Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ 22f0f71d

History | View | Annotate | Download (234.7 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=W0613,W0201
25

    
26
import os
27
import os.path
28
import sha
29
import time
30
import tempfile
31
import re
32
import platform
33
import logging
34
import copy
35
import random
36

    
37
from ganeti import ssh
38
from ganeti import utils
39
from ganeti import errors
40
from ganeti import hypervisor
41
from ganeti import locking
42
from ganeti import constants
43
from ganeti import objects
44
from ganeti import opcodes
45
from ganeti import serializer
46
from ganeti import ssconf
47

    
48

    
49
class LogicalUnit(object):
50
  """Logical Unit base class.
51

52
  Subclasses must follow these rules:
53
    - implement ExpandNames
54
    - implement CheckPrereq
55
    - implement Exec
56
    - implement BuildHooksEnv
57
    - redefine HPATH and HTYPE
58
    - optionally redefine their run requirements:
59
        REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
60

61
  Note that all commands require root permissions.
62

63
  """
64
  HPATH = None
65
  HTYPE = None
66
  _OP_REQP = []
67
  REQ_BGL = True
68

    
69
  def __init__(self, processor, op, context, rpc):
70
    """Constructor for LogicalUnit.
71

72
    This needs to be overriden in derived classes in order to check op
73
    validity.
74

75
    """
76
    self.proc = processor
77
    self.op = op
78
    self.cfg = context.cfg
79
    self.context = context
80
    self.rpc = rpc
81
    # Dicts used to declare locking needs to mcpu
82
    self.needed_locks = None
83
    self.acquired_locks = {}
84
    self.share_locks = dict(((i, 0) for i in locking.LEVELS))
85
    self.add_locks = {}
86
    self.remove_locks = {}
87
    # Used to force good behavior when calling helper functions
88
    self.recalculate_locks = {}
89
    self.__ssh = None
90
    # logging
91
    self.LogWarning = processor.LogWarning
92
    self.LogInfo = processor.LogInfo
93

    
94
    for attr_name in self._OP_REQP:
95
      attr_val = getattr(op, attr_name, None)
96
      if attr_val is None:
97
        raise errors.OpPrereqError("Required parameter '%s' missing" %
98
                                   attr_name)
99
    self.CheckArguments()
100

    
101
  def __GetSSH(self):
102
    """Returns the SshRunner object
103

104
    """
105
    if not self.__ssh:
106
      self.__ssh = ssh.SshRunner(self.cfg.GetClusterName())
107
    return self.__ssh
108

    
109
  ssh = property(fget=__GetSSH)
110

    
111
  def CheckArguments(self):
112
    """Check syntactic validity for the opcode arguments.
113

114
    This method is for doing a simple syntactic check and ensure
115
    validity of opcode parameters, without any cluster-related
116
    checks. While the same can be accomplished in ExpandNames and/or
117
    CheckPrereq, doing these separate is better because:
118

119
      - ExpandNames is left as as purely a lock-related function
120
      - CheckPrereq is run after we have aquired locks (and possible
121
        waited for them)
122

123
    The function is allowed to change the self.op attribute so that
124
    later methods can no longer worry about missing parameters.
125

126
    """
127
    pass
128

    
129
  def ExpandNames(self):
130
    """Expand names for this LU.
131

132
    This method is called before starting to execute the opcode, and it should
133
    update all the parameters of the opcode to their canonical form (e.g. a
134
    short node name must be fully expanded after this method has successfully
135
    completed). This way locking, hooks, logging, ecc. can work correctly.
136

137
    LUs which implement this method must also populate the self.needed_locks
138
    member, as a dict with lock levels as keys, and a list of needed lock names
139
    as values. Rules:
140

141
      - use an empty dict if you don't need any lock
142
      - if you don't need any lock at a particular level omit that level
143
      - don't put anything for the BGL level
144
      - if you want all locks at a level use locking.ALL_SET as a value
145

146
    If you need to share locks (rather than acquire them exclusively) at one
147
    level you can modify self.share_locks, setting a true value (usually 1) for
148
    that level. By default locks are not shared.
149

150
    Examples::
151

152
      # Acquire all nodes and one instance
153
      self.needed_locks = {
154
        locking.LEVEL_NODE: locking.ALL_SET,
155
        locking.LEVEL_INSTANCE: ['instance1.example.tld'],
156
      }
157
      # Acquire just two nodes
158
      self.needed_locks = {
159
        locking.LEVEL_NODE: ['node1.example.tld', 'node2.example.tld'],
160
      }
161
      # Acquire no locks
162
      self.needed_locks = {} # No, you can't leave it to the default value None
163

164
    """
165
    # The implementation of this method is mandatory only if the new LU is
166
    # concurrent, so that old LUs don't need to be changed all at the same
167
    # time.
168
    if self.REQ_BGL:
169
      self.needed_locks = {} # Exclusive LUs don't need locks.
170
    else:
171
      raise NotImplementedError
172

    
173
  def DeclareLocks(self, level):
174
    """Declare LU locking needs for a level
175

176
    While most LUs can just declare their locking needs at ExpandNames time,
177
    sometimes there's the need to calculate some locks after having acquired
178
    the ones before. This function is called just before acquiring locks at a
179
    particular level, but after acquiring the ones at lower levels, and permits
180
    such calculations. It can be used to modify self.needed_locks, and by
181
    default it does nothing.
182

183
    This function is only called if you have something already set in
184
    self.needed_locks for the level.
185

186
    @param level: Locking level which is going to be locked
187
    @type level: member of ganeti.locking.LEVELS
188

189
    """
190

    
191
  def CheckPrereq(self):
192
    """Check prerequisites for this LU.
193

194
    This method should check that the prerequisites for the execution
195
    of this LU are fulfilled. It can do internode communication, but
196
    it should be idempotent - no cluster or system changes are
197
    allowed.
198

199
    The method should raise errors.OpPrereqError in case something is
200
    not fulfilled. Its return value is ignored.
201

202
    This method should also update all the parameters of the opcode to
203
    their canonical form if it hasn't been done by ExpandNames before.
204

205
    """
206
    raise NotImplementedError
207

    
208
  def Exec(self, feedback_fn):
209
    """Execute the LU.
210

211
    This method should implement the actual work. It should raise
212
    errors.OpExecError for failures that are somewhat dealt with in
213
    code, or expected.
214

215
    """
216
    raise NotImplementedError
217

    
218
  def BuildHooksEnv(self):
219
    """Build hooks environment for this LU.
220

221
    This method should return a three-node tuple consisting of: a dict
222
    containing the environment that will be used for running the
223
    specific hook for this LU, a list of node names on which the hook
224
    should run before the execution, and a list of node names on which
225
    the hook should run after the execution.
226

227
    The keys of the dict must not have 'GANETI_' prefixed as this will
228
    be handled in the hooks runner. Also note additional keys will be
229
    added by the hooks runner. If the LU doesn't define any
230
    environment, an empty dict (and not None) should be returned.
231

232
    No nodes should be returned as an empty list (and not None).
233

234
    Note that if the HPATH for a LU class is None, this function will
235
    not be called.
236

237
    """
238
    raise NotImplementedError
239

    
240
  def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
241
    """Notify the LU about the results of its hooks.
242

243
    This method is called every time a hooks phase is executed, and notifies
244
    the Logical Unit about the hooks' result. The LU can then use it to alter
245
    its result based on the hooks.  By default the method does nothing and the
246
    previous result is passed back unchanged but any LU can define it if it
247
    wants to use the local cluster hook-scripts somehow.
248

249
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
250
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
251
    @param hook_results: the results of the multi-node hooks rpc call
252
    @param feedback_fn: function used send feedback back to the caller
253
    @param lu_result: the previous Exec result this LU had, or None
254
        in the PRE phase
255
    @return: the new Exec result, based on the previous result
256
        and hook results
257

258
    """
259
    return lu_result
260

    
261
  def _ExpandAndLockInstance(self):
262
    """Helper function to expand and lock an instance.
263

264
    Many LUs that work on an instance take its name in self.op.instance_name
265
    and need to expand it and then declare the expanded name for locking. This
266
    function does it, and then updates self.op.instance_name to the expanded
267
    name. It also initializes needed_locks as a dict, if this hasn't been done
268
    before.
269

270
    """
271
    if self.needed_locks is None:
272
      self.needed_locks = {}
273
    else:
274
      assert locking.LEVEL_INSTANCE not in self.needed_locks, \
275
        "_ExpandAndLockInstance called with instance-level locks set"
276
    expanded_name = self.cfg.ExpandInstanceName(self.op.instance_name)
277
    if expanded_name is None:
278
      raise errors.OpPrereqError("Instance '%s' not known" %
279
                                  self.op.instance_name)
280
    self.needed_locks[locking.LEVEL_INSTANCE] = expanded_name
281
    self.op.instance_name = expanded_name
282

    
283
  def _LockInstancesNodes(self, primary_only=False):
284
    """Helper function to declare instances' nodes for locking.
285

286
    This function should be called after locking one or more instances to lock
287
    their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
288
    with all primary or secondary nodes for instances already locked and
289
    present in self.needed_locks[locking.LEVEL_INSTANCE].
290

291
    It should be called from DeclareLocks, and for safety only works if
292
    self.recalculate_locks[locking.LEVEL_NODE] is set.
293

294
    In the future it may grow parameters to just lock some instance's nodes, or
295
    to just lock primaries or secondary nodes, if needed.
296

297
    If should be called in DeclareLocks in a way similar to::
298

299
      if level == locking.LEVEL_NODE:
300
        self._LockInstancesNodes()
301

302
    @type primary_only: boolean
303
    @param primary_only: only lock primary nodes of locked instances
304

305
    """
306
    assert locking.LEVEL_NODE in self.recalculate_locks, \
307
      "_LockInstancesNodes helper function called with no nodes to recalculate"
308

    
309
    # TODO: check if we're really been called with the instance locks held
310

    
311
    # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
312
    # future we might want to have different behaviors depending on the value
313
    # of self.recalculate_locks[locking.LEVEL_NODE]
314
    wanted_nodes = []
315
    for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
316
      instance = self.context.cfg.GetInstanceInfo(instance_name)
317
      wanted_nodes.append(instance.primary_node)
318
      if not primary_only:
319
        wanted_nodes.extend(instance.secondary_nodes)
320

    
321
    if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE:
322
      self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
323
    elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND:
324
      self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
325

    
326
    del self.recalculate_locks[locking.LEVEL_NODE]
327

    
328

    
329
class NoHooksLU(LogicalUnit):
330
  """Simple LU which runs no hooks.
331

332
  This LU is intended as a parent for other LogicalUnits which will
333
  run no hooks, in order to reduce duplicate code.
334

335
  """
336
  HPATH = None
337
  HTYPE = None
338

    
339

    
340
def _GetWantedNodes(lu, nodes):
341
  """Returns list of checked and expanded node names.
342

343
  @type lu: L{LogicalUnit}
344
  @param lu: the logical unit on whose behalf we execute
345
  @type nodes: list
346
  @param nodes: list of node names or None for all nodes
347
  @rtype: list
348
  @return: the list of nodes, sorted
349
  @raise errors.OpProgrammerError: if the nodes parameter is wrong type
350

351
  """
352
  if not isinstance(nodes, list):
353
    raise errors.OpPrereqError("Invalid argument type 'nodes'")
354

    
355
  if not nodes:
356
    raise errors.ProgrammerError("_GetWantedNodes should only be called with a"
357
      " non-empty list of nodes whose name is to be expanded.")
358

    
359
  wanted = []
360
  for name in nodes:
361
    node = lu.cfg.ExpandNodeName(name)
362
    if node is None:
363
      raise errors.OpPrereqError("No such node name '%s'" % name)
364
    wanted.append(node)
365

    
366
  return utils.NiceSort(wanted)
367

    
368

    
369
def _GetWantedInstances(lu, instances):
370
  """Returns list of checked and expanded instance names.
371

372
  @type lu: L{LogicalUnit}
373
  @param lu: the logical unit on whose behalf we execute
374
  @type instances: list
375
  @param instances: list of instance names or None for all instances
376
  @rtype: list
377
  @return: the list of instances, sorted
378
  @raise errors.OpPrereqError: if the instances parameter is wrong type
379
  @raise errors.OpPrereqError: if any of the passed instances is not found
380

381
  """
382
  if not isinstance(instances, list):
383
    raise errors.OpPrereqError("Invalid argument type 'instances'")
384

    
385
  if instances:
386
    wanted = []
387

    
388
    for name in instances:
389
      instance = lu.cfg.ExpandInstanceName(name)
390
      if instance is None:
391
        raise errors.OpPrereqError("No such instance name '%s'" % name)
392
      wanted.append(instance)
393

    
394
  else:
395
    wanted = utils.NiceSort(lu.cfg.GetInstanceList())
396
  return wanted
397

    
398

    
399
def _CheckOutputFields(static, dynamic, selected):
400
  """Checks whether all selected fields are valid.
401

402
  @type static: L{utils.FieldSet}
403
  @param static: static fields set
404
  @type dynamic: L{utils.FieldSet}
405
  @param dynamic: dynamic fields set
406

407
  """
408
  f = utils.FieldSet()
409
  f.Extend(static)
410
  f.Extend(dynamic)
411

    
412
  delta = f.NonMatching(selected)
413
  if delta:
414
    raise errors.OpPrereqError("Unknown output fields selected: %s"
415
                               % ",".join(delta))
416

    
417

    
418
def _CheckBooleanOpField(op, name):
419
  """Validates boolean opcode parameters.
420

421
  This will ensure that an opcode parameter is either a boolean value,
422
  or None (but that it always exists).
423

424
  """
425
  val = getattr(op, name, None)
426
  if not (val is None or isinstance(val, bool)):
427
    raise errors.OpPrereqError("Invalid boolean parameter '%s' (%s)" %
428
                               (name, str(val)))
429
  setattr(op, name, val)
430

    
431

    
432
def _CheckNodeOnline(lu, node):
433
  """Ensure that a given node is online.
434

435
  @param lu: the LU on behalf of which we make the check
436
  @param node: the node to check
437
  @raise errors.OpPrereqError: if the nodes is offline
438

439
  """
440
  if lu.cfg.GetNodeInfo(node).offline:
441
    raise errors.OpPrereqError("Can't use offline node %s" % node)
442

    
443

    
444
def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
445
                          memory, vcpus, nics):
446
  """Builds instance related env variables for hooks
447

448
  This builds the hook environment from individual variables.
449

450
  @type name: string
451
  @param name: the name of the instance
452
  @type primary_node: string
453
  @param primary_node: the name of the instance's primary node
454
  @type secondary_nodes: list
455
  @param secondary_nodes: list of secondary nodes as strings
456
  @type os_type: string
457
  @param os_type: the name of the instance's OS
458
  @type status: boolean
459
  @param status: the should_run status of the instance
460
  @type memory: string
461
  @param memory: the memory size of the instance
462
  @type vcpus: string
463
  @param vcpus: the count of VCPUs the instance has
464
  @type nics: list
465
  @param nics: list of tuples (ip, bridge, mac) representing
466
      the NICs the instance  has
467
  @rtype: dict
468
  @return: the hook environment for this instance
469

470
  """
471
  if status:
472
    str_status = "up"
473
  else:
474
    str_status = "down"
475
  env = {
476
    "OP_TARGET": name,
477
    "INSTANCE_NAME": name,
478
    "INSTANCE_PRIMARY": primary_node,
479
    "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
480
    "INSTANCE_OS_TYPE": os_type,
481
    "INSTANCE_STATUS": str_status,
482
    "INSTANCE_MEMORY": memory,
483
    "INSTANCE_VCPUS": vcpus,
484
  }
485

    
486
  if nics:
487
    nic_count = len(nics)
488
    for idx, (ip, bridge, mac) in enumerate(nics):
489
      if ip is None:
490
        ip = ""
491
      env["INSTANCE_NIC%d_IP" % idx] = ip
492
      env["INSTANCE_NIC%d_BRIDGE" % idx] = bridge
493
      env["INSTANCE_NIC%d_HWADDR" % idx] = mac
494
  else:
495
    nic_count = 0
496

    
497
  env["INSTANCE_NIC_COUNT"] = nic_count
498

    
499
  return env
500

    
501

    
502
def _BuildInstanceHookEnvByObject(lu, instance, override=None):
503
  """Builds instance related env variables for hooks from an object.
504

505
  @type lu: L{LogicalUnit}
506
  @param lu: the logical unit on whose behalf we execute
507
  @type instance: L{objects.Instance}
508
  @param instance: the instance for which we should build the
509
      environment
510
  @type override: dict
511
  @param override: dictionary with key/values that will override
512
      our values
513
  @rtype: dict
514
  @return: the hook environment dictionary
515

516
  """
517
  bep = lu.cfg.GetClusterInfo().FillBE(instance)
518
  args = {
519
    'name': instance.name,
520
    'primary_node': instance.primary_node,
521
    'secondary_nodes': instance.secondary_nodes,
522
    'os_type': instance.os,
523
    'status': instance.admin_up,
524
    'memory': bep[constants.BE_MEMORY],
525
    'vcpus': bep[constants.BE_VCPUS],
526
    'nics': [(nic.ip, nic.bridge, nic.mac) for nic in instance.nics],
527
  }
528
  if override:
529
    args.update(override)
530
  return _BuildInstanceHookEnv(**args)
531

    
532

    
533
def _AdjustCandidatePool(lu):
534
  """Adjust the candidate pool after node operations.
535

536
  """
537
  mod_list = lu.cfg.MaintainCandidatePool()
538
  if mod_list:
539
    lu.LogInfo("Promoted nodes to master candidate role: %s",
540
               ", ".join(node.name for node in mod_list))
541
    for name in mod_list:
542
      lu.context.ReaddNode(name)
543
  mc_now, mc_max = lu.cfg.GetMasterCandidateStats()
544
  if mc_now > mc_max:
545
    lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
546
               (mc_now, mc_max))
547

    
548

    
549
def _CheckInstanceBridgesExist(lu, instance):
550
  """Check that the brigdes needed by an instance exist.
551

552
  """
553
  # check bridges existance
554
  brlist = [nic.bridge for nic in instance.nics]
555
  result = lu.rpc.call_bridges_exist(instance.primary_node, brlist)
556
  result.Raise()
557
  if not result.data:
558
    raise errors.OpPrereqError("One or more target bridges %s does not"
559
                               " exist on destination node '%s'" %
560
                               (brlist, instance.primary_node))
561

    
562

    
563
class LUDestroyCluster(NoHooksLU):
564
  """Logical unit for destroying the cluster.
565

566
  """
567
  _OP_REQP = []
568

    
569
  def CheckPrereq(self):
570
    """Check prerequisites.
571

572
    This checks whether the cluster is empty.
573

574
    Any errors are signalled by raising errors.OpPrereqError.
575

576
    """
577
    master = self.cfg.GetMasterNode()
578

    
579
    nodelist = self.cfg.GetNodeList()
580
    if len(nodelist) != 1 or nodelist[0] != master:
581
      raise errors.OpPrereqError("There are still %d node(s) in"
582
                                 " this cluster." % (len(nodelist) - 1))
583
    instancelist = self.cfg.GetInstanceList()
584
    if instancelist:
585
      raise errors.OpPrereqError("There are still %d instance(s) in"
586
                                 " this cluster." % len(instancelist))
587

    
588
  def Exec(self, feedback_fn):
589
    """Destroys the cluster.
590

591
    """
592
    master = self.cfg.GetMasterNode()
593
    result = self.rpc.call_node_stop_master(master, False)
594
    result.Raise()
595
    if not result.data:
596
      raise errors.OpExecError("Could not disable the master role")
597
    priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
598
    utils.CreateBackup(priv_key)
599
    utils.CreateBackup(pub_key)
600
    return master
601

    
602

    
603
class LUVerifyCluster(LogicalUnit):
604
  """Verifies the cluster status.
605

606
  """
607
  HPATH = "cluster-verify"
608
  HTYPE = constants.HTYPE_CLUSTER
609
  _OP_REQP = ["skip_checks"]
610
  REQ_BGL = False
611

    
612
  def ExpandNames(self):
613
    self.needed_locks = {
614
      locking.LEVEL_NODE: locking.ALL_SET,
615
      locking.LEVEL_INSTANCE: locking.ALL_SET,
616
    }
617
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
618

    
619
  def _VerifyNode(self, nodeinfo, file_list, local_cksum,
620
                  node_result, feedback_fn, master_files,
621
                  drbd_map):
622
    """Run multiple tests against a node.
623

624
    Test list:
625

626
      - compares ganeti version
627
      - checks vg existance and size > 20G
628
      - checks config file checksum
629
      - checks ssh to other nodes
630

631
    @type nodeinfo: L{objects.Node}
632
    @param nodeinfo: the node to check
633
    @param file_list: required list of files
634
    @param local_cksum: dictionary of local files and their checksums
635
    @param node_result: the results from the node
636
    @param feedback_fn: function used to accumulate results
637
    @param master_files: list of files that only masters should have
638
    @param drbd_map: the useddrbd minors for this node, in
639
        form of minor: (instance, must_exist) which correspond to instances
640
        and their running status
641

642
    """
643
    node = nodeinfo.name
644

    
645
    # main result, node_result should be a non-empty dict
646
    if not node_result or not isinstance(node_result, dict):
647
      feedback_fn("  - ERROR: unable to verify node %s." % (node,))
648
      return True
649

    
650
    # compares ganeti version
651
    local_version = constants.PROTOCOL_VERSION
652
    remote_version = node_result.get('version', None)
653
    if not (remote_version and isinstance(remote_version, (list, tuple)) and
654
            len(remote_version) == 2):
655
      feedback_fn("  - ERROR: connection to %s failed" % (node))
656
      return True
657

    
658
    if local_version != remote_version[0]:
659
      feedback_fn("  - ERROR: incompatible protocol versions: master %s,"
660
                  " node %s %s" % (local_version, node, remote_version[0]))
661
      return True
662

    
663
    # node seems compatible, we can actually try to look into its results
664

    
665
    bad = False
666

    
667
    # full package version
668
    if constants.RELEASE_VERSION != remote_version[1]:
669
      feedback_fn("  - WARNING: software version mismatch: master %s,"
670
                  " node %s %s" %
671
                  (constants.RELEASE_VERSION, node, remote_version[1]))
672

    
673
    # checks vg existence and size > 20G
674

    
675
    vglist = node_result.get(constants.NV_VGLIST, None)
676
    if not vglist:
677
      feedback_fn("  - ERROR: unable to check volume groups on node %s." %
678
                      (node,))
679
      bad = True
680
    else:
681
      vgstatus = utils.CheckVolumeGroupSize(vglist, self.cfg.GetVGName(),
682
                                            constants.MIN_VG_SIZE)
683
      if vgstatus:
684
        feedback_fn("  - ERROR: %s on node %s" % (vgstatus, node))
685
        bad = True
686

    
687
    # checks config file checksum
688

    
689
    remote_cksum = node_result.get(constants.NV_FILELIST, None)
690
    if not isinstance(remote_cksum, dict):
691
      bad = True
692
      feedback_fn("  - ERROR: node hasn't returned file checksum data")
693
    else:
694
      for file_name in file_list:
695
        node_is_mc = nodeinfo.master_candidate
696
        must_have_file = file_name not in master_files
697
        if file_name not in remote_cksum:
698
          if node_is_mc or must_have_file:
699
            bad = True
700
            feedback_fn("  - ERROR: file '%s' missing" % file_name)
701
        elif remote_cksum[file_name] != local_cksum[file_name]:
702
          if node_is_mc or must_have_file:
703
            bad = True
704
            feedback_fn("  - ERROR: file '%s' has wrong checksum" % file_name)
705
          else:
706
            # not candidate and this is not a must-have file
707
            bad = True
708
            feedback_fn("  - ERROR: non master-candidate has old/wrong file"
709
                        " '%s'" % file_name)
710
        else:
711
          # all good, except non-master/non-must have combination
712
          if not node_is_mc and not must_have_file:
713
            feedback_fn("  - ERROR: file '%s' should not exist on non master"
714
                        " candidates" % file_name)
715

    
716
    # checks ssh to any
717

    
718
    if constants.NV_NODELIST not in node_result:
719
      bad = True
720
      feedback_fn("  - ERROR: node hasn't returned node ssh connectivity data")
721
    else:
722
      if node_result[constants.NV_NODELIST]:
723
        bad = True
724
        for node in node_result[constants.NV_NODELIST]:
725
          feedback_fn("  - ERROR: ssh communication with node '%s': %s" %
726
                          (node, node_result[constants.NV_NODELIST][node]))
727

    
728
    if constants.NV_NODENETTEST not in node_result:
729
      bad = True
730
      feedback_fn("  - ERROR: node hasn't returned node tcp connectivity data")
731
    else:
732
      if node_result[constants.NV_NODENETTEST]:
733
        bad = True
734
        nlist = utils.NiceSort(node_result[constants.NV_NODENETTEST].keys())
735
        for node in nlist:
736
          feedback_fn("  - ERROR: tcp communication with node '%s': %s" %
737
                          (node, node_result[constants.NV_NODENETTEST][node]))
738

    
739
    hyp_result = node_result.get(constants.NV_HYPERVISOR, None)
740
    if isinstance(hyp_result, dict):
741
      for hv_name, hv_result in hyp_result.iteritems():
742
        if hv_result is not None:
743
          feedback_fn("  - ERROR: hypervisor %s verify failure: '%s'" %
744
                      (hv_name, hv_result))
745

    
746
    # check used drbd list
747
    used_minors = node_result.get(constants.NV_DRBDLIST, [])
748
    for minor, (iname, must_exist) in drbd_map.items():
749
      if minor not in used_minors and must_exist:
750
        feedback_fn("  - ERROR: drbd minor %d of instance %s is not active" %
751
                    (minor, iname))
752
        bad = True
753
    for minor in used_minors:
754
      if minor not in drbd_map:
755
        feedback_fn("  - ERROR: unallocated drbd minor %d is in use" % minor)
756
        bad = True
757

    
758
    return bad
759

    
760
  def _VerifyInstance(self, instance, instanceconfig, node_vol_is,
761
                      node_instance, feedback_fn, n_offline):
762
    """Verify an instance.
763

764
    This function checks to see if the required block devices are
765
    available on the instance's node.
766

767
    """
768
    bad = False
769

    
770
    node_current = instanceconfig.primary_node
771

    
772
    node_vol_should = {}
773
    instanceconfig.MapLVsByNode(node_vol_should)
774

    
775
    for node in node_vol_should:
776
      if node in n_offline:
777
        # ignore missing volumes on offline nodes
778
        continue
779
      for volume in node_vol_should[node]:
780
        if node not in node_vol_is or volume not in node_vol_is[node]:
781
          feedback_fn("  - ERROR: volume %s missing on node %s" %
782
                          (volume, node))
783
          bad = True
784

    
785
    if instanceconfig.admin_up:
786
      if ((node_current not in node_instance or
787
          not instance in node_instance[node_current]) and
788
          node_current not in n_offline):
789
        feedback_fn("  - ERROR: instance %s not running on node %s" %
790
                        (instance, node_current))
791
        bad = True
792

    
793
    for node in node_instance:
794
      if (not node == node_current):
795
        if instance in node_instance[node]:
796
          feedback_fn("  - ERROR: instance %s should not run on node %s" %
797
                          (instance, node))
798
          bad = True
799

    
800
    return bad
801

    
802
  def _VerifyOrphanVolumes(self, node_vol_should, node_vol_is, feedback_fn):
803
    """Verify if there are any unknown volumes in the cluster.
804

805
    The .os, .swap and backup volumes are ignored. All other volumes are
806
    reported as unknown.
807

808
    """
809
    bad = False
810

    
811
    for node in node_vol_is:
812
      for volume in node_vol_is[node]:
813
        if node not in node_vol_should or volume not in node_vol_should[node]:
814
          feedback_fn("  - ERROR: volume %s on node %s should not exist" %
815
                      (volume, node))
816
          bad = True
817
    return bad
818

    
819
  def _VerifyOrphanInstances(self, instancelist, node_instance, feedback_fn):
820
    """Verify the list of running instances.
821

822
    This checks what instances are running but unknown to the cluster.
823

824
    """
825
    bad = False
826
    for node in node_instance:
827
      for runninginstance in node_instance[node]:
828
        if runninginstance not in instancelist:
829
          feedback_fn("  - ERROR: instance %s on node %s should not exist" %
830
                          (runninginstance, node))
831
          bad = True
832
    return bad
833

    
834
  def _VerifyNPlusOneMemory(self, node_info, instance_cfg, feedback_fn):
835
    """Verify N+1 Memory Resilience.
836

837
    Check that if one single node dies we can still start all the instances it
838
    was primary for.
839

840
    """
841
    bad = False
842

    
843
    for node, nodeinfo in node_info.iteritems():
844
      # This code checks that every node which is now listed as secondary has
845
      # enough memory to host all instances it is supposed to should a single
846
      # other node in the cluster fail.
847
      # FIXME: not ready for failover to an arbitrary node
848
      # FIXME: does not support file-backed instances
849
      # WARNING: we currently take into account down instances as well as up
850
      # ones, considering that even if they're down someone might want to start
851
      # them even in the event of a node failure.
852
      for prinode, instances in nodeinfo['sinst-by-pnode'].iteritems():
853
        needed_mem = 0
854
        for instance in instances:
855
          bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
856
          if bep[constants.BE_AUTO_BALANCE]:
857
            needed_mem += bep[constants.BE_MEMORY]
858
        if nodeinfo['mfree'] < needed_mem:
859
          feedback_fn("  - ERROR: not enough memory on node %s to accomodate"
860
                      " failovers should node %s fail" % (node, prinode))
861
          bad = True
862
    return bad
863

    
864
  def CheckPrereq(self):
865
    """Check prerequisites.
866

867
    Transform the list of checks we're going to skip into a set and check that
868
    all its members are valid.
869

870
    """
871
    self.skip_set = frozenset(self.op.skip_checks)
872
    if not constants.VERIFY_OPTIONAL_CHECKS.issuperset(self.skip_set):
873
      raise errors.OpPrereqError("Invalid checks to be skipped specified")
874

    
875
  def BuildHooksEnv(self):
876
    """Build hooks env.
877

878
    Cluster-Verify hooks just rone in the post phase and their failure makes
879
    the output be logged in the verify output and the verification to fail.
880

881
    """
882
    all_nodes = self.cfg.GetNodeList()
883
    # TODO: populate the environment with useful information for verify hooks
884
    env = {}
885
    return env, [], all_nodes
886

    
887
  def Exec(self, feedback_fn):
888
    """Verify integrity of cluster, performing various test on nodes.
889

890
    """
891
    bad = False
892
    feedback_fn("* Verifying global settings")
893
    for msg in self.cfg.VerifyConfig():
894
      feedback_fn("  - ERROR: %s" % msg)
895

    
896
    vg_name = self.cfg.GetVGName()
897
    hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
898
    nodelist = utils.NiceSort(self.cfg.GetNodeList())
899
    nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
900
    instancelist = utils.NiceSort(self.cfg.GetInstanceList())
901
    instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
902
                        for iname in instancelist)
903
    i_non_redundant = [] # Non redundant instances
904
    i_non_a_balanced = [] # Non auto-balanced instances
905
    n_offline = [] # List of offline nodes
906
    n_drained = [] # List of nodes being drained
907
    node_volume = {}
908
    node_instance = {}
909
    node_info = {}
910
    instance_cfg = {}
911

    
912
    # FIXME: verify OS list
913
    # do local checksums
914
    master_files = [constants.CLUSTER_CONF_FILE]
915

    
916
    file_names = ssconf.SimpleStore().GetFileList()
917
    file_names.append(constants.SSL_CERT_FILE)
918
    file_names.append(constants.RAPI_CERT_FILE)
919
    file_names.extend(master_files)
920

    
921
    local_checksums = utils.FingerprintFiles(file_names)
922

    
923
    feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
924
    node_verify_param = {
925
      constants.NV_FILELIST: file_names,
926
      constants.NV_NODELIST: [node.name for node in nodeinfo
927
                              if not node.offline],
928
      constants.NV_HYPERVISOR: hypervisors,
929
      constants.NV_NODENETTEST: [(node.name, node.primary_ip,
930
                                  node.secondary_ip) for node in nodeinfo
931
                                 if not node.offline],
932
      constants.NV_LVLIST: vg_name,
933
      constants.NV_INSTANCELIST: hypervisors,
934
      constants.NV_VGLIST: None,
935
      constants.NV_VERSION: None,
936
      constants.NV_HVINFO: self.cfg.GetHypervisorType(),
937
      constants.NV_DRBDLIST: None,
938
      }
939
    all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
940
                                           self.cfg.GetClusterName())
941

    
942
    cluster = self.cfg.GetClusterInfo()
943
    master_node = self.cfg.GetMasterNode()
944
    all_drbd_map = self.cfg.ComputeDRBDMap()
945

    
946
    for node_i in nodeinfo:
947
      node = node_i.name
948
      nresult = all_nvinfo[node].data
949

    
950
      if node_i.offline:
951
        feedback_fn("* Skipping offline node %s" % (node,))
952
        n_offline.append(node)
953
        continue
954

    
955
      if node == master_node:
956
        ntype = "master"
957
      elif node_i.master_candidate:
958
        ntype = "master candidate"
959
      elif node_i.drained:
960
        ntype = "drained"
961
        n_drained.append(node)
962
      else:
963
        ntype = "regular"
964
      feedback_fn("* Verifying node %s (%s)" % (node, ntype))
965

    
966
      if all_nvinfo[node].failed or not isinstance(nresult, dict):
967
        feedback_fn("  - ERROR: connection to %s failed" % (node,))
968
        bad = True
969
        continue
970

    
971
      node_drbd = {}
972
      for minor, instance in all_drbd_map[node].items():
973
        instance = instanceinfo[instance]
974
        node_drbd[minor] = (instance.name, instance.admin_up)
975
      result = self._VerifyNode(node_i, file_names, local_checksums,
976
                                nresult, feedback_fn, master_files,
977
                                node_drbd)
978
      bad = bad or result
979

    
980
      lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
981
      if isinstance(lvdata, basestring):
982
        feedback_fn("  - ERROR: LVM problem on node %s: %s" %
983
                    (node, utils.SafeEncode(lvdata)))
984
        bad = True
985
        node_volume[node] = {}
986
      elif not isinstance(lvdata, dict):
987
        feedback_fn("  - ERROR: connection to %s failed (lvlist)" % (node,))
988
        bad = True
989
        continue
990
      else:
991
        node_volume[node] = lvdata
992

    
993
      # node_instance
994
      idata = nresult.get(constants.NV_INSTANCELIST, None)
995
      if not isinstance(idata, list):
996
        feedback_fn("  - ERROR: connection to %s failed (instancelist)" %
997
                    (node,))
998
        bad = True
999
        continue
1000

    
1001
      node_instance[node] = idata
1002

    
1003
      # node_info
1004
      nodeinfo = nresult.get(constants.NV_HVINFO, None)
1005
      if not isinstance(nodeinfo, dict):
1006
        feedback_fn("  - ERROR: connection to %s failed (hvinfo)" % (node,))
1007
        bad = True
1008
        continue
1009

    
1010
      try:
1011
        node_info[node] = {
1012
          "mfree": int(nodeinfo['memory_free']),
1013
          "dfree": int(nresult[constants.NV_VGLIST][vg_name]),
1014
          "pinst": [],
1015
          "sinst": [],
1016
          # dictionary holding all instances this node is secondary for,
1017
          # grouped by their primary node. Each key is a cluster node, and each
1018
          # value is a list of instances which have the key as primary and the
1019
          # current node as secondary.  this is handy to calculate N+1 memory
1020
          # availability if you can only failover from a primary to its
1021
          # secondary.
1022
          "sinst-by-pnode": {},
1023
        }
1024
      except ValueError:
1025
        feedback_fn("  - ERROR: invalid value returned from node %s" % (node,))
1026
        bad = True
1027
        continue
1028

    
1029
    node_vol_should = {}
1030

    
1031
    for instance in instancelist:
1032
      feedback_fn("* Verifying instance %s" % instance)
1033
      inst_config = instanceinfo[instance]
1034
      result =  self._VerifyInstance(instance, inst_config, node_volume,
1035
                                     node_instance, feedback_fn, n_offline)
1036
      bad = bad or result
1037
      inst_nodes_offline = []
1038

    
1039
      inst_config.MapLVsByNode(node_vol_should)
1040

    
1041
      instance_cfg[instance] = inst_config
1042

    
1043
      pnode = inst_config.primary_node
1044
      if pnode in node_info:
1045
        node_info[pnode]['pinst'].append(instance)
1046
      elif pnode not in n_offline:
1047
        feedback_fn("  - ERROR: instance %s, connection to primary node"
1048
                    " %s failed" % (instance, pnode))
1049
        bad = True
1050

    
1051
      if pnode in n_offline:
1052
        inst_nodes_offline.append(pnode)
1053

    
1054
      # If the instance is non-redundant we cannot survive losing its primary
1055
      # node, so we are not N+1 compliant. On the other hand we have no disk
1056
      # templates with more than one secondary so that situation is not well
1057
      # supported either.
1058
      # FIXME: does not support file-backed instances
1059
      if len(inst_config.secondary_nodes) == 0:
1060
        i_non_redundant.append(instance)
1061
      elif len(inst_config.secondary_nodes) > 1:
1062
        feedback_fn("  - WARNING: multiple secondaries for instance %s"
1063
                    % instance)
1064

    
1065
      if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
1066
        i_non_a_balanced.append(instance)
1067

    
1068
      for snode in inst_config.secondary_nodes:
1069
        if snode in node_info:
1070
          node_info[snode]['sinst'].append(instance)
1071
          if pnode not in node_info[snode]['sinst-by-pnode']:
1072
            node_info[snode]['sinst-by-pnode'][pnode] = []
1073
          node_info[snode]['sinst-by-pnode'][pnode].append(instance)
1074
        elif snode not in n_offline:
1075
          feedback_fn("  - ERROR: instance %s, connection to secondary node"
1076
                      " %s failed" % (instance, snode))
1077
          bad = True
1078
        if snode in n_offline:
1079
          inst_nodes_offline.append(snode)
1080

    
1081
      if inst_nodes_offline:
1082
        # warn that the instance lives on offline nodes, and set bad=True
1083
        feedback_fn("  - ERROR: instance lives on offline node(s) %s" %
1084
                    ", ".join(inst_nodes_offline))
1085
        bad = True
1086

    
1087
    feedback_fn("* Verifying orphan volumes")
1088
    result = self._VerifyOrphanVolumes(node_vol_should, node_volume,
1089
                                       feedback_fn)
1090
    bad = bad or result
1091

    
1092
    feedback_fn("* Verifying remaining instances")
1093
    result = self._VerifyOrphanInstances(instancelist, node_instance,
1094
                                         feedback_fn)
1095
    bad = bad or result
1096

    
1097
    if constants.VERIFY_NPLUSONE_MEM not in self.skip_set:
1098
      feedback_fn("* Verifying N+1 Memory redundancy")
1099
      result = self._VerifyNPlusOneMemory(node_info, instance_cfg, feedback_fn)
1100
      bad = bad or result
1101

    
1102
    feedback_fn("* Other Notes")
1103
    if i_non_redundant:
1104
      feedback_fn("  - NOTICE: %d non-redundant instance(s) found."
1105
                  % len(i_non_redundant))
1106

    
1107
    if i_non_a_balanced:
1108
      feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
1109
                  % len(i_non_a_balanced))
1110

    
1111
    if n_offline:
1112
      feedback_fn("  - NOTICE: %d offline node(s) found." % len(n_offline))
1113

    
1114
    if n_drained:
1115
      feedback_fn("  - NOTICE: %d drained node(s) found." % len(n_drained))
1116

    
1117
    return not bad
1118

    
1119
  def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
1120
    """Analize the post-hooks' result
1121

1122
    This method analyses the hook result, handles it, and sends some
1123
    nicely-formatted feedback back to the user.
1124

1125
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
1126
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
1127
    @param hooks_results: the results of the multi-node hooks rpc call
1128
    @param feedback_fn: function used send feedback back to the caller
1129
    @param lu_result: previous Exec result
1130
    @return: the new Exec result, based on the previous result
1131
        and hook results
1132

1133
    """
1134
    # We only really run POST phase hooks, and are only interested in
1135
    # their results
1136
    if phase == constants.HOOKS_PHASE_POST:
1137
      # Used to change hooks' output to proper indentation
1138
      indent_re = re.compile('^', re.M)
1139
      feedback_fn("* Hooks Results")
1140
      if not hooks_results:
1141
        feedback_fn("  - ERROR: general communication failure")
1142
        lu_result = 1
1143
      else:
1144
        for node_name in hooks_results:
1145
          show_node_header = True
1146
          res = hooks_results[node_name]
1147
          if res.failed or res.data is False or not isinstance(res.data, list):
1148
            if res.offline:
1149
              # no need to warn or set fail return value
1150
              continue
1151
            feedback_fn("    Communication failure in hooks execution")
1152
            lu_result = 1
1153
            continue
1154
          for script, hkr, output in res.data:
1155
            if hkr == constants.HKR_FAIL:
1156
              # The node header is only shown once, if there are
1157
              # failing hooks on that node
1158
              if show_node_header:
1159
                feedback_fn("  Node %s:" % node_name)
1160
                show_node_header = False
1161
              feedback_fn("    ERROR: Script %s failed, output:" % script)
1162
              output = indent_re.sub('      ', output)
1163
              feedback_fn("%s" % output)
1164
              lu_result = 1
1165

    
1166
      return lu_result
1167

    
1168

    
1169
class LUVerifyDisks(NoHooksLU):
1170
  """Verifies the cluster disks status.
1171

1172
  """
1173
  _OP_REQP = []
1174
  REQ_BGL = False
1175

    
1176
  def ExpandNames(self):
1177
    self.needed_locks = {
1178
      locking.LEVEL_NODE: locking.ALL_SET,
1179
      locking.LEVEL_INSTANCE: locking.ALL_SET,
1180
    }
1181
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
1182

    
1183
  def CheckPrereq(self):
1184
    """Check prerequisites.
1185

1186
    This has no prerequisites.
1187

1188
    """
1189
    pass
1190

    
1191
  def Exec(self, feedback_fn):
1192
    """Verify integrity of cluster disks.
1193

1194
    """
1195
    result = res_nodes, res_nlvm, res_instances, res_missing = [], {}, [], {}
1196

    
1197
    vg_name = self.cfg.GetVGName()
1198
    nodes = utils.NiceSort(self.cfg.GetNodeList())
1199
    instances = [self.cfg.GetInstanceInfo(name)
1200
                 for name in self.cfg.GetInstanceList()]
1201

    
1202
    nv_dict = {}
1203
    for inst in instances:
1204
      inst_lvs = {}
1205
      if (not inst.admin_up or
1206
          inst.disk_template not in constants.DTS_NET_MIRROR):
1207
        continue
1208
      inst.MapLVsByNode(inst_lvs)
1209
      # transform { iname: {node: [vol,],},} to {(node, vol): iname}
1210
      for node, vol_list in inst_lvs.iteritems():
1211
        for vol in vol_list:
1212
          nv_dict[(node, vol)] = inst
1213

    
1214
    if not nv_dict:
1215
      return result
1216

    
1217
    node_lvs = self.rpc.call_volume_list(nodes, vg_name)
1218

    
1219
    to_act = set()
1220
    for node in nodes:
1221
      # node_volume
1222
      lvs = node_lvs[node]
1223
      if lvs.failed:
1224
        if not lvs.offline:
1225
          self.LogWarning("Connection to node %s failed: %s" %
1226
                          (node, lvs.data))
1227
        continue
1228
      lvs = lvs.data
1229
      if isinstance(lvs, basestring):
1230
        logging.warning("Error enumerating LVs on node %s: %s", node, lvs)
1231
        res_nlvm[node] = lvs
1232
      elif not isinstance(lvs, dict):
1233
        logging.warning("Connection to node %s failed or invalid data"
1234
                        " returned", node)
1235
        res_nodes.append(node)
1236
        continue
1237

    
1238
      for lv_name, (_, lv_inactive, lv_online) in lvs.iteritems():
1239
        inst = nv_dict.pop((node, lv_name), None)
1240
        if (not lv_online and inst is not None
1241
            and inst.name not in res_instances):
1242
          res_instances.append(inst.name)
1243

    
1244
    # any leftover items in nv_dict are missing LVs, let's arrange the
1245
    # data better
1246
    for key, inst in nv_dict.iteritems():
1247
      if inst.name not in res_missing:
1248
        res_missing[inst.name] = []
1249
      res_missing[inst.name].append(key)
1250

    
1251
    return result
1252

    
1253

    
1254
class LURenameCluster(LogicalUnit):
1255
  """Rename the cluster.
1256

1257
  """
1258
  HPATH = "cluster-rename"
1259
  HTYPE = constants.HTYPE_CLUSTER
1260
  _OP_REQP = ["name"]
1261

    
1262
  def BuildHooksEnv(self):
1263
    """Build hooks env.
1264

1265
    """
1266
    env = {
1267
      "OP_TARGET": self.cfg.GetClusterName(),
1268
      "NEW_NAME": self.op.name,
1269
      }
1270
    mn = self.cfg.GetMasterNode()
1271
    return env, [mn], [mn]
1272

    
1273
  def CheckPrereq(self):
1274
    """Verify that the passed name is a valid one.
1275

1276
    """
1277
    hostname = utils.HostInfo(self.op.name)
1278

    
1279
    new_name = hostname.name
1280
    self.ip = new_ip = hostname.ip
1281
    old_name = self.cfg.GetClusterName()
1282
    old_ip = self.cfg.GetMasterIP()
1283
    if new_name == old_name and new_ip == old_ip:
1284
      raise errors.OpPrereqError("Neither the name nor the IP address of the"
1285
                                 " cluster has changed")
1286
    if new_ip != old_ip:
1287
      if utils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
1288
        raise errors.OpPrereqError("The given cluster IP address (%s) is"
1289
                                   " reachable on the network. Aborting." %
1290
                                   new_ip)
1291

    
1292
    self.op.name = new_name
1293

    
1294
  def Exec(self, feedback_fn):
1295
    """Rename the cluster.
1296

1297
    """
1298
    clustername = self.op.name
1299
    ip = self.ip
1300

    
1301
    # shutdown the master IP
1302
    master = self.cfg.GetMasterNode()
1303
    result = self.rpc.call_node_stop_master(master, False)
1304
    if result.failed or not result.data:
1305
      raise errors.OpExecError("Could not disable the master role")
1306

    
1307
    try:
1308
      cluster = self.cfg.GetClusterInfo()
1309
      cluster.cluster_name = clustername
1310
      cluster.master_ip = ip
1311
      self.cfg.Update(cluster)
1312

    
1313
      # update the known hosts file
1314
      ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
1315
      node_list = self.cfg.GetNodeList()
1316
      try:
1317
        node_list.remove(master)
1318
      except ValueError:
1319
        pass
1320
      result = self.rpc.call_upload_file(node_list,
1321
                                         constants.SSH_KNOWN_HOSTS_FILE)
1322
      for to_node, to_result in result.iteritems():
1323
        if to_result.failed or not to_result.data:
1324
          logging.error("Copy of file %s to node %s failed",
1325
                        constants.SSH_KNOWN_HOSTS_FILE, to_node)
1326

    
1327
    finally:
1328
      result = self.rpc.call_node_start_master(master, False)
1329
      if result.failed or not result.data:
1330
        self.LogWarning("Could not re-enable the master role on"
1331
                        " the master, please restart manually.")
1332

    
1333

    
1334
def _RecursiveCheckIfLVMBased(disk):
1335
  """Check if the given disk or its children are lvm-based.
1336

1337
  @type disk: L{objects.Disk}
1338
  @param disk: the disk to check
1339
  @rtype: booleean
1340
  @return: boolean indicating whether a LD_LV dev_type was found or not
1341

1342
  """
1343
  if disk.children:
1344
    for chdisk in disk.children:
1345
      if _RecursiveCheckIfLVMBased(chdisk):
1346
        return True
1347
  return disk.dev_type == constants.LD_LV
1348

    
1349

    
1350
class LUSetClusterParams(LogicalUnit):
1351
  """Change the parameters of the cluster.
1352

1353
  """
1354
  HPATH = "cluster-modify"
1355
  HTYPE = constants.HTYPE_CLUSTER
1356
  _OP_REQP = []
1357
  REQ_BGL = False
1358

    
1359
  def CheckParameters(self):
1360
    """Check parameters
1361

1362
    """
1363
    if not hasattr(self.op, "candidate_pool_size"):
1364
      self.op.candidate_pool_size = None
1365
    if self.op.candidate_pool_size is not None:
1366
      try:
1367
        self.op.candidate_pool_size = int(self.op.candidate_pool_size)
1368
      except ValueError, err:
1369
        raise errors.OpPrereqError("Invalid candidate_pool_size value: %s" %
1370
                                   str(err))
1371
      if self.op.candidate_pool_size < 1:
1372
        raise errors.OpPrereqError("At least one master candidate needed")
1373

    
1374
  def ExpandNames(self):
1375
    # FIXME: in the future maybe other cluster params won't require checking on
1376
    # all nodes to be modified.
1377
    self.needed_locks = {
1378
      locking.LEVEL_NODE: locking.ALL_SET,
1379
    }
1380
    self.share_locks[locking.LEVEL_NODE] = 1
1381

    
1382
  def BuildHooksEnv(self):
1383
    """Build hooks env.
1384

1385
    """
1386
    env = {
1387
      "OP_TARGET": self.cfg.GetClusterName(),
1388
      "NEW_VG_NAME": self.op.vg_name,
1389
      }
1390
    mn = self.cfg.GetMasterNode()
1391
    return env, [mn], [mn]
1392

    
1393
  def CheckPrereq(self):
1394
    """Check prerequisites.
1395

1396
    This checks whether the given params don't conflict and
1397
    if the given volume group is valid.
1398

1399
    """
1400
    # FIXME: This only works because there is only one parameter that can be
1401
    # changed or removed.
1402
    if self.op.vg_name is not None and not self.op.vg_name:
1403
      instances = self.cfg.GetAllInstancesInfo().values()
1404
      for inst in instances:
1405
        for disk in inst.disks:
1406
          if _RecursiveCheckIfLVMBased(disk):
1407
            raise errors.OpPrereqError("Cannot disable lvm storage while"
1408
                                       " lvm-based instances exist")
1409

    
1410
    node_list = self.acquired_locks[locking.LEVEL_NODE]
1411

    
1412
    # if vg_name not None, checks given volume group on all nodes
1413
    if self.op.vg_name:
1414
      vglist = self.rpc.call_vg_list(node_list)
1415
      for node in node_list:
1416
        if vglist[node].failed:
1417
          # ignoring down node
1418
          self.LogWarning("Node %s unreachable/error, ignoring" % node)
1419
          continue
1420
        vgstatus = utils.CheckVolumeGroupSize(vglist[node].data,
1421
                                              self.op.vg_name,
1422
                                              constants.MIN_VG_SIZE)
1423
        if vgstatus:
1424
          raise errors.OpPrereqError("Error on node '%s': %s" %
1425
                                     (node, vgstatus))
1426

    
1427
    self.cluster = cluster = self.cfg.GetClusterInfo()
1428
    # validate beparams changes
1429
    if self.op.beparams:
1430
      utils.CheckBEParams(self.op.beparams)
1431
      self.new_beparams = cluster.FillDict(
1432
        cluster.beparams[constants.BEGR_DEFAULT], self.op.beparams)
1433

    
1434
    # hypervisor list/parameters
1435
    self.new_hvparams = cluster.FillDict(cluster.hvparams, {})
1436
    if self.op.hvparams:
1437
      if not isinstance(self.op.hvparams, dict):
1438
        raise errors.OpPrereqError("Invalid 'hvparams' parameter on input")
1439
      for hv_name, hv_dict in self.op.hvparams.items():
1440
        if hv_name not in self.new_hvparams:
1441
          self.new_hvparams[hv_name] = hv_dict
1442
        else:
1443
          self.new_hvparams[hv_name].update(hv_dict)
1444

    
1445
    if self.op.enabled_hypervisors is not None:
1446
      self.hv_list = self.op.enabled_hypervisors
1447
    else:
1448
      self.hv_list = cluster.enabled_hypervisors
1449

    
1450
    if self.op.hvparams or self.op.enabled_hypervisors is not None:
1451
      # either the enabled list has changed, or the parameters have, validate
1452
      for hv_name, hv_params in self.new_hvparams.items():
1453
        if ((self.op.hvparams and hv_name in self.op.hvparams) or
1454
            (self.op.enabled_hypervisors and
1455
             hv_name in self.op.enabled_hypervisors)):
1456
          # either this is a new hypervisor, or its parameters have changed
1457
          hv_class = hypervisor.GetHypervisor(hv_name)
1458
          hv_class.CheckParameterSyntax(hv_params)
1459
          _CheckHVParams(self, node_list, hv_name, hv_params)
1460

    
1461
  def Exec(self, feedback_fn):
1462
    """Change the parameters of the cluster.
1463

1464
    """
1465
    if self.op.vg_name is not None:
1466
      if self.op.vg_name != self.cfg.GetVGName():
1467
        self.cfg.SetVGName(self.op.vg_name)
1468
      else:
1469
        feedback_fn("Cluster LVM configuration already in desired"
1470
                    " state, not changing")
1471
    if self.op.hvparams:
1472
      self.cluster.hvparams = self.new_hvparams
1473
    if self.op.enabled_hypervisors is not None:
1474
      self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
1475
    if self.op.beparams:
1476
      self.cluster.beparams[constants.BEGR_DEFAULT] = self.new_beparams
1477
    if self.op.candidate_pool_size is not None:
1478
      self.cluster.candidate_pool_size = self.op.candidate_pool_size
1479

    
1480
    self.cfg.Update(self.cluster)
1481

    
1482
    # we want to update nodes after the cluster so that if any errors
1483
    # happen, we have recorded and saved the cluster info
1484
    if self.op.candidate_pool_size is not None:
1485
      _AdjustCandidatePool(self)
1486

    
1487

    
1488
class LURedistributeConfig(NoHooksLU):
1489
  """Force the redistribution of cluster configuration.
1490

1491
  This is a very simple LU.
1492

1493
  """
1494
  _OP_REQP = []
1495
  REQ_BGL = False
1496

    
1497
  def ExpandNames(self):
1498
    self.needed_locks = {
1499
      locking.LEVEL_NODE: locking.ALL_SET,
1500
    }
1501
    self.share_locks[locking.LEVEL_NODE] = 1
1502

    
1503
  def CheckPrereq(self):
1504
    """Check prerequisites.
1505

1506
    """
1507

    
1508
  def Exec(self, feedback_fn):
1509
    """Redistribute the configuration.
1510

1511
    """
1512
    self.cfg.Update(self.cfg.GetClusterInfo())
1513

    
1514

    
1515
def _WaitForSync(lu, instance, oneshot=False, unlock=False):
1516
  """Sleep and poll for an instance's disk to sync.
1517

1518
  """
1519
  if not instance.disks:
1520
    return True
1521

    
1522
  if not oneshot:
1523
    lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
1524

    
1525
  node = instance.primary_node
1526

    
1527
  for dev in instance.disks:
1528
    lu.cfg.SetDiskID(dev, node)
1529

    
1530
  retries = 0
1531
  while True:
1532
    max_time = 0
1533
    done = True
1534
    cumul_degraded = False
1535
    rstats = lu.rpc.call_blockdev_getmirrorstatus(node, instance.disks)
1536
    if rstats.failed or not rstats.data:
1537
      lu.LogWarning("Can't get any data from node %s", node)
1538
      retries += 1
1539
      if retries >= 10:
1540
        raise errors.RemoteError("Can't contact node %s for mirror data,"
1541
                                 " aborting." % node)
1542
      time.sleep(6)
1543
      continue
1544
    rstats = rstats.data
1545
    retries = 0
1546
    for i, mstat in enumerate(rstats):
1547
      if mstat is None:
1548
        lu.LogWarning("Can't compute data for node %s/%s",
1549
                           node, instance.disks[i].iv_name)
1550
        continue
1551
      # we ignore the ldisk parameter
1552
      perc_done, est_time, is_degraded, _ = mstat
1553
      cumul_degraded = cumul_degraded or (is_degraded and perc_done is None)
1554
      if perc_done is not None:
1555
        done = False
1556
        if est_time is not None:
1557
          rem_time = "%d estimated seconds remaining" % est_time
1558
          max_time = est_time
1559
        else:
1560
          rem_time = "no time estimate"
1561
        lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
1562
                        (instance.disks[i].iv_name, perc_done, rem_time))
1563
    if done or oneshot:
1564
      break
1565

    
1566
    time.sleep(min(60, max_time))
1567

    
1568
  if done:
1569
    lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
1570
  return not cumul_degraded
1571

    
1572

    
1573
def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
1574
  """Check that mirrors are not degraded.
1575

1576
  The ldisk parameter, if True, will change the test from the
1577
  is_degraded attribute (which represents overall non-ok status for
1578
  the device(s)) to the ldisk (representing the local storage status).
1579

1580
  """
1581
  lu.cfg.SetDiskID(dev, node)
1582
  if ldisk:
1583
    idx = 6
1584
  else:
1585
    idx = 5
1586

    
1587
  result = True
1588
  if on_primary or dev.AssembleOnSecondary():
1589
    rstats = lu.rpc.call_blockdev_find(node, dev)
1590
    msg = rstats.RemoteFailMsg()
1591
    if msg:
1592
      lu.LogWarning("Can't find disk on node %s: %s", node, msg)
1593
      result = False
1594
    elif not rstats.payload:
1595
      lu.LogWarning("Can't find disk on node %s", node)
1596
      result = False
1597
    else:
1598
      result = result and (not rstats.payload[idx])
1599
  if dev.children:
1600
    for child in dev.children:
1601
      result = result and _CheckDiskConsistency(lu, child, node, on_primary)
1602

    
1603
  return result
1604

    
1605

    
1606
class LUDiagnoseOS(NoHooksLU):
1607
  """Logical unit for OS diagnose/query.
1608

1609
  """
1610
  _OP_REQP = ["output_fields", "names"]
1611
  REQ_BGL = False
1612
  _FIELDS_STATIC = utils.FieldSet()
1613
  _FIELDS_DYNAMIC = utils.FieldSet("name", "valid", "node_status")
1614

    
1615
  def ExpandNames(self):
1616
    if self.op.names:
1617
      raise errors.OpPrereqError("Selective OS query not supported")
1618

    
1619
    _CheckOutputFields(static=self._FIELDS_STATIC,
1620
                       dynamic=self._FIELDS_DYNAMIC,
1621
                       selected=self.op.output_fields)
1622

    
1623
    # Lock all nodes, in shared mode
1624
    self.needed_locks = {}
1625
    self.share_locks[locking.LEVEL_NODE] = 1
1626
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
1627

    
1628
  def CheckPrereq(self):
1629
    """Check prerequisites.
1630

1631
    """
1632

    
1633
  @staticmethod
1634
  def _DiagnoseByOS(node_list, rlist):
1635
    """Remaps a per-node return list into an a per-os per-node dictionary
1636

1637
    @param node_list: a list with the names of all nodes
1638
    @param rlist: a map with node names as keys and OS objects as values
1639

1640
    @rtype: dict
1641
    @returns: a dictionary with osnames as keys and as value another map, with
1642
        nodes as keys and list of OS objects as values, eg::
1643

1644
          {"debian-etch": {"node1": [<object>,...],
1645
                           "node2": [<object>,]}
1646
          }
1647

1648
    """
1649
    all_os = {}
1650
    for node_name, nr in rlist.iteritems():
1651
      if nr.failed or not nr.data:
1652
        continue
1653
      for os_obj in nr.data:
1654
        if os_obj.name not in all_os:
1655
          # build a list of nodes for this os containing empty lists
1656
          # for each node in node_list
1657
          all_os[os_obj.name] = {}
1658
          for nname in node_list:
1659
            all_os[os_obj.name][nname] = []
1660
        all_os[os_obj.name][node_name].append(os_obj)
1661
    return all_os
1662

    
1663
  def Exec(self, feedback_fn):
1664
    """Compute the list of OSes.
1665

1666
    """
1667
    node_list = self.acquired_locks[locking.LEVEL_NODE]
1668
    valid_nodes = [node for node in self.cfg.GetOnlineNodeList()
1669
                   if node in node_list]
1670
    node_data = self.rpc.call_os_diagnose(valid_nodes)
1671
    if node_data == False:
1672
      raise errors.OpExecError("Can't gather the list of OSes")
1673
    pol = self._DiagnoseByOS(valid_nodes, node_data)
1674
    output = []
1675
    for os_name, os_data in pol.iteritems():
1676
      row = []
1677
      for field in self.op.output_fields:
1678
        if field == "name":
1679
          val = os_name
1680
        elif field == "valid":
1681
          val = utils.all([osl and osl[0] for osl in os_data.values()])
1682
        elif field == "node_status":
1683
          val = {}
1684
          for node_name, nos_list in os_data.iteritems():
1685
            val[node_name] = [(v.status, v.path) for v in nos_list]
1686
        else:
1687
          raise errors.ParameterError(field)
1688
        row.append(val)
1689
      output.append(row)
1690

    
1691
    return output
1692

    
1693

    
1694
class LURemoveNode(LogicalUnit):
1695
  """Logical unit for removing a node.
1696

1697
  """
1698
  HPATH = "node-remove"
1699
  HTYPE = constants.HTYPE_NODE
1700
  _OP_REQP = ["node_name"]
1701

    
1702
  def BuildHooksEnv(self):
1703
    """Build hooks env.
1704

1705
    This doesn't run on the target node in the pre phase as a failed
1706
    node would then be impossible to remove.
1707

1708
    """
1709
    env = {
1710
      "OP_TARGET": self.op.node_name,
1711
      "NODE_NAME": self.op.node_name,
1712
      }
1713
    all_nodes = self.cfg.GetNodeList()
1714
    all_nodes.remove(self.op.node_name)
1715
    return env, all_nodes, all_nodes
1716

    
1717
  def CheckPrereq(self):
1718
    """Check prerequisites.
1719

1720
    This checks:
1721
     - the node exists in the configuration
1722
     - it does not have primary or secondary instances
1723
     - it's not the master
1724

1725
    Any errors are signalled by raising errors.OpPrereqError.
1726

1727
    """
1728
    node = self.cfg.GetNodeInfo(self.cfg.ExpandNodeName(self.op.node_name))
1729
    if node is None:
1730
      raise errors.OpPrereqError, ("Node '%s' is unknown." % self.op.node_name)
1731

    
1732
    instance_list = self.cfg.GetInstanceList()
1733

    
1734
    masternode = self.cfg.GetMasterNode()
1735
    if node.name == masternode:
1736
      raise errors.OpPrereqError("Node is the master node,"
1737
                                 " you need to failover first.")
1738

    
1739
    for instance_name in instance_list:
1740
      instance = self.cfg.GetInstanceInfo(instance_name)
1741
      if node.name in instance.all_nodes:
1742
        raise errors.OpPrereqError("Instance %s is still running on the node,"
1743
                                   " please remove first." % instance_name)
1744
    self.op.node_name = node.name
1745
    self.node = node
1746

    
1747
  def Exec(self, feedback_fn):
1748
    """Removes the node from the cluster.
1749

1750
    """
1751
    node = self.node
1752
    logging.info("Stopping the node daemon and removing configs from node %s",
1753
                 node.name)
1754

    
1755
    self.context.RemoveNode(node.name)
1756

    
1757
    self.rpc.call_node_leave_cluster(node.name)
1758

    
1759
    # Promote nodes to master candidate as needed
1760
    _AdjustCandidatePool(self)
1761

    
1762

    
1763
class LUQueryNodes(NoHooksLU):
1764
  """Logical unit for querying nodes.
1765

1766
  """
1767
  _OP_REQP = ["output_fields", "names", "use_locking"]
1768
  REQ_BGL = False
1769
  _FIELDS_DYNAMIC = utils.FieldSet(
1770
    "dtotal", "dfree",
1771
    "mtotal", "mnode", "mfree",
1772
    "bootid",
1773
    "ctotal", "cnodes", "csockets",
1774
    )
1775

    
1776
  _FIELDS_STATIC = utils.FieldSet(
1777
    "name", "pinst_cnt", "sinst_cnt",
1778
    "pinst_list", "sinst_list",
1779
    "pip", "sip", "tags",
1780
    "serial_no",
1781
    "master_candidate",
1782
    "master",
1783
    "offline",
1784
    "drained",
1785
    )
1786

    
1787
  def ExpandNames(self):
1788
    _CheckOutputFields(static=self._FIELDS_STATIC,
1789
                       dynamic=self._FIELDS_DYNAMIC,
1790
                       selected=self.op.output_fields)
1791

    
1792
    self.needed_locks = {}
1793
    self.share_locks[locking.LEVEL_NODE] = 1
1794

    
1795
    if self.op.names:
1796
      self.wanted = _GetWantedNodes(self, self.op.names)
1797
    else:
1798
      self.wanted = locking.ALL_SET
1799

    
1800
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
1801
    self.do_locking = self.do_node_query and self.op.use_locking
1802
    if self.do_locking:
1803
      # if we don't request only static fields, we need to lock the nodes
1804
      self.needed_locks[locking.LEVEL_NODE] = self.wanted
1805

    
1806

    
1807
  def CheckPrereq(self):
1808
    """Check prerequisites.
1809

1810
    """
1811
    # The validation of the node list is done in the _GetWantedNodes,
1812
    # if non empty, and if empty, there's no validation to do
1813
    pass
1814

    
1815
  def Exec(self, feedback_fn):
1816
    """Computes the list of nodes and their attributes.
1817

1818
    """
1819
    all_info = self.cfg.GetAllNodesInfo()
1820
    if self.do_locking:
1821
      nodenames = self.acquired_locks[locking.LEVEL_NODE]
1822
    elif self.wanted != locking.ALL_SET:
1823
      nodenames = self.wanted
1824
      missing = set(nodenames).difference(all_info.keys())
1825
      if missing:
1826
        raise errors.OpExecError(
1827
          "Some nodes were removed before retrieving their data: %s" % missing)
1828
    else:
1829
      nodenames = all_info.keys()
1830

    
1831
    nodenames = utils.NiceSort(nodenames)
1832
    nodelist = [all_info[name] for name in nodenames]
1833

    
1834
    # begin data gathering
1835

    
1836
    if self.do_node_query:
1837
      live_data = {}
1838
      node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
1839
                                          self.cfg.GetHypervisorType())
1840
      for name in nodenames:
1841
        nodeinfo = node_data[name]
1842
        if not nodeinfo.failed and nodeinfo.data:
1843
          nodeinfo = nodeinfo.data
1844
          fn = utils.TryConvert
1845
          live_data[name] = {
1846
            "mtotal": fn(int, nodeinfo.get('memory_total', None)),
1847
            "mnode": fn(int, nodeinfo.get('memory_dom0', None)),
1848
            "mfree": fn(int, nodeinfo.get('memory_free', None)),
1849
            "dtotal": fn(int, nodeinfo.get('vg_size', None)),
1850
            "dfree": fn(int, nodeinfo.get('vg_free', None)),
1851
            "ctotal": fn(int, nodeinfo.get('cpu_total', None)),
1852
            "bootid": nodeinfo.get('bootid', None),
1853
            "cnodes": fn(int, nodeinfo.get('cpu_nodes', None)),
1854
            "csockets": fn(int, nodeinfo.get('cpu_sockets', None)),
1855
            }
1856
        else:
1857
          live_data[name] = {}
1858
    else:
1859
      live_data = dict.fromkeys(nodenames, {})
1860

    
1861
    node_to_primary = dict([(name, set()) for name in nodenames])
1862
    node_to_secondary = dict([(name, set()) for name in nodenames])
1863

    
1864
    inst_fields = frozenset(("pinst_cnt", "pinst_list",
1865
                             "sinst_cnt", "sinst_list"))
1866
    if inst_fields & frozenset(self.op.output_fields):
1867
      instancelist = self.cfg.GetInstanceList()
1868

    
1869
      for instance_name in instancelist:
1870
        inst = self.cfg.GetInstanceInfo(instance_name)
1871
        if inst.primary_node in node_to_primary:
1872
          node_to_primary[inst.primary_node].add(inst.name)
1873
        for secnode in inst.secondary_nodes:
1874
          if secnode in node_to_secondary:
1875
            node_to_secondary[secnode].add(inst.name)
1876

    
1877
    master_node = self.cfg.GetMasterNode()
1878

    
1879
    # end data gathering
1880

    
1881
    output = []
1882
    for node in nodelist:
1883
      node_output = []
1884
      for field in self.op.output_fields:
1885
        if field == "name":
1886
          val = node.name
1887
        elif field == "pinst_list":
1888
          val = list(node_to_primary[node.name])
1889
        elif field == "sinst_list":
1890
          val = list(node_to_secondary[node.name])
1891
        elif field == "pinst_cnt":
1892
          val = len(node_to_primary[node.name])
1893
        elif field == "sinst_cnt":
1894
          val = len(node_to_secondary[node.name])
1895
        elif field == "pip":
1896
          val = node.primary_ip
1897
        elif field == "sip":
1898
          val = node.secondary_ip
1899
        elif field == "tags":
1900
          val = list(node.GetTags())
1901
        elif field == "serial_no":
1902
          val = node.serial_no
1903
        elif field == "master_candidate":
1904
          val = node.master_candidate
1905
        elif field == "master":
1906
          val = node.name == master_node
1907
        elif field == "offline":
1908
          val = node.offline
1909
        elif field == "drained":
1910
          val = node.drained
1911
        elif self._FIELDS_DYNAMIC.Matches(field):
1912
          val = live_data[node.name].get(field, None)
1913
        else:
1914
          raise errors.ParameterError(field)
1915
        node_output.append(val)
1916
      output.append(node_output)
1917

    
1918
    return output
1919

    
1920

    
1921
class LUQueryNodeVolumes(NoHooksLU):
1922
  """Logical unit for getting volumes on node(s).
1923

1924
  """
1925
  _OP_REQP = ["nodes", "output_fields"]
1926
  REQ_BGL = False
1927
  _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
1928
  _FIELDS_STATIC = utils.FieldSet("node")
1929

    
1930
  def ExpandNames(self):
1931
    _CheckOutputFields(static=self._FIELDS_STATIC,
1932
                       dynamic=self._FIELDS_DYNAMIC,
1933
                       selected=self.op.output_fields)
1934

    
1935
    self.needed_locks = {}
1936
    self.share_locks[locking.LEVEL_NODE] = 1
1937
    if not self.op.nodes:
1938
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
1939
    else:
1940
      self.needed_locks[locking.LEVEL_NODE] = \
1941
        _GetWantedNodes(self, self.op.nodes)
1942

    
1943
  def CheckPrereq(self):
1944
    """Check prerequisites.
1945

1946
    This checks that the fields required are valid output fields.
1947

1948
    """
1949
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
1950

    
1951
  def Exec(self, feedback_fn):
1952
    """Computes the list of nodes and their attributes.
1953

1954
    """
1955
    nodenames = self.nodes
1956
    volumes = self.rpc.call_node_volumes(nodenames)
1957

    
1958
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
1959
             in self.cfg.GetInstanceList()]
1960

    
1961
    lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
1962

    
1963
    output = []
1964
    for node in nodenames:
1965
      if node not in volumes or volumes[node].failed or not volumes[node].data:
1966
        continue
1967

    
1968
      node_vols = volumes[node].data[:]
1969
      node_vols.sort(key=lambda vol: vol['dev'])
1970

    
1971
      for vol in node_vols:
1972
        node_output = []
1973
        for field in self.op.output_fields:
1974
          if field == "node":
1975
            val = node
1976
          elif field == "phys":
1977
            val = vol['dev']
1978
          elif field == "vg":
1979
            val = vol['vg']
1980
          elif field == "name":
1981
            val = vol['name']
1982
          elif field == "size":
1983
            val = int(float(vol['size']))
1984
          elif field == "instance":
1985
            for inst in ilist:
1986
              if node not in lv_by_node[inst]:
1987
                continue
1988
              if vol['name'] in lv_by_node[inst][node]:
1989
                val = inst.name
1990
                break
1991
            else:
1992
              val = '-'
1993
          else:
1994
            raise errors.ParameterError(field)
1995
          node_output.append(str(val))
1996

    
1997
        output.append(node_output)
1998

    
1999
    return output
2000

    
2001

    
2002
class LUAddNode(LogicalUnit):
2003
  """Logical unit for adding node to the cluster.
2004

2005
  """
2006
  HPATH = "node-add"
2007
  HTYPE = constants.HTYPE_NODE
2008
  _OP_REQP = ["node_name"]
2009

    
2010
  def BuildHooksEnv(self):
2011
    """Build hooks env.
2012

2013
    This will run on all nodes before, and on all nodes + the new node after.
2014

2015
    """
2016
    env = {
2017
      "OP_TARGET": self.op.node_name,
2018
      "NODE_NAME": self.op.node_name,
2019
      "NODE_PIP": self.op.primary_ip,
2020
      "NODE_SIP": self.op.secondary_ip,
2021
      }
2022
    nodes_0 = self.cfg.GetNodeList()
2023
    nodes_1 = nodes_0 + [self.op.node_name, ]
2024
    return env, nodes_0, nodes_1
2025

    
2026
  def CheckPrereq(self):
2027
    """Check prerequisites.
2028

2029
    This checks:
2030
     - the new node is not already in the config
2031
     - it is resolvable
2032
     - its parameters (single/dual homed) matches the cluster
2033

2034
    Any errors are signalled by raising errors.OpPrereqError.
2035

2036
    """
2037
    node_name = self.op.node_name
2038
    cfg = self.cfg
2039

    
2040
    dns_data = utils.HostInfo(node_name)
2041

    
2042
    node = dns_data.name
2043
    primary_ip = self.op.primary_ip = dns_data.ip
2044
    secondary_ip = getattr(self.op, "secondary_ip", None)
2045
    if secondary_ip is None:
2046
      secondary_ip = primary_ip
2047
    if not utils.IsValidIP(secondary_ip):
2048
      raise errors.OpPrereqError("Invalid secondary IP given")
2049
    self.op.secondary_ip = secondary_ip
2050

    
2051
    node_list = cfg.GetNodeList()
2052
    if not self.op.readd and node in node_list:
2053
      raise errors.OpPrereqError("Node %s is already in the configuration" %
2054
                                 node)
2055
    elif self.op.readd and node not in node_list:
2056
      raise errors.OpPrereqError("Node %s is not in the configuration" % node)
2057

    
2058
    for existing_node_name in node_list:
2059
      existing_node = cfg.GetNodeInfo(existing_node_name)
2060

    
2061
      if self.op.readd and node == existing_node_name:
2062
        if (existing_node.primary_ip != primary_ip or
2063
            existing_node.secondary_ip != secondary_ip):
2064
          raise errors.OpPrereqError("Readded node doesn't have the same IP"
2065
                                     " address configuration as before")
2066
        continue
2067

    
2068
      if (existing_node.primary_ip == primary_ip or
2069
          existing_node.secondary_ip == primary_ip or
2070
          existing_node.primary_ip == secondary_ip or
2071
          existing_node.secondary_ip == secondary_ip):
2072
        raise errors.OpPrereqError("New node ip address(es) conflict with"
2073
                                   " existing node %s" % existing_node.name)
2074

    
2075
    # check that the type of the node (single versus dual homed) is the
2076
    # same as for the master
2077
    myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
2078
    master_singlehomed = myself.secondary_ip == myself.primary_ip
2079
    newbie_singlehomed = secondary_ip == primary_ip
2080
    if master_singlehomed != newbie_singlehomed:
2081
      if master_singlehomed:
2082
        raise errors.OpPrereqError("The master has no private ip but the"
2083
                                   " new node has one")
2084
      else:
2085
        raise errors.OpPrereqError("The master has a private ip but the"
2086
                                   " new node doesn't have one")
2087

    
2088
    # checks reachablity
2089
    if not utils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
2090
      raise errors.OpPrereqError("Node not reachable by ping")
2091

    
2092
    if not newbie_singlehomed:
2093
      # check reachability from my secondary ip to newbie's secondary ip
2094
      if not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
2095
                           source=myself.secondary_ip):
2096
        raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
2097
                                   " based ping to noded port")
2098

    
2099
    cp_size = self.cfg.GetClusterInfo().candidate_pool_size
2100
    mc_now, _ = self.cfg.GetMasterCandidateStats()
2101
    master_candidate = mc_now < cp_size
2102

    
2103
    self.new_node = objects.Node(name=node,
2104
                                 primary_ip=primary_ip,
2105
                                 secondary_ip=secondary_ip,
2106
                                 master_candidate=master_candidate,
2107
                                 offline=False, drained=False)
2108

    
2109
  def Exec(self, feedback_fn):
2110
    """Adds the new node to the cluster.
2111

2112
    """
2113
    new_node = self.new_node
2114
    node = new_node.name
2115

    
2116
    # check connectivity
2117
    result = self.rpc.call_version([node])[node]
2118
    result.Raise()
2119
    if result.data:
2120
      if constants.PROTOCOL_VERSION == result.data:
2121
        logging.info("Communication to node %s fine, sw version %s match",
2122
                     node, result.data)
2123
      else:
2124
        raise errors.OpExecError("Version mismatch master version %s,"
2125
                                 " node version %s" %
2126
                                 (constants.PROTOCOL_VERSION, result.data))
2127
    else:
2128
      raise errors.OpExecError("Cannot get version from the new node")
2129

    
2130
    # setup ssh on node
2131
    logging.info("Copy ssh key to node %s", node)
2132
    priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
2133
    keyarray = []
2134
    keyfiles = [constants.SSH_HOST_DSA_PRIV, constants.SSH_HOST_DSA_PUB,
2135
                constants.SSH_HOST_RSA_PRIV, constants.SSH_HOST_RSA_PUB,
2136
                priv_key, pub_key]
2137

    
2138
    for i in keyfiles:
2139
      f = open(i, 'r')
2140
      try:
2141
        keyarray.append(f.read())
2142
      finally:
2143
        f.close()
2144

    
2145
    result = self.rpc.call_node_add(node, keyarray[0], keyarray[1],
2146
                                    keyarray[2],
2147
                                    keyarray[3], keyarray[4], keyarray[5])
2148

    
2149
    msg = result.RemoteFailMsg()
2150
    if msg:
2151
      raise errors.OpExecError("Cannot transfer ssh keys to the"
2152
                               " new node: %s" % msg)
2153

    
2154
    # Add node to our /etc/hosts, and add key to known_hosts
2155
    utils.AddHostToEtcHosts(new_node.name)
2156

    
2157
    if new_node.secondary_ip != new_node.primary_ip:
2158
      result = self.rpc.call_node_has_ip_address(new_node.name,
2159
                                                 new_node.secondary_ip)
2160
      if result.failed or not result.data:
2161
        raise errors.OpExecError("Node claims it doesn't have the secondary ip"
2162
                                 " you gave (%s). Please fix and re-run this"
2163
                                 " command." % new_node.secondary_ip)
2164

    
2165
    node_verify_list = [self.cfg.GetMasterNode()]
2166
    node_verify_param = {
2167
      'nodelist': [node],
2168
      # TODO: do a node-net-test as well?
2169
    }
2170

    
2171
    result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
2172
                                       self.cfg.GetClusterName())
2173
    for verifier in node_verify_list:
2174
      if result[verifier].failed or not result[verifier].data:
2175
        raise errors.OpExecError("Cannot communicate with %s's node daemon"
2176
                                 " for remote verification" % verifier)
2177
      if result[verifier].data['nodelist']:
2178
        for failed in result[verifier].data['nodelist']:
2179
          feedback_fn("ssh/hostname verification failed %s -> %s" %
2180
                      (verifier, result[verifier].data['nodelist'][failed]))
2181
        raise errors.OpExecError("ssh/hostname verification failed.")
2182

    
2183
    # Distribute updated /etc/hosts and known_hosts to all nodes,
2184
    # including the node just added
2185
    myself = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
2186
    dist_nodes = self.cfg.GetNodeList()
2187
    if not self.op.readd:
2188
      dist_nodes.append(node)
2189
    if myself.name in dist_nodes:
2190
      dist_nodes.remove(myself.name)
2191

    
2192
    logging.debug("Copying hosts and known_hosts to all nodes")
2193
    for fname in (constants.ETC_HOSTS, constants.SSH_KNOWN_HOSTS_FILE):
2194
      result = self.rpc.call_upload_file(dist_nodes, fname)
2195
      for to_node, to_result in result.iteritems():
2196
        if to_result.failed or not to_result.data:
2197
          logging.error("Copy of file %s to node %s failed", fname, to_node)
2198

    
2199
    to_copy = []
2200
    enabled_hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
2201
    if constants.HTS_USE_VNC.intersection(enabled_hypervisors):
2202
      to_copy.append(constants.VNC_PASSWORD_FILE)
2203

    
2204
    for fname in to_copy:
2205
      result = self.rpc.call_upload_file([node], fname)
2206
      if result[node].failed or not result[node]:
2207
        logging.error("Could not copy file %s to node %s", fname, node)
2208

    
2209
    if self.op.readd:
2210
      self.context.ReaddNode(new_node)
2211
    else:
2212
      self.context.AddNode(new_node)
2213

    
2214

    
2215
class LUSetNodeParams(LogicalUnit):
2216
  """Modifies the parameters of a node.
2217

2218
  """
2219
  HPATH = "node-modify"
2220
  HTYPE = constants.HTYPE_NODE
2221
  _OP_REQP = ["node_name"]
2222
  REQ_BGL = False
2223

    
2224
  def CheckArguments(self):
2225
    node_name = self.cfg.ExpandNodeName(self.op.node_name)
2226
    if node_name is None:
2227
      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name)
2228
    self.op.node_name = node_name
2229
    _CheckBooleanOpField(self.op, 'master_candidate')
2230
    _CheckBooleanOpField(self.op, 'offline')
2231
    if self.op.master_candidate is None and self.op.offline is None:
2232
      raise errors.OpPrereqError("Please pass at least one modification")
2233
    if self.op.offline == True and self.op.master_candidate == True:
2234
      raise errors.OpPrereqError("Can't set the node into offline and"
2235
                                 " master_candidate at the same time")
2236

    
2237
  def ExpandNames(self):
2238
    self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
2239

    
2240
  def BuildHooksEnv(self):
2241
    """Build hooks env.
2242

2243
    This runs on the master node.
2244

2245
    """
2246
    env = {
2247
      "OP_TARGET": self.op.node_name,
2248
      "MASTER_CANDIDATE": str(self.op.master_candidate),
2249
      "OFFLINE": str(self.op.offline),
2250
      }
2251
    nl = [self.cfg.GetMasterNode(),
2252
          self.op.node_name]
2253
    return env, nl, nl
2254

    
2255
  def CheckPrereq(self):
2256
    """Check prerequisites.
2257

2258
    This only checks the instance list against the existing names.
2259

2260
    """
2261
    node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
2262

    
2263
    if ((self.op.master_candidate == False or self.op.offline == True)
2264
        and node.master_candidate):
2265
      # we will demote the node from master_candidate
2266
      if self.op.node_name == self.cfg.GetMasterNode():
2267
        raise errors.OpPrereqError("The master node has to be a"
2268
                                   " master candidate and online")
2269
      cp_size = self.cfg.GetClusterInfo().candidate_pool_size
2270
      num_candidates, _ = self.cfg.GetMasterCandidateStats()
2271
      if num_candidates <= cp_size:
2272
        msg = ("Not enough master candidates (desired"
2273
               " %d, new value will be %d)" % (cp_size, num_candidates-1))
2274
        if self.op.force:
2275
          self.LogWarning(msg)
2276
        else:
2277
          raise errors.OpPrereqError(msg)
2278

    
2279
    if (self.op.master_candidate == True and node.offline and
2280
        not self.op.offline == False):
2281
      raise errors.OpPrereqError("Can't set an offline node to"
2282
                                 " master_candidate")
2283

    
2284
    return
2285

    
2286
  def Exec(self, feedback_fn):
2287
    """Modifies a node.
2288

2289
    """
2290
    node = self.node
2291

    
2292
    result = []
2293

    
2294
    if self.op.offline is not None:
2295
      node.offline = self.op.offline
2296
      result.append(("offline", str(self.op.offline)))
2297
      if self.op.offline == True and node.master_candidate:
2298
        node.master_candidate = False
2299
        result.append(("master_candidate", "auto-demotion due to offline"))
2300

    
2301
    if self.op.master_candidate is not None:
2302
      node.master_candidate = self.op.master_candidate
2303
      result.append(("master_candidate", str(self.op.master_candidate)))
2304
      if self.op.master_candidate == False:
2305
        rrc = self.rpc.call_node_demote_from_mc(node.name)
2306
        msg = rrc.RemoteFailMsg()
2307
        if msg:
2308
          self.LogWarning("Node failed to demote itself: %s" % msg)
2309

    
2310
    # this will trigger configuration file update, if needed
2311
    self.cfg.Update(node)
2312
    # this will trigger job queue propagation or cleanup
2313
    if self.op.node_name != self.cfg.GetMasterNode():
2314
      self.context.ReaddNode(node)
2315

    
2316
    return result
2317

    
2318

    
2319
class LUQueryClusterInfo(NoHooksLU):
2320
  """Query cluster configuration.
2321

2322
  """
2323
  _OP_REQP = []
2324
  REQ_BGL = False
2325

    
2326
  def ExpandNames(self):
2327
    self.needed_locks = {}
2328

    
2329
  def CheckPrereq(self):
2330
    """No prerequsites needed for this LU.
2331

2332
    """
2333
    pass
2334

    
2335
  def Exec(self, feedback_fn):
2336
    """Return cluster config.
2337

2338
    """
2339
    cluster = self.cfg.GetClusterInfo()
2340
    result = {
2341
      "software_version": constants.RELEASE_VERSION,
2342
      "protocol_version": constants.PROTOCOL_VERSION,
2343
      "config_version": constants.CONFIG_VERSION,
2344
      "os_api_version": constants.OS_API_VERSION,
2345
      "export_version": constants.EXPORT_VERSION,
2346
      "architecture": (platform.architecture()[0], platform.machine()),
2347
      "name": cluster.cluster_name,
2348
      "master": cluster.master_node,
2349
      "default_hypervisor": cluster.default_hypervisor,
2350
      "enabled_hypervisors": cluster.enabled_hypervisors,
2351
      "hvparams": dict([(hypervisor, cluster.hvparams[hypervisor])
2352
                        for hypervisor in cluster.enabled_hypervisors]),
2353
      "beparams": cluster.beparams,
2354
      "candidate_pool_size": cluster.candidate_pool_size,
2355
      }
2356

    
2357
    return result
2358

    
2359

    
2360
class LUQueryConfigValues(NoHooksLU):
2361
  """Return configuration values.
2362

2363
  """
2364
  _OP_REQP = []
2365
  REQ_BGL = False
2366
  _FIELDS_DYNAMIC = utils.FieldSet()
2367
  _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag")
2368

    
2369
  def ExpandNames(self):
2370
    self.needed_locks = {}
2371

    
2372
    _CheckOutputFields(static=self._FIELDS_STATIC,
2373
                       dynamic=self._FIELDS_DYNAMIC,
2374
                       selected=self.op.output_fields)
2375

    
2376
  def CheckPrereq(self):
2377
    """No prerequisites.
2378

2379
    """
2380
    pass
2381

    
2382
  def Exec(self, feedback_fn):
2383
    """Dump a representation of the cluster config to the standard output.
2384

2385
    """
2386
    values = []
2387
    for field in self.op.output_fields:
2388
      if field == "cluster_name":
2389
        entry = self.cfg.GetClusterName()
2390
      elif field == "master_node":
2391
        entry = self.cfg.GetMasterNode()
2392
      elif field == "drain_flag":
2393
        entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
2394
      else:
2395
        raise errors.ParameterError(field)
2396
      values.append(entry)
2397
    return values
2398

    
2399

    
2400
class LUActivateInstanceDisks(NoHooksLU):
2401
  """Bring up an instance's disks.
2402

2403
  """
2404
  _OP_REQP = ["instance_name"]
2405
  REQ_BGL = False
2406

    
2407
  def ExpandNames(self):
2408
    self._ExpandAndLockInstance()
2409
    self.needed_locks[locking.LEVEL_NODE] = []
2410
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2411

    
2412
  def DeclareLocks(self, level):
2413
    if level == locking.LEVEL_NODE:
2414
      self._LockInstancesNodes()
2415

    
2416
  def CheckPrereq(self):
2417
    """Check prerequisites.
2418

2419
    This checks that the instance is in the cluster.
2420

2421
    """
2422
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2423
    assert self.instance is not None, \
2424
      "Cannot retrieve locked instance %s" % self.op.instance_name
2425
    _CheckNodeOnline(self, self.instance.primary_node)
2426

    
2427
  def Exec(self, feedback_fn):
2428
    """Activate the disks.
2429

2430
    """
2431
    disks_ok, disks_info = _AssembleInstanceDisks(self, self.instance)
2432
    if not disks_ok:
2433
      raise errors.OpExecError("Cannot activate block devices")
2434

    
2435
    return disks_info
2436

    
2437

    
2438
def _AssembleInstanceDisks(lu, instance, ignore_secondaries=False):
2439
  """Prepare the block devices for an instance.
2440

2441
  This sets up the block devices on all nodes.
2442

2443
  @type lu: L{LogicalUnit}
2444
  @param lu: the logical unit on whose behalf we execute
2445
  @type instance: L{objects.Instance}
2446
  @param instance: the instance for whose disks we assemble
2447
  @type ignore_secondaries: boolean
2448
  @param ignore_secondaries: if true, errors on secondary nodes
2449
      won't result in an error return from the function
2450
  @return: False if the operation failed, otherwise a list of
2451
      (host, instance_visible_name, node_visible_name)
2452
      with the mapping from node devices to instance devices
2453

2454
  """
2455
  device_info = []
2456
  disks_ok = True
2457
  iname = instance.name
2458
  # With the two passes mechanism we try to reduce the window of
2459
  # opportunity for the race condition of switching DRBD to primary
2460
  # before handshaking occured, but we do not eliminate it
2461

    
2462
  # The proper fix would be to wait (with some limits) until the
2463
  # connection has been made and drbd transitions from WFConnection
2464
  # into any other network-connected state (Connected, SyncTarget,
2465
  # SyncSource, etc.)
2466

    
2467
  # 1st pass, assemble on all nodes in secondary mode
2468
  for inst_disk in instance.disks:
2469
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
2470
      lu.cfg.SetDiskID(node_disk, node)
2471
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
2472
      msg = result.RemoteFailMsg()
2473
      if msg:
2474
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
2475
                           " (is_primary=False, pass=1): %s",
2476
                           inst_disk.iv_name, node, msg)
2477
        if not ignore_secondaries:
2478
          disks_ok = False
2479

    
2480
  # FIXME: race condition on drbd migration to primary
2481

    
2482
  # 2nd pass, do only the primary node
2483
  for inst_disk in instance.disks:
2484
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
2485
      if node != instance.primary_node:
2486
        continue
2487
      lu.cfg.SetDiskID(node_disk, node)
2488
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
2489
      msg = result.RemoteFailMsg()
2490
      if msg:
2491
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
2492
                           " (is_primary=True, pass=2): %s",
2493
                           inst_disk.iv_name, node, msg)
2494
        disks_ok = False
2495
    device_info.append((instance.primary_node, inst_disk.iv_name, result.data))
2496

    
2497
  # leave the disks configured for the primary node
2498
  # this is a workaround that would be fixed better by
2499
  # improving the logical/physical id handling
2500
  for disk in instance.disks:
2501
    lu.cfg.SetDiskID(disk, instance.primary_node)
2502

    
2503
  return disks_ok, device_info
2504

    
2505

    
2506
def _StartInstanceDisks(lu, instance, force):
2507
  """Start the disks of an instance.
2508

2509
  """
2510
  disks_ok, dummy = _AssembleInstanceDisks(lu, instance,
2511
                                           ignore_secondaries=force)
2512
  if not disks_ok:
2513
    _ShutdownInstanceDisks(lu, instance)
2514
    if force is not None and not force:
2515
      lu.proc.LogWarning("", hint="If the message above refers to a"
2516
                         " secondary node,"
2517
                         " you can retry the operation using '--force'.")
2518
    raise errors.OpExecError("Disk consistency error")
2519

    
2520

    
2521
class LUDeactivateInstanceDisks(NoHooksLU):
2522
  """Shutdown an instance's disks.
2523

2524
  """
2525
  _OP_REQP = ["instance_name"]
2526
  REQ_BGL = False
2527

    
2528
  def ExpandNames(self):
2529
    self._ExpandAndLockInstance()
2530
    self.needed_locks[locking.LEVEL_NODE] = []
2531
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2532

    
2533
  def DeclareLocks(self, level):
2534
    if level == locking.LEVEL_NODE:
2535
      self._LockInstancesNodes()
2536

    
2537
  def CheckPrereq(self):
2538
    """Check prerequisites.
2539

2540
    This checks that the instance is in the cluster.
2541

2542
    """
2543
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2544
    assert self.instance is not None, \
2545
      "Cannot retrieve locked instance %s" % self.op.instance_name
2546

    
2547
  def Exec(self, feedback_fn):
2548
    """Deactivate the disks
2549

2550
    """
2551
    instance = self.instance
2552
    _SafeShutdownInstanceDisks(self, instance)
2553

    
2554

    
2555
def _SafeShutdownInstanceDisks(lu, instance):
2556
  """Shutdown block devices of an instance.
2557

2558
  This function checks if an instance is running, before calling
2559
  _ShutdownInstanceDisks.
2560

2561
  """
2562
  ins_l = lu.rpc.call_instance_list([instance.primary_node],
2563
                                      [instance.hypervisor])
2564
  ins_l = ins_l[instance.primary_node]
2565
  if ins_l.failed or not isinstance(ins_l.data, list):
2566
    raise errors.OpExecError("Can't contact node '%s'" %
2567
                             instance.primary_node)
2568

    
2569
  if instance.name in ins_l.data:
2570
    raise errors.OpExecError("Instance is running, can't shutdown"
2571
                             " block devices.")
2572

    
2573
  _ShutdownInstanceDisks(lu, instance)
2574

    
2575

    
2576
def _ShutdownInstanceDisks(lu, instance, ignore_primary=False):
2577
  """Shutdown block devices of an instance.
2578

2579
  This does the shutdown on all nodes of the instance.
2580

2581
  If the ignore_primary is false, errors on the primary node are
2582
  ignored.
2583

2584
  """
2585
  all_result = True
2586
  for disk in instance.disks:
2587
    for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
2588
      lu.cfg.SetDiskID(top_disk, node)
2589
      result = lu.rpc.call_blockdev_shutdown(node, top_disk)
2590
      msg = result.RemoteFailMsg()
2591
      if msg:
2592
        lu.LogWarning("Could not shutdown block device %s on node %s: %s",
2593
                      disk.iv_name, node, msg)
2594
        if not ignore_primary or node != instance.primary_node:
2595
          all_result = False
2596
  return all_result
2597

    
2598

    
2599
def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
2600
  """Checks if a node has enough free memory.
2601

2602
  This function check if a given node has the needed amount of free
2603
  memory. In case the node has less memory or we cannot get the
2604
  information from the node, this function raise an OpPrereqError
2605
  exception.
2606

2607
  @type lu: C{LogicalUnit}
2608
  @param lu: a logical unit from which we get configuration data
2609
  @type node: C{str}
2610
  @param node: the node to check
2611
  @type reason: C{str}
2612
  @param reason: string to use in the error message
2613
  @type requested: C{int}
2614
  @param requested: the amount of memory in MiB to check for
2615
  @type hypervisor_name: C{str}
2616
  @param hypervisor_name: the hypervisor to ask for memory stats
2617
  @raise errors.OpPrereqError: if the node doesn't have enough memory, or
2618
      we cannot check the node
2619

2620
  """
2621
  nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
2622
  nodeinfo[node].Raise()
2623
  free_mem = nodeinfo[node].data.get('memory_free')
2624
  if not isinstance(free_mem, int):
2625
    raise errors.OpPrereqError("Can't compute free memory on node %s, result"
2626
                             " was '%s'" % (node, free_mem))
2627
  if requested > free_mem:
2628
    raise errors.OpPrereqError("Not enough memory on node %s for %s:"
2629
                             " needed %s MiB, available %s MiB" %
2630
                             (node, reason, requested, free_mem))
2631

    
2632

    
2633
class LUStartupInstance(LogicalUnit):
2634
  """Starts an instance.
2635

2636
  """
2637
  HPATH = "instance-start"
2638
  HTYPE = constants.HTYPE_INSTANCE
2639
  _OP_REQP = ["instance_name", "force"]
2640
  REQ_BGL = False
2641

    
2642
  def ExpandNames(self):
2643
    self._ExpandAndLockInstance()
2644

    
2645
  def BuildHooksEnv(self):
2646
    """Build hooks env.
2647

2648
    This runs on master, primary and secondary nodes of the instance.
2649

2650
    """
2651
    env = {
2652
      "FORCE": self.op.force,
2653
      }
2654
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2655
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2656
    return env, nl, nl
2657

    
2658
  def CheckPrereq(self):
2659
    """Check prerequisites.
2660

2661
    This checks that the instance is in the cluster.
2662

2663
    """
2664
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2665
    assert self.instance is not None, \
2666
      "Cannot retrieve locked instance %s" % self.op.instance_name
2667

    
2668
    _CheckNodeOnline(self, instance.primary_node)
2669

    
2670
    bep = self.cfg.GetClusterInfo().FillBE(instance)
2671
    # check bridges existance
2672
    _CheckInstanceBridgesExist(self, instance)
2673

    
2674
    _CheckNodeFreeMemory(self, instance.primary_node,
2675
                         "starting instance %s" % instance.name,
2676
                         bep[constants.BE_MEMORY], instance.hypervisor)
2677

    
2678
  def Exec(self, feedback_fn):
2679
    """Start the instance.
2680

2681
    """
2682
    instance = self.instance
2683
    force = self.op.force
2684
    extra_args = getattr(self.op, "extra_args", "")
2685

    
2686
    self.cfg.MarkInstanceUp(instance.name)
2687

    
2688
    node_current = instance.primary_node
2689

    
2690
    _StartInstanceDisks(self, instance, force)
2691

    
2692
    result = self.rpc.call_instance_start(node_current, instance, extra_args)
2693
    msg = result.RemoteFailMsg()
2694
    if msg:
2695
      _ShutdownInstanceDisks(self, instance)
2696
      raise errors.OpExecError("Could not start instance: %s" % msg)
2697

    
2698

    
2699
class LURebootInstance(LogicalUnit):
2700
  """Reboot an instance.
2701

2702
  """
2703
  HPATH = "instance-reboot"
2704
  HTYPE = constants.HTYPE_INSTANCE
2705
  _OP_REQP = ["instance_name", "ignore_secondaries", "reboot_type"]
2706
  REQ_BGL = False
2707

    
2708
  def ExpandNames(self):
2709
    if self.op.reboot_type not in [constants.INSTANCE_REBOOT_SOFT,
2710
                                   constants.INSTANCE_REBOOT_HARD,
2711
                                   constants.INSTANCE_REBOOT_FULL]:
2712
      raise errors.ParameterError("reboot type not in [%s, %s, %s]" %
2713
                                  (constants.INSTANCE_REBOOT_SOFT,
2714
                                   constants.INSTANCE_REBOOT_HARD,
2715
                                   constants.INSTANCE_REBOOT_FULL))
2716
    self._ExpandAndLockInstance()
2717

    
2718
  def BuildHooksEnv(self):
2719
    """Build hooks env.
2720

2721
    This runs on master, primary and secondary nodes of the instance.
2722

2723
    """
2724
    env = {
2725
      "IGNORE_SECONDARIES": self.op.ignore_secondaries,
2726
      }
2727
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2728
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2729
    return env, nl, nl
2730

    
2731
  def CheckPrereq(self):
2732
    """Check prerequisites.
2733

2734
    This checks that the instance is in the cluster.
2735

2736
    """
2737
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2738
    assert self.instance is not None, \
2739
      "Cannot retrieve locked instance %s" % self.op.instance_name
2740

    
2741
    _CheckNodeOnline(self, instance.primary_node)
2742

    
2743
    # check bridges existance
2744
    _CheckInstanceBridgesExist(self, instance)
2745

    
2746
  def Exec(self, feedback_fn):
2747
    """Reboot the instance.
2748

2749
    """
2750
    instance = self.instance
2751
    ignore_secondaries = self.op.ignore_secondaries
2752
    reboot_type = self.op.reboot_type
2753
    extra_args = getattr(self.op, "extra_args", "")
2754

    
2755
    node_current = instance.primary_node
2756

    
2757
    if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
2758
                       constants.INSTANCE_REBOOT_HARD]:
2759
      result = self.rpc.call_instance_reboot(node_current, instance,
2760
                                             reboot_type, extra_args)
2761
      if result.failed or not result.data:
2762
        raise errors.OpExecError("Could not reboot instance")
2763
    else:
2764
      if not self.rpc.call_instance_shutdown(node_current, instance):
2765
        raise errors.OpExecError("could not shutdown instance for full reboot")
2766
      _ShutdownInstanceDisks(self, instance)
2767
      _StartInstanceDisks(self, instance, ignore_secondaries)
2768
      result = self.rpc.call_instance_start(node_current, instance, extra_args)
2769
      msg = result.RemoteFailMsg()
2770
      if msg:
2771
        _ShutdownInstanceDisks(self, instance)
2772
        raise errors.OpExecError("Could not start instance for"
2773
                                 " full reboot: %s" % msg)
2774

    
2775
    self.cfg.MarkInstanceUp(instance.name)
2776

    
2777

    
2778
class LUShutdownInstance(LogicalUnit):
2779
  """Shutdown an instance.
2780

2781
  """
2782
  HPATH = "instance-stop"
2783
  HTYPE = constants.HTYPE_INSTANCE
2784
  _OP_REQP = ["instance_name"]
2785
  REQ_BGL = False
2786

    
2787
  def ExpandNames(self):
2788
    self._ExpandAndLockInstance()
2789

    
2790
  def BuildHooksEnv(self):
2791
    """Build hooks env.
2792

2793
    This runs on master, primary and secondary nodes of the instance.
2794

2795
    """
2796
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2797
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2798
    return env, nl, nl
2799

    
2800
  def CheckPrereq(self):
2801
    """Check prerequisites.
2802

2803
    This checks that the instance is in the cluster.
2804

2805
    """
2806
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2807
    assert self.instance is not None, \
2808
      "Cannot retrieve locked instance %s" % self.op.instance_name
2809
    _CheckNodeOnline(self, self.instance.primary_node)
2810

    
2811
  def Exec(self, feedback_fn):
2812
    """Shutdown the instance.
2813

2814
    """
2815
    instance = self.instance
2816
    node_current = instance.primary_node
2817
    self.cfg.MarkInstanceDown(instance.name)
2818
    result = self.rpc.call_instance_shutdown(node_current, instance)
2819
    if result.failed or not result.data:
2820
      self.proc.LogWarning("Could not shutdown instance")
2821

    
2822
    _ShutdownInstanceDisks(self, instance)
2823

    
2824

    
2825
class LUReinstallInstance(LogicalUnit):
2826
  """Reinstall an instance.
2827

2828
  """
2829
  HPATH = "instance-reinstall"
2830
  HTYPE = constants.HTYPE_INSTANCE
2831
  _OP_REQP = ["instance_name"]
2832
  REQ_BGL = False
2833

    
2834
  def ExpandNames(self):
2835
    self._ExpandAndLockInstance()
2836

    
2837
  def BuildHooksEnv(self):
2838
    """Build hooks env.
2839

2840
    This runs on master, primary and secondary nodes of the instance.
2841

2842
    """
2843
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2844
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2845
    return env, nl, nl
2846

    
2847
  def CheckPrereq(self):
2848
    """Check prerequisites.
2849

2850
    This checks that the instance is in the cluster and is not running.
2851

2852
    """
2853
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2854
    assert instance is not None, \
2855
      "Cannot retrieve locked instance %s" % self.op.instance_name
2856
    _CheckNodeOnline(self, instance.primary_node)
2857

    
2858
    if instance.disk_template == constants.DT_DISKLESS:
2859
      raise errors.OpPrereqError("Instance '%s' has no disks" %
2860
                                 self.op.instance_name)
2861
    if instance.admin_up:
2862
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
2863
                                 self.op.instance_name)
2864
    remote_info = self.rpc.call_instance_info(instance.primary_node,
2865
                                              instance.name,
2866
                                              instance.hypervisor)
2867
    if remote_info.failed or remote_info.data:
2868
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
2869
                                 (self.op.instance_name,
2870
                                  instance.primary_node))
2871

    
2872
    self.op.os_type = getattr(self.op, "os_type", None)
2873
    if self.op.os_type is not None:
2874
      # OS verification
2875
      pnode = self.cfg.GetNodeInfo(
2876
        self.cfg.ExpandNodeName(instance.primary_node))
2877
      if pnode is None:
2878
        raise errors.OpPrereqError("Primary node '%s' is unknown" %
2879
                                   self.op.pnode)
2880
      result = self.rpc.call_os_get(pnode.name, self.op.os_type)
2881
      result.Raise()
2882
      if not isinstance(result.data, objects.OS):
2883
        raise errors.OpPrereqError("OS '%s' not in supported OS list for"
2884
                                   " primary node"  % self.op.os_type)
2885

    
2886
    self.instance = instance
2887

    
2888
  def Exec(self, feedback_fn):
2889
    """Reinstall the instance.
2890

2891
    """
2892
    inst = self.instance
2893

    
2894
    if self.op.os_type is not None:
2895
      feedback_fn("Changing OS to '%s'..." % self.op.os_type)
2896
      inst.os = self.op.os_type
2897
      self.cfg.Update(inst)
2898

    
2899
    _StartInstanceDisks(self, inst, None)
2900
    try:
2901
      feedback_fn("Running the instance OS create scripts...")
2902
      result = self.rpc.call_instance_os_add(inst.primary_node, inst)
2903
      msg = result.RemoteFailMsg()
2904
      if msg:
2905
        raise errors.OpExecError("Could not install OS for instance %s"
2906
                                 " on node %s: %s" %
2907
                                 (inst.name, inst.primary_node, msg))
2908
    finally:
2909
      _ShutdownInstanceDisks(self, inst)
2910

    
2911

    
2912
class LURenameInstance(LogicalUnit):
2913
  """Rename an instance.
2914

2915
  """
2916
  HPATH = "instance-rename"
2917
  HTYPE = constants.HTYPE_INSTANCE
2918
  _OP_REQP = ["instance_name", "new_name"]
2919

    
2920
  def BuildHooksEnv(self):
2921
    """Build hooks env.
2922

2923
    This runs on master, primary and secondary nodes of the instance.
2924

2925
    """
2926
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2927
    env["INSTANCE_NEW_NAME"] = self.op.new_name
2928
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2929
    return env, nl, nl
2930

    
2931
  def CheckPrereq(self):
2932
    """Check prerequisites.
2933

2934
    This checks that the instance is in the cluster and is not running.
2935

2936
    """
2937
    instance = self.cfg.GetInstanceInfo(
2938
      self.cfg.ExpandInstanceName(self.op.instance_name))
2939
    if instance is None:
2940
      raise errors.OpPrereqError("Instance '%s' not known" %
2941
                                 self.op.instance_name)
2942
    _CheckNodeOnline(self, instance.primary_node)
2943

    
2944
    if instance.admin_up:
2945
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
2946
                                 self.op.instance_name)
2947
    remote_info = self.rpc.call_instance_info(instance.primary_node,
2948
                                              instance.name,
2949
                                              instance.hypervisor)
2950
    remote_info.Raise()
2951
    if remote_info.data:
2952
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
2953
                                 (self.op.instance_name,
2954
                                  instance.primary_node))
2955
    self.instance = instance
2956

    
2957
    # new name verification
2958
    name_info = utils.HostInfo(self.op.new_name)
2959

    
2960
    self.op.new_name = new_name = name_info.name
2961
    instance_list = self.cfg.GetInstanceList()
2962
    if new_name in instance_list:
2963
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
2964
                                 new_name)
2965

    
2966
    if not getattr(self.op, "ignore_ip", False):
2967
      if utils.TcpPing(name_info.ip, constants.DEFAULT_NODED_PORT):
2968
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
2969
                                   (name_info.ip, new_name))
2970

    
2971

    
2972
  def Exec(self, feedback_fn):
2973
    """Reinstall the instance.
2974

2975
    """
2976
    inst = self.instance
2977
    old_name = inst.name
2978

    
2979
    if inst.disk_template == constants.DT_FILE:
2980
      old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
2981

    
2982
    self.cfg.RenameInstance(inst.name, self.op.new_name)
2983
    # Change the instance lock. This is definitely safe while we hold the BGL
2984
    self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
2985
    self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
2986

    
2987
    # re-read the instance from the configuration after rename
2988
    inst = self.cfg.GetInstanceInfo(self.op.new_name)
2989

    
2990
    if inst.disk_template == constants.DT_FILE:
2991
      new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
2992
      result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
2993
                                                     old_file_storage_dir,
2994
                                                     new_file_storage_dir)
2995
      result.Raise()
2996
      if not result.data:
2997
        raise errors.OpExecError("Could not connect to node '%s' to rename"
2998
                                 " directory '%s' to '%s' (but the instance"
2999
                                 " has been renamed in Ganeti)" % (
3000
                                 inst.primary_node, old_file_storage_dir,
3001
                                 new_file_storage_dir))
3002

    
3003
      if not result.data[0]:
3004
        raise errors.OpExecError("Could not rename directory '%s' to '%s'"
3005
                                 " (but the instance has been renamed in"
3006
                                 " Ganeti)" % (old_file_storage_dir,
3007
                                               new_file_storage_dir))
3008

    
3009
    _StartInstanceDisks(self, inst, None)
3010
    try:
3011
      result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
3012
                                                 old_name)
3013
      msg = result.RemoteFailMsg()
3014
      if msg:
3015
        msg = ("Could not run OS rename script for instance %s on node %s"
3016
               " (but the instance has been renamed in Ganeti): %s" %
3017
               (inst.name, inst.primary_node, msg))
3018
        self.proc.LogWarning(msg)
3019
    finally:
3020
      _ShutdownInstanceDisks(self, inst)
3021

    
3022

    
3023
class LURemoveInstance(LogicalUnit):
3024
  """Remove an instance.
3025

3026
  """
3027
  HPATH = "instance-remove"
3028
  HTYPE = constants.HTYPE_INSTANCE
3029
  _OP_REQP = ["instance_name", "ignore_failures"]
3030
  REQ_BGL = False
3031

    
3032
  def ExpandNames(self):
3033
    self._ExpandAndLockInstance()
3034
    self.needed_locks[locking.LEVEL_NODE] = []
3035
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3036

    
3037
  def DeclareLocks(self, level):
3038
    if level == locking.LEVEL_NODE:
3039
      self._LockInstancesNodes()
3040

    
3041
  def BuildHooksEnv(self):
3042
    """Build hooks env.
3043

3044
    This runs on master, primary and secondary nodes of the instance.
3045

3046
    """
3047
    env = _BuildInstanceHookEnvByObject(self, self.instance)
3048
    nl = [self.cfg.GetMasterNode()]
3049
    return env, nl, nl
3050

    
3051
  def CheckPrereq(self):
3052
    """Check prerequisites.
3053

3054
    This checks that the instance is in the cluster.
3055

3056
    """
3057
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3058
    assert self.instance is not None, \
3059
      "Cannot retrieve locked instance %s" % self.op.instance_name
3060

    
3061
  def Exec(self, feedback_fn):
3062
    """Remove the instance.
3063

3064
    """
3065
    instance = self.instance
3066
    logging.info("Shutting down instance %s on node %s",
3067
                 instance.name, instance.primary_node)
3068

    
3069
    result = self.rpc.call_instance_shutdown(instance.primary_node, instance)
3070
    if result.failed or not result.data:
3071
      if self.op.ignore_failures:
3072
        feedback_fn("Warning: can't shutdown instance")
3073
      else:
3074
        raise errors.OpExecError("Could not shutdown instance %s on node %s" %
3075
                                 (instance.name, instance.primary_node))
3076

    
3077
    logging.info("Removing block devices for instance %s", instance.name)
3078

    
3079
    if not _RemoveDisks(self, instance):
3080
      if self.op.ignore_failures:
3081
        feedback_fn("Warning: can't remove instance's disks")
3082
      else:
3083
        raise errors.OpExecError("Can't remove instance's disks")
3084

    
3085
    logging.info("Removing instance %s out of cluster config", instance.name)
3086

    
3087
    self.cfg.RemoveInstance(instance.name)
3088
    self.remove_locks[locking.LEVEL_INSTANCE] = instance.name
3089

    
3090

    
3091
class LUQueryInstances(NoHooksLU):
3092
  """Logical unit for querying instances.
3093

3094
  """
3095
  _OP_REQP = ["output_fields", "names", "use_locking"]
3096
  REQ_BGL = False
3097
  _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
3098
                                    "admin_state", "admin_ram",
3099
                                    "disk_template", "ip", "mac", "bridge",
3100
                                    "sda_size", "sdb_size", "vcpus", "tags",
3101
                                    "network_port", "beparams",
3102
                                    "(disk).(size)/([0-9]+)",
3103
                                    "(disk).(sizes)", "disk_usage",
3104
                                    "(nic).(mac|ip|bridge)/([0-9]+)",
3105
                                    "(nic).(macs|ips|bridges)",
3106
                                    "(disk|nic).(count)",
3107
                                    "serial_no", "hypervisor", "hvparams",] +
3108
                                  ["hv/%s" % name
3109
                                   for name in constants.HVS_PARAMETERS] +
3110
                                  ["be/%s" % name
3111
                                   for name in constants.BES_PARAMETERS])
3112
  _FIELDS_DYNAMIC = utils.FieldSet("oper_state", "oper_ram", "status")
3113

    
3114

    
3115
  def ExpandNames(self):
3116
    _CheckOutputFields(static=self._FIELDS_STATIC,
3117
                       dynamic=self._FIELDS_DYNAMIC,
3118
                       selected=self.op.output_fields)
3119

    
3120
    self.needed_locks = {}
3121
    self.share_locks[locking.LEVEL_INSTANCE] = 1
3122
    self.share_locks[locking.LEVEL_NODE] = 1
3123

    
3124
    if self.op.names:
3125
      self.wanted = _GetWantedInstances(self, self.op.names)
3126
    else:
3127
      self.wanted = locking.ALL_SET
3128

    
3129
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
3130
    self.do_locking = self.do_node_query and self.op.use_locking
3131
    if self.do_locking:
3132
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
3133
      self.needed_locks[locking.LEVEL_NODE] = []
3134
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3135

    
3136
  def DeclareLocks(self, level):
3137
    if level == locking.LEVEL_NODE and self.do_locking:
3138
      self._LockInstancesNodes()
3139

    
3140
  def CheckPrereq(self):
3141
    """Check prerequisites.
3142

3143
    """
3144
    pass
3145

    
3146
  def Exec(self, feedback_fn):
3147
    """Computes the list of nodes and their attributes.
3148

3149
    """
3150
    all_info = self.cfg.GetAllInstancesInfo()
3151
    if self.wanted == locking.ALL_SET:
3152
      # caller didn't specify instance names, so ordering is not important
3153
      if self.do_locking:
3154
        instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
3155
      else:
3156
        instance_names = all_info.keys()
3157
      instance_names = utils.NiceSort(instance_names)
3158
    else:
3159
      # caller did specify names, so we must keep the ordering
3160
      if self.do_locking:
3161
        tgt_set = self.acquired_locks[locking.LEVEL_INSTANCE]
3162
      else:
3163
        tgt_set = all_info.keys()
3164
      missing = set(self.wanted).difference(tgt_set)
3165
      if missing:
3166
        raise errors.OpExecError("Some instances were removed before"
3167
                                 " retrieving their data: %s" % missing)
3168
      instance_names = self.wanted
3169

    
3170
    instance_list = [all_info[iname] for iname in instance_names]
3171

    
3172
    # begin data gathering
3173

    
3174
    nodes = frozenset([inst.primary_node for inst in instance_list])
3175
    hv_list = list(set([inst.hypervisor for inst in instance_list]))
3176

    
3177
    bad_nodes = []
3178
    off_nodes = []
3179
    if self.do_node_query:
3180
      live_data = {}
3181
      node_data = self.rpc.call_all_instances_info(nodes, hv_list)
3182
      for name in nodes:
3183
        result = node_data[name]
3184
        if result.offline:
3185
          # offline nodes will be in both lists
3186
          off_nodes.append(name)
3187
        if result.failed:
3188
          bad_nodes.append(name)
3189
        else:
3190
          if result.data:
3191
            live_data.update(result.data)
3192
            # else no instance is alive
3193
    else:
3194
      live_data = dict([(name, {}) for name in instance_names])
3195

    
3196
    # end data gathering
3197

    
3198
    HVPREFIX = "hv/"
3199
    BEPREFIX = "be/"
3200
    output = []
3201
    for instance in instance_list:
3202
      iout = []
3203
      i_hv = self.cfg.GetClusterInfo().FillHV(instance)
3204
      i_be = self.cfg.GetClusterInfo().FillBE(instance)
3205
      for field in self.op.output_fields:
3206
        st_match = self._FIELDS_STATIC.Matches(field)
3207
        if field == "name":
3208
          val = instance.name
3209
        elif field == "os":
3210
          val = instance.os
3211
        elif field == "pnode":
3212
          val = instance.primary_node
3213
        elif field == "snodes":
3214
          val = list(instance.secondary_nodes)
3215
        elif field == "admin_state":
3216
          val = instance.admin_up
3217
        elif field == "oper_state":
3218
          if instance.primary_node in bad_nodes:
3219
            val = None
3220
          else:
3221
            val = bool(live_data.get(instance.name))
3222
        elif field == "status":
3223
          if instance.primary_node in off_nodes:
3224
            val = "ERROR_nodeoffline"
3225
          elif instance.primary_node in bad_nodes:
3226
            val = "ERROR_nodedown"
3227
          else:
3228
            running = bool(live_data.get(instance.name))
3229
            if running:
3230
              if instance.admin_up:
3231
                val = "running"
3232
              else:
3233
                val = "ERROR_up"
3234
            else:
3235
              if instance.admin_up:
3236
                val = "ERROR_down"
3237
              else:
3238
                val = "ADMIN_down"
3239
        elif field == "oper_ram":
3240
          if instance.primary_node in bad_nodes:
3241
            val = None
3242
          elif instance.name in live_data:
3243
            val = live_data[instance.name].get("memory", "?")
3244
          else:
3245
            val = "-"
3246
        elif field == "disk_template":
3247
          val = instance.disk_template
3248
        elif field == "ip":
3249
          val = instance.nics[0].ip
3250
        elif field == "bridge":
3251
          val = instance.nics[0].bridge
3252
        elif field == "mac":
3253
          val = instance.nics[0].mac
3254
        elif field == "sda_size" or field == "sdb_size":
3255
          idx = ord(field[2]) - ord('a')
3256
          try:
3257
            val = instance.FindDisk(idx).size
3258
          except errors.OpPrereqError:
3259
            val = None
3260
        elif field == "disk_usage": # total disk usage per node
3261
          disk_sizes = [{'size': disk.size} for disk in instance.disks]
3262
          val = _ComputeDiskSize(instance.disk_template, disk_sizes)
3263
        elif field == "tags":
3264
          val = list(instance.GetTags())
3265
        elif field == "serial_no":
3266
          val = instance.serial_no
3267
        elif field == "network_port":
3268
          val = instance.network_port
3269
        elif field == "hypervisor":
3270
          val = instance.hypervisor
3271
        elif field == "hvparams":
3272
          val = i_hv
3273
        elif (field.startswith(HVPREFIX) and
3274
              field[len(HVPREFIX):] in constants.HVS_PARAMETERS):
3275
          val = i_hv.get(field[len(HVPREFIX):], None)
3276
        elif field == "beparams":
3277
          val = i_be
3278
        elif (field.startswith(BEPREFIX) and
3279
              field[len(BEPREFIX):] in constants.BES_PARAMETERS):
3280
          val = i_be.get(field[len(BEPREFIX):], None)
3281
        elif st_match and st_match.groups():
3282
          # matches a variable list
3283
          st_groups = st_match.groups()
3284
          if st_groups and st_groups[0] == "disk":
3285
            if st_groups[1] == "count":
3286
              val = len(instance.disks)
3287
            elif st_groups[1] == "sizes":
3288
              val = [disk.size for disk in instance.disks]
3289
            elif st_groups[1] == "size":
3290
              try:
3291
                val = instance.FindDisk(st_groups[2]).size
3292
              except errors.OpPrereqError:
3293
                val = None
3294
            else:
3295
              assert False, "Unhandled disk parameter"
3296
          elif st_groups[0] == "nic":
3297
            if st_groups[1] == "count":
3298
              val = len(instance.nics)
3299
            elif st_groups[1] == "macs":
3300
              val = [nic.mac for nic in instance.nics]
3301
            elif st_groups[1] == "ips":
3302
              val = [nic.ip for nic in instance.nics]
3303
            elif st_groups[1] == "bridges":
3304
              val = [nic.bridge for nic in instance.nics]
3305
            else:
3306
              # index-based item
3307
              nic_idx = int(st_groups[2])
3308
              if nic_idx >= len(instance.nics):
3309
                val = None
3310
              else:
3311
                if st_groups[1] == "mac":
3312
                  val = instance.nics[nic_idx].mac
3313
                elif st_groups[1] == "ip":
3314
                  val = instance.nics[nic_idx].ip
3315
                elif st_groups[1] == "bridge":
3316
                  val = instance.nics[nic_idx].bridge
3317
                else:
3318
                  assert False, "Unhandled NIC parameter"
3319
          else:
3320
            assert False, "Unhandled variable parameter"
3321
        else:
3322
          raise errors.ParameterError(field)
3323
        iout.append(val)
3324
      output.append(iout)
3325

    
3326
    return output
3327

    
3328

    
3329
class LUFailoverInstance(LogicalUnit):
3330
  """Failover an instance.
3331

3332
  """
3333
  HPATH = "instance-failover"
3334
  HTYPE = constants.HTYPE_INSTANCE
3335
  _OP_REQP = ["instance_name", "ignore_consistency"]
3336
  REQ_BGL = False
3337

    
3338
  def ExpandNames(self):
3339
    self._ExpandAndLockInstance()
3340
    self.needed_locks[locking.LEVEL_NODE] = []
3341
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3342

    
3343
  def DeclareLocks(self, level):
3344
    if level == locking.LEVEL_NODE:
3345
      self._LockInstancesNodes()
3346

    
3347
  def BuildHooksEnv(self):
3348
    """Build hooks env.
3349

3350
    This runs on master, primary and secondary nodes of the instance.
3351

3352
    """
3353
    env = {
3354
      "IGNORE_CONSISTENCY": self.op.ignore_consistency,
3355
      }
3356
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
3357
    nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
3358
    return env, nl, nl
3359

    
3360
  def CheckPrereq(self):
3361
    """Check prerequisites.
3362

3363
    This checks that the instance is in the cluster.
3364

3365
    """
3366
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3367
    assert self.instance is not None, \
3368
      "Cannot retrieve locked instance %s" % self.op.instance_name
3369

    
3370
    bep = self.cfg.GetClusterInfo().FillBE(instance)
3371
    if instance.disk_template not in constants.DTS_NET_MIRROR:
3372
      raise errors.OpPrereqError("Instance's disk layout is not"
3373
                                 " network mirrored, cannot failover.")
3374

    
3375
    secondary_nodes = instance.secondary_nodes
3376
    if not secondary_nodes:
3377
      raise errors.ProgrammerError("no secondary node but using "
3378
                                   "a mirrored disk template")
3379

    
3380
    target_node = secondary_nodes[0]
3381
    _CheckNodeOnline(self, target_node)
3382
    # check memory requirements on the secondary node
3383
    _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
3384
                         instance.name, bep[constants.BE_MEMORY],
3385
                         instance.hypervisor)
3386

    
3387
    # check bridge existance
3388
    brlist = [nic.bridge for nic in instance.nics]
3389
    result = self.rpc.call_bridges_exist(target_node, brlist)
3390
    result.Raise()
3391
    if not result.data:
3392
      raise errors.OpPrereqError("One or more target bridges %s does not"
3393
                                 " exist on destination node '%s'" %
3394
                                 (brlist, target_node))
3395

    
3396
  def Exec(self, feedback_fn):
3397
    """Failover an instance.
3398

3399
    The failover is done by shutting it down on its present node and
3400
    starting it on the secondary.
3401

3402
    """
3403
    instance = self.instance
3404

    
3405
    source_node = instance.primary_node
3406
    target_node = instance.secondary_nodes[0]
3407

    
3408
    feedback_fn("* checking disk consistency between source and target")
3409
    for dev in instance.disks:
3410
      # for drbd, these are drbd over lvm
3411
      if not _CheckDiskConsistency(self, dev, target_node, False):
3412
        if instance.admin_up and not self.op.ignore_consistency:
3413
          raise errors.OpExecError("Disk %s is degraded on target node,"
3414
                                   " aborting failover." % dev.iv_name)
3415

    
3416
    feedback_fn("* shutting down instance on source node")
3417
    logging.info("Shutting down instance %s on node %s",
3418
                 instance.name, source_node)
3419

    
3420
    result = self.rpc.call_instance_shutdown(source_node, instance)
3421
    if result.failed or not result.data:
3422
      if self.op.ignore_consistency:
3423
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
3424
                             " Proceeding"
3425
                             " anyway. Please make sure node %s is down",
3426
                             instance.name, source_node, source_node)
3427
      else:
3428
        raise errors.OpExecError("Could not shutdown instance %s on node %s" %
3429
                                 (instance.name, source_node))
3430

    
3431
    feedback_fn("* deactivating the instance's disks on source node")
3432
    if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
3433
      raise errors.OpExecError("Can't shut down the instance's disks.")
3434

    
3435
    instance.primary_node = target_node
3436
    # distribute new instance config to the other nodes
3437
    self.cfg.Update(instance)
3438

    
3439
    # Only start the instance if it's marked as up
3440
    if instance.admin_up:
3441
      feedback_fn("* activating the instance's disks on target node")
3442
      logging.info("Starting instance %s on node %s",
3443
                   instance.name, target_node)
3444

    
3445
      disks_ok, dummy = _AssembleInstanceDisks(self, instance,
3446
                                               ignore_secondaries=True)
3447
      if not disks_ok:
3448
        _ShutdownInstanceDisks(self, instance)
3449
        raise errors.OpExecError("Can't activate the instance's disks")
3450

    
3451
      feedback_fn("* starting the instance on the target node")
3452
      result = self.rpc.call_instance_start(target_node, instance, None)
3453
      msg = result.RemoteFailMsg()
3454
      if msg:
3455
        _ShutdownInstanceDisks(self, instance)
3456
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
3457
                                 (instance.name, target_node, msg))
3458

    
3459

    
3460
class LUMigrateInstance(LogicalUnit):
3461
  """Migrate an instance.
3462

3463
  This is migration without shutting down, compared to the failover,
3464
  which is done with shutdown.
3465

3466
  """
3467
  HPATH = "instance-migrate"
3468
  HTYPE = constants.HTYPE_INSTANCE
3469
  _OP_REQP = ["instance_name", "live", "cleanup"]
3470

    
3471
  REQ_BGL = False
3472

    
3473
  def ExpandNames(self):
3474
    self._ExpandAndLockInstance()
3475
    self.needed_locks[locking.LEVEL_NODE] = []
3476
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3477

    
3478
  def DeclareLocks(self, level):
3479
    if level == locking.LEVEL_NODE:
3480
      self._LockInstancesNodes()
3481

    
3482
  def BuildHooksEnv(self):
3483
    """Build hooks env.
3484

3485
    This runs on master, primary and secondary nodes of the instance.
3486

3487
    """
3488
    env = _BuildInstanceHookEnvByObject(self, self.instance)
3489
    nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
3490
    return env, nl, nl
3491

    
3492
  def CheckPrereq(self):
3493
    """Check prerequisites.
3494

3495
    This checks that the instance is in the cluster.
3496

3497
    """
3498
    instance = self.cfg.GetInstanceInfo(
3499
      self.cfg.ExpandInstanceName(self.op.instance_name))
3500
    if instance is None:
3501
      raise errors.OpPrereqError("Instance '%s' not known" %
3502
                                 self.op.instance_name)
3503

    
3504
    if instance.disk_template != constants.DT_DRBD8:
3505
      raise errors.OpPrereqError("Instance's disk layout is not"
3506
                                 " drbd8, cannot migrate.")
3507

    
3508
    secondary_nodes = instance.secondary_nodes
3509
    if not secondary_nodes:
3510
      raise errors.ProgrammerError("no secondary node but using "
3511
                                   "drbd8 disk template")
3512

    
3513
    i_be = self.cfg.GetClusterInfo().FillBE(instance)
3514

    
3515
    target_node = secondary_nodes[0]
3516
    # check memory requirements on the secondary node
3517
    _CheckNodeFreeMemory(self, target_node, "migrating instance %s" %
3518
                         instance.name, i_be[constants.BE_MEMORY],
3519
                         instance.hypervisor)
3520

    
3521
    # check bridge existance
3522
    brlist = [nic.bridge for nic in instance.nics]
3523
    result = self.rpc.call_bridges_exist(target_node, brlist)
3524
    if result.failed or not result.data:
3525
      raise errors.OpPrereqError("One or more target bridges %s does not"
3526
                                 " exist on destination node '%s'" %
3527
                                 (brlist, target_node))
3528

    
3529
    if not self.op.cleanup:
3530
      result = self.rpc.call_instance_migratable(instance.primary_node,
3531
                                                 instance)
3532
      msg = result.RemoteFailMsg()
3533
      if msg:
3534
        raise errors.OpPrereqError("Can't migrate: %s - please use failover" %
3535
                                   msg)
3536

    
3537
    self.instance = instance
3538

    
3539
  def _WaitUntilSync(self):
3540
    """Poll with custom rpc for disk sync.
3541

3542
    This uses our own step-based rpc call.
3543

3544
    """
3545
    self.feedback_fn("* wait until resync is done")
3546
    all_done = False
3547
    while not all_done:
3548
      all_done = True
3549
      result = self.rpc.call_drbd_wait_sync(self.all_nodes,
3550
                                            self.nodes_ip,
3551
                                            self.instance.disks)
3552
      min_percent = 100
3553
      for node, nres in result.items():
3554
        msg = nres.RemoteFailMsg()
3555
        if msg:
3556
          raise errors.OpExecError("Cannot resync disks on node %s: %s" %
3557
                                   (node, msg))
3558
        node_done, node_percent = nres.payload
3559
        all_done = all_done and node_done
3560
        if node_percent is not None:
3561
          min_percent = min(min_percent, node_percent)
3562
      if not all_done:
3563
        if min_percent < 100:
3564
          self.feedback_fn("   - progress: %.1f%%" % min_percent)
3565
        time.sleep(2)
3566

    
3567
  def _EnsureSecondary(self, node):
3568
    """Demote a node to secondary.
3569

3570
    """
3571
    self.feedback_fn("* switching node %s to secondary mode" % node)
3572

    
3573
    for dev in self.instance.disks:
3574
      self.cfg.SetDiskID(dev, node)
3575

    
3576
    result = self.rpc.call_blockdev_close(node, self.instance.name,
3577
                                          self.instance.disks)
3578
    msg = result.RemoteFailMsg()
3579
    if msg:
3580
      raise errors.OpExecError("Cannot change disk to secondary on node %s,"
3581
                               " error %s" % (node, msg))
3582

    
3583
  def _GoStandalone(self):
3584
    """Disconnect from the network.
3585

3586
    """
3587
    self.feedback_fn("* changing into standalone mode")
3588
    result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
3589
                                               self.instance.disks)
3590
    for node, nres in result.items():
3591
      msg = nres.RemoteFailMsg()
3592
      if msg:
3593
        raise errors.OpExecError("Cannot disconnect disks node %s,"
3594
                                 " error %s" % (node, msg))
3595

    
3596
  def _GoReconnect(self, multimaster):
3597
    """Reconnect to the network.
3598

3599
    """
3600
    if multimaster:
3601
      msg = "dual-master"
3602
    else:
3603
      msg = "single-master"
3604
    self.feedback_fn("* changing disks into %s mode" % msg)
3605
    result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
3606
                                           self.instance.disks,
3607
                                           self.instance.name, multimaster)
3608
    for node, nres in result.items():
3609
      msg = nres.RemoteFailMsg()
3610
      if msg:
3611
        raise errors.OpExecError("Cannot change disks config on node %s,"
3612
                                 " error: %s" % (node, msg))
3613

    
3614
  def _ExecCleanup(self):
3615
    """Try to cleanup after a failed migration.
3616

3617
    The cleanup is done by:
3618
      - check that the instance is running only on one node
3619
        (and update the config if needed)
3620
      - change disks on its secondary node to secondary
3621
      - wait until disks are fully synchronized
3622
      - disconnect from the network
3623
      - change disks into single-master mode
3624
      - wait again until disks are fully synchronized
3625

3626
    """
3627
    instance = self.instance
3628
    target_node = self.target_node
3629
    source_node = self.source_node
3630

    
3631
    # check running on only one node
3632
    self.feedback_fn("* checking where the instance actually runs"
3633
                     " (if this hangs, the hypervisor might be in"
3634
                     " a bad state)")
3635
    ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
3636
    for node, result in ins_l.items():
3637
      result.Raise()
3638
      if not isinstance(result.data, list):
3639
        raise errors.OpExecError("Can't contact node '%s'" % node)
3640

    
3641
    runningon_source = instance.name in ins_l[source_node].data
3642
    runningon_target = instance.name in ins_l[target_node].data
3643

    
3644
    if runningon_source and runningon_target:
3645
      raise errors.OpExecError("Instance seems to be running on two nodes,"
3646
                               " or the hypervisor is confused. You will have"
3647
                               " to ensure manually that it runs only on one"
3648
                               " and restart this operation.")
3649

    
3650
    if not (runningon_source or runningon_target):
3651
      raise errors.OpExecError("Instance does not seem to be running at all."
3652
                               " In this case, it's safer to repair by"
3653
                               " running 'gnt-instance stop' to ensure disk"
3654
                               " shutdown, and then restarting it.")
3655

    
3656
    if runningon_target:
3657
      # the migration has actually succeeded, we need to update the config
3658
      self.feedback_fn("* instance running on secondary node (%s),"
3659
                       " updating config" % target_node)
3660
      instance.primary_node = target_node
3661
      self.cfg.Update(instance)
3662
      demoted_node = source_node
3663
    else:
3664
      self.feedback_fn("* instance confirmed to be running on its"
3665
                       " primary node (%s)" % source_node)
3666
      demoted_node = target_node
3667

    
3668
    self._EnsureSecondary(demoted_node)
3669
    try:
3670
      self._WaitUntilSync()
3671
    except errors.OpExecError:
3672
      # we ignore here errors, since if the device is standalone, it
3673
      # won't be able to sync
3674
      pass
3675
    self._GoStandalone()
3676
    self._GoReconnect(False)
3677
    self._WaitUntilSync()
3678

    
3679
    self.feedback_fn("* done")
3680

    
3681
  def _RevertDiskStatus(self):
3682
    """Try to revert the disk status after a failed migration.
3683

3684
    """
3685
    target_node = self.target_node
3686
    try:
3687
      self._EnsureSecondary(target_node)
3688
      self._GoStandalone()
3689
      self._GoReconnect(False)
3690
      self._WaitUntilSync()
3691
    except errors.OpExecError, err:
3692
      self.LogWarning("Migration failed and I can't reconnect the"
3693
                      " drives: error '%s'\n"
3694
                      "Please look and recover the instance status" %
3695
                      str(err))
3696

    
3697
  def _AbortMigration(self):
3698
    """Call the hypervisor code to abort a started migration.
3699

3700
    """
3701
    instance = self.instance
3702
    target_node = self.target_node
3703
    migration_info = self.migration_info
3704

    
3705
    abort_result = self.rpc.call_finalize_migration(target_node,
3706
                                                    instance,
3707
                                                    migration_info,
3708
                                                    False)
3709
    abort_msg = abort_result.RemoteFailMsg()
3710
    if abort_msg:
3711
      logging.error("Aborting migration failed on target node %s: %s" %
3712
                    (target_node, abort_msg))
3713
      # Don't raise an exception here, as we stil have to try to revert the
3714
      # disk status, even if this step failed.
3715

    
3716
  def _ExecMigration(self):
3717
    """Migrate an instance.
3718

3719
    The migrate is done by:
3720
      - change the disks into dual-master mode
3721
      - wait until disks are fully synchronized again
3722
      - migrate the instance
3723
      - change disks on the new secondary node (the old primary) to secondary
3724
      - wait until disks are fully synchronized
3725
      - change disks into single-master mode
3726

3727
    """
3728
    instance = self.instance
3729
    target_node = self.target_node
3730
    source_node = self.source_node
3731

    
3732
    self.feedback_fn("* checking disk consistency between source and target")
3733
    for dev in instance.disks:
3734
      if not _CheckDiskConsistency(self, dev, target_node, False):
3735
        raise errors.OpExecError("Disk %s is degraded or not fully"
3736
                                 " synchronized on target node,"
3737
                                 " aborting migrate." % dev.iv_name)
3738

    
3739
    # First get the migration information from the remote node
3740
    result = self.rpc.call_migration_info(source_node, instance)
3741
    msg = result.RemoteFailMsg()
3742
    if msg:
3743
      log_err = ("Failed fetching source migration information from %s: %s" %
3744
                 (source_node, msg))
3745
      logging.error(log_err)
3746
      raise errors.OpExecError(log_err)
3747

    
3748
    self.migration_info = migration_info = result.payload
3749

    
3750
    # Then switch the disks to master/master mode
3751
    self._EnsureSecondary(target_node)
3752
    self._GoStandalone()
3753
    self._GoReconnect(True)
3754
    self._WaitUntilSync()
3755

    
3756
    self.feedback_fn("* preparing %s to accept the instance" % target_node)
3757
    result = self.rpc.call_accept_instance(target_node,
3758
                                           instance,
3759
                                           migration_info,
3760
                                           self.nodes_ip[target_node])
3761

    
3762
    msg = result.RemoteFailMsg()
3763
    if msg:
3764
      logging.error("Instance pre-migration failed, trying to revert"
3765
                    " disk status: %s", msg)
3766
      self._AbortMigration()
3767
      self._RevertDiskStatus()
3768
      raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
3769
                               (instance.name, msg))
3770

    
3771
    self.feedback_fn("* migrating instance to %s" % target_node)
3772
    time.sleep(10)
3773
    result = self.rpc.call_instance_migrate(source_node, instance,
3774
                                            self.nodes_ip[target_node],
3775
                                            self.op.live)
3776
    msg = result.RemoteFailMsg()
3777
    if msg:
3778
      logging.error("Instance migration failed, trying to revert"
3779
                    " disk status: %s", msg)
3780
      self._AbortMigration()
3781
      self._RevertDiskStatus()
3782
      raise errors.OpExecError("Could not migrate instance %s: %s" %
3783
                               (instance.name, msg))
3784
    time.sleep(10)
3785

    
3786
    instance.primary_node = target_node
3787
    # distribute new instance config to the other nodes
3788
    self.cfg.Update(instance)
3789

    
3790
    result = self.rpc.call_finalize_migration(target_node,
3791
                                              instance,
3792
                                              migration_info,
3793
                                              True)
3794
    msg = result.RemoteFailMsg()
3795
    if msg:
3796
      logging.error("Instance migration succeeded, but finalization failed:"
3797
                    " %s" % msg)
3798
      raise errors.OpExecError("Could not finalize instance migration: %s" %
3799
                               msg)
3800

    
3801
    self._EnsureSecondary(source_node)
3802
    self._WaitUntilSync()
3803
    self._GoStandalone()
3804
    self._GoReconnect(False)
3805
    self._WaitUntilSync()
3806

    
3807
    self.feedback_fn("* done")
3808

    
3809
  def Exec(self, feedback_fn):
3810
    """Perform the migration.
3811

3812
    """
3813
    self.feedback_fn = feedback_fn
3814

    
3815
    self.source_node = self.instance.primary_node
3816
    self.target_node = self.instance.secondary_nodes[0]
3817
    self.all_nodes = [self.source_node, self.target_node]
3818
    self.nodes_ip = {
3819
      self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
3820
      self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
3821
      }
3822
    if self.op.cleanup:
3823
      return self._ExecCleanup()
3824
    else:
3825
      return self._ExecMigration()
3826

    
3827

    
3828
def _CreateBlockDev(lu, node, instance, device, force_create,
3829
                    info, force_open):
3830
  """Create a tree of block devices on a given node.
3831

3832
  If this device type has to be created on secondaries, create it and
3833
  all its children.
3834

3835
  If not, just recurse to children keeping the same 'force' value.
3836

3837
  @param lu: the lu on whose behalf we execute
3838
  @param node: the node on which to create the device
3839
  @type instance: L{objects.Instance}
3840
  @param instance: the instance which owns the device
3841
  @type device: L{objects.Disk}
3842
  @param device: the device to create
3843
  @type force_create: boolean
3844
  @param force_create: whether to force creation of this device; this
3845
      will be change to True whenever we find a device which has
3846
      CreateOnSecondary() attribute
3847
  @param info: the extra 'metadata' we should attach to the device
3848
      (this will be represented as a LVM tag)
3849
  @type force_open: boolean
3850
  @param force_open: this parameter will be passes to the
3851
      L{backend.BlockdevCreate} function where it specifies
3852
      whether we run on primary or not, and it affects both
3853
      the child assembly and the device own Open() execution
3854

3855
  """
3856
  if device.CreateOnSecondary():
3857
    force_create = True
3858

    
3859
  if device.children:
3860
    for child in device.children:
3861
      _CreateBlockDev(lu, node, instance, child, force_create,
3862
                      info, force_open)
3863

    
3864
  if not force_create:
3865
    return
3866

    
3867
  _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
3868

    
3869

    
3870
def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
3871
  """Create a single block device on a given node.
3872

3873
  This will not recurse over children of the device, so they must be
3874
  created in advance.
3875

3876
  @param lu: the lu on whose behalf we execute
3877
  @param node: the node on which to create the device
3878
  @type instance: L{objects.Instance}
3879
  @param instance: the instance which owns the device
3880
  @type device: L{objects.Disk}
3881
  @param device: the device to create
3882
  @param info: the extra 'metadata' we should attach to the device
3883
      (this will be represented as a LVM tag)
3884
  @type force_open: boolean
3885
  @param force_open: this parameter will be passes to the
3886
      L{backend.BlockdevCreate} function where it specifies
3887
      whether we run on primary or not, and it affects both
3888
      the child assembly and the device own Open() execution
3889

3890
  """
3891
  lu.cfg.SetDiskID(device, node)
3892
  result = lu.rpc.call_blockdev_create(node, device, device.size,
3893
                                       instance.name, force_open, info)
3894
  msg = result.RemoteFailMsg()
3895
  if msg:
3896
    raise errors.OpExecError("Can't create block device %s on"
3897
                             " node %s for instance %s: %s" %
3898
                             (device, node, instance.name, msg))
3899
  if device.physical_id is None:
3900
    device.physical_id = result.payload
3901

    
3902

    
3903
def _GenerateUniqueNames(lu, exts):
3904
  """Generate a suitable LV name.
3905

3906
  This will generate a logical volume name for the given instance.
3907

3908
  """
3909
  results = []
3910
  for val in exts:
3911
    new_id = lu.cfg.GenerateUniqueID()
3912
    results.append("%s%s" % (new_id, val))
3913
  return results
3914

    
3915

    
3916
def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
3917
                         p_minor, s_minor):
3918
  """Generate a drbd8 device complete with its children.
3919

3920
  """
3921
  port = lu.cfg.AllocatePort()
3922
  vgname = lu.cfg.GetVGName()
3923
  shared_secret = lu.cfg.GenerateDRBDSecret()
3924
  dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
3925
                          logical_id=(vgname, names[0]))
3926
  dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
3927
                          logical_id=(vgname, names[1]))
3928
  drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
3929
                          logical_id=(primary, secondary, port,
3930
                                      p_minor, s_minor,
3931
                                      shared_secret),
3932
                          children=[dev_data, dev_meta],
3933
                          iv_name=iv_name)
3934
  return drbd_dev
3935

    
3936

    
3937
def _GenerateDiskTemplate(lu, template_name,
3938
                          instance_name, primary_node,
3939
                          secondary_nodes, disk_info,
3940
                          file_storage_dir, file_driver,
3941
                          base_index):
3942
  """Generate the entire disk layout for a given template type.
3943

3944
  """
3945
  #TODO: compute space requirements
3946

    
3947
  vgname = lu.cfg.GetVGName()
3948
  disk_count = len(disk_info)
3949
  disks = []
3950
  if template_name == constants.DT_DISKLESS:
3951
    pass
3952
  elif template_name == constants.DT_PLAIN:
3953
    if len(secondary_nodes) != 0:
3954
      raise errors.ProgrammerError("Wrong template configuration")
3955

    
3956
    names = _GenerateUniqueNames(lu, [".disk%d" % i
3957
                                      for i in range(disk_count)])
3958
    for idx, disk in enumerate(disk_info):
3959
      disk_index = idx + base_index
3960
      disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
3961
                              logical_id=(vgname, names[idx]),
3962
                              iv_name="disk/%d" % disk_index,
3963
                              mode=disk["mode"])
3964
      disks.append(disk_dev)
3965
  elif template_name == constants.DT_DRBD8:
3966
    if len(secondary_nodes) != 1:
3967
      raise errors.ProgrammerError("Wrong template configuration")
3968
    remote_node = secondary_nodes[0]
3969
    minors = lu.cfg.AllocateDRBDMinor(
3970
      [primary_node, remote_node] * len(disk_info), instance_name)
3971

    
3972
    names = []
3973
    for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % i
3974
                                               for i in range(disk_count)]):
3975
      names.append(lv_prefix + "_data")
3976
      names.append(lv_prefix + "_meta")
3977
    for idx, disk in enumerate(disk_info):
3978
      disk_index = idx + base_index
3979
      disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
3980
                                      disk["size"], names[idx*2:idx*2+2],
3981
                                      "disk/%d" % disk_index,
3982
                                      minors[idx*2], minors[idx*2+1])
3983
      disk_dev.mode = disk["mode"]
3984
      disks.append(disk_dev)
3985
  elif template_name == constants.DT_FILE:
3986
    if len(secondary_nodes) != 0:
3987
      raise errors.ProgrammerError("Wrong template configuration")
3988

    
3989
    for idx, disk in enumerate(disk_info):
3990
      disk_index = idx + base_index
3991
      disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
3992
                              iv_name="disk/%d" % disk_index,
3993
                              logical_id=(file_driver,
3994
                                          "%s/disk%d" % (file_storage_dir,
3995
                                                         idx)),
3996
                              mode=disk["mode"])
3997
      disks.append(disk_dev)
3998
  else:
3999
    raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
4000
  return disks
4001

    
4002

    
4003
def _GetInstanceInfoText(instance):
4004
  """Compute that text that should be added to the disk's metadata.
4005

4006
  """
4007
  return "originstname+%s" % instance.name
4008

    
4009

    
4010
def _CreateDisks(lu, instance):
4011
  """Create all disks for an instance.
4012

4013
  This abstracts away some work from AddInstance.
4014

4015
  @type lu: L{LogicalUnit}