all_nodes)
7857
      check_nodes.discard(self.op.node_name)
7858
      for inst_node_name in check_nodes:
7859
        self._CheckFaultyDisks(inst, inst_node_name)
7860

    
7861
  def Exec(self, feedback_fn):
7862
    feedback_fn("Repairing storage unit '%s' on %s ..." %
7863
                (self.op.name, self.op.node_name))
7864

    
7865
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
7866
    result = self.rpc.call_storage_execute(self.op.node_name,
7867
                                           self.op.storage_type, st_args,
7868
                                           self.op.name,
7869
                                           constants.SO_FIX_CONSISTENCY)
7870
    result.Raise("Failed to repair storage unit '%s' on %s" %
7871
                 (self.op.name, self.op.node_name))
7872

    
7873

    
7874
class LUNodeEvacuationStrategy(NoHooksLU):
7875
  """Computes the node evacuation strategy.
7876

7877
  """
7878
  _OP_REQP = ["nodes"]
7879
  REQ_BGL = False
7880

    
7881
  def CheckArguments(self):
7882
    if not hasattr(self.op, "remote_node"):
7883
      self.op.remote_node = None
7884
    if not hasattr(self.op, "iallocator"):
7885
      self.op.iallocator = None
7886
    if self.op.remote_node is not None and self.op.iallocator is not None:
7887
      raise errors.OpPrereqError("Give either the iallocator or the new"
7888
                                 " secondary, not both", errors.ECODE_INVAL)
7889

    
7890
  def ExpandNames(self):
7891
    self.op.nodes = _GetWantedNodes(self, self.op.nodes)
7892
    self.needed_locks = locks = {}
7893
    if self.op.remote_node is None:
7894
      locks[locking.LEVEL_NODE] = locking.ALL_SET
7895
    else:
7896
      self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
7897
      locks[locking.LEVEL_NODE] = self.op.nodes + [self.op.remote_node]
7898

    
7899
  def CheckPrereq(self):
7900
    pass
7901

    
7902
  def Exec(self, feedback_fn):
7903
    if self.op.remote_node is not None:
7904
      instances = []
7905
      for node in self.op.nodes:
7906
        instances.extend(_GetNodeSecondaryInstances(self.cfg, node))
7907
      result = []
7908
      for i in instances:
7909
        if i.primary_node == self.op.remote_node:
7910
          raise errors.OpPrereqError("Node %s is the primary node of"
7911
                                     " instance %s, cannot use it as"
7912
                                     " secondary" %
7913
                                     (self.op.remote_node, i.name),
7914
                                     errors.ECODE_INVAL)
7915
        result.append([i.name, self.op.remote_node])
7916
    else:
7917
      ial = IAllocator(self.cfg, self.rpc,
7918
                       mode=constants.IALLOCATOR_MODE_MEVAC,
7919
                       evac_nodes=self.op.nodes)
7920
      ial.Run(self.op.iallocator, validate=True)
7921
      if not ial.success:
7922
        raise errors.OpExecError("No valid evacuation solution: %s" % ial.info,
7923
                                 errors.ECODE_NORES)
7924
      result = ial.result
7925
    return result
7926

    
7927

    
7928
class LUGrowDisk(LogicalUnit):
7929
  """Grow a disk of an instance.
7930

7931
  """
7932
  HPATH = "disk-grow"
7933
  HTYPE = constants.HTYPE_INSTANCE
7934
  _OP_REQP = ["instance_name", "disk", "amount", "wait_for_sync"]
7935
  REQ_BGL = False
7936

    
7937
  def ExpandNames(self):
7938
    self._ExpandAndLockInstance()
7939
    self.needed_locks[locking.LEVEL_NODE] = []
7940
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7941

    
7942
  def DeclareLocks(self, level):
7943
    if level == locking.LEVEL_NODE:
7944
      self._LockInstancesNodes()
7945

    
7946
  def BuildHooksEnv(self):
7947
    """Build hooks env.
7948

7949
    This runs on the master, the primary and all the secondaries.
7950

7951
    """
7952
    env = {
7953
      "DISK": self.op.disk,
7954
      "AMOUNT": self.op.amount,
7955
      }
7956
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
7957
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
7958
    return env, nl, nl
7959

    
7960
  def CheckPrereq(self):
7961
    """Check prerequisites.
7962

7963
    This checks that the instance is in the cluster.
7964

7965
    """
7966
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7967
    assert instance is not None, \
7968
      "Cannot retrieve locked instance %s" % self.op.instance_name
7969
    nodenames = list(instance.all_nodes)
7970
    for node in nodenames:
7971
      _CheckNodeOnline(self, node)
7972

    
7973

    
7974
    self.instance = instance
7975

    
7976
    if instance.disk_template not in constants.DTS_GROWABLE:
7977
      raise errors.OpPrereqError("Instance's disk layout does not support"
7978
                                 " growing.", errors.ECODE_INVAL)
7979

    
7980
    self.disk = instance.FindDisk(self.op.disk)
7981

    
7982
    if instance.disk_template != constants.DT_FILE:
7983
      # TODO: check the free disk space for file, when that feature will be
7984
      # supported
7985
      _CheckNodesFreeDisk(self, nodenames, self.op.amount)
7986

    
7987
  def Exec(self, feedback_fn):
7988
    """Execute disk grow.
7989

7990
    """
7991
    instance = self.instance
7992
    disk = self.disk
7993
    for node in instance.all_nodes:
7994
      self.cfg.SetDiskID(disk, node)
7995
      result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
7996
      result.Raise("Grow request failed to node %s" % node)
7997

    
7998
      # TODO: Rewrite code to work properly
7999
      # DRBD goes into sync mode for a short amount of time after executing the
8000
      # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
8001
      # calling "resize" in sync mode fails. Sleeping for a short amount of
8002
      # time is a work-around.
8003
      time.sleep(5)
8004

    
8005
    disk.RecordGrow(self.op.amount)
8006
    self.cfg.Update(instance, feedback_fn)
8007
    if self.op.wait_for_sync:
8008
      disk_abort = not _WaitForSync(self, instance)
8009
      if disk_abort:
8010
        self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
8011
                             " status.\nPlease check the instance.")
8012

    
8013

    
8014
class LUQueryInstanceData(NoHooksLU):
8015
  """Query runtime instance data.
8016

8017
  """
8018
  _OP_REQP = ["instances", "static"]
8019
  REQ_BGL = False
8020

    
8021
  def ExpandNames(self):
8022
    self.needed_locks = {}
8023
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
8024

    
8025
    if not isinstance(self.op.instances, list):
8026
      raise errors.OpPrereqError("Invalid argument type 'instances'",
8027
                                 errors.ECODE_INVAL)
8028

    
8029
    if self.op.instances:
8030
      self.wanted_names = []
8031
      for name in self.op.instances:
8032
        full_name = _ExpandInstanceName(self.cfg, name)
8033
        self.wanted_names.append(full_name)
8034
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
8035
    else:
8036
      self.wanted_names = None
8037
      self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
8038

    
8039
    self.needed_locks[locking.LEVEL_NODE] = []
8040
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8041

    
8042
  def DeclareLocks(self, level):
8043
    if level == locking.LEVEL_NODE:
8044
      self._LockInstancesNodes()
8045

    
8046
  def CheckPrereq(self):
8047
    """Check prerequisites.
8048

8049
    This only checks the optional instance list against the existing names.
8050

8051
    """
8052
    if self.wanted_names is None:
8053
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
8054

    
8055
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
8056
                             in self.wanted_names]
8057
    return
8058

    
8059
  def _ComputeBlockdevStatus(self, node, instance_name, dev):
8060
    """Returns the status of a block device
8061

8062
    """
8063
    if self.op.static or not node:
8064
      return None
8065

    
8066
    self.cfg.SetDiskID(dev, node)
8067

    
8068
    result = self.rpc.call_blockdev_find(node, dev)
8069
    if result.offline:
8070
      return None
8071

    
8072
    result.Raise("Can't compute disk status for %s" % instance_name)
8073

    
8074
    status = result.payload
8075
    if status is None:
8076
      return None
8077

    
8078
    return (status.dev_path, status.major, status.minor,
8079
            status.sync_percent, status.estimated_time,
8080
            status.is_degraded, status.ldisk_status)
8081

    
8082
  def _ComputeDiskStatus(self, instance, snode, dev):
8083
    """Compute block device status.
8084

8085
    """
8086
    if dev.dev_type in constants.LDS_DRBD:
8087
      # we change the snode then (otherwise we use the one passed in)
8088
      if dev.logical_id[0] == instance.primary_node:
8089
        snode = dev.logical_id[1]
8090
      else:
8091
        snode = dev.logical_id[0]
8092

    
8093
    dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
8094
                                              instance.name, dev)
8095
    dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
8096

    
8097
    if dev.children:
8098
      dev_children = [self._ComputeDiskStatus(instance, snode, child)
8099
                      for child in dev.children]
8100
    else:
8101
      dev_children = []
8102

    
8103
    data = {
8104
      "iv_name": dev.iv_name,
8105
      "dev_type": dev.dev_type,
8106
      "logical_id": dev.logical_id,
8107
      "physical_id": dev.physical_id,
8108
      "pstatus": dev_pstatus,
8109
      "sstatus": dev_sstatus,
8110
      "children": dev_children,
8111
      "mode": dev.mode,
8112
      "size": dev.size,
8113
      }
8114

    
8115
    return data
8116

    
8117
  def Exec(self, feedback_fn):
8118
    """Gather and return data"""
8119
    result = {}
8120

    
8121
    cluster = self.cfg.GetClusterInfo()
8122

    
8123
    for instance in self.wanted_instances:
8124
      if not self.op.static:
8125
        remote_info = self.rpc.call_instance_info(instance.primary_node,
8126
                                                  instance.name,
8127
                                                  instance.hypervisor)
8128
        remote_info.Raise("Error checking node %s" % instance.primary_node)
8129
        remote_info = remote_info.payload
8130
        if remote_info and "state" in remote_info:
8131
          remote_state = "up"
8132
        else:
8133
          remote_state = "down"
8134
      else:
8135
        remote_state = None
8136
      if instance.admin_up:
8137
        config_state = "up"
8138
      else:
8139
        config_state = "down"
8140

    
8141
      disks = [self._ComputeDiskStatus(instance, None, device)
8142
               for device in instance.disks]
8143

    
8144
      idict = {
8145
        "name": instance.name,
8146
        "config_state": config_state,
8147
        "run_state": remote_state,
8148
        "pnode": instance.primary_node,
8149
        "snodes": instance.secondary_nodes,
8150
        "os": instance.os,
8151
        # this happens to be the same format used for hooks
8152
        "nics": _NICListToTuple(self, instance.nics),
8153
        "disk_template": instance.disk_template,
8154
        "disks": disks,
8155
        "hypervisor": instance.hypervisor,
8156
        "network_port": instance.network_port,
8157
        "hv_instance": instance.hvparams,
8158
        "hv_actual": cluster.FillHV(instance, skip_globals=True),
8159
        "be_instance": instance.beparams,
8160
        "be_actual": cluster.FillBE(instance),
8161
        "serial_no": instance.serial_no,
8162
        "mtime": instance.mtime,
8163
        "ctime": instance.ctime,
8164
        "uuid": instance.uuid,
8165
        }
8166

    
8167
      result[instance.name] = idict
8168

    
8169
    return result
8170

    
8171

    
8172
class LUSetInstanceParams(LogicalUnit):
8173
  """Modifies an instances's parameters.
8174

8175
  """
8176
  HPATH = "instance-modify"
8177
  HTYPE = constants.HTYPE_INSTANCE
8178
  _OP_REQP = ["instance_name"]
8179
  REQ_BGL = False
8180

    
8181
  def CheckArguments(self):
8182
    if not hasattr(self.op, 'nics'):
8183
      self.op.nics = []
8184
    if not hasattr(self.op, 'disks'):
8185
      self.op.disks = []
8186
    if not hasattr(self.op, 'beparams'):
8187
      self.op.beparams = {}
8188
    if not hasattr(self.op, 'hvparams'):
8189
      self.op.hvparams = {}
8190
    if not hasattr(self.op, "disk_template"):
8191
      self.op.disk_template = None
8192
    if not hasattr(self.op, "remote_node"):
8193
      self.op.remote_node = None
8194
    if not hasattr(self.op, "os_name"):
8195
      self.op.os_name = None
8196
    if not hasattr(self.op, "force_variant"):
8197
      self.op.force_variant = False
8198
    self.op.force = getattr(self.op, "force", False)
8199
    if not (self.op.nics or self.op.disks or self.op.disk_template or
8200
            self.op.hvparams or self.op.beparams or self.op.os_name):
8201
      raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
8202

    
8203
    if self.op.hvparams:
8204
      _CheckGlobalHvParams(self.op.hvparams)
8205

    
8206
    # Disk validation
8207
    disk_addremove = 0
8208
    for disk_op, disk_dict in self.op.disks:
8209
      if disk_op == constants.DDM_REMOVE:
8210
        disk_addremove += 1
8211
        continue
8212
      elif disk_op == constants.DDM_ADD:
8213
        disk_addremove += 1
8214
      else:
8215
        if not isinstance(disk_op, int):
8216
          raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL)
8217
        if not isinstance(disk_dict, dict):
8218
          msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
8219
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
8220

    
8221
      if disk_op == constants.DDM_ADD:
8222
        mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
8223
        if mode not in constants.DISK_ACCESS_SET:
8224
          raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
8225
                                     errors.ECODE_INVAL)
8226
        size = disk_dict.get('size', None)
8227
        if size is None:
8228
          raise errors.OpPrereqError("Required disk parameter size missing",
8229
                                     errors.ECODE_INVAL)
8230
        try:
8231
          size = int(size)
8232
        except (TypeError, ValueError), err:
8233
          raise errors.OpPrereqError("Invalid disk size parameter: %s" %
8234
                                     str(err), errors.ECODE_INVAL)
8235
        disk_dict['size'] = size
8236
      else:
8237
        # modification of disk
8238
        if 'size' in disk_dict:
8239
          raise errors.OpPrereqError("Disk size change not possible, use"
8240
                                     " grow-disk", errors.ECODE_INVAL)
8241

    
8242
    if disk_addremove > 1:
8243
      raise errors.OpPrereqError("Only one disk add or remove operation"
8244
                                 " supported at a time", errors.ECODE_INVAL)
8245

    
8246
    if self.op.disks and self.op.disk_template is not None:
8247
      raise errors.OpPrereqError("Disk template conversion and other disk"
8248
                                 " changes not supported at the same time",
8249
                                 errors.ECODE_INVAL)
8250

    
8251
    if self.op.disk_template:
8252
      _CheckDiskTemplate(self.op.disk_template)
8253
      if (self.op.disk_template in constants.DTS_NET_MIRROR and
8254
          self.op.remote_node is None):
8255
        raise errors.OpPrereqError("Changing the disk template to a mirrored"
8256
                                   " one requires specifying a secondary node",
8257
                                   errors.ECODE_INVAL)
8258

    
8259
    # NIC validation
8260
    nic_addremove = 0
8261
    for nic_op, nic_dict in self.op.nics:
8262
      if nic_op == constants.DDM_REMOVE:
8263
        nic_addremove += 1
8264
        continue
8265
      elif nic_op == constants.DDM_ADD:
8266
        nic_addremove += 1
8267
      else:
8268
        if not isinstance(nic_op, int):
8269
          raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL)
8270
        if not isinstance(nic_dict, dict):
8271
          msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
8272
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
8273

    
8274
      # nic_dict should be a dict
8275
      nic_ip = nic_dict.get('ip', None)
8276
      if nic_ip is not None:
8277
        if nic_ip.lower() == constants.VALUE_NONE:
8278
          nic_dict['ip'] = None
8279
        else:
8280
          if not utils.IsValidIP(nic_ip):
8281
            raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip,
8282
                                       errors.ECODE_INVAL)
8283

    
8284
      nic_bridge = nic_dict.get('bridge', None)
8285
      nic_link = nic_dict.get('link', None)
8286
      if nic_bridge and nic_link:
8287
        raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
8288
                                   " at the same time", errors.ECODE_INVAL)
8289
      elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
8290
        nic_dict['bridge'] = None
8291
      elif nic_link and nic_link.lower() == constants.VALUE_NONE:
8292
        nic_dict['link'] = None
8293

    
8294
      if nic_op == constants.DDM_ADD:
8295
        nic_mac = nic_dict.get('mac', None)
8296
        if nic_mac is None:
8297
          nic_dict['mac'] = constants.VALUE_AUTO
8298

    
8299
      if 'mac' in nic_dict:
8300
        nic_mac = nic_dict['mac']
8301
        if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
8302
          nic_mac = utils.NormalizeAndValidateMac(nic_mac)
8303

    
8304
        if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
8305
          raise errors.OpPrereqError("'auto' is not a valid MAC address when"
8306
                                     " modifying an existing nic",
8307
                                     errors.ECODE_INVAL)
8308

    
8309
    if nic_addremove > 1:
8310
      raise errors.OpPrereqError("Only one NIC add or remove operation"
8311
                                 " supported at a time", errors.ECODE_INVAL)
8312

    
8313
  def ExpandNames(self):
8314
    self._ExpandAndLockInstance()
8315
    self.needed_locks[locking.LEVEL_NODE] = []
8316
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
8317

    
8318
  def DeclareLocks(self, level):
8319
    if level == locking.LEVEL_NODE:
8320
      self._LockInstancesNodes()
8321
      if self.op.disk_template and self.op.remote_node:
8322
        self.op.remote_node = _ExpandNodeName(self.cfg, self.op.remote_node)
8323
        self.needed_locks[locking.LEVEL_NODE].append(self.op.remote_node)
8324

    
8325
  def BuildHooksEnv(self):
8326
    """Build hooks env.
8327

8328
    This runs on the master, primary and secondaries.
8329

8330
    """
8331
    args = dict()
8332
    if constants.BE_MEMORY in self.be_new:
8333
      args['memory'] = self.be_new[constants.BE_MEMORY]
8334
    if constants.BE_VCPUS in self.be_new:
8335
      args['vcpus'] = self.be_new[constants.BE_VCPUS]
8336
    # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
8337
    # information at all.
8338
    if self.op.nics:
8339
      args['nics'] = []
8340
      nic_override = dict(self.op.nics)
8341
      c_nicparams = self.cluster.nicparams[constants.PP_DEFAULT]
8342
      for idx, nic in enumerate(self.instance.nics):
8343
        if idx in nic_override:
8344
          this_nic_override = nic_override[idx]
8345
        else:
8346
          this_nic_override = {}
8347
        if 'ip' in this_nic_override:
8348
          ip = this_nic_override['ip']
8349
        else:
8350
          ip = nic.ip
8351
        if 'mac' in this_nic_override:
8352
          mac = this_nic_override['mac']
8353
        else:
8354
          mac = nic.mac
8355
        if idx in self.nic_pnew:
8356
          nicparams = self.nic_pnew[idx]
8357
        else:
8358
          nicparams = objects.FillDict(c_nicparams, nic.nicparams)
8359
        mode = nicparams[constants.NIC_MODE]
8360
        link = nicparams[constants.NIC_LINK]
8361
        args['nics'].append((ip, mac, mode, link))
8362
      if constants.DDM_ADD in nic_override:
8363
        ip = nic_override[constants.DDM_ADD].get('ip', None)
8364
        mac = nic_override[constants.DDM_ADD]['mac']
8365
        nicparams = self.nic_pnew[constants.DDM_ADD]
8366
        mode = nicparams[constants.NIC_MODE]
8367
        link = nicparams[constants.NIC_LINK]
8368
        args['nics'].append((ip, mac, mode, link))
8369
      elif constants.DDM_REMOVE in nic_override:
8370
        del args['nics'][-1]
8371

    
8372
    env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
8373
    if self.op.disk_template:
8374
      env["NEW_DISK_TEMPLATE"] = self.op.disk_template
8375
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
8376
    return env, nl, nl
8377

    
8378
  @staticmethod
8379
  def _GetUpdatedParams(old_params, update_dict,
8380
                        default_values, parameter_types):
8381
    """Return the new params dict for the given params.
8382

8383
    @type old_params: dict
8384
    @param old_params: old parameters
8385
    @type update_dict: dict
8386
    @param update_dict: dict containing new parameter values,
8387
                        or constants.VALUE_DEFAULT to reset the
8388
                        parameter to its default value
8389
    @type default_values: dict
8390
    @param default_values: default values for the filled parameters
8391
    @type parameter_types: dict
8392
    @param parameter_types: dict mapping target dict keys to types
8393
                            in constants.ENFORCEABLE_TYPES
8394
    @rtype: (dict, dict)
8395
    @return: (new_parameters, filled_parameters)
8396

8397
    """
8398
    params_copy = copy.deepcopy(old_params)
8399
    for key, val in update_dict.iteritems():
8400
      if val == constants.VALUE_DEFAULT:
8401
        try:
8402
          del params_copy[key]
8403
        except KeyError:
8404
          pass
8405
      else:
8406
        params_copy[key] = val
8407
    utils.ForceDictType(params_copy, parameter_types)
8408
    params_filled = objects.FillDict(default_values, params_copy)
8409
    return (params_copy, params_filled)
8410

    
8411
  def CheckPrereq(self):
8412
    """Check prerequisites.
8413

8414
    This only checks the instance list against the existing names.
8415

8416
    """
8417
    self.force = self.op.force
8418

    
8419
    # checking the new params on the primary/secondary nodes
8420

    
8421
    instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
8422
    cluster = self.cluster = self.cfg.GetClusterInfo()
8423
    assert self.instance is not None, \
8424
      "Cannot retrieve locked instance %s" % self.op.instance_name
8425
    pnode = instance.primary_node
8426
    nodelist = list(instance.all_nodes)
8427

    
8428
    if self.op.disk_template:
8429
      if instance.disk_template == self.op.disk_template:
8430
        raise errors.OpPrereqError("Instance already has disk template %s" %
8431
                                   instance.disk_template, errors.ECODE_INVAL)
8432

    
8433
      if (instance.disk_template,
8434
          self.op.disk_template) not in self._DISK_CONVERSIONS:
8435
        raise errors.OpPrereqError("Unsupported disk template conversion from"
8436
                                   " %s to %s" % (instance.disk_template,
8437
                                                  self.op.disk_template),
8438
                                   errors.ECODE_INVAL)
8439
      if self.op.disk_template in constants.DTS_NET_MIRROR:
8440
        _CheckNodeOnline(self, self.op.remote_node)
8441
        _CheckNodeNotDrained(self, self.op.remote_node)
8442
        disks = [{"size": d.size} for d in instance.disks]
8443
        required = _ComputeDiskSize(self.op.disk_template, disks)
8444
        _CheckNodesFreeDisk(self, [self.op.remote_node], required)
8445
        _CheckInstanceDown(self, instance, "cannot change disk template")
8446

    
8447
    # hvparams processing
8448
    if self.op.hvparams:
8449
      i_hvdict, hv_new = self._GetUpdatedParams(
8450
                             instance.hvparams, self.op.hvparams,
8451
                             cluster.hvparams[instance.hypervisor],
8452
                             constants.HVS_PARAMETER_TYPES)
8453
      # local check
8454
      hypervisor.GetHypervisor(
8455
        instance.hypervisor).CheckParameterSyntax(hv_new)
8456
      _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
8457
      self.hv_new = hv_new # the new actual values
8458
      self.hv_inst = i_hvdict # the new dict (without defaults)
8459
    else:
8460
      self.hv_new = self.hv_inst = {}
8461

    
8462
    # beparams processing
8463
    if self.op.beparams:
8464
      i_bedict, be_new = self._GetUpdatedParams(
8465
                             instance.beparams, self.op.beparams,
8466
                             cluster.beparams[constants.PP_DEFAULT],
8467
                             constants.BES_PARAMETER_TYPES)
8468
      self.be_new = be_new # the new actual values
8469
      self.be_inst = i_bedict # the new dict (without defaults)
8470
    else:
8471
      self.be_new = self.be_inst = {}
8472

    
8473
    self.warn = []
8474

    
8475
    if constants.BE_MEMORY in self.op.beparams and not self.force:
8476
      mem_check_list = [pnode]
8477
      if be_new[constants.BE_AUTO_BALANCE]:
8478
        # either we changed auto_balance to yes or it was from before
8479
        mem_check_list.extend(instance.secondary_nodes)
8480
      instance_info = self.rpc.call_instance_info(pnode, instance.name,
8481
                                                  instance.hypervisor)
8482
      nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
8483
                                         instance.hypervisor)
8484
      pninfo = nodeinfo[pnode]
8485
      msg = pninfo.fail_msg
8486
      if msg:
8487
        # Assume the primary node is unreachable and go ahead
8488
        self.warn.append("Can't get info from primary node %s: %s" %
8489
                         (pnode,  msg))
8490
      elif not isinstance(pninfo.payload.get('memory_free', None), int):
8491
        self.warn.append("Node data from primary node %s doesn't contain"
8492
                         " free memory information" % pnode)
8493
      elif instance_info.fail_msg:
8494
        self.warn.append("Can't get instance runtime information: %s" %
8495
                        instance_info.fail_msg)
8496
      else:
8497
        if instance_info.payload:
8498
          current_mem = int(instance_info.payload['memory'])
8499
        else:
8500
          # Assume instance not running
8501
          # (there is a slight race condition here, but it's not very probable,
8502
          # and we have no other way to check)
8503
          current_mem = 0
8504
        miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
8505
                    pninfo.payload['memory_free'])
8506
        if miss_mem > 0:
8507
          raise errors.OpPrereqError("This change will prevent the instance"
8508
                                     " from starting, due to %d MB of memory"
8509
                                     " missing on its primary node" % miss_mem,
8510
                                     errors.ECODE_NORES)
8511

    
8512
      if be_new[constants.BE_AUTO_BALANCE]:
8513
        for node, nres in nodeinfo.items():
8514
          if node not in instance.secondary_nodes:
8515
            continue
8516
          msg = nres.fail_msg
8517
          if msg:
8518
            self.warn.append("Can't get info from secondary node %s: %s" %
8519
                             (node, msg))
8520
          elif not isinstance(nres.payload.get('memory_free', None), int):
8521
            self.warn.append("Secondary node %s didn't return free"
8522
                             " memory information" % node)
8523
          elif be_new[constants.BE_MEMORY] > nres.payload['memory_free']:
8524
            self.warn.append("Not enough memory to failover instance to"
8525
                             " secondary node %s" % node)
8526

    
8527
    # NIC processing
8528
    self.nic_pnew = {}
8529
    self.nic_pinst = {}
8530
    for nic_op, nic_dict in self.op.nics:
8531
      if nic_op == constants.DDM_REMOVE:
8532
        if not instance.nics:
8533
          raise errors.OpPrereqError("Instance has no NICs, cannot remove",
8534
                                     errors.ECODE_INVAL)
8535
        continue
8536
      if nic_op != constants.DDM_ADD:
8537
        # an existing nic
8538
        if not instance.nics:
8539
          raise errors.OpPrereqError("Invalid NIC index %s, instance has"
8540
                                     " no NICs" % nic_op,
8541
                                     errors.ECODE_INVAL)
8542
        if nic_op < 0 or nic_op >= len(instance.nics):
8543
          raise errors.OpPrereqError("Invalid NIC index %s, valid values"
8544
                                     " are 0 to %d" %
8545
                                     (nic_op, len(instance.nics) - 1),
8546
                                     errors.ECODE_INVAL)
8547
        old_nic_params = instance.nics[nic_op].nicparams
8548
        old_nic_ip = instance.nics[nic_op].ip
8549
      else:
8550
        old_nic_params = {}
8551
        old_nic_ip = None
8552

    
8553
      update_params_dict = dict([(key, nic_dict[key])
8554
                                 for key in constants.NICS_PARAMETERS
8555
                                 if key in nic_dict])
8556

    
8557
      if 'bridge' in nic_dict:
8558
        update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
8559

    
8560
      new_nic_params, new_filled_nic_params = \
8561
          self._GetUpdatedParams(old_nic_params, update_params_dict,
8562
                                 cluster.nicparams[constants.PP_DEFAULT],
8563
                                 constants.NICS_PARAMETER_TYPES)
8564
      objects.NIC.CheckParameterSyntax(new_filled_nic_params)
8565
      self.nic_pinst[nic_op] = new_nic_params
8566
      self.nic_pnew[nic_op] = new_filled_nic_params
8567
      new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
8568

    
8569
      if new_nic_mode == constants.NIC_MODE_BRIDGED:
8570
        nic_bridge = new_filled_nic_params[constants.NIC_LINK]
8571
        msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
8572
        if msg:
8573
          msg = "Error checking bridges on node %s: %s" % (pnode, msg)
8574
          if self.force:
8575
            self.warn.append(msg)
8576
          else:
8577
            raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
8578
      if new_nic_mode == constants.NIC_MODE_ROUTED:
8579
        if 'ip' in nic_dict:
8580
          nic_ip = nic_dict['ip']
8581
        else:
8582
          nic_ip = old_nic_ip
8583
        if nic_ip is None:
8584
          raise errors.OpPrereqError('Cannot set the nic ip to None'
8585
                                     ' on a routed nic', errors.ECODE_INVAL)
8586
      if 'mac' in nic_dict:
8587
        nic_mac = nic_dict['mac']
8588
        if nic_mac is None:
8589
          raise errors.OpPrereqError('Cannot set the nic mac to None',
8590
                                     errors.ECODE_INVAL)
8591
        elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
8592
          # otherwise generate the mac
8593
          nic_dict['mac'] = self.cfg.GenerateMAC(self.proc.GetECId())
8594
        else:
8595
          # or validate/reserve the current one
8596
          try:
8597
            self.cfg.ReserveMAC(nic_mac, self.proc.GetECId())
8598
          except errors.ReservationError:
8599
            raise errors.OpPrereqError("MAC address %s already in use"
8600
                                       " in cluster" % nic_mac,
8601
                                       errors.ECODE_NOTUNIQUE)
8602

    
8603
    # DISK processing
8604
    if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
8605
      raise errors.OpPrereqError("Disk operations not supported for"
8606
                                 " diskless instances",
8607
                                 errors.ECODE_INVAL)
8608
    for disk_op, _ in self.op.disks:
8609
      if disk_op == constants.DDM_REMOVE:
8610
        if len(instance.disks) == 1:
8611
          raise errors.OpPrereqError("Cannot remove the last disk of"
8612
                                     " an instance", errors.ECODE_INVAL)
8613
        _CheckInstanceDown(self, instance, "cannot remove disks")
8614

    
8615
      if (disk_op == constants.DDM_ADD and
8616
          len(instance.nics) >= constants.MAX_DISKS):
8617
        raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
8618
                                   " add more" % constants.MAX_DISKS,
8619
                                   errors.ECODE_STATE)
8620
      if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
8621
        # an existing disk
8622
        if disk_op < 0 or disk_op >= len(instance.disks):
8623
          raise errors.OpPrereqError("Invalid disk index %s, valid values"
8624
                                     " are 0 to %d" %
8625
                                     (disk_op, len(instance.disks)),
8626
                                     errors.ECODE_INVAL)
8627

    
8628
    # OS change
8629
    if self.op.os_name and not self.op.force:
8630
      _CheckNodeHasOS(self, instance.primary_node, self.op.os_name,
8631
                      self.op.force_variant)
8632

    
8633
    return
8634

    
8635
  def _ConvertPlainToDrbd(self, feedback_fn):
8636
    """Converts an instance from plain to drbd.
8637

8638
    """
8639
    feedback_fn("Converting template to drbd")
8640
    instance = self.instance
8641
    pnode = instance.primary_node
8642
    snode = self.op.remote_node
8643

    
8644
    # create a fake disk info for _GenerateDiskTemplate
8645
    disk_info = [{"size": d.size, "mode": d.mode} for d in instance.disks]
8646
    new_disks = _GenerateDiskTemplate(self, self.op.disk_template,
8647
                                      instance.name, pnode, [snode],
8648
                                      disk_info, None, None, 0)
8649
    info = _GetInstanceInfoText(instance)
8650
    feedback_fn("Creating aditional volumes...")
8651
    # first, create the missing data and meta devices
8652
    for disk in new_disks:
8653
      # unfortunately this is... not too nice
8654
      _CreateSingleBlockDev(self, pnode, instance, disk.children[1],
8655
                            info, True)
8656
      for child in disk.children:
8657
        _CreateSingleBlockDev(self, snode, instance, child, info, True)
8658
    # at this stage, all new LVs have been created, we can rename the
8659
    # old ones
8660
    feedback_fn("Renaming original volumes...")
8661
    rename_list = [(o, n.children[0].logical_id)
8662
                   for (o, n) in zip(instance.disks, new_disks)]
8663
    result = self.rpc.call_blockdev_rename(pnode, rename_list)
8664
    result.Raise("Failed to rename original LVs")
8665

    
8666
    feedback_fn("Initializing DRBD devices...")
8667
    # all child devices are in place, we can now create the DRBD devices
8668
    for disk in new_disks:
8669
      for node in [pnode, snode]:
8670
        f_create = node == pnode
8671
        _CreateSingleBlockDev(self, node, instance, disk, info, f_create)
8672

    
8673
    # at this point, the instance has been modified
8674
    instance.disk_template = constants.DT_DRBD8
8675
    instance.disks = new_disks
8676
    self.cfg.Update(instance, feedback_fn)
8677

    
8678
    # disks are created, waiting for sync
8679
    disk_abort = not _WaitForSync(self, instance)
8680
    if disk_abort:
8681
      raise errors.OpExecError("There are some degraded disks for"
8682
                               " this instance, please cleanup manually")
8683

    
8684
  def _ConvertDrbdToPlain(self, feedback_fn):
8685
    """Converts an instance from drbd to plain.
8686

8687
    """
8688
    instance = self.instance
8689
    assert len(instance.secondary_nodes) == 1
8690
    pnode = instance.primary_node
8691
    snode = instance.secondary_nodes[0]
8692
    feedback_fn("Converting template to plain")
8693

    
8694
    old_disks = instance.disks
8695
    new_disks = [d.children[0] for d in old_disks]
8696

    
8697
    # copy over size and mode
8698
    for parent, child in zip(old_disks, new_disks):
8699
      child.size = parent.size
8700
      child.mode = parent.mode
8701

    
8702
    # update instance structure
8703
    instance.disks = new_disks
8704
    instance.disk_template = constants.DT_PLAIN
8705
    self.cfg.Update(instance, feedback_fn)
8706

    
8707
    feedback_fn("Removing volumes on the secondary node...")
8708
    for disk in old_disks:
8709
      self.cfg.SetDiskID(disk, snode)
8710
      msg = self.rpc.call_blockdev_remove(snode, disk).fail_msg
8711
      if msg:
8712
        self.LogWarning("Could not remove block device %s on node %s,"
8713
                        " continuing anyway: %s", disk.iv_name, snode, msg)
8714

    
8715
    feedback_fn("Removing unneeded volumes on the primary node...")
8716
    for idx, disk in enumerate(old_disks):
8717
      meta = disk.children[1]
8718
      self.cfg.SetDiskID(meta, pnode)
8719
      msg = self.rpc.call_blockdev_remove(pnode, meta).fail_msg
8720
      if msg:
8721
        self.LogWarning("Could not remove metadata for disk %d on node %s,"
8722
                        " continuing anyway: %s", idx, pnode, msg)
8723

    
8724

    
8725
  def Exec(self, feedback_fn):
8726
    """Modifies an instance.
8727

8728
    All parameters take effect only at the next restart of the instance.
8729

8730
    """
8731
    # Process here the warnings from CheckPrereq, as we don't have a
8732
    # feedback_fn there.
8733
    for warn in self.warn:
8734
      feedback_fn("WARNING: %s" % warn)
8735

    
8736
    result = []
8737
    instance = self.instance
8738
    # disk changes
8739
    for disk_op, disk_dict in self.op.disks:
8740
      if disk_op == constants.DDM_REMOVE:
8741
        # remove the last disk
8742
        device = instance.disks.pop()
8743
        device_idx = len(instance.disks)
8744
        for node, disk in device.ComputeNodeTree(instance.primary_node):
8745
          self.cfg.SetDiskID(disk, node)
8746
          msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
8747
          if msg:
8748
            self.LogWarning("Could not remove disk/%d on node %s: %s,"
8749
                            " continuing anyway", device_idx, node, msg)
8750
        result.append(("disk/%d" % device_idx, "remove"))
8751
      elif disk_op == constants.DDM_ADD:
8752
        # add a new disk
8753
        if instance.disk_template == constants.DT_FILE:
8754
          file_driver, file_path = instance.disks[0].logical_id
8755
          file_path = os.path.dirname(file_path)
8756
        else:
8757
          file_driver = file_path = None
8758
        disk_idx_base = len(instance.disks)
8759
        new_disk = _GenerateDiskTemplate(self,
8760
                                         instance.disk_template,
8761
                                         instance.name, instance.primary_node,
8762
                                         instance.secondary_nodes,
8763
                                         [disk_dict],
8764
                                         file_path,
8765
                                         file_driver,
8766
                                         disk_idx_base)[0]
8767
        instance.disks.append(new_disk)
8768
        info = _GetInstanceInfoText(instance)
8769

    
8770
        logging.info("Creating volume %s for instance %s",
8771
                     new_disk.iv_name, instance.name)
8772
        # Note: this needs to be kept in sync with _CreateDisks
8773
        #HARDCODE
8774
        for node in instance.all_nodes:
8775
          f_create = node == instance.primary_node
8776
          try:
8777
            _CreateBlockDev(self, node, instance, new_disk,
8778
                            f_create, info, f_create)
8779
          except errors.OpExecError, err:
8780
            self.LogWarning("Failed to create volume %s (%s) on"
8781
                            " node %s: %s",
8782
                            new_disk.iv_name, new_disk, node, err)
8783
        result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
8784
                       (new_disk.size, new_disk.mode)))
8785
      else:
8786
        # change a given disk
8787
        instance.disks[disk_op].mode = disk_dict['mode']
8788
        result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
8789

    
8790
    if self.op.disk_template:
8791
      r_shut = _ShutdownInstanceDisks(self, instance)
8792
      if not r_shut:
8793
        raise errors.OpExecError("Cannot shutdow instance disks, unable to"
8794
                                 " proceed with disk template conversion")
8795
      mode = (instance.disk_template, self.op.disk_template)
8796
      try:
8797
        self._DISK_CONVERSIONS[mode](self, feedback_fn)
8798
      except:
8799
        self.cfg.ReleaseDRBDMinors(instance.name)
8800
        raise
8801
      result.append(("disk_template", self.op.disk_template))
8802

    
8803
    # NIC changes
8804
    for nic_op, nic_dict in self.op.nics:
8805
      if nic_op == constants.DDM_REMOVE:
8806
        # remove the last nic
8807
        del instance.nics[-1]
8808
        result.append(("nic.%d" % len(instance.nics), "remove"))
8809
      elif nic_op == constants.DDM_ADD:
8810
        # mac and bridge should be set, by now
8811
        mac = nic_dict['mac']
8812
        ip = nic_dict.get('ip', None)
8813
        nicparams = self.nic_pinst[constants.DDM_ADD]
8814
        new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
8815
        instance.nics.append(new_nic)
8816
        result.append(("nic.%d" % (len(instance.nics) - 1),
8817
                       "add:mac=%s,ip=%s,mode=%s,link=%s" %
8818
                       (new_nic.mac, new_nic.ip,
8819
                        self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
8820
                        self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
8821
                       )))
8822
      else:
8823
        for key in 'mac', 'ip':
8824
          if key in nic_dict:
8825
            setattr(instance.nics[nic_op], key, nic_dict[key])
8826
        if nic_op in self.nic_pinst:
8827
          instance.nics[nic_op].nicparams = self.nic_pinst[nic_op]
8828
        for key, val in nic_dict.iteritems():
8829
          result.append(("nic.%s/%d" % (key, nic_op), val))
8830

    
8831
    # hvparams changes
8832
    if self.op.hvparams:
8833
      instance.hvparams = self.hv_inst
8834
      for key, val in self.op.hvparams.iteritems():
8835
        result.append(("hv/%s" % key, val))
8836

    
8837
    # beparams changes
8838
    if self.op.beparams:
8839
      instance.beparams = self.be_inst
8840
      for key, val in self.op.beparams.iteritems():
8841
        result.append(("be/%s" % key, val))
8842

    
8843
    # OS change
8844
    if self.op.os_name:
8845
      instance.os = self.op.os_name
8846

    
8847
    self.cfg.Update(instance, feedback_fn)
8848

    
8849
    return result
8850

    
8851
  _DISK_CONVERSIONS = {
8852
    (constants.DT_PLAIN, constants.DT_DRBD8): _ConvertPlainToDrbd,
8853
    (constants.DT_DRBD8, constants.DT_PLAIN): _ConvertDrbdToPlain,
8854
    }
8855

    
8856

    
8857
class LUQueryExports(NoHooksLU):
8858
  """Query the exports list
8859

8860
  """
8861
  _OP_REQP = ['nodes']
8862
  REQ_BGL = False
8863

    
8864
  def ExpandNames(self):
8865
    self.needed_locks = {}
8866
    self.share_locks[locking.LEVEL_NODE] = 1
8867
    if not self.op.nodes:
8868
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
8869
    else:
8870
      self.needed_locks[locking.LEVEL_NODE] = \
8871
        _GetWantedNodes(self, self.op.nodes)
8872

    
8873
  def CheckPrereq(self):
8874
    """Check prerequisites.
8875

8876
    """
8877
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
8878

    
8879
  def Exec(self, feedback_fn):
8880
    """Compute the list of all the exported system images.
8881

8882
    @rtype: dict
8883
    @return: a dictionary with the structure node->(export-list)
8884
        where export-list is a list of the instances exported on
8885
        that node.
8886

8887
    """
8888
    rpcresult = self.rpc.call_export_list(self.nodes)
8889
    result = {}
8890
    for node in rpcresult:
8891
      if rpcresult[node].fail_msg:
8892
        result[node] = False
8893
      else:
8894
        result[node] = rpcresult[node].payload
8895

    
8896
    return result
8897

    
8898

    
8899
class LUPrepareExport(NoHooksLU):
8900
  """Prepares an instance for an export and returns useful information.
8901

8902
  """
8903
  _OP_REQP = ["instance_name", "mode"]
8904
  REQ_BGL = False
8905

    
8906
  def CheckArguments(self):
8907
    """Check the arguments.
8908

8909
    """
8910
    if self.op.mode not in constants.EXPORT_MODES:
8911
      raise errors.OpPrereqError("Invalid export mode %r" % self.op.mode,
8912
                                 errors.ECODE_INVAL)
8913

    
8914
  def ExpandNames(self):
8915
    self._ExpandAndLockInstance()
8916

    
8917
  def CheckPrereq(self):
8918
    """Check prerequisites.
8919

8920
    """
8921
    instance_name = self.op.instance_name
8922

    
8923
    self.instance = self.cfg.GetInstanceInfo(instance_name)
8924
    assert self.instance is not None, \
8925
          "Cannot retrieve locked instance %s" % self.op.instance_name
8926
    _CheckNodeOnline(self, self.instance.primary_node)
8927

    
8928
    self._cds = _GetClusterDomainSecret()
8929

    
8930
  def Exec(self, feedback_fn):
8931
    """Prepares an instance for an export.
8932

8933
    """
8934
    instance = self.instance
8935

    
8936
    if self.op.mode == constants.EXPORT_MODE_REMOTE:
8937
      salt = utils.GenerateSecret(8)
8938

    
8939
      feedback_fn("Generating X509 certificate on %s" % instance.primary_node)
8940
      result = self.rpc.call_x509_cert_create(instance.primary_node,
8941
                                              constants.RIE_CERT_VALIDITY)
8942
      result.Raise("Can't create X509 key and certificate on %s" % result.node)
8943

    
8944
      (name, cert_pem) = result.payload
8945

    
8946
      cert = OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
8947
                                             cert_pem)
8948

    
8949
      return {
8950
        "handshake": masterd.instance.ComputeRemoteExportHandshake(self._cds),
8951
        "x509_key_name": (name, utils.Sha1Hmac(self._cds, name, salt=salt),
8952
                          salt),
8953
        "x509_ca": utils.SignX509Certificate(cert, self._cds, salt),
8954
        }
8955

    
8956
    return None
8957

    
8958

    
8959
class LUExportInstance(LogicalUnit):
8960
  """Export an instance to an image in the cluster.
8961

8962
  """
8963
  HPATH = "instance-export"
8964
  HTYPE = constants.HTYPE_INSTANCE
8965
  _OP_REQP = ["instance_name", "target_node", "shutdown"]
8966
  REQ_BGL = False
8967

    
8968
  def CheckArguments(self):
8969
    """Check the arguments.
8970

8971
    """
8972
    _CheckBooleanOpField(self.op, "remove_instance")
8973
    _CheckBooleanOpField(self.op, "ignore_remove_failures")
8974

    
8975
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
8976
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
8977
    self.remove_instance = getattr(self.op, "remove_instance", False)
8978
    self.ignore_remove_failures = getattr(self.op, "ignore_remove_failures",
8979
                                          False)
8980
    self.export_mode = getattr(self.op, "mode", constants.EXPORT_MODE_LOCAL)
8981
    self.x509_key_name = getattr(self.op, "x509_key_name", None)
8982
    self.dest_x509_ca_pem = getattr(self.op, "destination_x509_ca", None)
8983

    
8984
    if self.remove_instance and not self.op.shutdown:
8985
      raise errors.OpPrereqError("Can not remove instance without shutting it"
8986
                                 " down before")
8987

    
8988
    if self.export_mode not in constants.EXPORT_MODES:
8989
      raise errors.OpPrereqError("Invalid export mode %r" % self.export_mode,
8990
                                 errors.ECODE_INVAL)
8991

    
8992
    if self.export_mode == constants.EXPORT_MODE_REMOTE:
8993
      if not self.x509_key_name:
8994
        raise errors.OpPrereqError("Missing X509 key name for encryption",
8995
                                   errors.ECODE_INVAL)
8996

    
8997
      if not self.dest_x509_ca_pem:
8998
        raise errors.OpPrereqError("Missing destination X509 CA",
8999
                                   errors.ECODE_INVAL)
9000

    
9001
  def ExpandNames(self):
9002
    self._ExpandAndLockInstance()
9003

    
9004
    # Lock all nodes for local exports
9005
    if self.export_mode == constants.EXPORT_MODE_LOCAL:
9006
      # FIXME: lock only instance primary and destination node
9007
      #
9008
      # Sad but true, for now we have do lock all nodes, as we don't know where
9009
      # the previous export might be, and in this LU we search for it and
9010
      # remove it from its current node. In the future we could fix this by:
9011
      #  - making a tasklet to search (share-lock all), then create the new one,
9012
      #    then one to remove, after
9013
      #  - removing the removal operation altogether
9014
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9015

    
9016
  def DeclareLocks(self, level):
9017
    """Last minute lock declaration."""
9018
    # All nodes are locked anyway, so nothing to do here.
9019

    
9020
  def BuildHooksEnv(self):
9021
    """Build hooks env.
9022

9023
    This will run on the master, primary node and target node.
9024

9025
    """
9026
    env = {
9027
      "EXPORT_MODE": self.export_mode,
9028
      "EXPORT_NODE": self.op.target_node,
9029
      "EXPORT_DO_SHUTDOWN": self.op.shutdown,
9030
      "SHUTDOWN_TIMEOUT": self.shutdown_timeout,
9031
      # TODO: Generic function for boolean env variables
9032
      "REMOVE_INSTANCE": str(bool(self.remove_instance)),
9033
      }
9034

    
9035
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
9036

    
9037
    nl = [self.cfg.GetMasterNode(), self.instance.primary_node]
9038

    
9039
    if self.export_mode == constants.EXPORT_MODE_LOCAL:
9040
      nl.append(self.op.target_node)
9041

    
9042
    return env, nl, nl
9043

    
9044
  def CheckPrereq(self):
9045
    """Check prerequisites.
9046

9047
    This checks that the instance and node names are valid.
9048

9049
    """
9050
    instance_name = self.op.instance_name
9051

    
9052
    self.instance = self.cfg.GetInstanceInfo(instance_name)
9053
    assert self.instance is not None, \
9054
          "Cannot retrieve locked instance %s" % self.op.instance_name
9055
    _CheckNodeOnline(self, self.instance.primary_node)
9056

    
9057
    if self.export_mode == constants.EXPORT_MODE_LOCAL:
9058
      self.op.target_node = _ExpandNodeName(self.cfg, self.op.target_node)
9059
      self.dst_node = self.cfg.GetNodeInfo(self.op.target_node)
9060
      assert self.dst_node is not None
9061

    
9062
      _CheckNodeOnline(self, self.dst_node.name)
9063
      _CheckNodeNotDrained(self, self.dst_node.name)
9064

    
9065
      self._cds = None
9066
      self.dest_x509_ca = None
9067

    
9068
    elif self.export_mode == constants.EXPORT_MODE_REMOTE:
9069
      self.dst_node = None
9070

    
9071
      if len(self.op.target_node) != len(self.instance.disks):
9072
        raise errors.OpPrereqError(("Received destination information for %s"
9073
                                    " disks, but instance %s has %s disks") %
9074
                                   (len(self.op.target_node), instance_name,
9075
                                    len(self.instance.disks)),
9076
                                   errors.ECODE_INVAL)
9077

    
9078
      cds = _GetClusterDomainSecret()
9079

    
9080
      # Check X509 key name
9081
      try:
9082
        (key_name, hmac_digest, hmac_salt) = self.x509_key_name
9083
      except (TypeError, ValueError), err:
9084
        raise errors.OpPrereqError("Invalid data for X509 key name: %s" % err)
9085

    
9086
      if not utils.VerifySha1Hmac(cds, key_name, hmac_digest, salt=hmac_salt):
9087
        raise errors.OpPrereqError("HMAC for X509 key name is wrong",
9088
                                   errors.ECODE_INVAL)
9089

    
9090
      # Load and verify CA
9091
      try:
9092
        (cert, _) = utils.LoadSignedX509Certificate(self.dest_x509_ca_pem, cds)
9093
      except OpenSSL.crypto.Error, err:
9094
        raise errors.OpPrereqError("Unable to load destination X509 CA (%s)" %
9095
                                   (err, ), errors.ECODE_INVAL)
9096

    
9097
      (errcode, msg) = utils.VerifyX509Certificate(cert, None, None)
9098
      if errcode is not None:
9099
        raise errors.OpPrereqError("Invalid destination X509 CA (%s)" % (msg, ),
9100
                                   errors.ECODE_INVAL)
9101

    
9102
      self.dest_x509_ca = cert
9103

    
9104
      # Verify target information
9105
      for idx, disk_data in enumerate(self.op.target_node):
9106
        try:
9107
          masterd.instance.CheckRemoteExportDiskInfo(cds, idx, disk_data)
9108
        except errors.GenericError, err:
9109
          raise errors.OpPrereqError("Target info for disk %s: %s" % (idx, err),
9110
                                     errors.ECODE_INVAL)
9111

    
9112
    else:
9113
      raise errors.ProgrammerError("Unhandled export mode %r" %
9114
                                   self.export_mode)
9115

    
9116
    # instance disk type verification
9117
    # TODO: Implement export support for file-based disks
9118
    for disk in self.instance.disks:
9119
      if disk.dev_type == constants.LD_FILE:
9120
        raise errors.OpPrereqError("Export not supported for instances with"
9121
                                   " file-based disks", errors.ECODE_INVAL)
9122

    
9123
  def _CleanupExports(self, feedback_fn):
9124
    """Removes exports of current instance from all other nodes.
9125

9126
    If an instance in a cluster with nodes A..D was exported to node C, its
9127
    exports will be removed from the nodes A, B and D.
9128

9129
    """
9130
    assert self.export_mode != constants.EXPORT_MODE_REMOTE
9131

    
9132
    nodelist = self.cfg.GetNodeList()
9133
    nodelist.remove(self.dst_node.name)
9134

    
9135
    # on one-node clusters nodelist will be empty after the removal
9136
    # if we proceed the backup would be removed because OpQueryExports
9137
    # substitutes an empty list with the full cluster node list.
9138
    iname = self.instance.name
9139
    if nodelist:
9140
      feedback_fn("Removing old exports for instance %s" % iname)
9141
      exportlist = self.rpc.call_export_list(nodelist)
9142
      for node in exportlist:
9143
        if exportlist[node].fail_msg:
9144
          continue
9145
        if iname in exportlist[node].payload:
9146
          msg = self.rpc.call_export_remove(node, iname).fail_msg
9147
          if msg:
9148
            self.LogWarning("Could not remove older export for instance %s"
9149
                            " on node %s: %s", iname, node, msg)
9150

    
9151
  def Exec(self, feedback_fn):
9152
    """Export an instance to an image in the cluster.
9153

9154
    """
9155
    assert self.export_mode in constants.EXPORT_MODES
9156

    
9157
    instance = self.instance
9158
    src_node = instance.primary_node
9159

    
9160
    if self.op.shutdown:
9161
      # shutdown the instance, but not the disks
9162
      feedback_fn("Shutting down instance %s" % instance.name)
9163
      result = self.rpc.call_instance_shutdown(src_node, instance,
9164
                                               self.shutdown_timeout)
9165
      # TODO: Maybe ignore failures if ignore_remove_failures is set
9166
      result.Raise("Could not shutdown instance %s on"
9167
                   " node %s" % (instance.name, src_node))
9168

    
9169
    # set the disks ID correctly since call_instance_start needs the
9170
    # correct drbd minor to create the symlinks
9171
    for disk in instance.disks:
9172
      self.cfg.SetDiskID(disk, src_node)
9173

    
9174
    activate_disks = (not instance.admin_up)
9175

    
9176
    if activate_disks:
9177
      # Activate the instance disks if we'exporting a stopped instance
9178
      feedback_fn("Activating disks for %s" % instance.name)
9179
      _StartInstanceDisks(self, instance, None)
9180

    
9181
    try:
9182
      helper = masterd.instance.ExportInstanceHelper(self, feedback_fn,
9183
                                                     instance)
9184

    
9185
      helper.CreateSnapshots()
9186
      try:
9187
        if self.export_mode == constants.EXPORT_MODE_LOCAL:
9188
          (fin_resu, dresults) = helper.LocalExport(self.dst_node)
9189
        elif self.export_mode == constants.EXPORT_MODE_REMOTE:
9190
          connect_timeout = constants.RIE_CONNECT_TIMEOUT
9191
          timeouts = masterd.instance.ImportExportTimeouts(connect_timeout)
9192

    
9193
          (key_name, _, _) = self.x509_key_name
9194

    
9195
          dest_ca_pem = \
9196
            OpenSSL.crypto.dump_certificate(OpenSSL.crypto.FILETYPE_PEM,
9197
                                            self.dest_x509_ca)
9198

    
9199
          opts = objects.ImportExportOptions(key_name=key_name,
9200
                                             ca_pem=dest_ca_pem)
9201

    
9202
          (fin_resu, dresults) = helper.RemoteExport(opts, self.op.target_node,
9203
                                                     timeouts)
9204
      finally:
9205
        helper.Cleanup()
9206

    
9207
      # Check for backwards compatibility
9208
      assert len(dresults) == len(instance.disks)
9209
      assert compat.all(isinstance(i, bool) for i in dresults), \
9210
             "Not all results are boolean: %r" % dresults
9211

    
9212
    finally:
9213
      if activate_disks:
9214
        feedback_fn("Deactivating disks for %s" % instance.name)
9215
        _ShutdownInstanceDisks(self, instance)
9216

    
9217
    # Remove instance if requested
9218
    if self.remove_instance:
9219
      if not (compat.all(dresults) and fin_resu):
9220
        feedback_fn("Not removing instance %s as parts of the export failed" %
9221
                    instance.name)
9222
      else:
9223
        feedback_fn("Removing instance %s" % instance.name)
9224
        _RemoveInstance(self, feedback_fn, instance,
9225
                        self.ignore_remove_failures)
9226

    
9227
    if self.export_mode == constants.EXPORT_MODE_LOCAL:
9228
      self._CleanupExports(feedback_fn)
9229

    
9230
    return fin_resu, dresults
9231

    
9232

    
9233
class LURemoveExport(NoHooksLU):
9234
  """Remove exports related to the named instance.
9235

9236
  """
9237
  _OP_REQP = ["instance_name"]
9238
  REQ_BGL = False
9239

    
9240
  def ExpandNames(self):
9241
    self.needed_locks = {}
9242
    # We need all nodes to be locked in order for RemoveExport to work, but we
9243
    # don't need to lock the instance itself, as nothing will happen to it (and
9244
    # we can remove exports also for a removed instance)
9245
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
9246

    
9247
  def CheckPrereq(self):
9248
    """Check prerequisites.
9249
    """
9250
    pass
9251

    
9252
  def Exec(self, feedback_fn):
9253
    """Remove any export.
9254

9255
    """
9256
    instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
9257
    # If the instance was not found we'll try with the name that was passed in.
9258
    # This will only work if it was an FQDN, though.
9259
    fqdn_warn = False
9260
    if not instance_name:
9261
      fqdn_warn = True
9262
      instance_name = self.op.instance_name
9263

    
9264
    locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
9265
    exportlist = self.rpc.call_export_list(locked_nodes)
9266
    found = False
9267
    for node in exportlist:
9268
      msg = exportlist[node].fail_msg
9269
      if msg:
9270
        self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
9271
        continue
9272
      if instance_name in exportlist[node].payload:
9273
        found = True
9274
        result = self.rpc.call_export_remove(node, instance_name)
9275
        msg = result.fail_msg
9276
        if msg:
9277
          logging.error("Could not remove export for instance %s"
9278
                        " on node %s: %s", instance_name, node, msg)
9279

    
9280
    if fqdn_warn and not found:
9281
      feedback_fn("Export not found. If trying to remove an export belonging"
9282
                  " to a deleted instance please use its Fully Qualified"
9283
                  " Domain Name.")
9284

    
9285

    
9286
class TagsLU(NoHooksLU): # pylint: disable-msg=W0223
9287
  """Generic tags LU.
9288

9289
  This is an abstract class which is the parent of all the other tags LUs.
9290

9291
  """
9292

    
9293
  def ExpandNames(self):
9294
    self.needed_locks = {}
9295
    if self.op.kind == constants.TAG_NODE:
9296
      self.op.name = _ExpandNodeName(self.cfg, self.op.name)
9297
      self.needed_locks[locking.LEVEL_NODE] = self.op.name
9298
    elif self.op.kind == constants.TAG_INSTANCE:
9299
      self.op.name = _ExpandInstanceName(self.cfg, self.op.name)
9300
      self.needed_locks[locking.LEVEL_INSTANCE] = self.op.name
9301

    
9302
  def CheckPrereq(self):
9303
    """Check prerequisites.
9304

9305
    """
9306
    if self.op.kind == constants.TAG_CLUSTER:
9307
      self.target = self.cfg.GetClusterInfo()
9308
    elif self.op.kind == constants.TAG_NODE:
9309
      self.target = self.cfg.GetNodeInfo(self.op.name)
9310
    elif self.op.kind == constants.TAG_INSTANCE:
9311
      self.target = self.cfg.GetInstanceInfo(self.op.name)
9312
    else:
9313
      raise errors.OpPrereqError("Wrong tag type requested (%s)" %
9314
                                 str(self.op.kind), errors.ECODE_INVAL)
9315

    
9316

    
9317
class LUGetTags(TagsLU):
9318
  """Returns the tags of a given object.
9319

9320
  """
9321
  _OP_REQP = ["kind", "name"]
9322
  REQ_BGL = False
9323

    
9324
  def Exec(self, feedback_fn):
9325
    """Returns the tag list.
9326

9327
    """
9328
    return list(self.target.GetTags())
9329

    
9330

    
9331
class LUSearchTags(NoHooksLU):
9332
  """Searches the tags for a given pattern.
9333

9334
  """
9335
  _OP_REQP = ["pattern"]
9336
  REQ_BGL = False
9337

    
9338
  def ExpandNames(self):
9339
    self.needed_locks = {}
9340

    
9341
  def CheckPrereq(self):
9342
    """Check prerequisites.
9343

9344
    This checks the pattern passed for validity by compiling it.
9345

9346
    """
9347
    try:
9348
      self.re = re.compile(self.op.pattern)
9349
    except re.error, err:
9350
      raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
9351
                                 (self.op.pattern, err), errors.ECODE_INVAL)
9352

    
9353
  def Exec(self, feedback_fn):
9354
    """Returns the tag list.
9355

9356
    """
9357
    cfg = self.cfg
9358
    tgts = [("/cluster", cfg.GetClusterInfo())]
9359
    ilist = cfg.GetAllInstancesInfo().values()
9360
    tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
9361
    nlist = cfg.GetAllNodesInfo().values()
9362
    tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
9363
    results = []
9364
    for path, target in tgts:
9365
      for tag in target.GetTags():
9366
        if self.re.search(tag):
9367
          results.append((path, tag))
9368
    return results
9369

    
9370

    
9371
class LUAddTags(TagsLU):
9372
  """Sets a tag on a given object.
9373

9374
  """
9375
  _OP_REQP = ["kind", "name", "tags"]
9376
  REQ_BGL = False
9377

    
9378
  def CheckPrereq(self):
9379
    """Check prerequisites.
9380

9381
    This checks the type and length of the tag name and value.
9382

9383
    """
9384
    TagsLU.CheckPrereq(self)
9385
    for tag in self.op.tags:
9386
      objects.TaggableObject.ValidateTag(tag)
9387

    
9388
  def Exec(self, feedback_fn):
9389
    """Sets the tag.
9390

9391
    """
9392
    try:
9393
      for tag in self.op.tags:
9394
        self.target.AddTag(tag)
9395
    except errors.TagError, err:
9396
      raise errors.OpExecError("Error while setting tag: %s" % str(err))
9397
    self.cfg.Update(self.target, feedback_fn)
9398

    
9399

    
9400
class LUDelTags(TagsLU):
9401
  """Delete a list of tags from a given object.
9402

9403
  """
9404
  _OP_REQP = ["kind", "name", "tags"]
9405
  REQ_BGL = False
9406

    
9407
  def CheckPrereq(self):
9408
    """Check prerequisites.
9409

9410
    This checks that we have the given tag.
9411

9412
    """
9413
    TagsLU.CheckPrereq(self)
9414
    for tag in self.op.tags:
9415
      objects.TaggableObject.ValidateTag(tag)
9416
    del_tags = frozenset(self.op.tags)
9417
    cur_tags = self.target.GetTags()
9418
    if not del_tags <= cur_tags:
9419
      diff_tags = del_tags - cur_tags
9420
      diff_names = ["'%s'" % tag for tag in diff_tags]
9421
      diff_names.sort()
9422
      raise errors.OpPrereqError("Tag(s) %s not found" %
9423
                                 (",".join(diff_names)), errors.ECODE_NOENT)
9424

    
9425
  def Exec(self, feedback_fn):
9426
    """Remove the tag from the object.
9427

9428
    """
9429
    for tag in self.op.tags:
9430
      self.target.RemoveTag(tag)
9431
    self.cfg.Update(self.target, feedback_fn)
9432

    
9433

    
9434
class LUTestDelay(NoHooksLU):
9435
  """Sleep for a specified amount of time.
9436

9437
  This LU sleeps on the master and/or nodes for a specified amount of
9438
  time.
9439

9440
  """
9441
  _OP_REQP = ["duration", "on_master", "on_nodes"]
9442
  REQ_BGL = False
9443

    
9444
  def ExpandNames(self):
9445
    """Expand names and set required locks.
9446

9447
    This expands the node list, if any.
9448

9449
    """
9450
    self.needed_locks = {}
9451
    if self.op.on_nodes:
9452
      # _GetWantedNodes can be used here, but is not always appropriate to use
9453
      # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
9454
      # more information.
9455
      self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
9456
      self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
9457

    
9458
  def CheckPrereq(self):
9459
    """Check prerequisites.
9460

9461
    """
9462

    
9463
  def Exec(self, feedback_fn):
9464
    """Do the actual sleep.
9465

9466
    """
9467
    if self.op.on_master:
9468
      if not utils.TestDelay(self.op.duration):
9469
        raise errors.OpExecError("Error during master delay test")
9470
    if self.op.on_nodes:
9471
      result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
9472
      for node, node_result in result.items():
9473
        node_result.Raise("Failure during rpc call to node %s" % node)
9474

    
9475

    
9476
class IAllocator(object):
9477
  """IAllocator framework.
9478

9479
  An IAllocator instance has three sets of attributes:
9480
    - cfg that is needed to query the cluster
9481
    - input data (all members of the _KEYS class attribute are required)
9482
    - four buffer attributes (in|out_data|text), that represent the
9483
      input (to the external script) in text and data structure format,
9484
      and the output from it, again in two formats
9485
    - the result variables from the script (success, info, nodes) for
9486
      easy usage
9487

9488
  """
9489
  # pylint: disable-msg=R0902
9490
  # lots of instance attributes
9491
  _ALLO_KEYS = [
9492
    "name", "mem_size", "disks", "disk_template",
9493
    "os", "tags", "nics", "vcpus", "hypervisor",
9494
    ]
9495
  _RELO_KEYS = [
9496
    "name", "relocate_from",
9497
    ]
9498
  _EVAC_KEYS = [
9499
    "evac_nodes",
9500
    ]
9501

    
9502
  def __init__(self, cfg, rpc, mode, **kwargs):
9503
    self.cfg = cfg
9504
    self.rpc = rpc
9505
    # init buffer variables
9506
    self.in_text = self.out_text = self.in_data = self.out_data = None
9507
    # init all input fields so that pylint is happy
9508
    self.mode = mode
9509
    self.mem_size = self.disks = self.disk_template = None
9510
    self.os = self.tags = self.nics = self.vcpus = None
9511
    self.hypervisor = None
9512
    self.relocate_from = None
9513
    self.name = None
9514
    self.evac_nodes = None
9515
    # computed fields
9516
    self.required_nodes = None
9517
    # init result fields
9518
    self.success = self.info = self.result = None
9519
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
9520
      keyset = self._ALLO_KEYS
9521
      fn = self._AddNewInstance
9522
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
9523
      keyset = self._RELO_KEYS
9524
      fn = self._AddRelocateInstance
9525
    elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
9526
      keyset = self._EVAC_KEYS
9527
      fn = self._AddEvacuateNodes
9528
    else:
9529
      raise errors.ProgrammerError("Unknown mode '%s' passed to the"
9530
                                   " IAllocator" % self.mode)
9531
    for key in kwargs:
9532
      if key not in keyset:
9533
        raise errors.ProgrammerError("Invalid input parameter '%s' to"
9534
                                     " IAllocator" % key)
9535
      setattr(self, key, kwargs[key])
9536

    
9537
    for key in keyset:
9538
      if key not in kwargs:
9539
        raise errors.ProgrammerError("Missing input parameter '%s' to"
9540
                                     " IAllocator" % key)
9541
    self._BuildInputData(fn)
9542

    
9543
  def _ComputeClusterData(self):
9544
    """Compute the generic allocator input data.
9545

9546
    This is the data that is independent of the actual operation.
9547

9548
    """
9549
    cfg = self.cfg
9550
    cluster_info = cfg.GetClusterInfo()
9551
    # cluster data
9552
    data = {
9553
      "version": constants.IALLOCATOR_VERSION,
9554
      "cluster_name": cfg.GetClusterName(),
9555
      "cluster_tags": list(cluster_info.GetTags()),
9556
      "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
9557
      # we don't have job IDs
9558
      }
9559
    iinfo = cfg.GetAllInstancesInfo().values()
9560
    i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
9561

    
9562
    # node data
9563
    node_results = {}
9564
    node_list = cfg.GetNodeList()
9565

    
9566
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
9567
      hypervisor_name = self.hypervisor
9568
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
9569
      hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
9570
    elif self.mode == constants.IALLOCATOR_MODE_MEVAC:
9571
      hypervisor_name = cluster_info.enabled_hypervisors[0]
9572

    
9573
    node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(),
9574
                                        hypervisor_name)
9575
    node_iinfo = \
9576
      self.rpc.call_all_instances_info(node_list,
9577
                                       cluster_info.enabled_hypervisors)
9578
    for nname, nresult in node_data.items():
9579
      # first fill in static (config-based) values
9580
      ninfo = cfg.GetNodeInfo(nname)
9581
      pnr = {
9582
        "tags": list(ninfo.GetTags()),
9583
        "primary_ip": ninfo.primary_ip,
9584
        "secondary_ip": ninfo.secondary_ip,
9585
        "offline": ninfo.offline,
9586
        "drained": ninfo.drained,
9587
        "master_candidate": ninfo.master_candidate,
9588
        }
9589

    
9590
      if not (ninfo.offline or ninfo.drained):
9591
        nresult.Raise("Can't get data for node %s" % nname)
9592
        node_iinfo[nname].Raise("Can't get node instance info from node %s" %
9593
                                nname)
9594
        remote_info = nresult.payload
9595

    
9596
        for attr in ['memory_total', 'memory_free', 'memory_dom0',
9597
                     'vg_size', 'vg_free', 'cpu_total']:
9598
          if attr not in remote_info:
9599
            raise errors.OpExecError("Node '%s' didn't return attribute"
9600
                                     " '%s'" % (nname, attr))
9601
          if not isinstance(remote_info[attr], int):
9602
            raise errors.OpExecError("Node '%s' returned invalid value"
9603
                                     " for '%s': %s" %
9604
                                     (nname, attr, remote_info[attr]))
9605
        # compute memory used by primary instances
9606
        i_p_mem = i_p_up_mem = 0
9607
        for iinfo, beinfo in i_list:
9608
          if iinfo.primary_node == nname:
9609
            i_p_mem += beinfo[constants.BE_MEMORY]
9610
            if iinfo.name not in node_iinfo[nname].payload:
9611
              i_used_mem = 0
9612
            else:
9613
              i_used_mem = int(node_iinfo[nname].payload[iinfo.name]['memory'])
9614
            i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
9615
            remote_info['memory_free'] -= max(0, i_mem_diff)
9616

    
9617
            if iinfo.admin_up:
9618
              i_p_up_mem += beinfo[constants.BE_MEMORY]
9619

    
9620
        # compute memory used by instances
9621
        pnr_dyn = {
9622
          "total_memory": remote_info['memory_total'],
9623
          "reserved_memory": remote_info['memory_dom0'],
9624
          "free_memory": remote_info['memory_free'],
9625
          "total_disk": remote_info['vg_size'],
9626
          "free_disk": remote_info['vg_free'],
9627
          "total_cpus": remote_info['cpu_total'],
9628
          "i_pri_memory": i_p_mem,
9629
          "i_pri_up_memory": i_p_up_mem,
9630
          }
9631
        pnr.update(pnr_dyn)
9632

    
9633
      node_results[nname] = pnr
9634
    data["nodes"] = node_results
9635

    
9636
    # instance data
9637
    instance_data = {}
9638
    for iinfo, beinfo in i_list:
9639
      nic_data = []
9640
      for nic in iinfo.nics:
9641
        filled_params = objects.FillDict(
9642
            cluster_info.nicparams[constants.PP_DEFAULT],
9643
            nic.nicparams)
9644
        nic_dict = {"mac": nic.mac,
9645
                    "ip": nic.ip,
9646
                    "mode": filled_params[constants.NIC_MODE],
9647
                    "link": filled_params[constants.NIC_LINK],
9648
                   }
9649
        if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
9650
          nic_dict["bridge"] = filled_params[constants.NIC_LINK]
9651
        nic_data.append(nic_dict)
9652
      pir = {
9653
        "tags": list(iinfo.GetTags()),
9654
        "admin_up": iinfo.admin_up,
9655
        "vcpus": beinfo[constants.BE_VCPUS],
9656
        "memory": beinfo[constants.BE_MEMORY],
9657
        "os": iinfo.os,
9658
        "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
9659
        "nics": nic_data,
9660
        "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
9661
        "disk_template": iinfo.disk_template,
9662
        "hypervisor": iinfo.hypervisor,
9663
        }
9664
      pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
9665
                                                 pir["disks"])
9666
      instance_data[iinfo.name] = pir
9667

    
9668
    data["instances"] = instance_data
9669

    
9670
    self.in_data = data
9671

    
9672
  def _AddNewInstance(self):
9673
    """Add new instance data to allocator structure.
9674

9675
    This in combination with _AllocatorGetClusterData will create the
9676
    correct structure needed as input for the allocator.
9677

9678
    The checks for the completeness of the opcode must have already been
9679
    done.
9680

9681
    """
9682
    disk_space = _ComputeDiskSize(self.disk_template, self.disks)
9683

    
9684
    if self.disk_template in constants.DTS_NET_MIRROR:
9685
      self.required_nodes = 2
9686
    else:
9687
      self.required_nodes = 1
9688
    request = {
9689
      "name": self.name,
9690
      "disk_template": self.disk_template,
9691
      "tags": self.tags,
9692
      "os": self.os,
9693
      "vcpus": self.vcpus,
9694
      "memory": self.mem_size,
9695
      "disks": self.disks,
9696
      "disk_space_total": disk_space,
9697
      "nics": self.nics,
9698
      "required_nodes": self.required_nodes,
9699
      }
9700
    return request
9701

    
9702
  def _AddRelocateInstance(self):
9703
    """Add relocate instance data to allocator structure.
9704

9705
    This in combination with _IAllocatorGetClusterData will create the
9706
    correct structure needed as input for the allocator.
9707

9708
    The checks for the completeness of the opcode must have already been
9709
    done.
9710

9711
    """
9712
    instance = self.cfg.GetInstanceInfo(self.name)
9713
    if instance is None:
9714
      raise errors.ProgrammerError("Unknown instance '%s' passed to"
9715
                                   " IAllocator" % self.name)
9716

    
9717
    if instance.disk_template not in constants.DTS_NET_MIRROR:
9718
      raise errors.OpPrereqError("Can't relocate non-mirrored instances",
9719
                                 errors.ECODE_INVAL)
9720

    
9721
    if len(instance.secondary_nodes) != 1:
9722
      raise errors.OpPrereqError("Instance has not exactly one secondary node",
9723
                                 errors.ECODE_STATE)
9724

    
9725
    self.required_nodes = 1
9726
    disk_sizes = [{'size': disk.size} for disk in instance.disks]
9727
    disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
9728

    
9729
    request = {
9730
      "name": self.name,
9731
      "disk_space_total": disk_space,
9732
      "required_nodes": self.required_nodes,
9733
      "relocate_from": self.relocate_from,
9734
      }
9735
    return request
9736

    
9737
  def _AddEvacuateNodes(self):
9738
    """Add evacuate nodes data to allocator structure.
9739

9740
    """
9741
    request = {
9742
      "evac_nodes": self.evac_nodes
9743
      }
9744
    return request
9745

    
9746
  def _BuildInputData(self, fn):
9747
    """Build input data structures.
9748

9749
    """
9750
    self._ComputeClusterData()
9751

    
9752
    request = fn()
9753
    request["type"] = self.mode
9754
    self.in_data["request"] = request
9755

    
9756
    self.in_text = serializer.Dump(self.in_data)
9757

    
9758
  def Run(self, name, validate=True, call_fn=None):
9759
    """Run an instance allocator and return the results.
9760

9761
    """
9762
    if call_fn is None:
9763
      call_fn = self.rpc.call_iallocator_runner
9764

    
9765
    result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
9766
    result.Raise("Failure while running the iallocator script")
9767

    
9768
    self.out_text = result.payload
9769
    if validate:
9770
      self._ValidateResult()
9771

    
9772
  def _ValidateResult(self):
9773
    """Process the allocator results.
9774

9775
    This will process and if successful save the result in
9776
    self.out_data and the other parameters.
9777

9778
    """
9779
    try:
9780
      rdict = serializer.Load(self.out_text)
9781
    except Exception, err:
9782
      raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
9783

    
9784
    if not isinstance(rdict, dict):
9785
      raise errors.OpExecError("Can't parse iallocator results: not a dict")
9786

    
9787
    # TODO: remove backwards compatiblity in later versions
9788
    if "nodes" in rdict and "result" not in rdict:
9789
      rdict["result"] = rdict["nodes"]
9790
      del rdict["nodes"]
9791

    
9792
    for key in "success", "info", "result":
9793
      if key not in rdict:
9794
        raise errors.OpExecError("Can't parse iallocator results:"
9795
                                 " missing key '%s'" % key)
9796
      setattr(self, key, rdict[key])
9797

    
9798
    if not isinstance(rdict["result"], list):
9799
      raise errors.OpExecError("Can't parse iallocator results: 'result' key"
9800
                               " is not a list")
9801
    self.out_data = rdict
9802

    
9803

    
9804
class LUTestAllocator(NoHooksLU):
9805
  """Run allocator tests.
9806

9807
  This LU runs the allocator tests
9808

9809
  """
9810
  _OP_REQP = ["direction", "mode", "name"]
9811

    
9812
  def CheckPrereq(self):
9813
    """Check prerequisites.
9814

9815
    This checks the opcode parameters depending on the director and mode test.
9816

9817
    """
9818
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
9819
      for attr in ["name", "mem_size", "disks", "disk_template",
9820
                   "os", "tags", "nics", "vcpus"]:
9821
        if not hasattr(self.op, attr):
9822
          raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
9823
                                     attr, errors.ECODE_INVAL)
9824
      iname = self.cfg.ExpandInstanceName(self.op.name)
9825
      if iname is not None:
9826
        raise errors.OpPrereqError("Instance '%s' already in the cluster" %
9827
                                   iname, errors.ECODE_EXISTS)
9828
      if not isinstance(self.op.nics, list):
9829
        raise errors.OpPrereqError("Invalid parameter 'nics'",
9830
                                   errors.ECODE_INVAL)
9831
      for row in self.op.nics:
9832
        if (not isinstance(row, dict) or
9833
            "mac" not in row or
9834
            "ip" not in row or
9835
            "bridge" not in row):
9836
          raise errors.OpPrereqError("Invalid contents of the 'nics'"
9837
                                     " parameter", errors.ECODE_INVAL)
9838
      if not isinstance(self.op.disks, list):
9839
        raise errors.OpPrereqError("Invalid parameter 'disks'",
9840
                                   errors.ECODE_INVAL)
9841
      for row in self.op.disks:
9842
        if (not isinstance(row, dict) or
9843
            "size" not in row or
9844
            not isinstance(row["size"], int) or
9845
            "mode" not in row or
9846
            row["mode"] not in ['r', 'w']):
9847
          raise errors.OpPrereqError("Invalid contents of the 'disks'"
9848
                                     " parameter", errors.ECODE_INVAL)
9849
      if not hasattr(self.op, "hypervisor") or self.op.hypervisor is None:
9850
        self.op.hypervisor = self.cfg.GetHypervisorType()
9851
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
9852
      if not hasattr(self.op, "name"):
9853
        raise errors.OpPrereqError("Missing attribute 'name' on opcode input",
9854
                                   errors.ECODE_INVAL)
9855
      fname = _ExpandInstanceName(self.cfg, self.op.name)
9856
      self.op.name = fname
9857
      self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
9858
    elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
9859
      if not hasattr(self.op, "evac_nodes"):
9860
        raise errors.OpPrereqError("Missing attribute 'evac_nodes' on"
9861
                                   " opcode input", errors.ECODE_INVAL)
9862
    else:
9863
      raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
9864
                                 self.op.mode, errors.ECODE_INVAL)
9865

    
9866
    if self.op.direction == constants.IALLOCATOR_DIR_OUT:
9867
      if not hasattr(self.op, "allocator") or self.op.allocator is None:
9868
        raise errors.OpPrereqError("Missing allocator name",
9869
                                   errors.ECODE_INVAL)
9870
    elif self.op.direction != constants.IALLOCATOR_DIR_IN:
9871
      raise errors.OpPrereqError("Wrong allocator test '%s'" %
9872
                                 self.op.direction, errors.ECODE_INVAL)
9873

    
9874
  def Exec(self, feedback_fn):
9875
    """Run the allocator test.
9876

9877
    """
9878
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
9879
      ial = IAllocator(self.cfg, self.rpc,
9880
                       mode=self.op.mode,
9881
                       name=self.op.name,
9882
                       mem_size=self.op.mem_size,
9883
                       disks=self.op.disks,
9884
                       disk_template=self.op.disk_template,
9885
                       os=self.op.os,
9886
                       tags=self.op.tags,
9887
                       nics=self.op.nics,
9888
                       vcpus=self.op.vcpus,
9889
                       hypervisor=self.op.hypervisor,
9890
                       )
9891
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
9892
      ial = IAllocator(self.cfg, self.rpc,
9893
                       mode=self.op.mode,
9894
                       name=self.op.name,
9895
                       relocate_from=list(self.relocate_from),
9896
                       )
9897
    elif self.op.mode == constants.IALLOCATOR_MODE_MEVAC:
9898
      ial = IAllocator(self.cfg, self.rpc,
9899
                       mode=self.op.mode,
9900
                       evac_nodes=self.op.evac_nodes)
9901
    else:
9902
      raise errors.ProgrammerError("Uncatched mode %s in"
9903
                                   " LUTestAllocator.Exec", self.op.mode)
9904

    
9905
    if self.op.direction == constants.IALLOCATOR_DIR_IN:
9906
      result = ial.in_text
9907
    else:
9908
      ial.Run(self.op.allocator, validate=False)
9909
      result = ial.out_text
9910
    return result