Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ 0105bad3

History | View | Annotate | Download (233 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
    node_volume = {}
907
    node_instance = {}
908
    node_info = {}
909
    instance_cfg = {}
910

    
911
    # FIXME: verify OS list
912
    # do local checksums
913
    master_files = [constants.CLUSTER_CONF_FILE]
914

    
915
    file_names = ssconf.SimpleStore().GetFileList()
916
    file_names.append(constants.SSL_CERT_FILE)
917
    file_names.append(constants.RAPI_CERT_FILE)
918
    file_names.extend(master_files)
919

    
920
    local_checksums = utils.FingerprintFiles(file_names)
921

    
922
    feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
923
    node_verify_param = {
924
      constants.NV_FILELIST: file_names,
925
      constants.NV_NODELIST: [node.name for node in nodeinfo
926
                              if not node.offline],
927
      constants.NV_HYPERVISOR: hypervisors,
928
      constants.NV_NODENETTEST: [(node.name, node.primary_ip,
929
                                  node.secondary_ip) for node in nodeinfo
930
                                 if not node.offline],
931
      constants.NV_LVLIST: vg_name,
932
      constants.NV_INSTANCELIST: hypervisors,
933
      constants.NV_VGLIST: None,
934
      constants.NV_VERSION: None,
935
      constants.NV_HVINFO: self.cfg.GetHypervisorType(),
936
      constants.NV_DRBDLIST: None,
937
      }
938
    all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
939
                                           self.cfg.GetClusterName())
940

    
941
    cluster = self.cfg.GetClusterInfo()
942
    master_node = self.cfg.GetMasterNode()
943
    all_drbd_map = self.cfg.ComputeDRBDMap()
944

    
945
    for node_i in nodeinfo:
946
      node = node_i.name
947
      nresult = all_nvinfo[node].data
948

    
949
      if node_i.offline:
950
        feedback_fn("* Skipping offline node %s" % (node,))
951
        n_offline.append(node)
952
        continue
953

    
954
      if node == master_node:
955
        ntype = "master"
956
      elif node_i.master_candidate:
957
        ntype = "master candidate"
958
      else:
959
        ntype = "regular"
960
      feedback_fn("* Verifying node %s (%s)" % (node, ntype))
961

    
962
      if all_nvinfo[node].failed or not isinstance(nresult, dict):
963
        feedback_fn("  - ERROR: connection to %s failed" % (node,))
964
        bad = True
965
        continue
966

    
967
      node_drbd = {}
968
      for minor, instance in all_drbd_map[node].items():
969
        instance = instanceinfo[instance]
970
        node_drbd[minor] = (instance.name, instance.admin_up)
971
      result = self._VerifyNode(node_i, file_names, local_checksums,
972
                                nresult, feedback_fn, master_files,
973
                                node_drbd)
974
      bad = bad or result
975

    
976
      lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
977
      if isinstance(lvdata, basestring):
978
        feedback_fn("  - ERROR: LVM problem on node %s: %s" %
979
                    (node, utils.SafeEncode(lvdata)))
980
        bad = True
981
        node_volume[node] = {}
982
      elif not isinstance(lvdata, dict):
983
        feedback_fn("  - ERROR: connection to %s failed (lvlist)" % (node,))
984
        bad = True
985
        continue
986
      else:
987
        node_volume[node] = lvdata
988

    
989
      # node_instance
990
      idata = nresult.get(constants.NV_INSTANCELIST, None)
991
      if not isinstance(idata, list):
992
        feedback_fn("  - ERROR: connection to %s failed (instancelist)" %
993
                    (node,))
994
        bad = True
995
        continue
996

    
997
      node_instance[node] = idata
998

    
999
      # node_info
1000
      nodeinfo = nresult.get(constants.NV_HVINFO, None)
1001
      if not isinstance(nodeinfo, dict):
1002
        feedback_fn("  - ERROR: connection to %s failed (hvinfo)" % (node,))
1003
        bad = True
1004
        continue
1005

    
1006
      try:
1007
        node_info[node] = {
1008
          "mfree": int(nodeinfo['memory_free']),
1009
          "dfree": int(nresult[constants.NV_VGLIST][vg_name]),
1010
          "pinst": [],
1011
          "sinst": [],
1012
          # dictionary holding all instances this node is secondary for,
1013
          # grouped by their primary node. Each key is a cluster node, and each
1014
          # value is a list of instances which have the key as primary and the
1015
          # current node as secondary.  this is handy to calculate N+1 memory
1016
          # availability if you can only failover from a primary to its
1017
          # secondary.
1018
          "sinst-by-pnode": {},
1019
        }
1020
      except ValueError:
1021
        feedback_fn("  - ERROR: invalid value returned from node %s" % (node,))
1022
        bad = True
1023
        continue
1024

    
1025
    node_vol_should = {}
1026

    
1027
    for instance in instancelist:
1028
      feedback_fn("* Verifying instance %s" % instance)
1029
      inst_config = instanceinfo[instance]
1030
      result =  self._VerifyInstance(instance, inst_config, node_volume,
1031
                                     node_instance, feedback_fn, n_offline)
1032
      bad = bad or result
1033
      inst_nodes_offline = []
1034

    
1035
      inst_config.MapLVsByNode(node_vol_should)
1036

    
1037
      instance_cfg[instance] = inst_config
1038

    
1039
      pnode = inst_config.primary_node
1040
      if pnode in node_info:
1041
        node_info[pnode]['pinst'].append(instance)
1042
      elif pnode not in n_offline:
1043
        feedback_fn("  - ERROR: instance %s, connection to primary node"
1044
                    " %s failed" % (instance, pnode))
1045
        bad = True
1046

    
1047
      if pnode in n_offline:
1048
        inst_nodes_offline.append(pnode)
1049

    
1050
      # If the instance is non-redundant we cannot survive losing its primary
1051
      # node, so we are not N+1 compliant. On the other hand we have no disk
1052
      # templates with more than one secondary so that situation is not well
1053
      # supported either.
1054
      # FIXME: does not support file-backed instances
1055
      if len(inst_config.secondary_nodes) == 0:
1056
        i_non_redundant.append(instance)
1057
      elif len(inst_config.secondary_nodes) > 1:
1058
        feedback_fn("  - WARNING: multiple secondaries for instance %s"
1059
                    % instance)
1060

    
1061
      if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
1062
        i_non_a_balanced.append(instance)
1063

    
1064
      for snode in inst_config.secondary_nodes:
1065
        if snode in node_info:
1066
          node_info[snode]['sinst'].append(instance)
1067
          if pnode not in node_info[snode]['sinst-by-pnode']:
1068
            node_info[snode]['sinst-by-pnode'][pnode] = []
1069
          node_info[snode]['sinst-by-pnode'][pnode].append(instance)
1070
        elif snode not in n_offline:
1071
          feedback_fn("  - ERROR: instance %s, connection to secondary node"
1072
                      " %s failed" % (instance, snode))
1073
          bad = True
1074
        if snode in n_offline:
1075
          inst_nodes_offline.append(snode)
1076

    
1077
      if inst_nodes_offline:
1078
        # warn that the instance lives on offline nodes, and set bad=True
1079
        feedback_fn("  - ERROR: instance lives on offline node(s) %s" %
1080
                    ", ".join(inst_nodes_offline))
1081
        bad = True
1082

    
1083
    feedback_fn("* Verifying orphan volumes")
1084
    result = self._VerifyOrphanVolumes(node_vol_should, node_volume,
1085
                                       feedback_fn)
1086
    bad = bad or result
1087

    
1088
    feedback_fn("* Verifying remaining instances")
1089
    result = self._VerifyOrphanInstances(instancelist, node_instance,
1090
                                         feedback_fn)
1091
    bad = bad or result
1092

    
1093
    if constants.VERIFY_NPLUSONE_MEM not in self.skip_set:
1094
      feedback_fn("* Verifying N+1 Memory redundancy")
1095
      result = self._VerifyNPlusOneMemory(node_info, instance_cfg, feedback_fn)
1096
      bad = bad or result
1097

    
1098
    feedback_fn("* Other Notes")
1099
    if i_non_redundant:
1100
      feedback_fn("  - NOTICE: %d non-redundant instance(s) found."
1101
                  % len(i_non_redundant))
1102

    
1103
    if i_non_a_balanced:
1104
      feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
1105
                  % len(i_non_a_balanced))
1106

    
1107
    if n_offline:
1108
      feedback_fn("  - NOTICE: %d offline node(s) found." % len(n_offline))
1109

    
1110
    return not bad
1111

    
1112
  def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
1113
    """Analize the post-hooks' result
1114

1115
    This method analyses the hook result, handles it, and sends some
1116
    nicely-formatted feedback back to the user.
1117

1118
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
1119
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
1120
    @param hooks_results: the results of the multi-node hooks rpc call
1121
    @param feedback_fn: function used send feedback back to the caller
1122
    @param lu_result: previous Exec result
1123
    @return: the new Exec result, based on the previous result
1124
        and hook results
1125

1126
    """
1127
    # We only really run POST phase hooks, and are only interested in
1128
    # their results
1129
    if phase == constants.HOOKS_PHASE_POST:
1130
      # Used to change hooks' output to proper indentation
1131
      indent_re = re.compile('^', re.M)
1132
      feedback_fn("* Hooks Results")
1133
      if not hooks_results:
1134
        feedback_fn("  - ERROR: general communication failure")
1135
        lu_result = 1
1136
      else:
1137
        for node_name in hooks_results:
1138
          show_node_header = True
1139
          res = hooks_results[node_name]
1140
          if res.failed or res.data is False or not isinstance(res.data, list):
1141
            if res.offline:
1142
              # no need to warn or set fail return value
1143
              continue
1144
            feedback_fn("    Communication failure in hooks execution")
1145
            lu_result = 1
1146
            continue
1147
          for script, hkr, output in res.data:
1148
            if hkr == constants.HKR_FAIL:
1149
              # The node header is only shown once, if there are
1150
              # failing hooks on that node
1151
              if show_node_header:
1152
                feedback_fn("  Node %s:" % node_name)
1153
                show_node_header = False
1154
              feedback_fn("    ERROR: Script %s failed, output:" % script)
1155
              output = indent_re.sub('      ', output)
1156
              feedback_fn("%s" % output)
1157
              lu_result = 1
1158

    
1159
      return lu_result
1160

    
1161

    
1162
class LUVerifyDisks(NoHooksLU):
1163
  """Verifies the cluster disks status.
1164

1165
  """
1166
  _OP_REQP = []
1167
  REQ_BGL = False
1168

    
1169
  def ExpandNames(self):
1170
    self.needed_locks = {
1171
      locking.LEVEL_NODE: locking.ALL_SET,
1172
      locking.LEVEL_INSTANCE: locking.ALL_SET,
1173
    }
1174
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
1175

    
1176
  def CheckPrereq(self):
1177
    """Check prerequisites.
1178

1179
    This has no prerequisites.
1180

1181
    """
1182
    pass
1183

    
1184
  def Exec(self, feedback_fn):
1185
    """Verify integrity of cluster disks.
1186

1187
    """
1188
    result = res_nodes, res_nlvm, res_instances, res_missing = [], {}, [], {}
1189

    
1190
    vg_name = self.cfg.GetVGName()
1191
    nodes = utils.NiceSort(self.cfg.GetNodeList())
1192
    instances = [self.cfg.GetInstanceInfo(name)
1193
                 for name in self.cfg.GetInstanceList()]
1194

    
1195
    nv_dict = {}
1196
    for inst in instances:
1197
      inst_lvs = {}
1198
      if (not inst.admin_up or
1199
          inst.disk_template not in constants.DTS_NET_MIRROR):
1200
        continue
1201
      inst.MapLVsByNode(inst_lvs)
1202
      # transform { iname: {node: [vol,],},} to {(node, vol): iname}
1203
      for node, vol_list in inst_lvs.iteritems():
1204
        for vol in vol_list:
1205
          nv_dict[(node, vol)] = inst
1206

    
1207
    if not nv_dict:
1208
      return result
1209

    
1210
    node_lvs = self.rpc.call_volume_list(nodes, vg_name)
1211

    
1212
    to_act = set()
1213
    for node in nodes:
1214
      # node_volume
1215
      lvs = node_lvs[node]
1216
      if lvs.failed:
1217
        if not lvs.offline:
1218
          self.LogWarning("Connection to node %s failed: %s" %
1219
                          (node, lvs.data))
1220
        continue
1221
      lvs = lvs.data
1222
      if isinstance(lvs, basestring):
1223
        logging.warning("Error enumerating LVs on node %s: %s", node, lvs)
1224
        res_nlvm[node] = lvs
1225
      elif not isinstance(lvs, dict):
1226
        logging.warning("Connection to node %s failed or invalid data"
1227
                        " returned", node)
1228
        res_nodes.append(node)
1229
        continue
1230

    
1231
      for lv_name, (_, lv_inactive, lv_online) in lvs.iteritems():
1232
        inst = nv_dict.pop((node, lv_name), None)
1233
        if (not lv_online and inst is not None
1234
            and inst.name not in res_instances):
1235
          res_instances.append(inst.name)
1236

    
1237
    # any leftover items in nv_dict are missing LVs, let's arrange the
1238
    # data better
1239
    for key, inst in nv_dict.iteritems():
1240
      if inst.name not in res_missing:
1241
        res_missing[inst.name] = []
1242
      res_missing[inst.name].append(key)
1243

    
1244
    return result
1245

    
1246

    
1247
class LURenameCluster(LogicalUnit):
1248
  """Rename the cluster.
1249

1250
  """
1251
  HPATH = "cluster-rename"
1252
  HTYPE = constants.HTYPE_CLUSTER
1253
  _OP_REQP = ["name"]
1254

    
1255
  def BuildHooksEnv(self):
1256
    """Build hooks env.
1257

1258
    """
1259
    env = {
1260
      "OP_TARGET": self.cfg.GetClusterName(),
1261
      "NEW_NAME": self.op.name,
1262
      }
1263
    mn = self.cfg.GetMasterNode()
1264
    return env, [mn], [mn]
1265

    
1266
  def CheckPrereq(self):
1267
    """Verify that the passed name is a valid one.
1268

1269
    """
1270
    hostname = utils.HostInfo(self.op.name)
1271

    
1272
    new_name = hostname.name
1273
    self.ip = new_ip = hostname.ip
1274
    old_name = self.cfg.GetClusterName()
1275
    old_ip = self.cfg.GetMasterIP()
1276
    if new_name == old_name and new_ip == old_ip:
1277
      raise errors.OpPrereqError("Neither the name nor the IP address of the"
1278
                                 " cluster has changed")
1279
    if new_ip != old_ip:
1280
      if utils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
1281
        raise errors.OpPrereqError("The given cluster IP address (%s) is"
1282
                                   " reachable on the network. Aborting." %
1283
                                   new_ip)
1284

    
1285
    self.op.name = new_name
1286

    
1287
  def Exec(self, feedback_fn):
1288
    """Rename the cluster.
1289

1290
    """
1291
    clustername = self.op.name
1292
    ip = self.ip
1293

    
1294
    # shutdown the master IP
1295
    master = self.cfg.GetMasterNode()
1296
    result = self.rpc.call_node_stop_master(master, False)
1297
    if result.failed or not result.data:
1298
      raise errors.OpExecError("Could not disable the master role")
1299

    
1300
    try:
1301
      cluster = self.cfg.GetClusterInfo()
1302
      cluster.cluster_name = clustername
1303
      cluster.master_ip = ip
1304
      self.cfg.Update(cluster)
1305

    
1306
      # update the known hosts file
1307
      ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
1308
      node_list = self.cfg.GetNodeList()
1309
      try:
1310
        node_list.remove(master)
1311
      except ValueError:
1312
        pass
1313
      result = self.rpc.call_upload_file(node_list,
1314
                                         constants.SSH_KNOWN_HOSTS_FILE)
1315
      for to_node, to_result in result.iteritems():
1316
        if to_result.failed or not to_result.data:
1317
          logging.error("Copy of file %s to node %s failed",
1318
                        constants.SSH_KNOWN_HOSTS_FILE, to_node)
1319

    
1320
    finally:
1321
      result = self.rpc.call_node_start_master(master, False)
1322
      if result.failed or not result.data:
1323
        self.LogWarning("Could not re-enable the master role on"
1324
                        " the master, please restart manually.")
1325

    
1326

    
1327
def _RecursiveCheckIfLVMBased(disk):
1328
  """Check if the given disk or its children are lvm-based.
1329

1330
  @type disk: L{objects.Disk}
1331
  @param disk: the disk to check
1332
  @rtype: booleean
1333
  @return: boolean indicating whether a LD_LV dev_type was found or not
1334

1335
  """
1336
  if disk.children:
1337
    for chdisk in disk.children:
1338
      if _RecursiveCheckIfLVMBased(chdisk):
1339
        return True
1340
  return disk.dev_type == constants.LD_LV
1341

    
1342

    
1343
class LUSetClusterParams(LogicalUnit):
1344
  """Change the parameters of the cluster.
1345

1346
  """
1347
  HPATH = "cluster-modify"
1348
  HTYPE = constants.HTYPE_CLUSTER
1349
  _OP_REQP = []
1350
  REQ_BGL = False
1351

    
1352
  def CheckParameters(self):
1353
    """Check parameters
1354

1355
    """
1356
    if not hasattr(self.op, "candidate_pool_size"):
1357
      self.op.candidate_pool_size = None
1358
    if self.op.candidate_pool_size is not None:
1359
      try:
1360
        self.op.candidate_pool_size = int(self.op.candidate_pool_size)
1361
      except ValueError, err:
1362
        raise errors.OpPrereqError("Invalid candidate_pool_size value: %s" %
1363
                                   str(err))
1364
      if self.op.candidate_pool_size < 1:
1365
        raise errors.OpPrereqError("At least one master candidate needed")
1366

    
1367
  def ExpandNames(self):
1368
    # FIXME: in the future maybe other cluster params won't require checking on
1369
    # all nodes to be modified.
1370
    self.needed_locks = {
1371
      locking.LEVEL_NODE: locking.ALL_SET,
1372
    }
1373
    self.share_locks[locking.LEVEL_NODE] = 1
1374

    
1375
  def BuildHooksEnv(self):
1376
    """Build hooks env.
1377

1378
    """
1379
    env = {
1380
      "OP_TARGET": self.cfg.GetClusterName(),
1381
      "NEW_VG_NAME": self.op.vg_name,
1382
      }
1383
    mn = self.cfg.GetMasterNode()
1384
    return env, [mn], [mn]
1385

    
1386
  def CheckPrereq(self):
1387
    """Check prerequisites.
1388

1389
    This checks whether the given params don't conflict and
1390
    if the given volume group is valid.
1391

1392
    """
1393
    # FIXME: This only works because there is only one parameter that can be
1394
    # changed or removed.
1395
    if self.op.vg_name is not None and not self.op.vg_name:
1396
      instances = self.cfg.GetAllInstancesInfo().values()
1397
      for inst in instances:
1398
        for disk in inst.disks:
1399
          if _RecursiveCheckIfLVMBased(disk):
1400
            raise errors.OpPrereqError("Cannot disable lvm storage while"
1401
                                       " lvm-based instances exist")
1402

    
1403
    node_list = self.acquired_locks[locking.LEVEL_NODE]
1404

    
1405
    # if vg_name not None, checks given volume group on all nodes
1406
    if self.op.vg_name:
1407
      vglist = self.rpc.call_vg_list(node_list)
1408
      for node in node_list:
1409
        if vglist[node].failed:
1410
          # ignoring down node
1411
          self.LogWarning("Node %s unreachable/error, ignoring" % node)
1412
          continue
1413
        vgstatus = utils.CheckVolumeGroupSize(vglist[node].data,
1414
                                              self.op.vg_name,
1415
                                              constants.MIN_VG_SIZE)
1416
        if vgstatus:
1417
          raise errors.OpPrereqError("Error on node '%s': %s" %
1418
                                     (node, vgstatus))
1419

    
1420
    self.cluster = cluster = self.cfg.GetClusterInfo()
1421
    # validate beparams changes
1422
    if self.op.beparams:
1423
      utils.CheckBEParams(self.op.beparams)
1424
      self.new_beparams = cluster.FillDict(
1425
        cluster.beparams[constants.BEGR_DEFAULT], self.op.beparams)
1426

    
1427
    # hypervisor list/parameters
1428
    self.new_hvparams = cluster.FillDict(cluster.hvparams, {})
1429
    if self.op.hvparams:
1430
      if not isinstance(self.op.hvparams, dict):
1431
        raise errors.OpPrereqError("Invalid 'hvparams' parameter on input")
1432
      for hv_name, hv_dict in self.op.hvparams.items():
1433
        if hv_name not in self.new_hvparams:
1434
          self.new_hvparams[hv_name] = hv_dict
1435
        else:
1436
          self.new_hvparams[hv_name].update(hv_dict)
1437

    
1438
    if self.op.enabled_hypervisors is not None:
1439
      self.hv_list = self.op.enabled_hypervisors
1440
    else:
1441
      self.hv_list = cluster.enabled_hypervisors
1442

    
1443
    if self.op.hvparams or self.op.enabled_hypervisors is not None:
1444
      # either the enabled list has changed, or the parameters have, validate
1445
      for hv_name, hv_params in self.new_hvparams.items():
1446
        if ((self.op.hvparams and hv_name in self.op.hvparams) or
1447
            (self.op.enabled_hypervisors and
1448
             hv_name in self.op.enabled_hypervisors)):
1449
          # either this is a new hypervisor, or its parameters have changed
1450
          hv_class = hypervisor.GetHypervisor(hv_name)
1451
          hv_class.CheckParameterSyntax(hv_params)
1452
          _CheckHVParams(self, node_list, hv_name, hv_params)
1453

    
1454
  def Exec(self, feedback_fn):
1455
    """Change the parameters of the cluster.
1456

1457
    """
1458
    if self.op.vg_name is not None:
1459
      if self.op.vg_name != self.cfg.GetVGName():
1460
        self.cfg.SetVGName(self.op.vg_name)
1461
      else:
1462
        feedback_fn("Cluster LVM configuration already in desired"
1463
                    " state, not changing")
1464
    if self.op.hvparams:
1465
      self.cluster.hvparams = self.new_hvparams
1466
    if self.op.enabled_hypervisors is not None:
1467
      self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
1468
    if self.op.beparams:
1469
      self.cluster.beparams[constants.BEGR_DEFAULT] = self.new_beparams
1470
    if self.op.candidate_pool_size is not None:
1471
      self.cluster.candidate_pool_size = self.op.candidate_pool_size
1472

    
1473
    self.cfg.Update(self.cluster)
1474

    
1475
    # we want to update nodes after the cluster so that if any errors
1476
    # happen, we have recorded and saved the cluster info
1477
    if self.op.candidate_pool_size is not None:
1478
      _AdjustCandidatePool(self)
1479

    
1480

    
1481
class LURedistributeConfig(NoHooksLU):
1482
  """Force the redistribution of cluster configuration.
1483

1484
  This is a very simple LU.
1485

1486
  """
1487
  _OP_REQP = []
1488
  REQ_BGL = False
1489

    
1490
  def ExpandNames(self):
1491
    self.needed_locks = {
1492
      locking.LEVEL_NODE: locking.ALL_SET,
1493
    }
1494
    self.share_locks[locking.LEVEL_NODE] = 1
1495

    
1496
  def CheckPrereq(self):
1497
    """Check prerequisites.
1498

1499
    """
1500

    
1501
  def Exec(self, feedback_fn):
1502
    """Redistribute the configuration.
1503

1504
    """
1505
    self.cfg.Update(self.cfg.GetClusterInfo())
1506

    
1507

    
1508
def _WaitForSync(lu, instance, oneshot=False, unlock=False):
1509
  """Sleep and poll for an instance's disk to sync.
1510

1511
  """
1512
  if not instance.disks:
1513
    return True
1514

    
1515
  if not oneshot:
1516
    lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
1517

    
1518
  node = instance.primary_node
1519

    
1520
  for dev in instance.disks:
1521
    lu.cfg.SetDiskID(dev, node)
1522

    
1523
  retries = 0
1524
  while True:
1525
    max_time = 0
1526
    done = True
1527
    cumul_degraded = False
1528
    rstats = lu.rpc.call_blockdev_getmirrorstatus(node, instance.disks)
1529
    if rstats.failed or not rstats.data:
1530
      lu.LogWarning("Can't get any data from node %s", node)
1531
      retries += 1
1532
      if retries >= 10:
1533
        raise errors.RemoteError("Can't contact node %s for mirror data,"
1534
                                 " aborting." % node)
1535
      time.sleep(6)
1536
      continue
1537
    rstats = rstats.data
1538
    retries = 0
1539
    for i, mstat in enumerate(rstats):
1540
      if mstat is None:
1541
        lu.LogWarning("Can't compute data for node %s/%s",
1542
                           node, instance.disks[i].iv_name)
1543
        continue
1544
      # we ignore the ldisk parameter
1545
      perc_done, est_time, is_degraded, _ = mstat
1546
      cumul_degraded = cumul_degraded or (is_degraded and perc_done is None)
1547
      if perc_done is not None:
1548
        done = False
1549
        if est_time is not None:
1550
          rem_time = "%d estimated seconds remaining" % est_time
1551
          max_time = est_time
1552
        else:
1553
          rem_time = "no time estimate"
1554
        lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
1555
                        (instance.disks[i].iv_name, perc_done, rem_time))
1556
    if done or oneshot:
1557
      break
1558

    
1559
    time.sleep(min(60, max_time))
1560

    
1561
  if done:
1562
    lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
1563
  return not cumul_degraded
1564

    
1565

    
1566
def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
1567
  """Check that mirrors are not degraded.
1568

1569
  The ldisk parameter, if True, will change the test from the
1570
  is_degraded attribute (which represents overall non-ok status for
1571
  the device(s)) to the ldisk (representing the local storage status).
1572

1573
  """
1574
  lu.cfg.SetDiskID(dev, node)
1575
  if ldisk:
1576
    idx = 6
1577
  else:
1578
    idx = 5
1579

    
1580
  result = True
1581
  if on_primary or dev.AssembleOnSecondary():
1582
    rstats = lu.rpc.call_blockdev_find(node, dev)
1583
    if rstats.failed or not rstats.data:
1584
      logging.warning("Node %s: disk degraded, not found or node down", node)
1585
      result = False
1586
    else:
1587
      result = result and (not rstats.data[idx])
1588
  if dev.children:
1589
    for child in dev.children:
1590
      result = result and _CheckDiskConsistency(lu, child, node, on_primary)
1591

    
1592
  return result
1593

    
1594

    
1595
class LUDiagnoseOS(NoHooksLU):
1596
  """Logical unit for OS diagnose/query.
1597

1598
  """
1599
  _OP_REQP = ["output_fields", "names"]
1600
  REQ_BGL = False
1601
  _FIELDS_STATIC = utils.FieldSet()
1602
  _FIELDS_DYNAMIC = utils.FieldSet("name", "valid", "node_status")
1603

    
1604
  def ExpandNames(self):
1605
    if self.op.names:
1606
      raise errors.OpPrereqError("Selective OS query not supported")
1607

    
1608
    _CheckOutputFields(static=self._FIELDS_STATIC,
1609
                       dynamic=self._FIELDS_DYNAMIC,
1610
                       selected=self.op.output_fields)
1611

    
1612
    # Lock all nodes, in shared mode
1613
    self.needed_locks = {}
1614
    self.share_locks[locking.LEVEL_NODE] = 1
1615
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
1616

    
1617
  def CheckPrereq(self):
1618
    """Check prerequisites.
1619

1620
    """
1621

    
1622
  @staticmethod
1623
  def _DiagnoseByOS(node_list, rlist):
1624
    """Remaps a per-node return list into an a per-os per-node dictionary
1625

1626
    @param node_list: a list with the names of all nodes
1627
    @param rlist: a map with node names as keys and OS objects as values
1628

1629
    @rtype: dict
1630
    @returns: a dictionary with osnames as keys and as value another map, with
1631
        nodes as keys and list of OS objects as values, eg::
1632

1633
          {"debian-etch": {"node1": [<object>,...],
1634
                           "node2": [<object>,]}
1635
          }
1636

1637
    """
1638
    all_os = {}
1639
    for node_name, nr in rlist.iteritems():
1640
      if nr.failed or not nr.data:
1641
        continue
1642
      for os_obj in nr.data:
1643
        if os_obj.name not in all_os:
1644
          # build a list of nodes for this os containing empty lists
1645
          # for each node in node_list
1646
          all_os[os_obj.name] = {}
1647
          for nname in node_list:
1648
            all_os[os_obj.name][nname] = []
1649
        all_os[os_obj.name][node_name].append(os_obj)
1650
    return all_os
1651

    
1652
  def Exec(self, feedback_fn):
1653
    """Compute the list of OSes.
1654

1655
    """
1656
    node_list = self.acquired_locks[locking.LEVEL_NODE]
1657
    valid_nodes = [node for node in self.cfg.GetOnlineNodeList()
1658
                   if node in node_list]
1659
    node_data = self.rpc.call_os_diagnose(valid_nodes)
1660
    if node_data == False:
1661
      raise errors.OpExecError("Can't gather the list of OSes")
1662
    pol = self._DiagnoseByOS(valid_nodes, node_data)
1663
    output = []
1664
    for os_name, os_data in pol.iteritems():
1665
      row = []
1666
      for field in self.op.output_fields:
1667
        if field == "name":
1668
          val = os_name
1669
        elif field == "valid":
1670
          val = utils.all([osl and osl[0] for osl in os_data.values()])
1671
        elif field == "node_status":
1672
          val = {}
1673
          for node_name, nos_list in os_data.iteritems():
1674
            val[node_name] = [(v.status, v.path) for v in nos_list]
1675
        else:
1676
          raise errors.ParameterError(field)
1677
        row.append(val)
1678
      output.append(row)
1679

    
1680
    return output
1681

    
1682

    
1683
class LURemoveNode(LogicalUnit):
1684
  """Logical unit for removing a node.
1685

1686
  """
1687
  HPATH = "node-remove"
1688
  HTYPE = constants.HTYPE_NODE
1689
  _OP_REQP = ["node_name"]
1690

    
1691
  def BuildHooksEnv(self):
1692
    """Build hooks env.
1693

1694
    This doesn't run on the target node in the pre phase as a failed
1695
    node would then be impossible to remove.
1696

1697
    """
1698
    env = {
1699
      "OP_TARGET": self.op.node_name,
1700
      "NODE_NAME": self.op.node_name,
1701
      }
1702
    all_nodes = self.cfg.GetNodeList()
1703
    all_nodes.remove(self.op.node_name)
1704
    return env, all_nodes, all_nodes
1705

    
1706
  def CheckPrereq(self):
1707
    """Check prerequisites.
1708

1709
    This checks:
1710
     - the node exists in the configuration
1711
     - it does not have primary or secondary instances
1712
     - it's not the master
1713

1714
    Any errors are signalled by raising errors.OpPrereqError.
1715

1716
    """
1717
    node = self.cfg.GetNodeInfo(self.cfg.ExpandNodeName(self.op.node_name))
1718
    if node is None:
1719
      raise errors.OpPrereqError, ("Node '%s' is unknown." % self.op.node_name)
1720

    
1721
    instance_list = self.cfg.GetInstanceList()
1722

    
1723
    masternode = self.cfg.GetMasterNode()
1724
    if node.name == masternode:
1725
      raise errors.OpPrereqError("Node is the master node,"
1726
                                 " you need to failover first.")
1727

    
1728
    for instance_name in instance_list:
1729
      instance = self.cfg.GetInstanceInfo(instance_name)
1730
      if node.name in instance.all_nodes:
1731
        raise errors.OpPrereqError("Instance %s is still running on the node,"
1732
                                   " please remove first." % instance_name)
1733
    self.op.node_name = node.name
1734
    self.node = node
1735

    
1736
  def Exec(self, feedback_fn):
1737
    """Removes the node from the cluster.
1738

1739
    """
1740
    node = self.node
1741
    logging.info("Stopping the node daemon and removing configs from node %s",
1742
                 node.name)
1743

    
1744
    self.context.RemoveNode(node.name)
1745

    
1746
    self.rpc.call_node_leave_cluster(node.name)
1747

    
1748
    # Promote nodes to master candidate as needed
1749
    _AdjustCandidatePool(self)
1750

    
1751

    
1752
class LUQueryNodes(NoHooksLU):
1753
  """Logical unit for querying nodes.
1754

1755
  """
1756
  _OP_REQP = ["output_fields", "names", "use_locking"]
1757
  REQ_BGL = False
1758
  _FIELDS_DYNAMIC = utils.FieldSet(
1759
    "dtotal", "dfree",
1760
    "mtotal", "mnode", "mfree",
1761
    "bootid",
1762
    "ctotal", "cnodes", "csockets",
1763
    )
1764

    
1765
  _FIELDS_STATIC = utils.FieldSet(
1766
    "name", "pinst_cnt", "sinst_cnt",
1767
    "pinst_list", "sinst_list",
1768
    "pip", "sip", "tags",
1769
    "serial_no",
1770
    "master_candidate",
1771
    "master",
1772
    "offline",
1773
    )
1774

    
1775
  def ExpandNames(self):
1776
    _CheckOutputFields(static=self._FIELDS_STATIC,
1777
                       dynamic=self._FIELDS_DYNAMIC,
1778
                       selected=self.op.output_fields)
1779

    
1780
    self.needed_locks = {}
1781
    self.share_locks[locking.LEVEL_NODE] = 1
1782

    
1783
    if self.op.names:
1784
      self.wanted = _GetWantedNodes(self, self.op.names)
1785
    else:
1786
      self.wanted = locking.ALL_SET
1787

    
1788
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
1789
    self.do_locking = self.do_node_query and self.op.use_locking
1790
    if self.do_locking:
1791
      # if we don't request only static fields, we need to lock the nodes
1792
      self.needed_locks[locking.LEVEL_NODE] = self.wanted
1793

    
1794

    
1795
  def CheckPrereq(self):
1796
    """Check prerequisites.
1797

1798
    """
1799
    # The validation of the node list is done in the _GetWantedNodes,
1800
    # if non empty, and if empty, there's no validation to do
1801
    pass
1802

    
1803
  def Exec(self, feedback_fn):
1804
    """Computes the list of nodes and their attributes.
1805

1806
    """
1807
    all_info = self.cfg.GetAllNodesInfo()
1808
    if self.do_locking:
1809
      nodenames = self.acquired_locks[locking.LEVEL_NODE]
1810
    elif self.wanted != locking.ALL_SET:
1811
      nodenames = self.wanted
1812
      missing = set(nodenames).difference(all_info.keys())
1813
      if missing:
1814
        raise errors.OpExecError(
1815
          "Some nodes were removed before retrieving their data: %s" % missing)
1816
    else:
1817
      nodenames = all_info.keys()
1818

    
1819
    nodenames = utils.NiceSort(nodenames)
1820
    nodelist = [all_info[name] for name in nodenames]
1821

    
1822
    # begin data gathering
1823

    
1824
    if self.do_node_query:
1825
      live_data = {}
1826
      node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
1827
                                          self.cfg.GetHypervisorType())
1828
      for name in nodenames:
1829
        nodeinfo = node_data[name]
1830
        if not nodeinfo.failed and nodeinfo.data:
1831
          nodeinfo = nodeinfo.data
1832
          fn = utils.TryConvert
1833
          live_data[name] = {
1834
            "mtotal": fn(int, nodeinfo.get('memory_total', None)),
1835
            "mnode": fn(int, nodeinfo.get('memory_dom0', None)),
1836
            "mfree": fn(int, nodeinfo.get('memory_free', None)),
1837
            "dtotal": fn(int, nodeinfo.get('vg_size', None)),
1838
            "dfree": fn(int, nodeinfo.get('vg_free', None)),
1839
            "ctotal": fn(int, nodeinfo.get('cpu_total', None)),
1840
            "bootid": nodeinfo.get('bootid', None),
1841
            "cnodes": fn(int, nodeinfo.get('cpu_nodes', None)),
1842
            "csockets": fn(int, nodeinfo.get('cpu_sockets', None)),
1843
            }
1844
        else:
1845
          live_data[name] = {}
1846
    else:
1847
      live_data = dict.fromkeys(nodenames, {})
1848

    
1849
    node_to_primary = dict([(name, set()) for name in nodenames])
1850
    node_to_secondary = dict([(name, set()) for name in nodenames])
1851

    
1852
    inst_fields = frozenset(("pinst_cnt", "pinst_list",
1853
                             "sinst_cnt", "sinst_list"))
1854
    if inst_fields & frozenset(self.op.output_fields):
1855
      instancelist = self.cfg.GetInstanceList()
1856

    
1857
      for instance_name in instancelist:
1858
        inst = self.cfg.GetInstanceInfo(instance_name)
1859
        if inst.primary_node in node_to_primary:
1860
          node_to_primary[inst.primary_node].add(inst.name)
1861
        for secnode in inst.secondary_nodes:
1862
          if secnode in node_to_secondary:
1863
            node_to_secondary[secnode].add(inst.name)
1864

    
1865
    master_node = self.cfg.GetMasterNode()
1866

    
1867
    # end data gathering
1868

    
1869
    output = []
1870
    for node in nodelist:
1871
      node_output = []
1872
      for field in self.op.output_fields:
1873
        if field == "name":
1874
          val = node.name
1875
        elif field == "pinst_list":
1876
          val = list(node_to_primary[node.name])
1877
        elif field == "sinst_list":
1878
          val = list(node_to_secondary[node.name])
1879
        elif field == "pinst_cnt":
1880
          val = len(node_to_primary[node.name])
1881
        elif field == "sinst_cnt":
1882
          val = len(node_to_secondary[node.name])
1883
        elif field == "pip":
1884
          val = node.primary_ip
1885
        elif field == "sip":
1886
          val = node.secondary_ip
1887
        elif field == "tags":
1888
          val = list(node.GetTags())
1889
        elif field == "serial_no":
1890
          val = node.serial_no
1891
        elif field == "master_candidate":
1892
          val = node.master_candidate
1893
        elif field == "master":
1894
          val = node.name == master_node
1895
        elif field == "offline":
1896
          val = node.offline
1897
        elif self._FIELDS_DYNAMIC.Matches(field):
1898
          val = live_data[node.name].get(field, None)
1899
        else:
1900
          raise errors.ParameterError(field)
1901
        node_output.append(val)
1902
      output.append(node_output)
1903

    
1904
    return output
1905

    
1906

    
1907
class LUQueryNodeVolumes(NoHooksLU):
1908
  """Logical unit for getting volumes on node(s).
1909

1910
  """
1911
  _OP_REQP = ["nodes", "output_fields"]
1912
  REQ_BGL = False
1913
  _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
1914
  _FIELDS_STATIC = utils.FieldSet("node")
1915

    
1916
  def ExpandNames(self):
1917
    _CheckOutputFields(static=self._FIELDS_STATIC,
1918
                       dynamic=self._FIELDS_DYNAMIC,
1919
                       selected=self.op.output_fields)
1920

    
1921
    self.needed_locks = {}
1922
    self.share_locks[locking.LEVEL_NODE] = 1
1923
    if not self.op.nodes:
1924
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
1925
    else:
1926
      self.needed_locks[locking.LEVEL_NODE] = \
1927
        _GetWantedNodes(self, self.op.nodes)
1928

    
1929
  def CheckPrereq(self):
1930
    """Check prerequisites.
1931

1932
    This checks that the fields required are valid output fields.
1933

1934
    """
1935
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
1936

    
1937
  def Exec(self, feedback_fn):
1938
    """Computes the list of nodes and their attributes.
1939

1940
    """
1941
    nodenames = self.nodes
1942
    volumes = self.rpc.call_node_volumes(nodenames)
1943

    
1944
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
1945
             in self.cfg.GetInstanceList()]
1946

    
1947
    lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
1948

    
1949
    output = []
1950
    for node in nodenames:
1951
      if node not in volumes or volumes[node].failed or not volumes[node].data:
1952
        continue
1953

    
1954
      node_vols = volumes[node].data[:]
1955
      node_vols.sort(key=lambda vol: vol['dev'])
1956

    
1957
      for vol in node_vols:
1958
        node_output = []
1959
        for field in self.op.output_fields:
1960
          if field == "node":
1961
            val = node
1962
          elif field == "phys":
1963
            val = vol['dev']
1964
          elif field == "vg":
1965
            val = vol['vg']
1966
          elif field == "name":
1967
            val = vol['name']
1968
          elif field == "size":
1969
            val = int(float(vol['size']))
1970
          elif field == "instance":
1971
            for inst in ilist:
1972
              if node not in lv_by_node[inst]:
1973
                continue
1974
              if vol['name'] in lv_by_node[inst][node]:
1975
                val = inst.name
1976
                break
1977
            else:
1978
              val = '-'
1979
          else:
1980
            raise errors.ParameterError(field)
1981
          node_output.append(str(val))
1982

    
1983
        output.append(node_output)
1984

    
1985
    return output
1986

    
1987

    
1988
class LUAddNode(LogicalUnit):
1989
  """Logical unit for adding node to the cluster.
1990

1991
  """
1992
  HPATH = "node-add"
1993
  HTYPE = constants.HTYPE_NODE
1994
  _OP_REQP = ["node_name"]
1995

    
1996
  def BuildHooksEnv(self):
1997
    """Build hooks env.
1998

1999
    This will run on all nodes before, and on all nodes + the new node after.
2000

2001
    """
2002
    env = {
2003
      "OP_TARGET": self.op.node_name,
2004
      "NODE_NAME": self.op.node_name,
2005
      "NODE_PIP": self.op.primary_ip,
2006
      "NODE_SIP": self.op.secondary_ip,
2007
      }
2008
    nodes_0 = self.cfg.GetNodeList()
2009
    nodes_1 = nodes_0 + [self.op.node_name, ]
2010
    return env, nodes_0, nodes_1
2011

    
2012
  def CheckPrereq(self):
2013
    """Check prerequisites.
2014

2015
    This checks:
2016
     - the new node is not already in the config
2017
     - it is resolvable
2018
     - its parameters (single/dual homed) matches the cluster
2019

2020
    Any errors are signalled by raising errors.OpPrereqError.
2021

2022
    """
2023
    node_name = self.op.node_name
2024
    cfg = self.cfg
2025

    
2026
    dns_data = utils.HostInfo(node_name)
2027

    
2028
    node = dns_data.name
2029
    primary_ip = self.op.primary_ip = dns_data.ip
2030
    secondary_ip = getattr(self.op, "secondary_ip", None)
2031
    if secondary_ip is None:
2032
      secondary_ip = primary_ip
2033
    if not utils.IsValidIP(secondary_ip):
2034
      raise errors.OpPrereqError("Invalid secondary IP given")
2035
    self.op.secondary_ip = secondary_ip
2036

    
2037
    node_list = cfg.GetNodeList()
2038
    if not self.op.readd and node in node_list:
2039
      raise errors.OpPrereqError("Node %s is already in the configuration" %
2040
                                 node)
2041
    elif self.op.readd and node not in node_list:
2042
      raise errors.OpPrereqError("Node %s is not in the configuration" % node)
2043

    
2044
    for existing_node_name in node_list:
2045
      existing_node = cfg.GetNodeInfo(existing_node_name)
2046

    
2047
      if self.op.readd and node == existing_node_name:
2048
        if (existing_node.primary_ip != primary_ip or
2049
            existing_node.secondary_ip != secondary_ip):
2050
          raise errors.OpPrereqError("Readded node doesn't have the same IP"
2051
                                     " address configuration as before")
2052
        continue
2053

    
2054
      if (existing_node.primary_ip == primary_ip or
2055
          existing_node.secondary_ip == primary_ip or
2056
          existing_node.primary_ip == secondary_ip or
2057
          existing_node.secondary_ip == secondary_ip):
2058
        raise errors.OpPrereqError("New node ip address(es) conflict with"
2059
                                   " existing node %s" % existing_node.name)
2060

    
2061
    # check that the type of the node (single versus dual homed) is the
2062
    # same as for the master
2063
    myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
2064
    master_singlehomed = myself.secondary_ip == myself.primary_ip
2065
    newbie_singlehomed = secondary_ip == primary_ip
2066
    if master_singlehomed != newbie_singlehomed:
2067
      if master_singlehomed:
2068
        raise errors.OpPrereqError("The master has no private ip but the"
2069
                                   " new node has one")
2070
      else:
2071
        raise errors.OpPrereqError("The master has a private ip but the"
2072
                                   " new node doesn't have one")
2073

    
2074
    # checks reachablity
2075
    if not utils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
2076
      raise errors.OpPrereqError("Node not reachable by ping")
2077

    
2078
    if not newbie_singlehomed:
2079
      # check reachability from my secondary ip to newbie's secondary ip
2080
      if not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
2081
                           source=myself.secondary_ip):
2082
        raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
2083
                                   " based ping to noded port")
2084

    
2085
    cp_size = self.cfg.GetClusterInfo().candidate_pool_size
2086
    mc_now, _ = self.cfg.GetMasterCandidateStats()
2087
    master_candidate = mc_now < cp_size
2088

    
2089
    self.new_node = objects.Node(name=node,
2090
                                 primary_ip=primary_ip,
2091
                                 secondary_ip=secondary_ip,
2092
                                 master_candidate=master_candidate,
2093
                                 offline=False)
2094

    
2095
  def Exec(self, feedback_fn):
2096
    """Adds the new node to the cluster.
2097

2098
    """
2099
    new_node = self.new_node
2100
    node = new_node.name
2101

    
2102
    # check connectivity
2103
    result = self.rpc.call_version([node])[node]
2104
    result.Raise()
2105
    if result.data:
2106
      if constants.PROTOCOL_VERSION == result.data:
2107
        logging.info("Communication to node %s fine, sw version %s match",
2108
                     node, result.data)
2109
      else:
2110
        raise errors.OpExecError("Version mismatch master version %s,"
2111
                                 " node version %s" %
2112
                                 (constants.PROTOCOL_VERSION, result.data))
2113
    else:
2114
      raise errors.OpExecError("Cannot get version from the new node")
2115

    
2116
    # setup ssh on node
2117
    logging.info("Copy ssh key to node %s", node)
2118
    priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
2119
    keyarray = []
2120
    keyfiles = [constants.SSH_HOST_DSA_PRIV, constants.SSH_HOST_DSA_PUB,
2121
                constants.SSH_HOST_RSA_PRIV, constants.SSH_HOST_RSA_PUB,
2122
                priv_key, pub_key]
2123

    
2124
    for i in keyfiles:
2125
      f = open(i, 'r')
2126
      try:
2127
        keyarray.append(f.read())
2128
      finally:
2129
        f.close()
2130

    
2131
    result = self.rpc.call_node_add(node, keyarray[0], keyarray[1],
2132
                                    keyarray[2],
2133
                                    keyarray[3], keyarray[4], keyarray[5])
2134

    
2135
    msg = result.RemoteFailMsg()
2136
    if msg:
2137
      raise errors.OpExecError("Cannot transfer ssh keys to the"
2138
                               " new node: %s" % msg)
2139

    
2140
    # Add node to our /etc/hosts, and add key to known_hosts
2141
    utils.AddHostToEtcHosts(new_node.name)
2142

    
2143
    if new_node.secondary_ip != new_node.primary_ip:
2144
      result = self.rpc.call_node_has_ip_address(new_node.name,
2145
                                                 new_node.secondary_ip)
2146
      if result.failed or not result.data:
2147
        raise errors.OpExecError("Node claims it doesn't have the secondary ip"
2148
                                 " you gave (%s). Please fix and re-run this"
2149
                                 " command." % new_node.secondary_ip)
2150

    
2151
    node_verify_list = [self.cfg.GetMasterNode()]
2152
    node_verify_param = {
2153
      'nodelist': [node],
2154
      # TODO: do a node-net-test as well?
2155
    }
2156

    
2157
    result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
2158
                                       self.cfg.GetClusterName())
2159
    for verifier in node_verify_list:
2160
      if result[verifier].failed or not result[verifier].data:
2161
        raise errors.OpExecError("Cannot communicate with %s's node daemon"
2162
                                 " for remote verification" % verifier)
2163
      if result[verifier].data['nodelist']:
2164
        for failed in result[verifier].data['nodelist']:
2165
          feedback_fn("ssh/hostname verification failed %s -> %s" %
2166
                      (verifier, result[verifier].data['nodelist'][failed]))
2167
        raise errors.OpExecError("ssh/hostname verification failed.")
2168

    
2169
    # Distribute updated /etc/hosts and known_hosts to all nodes,
2170
    # including the node just added
2171
    myself = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
2172
    dist_nodes = self.cfg.GetNodeList()
2173
    if not self.op.readd:
2174
      dist_nodes.append(node)
2175
    if myself.name in dist_nodes:
2176
      dist_nodes.remove(myself.name)
2177

    
2178
    logging.debug("Copying hosts and known_hosts to all nodes")
2179
    for fname in (constants.ETC_HOSTS, constants.SSH_KNOWN_HOSTS_FILE):
2180
      result = self.rpc.call_upload_file(dist_nodes, fname)
2181
      for to_node, to_result in result.iteritems():
2182
        if to_result.failed or not to_result.data:
2183
          logging.error("Copy of file %s to node %s failed", fname, to_node)
2184

    
2185
    to_copy = []
2186
    enabled_hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
2187
    if constants.HTS_USE_VNC.intersection(enabled_hypervisors):
2188
      to_copy.append(constants.VNC_PASSWORD_FILE)
2189

    
2190
    for fname in to_copy:
2191
      result = self.rpc.call_upload_file([node], fname)
2192
      if result[node].failed or not result[node]:
2193
        logging.error("Could not copy file %s to node %s", fname, node)
2194

    
2195
    if self.op.readd:
2196
      self.context.ReaddNode(new_node)
2197
    else:
2198
      self.context.AddNode(new_node)
2199

    
2200

    
2201
class LUSetNodeParams(LogicalUnit):
2202
  """Modifies the parameters of a node.
2203

2204
  """
2205
  HPATH = "node-modify"
2206
  HTYPE = constants.HTYPE_NODE
2207
  _OP_REQP = ["node_name"]
2208
  REQ_BGL = False
2209

    
2210
  def CheckArguments(self):
2211
    node_name = self.cfg.ExpandNodeName(self.op.node_name)
2212
    if node_name is None:
2213
      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name)
2214
    self.op.node_name = node_name
2215
    _CheckBooleanOpField(self.op, 'master_candidate')
2216
    _CheckBooleanOpField(self.op, 'offline')
2217
    if self.op.master_candidate is None and self.op.offline is None:
2218
      raise errors.OpPrereqError("Please pass at least one modification")
2219
    if self.op.offline == True and self.op.master_candidate == True:
2220
      raise errors.OpPrereqError("Can't set the node into offline and"
2221
                                 " master_candidate at the same time")
2222

    
2223
  def ExpandNames(self):
2224
    self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
2225

    
2226
  def BuildHooksEnv(self):
2227
    """Build hooks env.
2228

2229
    This runs on the master node.
2230

2231
    """
2232
    env = {
2233
      "OP_TARGET": self.op.node_name,
2234
      "MASTER_CANDIDATE": str(self.op.master_candidate),
2235
      "OFFLINE": str(self.op.offline),
2236
      }
2237
    nl = [self.cfg.GetMasterNode(),
2238
          self.op.node_name]
2239
    return env, nl, nl
2240

    
2241
  def CheckPrereq(self):
2242
    """Check prerequisites.
2243

2244
    This only checks the instance list against the existing names.
2245

2246
    """
2247
    node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
2248

    
2249
    if ((self.op.master_candidate == False or self.op.offline == True)
2250
        and node.master_candidate):
2251
      # we will demote the node from master_candidate
2252
      if self.op.node_name == self.cfg.GetMasterNode():
2253
        raise errors.OpPrereqError("The master node has to be a"
2254
                                   " master candidate and online")
2255
      cp_size = self.cfg.GetClusterInfo().candidate_pool_size
2256
      num_candidates, _ = self.cfg.GetMasterCandidateStats()
2257
      if num_candidates <= cp_size:
2258
        msg = ("Not enough master candidates (desired"
2259
               " %d, new value will be %d)" % (cp_size, num_candidates-1))
2260
        if self.op.force:
2261
          self.LogWarning(msg)
2262
        else:
2263
          raise errors.OpPrereqError(msg)
2264

    
2265
    if (self.op.master_candidate == True and node.offline and
2266
        not self.op.offline == False):
2267
      raise errors.OpPrereqError("Can't set an offline node to"
2268
                                 " master_candidate")
2269

    
2270
    return
2271

    
2272
  def Exec(self, feedback_fn):
2273
    """Modifies a node.
2274

2275
    """
2276
    node = self.node
2277

    
2278
    result = []
2279

    
2280
    if self.op.offline is not None:
2281
      node.offline = self.op.offline
2282
      result.append(("offline", str(self.op.offline)))
2283
      if self.op.offline == True and node.master_candidate:
2284
        node.master_candidate = False
2285
        result.append(("master_candidate", "auto-demotion due to offline"))
2286

    
2287
    if self.op.master_candidate is not None:
2288
      node.master_candidate = self.op.master_candidate
2289
      result.append(("master_candidate", str(self.op.master_candidate)))
2290
      if self.op.master_candidate == False:
2291
        rrc = self.rpc.call_node_demote_from_mc(node.name)
2292
        msg = rrc.RemoteFailMsg()
2293
        if msg:
2294
          self.LogWarning("Node failed to demote itself: %s" % msg)
2295

    
2296
    # this will trigger configuration file update, if needed
2297
    self.cfg.Update(node)
2298
    # this will trigger job queue propagation or cleanup
2299
    if self.op.node_name != self.cfg.GetMasterNode():
2300
      self.context.ReaddNode(node)
2301

    
2302
    return result
2303

    
2304

    
2305
class LUQueryClusterInfo(NoHooksLU):
2306
  """Query cluster configuration.
2307

2308
  """
2309
  _OP_REQP = []
2310
  REQ_BGL = False
2311

    
2312
  def ExpandNames(self):
2313
    self.needed_locks = {}
2314

    
2315
  def CheckPrereq(self):
2316
    """No prerequsites needed for this LU.
2317

2318
    """
2319
    pass
2320

    
2321
  def Exec(self, feedback_fn):
2322
    """Return cluster config.
2323

2324
    """
2325
    cluster = self.cfg.GetClusterInfo()
2326
    result = {
2327
      "software_version": constants.RELEASE_VERSION,
2328
      "protocol_version": constants.PROTOCOL_VERSION,
2329
      "config_version": constants.CONFIG_VERSION,
2330
      "os_api_version": constants.OS_API_VERSION,
2331
      "export_version": constants.EXPORT_VERSION,
2332
      "architecture": (platform.architecture()[0], platform.machine()),
2333
      "name": cluster.cluster_name,
2334
      "master": cluster.master_node,
2335
      "default_hypervisor": cluster.default_hypervisor,
2336
      "enabled_hypervisors": cluster.enabled_hypervisors,
2337
      "hvparams": dict([(hypervisor, cluster.hvparams[hypervisor])
2338
                        for hypervisor in cluster.enabled_hypervisors]),
2339
      "beparams": cluster.beparams,
2340
      "candidate_pool_size": cluster.candidate_pool_size,
2341
      }
2342

    
2343
    return result
2344

    
2345

    
2346
class LUQueryConfigValues(NoHooksLU):
2347
  """Return configuration values.
2348

2349
  """
2350
  _OP_REQP = []
2351
  REQ_BGL = False
2352
  _FIELDS_DYNAMIC = utils.FieldSet()
2353
  _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag")
2354

    
2355
  def ExpandNames(self):
2356
    self.needed_locks = {}
2357

    
2358
    _CheckOutputFields(static=self._FIELDS_STATIC,
2359
                       dynamic=self._FIELDS_DYNAMIC,
2360
                       selected=self.op.output_fields)
2361

    
2362
  def CheckPrereq(self):
2363
    """No prerequisites.
2364

2365
    """
2366
    pass
2367

    
2368
  def Exec(self, feedback_fn):
2369
    """Dump a representation of the cluster config to the standard output.
2370

2371
    """
2372
    values = []
2373
    for field in self.op.output_fields:
2374
      if field == "cluster_name":
2375
        entry = self.cfg.GetClusterName()
2376
      elif field == "master_node":
2377
        entry = self.cfg.GetMasterNode()
2378
      elif field == "drain_flag":
2379
        entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
2380
      else:
2381
        raise errors.ParameterError(field)
2382
      values.append(entry)
2383
    return values
2384

    
2385

    
2386
class LUActivateInstanceDisks(NoHooksLU):
2387
  """Bring up an instance's disks.
2388

2389
  """
2390
  _OP_REQP = ["instance_name"]
2391
  REQ_BGL = False
2392

    
2393
  def ExpandNames(self):
2394
    self._ExpandAndLockInstance()
2395
    self.needed_locks[locking.LEVEL_NODE] = []
2396
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2397

    
2398
  def DeclareLocks(self, level):
2399
    if level == locking.LEVEL_NODE:
2400
      self._LockInstancesNodes()
2401

    
2402
  def CheckPrereq(self):
2403
    """Check prerequisites.
2404

2405
    This checks that the instance is in the cluster.
2406

2407
    """
2408
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2409
    assert self.instance is not None, \
2410
      "Cannot retrieve locked instance %s" % self.op.instance_name
2411
    _CheckNodeOnline(self, self.instance.primary_node)
2412

    
2413
  def Exec(self, feedback_fn):
2414
    """Activate the disks.
2415

2416
    """
2417
    disks_ok, disks_info = _AssembleInstanceDisks(self, self.instance)
2418
    if not disks_ok:
2419
      raise errors.OpExecError("Cannot activate block devices")
2420

    
2421
    return disks_info
2422

    
2423

    
2424
def _AssembleInstanceDisks(lu, instance, ignore_secondaries=False):
2425
  """Prepare the block devices for an instance.
2426

2427
  This sets up the block devices on all nodes.
2428

2429
  @type lu: L{LogicalUnit}
2430
  @param lu: the logical unit on whose behalf we execute
2431
  @type instance: L{objects.Instance}
2432
  @param instance: the instance for whose disks we assemble
2433
  @type ignore_secondaries: boolean
2434
  @param ignore_secondaries: if true, errors on secondary nodes
2435
      won't result in an error return from the function
2436
  @return: False if the operation failed, otherwise a list of
2437
      (host, instance_visible_name, node_visible_name)
2438
      with the mapping from node devices to instance devices
2439

2440
  """
2441
  device_info = []
2442
  disks_ok = True
2443
  iname = instance.name
2444
  # With the two passes mechanism we try to reduce the window of
2445
  # opportunity for the race condition of switching DRBD to primary
2446
  # before handshaking occured, but we do not eliminate it
2447

    
2448
  # The proper fix would be to wait (with some limits) until the
2449
  # connection has been made and drbd transitions from WFConnection
2450
  # into any other network-connected state (Connected, SyncTarget,
2451
  # SyncSource, etc.)
2452

    
2453
  # 1st pass, assemble on all nodes in secondary mode
2454
  for inst_disk in instance.disks:
2455
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
2456
      lu.cfg.SetDiskID(node_disk, node)
2457
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
2458
      if result.failed or not result:
2459
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
2460
                           " (is_primary=False, pass=1)",
2461
                           inst_disk.iv_name, node)
2462
        if not ignore_secondaries:
2463
          disks_ok = False
2464

    
2465
  # FIXME: race condition on drbd migration to primary
2466

    
2467
  # 2nd pass, do only the primary node
2468
  for inst_disk in instance.disks:
2469
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
2470
      if node != instance.primary_node:
2471
        continue
2472
      lu.cfg.SetDiskID(node_disk, node)
2473
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
2474
      if result.failed or not result:
2475
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
2476
                           " (is_primary=True, pass=2)",
2477
                           inst_disk.iv_name, node)
2478
        disks_ok = False
2479
    device_info.append((instance.primary_node, inst_disk.iv_name, result.data))
2480

    
2481
  # leave the disks configured for the primary node
2482
  # this is a workaround that would be fixed better by
2483
  # improving the logical/physical id handling
2484
  for disk in instance.disks:
2485
    lu.cfg.SetDiskID(disk, instance.primary_node)
2486

    
2487
  return disks_ok, device_info
2488

    
2489

    
2490
def _StartInstanceDisks(lu, instance, force):
2491
  """Start the disks of an instance.
2492

2493
  """
2494
  disks_ok, dummy = _AssembleInstanceDisks(lu, instance,
2495
                                           ignore_secondaries=force)
2496
  if not disks_ok:
2497
    _ShutdownInstanceDisks(lu, instance)
2498
    if force is not None and not force:
2499
      lu.proc.LogWarning("", hint="If the message above refers to a"
2500
                         " secondary node,"
2501
                         " you can retry the operation using '--force'.")
2502
    raise errors.OpExecError("Disk consistency error")
2503

    
2504

    
2505
class LUDeactivateInstanceDisks(NoHooksLU):
2506
  """Shutdown an instance's disks.
2507

2508
  """
2509
  _OP_REQP = ["instance_name"]
2510
  REQ_BGL = False
2511

    
2512
  def ExpandNames(self):
2513
    self._ExpandAndLockInstance()
2514
    self.needed_locks[locking.LEVEL_NODE] = []
2515
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2516

    
2517
  def DeclareLocks(self, level):
2518
    if level == locking.LEVEL_NODE:
2519
      self._LockInstancesNodes()
2520

    
2521
  def CheckPrereq(self):
2522
    """Check prerequisites.
2523

2524
    This checks that the instance is in the cluster.
2525

2526
    """
2527
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2528
    assert self.instance is not None, \
2529
      "Cannot retrieve locked instance %s" % self.op.instance_name
2530

    
2531
  def Exec(self, feedback_fn):
2532
    """Deactivate the disks
2533

2534
    """
2535
    instance = self.instance
2536
    _SafeShutdownInstanceDisks(self, instance)
2537

    
2538

    
2539
def _SafeShutdownInstanceDisks(lu, instance):
2540
  """Shutdown block devices of an instance.
2541

2542
  This function checks if an instance is running, before calling
2543
  _ShutdownInstanceDisks.
2544

2545
  """
2546
  ins_l = lu.rpc.call_instance_list([instance.primary_node],
2547
                                      [instance.hypervisor])
2548
  ins_l = ins_l[instance.primary_node]
2549
  if ins_l.failed or not isinstance(ins_l.data, list):
2550
    raise errors.OpExecError("Can't contact node '%s'" %
2551
                             instance.primary_node)
2552

    
2553
  if instance.name in ins_l.data:
2554
    raise errors.OpExecError("Instance is running, can't shutdown"
2555
                             " block devices.")
2556

    
2557
  _ShutdownInstanceDisks(lu, instance)
2558

    
2559

    
2560
def _ShutdownInstanceDisks(lu, instance, ignore_primary=False):
2561
  """Shutdown block devices of an instance.
2562

2563
  This does the shutdown on all nodes of the instance.
2564

2565
  If the ignore_primary is false, errors on the primary node are
2566
  ignored.
2567

2568
  """
2569
  result = True
2570
  for disk in instance.disks:
2571
    for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
2572
      lu.cfg.SetDiskID(top_disk, node)
2573
      result = lu.rpc.call_blockdev_shutdown(node, top_disk)
2574
      if result.failed or not result.data:
2575
        logging.error("Could not shutdown block device %s on node %s",
2576
                      disk.iv_name, node)
2577
        if not ignore_primary or node != instance.primary_node:
2578
          result = False
2579
  return result
2580

    
2581

    
2582
def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
2583
  """Checks if a node has enough free memory.
2584

2585
  This function check if a given node has the needed amount of free
2586
  memory. In case the node has less memory or we cannot get the
2587
  information from the node, this function raise an OpPrereqError
2588
  exception.
2589

2590
  @type lu: C{LogicalUnit}
2591
  @param lu: a logical unit from which we get configuration data
2592
  @type node: C{str}
2593
  @param node: the node to check
2594
  @type reason: C{str}
2595
  @param reason: string to use in the error message
2596
  @type requested: C{int}
2597
  @param requested: the amount of memory in MiB to check for
2598
  @type hypervisor_name: C{str}
2599
  @param hypervisor_name: the hypervisor to ask for memory stats
2600
  @raise errors.OpPrereqError: if the node doesn't have enough memory, or
2601
      we cannot check the node
2602

2603
  """
2604
  nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
2605
  nodeinfo[node].Raise()
2606
  free_mem = nodeinfo[node].data.get('memory_free')
2607
  if not isinstance(free_mem, int):
2608
    raise errors.OpPrereqError("Can't compute free memory on node %s, result"
2609
                             " was '%s'" % (node, free_mem))
2610
  if requested > free_mem:
2611
    raise errors.OpPrereqError("Not enough memory on node %s for %s:"
2612
                             " needed %s MiB, available %s MiB" %
2613
                             (node, reason, requested, free_mem))
2614

    
2615

    
2616
class LUStartupInstance(LogicalUnit):
2617
  """Starts an instance.
2618

2619
  """
2620
  HPATH = "instance-start"
2621
  HTYPE = constants.HTYPE_INSTANCE
2622
  _OP_REQP = ["instance_name", "force"]
2623
  REQ_BGL = False
2624

    
2625
  def ExpandNames(self):
2626
    self._ExpandAndLockInstance()
2627

    
2628
  def BuildHooksEnv(self):
2629
    """Build hooks env.
2630

2631
    This runs on master, primary and secondary nodes of the instance.
2632

2633
    """
2634
    env = {
2635
      "FORCE": self.op.force,
2636
      }
2637
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2638
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2639
    return env, nl, nl
2640

    
2641
  def CheckPrereq(self):
2642
    """Check prerequisites.
2643

2644
    This checks that the instance is in the cluster.
2645

2646
    """
2647
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2648
    assert self.instance is not None, \
2649
      "Cannot retrieve locked instance %s" % self.op.instance_name
2650

    
2651
    _CheckNodeOnline(self, instance.primary_node)
2652

    
2653
    bep = self.cfg.GetClusterInfo().FillBE(instance)
2654
    # check bridges existance
2655
    _CheckInstanceBridgesExist(self, instance)
2656

    
2657
    _CheckNodeFreeMemory(self, instance.primary_node,
2658
                         "starting instance %s" % instance.name,
2659
                         bep[constants.BE_MEMORY], instance.hypervisor)
2660

    
2661
  def Exec(self, feedback_fn):
2662
    """Start the instance.
2663

2664
    """
2665
    instance = self.instance
2666
    force = self.op.force
2667
    extra_args = getattr(self.op, "extra_args", "")
2668

    
2669
    self.cfg.MarkInstanceUp(instance.name)
2670

    
2671
    node_current = instance.primary_node
2672

    
2673
    _StartInstanceDisks(self, instance, force)
2674

    
2675
    result = self.rpc.call_instance_start(node_current, instance, extra_args)
2676
    msg = result.RemoteFailMsg()
2677
    if msg:
2678
      _ShutdownInstanceDisks(self, instance)
2679
      raise errors.OpExecError("Could not start instance: %s" % msg)
2680

    
2681

    
2682
class LURebootInstance(LogicalUnit):
2683
  """Reboot an instance.
2684

2685
  """
2686
  HPATH = "instance-reboot"
2687
  HTYPE = constants.HTYPE_INSTANCE
2688
  _OP_REQP = ["instance_name", "ignore_secondaries", "reboot_type"]
2689
  REQ_BGL = False
2690

    
2691
  def ExpandNames(self):
2692
    if self.op.reboot_type not in [constants.INSTANCE_REBOOT_SOFT,
2693
                                   constants.INSTANCE_REBOOT_HARD,
2694
                                   constants.INSTANCE_REBOOT_FULL]:
2695
      raise errors.ParameterError("reboot type not in [%s, %s, %s]" %
2696
                                  (constants.INSTANCE_REBOOT_SOFT,
2697
                                   constants.INSTANCE_REBOOT_HARD,
2698
                                   constants.INSTANCE_REBOOT_FULL))
2699
    self._ExpandAndLockInstance()
2700

    
2701
  def BuildHooksEnv(self):
2702
    """Build hooks env.
2703

2704
    This runs on master, primary and secondary nodes of the instance.
2705

2706
    """
2707
    env = {
2708
      "IGNORE_SECONDARIES": self.op.ignore_secondaries,
2709
      }
2710
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2711
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2712
    return env, nl, nl
2713

    
2714
  def CheckPrereq(self):
2715
    """Check prerequisites.
2716

2717
    This checks that the instance is in the cluster.
2718

2719
    """
2720
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2721
    assert self.instance is not None, \
2722
      "Cannot retrieve locked instance %s" % self.op.instance_name
2723

    
2724
    _CheckNodeOnline(self, instance.primary_node)
2725

    
2726
    # check bridges existance
2727
    _CheckInstanceBridgesExist(self, instance)
2728

    
2729
  def Exec(self, feedback_fn):
2730
    """Reboot the instance.
2731

2732
    """
2733
    instance = self.instance
2734
    ignore_secondaries = self.op.ignore_secondaries
2735
    reboot_type = self.op.reboot_type
2736
    extra_args = getattr(self.op, "extra_args", "")
2737

    
2738
    node_current = instance.primary_node
2739

    
2740
    if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
2741
                       constants.INSTANCE_REBOOT_HARD]:
2742
      result = self.rpc.call_instance_reboot(node_current, instance,
2743
                                             reboot_type, extra_args)
2744
      if result.failed or not result.data:
2745
        raise errors.OpExecError("Could not reboot instance")
2746
    else:
2747
      if not self.rpc.call_instance_shutdown(node_current, instance):
2748
        raise errors.OpExecError("could not shutdown instance for full reboot")
2749
      _ShutdownInstanceDisks(self, instance)
2750
      _StartInstanceDisks(self, instance, ignore_secondaries)
2751
      result = self.rpc.call_instance_start(node_current, instance, extra_args)
2752
      msg = result.RemoteFailMsg()
2753
      if msg:
2754
        _ShutdownInstanceDisks(self, instance)
2755
        raise errors.OpExecError("Could not start instance for"
2756
                                 " full reboot: %s" % msg)
2757

    
2758
    self.cfg.MarkInstanceUp(instance.name)
2759

    
2760

    
2761
class LUShutdownInstance(LogicalUnit):
2762
  """Shutdown an instance.
2763

2764
  """
2765
  HPATH = "instance-stop"
2766
  HTYPE = constants.HTYPE_INSTANCE
2767
  _OP_REQP = ["instance_name"]
2768
  REQ_BGL = False
2769

    
2770
  def ExpandNames(self):
2771
    self._ExpandAndLockInstance()
2772

    
2773
  def BuildHooksEnv(self):
2774
    """Build hooks env.
2775

2776
    This runs on master, primary and secondary nodes of the instance.
2777

2778
    """
2779
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2780
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2781
    return env, nl, nl
2782

    
2783
  def CheckPrereq(self):
2784
    """Check prerequisites.
2785

2786
    This checks that the instance is in the cluster.
2787

2788
    """
2789
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2790
    assert self.instance is not None, \
2791
      "Cannot retrieve locked instance %s" % self.op.instance_name
2792
    _CheckNodeOnline(self, self.instance.primary_node)
2793

    
2794
  def Exec(self, feedback_fn):
2795
    """Shutdown the instance.
2796

2797
    """
2798
    instance = self.instance
2799
    node_current = instance.primary_node
2800
    self.cfg.MarkInstanceDown(instance.name)
2801
    result = self.rpc.call_instance_shutdown(node_current, instance)
2802
    if result.failed or not result.data:
2803
      self.proc.LogWarning("Could not shutdown instance")
2804

    
2805
    _ShutdownInstanceDisks(self, instance)
2806

    
2807

    
2808
class LUReinstallInstance(LogicalUnit):
2809
  """Reinstall an instance.
2810

2811
  """
2812
  HPATH = "instance-reinstall"
2813
  HTYPE = constants.HTYPE_INSTANCE
2814
  _OP_REQP = ["instance_name"]
2815
  REQ_BGL = False
2816

    
2817
  def ExpandNames(self):
2818
    self._ExpandAndLockInstance()
2819

    
2820
  def BuildHooksEnv(self):
2821
    """Build hooks env.
2822

2823
    This runs on master, primary and secondary nodes of the instance.
2824

2825
    """
2826
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2827
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2828
    return env, nl, nl
2829

    
2830
  def CheckPrereq(self):
2831
    """Check prerequisites.
2832

2833
    This checks that the instance is in the cluster and is not running.
2834

2835
    """
2836
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2837
    assert instance is not None, \
2838
      "Cannot retrieve locked instance %s" % self.op.instance_name
2839
    _CheckNodeOnline(self, instance.primary_node)
2840

    
2841
    if instance.disk_template == constants.DT_DISKLESS:
2842
      raise errors.OpPrereqError("Instance '%s' has no disks" %
2843
                                 self.op.instance_name)
2844
    if instance.admin_up:
2845
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
2846
                                 self.op.instance_name)
2847
    remote_info = self.rpc.call_instance_info(instance.primary_node,
2848
                                              instance.name,
2849
                                              instance.hypervisor)
2850
    if remote_info.failed or remote_info.data:
2851
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
2852
                                 (self.op.instance_name,
2853
                                  instance.primary_node))
2854

    
2855
    self.op.os_type = getattr(self.op, "os_type", None)
2856
    if self.op.os_type is not None:
2857
      # OS verification
2858
      pnode = self.cfg.GetNodeInfo(
2859
        self.cfg.ExpandNodeName(instance.primary_node))
2860
      if pnode is None:
2861
        raise errors.OpPrereqError("Primary node '%s' is unknown" %
2862
                                   self.op.pnode)
2863
      result = self.rpc.call_os_get(pnode.name, self.op.os_type)
2864
      result.Raise()
2865
      if not isinstance(result.data, objects.OS):
2866
        raise errors.OpPrereqError("OS '%s' not in supported OS list for"
2867
                                   " primary node"  % self.op.os_type)
2868

    
2869
    self.instance = instance
2870

    
2871
  def Exec(self, feedback_fn):
2872
    """Reinstall the instance.
2873

2874
    """
2875
    inst = self.instance
2876

    
2877
    if self.op.os_type is not None:
2878
      feedback_fn("Changing OS to '%s'..." % self.op.os_type)
2879
      inst.os = self.op.os_type
2880
      self.cfg.Update(inst)
2881

    
2882
    _StartInstanceDisks(self, inst, None)
2883
    try:
2884
      feedback_fn("Running the instance OS create scripts...")
2885
      result = self.rpc.call_instance_os_add(inst.primary_node, inst)
2886
      msg = result.RemoteFailMsg()
2887
      if msg:
2888
        raise errors.OpExecError("Could not install OS for instance %s"
2889
                                 " on node %s: %s" %
2890
                                 (inst.name, inst.primary_node, msg))
2891
    finally:
2892
      _ShutdownInstanceDisks(self, inst)
2893

    
2894

    
2895
class LURenameInstance(LogicalUnit):
2896
  """Rename an instance.
2897

2898
  """
2899
  HPATH = "instance-rename"
2900
  HTYPE = constants.HTYPE_INSTANCE
2901
  _OP_REQP = ["instance_name", "new_name"]
2902

    
2903
  def BuildHooksEnv(self):
2904
    """Build hooks env.
2905

2906
    This runs on master, primary and secondary nodes of the instance.
2907

2908
    """
2909
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2910
    env["INSTANCE_NEW_NAME"] = self.op.new_name
2911
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2912
    return env, nl, nl
2913

    
2914
  def CheckPrereq(self):
2915
    """Check prerequisites.
2916

2917
    This checks that the instance is in the cluster and is not running.
2918

2919
    """
2920
    instance = self.cfg.GetInstanceInfo(
2921
      self.cfg.ExpandInstanceName(self.op.instance_name))
2922
    if instance is None:
2923
      raise errors.OpPrereqError("Instance '%s' not known" %
2924
                                 self.op.instance_name)
2925
    _CheckNodeOnline(self, instance.primary_node)
2926

    
2927
    if instance.admin_up:
2928
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
2929
                                 self.op.instance_name)
2930
    remote_info = self.rpc.call_instance_info(instance.primary_node,
2931
                                              instance.name,
2932
                                              instance.hypervisor)
2933
    remote_info.Raise()
2934
    if remote_info.data:
2935
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
2936
                                 (self.op.instance_name,
2937
                                  instance.primary_node))
2938
    self.instance = instance
2939

    
2940
    # new name verification
2941
    name_info = utils.HostInfo(self.op.new_name)
2942

    
2943
    self.op.new_name = new_name = name_info.name
2944
    instance_list = self.cfg.GetInstanceList()
2945
    if new_name in instance_list:
2946
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
2947
                                 new_name)
2948

    
2949
    if not getattr(self.op, "ignore_ip", False):
2950
      if utils.TcpPing(name_info.ip, constants.DEFAULT_NODED_PORT):
2951
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
2952
                                   (name_info.ip, new_name))
2953

    
2954

    
2955
  def Exec(self, feedback_fn):
2956
    """Reinstall the instance.
2957

2958
    """
2959
    inst = self.instance
2960
    old_name = inst.name
2961

    
2962
    if inst.disk_template == constants.DT_FILE:
2963
      old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
2964

    
2965
    self.cfg.RenameInstance(inst.name, self.op.new_name)
2966
    # Change the instance lock. This is definitely safe while we hold the BGL
2967
    self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
2968
    self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
2969

    
2970
    # re-read the instance from the configuration after rename
2971
    inst = self.cfg.GetInstanceInfo(self.op.new_name)
2972

    
2973
    if inst.disk_template == constants.DT_FILE:
2974
      new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
2975
      result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
2976
                                                     old_file_storage_dir,
2977
                                                     new_file_storage_dir)
2978
      result.Raise()
2979
      if not result.data:
2980
        raise errors.OpExecError("Could not connect to node '%s' to rename"
2981
                                 " directory '%s' to '%s' (but the instance"
2982
                                 " has been renamed in Ganeti)" % (
2983
                                 inst.primary_node, old_file_storage_dir,
2984
                                 new_file_storage_dir))
2985

    
2986
      if not result.data[0]:
2987
        raise errors.OpExecError("Could not rename directory '%s' to '%s'"
2988
                                 " (but the instance has been renamed in"
2989
                                 " Ganeti)" % (old_file_storage_dir,
2990
                                               new_file_storage_dir))
2991

    
2992
    _StartInstanceDisks(self, inst, None)
2993
    try:
2994
      result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
2995
                                                 old_name)
2996
      msg = result.RemoteFailMsg()
2997
      if msg:
2998
        msg = ("Could not run OS rename script for instance %s on node %s"
2999
               " (but the instance has been renamed in Ganeti): %s" %
3000
               (inst.name, inst.primary_node, msg))
3001
        self.proc.LogWarning(msg)
3002
    finally:
3003
      _ShutdownInstanceDisks(self, inst)
3004

    
3005

    
3006
class LURemoveInstance(LogicalUnit):
3007
  """Remove an instance.
3008

3009
  """
3010
  HPATH = "instance-remove"
3011
  HTYPE = constants.HTYPE_INSTANCE
3012
  _OP_REQP = ["instance_name", "ignore_failures"]
3013
  REQ_BGL = False
3014

    
3015
  def ExpandNames(self):
3016
    self._ExpandAndLockInstance()
3017
    self.needed_locks[locking.LEVEL_NODE] = []
3018
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3019

    
3020
  def DeclareLocks(self, level):
3021
    if level == locking.LEVEL_NODE:
3022
      self._LockInstancesNodes()
3023

    
3024
  def BuildHooksEnv(self):
3025
    """Build hooks env.
3026

3027
    This runs on master, primary and secondary nodes of the instance.
3028

3029
    """
3030
    env = _BuildInstanceHookEnvByObject(self, self.instance)
3031
    nl = [self.cfg.GetMasterNode()]
3032
    return env, nl, nl
3033

    
3034
  def CheckPrereq(self):
3035
    """Check prerequisites.
3036

3037
    This checks that the instance is in the cluster.
3038

3039
    """
3040
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3041
    assert self.instance is not None, \
3042
      "Cannot retrieve locked instance %s" % self.op.instance_name
3043

    
3044
  def Exec(self, feedback_fn):
3045
    """Remove the instance.
3046

3047
    """
3048
    instance = self.instance
3049
    logging.info("Shutting down instance %s on node %s",
3050
                 instance.name, instance.primary_node)
3051

    
3052
    result = self.rpc.call_instance_shutdown(instance.primary_node, instance)
3053
    if result.failed or not result.data:
3054
      if self.op.ignore_failures:
3055
        feedback_fn("Warning: can't shutdown instance")
3056
      else:
3057
        raise errors.OpExecError("Could not shutdown instance %s on node %s" %
3058
                                 (instance.name, instance.primary_node))
3059

    
3060
    logging.info("Removing block devices for instance %s", instance.name)
3061

    
3062
    if not _RemoveDisks(self, instance):
3063
      if self.op.ignore_failures:
3064
        feedback_fn("Warning: can't remove instance's disks")
3065
      else:
3066
        raise errors.OpExecError("Can't remove instance's disks")
3067

    
3068
    logging.info("Removing instance %s out of cluster config", instance.name)
3069

    
3070
    self.cfg.RemoveInstance(instance.name)
3071
    self.remove_locks[locking.LEVEL_INSTANCE] = instance.name
3072

    
3073

    
3074
class LUQueryInstances(NoHooksLU):
3075
  """Logical unit for querying instances.
3076

3077
  """
3078
  _OP_REQP = ["output_fields", "names", "use_locking"]
3079
  REQ_BGL = False
3080
  _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
3081
                                    "admin_state", "admin_ram",
3082
                                    "disk_template", "ip", "mac", "bridge",
3083
                                    "sda_size", "sdb_size", "vcpus", "tags",
3084
                                    "network_port", "beparams",
3085
                                    "(disk).(size)/([0-9]+)",
3086
                                    "(disk).(sizes)",
3087
                                    "(nic).(mac|ip|bridge)/([0-9]+)",
3088
                                    "(nic).(macs|ips|bridges)",
3089
                                    "(disk|nic).(count)",
3090
                                    "serial_no", "hypervisor", "hvparams",] +
3091
                                  ["hv/%s" % name
3092
                                   for name in constants.HVS_PARAMETERS] +
3093
                                  ["be/%s" % name
3094
                                   for name in constants.BES_PARAMETERS])
3095
  _FIELDS_DYNAMIC = utils.FieldSet("oper_state", "oper_ram", "status")
3096

    
3097

    
3098
  def ExpandNames(self):
3099
    _CheckOutputFields(static=self._FIELDS_STATIC,
3100
                       dynamic=self._FIELDS_DYNAMIC,
3101
                       selected=self.op.output_fields)
3102

    
3103
    self.needed_locks = {}
3104
    self.share_locks[locking.LEVEL_INSTANCE] = 1
3105
    self.share_locks[locking.LEVEL_NODE] = 1
3106

    
3107
    if self.op.names:
3108
      self.wanted = _GetWantedInstances(self, self.op.names)
3109
    else:
3110
      self.wanted = locking.ALL_SET
3111

    
3112
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
3113
    self.do_locking = self.do_node_query and self.op.use_locking
3114
    if self.do_locking:
3115
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
3116
      self.needed_locks[locking.LEVEL_NODE] = []
3117
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3118

    
3119
  def DeclareLocks(self, level):
3120
    if level == locking.LEVEL_NODE and self.do_locking:
3121
      self._LockInstancesNodes()
3122

    
3123
  def CheckPrereq(self):
3124
    """Check prerequisites.
3125

3126
    """
3127
    pass
3128

    
3129
  def Exec(self, feedback_fn):
3130
    """Computes the list of nodes and their attributes.
3131

3132
    """
3133
    all_info = self.cfg.GetAllInstancesInfo()
3134
    if self.wanted == locking.ALL_SET:
3135
      # caller didn't specify instance names, so ordering is not important
3136
      if self.do_locking:
3137
        instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
3138
      else:
3139
        instance_names = all_info.keys()
3140
      instance_names = utils.NiceSort(instance_names)
3141
    else:
3142
      # caller did specify names, so we must keep the ordering
3143
      if self.do_locking:
3144
        tgt_set = self.acquired_locks[locking.LEVEL_INSTANCE]
3145
      else:
3146
        tgt_set = all_info.keys()
3147
      missing = set(self.wanted).difference(tgt_set)
3148
      if missing:
3149
        raise errors.OpExecError("Some instances were removed before"
3150
                                 " retrieving their data: %s" % missing)
3151
      instance_names = self.wanted
3152

    
3153
    instance_list = [all_info[iname] for iname in instance_names]
3154

    
3155
    # begin data gathering
3156

    
3157
    nodes = frozenset([inst.primary_node for inst in instance_list])
3158
    hv_list = list(set([inst.hypervisor for inst in instance_list]))
3159

    
3160
    bad_nodes = []
3161
    off_nodes = []
3162
    if self.do_node_query:
3163
      live_data = {}
3164
      node_data = self.rpc.call_all_instances_info(nodes, hv_list)
3165
      for name in nodes:
3166
        result = node_data[name]
3167
        if result.offline:
3168
          # offline nodes will be in both lists
3169
          off_nodes.append(name)
3170
        if result.failed:
3171
          bad_nodes.append(name)
3172
        else:
3173
          if result.data:
3174
            live_data.update(result.data)
3175
            # else no instance is alive
3176
    else:
3177
      live_data = dict([(name, {}) for name in instance_names])
3178

    
3179
    # end data gathering
3180

    
3181
    HVPREFIX = "hv/"
3182
    BEPREFIX = "be/"
3183
    output = []
3184
    for instance in instance_list:
3185
      iout = []
3186
      i_hv = self.cfg.GetClusterInfo().FillHV(instance)
3187
      i_be = self.cfg.GetClusterInfo().FillBE(instance)
3188
      for field in self.op.output_fields:
3189
        st_match = self._FIELDS_STATIC.Matches(field)
3190
        if field == "name":
3191
          val = instance.name
3192
        elif field == "os":
3193
          val = instance.os
3194
        elif field == "pnode":
3195
          val = instance.primary_node
3196
        elif field == "snodes":
3197
          val = list(instance.secondary_nodes)
3198
        elif field == "admin_state":
3199
          val = instance.admin_up
3200
        elif field == "oper_state":
3201
          if instance.primary_node in bad_nodes:
3202
            val = None
3203
          else:
3204
            val = bool(live_data.get(instance.name))
3205
        elif field == "status":
3206
          if instance.primary_node in off_nodes:
3207
            val = "ERROR_nodeoffline"
3208
          elif instance.primary_node in bad_nodes:
3209
            val = "ERROR_nodedown"
3210
          else:
3211
            running = bool(live_data.get(instance.name))
3212
            if running:
3213
              if instance.admin_up:
3214
                val = "running"
3215
              else:
3216
                val = "ERROR_up"
3217
            else:
3218
              if instance.admin_up:
3219
                val = "ERROR_down"
3220
              else:
3221
                val = "ADMIN_down"
3222
        elif field == "oper_ram":
3223
          if instance.primary_node in bad_nodes:
3224
            val = None
3225
          elif instance.name in live_data:
3226
            val = live_data[instance.name].get("memory", "?")
3227
          else:
3228
            val = "-"
3229
        elif field == "disk_template":
3230
          val = instance.disk_template
3231
        elif field == "ip":
3232
          val = instance.nics[0].ip
3233
        elif field == "bridge":
3234
          val = instance.nics[0].bridge
3235
        elif field == "mac":
3236
          val = instance.nics[0].mac
3237
        elif field == "sda_size" or field == "sdb_size":
3238
          idx = ord(field[2]) - ord('a')
3239
          try:
3240
            val = instance.FindDisk(idx).size
3241
          except errors.OpPrereqError:
3242
            val = None
3243
        elif field == "tags":
3244
          val = list(instance.GetTags())
3245
        elif field == "serial_no":
3246
          val = instance.serial_no
3247
        elif field == "network_port":
3248
          val = instance.network_port
3249
        elif field == "hypervisor":
3250
          val = instance.hypervisor
3251
        elif field == "hvparams":
3252
          val = i_hv
3253
        elif (field.startswith(HVPREFIX) and
3254
              field[len(HVPREFIX):] in constants.HVS_PARAMETERS):
3255
          val = i_hv.get(field[len(HVPREFIX):], None)
3256
        elif field == "beparams":
3257
          val = i_be
3258
        elif (field.startswith(BEPREFIX) and
3259
              field[len(BEPREFIX):] in constants.BES_PARAMETERS):
3260
          val = i_be.get(field[len(BEPREFIX):], None)
3261
        elif st_match and st_match.groups():
3262
          # matches a variable list
3263
          st_groups = st_match.groups()
3264
          if st_groups and st_groups[0] == "disk":
3265
            if st_groups[1] == "count":
3266
              val = len(instance.disks)
3267
            elif st_groups[1] == "sizes":
3268
              val = [disk.size for disk in instance.disks]
3269
            elif st_groups[1] == "size":
3270
              try:
3271
                val = instance.FindDisk(st_groups[2]).size
3272
              except errors.OpPrereqError:
3273
                val = None
3274
            else:
3275
              assert False, "Unhandled disk parameter"
3276
          elif st_groups[0] == "nic":
3277
            if st_groups[1] == "count":
3278
              val = len(instance.nics)
3279
            elif st_groups[1] == "macs":
3280
              val = [nic.mac for nic in instance.nics]
3281
            elif st_groups[1] == "ips":
3282
              val = [nic.ip for nic in instance.nics]
3283
            elif st_groups[1] == "bridges":
3284
              val = [nic.bridge for nic in instance.nics]
3285
            else:
3286
              # index-based item
3287
              nic_idx = int(st_groups[2])
3288
              if nic_idx >= len(instance.nics):
3289
                val = None
3290
              else:
3291
                if st_groups[1] == "mac":
3292
                  val = instance.nics[nic_idx].mac
3293
                elif st_groups[1] == "ip":
3294
                  val = instance.nics[nic_idx].ip
3295
                elif st_groups[1] == "bridge":
3296
                  val = instance.nics[nic_idx].bridge
3297
                else:
3298
                  assert False, "Unhandled NIC parameter"
3299
          else:
3300
            assert False, "Unhandled variable parameter"
3301
        else:
3302
          raise errors.ParameterError(field)
3303
        iout.append(val)
3304
      output.append(iout)
3305

    
3306
    return output
3307

    
3308

    
3309
class LUFailoverInstance(LogicalUnit):
3310
  """Failover an instance.
3311

3312
  """
3313
  HPATH = "instance-failover"
3314
  HTYPE = constants.HTYPE_INSTANCE
3315
  _OP_REQP = ["instance_name", "ignore_consistency"]
3316
  REQ_BGL = False
3317

    
3318
  def ExpandNames(self):
3319
    self._ExpandAndLockInstance()
3320
    self.needed_locks[locking.LEVEL_NODE] = []
3321
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3322

    
3323
  def DeclareLocks(self, level):
3324
    if level == locking.LEVEL_NODE:
3325
      self._LockInstancesNodes()
3326

    
3327
  def BuildHooksEnv(self):
3328
    """Build hooks env.
3329

3330
    This runs on master, primary and secondary nodes of the instance.
3331

3332
    """
3333
    env = {
3334
      "IGNORE_CONSISTENCY": self.op.ignore_consistency,
3335
      }
3336
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
3337
    nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
3338
    return env, nl, nl
3339

    
3340
  def CheckPrereq(self):
3341
    """Check prerequisites.
3342

3343
    This checks that the instance is in the cluster.
3344

3345
    """
3346
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3347
    assert self.instance is not None, \
3348
      "Cannot retrieve locked instance %s" % self.op.instance_name
3349

    
3350
    bep = self.cfg.GetClusterInfo().FillBE(instance)
3351
    if instance.disk_template not in constants.DTS_NET_MIRROR:
3352
      raise errors.OpPrereqError("Instance's disk layout is not"
3353
                                 " network mirrored, cannot failover.")
3354

    
3355
    secondary_nodes = instance.secondary_nodes
3356
    if not secondary_nodes:
3357
      raise errors.ProgrammerError("no secondary node but using "
3358
                                   "a mirrored disk template")
3359

    
3360
    target_node = secondary_nodes[0]
3361
    _CheckNodeOnline(self, target_node)
3362
    # check memory requirements on the secondary node
3363
    _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
3364
                         instance.name, bep[constants.BE_MEMORY],
3365
                         instance.hypervisor)
3366

    
3367
    # check bridge existance
3368
    brlist = [nic.bridge for nic in instance.nics]
3369
    result = self.rpc.call_bridges_exist(target_node, brlist)
3370
    result.Raise()
3371
    if not result.data:
3372
      raise errors.OpPrereqError("One or more target bridges %s does not"
3373
                                 " exist on destination node '%s'" %
3374
                                 (brlist, target_node))
3375

    
3376
  def Exec(self, feedback_fn):
3377
    """Failover an instance.
3378

3379
    The failover is done by shutting it down on its present node and
3380
    starting it on the secondary.
3381

3382
    """
3383
    instance = self.instance
3384

    
3385
    source_node = instance.primary_node
3386
    target_node = instance.secondary_nodes[0]
3387

    
3388
    feedback_fn("* checking disk consistency between source and target")
3389
    for dev in instance.disks:
3390
      # for drbd, these are drbd over lvm
3391
      if not _CheckDiskConsistency(self, dev, target_node, False):
3392
        if instance.admin_up and not self.op.ignore_consistency:
3393
          raise errors.OpExecError("Disk %s is degraded on target node,"
3394
                                   " aborting failover." % dev.iv_name)
3395

    
3396
    feedback_fn("* shutting down instance on source node")
3397
    logging.info("Shutting down instance %s on node %s",
3398
                 instance.name, source_node)
3399

    
3400
    result = self.rpc.call_instance_shutdown(source_node, instance)
3401
    if result.failed or not result.data:
3402
      if self.op.ignore_consistency:
3403
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
3404
                             " Proceeding"
3405
                             " anyway. Please make sure node %s is down",
3406
                             instance.name, source_node, source_node)
3407
      else:
3408
        raise errors.OpExecError("Could not shutdown instance %s on node %s" %
3409
                                 (instance.name, source_node))
3410

    
3411
    feedback_fn("* deactivating the instance's disks on source node")
3412
    if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
3413
      raise errors.OpExecError("Can't shut down the instance's disks.")
3414

    
3415
    instance.primary_node = target_node
3416
    # distribute new instance config to the other nodes
3417
    self.cfg.Update(instance)
3418

    
3419
    # Only start the instance if it's marked as up
3420
    if instance.admin_up:
3421
      feedback_fn("* activating the instance's disks on target node")
3422
      logging.info("Starting instance %s on node %s",
3423
                   instance.name, target_node)
3424

    
3425
      disks_ok, dummy = _AssembleInstanceDisks(self, instance,
3426
                                               ignore_secondaries=True)
3427
      if not disks_ok:
3428
        _ShutdownInstanceDisks(self, instance)
3429
        raise errors.OpExecError("Can't activate the instance's disks")
3430

    
3431
      feedback_fn("* starting the instance on the target node")
3432
      result = self.rpc.call_instance_start(target_node, instance, None)
3433
      msg = result.RemoteFailMsg()
3434
      if msg:
3435
        _ShutdownInstanceDisks(self, instance)
3436
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
3437
                                 (instance.name, target_node, msg))
3438

    
3439

    
3440
class LUMigrateInstance(LogicalUnit):
3441
  """Migrate an instance.
3442

3443
  This is migration without shutting down, compared to the failover,
3444
  which is done with shutdown.
3445

3446
  """
3447
  HPATH = "instance-migrate"
3448
  HTYPE = constants.HTYPE_INSTANCE
3449
  _OP_REQP = ["instance_name", "live", "cleanup"]
3450

    
3451
  REQ_BGL = False
3452

    
3453
  def ExpandNames(self):
3454
    self._ExpandAndLockInstance()
3455
    self.needed_locks[locking.LEVEL_NODE] = []
3456
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3457

    
3458
  def DeclareLocks(self, level):
3459
    if level == locking.LEVEL_NODE:
3460
      self._LockInstancesNodes()
3461

    
3462
  def BuildHooksEnv(self):
3463
    """Build hooks env.
3464

3465
    This runs on master, primary and secondary nodes of the instance.
3466

3467
    """
3468
    env = _BuildInstanceHookEnvByObject(self, self.instance)
3469
    nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
3470
    return env, nl, nl
3471

    
3472
  def CheckPrereq(self):
3473
    """Check prerequisites.
3474

3475
    This checks that the instance is in the cluster.
3476

3477
    """
3478
    instance = self.cfg.GetInstanceInfo(
3479
      self.cfg.ExpandInstanceName(self.op.instance_name))
3480
    if instance is None:
3481
      raise errors.OpPrereqError("Instance '%s' not known" %
3482
                                 self.op.instance_name)
3483

    
3484
    if instance.disk_template != constants.DT_DRBD8:
3485
      raise errors.OpPrereqError("Instance's disk layout is not"
3486
                                 " drbd8, cannot migrate.")
3487

    
3488
    secondary_nodes = instance.secondary_nodes
3489
    if not secondary_nodes:
3490
      raise errors.ProgrammerError("no secondary node but using "
3491
                                   "drbd8 disk template")
3492

    
3493
    i_be = self.cfg.GetClusterInfo().FillBE(instance)
3494

    
3495
    target_node = secondary_nodes[0]
3496
    # check memory requirements on the secondary node
3497
    _CheckNodeFreeMemory(self, target_node, "migrating instance %s" %
3498
                         instance.name, i_be[constants.BE_MEMORY],
3499
                         instance.hypervisor)
3500

    
3501
    # check bridge existance
3502
    brlist = [nic.bridge for nic in instance.nics]
3503
    result = self.rpc.call_bridges_exist(target_node, brlist)
3504
    if result.failed or not result.data:
3505
      raise errors.OpPrereqError("One or more target bridges %s does not"
3506
                                 " exist on destination node '%s'" %
3507
                                 (brlist, target_node))
3508

    
3509
    if not self.op.cleanup:
3510
      result = self.rpc.call_instance_migratable(instance.primary_node,
3511
                                                 instance)
3512
      msg = result.RemoteFailMsg()
3513
      if msg:
3514
        raise errors.OpPrereqError("Can't migrate: %s - please use failover" %
3515
                                   msg)
3516

    
3517
    self.instance = instance
3518

    
3519
  def _WaitUntilSync(self):
3520
    """Poll with custom rpc for disk sync.
3521

3522
    This uses our own step-based rpc call.
3523

3524
    """
3525
    self.feedback_fn("* wait until resync is done")
3526
    all_done = False
3527
    while not all_done:
3528
      all_done = True
3529
      result = self.rpc.call_drbd_wait_sync(self.all_nodes,
3530
                                            self.nodes_ip,
3531
                                            self.instance.disks)
3532
      min_percent = 100
3533
      for node, nres in result.items():
3534
        msg = nres.RemoteFailMsg()
3535
        if msg:
3536
          raise errors.OpExecError("Cannot resync disks on node %s: %s" %
3537
                                   (node, msg))
3538
        node_done, node_percent = nres.payload
3539
        all_done = all_done and node_done
3540
        if node_percent is not None:
3541
          min_percent = min(min_percent, node_percent)
3542
      if not all_done:
3543
        if min_percent < 100:
3544
          self.feedback_fn("   - progress: %.1f%%" % min_percent)
3545
        time.sleep(2)
3546

    
3547
  def _EnsureSecondary(self, node):
3548
    """Demote a node to secondary.
3549

3550
    """
3551
    self.feedback_fn("* switching node %s to secondary mode" % node)
3552

    
3553
    for dev in self.instance.disks:
3554
      self.cfg.SetDiskID(dev, node)
3555

    
3556
    result = self.rpc.call_blockdev_close(node, self.instance.name,
3557
                                          self.instance.disks)
3558
    msg = result.RemoteFailMsg()
3559
    if msg:
3560
      raise errors.OpExecError("Cannot change disk to secondary on node %s,"
3561
                               " error %s" % (node, msg))
3562

    
3563
  def _GoStandalone(self):
3564
    """Disconnect from the network.
3565

3566
    """
3567
    self.feedback_fn("* changing into standalone mode")
3568
    result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
3569
                                               self.instance.disks)
3570
    for node, nres in result.items():
3571
      msg = nres.RemoteFailMsg()
3572
      if msg:
3573
        raise errors.OpExecError("Cannot disconnect disks node %s,"
3574
                                 " error %s" % (node, msg))
3575

    
3576
  def _GoReconnect(self, multimaster):
3577
    """Reconnect to the network.
3578

3579
    """
3580
    if multimaster:
3581
      msg = "dual-master"
3582
    else:
3583
      msg = "single-master"
3584
    self.feedback_fn("* changing disks into %s mode" % msg)
3585
    result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
3586
                                           self.instance.disks,
3587
                                           self.instance.name, multimaster)
3588
    for node, nres in result.items():
3589
      msg = nres.RemoteFailMsg()
3590
      if msg:
3591
        raise errors.OpExecError("Cannot change disks config on node %s,"
3592
                                 " error: %s" % (node, msg))
3593

    
3594
  def _ExecCleanup(self):
3595
    """Try to cleanup after a failed migration.
3596

3597
    The cleanup is done by:
3598
      - check that the instance is running only on one node
3599
        (and update the config if needed)
3600
      - change disks on its secondary node to secondary
3601
      - wait until disks are fully synchronized
3602
      - disconnect from the network
3603
      - change disks into single-master mode
3604
      - wait again until disks are fully synchronized
3605

3606
    """
3607
    instance = self.instance
3608
    target_node = self.target_node
3609
    source_node = self.source_node
3610

    
3611
    # check running on only one node
3612
    self.feedback_fn("* checking where the instance actually runs"
3613
                     " (if this hangs, the hypervisor might be in"
3614
                     " a bad state)")
3615
    ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
3616
    for node, result in ins_l.items():
3617
      result.Raise()
3618
      if not isinstance(result.data, list):
3619
        raise errors.OpExecError("Can't contact node '%s'" % node)
3620

    
3621
    runningon_source = instance.name in ins_l[source_node].data
3622
    runningon_target = instance.name in ins_l[target_node].data
3623

    
3624
    if runningon_source and runningon_target:
3625
      raise errors.OpExecError("Instance seems to be running on two nodes,"
3626
                               " or the hypervisor is confused. You will have"
3627
                               " to ensure manually that it runs only on one"
3628
                               " and restart this operation.")
3629

    
3630
    if not (runningon_source or runningon_target):
3631
      raise errors.OpExecError("Instance does not seem to be running at all."
3632
                               " In this case, it's safer to repair by"
3633
                               " running 'gnt-instance stop' to ensure disk"
3634
                               " shutdown, and then restarting it.")
3635

    
3636
    if runningon_target:
3637
      # the migration has actually succeeded, we need to update the config
3638
      self.feedback_fn("* instance running on secondary node (%s),"
3639
                       " updating config" % target_node)
3640
      instance.primary_node = target_node
3641
      self.cfg.Update(instance)
3642
      demoted_node = source_node
3643
    else:
3644
      self.feedback_fn("* instance confirmed to be running on its"
3645
                       " primary node (%s)" % source_node)
3646
      demoted_node = target_node
3647

    
3648
    self._EnsureSecondary(demoted_node)
3649
    try:
3650
      self._WaitUntilSync()
3651
    except errors.OpExecError:
3652
      # we ignore here errors, since if the device is standalone, it
3653
      # won't be able to sync
3654
      pass
3655
    self._GoStandalone()
3656
    self._GoReconnect(False)
3657
    self._WaitUntilSync()
3658

    
3659
    self.feedback_fn("* done")
3660

    
3661
  def _RevertDiskStatus(self):
3662
    """Try to revert the disk status after a failed migration.
3663

3664
    """
3665
    target_node = self.target_node
3666
    try:
3667
      self._EnsureSecondary(target_node)
3668
      self._GoStandalone()
3669
      self._GoReconnect(False)
3670
      self._WaitUntilSync()
3671
    except errors.OpExecError, err:
3672
      self.LogWarning("Migration failed and I can't reconnect the"
3673
                      " drives: error '%s'\n"
3674
                      "Please look and recover the instance status" %
3675
                      str(err))
3676

    
3677
  def _AbortMigration(self):
3678
    """Call the hypervisor code to abort a started migration.
3679

3680
    """
3681
    instance = self.instance
3682
    target_node = self.target_node
3683
    migration_info = self.migration_info
3684

    
3685
    abort_result = self.rpc.call_finalize_migration(target_node,
3686
                                                    instance,
3687
                                                    migration_info,
3688
                                                    False)
3689
    abort_msg = abort_result.RemoteFailMsg()
3690
    if abort_msg:
3691
      logging.error("Aborting migration failed on target node %s: %s" %
3692
                    (target_node, abort_msg))
3693
      # Don't raise an exception here, as we stil have to try to revert the
3694
      # disk status, even if this step failed.
3695

    
3696
  def _ExecMigration(self):
3697
    """Migrate an instance.
3698

3699
    The migrate is done by:
3700
      - change the disks into dual-master mode
3701
      - wait until disks are fully synchronized again
3702
      - migrate the instance
3703
      - change disks on the new secondary node (the old primary) to secondary
3704
      - wait until disks are fully synchronized
3705
      - change disks into single-master mode
3706

3707
    """
3708
    instance = self.instance
3709
    target_node = self.target_node
3710
    source_node = self.source_node
3711

    
3712
    self.feedback_fn("* checking disk consistency between source and target")
3713
    for dev in instance.disks:
3714
      if not _CheckDiskConsistency(self, dev, target_node, False):
3715
        raise errors.OpExecError("Disk %s is degraded or not fully"
3716
                                 " synchronized on target node,"
3717
                                 " aborting migrate." % dev.iv_name)
3718

    
3719
    # First get the migration information from the remote node
3720
    result = self.rpc.call_migration_info(source_node, instance)
3721
    msg = result.RemoteFailMsg()
3722
    if msg:
3723
      log_err = ("Failed fetching source migration information from %s: %s" %
3724
                 (source_node, msg))
3725
      logging.error(log_err)
3726
      raise errors.OpExecError(log_err)
3727

    
3728
    self.migration_info = migration_info = result.payload
3729

    
3730
    # Then switch the disks to master/master mode
3731
    self._EnsureSecondary(target_node)
3732
    self._GoStandalone()
3733
    self._GoReconnect(True)
3734
    self._WaitUntilSync()
3735

    
3736
    self.feedback_fn("* preparing %s to accept the instance" % target_node)
3737
    result = self.rpc.call_accept_instance(target_node,
3738
                                           instance,
3739
                                           migration_info,
3740
                                           self.nodes_ip[target_node])
3741

    
3742
    msg = result.RemoteFailMsg()
3743
    if msg:
3744
      logging.error("Instance pre-migration failed, trying to revert"
3745
                    " disk status: %s", msg)
3746
      self._AbortMigration()
3747
      self._RevertDiskStatus()
3748
      raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
3749
                               (instance.name, msg))
3750

    
3751
    self.feedback_fn("* migrating instance to %s" % target_node)
3752
    time.sleep(10)
3753
    result = self.rpc.call_instance_migrate(source_node, instance,
3754
                                            self.nodes_ip[target_node],
3755
                                            self.op.live)
3756
    msg = result.RemoteFailMsg()
3757
    if msg:
3758
      logging.error("Instance migration failed, trying to revert"
3759
                    " disk status: %s", msg)
3760
      self._AbortMigration()
3761
      self._RevertDiskStatus()
3762
      raise errors.OpExecError("Could not migrate instance %s: %s" %
3763
                               (instance.name, msg))
3764
    time.sleep(10)
3765

    
3766
    instance.primary_node = target_node
3767
    # distribute new instance config to the other nodes
3768
    self.cfg.Update(instance)
3769

    
3770
    result = self.rpc.call_finalize_migration(target_node,
3771
                                              instance,
3772
                                              migration_info,
3773
                                              True)
3774
    msg = result.RemoteFailMsg()
3775
    if msg:
3776
      logging.error("Instance migration succeeded, but finalization failed:"
3777
                    " %s" % msg)
3778
      raise errors.OpExecError("Could not finalize instance migration: %s" %
3779
                               msg)
3780

    
3781
    self._EnsureSecondary(source_node)
3782
    self._WaitUntilSync()
3783
    self._GoStandalone()
3784
    self._GoReconnect(False)
3785
    self._WaitUntilSync()
3786

    
3787
    self.feedback_fn("* done")
3788

    
3789
  def Exec(self, feedback_fn):
3790
    """Perform the migration.
3791

3792
    """
3793
    self.feedback_fn = feedback_fn
3794

    
3795
    self.source_node = self.instance.primary_node
3796
    self.target_node = self.instance.secondary_nodes[0]
3797
    self.all_nodes = [self.source_node, self.target_node]
3798
    self.nodes_ip = {
3799
      self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
3800
      self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
3801
      }
3802
    if self.op.cleanup:
3803
      return self._ExecCleanup()
3804
    else:
3805
      return self._ExecMigration()
3806

    
3807

    
3808
def _CreateBlockDev(lu, node, instance, device, force_create,
3809
                    info, force_open):
3810
  """Create a tree of block devices on a given node.
3811

3812
  If this device type has to be created on secondaries, create it and
3813
  all its children.
3814

3815
  If not, just recurse to children keeping the same 'force' value.
3816

3817
  @param lu: the lu on whose behalf we execute
3818
  @param node: the node on which to create the device
3819
  @type instance: L{objects.Instance}
3820
  @param instance: the instance which owns the device
3821
  @type device: L{objects.Disk}
3822
  @param device: the device to create
3823
  @type force_create: boolean
3824
  @param force_create: whether to force creation of this device; this
3825
      will be change to True whenever we find a device which has
3826
      CreateOnSecondary() attribute
3827
  @param info: the extra 'metadata' we should attach to the device
3828
      (this will be represented as a LVM tag)
3829
  @type force_open: boolean
3830
  @param force_open: this parameter will be passes to the
3831
      L{backend.CreateBlockDevice} function where it specifies
3832
      whether we run on primary or not, and it affects both
3833
      the child assembly and the device own Open() execution
3834

3835
  """
3836
  if device.CreateOnSecondary():
3837
    force_create = True
3838

    
3839
  if device.children:
3840
    for child in device.children:
3841
      _CreateBlockDev(lu, node, instance, child, force_create,
3842
                      info, force_open)
3843

    
3844
  if not force_create:
3845
    return
3846

    
3847
  _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
3848

    
3849

    
3850
def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
3851
  """Create a single block device on a given node.
3852

3853
  This will not recurse over children of the device, so they must be
3854
  created in advance.
3855

3856
  @param lu: the lu on whose behalf we execute
3857
  @param node: the node on which to create the device
3858
  @type instance: L{objects.Instance}
3859
  @param instance: the instance which owns the device
3860
  @type device: L{objects.Disk}
3861
  @param device: the device to create
3862
  @param info: the extra 'metadata' we should attach to the device
3863
      (this will be represented as a LVM tag)
3864
  @type force_open: boolean
3865
  @param force_open: this parameter will be passes to the
3866
      L{backend.CreateBlockDevice} function where it specifies
3867
      whether we run on primary or not, and it affects both
3868
      the child assembly and the device own Open() execution
3869

3870
  """
3871
  lu.cfg.SetDiskID(device, node)
3872
  result = lu.rpc.call_blockdev_create(node, device, device.size,
3873
                                       instance.name, force_open, info)
3874
  msg = result.RemoteFailMsg()
3875
  if msg:
3876
    raise errors.OpExecError("Can't create block device %s on"
3877
                             " node %s for instance %s: %s" %
3878
                             (device, node, instance.name, msg))
3879
  if device.physical_id is None:
3880
    device.physical_id = result.payload
3881

    
3882

    
3883
def _GenerateUniqueNames(lu, exts):
3884
  """Generate a suitable LV name.
3885

3886
  This will generate a logical volume name for the given instance.
3887

3888
  """
3889
  results = []
3890
  for val in exts:
3891
    new_id = lu.cfg.GenerateUniqueID()
3892
    results.append("%s%s" % (new_id, val))
3893
  return results
3894

    
3895

    
3896
def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
3897
                         p_minor, s_minor):
3898
  """Generate a drbd8 device complete with its children.
3899

3900
  """
3901
  port = lu.cfg.AllocatePort()
3902
  vgname = lu.cfg.GetVGName()
3903
  shared_secret = lu.cfg.GenerateDRBDSecret()
3904
  dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
3905
                          logical_id=(vgname, names[0]))
3906
  dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
3907
                          logical_id=(vgname, names[1]))
3908
  drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
3909
                          logical_id=(primary, secondary, port,
3910
                                      p_minor, s_minor,
3911
                                      shared_secret),
3912
                          children=[dev_data, dev_meta],
3913
                          iv_name=iv_name)
3914
  return drbd_dev
3915

    
3916

    
3917
def _GenerateDiskTemplate(lu, template_name,
3918
                          instance_name, primary_node,
3919
                          secondary_nodes, disk_info,
3920
                          file_storage_dir, file_driver,
3921
                          base_index):
3922
  """Generate the entire disk layout for a given template type.
3923

3924
  """
3925
  #TODO: compute space requirements
3926

    
3927
  vgname = lu.cfg.GetVGName()
3928
  disk_count = len(disk_info)
3929
  disks = []
3930
  if template_name == constants.DT_DISKLESS:
3931
    pass
3932
  elif template_name == constants.DT_PLAIN:
3933
    if len(secondary_nodes) != 0:
3934
      raise errors.ProgrammerError("Wrong template configuration")
3935

    
3936
    names = _GenerateUniqueNames(lu, [".disk%d" % i
3937
                                      for i in range(disk_count)])
3938
    for idx, disk in enumerate(disk_info):
3939
      disk_index = idx + base_index
3940
      disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
3941
                              logical_id=(vgname, names[idx]),
3942
                              iv_name="disk/%d" % disk_index,
3943
                              mode=disk["mode"])
3944
      disks.append(disk_dev)
3945
  elif template_name == constants.DT_DRBD8:
3946
    if len(secondary_nodes) != 1:
3947
      raise errors.ProgrammerError("Wrong template configuration")
3948
    remote_node = secondary_nodes[0]
3949
    minors = lu.cfg.AllocateDRBDMinor(
3950
      [primary_node, remote_node] * len(disk_info), instance_name)
3951

    
3952
    names = []
3953
    for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % i
3954
                                               for i in range(disk_count)]):
3955
      names.append(lv_prefix + "_data")
3956
      names.append(lv_prefix + "_meta")
3957
    for idx, disk in enumerate(disk_info):
3958
      disk_index = idx + base_index
3959
      disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
3960
                                      disk["size"], names[idx*2:idx*2+2],
3961
                                      "disk/%d" % disk_index,
3962
                                      minors[idx*2], minors[idx*2+1])
3963
      disk_dev.mode = disk["mode"]
3964
      disks.append(disk_dev)
3965
  elif template_name == constants.DT_FILE:
3966
    if len(secondary_nodes) != 0:
3967
      raise errors.ProgrammerError("Wrong template configuration")
3968

    
3969
    for idx, disk in enumerate(disk_info):
3970
      disk_index = idx + base_index
3971
      disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
3972
                              iv_name="disk/%d" % disk_index,
3973
                              logical_id=(file_driver,
3974
                                          "%s/disk%d" % (file_storage_dir,
3975
                                                         idx)),
3976
                              mode=disk["mode"])
3977
      disks.append(disk_dev)
3978
  else:
3979
    raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
3980
  return disks
3981

    
3982

    
3983
def _GetInstanceInfoText(instance):
3984
  """Compute that text that should be added to the disk's metadata.
3985

3986
  """
3987
  return "originstname+%s" % instance.name
3988

    
3989

    
3990
def _CreateDisks(lu, instance):
3991
  """Create all disks for an instance.
3992

3993
  This abstracts away some work from AddInstance.
3994

3995
  @type lu: L{LogicalUnit}
3996
  @param lu: the logical unit on whose behalf we execute
3997
  @type instance: L{objects.Instance}
3998
  @param instance: the instance whose disks we should create
3999
  @rtype: boolean
4000
  @return: the success of the creation
4001

4002
  """
4003
  info = _GetInstanceInfoText(instance)
4004
  pnode = instance.primary_node
4005

    
4006
  if instance.disk_template == constants.DT_FILE:
4007
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
4008
    result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
4009

    
4010
    if result.failed or not result.data:
4011
      raise errors.OpExecError("Could not connect to node '%s'" % pnode)
4012

    
4013
    if not result.data[0]:
4014
      raise errors.OpExecError("Failed to create directory '%s'" %
4015
                               file_storage_dir)
4016

    
4017
  # Note: this needs to be kept in sync with adding of disks in
4018
  # LUSetInstanceParams
4019
  for device in instance.disks:
4020
    logging.info("Creating volume %s for instance %s",
4021
                 device.iv_name, instance.name)
4022
    #HARDCODE
4023
    for node in instance.all_nodes:
4024
      f_create = node == pnode
4025
      _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
4026

    
4027

    
4028
def _RemoveDisks(lu, instance):
4029
  """Remove all disks for an instance.
4030

4031
  This abstracts away some work from `AddInstance()` and
4032
  `RemoveInstance()`. Note that in case some of the devices couldn't
4033
  be removed, the removal will continue with the other ones (compare
4034
  with `_CreateDisks()`).
4035

4036
  @type lu: L{LogicalUnit}
4037
  @param lu: the logical unit on whose behalf we execute
4038
  @type instance: L{objects.Instance}
4039
  @param instance: the instance whose disks we should remove
4040
  @rtype: boolean
4041
  @return: the success of the removal
4042

4043
  """
4044
  logging.info("Removing block devices for instance %s", instance.name)
4045

    
4046
  result = True
4047
  for device in instance.disks:
4048
    for node, disk in device.ComputeNodeTree(instance.primary_node):
4049
      lu.cfg.SetDiskID(disk, node)
4050
      result = lu.rpc.call_blockdev_remove(node, disk)
4051
      if result.failed or not result.data:
4052
        lu.proc.LogWarning("Could not remove block device %s on node %s,"
4053
                           " continuing anyway", device.iv_name, node)
4054
        result = False
4055

    
4056
  if instance.disk_template == constants.DT_FILE:
4057
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
4058
    result = lu.rpc.call_file_storage_dir_remove(instance.primary_node,
4059
                                                 file_storage_dir)
4060
    if result.failed or not result.data:
4061
      logging.error("Could not remove directory '%s'", file_storage_dir)
4062
      result = False
4063

    
4064
  return result
4065

    
4066

    
4067
def _ComputeDiskSize(disk_template, disks):
4068
  """Compute disk size requirements in the volume group
4069

4070
  """
4071
  # Required free disk space as a function of disk and swap space
4072
  req_size_dict = {
4073
    constants.DT_DISKLESS: None,
4074
    constants.DT_PLAIN: sum(d["size"] for d in disks),
4075
    # 128 MB are added for drbd metadata for each disk
4076
    constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
4077
    constants.DT_FILE: None,
4078
  }
4079

    
4080
  if disk_template not in req_size_dict:
4081
    raise errors.ProgrammerError("Disk template '%s' size requirement"
4082
                                 " is unknown" %  disk_template)
4083

    
4084
  return req_size_dict[disk_template]
4085

    
4086

    
4087
def _CheckHVParams(lu, nodenames, hvname, hvparams):
4088
  """Hypervisor parameter validation.
4089

4090
  This function abstract the hypervisor parameter validation to be
4091
  used in both instance create and instance modify.
4092

4093
  @type lu: L{LogicalUnit}
4094
  @param lu: the logical unit for which we check
4095
  @type nodenames: list
4096
  @param nodenames: the list of nodes on which we should check
4097
  @type hvname: string
4098
  @param hvname: the name of the hypervisor we should use
4099
  @type hvparams: dict
4100
  @param hvparams: the parameters which we need to check
4101
  @raise errors.OpPrereqError: if the parameters are not valid
4102

4103
  """
4104
  hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
4105
                                                  hvname,
4106
                                                  hvparams)
4107
  for node in nodenames:
4108
    info = hvinfo[node]
4109
    if info.offline:
4110
      continue
4111
    msg = info.RemoteFailMsg()
4112
    if msg:
4113
      raise errors.OpPrereqError("Hypervisor parameter validation failed:"
4114
                                 " %s" % msg)
4115

    
4116

    
4117
class LUCreateInstance(LogicalUnit):
4118
  """Create an instance.
4119

4120
  """
4121
  HPATH = "instance-add"
4122
  HTYPE = constants.HTYPE_INSTANCE
4123
  _OP_REQP = ["instance_name", "disks", "disk_template",
4124
              "mode", "start",
4125
              "wait_for_sync", "ip_check", "nics",
4126
              "hvparams", "beparams"]
4127
  REQ_BGL = False
4128

    
4129
  def _ExpandNode(self, node):
4130
    """Expands and checks one node name.
4131

4132
    """
4133
    node_full = self.cfg.ExpandNodeName(node)
4134
    if node_full is None:
4135
      raise errors.OpPrereqError("Unknown node %s" % node)
4136
    return node_full
4137

    
4138
  def ExpandNames(self):
4139
    """ExpandNames for CreateInstance.
4140

4141
    Figure out the right locks for instance creation.
4142

4143
    """
4144
    self.needed_locks = {}
4145

    
4146
    # set optional parameters to none if they don't exist
4147
    for attr in ["pnode", "snode", "iallocator", "hypervisor"]:
4148
      if not hasattr(self.op, attr):
4149
        setattr(self.op, attr, None)
4150

    
4151
    # cheap checks, mostly valid constants given
4152

    
4153
    # verify creation mode
4154
    if self.op.mode not in (constants.INSTANCE_CREATE,
4155
                            constants.INSTANCE_IMPORT):
4156
      raise errors.OpPrereqError("Invalid instance creation mode '%s'" %
4157
                                 self.op.mode)
4158

    
4159
    # disk template and mirror node verification
4160
    if self.op.disk_template not in constants.DISK_TEMPLATES:
4161
      raise errors.OpPrereqError("Invalid disk template name")
4162

    
4163
    if self.op.hypervisor is None:
4164
      self.op.hypervisor = self.cfg.GetHypervisorType()
4165

    
4166
    cluster = self.cfg.GetClusterInfo()
4167
    enabled_hvs = cluster.enabled_hypervisors
4168
    if self.op.hypervisor not in enabled_hvs:
4169
      raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
4170
                                 " cluster (%s)" % (self.op.hypervisor,
4171
                                  ",".join(enabled_hvs)))
4172

    
4173
    # check hypervisor parameter syntax (locally)
4174

    
4175
    filled_hvp = cluster.FillDict(cluster.hvparams[self.op.hypervisor],
4176
                                  self.op.hvparams)
4177
    hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
4178
    hv_type.CheckParameterSyntax(filled_hvp)
4179

    
4180
    # fill and remember the beparams dict
4181
    utils.CheckBEParams(self.op.beparams)
4182
    self.be_full = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
4183
                                    self.op.beparams)
4184

    
4185
    #### instance parameters check
4186

    
4187
    # instance name verification
4188
    hostname1 = utils.HostInfo(self.op.instance_name)
4189
    self.op.instance_name = instance_name = hostname1.name
4190

    
4191
    # this is just a preventive check, but someone might still add this
4192
    # instance in the meantime, and creation will fail at lock-add time
4193
    if instance_name in self.cfg.GetInstanceList():
4194
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
4195
                                 instance_name)
4196

    
4197
    self.add_locks[locking.LEVEL_INSTANCE] = instance_name
4198

    
4199
    # NIC buildup
4200
    self.nics = []
4201
    for nic in self.op.nics:
4202
      # ip validity checks
4203
      ip = nic.get("ip", None)
4204
      if ip is None or ip.lower() == "none":
4205
        nic_ip = None
4206
      elif ip.lower() == constants.VALUE_AUTO:
4207
        nic_ip = hostname1.ip
4208
      else:
4209
        if not utils.IsValidIP(ip):
4210
          raise errors.OpPrereqError("Given IP address '%s' doesn't look"
4211
                                     " like a valid IP" % ip)
4212
        nic_ip = ip
4213

    
4214
      # MAC address verification
4215
      mac = nic.get("mac", constants.VALUE_AUTO)
4216
      if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
4217
        if not utils.IsValidMac(mac.lower()):
4218
          raise errors.OpPrereqError("Invalid MAC address specified: %s" %
4219
                                     mac)
4220
      # bridge verification
4221
      bridge = nic.get("bridge", None)
4222
      if bridge is None:
4223
        bridge = self.cfg.GetDefBridge()
4224
      self.nics.append(objects.NIC(mac=mac, ip=nic_ip, bridge=bridge))
4225

    
4226
    # disk checks/pre-build
4227
    self.disks = []
4228
    for disk in self.op.disks:
4229
      mode = disk.get("mode", constants.DISK_RDWR)
4230
      if mode not in constants.DISK_ACCESS_SET:
4231
        raise errors.OpPrereqError("Invalid disk access mode '%s'" %
4232
                                   mode)
4233
      size = disk.get("size", None)
4234
      if size is None:
4235
        raise errors.OpPrereqError("Missing disk size")
4236
      try:
4237
        size = int(size)
4238
      except ValueError:
4239
        raise errors.OpPrereqError("Invalid disk size '%s'" % size)
4240
      self.disks.append({"size": size, "mode": mode})
4241

    
4242
    # used in CheckPrereq for ip ping check
4243
    self.check_ip = hostname1.ip
4244

    
4245
    # file storage checks
4246
    if (self.op.file_driver and
4247
        not self.op.file_driver in constants.FILE_DRIVER):
4248
      raise errors.OpPrereqError("Invalid file driver name '%s'" %
4249
                                 self.op.file_driver)
4250

    
4251
    if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
4252
      raise errors.OpPrereqError("File storage directory path not absolute")
4253

    
4254
    ### Node/iallocator related checks
4255
    if [self.op.iallocator, self.op.pnode].count(None) != 1:
4256
      raise errors.OpPrereqError("One and only one of iallocator and primary"
4257
                                 " node must be given")
4258

    
4259
    if self.op.iallocator:
4260
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4261
    else:
4262
      self.op.pnode = self._ExpandNode(self.op.pnode)
4263
      nodelist = [self.op.pnode]
4264
      if self.op.snode is not None:
4265
        self.op.snode = self._ExpandNode(self.op.snode)
4266
        nodelist.append(self.op.snode)
4267
      self.needed_locks[locking.LEVEL_NODE] = nodelist
4268

    
4269
    # in case of import lock the source node too
4270
    if self.op.mode == constants.INSTANCE_IMPORT:
4271
      src_node = getattr(self.op, "src_node", None)
4272
      src_path = getattr(self.op, "src_path", None)
4273

    
4274
      if src_path is None:
4275
        self.op.src_path = src_path = self.op.instance_name
4276

    
4277
      if src_node is None:
4278
        self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4279
        self.op.src_node = None
4280
        if os.path.isabs(src_path):
4281
          raise errors.OpPrereqError("Importing an instance from an absolute"
4282
                                     " path requires a source node option.")
4283
      else:
4284
        self.op.src_node = src_node = self._ExpandNode(src_node)
4285
        if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
4286
          self.needed_locks[locking.LEVEL_NODE].append(src_node)
4287
        if not os.path.isabs(src_path):
4288
          self.op.src_path = src_path = \
4289
            os.path.join(constants.EXPORT_DIR, src_path)
4290

    
4291
    else: # INSTANCE_CREATE
4292
      if getattr(self.op, "os_type", None) is None:
4293
        raise errors.OpPrereqError("No guest OS specified")
4294

    
4295
  def _RunAllocator(self):
4296
    """Run the allocator based on input opcode.
4297

4298
    """
4299
    nics = [n.ToDict() for n in self.nics]
4300
    ial = IAllocator(self,
4301
                     mode=constants.IALLOCATOR_MODE_ALLOC,
4302
                     name=self.op.instance_name,
4303
                     disk_template=self.op.disk_template,
4304
                     tags=[],
4305
                     os=self.op.os_type,
4306
                     vcpus=self.be_full[constants.BE_VCPUS],
4307
                     mem_size=self.be_full[constants.BE_MEMORY],
4308
                     disks=self.disks,
4309
                     nics=nics,
4310
                     hypervisor=self.op.hypervisor,
4311
                     )
4312

    
4313
    ial.Run(self.op.iallocator)
4314

    
4315
    if not ial.success:
4316
      raise errors.OpPrereqError("Can't compute nodes using"
4317
                                 " iallocator '%s': %s" % (self.op.iallocator,
4318
                                                           ial.info))
4319
    if len(ial.nodes) != ial.required_nodes:
4320
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
4321
                                 " of nodes (%s), required %s" %
4322
                                 (self.op.iallocator, len(ial.nodes),
4323
                                  ial.required_nodes))
4324
    self.op.pnode = ial.nodes[0]
4325
    self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
4326
                 self.op.instance_name, self.op.iallocator,
4327
                 ", ".join(ial.nodes))
4328
    if ial.required_nodes == 2:
4329
      self.op.snode = ial.nodes[1]
4330

    
4331
  def BuildHooksEnv(self):
4332
    """Build hooks env.
4333

4334
    This runs on master, primary and secondary nodes of the instance.
4335

4336
    """
4337
    env = {
4338
      "INSTANCE_DISK_TEMPLATE": self.op.disk_template,
4339
      "INSTANCE_DISK_SIZE": ",".join(str(d["size"]) for d in self.disks),
4340
      "INSTANCE_ADD_MODE": self.op.mode,
4341
      }
4342
    if self.op.mode == constants.INSTANCE_IMPORT:
4343
      env["INSTANCE_SRC_NODE"] = self.op.src_node
4344
      env["INSTANCE_SRC_PATH"] = self.op.src_path
4345
      env["INSTANCE_SRC_IMAGES"] = self.src_images
4346

    
4347
    env.update(_BuildInstanceHookEnv(name=self.op.instance_name,
4348
      primary_node=self.op.pnode,
4349
      secondary_nodes=self.secondaries,
4350
      status=self.op.start,
4351
      os_type=self.op.os_type,
4352
      memory=self.be_full[constants.BE_MEMORY],
4353
      vcpus=self.be_full[constants.BE_VCPUS],
4354
      nics=[(n.ip, n.bridge, n.mac) for n in self.nics],
4355
    ))
4356

    
4357
    nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
4358
          self.secondaries)
4359
    return env, nl, nl
4360

    
4361

    
4362
  def CheckPrereq(self):
4363
    """Check prerequisites.
4364

4365
    """
4366
    if (not self.cfg.GetVGName() and
4367
        self.op.disk_template not in constants.DTS_NOT_LVM):
4368
      raise errors.OpPrereqError("Cluster does not support lvm-based"
4369
                                 " instances")
4370

    
4371

    
4372
    if self.op.mode == constants.INSTANCE_IMPORT:
4373
      src_node = self.op.src_node
4374
      src_path = self.op.src_path
4375

    
4376
      if src_node is None:
4377
        exp_list = self.rpc.call_export_list(
4378
          self.acquired_locks[locking.LEVEL_NODE])
4379
        found = False
4380
        for node in exp_list:
4381
          if not exp_list[node].failed and src_path in exp_list[node].data:
4382
            found = True
4383
            self.op.src_node = src_node = node
4384
            self.op.src_path = src_path = os.path.join(constants.EXPORT_DIR,
4385
                                                       src_path)
4386
            break
4387
        if not found:
4388
          raise errors.OpPrereqError("No export found for relative path %s" %
4389
                                      src_path)
4390

    
4391
      _CheckNodeOnline(self, src_node)
4392
      result = self.rpc.call_export_info(src_node, src_path)
4393
      result.Raise()
4394
      if not result.data:
4395
        raise errors.OpPrereqError("No export found in dir %s" % src_path)
4396

    
4397
      export_info = result.data
4398
      if not export_info.has_section(constants.INISECT_EXP):
4399
        raise errors.ProgrammerError("Corrupted export config")
4400

    
4401
      ei_version = export_info.get(constants.INISECT_EXP, 'version')
4402
      if (int(ei_version) != constants.EXPORT_VERSION):
4403
        raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
4404
                                   (ei_version, constants.EXPORT_VERSION))
4405

    
4406
      # Check that the new instance doesn't have less disks than the export
4407
      instance_disks = len(self.disks)
4408
      export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
4409
      if instance_disks < export_disks:
4410
        raise errors.OpPrereqError("Not enough disks to import."
4411
                                   " (instance: %d, export: %d)" %
4412
                                   (instance_disks, export_disks))
4413

    
4414
      self.op.os_type = export_info.get(constants.INISECT_EXP, 'os')
4415
      disk_images = []
4416
      for idx in range(export_disks):
4417
        option = 'disk%d_dump' % idx
4418
        if export_info.has_option(constants.INISECT_INS, option):
4419
          # FIXME: are the old os-es, disk sizes, etc. useful?
4420
          export_name = export_info.get(constants.INISECT_INS, option)
4421
          image = os.path.join(src_path, export_name)
4422
          disk_images.append(image)
4423
        else:
4424
          disk_images.append(False)
4425

    
4426
      self.src_images = disk_images
4427

    
4428
      old_name = export_info.get(constants.INISECT_INS, 'name')
4429
      # FIXME: int() here could throw a ValueError on broken exports
4430
      exp_nic_count = int(export_info.get(constants.INISECT_INS, 'nic_count'))
4431
      if self.op.instance_name == old_name:
4432
        for idx, nic in enumerate(self.nics):
4433
          if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
4434
            nic_mac_ini = 'nic%d_mac' % idx
4435
            nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
4436

    
4437
    # ip ping checks (we use the same ip that was resolved in ExpandNames)
4438
    if self.op.start and not self.op.ip_check:
4439
      raise errors.OpPrereqError("Cannot ignore IP address conflicts when"
4440
                                 " adding an instance in start mode")
4441

    
4442
    if self.op.ip_check:
4443
      if utils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
4444
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
4445
                                   (self.check_ip, self.op.instance_name))
4446

    
4447
    #### allocator run
4448

    
4449
    if self.op.iallocator is not None:
4450
      self._RunAllocator()
4451

    
4452
    #### node related checks
4453

    
4454
    # check primary node
4455
    self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
4456
    assert self.pnode is not None, \
4457
      "Cannot retrieve locked node %s" % self.op.pnode
4458
    if pnode.offline:
4459
      raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
4460
                                 pnode.name)
4461

    
4462
    self.secondaries = []
4463

    
4464
    # mirror node verification
4465
    if self.op.disk_template in constants.DTS_NET_MIRROR:
4466
      if self.op.snode is None:
4467
        raise errors.OpPrereqError("The networked disk templates need"
4468
                                   " a mirror node")
4469
      if self.op.snode == pnode.name:
4470
        raise errors.OpPrereqError("The secondary node cannot be"
4471
                                   " the primary node.")
4472
      self.secondaries.append(self.op.snode)
4473
      _CheckNodeOnline(self, self.op.snode)
4474

    
4475
    nodenames = [pnode.name] + self.secondaries
4476

    
4477
    req_size = _ComputeDiskSize(self.op.disk_template,
4478
                                self.disks)
4479

    
4480
    # Check lv size requirements
4481
    if req_size is not None:
4482
      nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
4483
                                         self.op.hypervisor)
4484
      for node in nodenames:
4485
        info = nodeinfo[node]
4486
        info.Raise()
4487
        info = info.data
4488
        if not info:
4489
          raise errors.OpPrereqError("Cannot get current information"
4490
                                     " from node '%s'" % node)
4491
        vg_free = info.get('vg_free', None)
4492
        if not isinstance(vg_free, int):
4493
          raise errors.OpPrereqError("Can't compute free disk space on"
4494
                                     " node %s" % node)
4495
        if req_size > info['vg_free']:
4496
          raise errors.OpPrereqError("Not enough disk space on target node %s."
4497
                                     " %d MB available, %d MB required" %
4498
                                     (node, info['vg_free'], req_size))
4499

    
4500
    _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
4501

    
4502
    # os verification
4503
    result = self.rpc.call_os_get(pnode.name, self.op.os_type)
4504
    result.Raise()
4505
    if not isinstance(result.data, objects.OS):
4506
      raise errors.OpPrereqError("OS '%s' not in supported os list for"
4507
                                 " primary node"  % self.op.os_type)
4508

    
4509
    # bridge check on primary node
4510
    bridges = [n.bridge for n in self.nics]
4511
    result = self.rpc.call_bridges_exist(self.pnode.name, bridges)
4512
    result.Raise()
4513
    if not result.data:
4514
      raise errors.OpPrereqError("One of the target bridges '%s' does not"
4515
                                 " exist on destination node '%s'" %
4516
                                 (",".join(bridges), pnode.name))
4517

    
4518
    # memory check on primary node
4519
    if self.op.start:
4520
      _CheckNodeFreeMemory(self, self.pnode.name,
4521
                           "creating instance %s" % self.op.instance_name,
4522
                           self.be_full[constants.BE_MEMORY],
4523
                           self.op.hypervisor)
4524

    
4525
  def Exec(self, feedback_fn):
4526
    """Create and add the instance to the cluster.
4527

4528
    """
4529
    instance = self.op.instance_name
4530
    pnode_name = self.pnode.name
4531

    
4532
    for nic in self.nics:
4533
      if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
4534
        nic.mac = self.cfg.GenerateMAC()
4535

    
4536
    ht_kind = self.op.hypervisor
4537
    if ht_kind in constants.HTS_REQ_PORT:
4538
      network_port = self.cfg.AllocatePort()
4539
    else:
4540
      network_port = None
4541

    
4542
    ##if self.op.vnc_bind_address is None:
4543
    ##  self.op.vnc_bind_address = constants.VNC_DEFAULT_BIND_ADDRESS
4544

    
4545
    # this is needed because os.path.join does not accept None arguments
4546
    if self.op.file_storage_dir is None:
4547
      string_file_storage_dir = ""
4548
    else:
4549
      string_file_storage_dir = self.op.file_storage_dir
4550

    
4551
    # build the full file storage dir path
4552
    file_storage_dir = os.path.normpath(os.path.join(
4553
                                        self.cfg.GetFileStorageDir(),
4554
                                        string_file_storage_dir, instance))
4555

    
4556

    
4557
    disks = _GenerateDiskTemplate(self,
4558
                                  self.op.disk_template,
4559
                                  instance, pnode_name,
4560
                                  self.secondaries,
4561
                                  self.disks,
4562
                                  file_storage_dir,
4563
                                  self.op.file_driver,
4564
                                  0)
4565

    
4566
    iobj = objects.Instance(name=instance, os=self.op.os_type,
4567
                            primary_node=pnode_name,
4568
                            nics=self.nics, disks=disks,
4569
                            disk_template=self.op.disk_template,
4570
                            admin_up=False,
4571
                            network_port=network_port,
4572
                            beparams=self.op.beparams,
4573
                            hvparams=self.op.hvparams,
4574
                            hypervisor=self.op.hypervisor,
4575
                            )
4576

    
4577
    feedback_fn("* creating instance disks...")
4578
    try:
4579
      _CreateDisks(self, iobj)
4580
    except errors.OpExecError:
4581
      self.LogWarning("Device creation failed, reverting...")
4582
      try:
4583
        _RemoveDisks(self, iobj)
4584
      finally:
4585
        self.cfg.ReleaseDRBDMinors(instance)
4586
        raise
4587

    
4588
    feedback_fn("adding instance %s to cluster config" % instance)
4589

    
4590
    self.cfg.AddInstance(iobj)
4591
    # Declare that we don't want to remove the instance lock anymore, as we've
4592
    # added the instance to the config
4593
    del self.remove_locks[locking.LEVEL_INSTANCE]
4594
    # Unlock all the nodes
4595
    if self.op.mode == constants.INSTANCE_IMPORT:
4596
      nodes_keep = [self.op.src_node]
4597
      nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
4598
                       if node != self.op.src_node]
4599
      self.context.glm.release(locking.LEVEL_NODE, nodes_release)
4600
      self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
4601
    else:
4602
      self.context.glm.release(locking.LEVEL_NODE)
4603
      del self.acquired_locks[locking.LEVEL_NODE]
4604

    
4605
    if self.op.wait_for_sync:
4606
      disk_abort = not _WaitForSync(self, iobj)
4607
    elif iobj.disk_template in constants.DTS_NET_MIRROR:
4608
      # make sure the disks are not degraded (still sync-ing is ok)
4609
      time.sleep(15)
4610
      feedback_fn("* checking mirrors status")
4611
      disk_abort = not _WaitForSync(self, iobj, oneshot=True)
4612
    else:
4613
      disk_abort = False
4614

    
4615
    if disk_abort:
4616
      _RemoveDisks(self, iobj)
4617
      self.cfg.RemoveInstance(iobj.name)
4618
      # Make sure the instance lock gets removed
4619
      self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
4620
      raise errors.OpExecError("There are some degraded disks for"
4621
                               " this instance")
4622

    
4623
    feedback_fn("creating os for instance %s on node %s" %
4624
                (instance, pnode_name))
4625

    
4626
    if iobj.disk_template != constants.DT_DISKLESS:
4627
      if self.op.mode == constants.INSTANCE_CREATE:
4628
        feedback_fn("* running the instance OS create scripts...")
4629
        result = self.rpc.call_instance_os_add(pnode_name, iobj)
4630
        msg = result.RemoteFailMsg()
4631
        if msg:
4632
          raise errors.OpExecError("Could not add os for instance %s"
4633
                                   " on node %s: %s" %
4634
                                   (instance, pnode_name, msg))
4635

    
4636
      elif self.op.mode == constants.INSTANCE_IMPORT:
4637
        feedback_fn("* running the instance OS import scripts...")
4638
        src_node = self.op.src_node
4639
        src_images = self.src_images
4640
        cluster_name = self.cfg.GetClusterName()
4641
        import_result = self.rpc.call_instance_os_import(pnode_name, iobj,
4642
                                                         src_node, src_images,
4643
                                                         cluster_name)
4644
        import_result.Raise()
4645
        for idx, result in enumerate(import_result.data):
4646
          if not result:
4647
            self.LogWarning("Could not import the image %s for instance"
4648
                            " %s, disk %d, on node %s" %
4649
                            (src_images[idx], instance, idx, pnode_name))
4650
      else:
4651
        # also checked in the prereq part
4652
        raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
4653
                                     % self.op.mode)
4654

    
4655
    if self.op.start:
4656
      iobj.admin_up = True
4657
      self.cfg.Update(iobj)
4658
      logging.info("Starting instance %s on node %s", instance, pnode_name)
4659
      feedback_fn("* starting instance...")
4660
      result = self.rpc.call_instance_start(pnode_name, iobj, None)
4661
      msg = result.RemoteFailMsg()
4662
      if msg:
4663
        raise errors.OpExecError("Could not start instance: %s" % msg)
4664

    
4665

    
4666
class LUConnectConsole(NoHooksLU):
4667
  """Connect to an instance's console.
4668

4669
  This is somewhat special in that it returns the command line that
4670
  you need to run on the master node in order to connect to the
4671
  console.
4672

4673
  """
4674
  _OP_REQP = ["instance_name"]
4675
  REQ_BGL = False
4676

    
4677
  def ExpandNames(self):
4678
    self._ExpandAndLockInstance()
4679

    
4680
  def CheckPrereq(self):
4681
    """Check prerequisites.
4682

4683
    This checks that the instance is in the cluster.
4684

4685
    """
4686
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4687
    assert self.instance is not None, \
4688
      "Cannot retrieve locked instance %s" % self.op.instance_name
4689
    _CheckNodeOnline(self, self.instance.primary_node)
4690

    
4691
  def Exec(self, feedback_fn):
4692
    """Connect to the console of an instance
4693

4694
    """
4695
    instance = self.instance
4696
    node = instance.primary_node
4697

    
4698
    node_insts = self.rpc.call_instance_list([node],
4699
                                             [instance.hypervisor])[node]
4700
    node_insts.Raise()
4701

    
4702
    if instance.name not in node_insts.data:
4703
      raise errors.OpExecError("Instance %s is not running." % instance.name)
4704

    
4705
    logging.debug("Connecting to console of %s on %s", instance.name, node)
4706

    
4707
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
4708
    cluster = self.cfg.GetClusterInfo()
4709
    # beparams and hvparams are passed separately, to avoid editing the
4710
    # instance and then saving the defaults in the instance itself.
4711
    hvparams = cluster.FillHV(instance)
4712
    beparams = cluster.FillBE(instance)
4713
    console_cmd = hyper.GetShellCommandForConsole(instance, hvparams, beparams)
4714

    
4715
    # build ssh cmdline
4716
    return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
4717

    
4718

    
4719
class LUReplaceDisks(LogicalUnit):
4720
  """Replace the disks of an instance.
4721

4722
  """
4723
  HPATH = "mirrors-replace"
4724
  HTYPE = constants.HTYPE_INSTANCE
4725
  _OP_REQP = ["instance_name", "mode", "disks"]
4726
  REQ_BGL = False
4727

    
4728
  def CheckArguments(self):
4729
    if not hasattr(self.op, "remote_node"):
4730
      self.op.remote_node = None
4731
    if not hasattr(self.op, "iallocator"):
4732
      self.op.iallocator = None
4733

    
4734
    # check for valid parameter combination
4735
    cnt = [self.op.remote_node, self.op.iallocator].count(None)
4736
    if self.op.mode == constants.REPLACE_DISK_CHG:
4737
      if cnt == 2:
4738
        raise errors.OpPrereqError("When changing the secondary either an"
4739
                                   " iallocator script must be used or the"
4740
                                   " new node given")
4741
      elif cnt == 0:
4742
        raise errors.OpPrereqError("Give either the iallocator or the new"
4743
                                   " secondary, not both")
4744
    else: # not replacing the secondary
4745
      if cnt != 2:
4746
        raise errors.OpPrereqError("The iallocator and new node options can"
4747
                                   " be used only when changing the"
4748
                                   " secondary node")
4749

    
4750
  def ExpandNames(self):
4751
    self._ExpandAndLockInstance()
4752

    
4753
    if self.op.iallocator is not None:
4754
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4755
    elif self.op.remote_node is not None:
4756
      remote_node = self.cfg.ExpandNodeName(self.op.remote_node)
4757
      if remote_node is None:
4758
        raise errors.OpPrereqError("Node '%s' not known" %
4759
                                   self.op.remote_node)
4760
      self.op.remote_node = remote_node
4761
      # Warning: do not remove the locking of the new secondary here
4762
      # unless DRBD8.AddChildren is changed to work in parallel;
4763
      # currently it doesn't since parallel invocations of
4764
      # FindUnusedMinor will conflict
4765
      self.needed_locks[locking.LEVEL_NODE] = [remote_node]
4766
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
4767
    else:
4768
      self.needed_locks[locking.LEVEL_NODE] = []
4769
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4770

    
4771
  def DeclareLocks(self, level):
4772
    # If we're not already locking all nodes in the set we have to declare the
4773
    # instance's primary/secondary nodes.
4774
    if (level == locking.LEVEL_NODE and
4775
        self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
4776
      self._LockInstancesNodes()
4777

    
4778
  def _RunAllocator(self):
4779
    """Compute a new secondary node using an IAllocator.
4780

4781
    """
4782
    ial = IAllocator(self,
4783
                     mode=constants.IALLOCATOR_MODE_RELOC,
4784
                     name=self.op.instance_name,
4785
                     relocate_from=[self.sec_node])
4786

    
4787
    ial.Run(self.op.iallocator)
4788

    
4789
    if not ial.success:
4790
      raise errors.OpPrereqError("Can't compute nodes using"
4791
                                 " iallocator '%s': %s" % (self.op.iallocator,
4792
                                                           ial.info))
4793
    if len(ial.nodes) != ial.required_nodes:
4794
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
4795
                                 " of nodes (%s), required %s" %
4796
                                 (len(ial.nodes), ial.required_nodes))
4797
    self.op.remote_node = ial.nodes[0]
4798
    self.LogInfo("Selected new secondary for the instance: %s",
4799
                 self.op.remote_node)
4800

    
4801
  def BuildHooksEnv(self):
4802
    """Build hooks env.
4803

4804
    This runs on the master, the primary and all the secondaries.
4805

4806
    """
4807
    env = {
4808
      "MODE": self.op.mode,
4809
      "NEW_SECONDARY": self.op.remote_node,
4810
      "OLD_SECONDARY": self.instance.secondary_nodes[0],
4811
      }
4812
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4813
    nl = [
4814
      self.cfg.GetMasterNode(),
4815
      self.instance.primary_node,
4816
      ]
4817
    if self.op.remote_node is not None:
4818
      nl.append(self.op.remote_node)
4819
    return env, nl, nl
4820

    
4821
  def CheckPrereq(self):
4822
    """Check prerequisites.
4823

4824
    This checks that the instance is in the cluster.
4825

4826
    """
4827
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4828
    assert instance is not None, \
4829
      "Cannot retrieve locked instance %s" % self.op.instance_name
4830
    self.instance = instance
4831

    
4832
    if instance.disk_template != constants.DT_DRBD8:
4833
      raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
4834
                                 " instances")
4835

    
4836
    if len(instance.secondary_nodes) != 1:
4837
      raise errors.OpPrereqError("The instance has a strange layout,"
4838
                                 " expected one secondary but found %d" %
4839
                                 len(instance.secondary_nodes))
4840

    
4841
    self.sec_node = instance.secondary_nodes[0]
4842

    
4843
    if self.op.iallocator is not None:
4844
      self._RunAllocator()
4845

    
4846
    remote_node = self.op.remote_node
4847
    if remote_node is not None:
4848
      self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
4849
      assert self.remote_node_info is not None, \
4850
        "Cannot retrieve locked node %s" % remote_node
4851
    else:
4852
      self.remote_node_info = None
4853
    if remote_node == instance.primary_node:
4854
      raise errors.OpPrereqError("The specified node is the primary node of"
4855
                                 " the instance.")
4856
    elif remote_node == self.sec_node:
4857
      raise errors.OpPrereqError("The specified node is already the"
4858
                                 " secondary node of the instance.")
4859

    
4860
    if self.op.mode == constants.REPLACE_DISK_PRI:
4861
      n1 = self.tgt_node = instance.primary_node
4862
      n2 = self.oth_node = self.sec_node
4863
    elif self.op.mode == constants.REPLACE_DISK_SEC:
4864
      n1 = self.tgt_node = self.sec_node
4865
      n2 = self.oth_node = instance.primary_node
4866
    elif self.op.mode == constants.REPLACE_DISK_CHG:
4867
      n1 = self.new_node = remote_node
4868
      n2 = self.oth_node = instance.primary_node
4869
      self.tgt_node = self.sec_node
4870
    else:
4871
      raise errors.ProgrammerError("Unhandled disk replace mode")
4872

    
4873
    _CheckNodeOnline(self, n1)
4874
    _CheckNodeOnline(self, n2)
4875

    
4876
    if not self.op.disks:
4877
      self.op.disks = range(len(instance.disks))
4878

    
4879
    for disk_idx in self.op.disks:
4880
      instance.FindDisk(disk_idx)
4881

    
4882
  def _ExecD8DiskOnly(self, feedback_fn):
4883
    """Replace a disk on the primary or secondary for dbrd8.
4884

4885
    The algorithm for replace is quite complicated:
4886

4887
      1. for each disk to be replaced:
4888

4889
        1. create new LVs on the target node with unique names
4890
        1. detach old LVs from the drbd device
4891
        1. rename old LVs to name_replaced.<time_t>
4892
        1. rename new LVs to old LVs
4893
        1. attach the new LVs (with the old names now) to the drbd device
4894

4895
      1. wait for sync across all devices
4896

4897
      1. for each modified disk:
4898

4899
        1. remove old LVs (which have the name name_replaces.<time_t>)
4900

4901
    Failures are not very well handled.
4902

4903
    """
4904
    steps_total = 6
4905
    warning, info = (self.proc.LogWarning, self.proc.LogInfo)
4906
    instance = self.instance
4907
    iv_names = {}
4908
    vgname = self.cfg.GetVGName()
4909
    # start of work
4910
    cfg = self.cfg
4911
    tgt_node = self.tgt_node
4912
    oth_node = self.oth_node
4913

    
4914
    # Step: check device activation
4915
    self.proc.LogStep(1, steps_total, "check device existence")
4916
    info("checking volume groups")
4917
    my_vg = cfg.GetVGName()
4918
    results = self.rpc.call_vg_list([oth_node, tgt_node])
4919
    if not results:
4920
      raise errors.OpExecError("Can't list volume groups on the nodes")
4921
    for node in oth_node, tgt_node:
4922
      res = results[node]
4923
      if res.failed or not res.data or my_vg not in res.data:
4924
        raise errors.OpExecError("Volume group '%s' not found on %s" %
4925
                                 (my_vg, node))
4926
    for idx, dev in enumerate(instance.disks):
4927
      if idx not in self.op.disks:
4928
        continue
4929
      for node in tgt_node, oth_node:
4930
        info("checking disk/%d on %s" % (idx, node))
4931
        cfg.SetDiskID(dev, node)
4932
        if not self.rpc.call_blockdev_find(node, dev):
4933
          raise errors.OpExecError("Can't find disk/%d on node %s" %
4934
                                   (idx, node))
4935

    
4936
    # Step: check other node consistency
4937
    self.proc.LogStep(2, steps_total, "check peer consistency")
4938
    for idx, dev in enumerate(instance.disks):
4939
      if idx not in self.op.disks:
4940
        continue
4941
      info("checking disk/%d consistency on %s" % (idx, oth_node))
4942
      if not _CheckDiskConsistency(self, dev, oth_node,
4943
                                   oth_node==instance.primary_node):
4944
        raise errors.OpExecError("Peer node (%s) has degraded storage, unsafe"
4945
                                 " to replace disks on this node (%s)" %
4946
                                 (oth_node, tgt_node))
4947

    
4948
    # Step: create new storage
4949
    self.proc.LogStep(3, steps_total, "allocate new storage")
4950
    for idx, dev in enumerate(instance.disks):
4951
      if idx not in self.op.disks:
4952
        continue
4953
      size = dev.size
4954
      cfg.SetDiskID(dev, tgt_node)
4955
      lv_names = [".disk%d_%s" % (idx, suf)
4956
                  for suf in ["data", "meta"]]
4957
      names = _GenerateUniqueNames(self, lv_names)
4958
      lv_data = objects.Disk(dev_type=constants.LD_LV, size=size,
4959
                             logical_id=(vgname, names[0]))
4960
      lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
4961
                             logical_id=(vgname, names[1]))
4962
      new_lvs = [lv_data, lv_meta]
4963
      old_lvs = dev.children
4964
      iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
4965
      info("creating new local storage on %s for %s" %
4966
           (tgt_node, dev.iv_name))
4967
      # we pass force_create=True to force the LVM creation
4968
      for new_lv in new_lvs:
4969
        _CreateBlockDev(self, tgt_node, instance, new_lv, True,
4970
                        _GetInstanceInfoText(instance), False)
4971

    
4972
    # Step: for each lv, detach+rename*2+attach
4973
    self.proc.LogStep(4, steps_total, "change drbd configuration")
4974
    for dev, old_lvs, new_lvs in iv_names.itervalues():
4975
      info("detaching %s drbd from local storage" % dev.iv_name)
4976
      result = self.rpc.call_blockdev_removechildren(tgt_node, dev, old_lvs)
4977
      result.Raise()
4978
      if not result.data:
4979
        raise errors.OpExecError("Can't detach drbd from local storage on node"
4980
                                 " %s for device %s" % (tgt_node, dev.iv_name))
4981
      #dev.children = []
4982
      #cfg.Update(instance)
4983

    
4984
      # ok, we created the new LVs, so now we know we have the needed
4985
      # storage; as such, we proceed on the target node to rename
4986
      # old_lv to _old, and new_lv to old_lv; note that we rename LVs
4987
      # using the assumption that logical_id == physical_id (which in
4988
      # turn is the unique_id on that node)
4989

    
4990
      # FIXME(iustin): use a better name for the replaced LVs
4991
      temp_suffix = int(time.time())
4992
      ren_fn = lambda d, suff: (d.physical_id[0],
4993
                                d.physical_id[1] + "_replaced-%s" % suff)
4994
      # build the rename list based on what LVs exist on the node
4995
      rlist = []
4996
      for to_ren in old_lvs:
4997
        find_res = self.rpc.call_blockdev_find(tgt_node, to_ren)
4998
        if not find_res.failed and find_res.data is not None: # device exists
4999
          rlist.append((to_ren, ren_fn(to_ren, temp_suffix)))
5000

    
5001
      info("renaming the old LVs on the target node")
5002
      result = self.rpc.call_blockdev_rename(tgt_node, rlist)
5003
      result.Raise()
5004
      if not result.data:
5005
        raise errors.OpExecError("Can't rename old LVs on node %s" % tgt_node)
5006
      # now we rename the new LVs to the old LVs
5007
      info("renaming the new LVs on the target node")
5008
      rlist = [(new, old.physical_id) for old, new in zip(old_lvs, new_lvs)]
5009
      result = self.rpc.call_blockdev_rename(tgt_node, rlist)
5010
      result.Raise()
5011
      if not result.data:
5012
        raise errors.OpExecError("Can't rename new LVs on node %s" % tgt_node)
5013

    
5014
      for old, new in zip(old_lvs, new_lvs):
5015
        new.logical_id = old.logical_id
5016
        cfg.SetDiskID(new, tgt_node)
5017

    
5018
      for disk in old_lvs:
5019
        disk.logical_id = ren_fn(disk, temp_suffix)
5020
        cfg.SetDiskID(disk, tgt_node)
5021

    
5022
      # now that the new lvs have the old name, we can add them to the device
5023
      info("adding new mirror component on %s" % tgt_node)
5024
      result = self.rpc.call_blockdev_addchildren(tgt_node, dev, new_lvs)
5025
      if result.failed or not result.data:
5026
        for new_lv in new_lvs:
5027
          result = self.rpc.call_blockdev_remove(tgt_node, new_lv)
5028
          if result.failed or not result.data:
5029
            warning("Can't rollback device %s", hint="manually cleanup unused"
5030
                    " logical volumes")
5031
        raise errors.OpExecError("Can't add local storage to drbd")
5032

    
5033
      dev.children = new_lvs
5034
      cfg.Update(instance)
5035

    
5036
    # Step: wait for sync
5037

    
5038
    # this can fail as the old devices are degraded and _WaitForSync
5039
    # does a combined result over all disks, so we don't check its
5040
    # return value
5041
    self.proc.LogStep(5, steps_total, "sync devices")
5042
    _WaitForSync(self, instance, unlock=True)
5043

    
5044
    # so check manually all the devices
5045
    for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
5046
      cfg.SetDiskID(dev, instance.primary_node)
5047
      result = self.rpc.call_blockdev_find(instance.primary_node, dev)
5048
      if result.failed or result.data[5]:
5049
        raise errors.OpExecError("DRBD device %s is degraded!" % name)
5050

    
5051
    # Step: remove old storage
5052
    self.proc.LogStep(6, steps_total, "removing old storage")
5053
    for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
5054
      info("remove logical volumes for %s" % name)
5055
      for lv in old_lvs:
5056
        cfg.SetDiskID(lv, tgt_node)
5057
        result = self.rpc.call_blockdev_remove(tgt_node, lv)
5058
        if result.failed or not result.data:
5059
          warning("Can't remove old LV", hint="manually remove unused LVs")
5060
          continue
5061

    
5062
  def _ExecD8Secondary(self, feedback_fn):
5063
    """Replace the secondary node for drbd8.
5064

5065
    The algorithm for replace is quite complicated:
5066
      - for all disks of the instance:
5067
        - create new LVs on the new node with same names
5068
        - shutdown the drbd device on the old secondary
5069
        - disconnect the drbd network on the primary
5070
        - create the drbd device on the new secondary
5071
        - network attach the drbd on the primary, using an artifice:
5072
          the drbd code for Attach() will connect to the network if it
5073
          finds a device which is connected to the good local disks but
5074
          not network enabled
5075
      - wait for sync across all devices
5076
      - remove all disks from the old secondary
5077

5078
    Failures are not very well handled.
5079

5080
    """
5081
    steps_total = 6
5082
    warning, info = (self.proc.LogWarning, self.proc.LogInfo)
5083
    instance = self.instance
5084
    iv_names = {}
5085
    # start of work
5086
    cfg = self.cfg
5087
    old_node = self.tgt_node
5088
    new_node = self.new_node
5089
    pri_node = instance.primary_node
5090
    nodes_ip = {
5091
      old_node: self.cfg.GetNodeInfo(old_node).secondary_ip,
5092
      new_node: self.cfg.GetNodeInfo(new_node).secondary_ip,
5093
      pri_node: self.cfg.GetNodeInfo(pri_node).secondary_ip,
5094
      }
5095

    
5096
    # Step: check device activation
5097
    self.proc.LogStep(1, steps_total, "check device existence")
5098
    info("checking volume groups")
5099
    my_vg = cfg.GetVGName()
5100
    results = self.rpc.call_vg_list([pri_node, new_node])
5101
    for node in pri_node, new_node:
5102
      res = results[node]
5103
      if res.failed or not res.data or my_vg not in res.data:
5104
        raise errors.OpExecError("Volume group '%s' not found on %s" %
5105
                                 (my_vg, node))
5106
    for idx, dev in enumerate(instance.disks):
5107
      if idx not in self.op.disks:
5108
        continue
5109
      info("checking disk/%d on %s" % (idx, pri_node))
5110
      cfg.SetDiskID(dev, pri_node)
5111
      result = self.rpc.call_blockdev_find(pri_node, dev)
5112
      result.Raise()
5113
      if not result.data:
5114
        raise errors.OpExecError("Can't find disk/%d on node %s" %
5115
                                 (idx, pri_node))
5116

    
5117
    # Step: check other node consistency
5118
    self.proc.LogStep(2, steps_total, "check peer consistency")
5119
    for idx, dev in enumerate(instance.disks):
5120
      if idx not in self.op.disks:
5121
        continue
5122
      info("checking disk/%d consistency on %s" % (idx, pri_node))
5123
      if not _CheckDiskConsistency(self, dev, pri_node, True, ldisk=True):
5124
        raise errors.OpExecError("Primary node (%s) has degraded storage,"
5125
                                 " unsafe to replace the secondary" %
5126
                                 pri_node)
5127

    
5128
    # Step: create new storage
5129
    self.proc.LogStep(3, steps_total, "allocate new storage")
5130
    for idx, dev in enumerate(instance.disks):
5131
      info("adding new local storage on %s for disk/%d" %
5132
           (new_node, idx))
5133
      # we pass force_create=True to force LVM creation
5134
      for new_lv in dev.children:
5135
        _CreateBlockDev(self, new_node, instance, new_lv, True,
5136
                        _GetInstanceInfoText(instance), False)
5137

    
5138
    # Step 4: dbrd minors and drbd setups changes
5139
    # after this, we must manually remove the drbd minors on both the
5140
    # error and the success paths
5141
    minors = cfg.AllocateDRBDMinor([new_node for dev in instance.disks],
5142
                                   instance.name)
5143
    logging.debug("Allocated minors %s" % (minors,))
5144
    self.proc.LogStep(4, steps_total, "changing drbd configuration")
5145
    for idx, (dev, new_minor) in enumerate(zip(instance.disks, minors)):
5146
      size = dev.size
5147
      info("activating a new drbd on %s for disk/%d" % (new_node, idx))
5148
      # create new devices on new_node; note that we create two IDs:
5149
      # one without port, so the drbd will be activated without
5150
      # networking information on the new node at this stage, and one
5151
      # with network, for the latter activation in step 4
5152
      (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
5153
      if pri_node == o_node1:
5154
        p_minor = o_minor1
5155
      else:
5156
        p_minor = o_minor2
5157

    
5158
      new_alone_id = (pri_node, new_node, None, p_minor, new_minor, o_secret)
5159
      new_net_id = (pri_node, new_node, o_port, p_minor, new_minor, o_secret)
5160

    
5161
      iv_names[idx] = (dev, dev.children, new_net_id)
5162
      logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
5163
                    new_net_id)
5164
      new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
5165
                              logical_id=new_alone_id,
5166
                              children=dev.children)
5167
      try:
5168
        _CreateSingleBlockDev(self, new_node, instance, new_drbd,
5169
                              _GetInstanceInfoText(instance), False)
5170
      except errors.BlockDeviceError:
5171
        self.cfg.ReleaseDRBDMinors(instance.name)
5172
        raise
5173

    
5174
    for idx, dev in enumerate(instance.disks):
5175
      # we have new devices, shutdown the drbd on the old secondary
5176
      info("shutting down drbd for disk/%d on old node" % idx)
5177
      cfg.SetDiskID(dev, old_node)
5178
      result = self.rpc.call_blockdev_shutdown(old_node, dev)
5179
      if result.failed or not result.data:
5180
        warning("Failed to shutdown drbd for disk/%d on old node" % idx,
5181
                hint="Please cleanup this device manually as soon as possible")
5182

    
5183
    info("detaching primary drbds from the network (=> standalone)")
5184
    result = self.rpc.call_drbd_disconnect_net([pri_node], nodes_ip,
5185
                                               instance.disks)[pri_node]
5186

    
5187
    msg = result.RemoteFailMsg()
5188
    if msg:
5189
      # detaches didn't succeed (unlikely)
5190
      self.cfg.ReleaseDRBDMinors(instance.name)
5191
      raise errors.OpExecError("Can't detach the disks from the network on"
5192
                               " old node: %s" % (msg,))
5193

    
5194
    # if we managed to detach at least one, we update all the disks of
5195
    # the instance to point to the new secondary
5196
    info("updating instance configuration")
5197
    for dev, _, new_logical_id in iv_names.itervalues():
5198
      dev.logical_id = new_logical_id
5199
      cfg.SetDiskID(dev, pri_node)
5200
    cfg.Update(instance)
5201

    
5202
    # and now perform the drbd attach
5203
    info("attaching primary drbds to new secondary (standalone => connected)")
5204
    result = self.rpc.call_drbd_attach_net([pri_node, new_node], nodes_ip,
5205
                                           instance.disks, instance.name,
5206
                                           False)
5207
    for to_node, to_result in result.items():
5208
      msg = to_result.RemoteFailMsg()
5209
      if msg:
5210
        warning("can't attach drbd disks on node %s: %s", to_node, msg,
5211
                hint="please do a gnt-instance info to see the"
5212
                " status of disks")
5213

    
5214
    # this can fail as the old devices are degraded and _WaitForSync
5215
    # does a combined result over all disks, so we don't check its
5216
    # return value
5217
    self.proc.LogStep(5, steps_total, "sync devices")
5218
    _WaitForSync(self, instance, unlock=True)
5219

    
5220
    # so check manually all the devices
5221
    for idx, (dev, old_lvs, _) in iv_names.iteritems():
5222
      cfg.SetDiskID(dev, pri_node)
5223
      result = self.rpc.call_blockdev_find(pri_node, dev)
5224
      result.Raise()
5225
      if result.data[5]:
5226
        raise errors.OpExecError("DRBD device disk/%d is degraded!" % idx)
5227

    
5228
    self.proc.LogStep(6, steps_total, "removing old storage")
5229
    for idx, (dev, old_lvs, _) in iv_names.iteritems():
5230
      info("remove logical volumes for disk/%d" % idx)
5231
      for lv in old_lvs:
5232
        cfg.SetDiskID(lv, old_node)
5233
        result = self.rpc.call_blockdev_remove(old_node, lv)
5234
        if result.failed or not result.data:
5235
          warning("Can't remove LV on old secondary",
5236
                  hint="Cleanup stale volumes by hand")
5237

    
5238
  def Exec(self, feedback_fn):
5239
    """Execute disk replacement.
5240

5241
    This dispatches the disk replacement to the appropriate handler.
5242

5243
    """
5244
    instance = self.instance
5245

    
5246
    # Activate the instance disks if we're replacing them on a down instance
5247
    if not instance.admin_up:
5248
      _StartInstanceDisks(self, instance, True)
5249

    
5250
    if self.op.mode == constants.REPLACE_DISK_CHG:
5251
      fn = self._ExecD8Secondary
5252
    else:
5253
      fn = self._ExecD8DiskOnly
5254

    
5255
    ret = fn(feedback_fn)
5256

    
5257
    # Deactivate the instance disks if we're replacing them on a down instance
5258
    if not instance.admin_up:
5259
      _SafeShutdownInstanceDisks(self, instance)
5260

    
5261
    return ret
5262

    
5263

    
5264
class LUGrowDisk(LogicalUnit):
5265
  """Grow a disk of an instance.
5266

5267
  """
5268
  HPATH = "disk-grow"
5269
  HTYPE = constants.HTYPE_INSTANCE
5270
  _OP_REQP = ["instance_name", "disk", "amount", "wait_for_sync"]
5271
  REQ_BGL = False
5272

    
5273
  def ExpandNames(self):
5274
    self._ExpandAndLockInstance()
5275
    self.needed_locks[locking.LEVEL_NODE] = []
5276
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5277

    
5278
  def DeclareLocks(self, level):
5279
    if level == locking.LEVEL_NODE:
5280
      self._LockInstancesNodes()
5281

    
5282
  def BuildHooksEnv(self):
5283
    """Build hooks env.
5284

5285
    This runs on the master, the primary and all the secondaries.
5286

5287
    """
5288
    env = {
5289
      "DISK": self.op.disk,
5290
      "AMOUNT": self.op.amount,
5291
      }
5292
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5293
    nl = [
5294
      self.cfg.GetMasterNode(),
5295
      self.instance.primary_node,
5296
      ]
5297
    return env, nl, nl
5298

    
5299
  def CheckPrereq(self):
5300
    """Check prerequisites.
5301

5302
    This checks that the instance is in the cluster.
5303

5304
    """
5305
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5306
    assert instance is not None, \
5307
      "Cannot retrieve locked instance %s" % self.op.instance_name
5308
    nodenames = list(instance.all_nodes)
5309
    for node in nodenames:
5310
      _CheckNodeOnline(self, node)
5311

    
5312

    
5313
    self.instance = instance
5314

    
5315
    if instance.disk_template not in (constants.DT_PLAIN, constants.DT_DRBD8):
5316
      raise errors.OpPrereqError("Instance's disk layout does not support"
5317
                                 " growing.")
5318

    
5319
    self.disk = instance.FindDisk(self.op.disk)
5320

    
5321
    nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
5322
                                       instance.hypervisor)
5323
    for node in nodenames:
5324
      info = nodeinfo[node]
5325
      if info.failed or not info.data:
5326
        raise errors.OpPrereqError("Cannot get current information"
5327
                                   " from node '%s'" % node)
5328
      vg_free = info.data.get('vg_free', None)
5329
      if not isinstance(vg_free, int):
5330
        raise errors.OpPrereqError("Can't compute free disk space on"
5331
                                   " node %s" % node)
5332
      if self.op.amount > vg_free:
5333
        raise errors.OpPrereqError("Not enough disk space on target node %s:"
5334
                                   " %d MiB available, %d MiB required" %
5335
                                   (node, vg_free, self.op.amount))
5336

    
5337
  def Exec(self, feedback_fn):
5338
    """Execute disk grow.
5339

5340
    """
5341
    instance = self.instance
5342
    disk = self.disk
5343
    for node in instance.all_nodes:
5344
      self.cfg.SetDiskID(disk, node)
5345
      result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
5346
      msg = result.RemoteFailMsg()
5347
      if msg:
5348
        raise errors.OpExecError("Grow request failed to node %s: %s" %
5349
                                 (node, msg))
5350
    disk.RecordGrow(self.op.amount)
5351
    self.cfg.Update(instance)
5352
    if self.op.wait_for_sync:
5353
      disk_abort = not _WaitForSync(self, instance)
5354
      if disk_abort:
5355
        self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
5356
                             " status.\nPlease check the instance.")
5357

    
5358

    
5359
class LUQueryInstanceData(NoHooksLU):
5360
  """Query runtime instance data.
5361

5362
  """
5363
  _OP_REQP = ["instances", "static"]
5364
  REQ_BGL = False
5365

    
5366
  def ExpandNames(self):
5367
    self.needed_locks = {}
5368
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
5369

    
5370
    if not isinstance(self.op.instances, list):
5371
      raise errors.OpPrereqError("Invalid argument type 'instances'")
5372

    
5373
    if self.op.instances:
5374
      self.wanted_names = []
5375
      for name in self.op.instances:
5376
        full_name = self.cfg.ExpandInstanceName(name)
5377
        if full_name is None:
5378
          raise errors.OpPrereqError("Instance '%s' not known" % name)
5379
        self.wanted_names.append(full_name)
5380
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
5381
    else:
5382
      self.wanted_names = None
5383
      self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
5384

    
5385
    self.needed_locks[locking.LEVEL_NODE] = []
5386
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5387

    
5388
  def DeclareLocks(self, level):
5389
    if level == locking.LEVEL_NODE:
5390
      self._LockInstancesNodes()
5391

    
5392
  def CheckPrereq(self):
5393
    """Check prerequisites.
5394

5395
    This only checks the optional instance list against the existing names.
5396

5397
    """
5398
    if self.wanted_names is None:
5399
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
5400

    
5401
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
5402
                             in self.wanted_names]
5403
    return
5404

    
5405
  def _ComputeDiskStatus(self, instance, snode, dev):
5406
    """Compute block device status.
5407

5408
    """
5409
    static = self.op.static
5410
    if not static:
5411
      self.cfg.SetDiskID(dev, instance.primary_node)
5412
      dev_pstatus = self.rpc.call_blockdev_find(instance.primary_node, dev)
5413
      dev_pstatus.Raise()
5414
      dev_pstatus = dev_pstatus.data
5415
    else:
5416
      dev_pstatus = None
5417

    
5418
    if dev.dev_type in constants.LDS_DRBD:
5419
      # we change the snode then (otherwise we use the one passed in)
5420
      if dev.logical_id[0] == instance.primary_node:
5421
        snode = dev.logical_id[1]
5422
      else:
5423
        snode = dev.logical_id[0]
5424

    
5425
    if snode and not static:
5426
      self.cfg.SetDiskID(dev, snode)
5427
      dev_sstatus = self.rpc.call_blockdev_find(snode, dev)
5428
      dev_sstatus.Raise()
5429
      dev_sstatus = dev_sstatus.data
5430
    else:
5431
      dev_sstatus = None
5432

    
5433
    if dev.children:
5434
      dev_children = [self._ComputeDiskStatus(instance, snode, child)
5435
                      for child in dev.children]
5436
    else:
5437
      dev_children = []
5438

    
5439
    data = {
5440
      "iv_name": dev.iv_name,
5441
      "dev_type": dev.dev_type,
5442
      "logical_id": dev.logical_id,
5443
      "physical_id": dev.physical_id,
5444
      "pstatus": dev_pstatus,
5445
      "sstatus": dev_sstatus,
5446
      "children": dev_children,
5447
      "mode": dev.mode,
5448
      }
5449

    
5450
    return data
5451

    
5452
  def Exec(self, feedback_fn):
5453
    """Gather and return data"""
5454
    result = {}
5455

    
5456
    cluster = self.cfg.GetClusterInfo()
5457

    
5458
    for instance in self.wanted_instances:
5459
      if not self.op.static:
5460
        remote_info = self.rpc.call_instance_info(instance.primary_node,
5461
                                                  instance.name,
5462
                                                  instance.hypervisor)
5463
        remote_info.Raise()
5464
        remote_info = remote_info.data
5465
        if remote_info and "state" in remote_info:
5466
          remote_state = "up"
5467
        else:
5468
          remote_state = "down"
5469
      else:
5470
        remote_state = None
5471
      if instance.admin_up:
5472
        config_state = "up"
5473
      else:
5474
        config_state = "down"
5475

    
5476
      disks = [self._ComputeDiskStatus(instance, None, device)
5477
               for device in instance.disks]
5478

    
5479
      idict = {
5480
        "name": instance.name,
5481
        "config_state": config_state,
5482
        "run_state": remote_state,
5483
        "pnode": instance.primary_node,
5484
        "snodes": instance.secondary_nodes,
5485
        "os": instance.os,
5486
        "nics": [(nic.mac, nic.ip, nic.bridge) for nic in instance.nics],
5487
        "disks": disks,
5488
        "hypervisor": instance.hypervisor,
5489
        "network_port": instance.network_port,
5490
        "hv_instance": instance.hvparams,
5491
        "hv_actual": cluster.FillHV(instance),
5492
        "be_instance": instance.beparams,
5493
        "be_actual": cluster.FillBE(instance),
5494
        }
5495

    
5496
      result[instance.name] = idict
5497

    
5498
    return result
5499

    
5500

    
5501
class LUSetInstanceParams(LogicalUnit):
5502
  """Modifies an instances's parameters.
5503

5504
  """
5505
  HPATH = "instance-modify"
5506
  HTYPE = constants.HTYPE_INSTANCE
5507
  _OP_REQP = ["instance_name"]
5508
  REQ_BGL = False
5509

    
5510
  def CheckArguments(self):
5511
    if not hasattr(self.op, 'nics'):
5512
      self.op.nics = []
5513
    if not hasattr(self.op, 'disks'):
5514
      self.op.disks = []
5515
    if not hasattr(self.op, 'beparams'):
5516
      self.op.beparams = {}
5517
    if not hasattr(self.op, 'hvparams'):
5518
      self.op.hvparams = {}
5519
    self.op.force = getattr(self.op, "force", False)
5520
    if not (self.op.nics or self.op.disks or
5521
            self.op.hvparams or self.op.beparams):
5522
      raise errors.OpPrereqError("No changes submitted")
5523

    
5524
    utils.CheckBEParams(self.op.beparams)
5525

    
5526
    # Disk validation
5527
    disk_addremove = 0
5528
    for disk_op, disk_dict in self.op.disks:
5529
      if disk_op == constants.DDM_REMOVE:
5530
        disk_addremove += 1
5531
        continue
5532
      elif disk_op == constants.DDM_ADD:
5533
        disk_addremove += 1
5534
      else:
5535
        if not isinstance(disk_op, int):
5536
          raise errors.OpPrereqError("Invalid disk index")
5537
      if disk_op == constants.DDM_ADD:
5538
        mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
5539
        if mode not in constants.DISK_ACCESS_SET:
5540
          raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode)
5541
        size = disk_dict.get('size', None)
5542
        if size is None:
5543
          raise errors.OpPrereqError("Required disk parameter size missing")
5544
        try:
5545
          size = int(size)
5546
        except ValueError, err:
5547
          raise errors.OpPrereqError("Invalid disk size parameter: %s" %
5548
                                     str(err))
5549
        disk_dict['size'] = size
5550
      else:
5551
        # modification of disk
5552
        if 'size' in disk_dict:
5553
          raise errors.OpPrereqError("Disk size change not possible, use"
5554
                                     " grow-disk")
5555

    
5556
    if disk_addremove > 1:
5557
      raise errors.OpPrereqError("Only one disk add or remove operation"
5558
                                 " supported at a time")
5559

    
5560
    # NIC validation
5561
    nic_addremove = 0
5562
    for nic_op, nic_dict in self.op.nics:
5563
      if nic_op == constants.DDM_REMOVE:
5564
        nic_addremove += 1
5565
        continue
5566
      elif nic_op == constants.DDM_ADD:
5567
        nic_addremove += 1
5568
      else:
5569
        if not isinstance(nic_op, int):
5570
          raise errors.OpPrereqError("Invalid nic index")
5571

    
5572
      # nic_dict should be a dict
5573
      nic_ip = nic_dict.get('ip', None)
5574
      if nic_ip is not None:
5575
        if nic_ip.lower() == "none":
5576
          nic_dict['ip'] = None
5577
        else:
5578
          if not utils.IsValidIP(nic_ip):
5579
            raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip)
5580
      # we can only check None bridges and assign the default one
5581
      nic_bridge = nic_dict.get('bridge', None)
5582
      if nic_bridge is None:
5583
        nic_dict['bridge'] = self.cfg.GetDefBridge()
5584
      # but we can validate MACs
5585
      nic_mac = nic_dict.get('mac', None)
5586
      if nic_mac is not None:
5587
        if self.cfg.IsMacInUse(nic_mac):
5588
          raise errors.OpPrereqError("MAC address %s already in use"
5589
                                     " in cluster" % nic_mac)
5590
        if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
5591
          if not utils.IsValidMac(nic_mac):
5592
            raise errors.OpPrereqError("Invalid MAC address %s" % nic_mac)
5593
    if nic_addremove > 1:
5594
      raise errors.OpPrereqError("Only one NIC add or remove operation"
5595
                                 " supported at a time")
5596

    
5597
  def ExpandNames(self):
5598
    self._ExpandAndLockInstance()
5599
    self.needed_locks[locking.LEVEL_NODE] = []
5600
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5601

    
5602
  def DeclareLocks(self, level):
5603
    if level == locking.LEVEL_NODE:
5604
      self._LockInstancesNodes()
5605

    
5606
  def BuildHooksEnv(self):
5607
    """Build hooks env.
5608

5609
    This runs on the master, primary and secondaries.
5610

5611
    """
5612
    args = dict()
5613
    if constants.BE_MEMORY in self.be_new:
5614
      args['memory'] = self.be_new[constants.BE_MEMORY]
5615
    if constants.BE_VCPUS in self.be_new:
5616
      args['vcpus'] = self.be_new[constants.BE_VCPUS]
5617
    # FIXME: readd disk/nic changes
5618
    env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
5619
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5620
    return env, nl, nl
5621

    
5622
  def CheckPrereq(self):
5623
    """Check prerequisites.
5624

5625
    This only checks the instance list against the existing names.
5626

5627
    """
5628
    force = self.force = self.op.force
5629

    
5630
    # checking the new params on the primary/secondary nodes
5631

    
5632
    instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5633
    assert self.instance is not None, \
5634
      "Cannot retrieve locked instance %s" % self.op.instance_name
5635
    pnode = instance.primary_node
5636
    nodelist = list(instance.all_nodes)
5637

    
5638
    # hvparams processing
5639
    if self.op.hvparams:
5640
      i_hvdict = copy.deepcopy(instance.hvparams)
5641
      for key, val in self.op.hvparams.iteritems():
5642
        if val == constants.VALUE_DEFAULT:
5643
          try:
5644
            del i_hvdict[key]
5645
          except KeyError:
5646
            pass
5647
        elif val == constants.VALUE_NONE:
5648
          i_hvdict[key] = None
5649
        else:
5650
          i_hvdict[key] = val
5651
      cluster = self.cfg.GetClusterInfo()
5652
      hv_new = cluster.FillDict(cluster.hvparams[instance.hypervisor],
5653
                                i_hvdict)
5654
      # local check
5655
      hypervisor.GetHypervisor(
5656
        instance.hypervisor).CheckParameterSyntax(hv_new)
5657
      _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
5658
      self.hv_new = hv_new # the new actual values
5659
      self.hv_inst = i_hvdict # the new dict (without defaults)
5660
    else:
5661
      self.hv_new = self.hv_inst = {}
5662

    
5663
    # beparams processing
5664
    if self.op.beparams:
5665
      i_bedict = copy.deepcopy(instance.beparams)
5666
      for key, val in self.op.beparams.iteritems():
5667
        if val == constants.VALUE_DEFAULT:
5668
          try:
5669
            del i_bedict[key]
5670
          except KeyError:
5671
            pass
5672
        else:
5673
          i_bedict[key] = val
5674
      cluster = self.cfg.GetClusterInfo()
5675
      be_new = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
5676
                                i_bedict)
5677
      self.be_new = be_new # the new actual values
5678
      self.be_inst = i_bedict # the new dict (without defaults)
5679
    else:
5680
      self.be_new = self.be_inst = {}
5681

    
5682
    self.warn = []
5683

    
5684
    if constants.BE_MEMORY in self.op.beparams and not self.force:
5685
      mem_check_list = [pnode]
5686
      if be_new[constants.BE_AUTO_BALANCE]:
5687
        # either we changed auto_balance to yes or it was from before
5688
        mem_check_list.extend(instance.secondary_nodes)
5689
      instance_info = self.rpc.call_instance_info(pnode, instance.name,
5690
                                                  instance.hypervisor)
5691
      nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
5692
                                         instance.hypervisor)
5693
      if nodeinfo[pnode].failed or not isinstance(nodeinfo[pnode].data, dict):
5694
        # Assume the primary node is unreachable and go ahead
5695
        self.warn.append("Can't get info from primary node %s" % pnode)
5696
      else:
5697
        if not instance_info.failed and instance_info.data:
5698
          current_mem = instance_info.data['memory']
5699
        else:
5700
          # Assume instance not running
5701
          # (there is a slight race condition here, but it's not very probable,
5702
          # and we have no other way to check)
5703
          current_mem = 0
5704
        miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
5705
                    nodeinfo[pnode].data['memory_free'])
5706
        if miss_mem > 0:
5707
          raise errors.OpPrereqError("This change will prevent the instance"
5708
                                     " from starting, due to %d MB of memory"
5709
                                     " missing on its primary node" % miss_mem)
5710

    
5711
      if be_new[constants.BE_AUTO_BALANCE]:
5712
        for node, nres in nodeinfo.iteritems():
5713
          if node not in instance.secondary_nodes:
5714
            continue
5715
          if nres.failed or not isinstance(nres.data, dict):
5716
            self.warn.append("Can't get info from secondary node %s" % node)
5717
          elif be_new[constants.BE_MEMORY] > nres.data['memory_free']:
5718
            self.warn.append("Not enough memory to failover instance to"
5719
                             " secondary node %s" % node)
5720

    
5721
    # NIC processing
5722
    for nic_op, nic_dict in self.op.nics:
5723
      if nic_op == constants.DDM_REMOVE:
5724
        if not instance.nics:
5725
          raise errors.OpPrereqError("Instance has no NICs, cannot remove")
5726
        continue
5727
      if nic_op != constants.DDM_ADD:
5728
        # an existing nic
5729
        if nic_op < 0 or nic_op >= len(instance.nics):
5730
          raise errors.OpPrereqError("Invalid NIC index %s, valid values"
5731
                                     " are 0 to %d" %
5732
                                     (nic_op, len(instance.nics)))
5733
      nic_bridge = nic_dict.get('bridge', None)
5734
      if nic_bridge is not None:
5735
        if not self.rpc.call_bridges_exist(pnode, [nic_bridge]):
5736
          msg = ("Bridge '%s' doesn't exist on one of"
5737
                 " the instance nodes" % nic_bridge)
5738
          if self.force:
5739
            self.warn.append(msg)
5740
          else:
5741
            raise errors.OpPrereqError(msg)
5742

    
5743
    # DISK processing
5744
    if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
5745
      raise errors.OpPrereqError("Disk operations not supported for"
5746
                                 " diskless instances")
5747
    for disk_op, disk_dict in self.op.disks:
5748
      if disk_op == constants.DDM_REMOVE:
5749
        if len(instance.disks) == 1:
5750
          raise errors.OpPrereqError("Cannot remove the last disk of"
5751
                                     " an instance")
5752
        ins_l = self.rpc.call_instance_list([pnode], [instance.hypervisor])
5753
        ins_l = ins_l[pnode]
5754
        if ins_l.failed or not isinstance(ins_l.data, list):
5755
          raise errors.OpPrereqError("Can't contact node '%s'" % pnode)
5756
        if instance.name in ins_l.data:
5757
          raise errors.OpPrereqError("Instance is running, can't remove"
5758
                                     " disks.")
5759

    
5760
      if (disk_op == constants.DDM_ADD and
5761
          len(instance.nics) >= constants.MAX_DISKS):
5762
        raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
5763
                                   " add more" % constants.MAX_DISKS)
5764
      if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
5765
        # an existing disk
5766
        if disk_op < 0 or disk_op >= len(instance.disks):
5767
          raise errors.OpPrereqError("Invalid disk index %s, valid values"
5768
                                     " are 0 to %d" %
5769
                                     (disk_op, len(instance.disks)))
5770

    
5771
    return
5772

    
5773
  def Exec(self, feedback_fn):
5774
    """Modifies an instance.
5775

5776
    All parameters take effect only at the next restart of the instance.
5777

5778
    """
5779
    # Process here the warnings from CheckPrereq, as we don't have a
5780
    # feedback_fn there.
5781
    for warn in self.warn:
5782
      feedback_fn("WARNING: %s" % warn)
5783

    
5784
    result = []
5785
    instance = self.instance
5786
    # disk changes
5787
    for disk_op, disk_dict in self.op.disks:
5788
      if disk_op == constants.DDM_REMOVE:
5789
        # remove the last disk
5790
        device = instance.disks.pop()
5791
        device_idx = len(instance.disks)
5792
        for node, disk in device.ComputeNodeTree(instance.primary_node):
5793
          self.cfg.SetDiskID(disk, node)
5794
          rpc_result = self.rpc.call_blockdev_remove(node, disk)
5795
          if rpc_result.failed or not rpc_result.data:
5796
            self.proc.LogWarning("Could not remove disk/%d on node %s,"
5797
                                 " continuing anyway", device_idx, node)
5798
        result.append(("disk/%d" % device_idx, "remove"))
5799
      elif disk_op == constants.DDM_ADD:
5800
        # add a new disk
5801
        if instance.disk_template == constants.DT_FILE:
5802
          file_driver, file_path = instance.disks[0].logical_id
5803
          file_path = os.path.dirname(file_path)
5804
        else:
5805
          file_driver = file_path = None
5806
        disk_idx_base = len(instance.disks)
5807
        new_disk = _GenerateDiskTemplate(self,
5808
                                         instance.disk_template,
5809
                                         instance.name, instance.primary_node,
5810
                                         instance.secondary_nodes,
5811
                                         [disk_dict],
5812
                                         file_path,
5813
                                         file_driver,
5814
                                         disk_idx_base)[0]
5815
        instance.disks.append(new_disk)
5816
        info = _GetInstanceInfoText(instance)
5817

    
5818
        logging.info("Creating volume %s for instance %s",
5819
                     new_disk.iv_name, instance.name)
5820
        # Note: this needs to be kept in sync with _CreateDisks
5821
        #HARDCODE
5822
        for node in instance.all_nodes:
5823
          f_create = node == instance.primary_node
5824
          try:
5825
            _CreateBlockDev(self, node, instance, new_disk,
5826
                            f_create, info, f_create)
5827
          except errors.OpExecError, err:
5828
            self.LogWarning("Failed to create volume %s (%s) on"
5829
                            " node %s: %s",
5830
                            new_disk.iv_name, new_disk, node, err)
5831
        result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
5832
                       (new_disk.size, new_disk.mode)))
5833
      else:
5834
        # change a given disk
5835
        instance.disks[disk_op].mode = disk_dict['mode']
5836
        result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
5837
    # NIC changes
5838
    for nic_op, nic_dict in self.op.nics:
5839
      if nic_op == constants.DDM_REMOVE:
5840
        # remove the last nic
5841
        del instance.nics[-1]
5842
        result.append(("nic.%d" % len(instance.nics), "remove"))
5843
      elif nic_op == constants.DDM_ADD:
5844
        # add a new nic
5845
        if 'mac' not in nic_dict:
5846
          mac = constants.VALUE_GENERATE
5847
        else:
5848
          mac = nic_dict['mac']
5849
        if mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
5850
          mac = self.cfg.GenerateMAC()
5851
        new_nic = objects.NIC(mac=mac, ip=nic_dict.get('ip', None),
5852
                              bridge=nic_dict.get('bridge', None))
5853
        instance.nics.append(new_nic)
5854
        result.append(("nic.%d" % (len(instance.nics) - 1),
5855
                       "add:mac=%s,ip=%s,bridge=%s" %
5856
                       (new_nic.mac, new_nic.ip, new_nic.bridge)))
5857
      else:
5858
        # change a given nic
5859
        for key in 'mac', 'ip', 'bridge':
5860
          if key in nic_dict:
5861
            setattr(instance.nics[nic_op], key, nic_dict[key])
5862
            result.append(("nic.%s/%d" % (key, nic_op), nic_dict[key]))
5863

    
5864
    # hvparams changes
5865
    if self.op.hvparams:
5866
      instance.hvparams = self.hv_new
5867
      for key, val in self.op.hvparams.iteritems():
5868
        result.append(("hv/%s" % key, val))
5869

    
5870
    # beparams changes
5871
    if self.op.beparams:
5872
      instance.beparams = self.be_inst
5873
      for key, val in self.op.beparams.iteritems():
5874
        result.append(("be/%s" % key, val))
5875

    
5876
    self.cfg.Update(instance)
5877

    
5878
    return result
5879

    
5880

    
5881
class LUQueryExports(NoHooksLU):
5882
  """Query the exports list
5883

5884
  """
5885
  _OP_REQP = ['nodes']
5886
  REQ_BGL = False
5887

    
5888
  def ExpandNames(self):
5889
    self.needed_locks = {}
5890
    self.share_locks[locking.LEVEL_NODE] = 1
5891
    if not self.op.nodes:
5892
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5893
    else:
5894
      self.needed_locks[locking.LEVEL_NODE] = \
5895
        _GetWantedNodes(self, self.op.nodes)
5896

    
5897
  def CheckPrereq(self):
5898
    """Check prerequisites.
5899

5900
    """
5901
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
5902

    
5903
  def Exec(self, feedback_fn):
5904
    """Compute the list of all the exported system images.
5905

5906
    @rtype: dict
5907
    @return: a dictionary with the structure node->(export-list)
5908
        where export-list is a list of the instances exported on
5909
        that node.
5910

5911
    """
5912
    rpcresult = self.rpc.call_export_list(self.nodes)
5913
    result = {}
5914
    for node in rpcresult:
5915
      if rpcresult[node].failed:
5916
        result[node] = False
5917
      else:
5918
        result[node] = rpcresult[node].data
5919

    
5920
    return result
5921

    
5922

    
5923
class LUExportInstance(LogicalUnit):
5924
  """Export an instance to an image in the cluster.
5925

5926
  """
5927
  HPATH = "instance-export"
5928
  HTYPE = constants.HTYPE_INSTANCE
5929
  _OP_REQP = ["instance_name", "target_node", "shutdown"]
5930
  REQ_BGL = False
5931

    
5932
  def ExpandNames(self):
5933
    self._ExpandAndLockInstance()
5934
    # FIXME: lock only instance primary and destination node
5935
    #
5936
    # Sad but true, for now we have do lock all nodes, as we don't know where
5937
    # the previous export might be, and and in this LU we search for it and
5938
    # remove it from its current node. In the future we could fix this by:
5939
    #  - making a tasklet to search (share-lock all), then create the new one,
5940
    #    then one to remove, after
5941
    #  - removing the removal operation altoghether
5942
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5943

    
5944
  def DeclareLocks(self, level):
5945
    """Last minute lock declaration."""
5946
    # All nodes are locked anyway, so nothing to do here.
5947

    
5948
  def BuildHooksEnv(self):
5949
    """Build hooks env.
5950

5951
    This will run on the master, primary node and target node.
5952

5953
    """
5954
    env = {
5955
      "EXPORT_NODE": self.op.target_node,
5956
      "EXPORT_DO_SHUTDOWN": self.op.shutdown,
5957
      }
5958
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5959
    nl = [self.cfg.GetMasterNode(), self.instance.primary_node,
5960
          self.op.target_node]
5961
    return env, nl, nl
5962

    
5963
  def CheckPrereq(self):
5964
    """Check prerequisites.
5965

5966
    This checks that the instance and node names are valid.
5967

5968
    """
5969
    instance_name = self.op.instance_name
5970
    self.instance = self.cfg.GetInstanceInfo(instance_name)
5971
    assert self.instance is not None, \
5972
          "Cannot retrieve locked instance %s" % self.op.instance_name
5973
    _CheckNodeOnline(self, self.instance.primary_node)
5974

    
5975
    self.dst_node = self.cfg.GetNodeInfo(
5976
      self.cfg.ExpandNodeName(self.op.target_node))
5977

    
5978
    if self.dst_node is None:
5979
      # This is wrong node name, not a non-locked node
5980
      raise errors.OpPrereqError("Wrong node name %s" % self.op.target_node)
5981
    _CheckNodeOnline(self, self.dst_node.name)
5982

    
5983
    # instance disk type verification
5984
    for disk in self.instance.disks:
5985
      if disk.dev_type == constants.LD_FILE:
5986
        raise errors.OpPrereqError("Export not supported for instances with"
5987
                                   " file-based disks")
5988

    
5989
  def Exec(self, feedback_fn):
5990
    """Export an instance to an image in the cluster.
5991

5992
    """
5993
    instance = self.instance
5994
    dst_node = self.dst_node
5995
    src_node = instance.primary_node
5996
    if self.op.shutdown:
5997
      # shutdown the instance, but not the disks
5998
      result = self.rpc.call_instance_shutdown(src_node, instance)
5999
      result.Raise()
6000
      if not result.data:
6001
        raise errors.OpExecError("Could not shutdown instance %s on node %s" %
6002
                                 (instance.name, src_node))
6003

    
6004
    vgname = self.cfg.GetVGName()
6005

    
6006
    snap_disks = []
6007

    
6008
    # set the disks ID correctly since call_instance_start needs the
6009
    # correct drbd minor to create the symlinks
6010
    for disk in instance.disks:
6011
      self.cfg.SetDiskID(disk, src_node)
6012

    
6013
    try:
6014
      for disk in instance.disks:
6015
        # new_dev_name will be a snapshot of an lvm leaf of the one we passed
6016
        new_dev_name = self.rpc.call_blockdev_snapshot(src_node, disk)
6017
        if new_dev_name.failed or not new_dev_name.data:
6018
          self.LogWarning("Could not snapshot block device %s on node %s",
6019
                          disk.logical_id[1], src_node)
6020
          snap_disks.append(False)
6021
        else:
6022
          new_dev = objects.Disk(dev_type=constants.LD_LV, size=disk.size,
6023
                                 logical_id=(vgname, new_dev_name.data),
6024
                                 physical_id=(vgname, new_dev_name.data),
6025
                                 iv_name=disk.iv_name)
6026
          snap_disks.append(new_dev)
6027

    
6028
    finally:
6029
      if self.op.shutdown and instance.admin_up:
6030
        result = self.rpc.call_instance_start(src_node, instance, None)
6031
        msg = result.RemoteFailMsg()
6032
        if msg:
6033
          _ShutdownInstanceDisks(self, instance)
6034
          raise errors.OpExecError("Could not start instance: %s" % msg)
6035

    
6036
    # TODO: check for size
6037

    
6038
    cluster_name = self.cfg.GetClusterName()
6039
    for idx, dev in enumerate(snap_disks):
6040
      if dev:
6041
        result = self.rpc.call_snapshot_export(src_node, dev, dst_node.name,
6042
                                               instance, cluster_name, idx)
6043
        if result.failed or not result.data:
6044
          self.LogWarning("Could not export block device %s from node %s to"
6045
                          " node %s", dev.logical_id[1], src_node,
6046
                          dst_node.name)
6047
        result = self.rpc.call_blockdev_remove(src_node, dev)
6048
        if result.failed or not result.data:
6049
          self.LogWarning("Could not remove snapshot block device %s from node"
6050
                          " %s", dev.logical_id[1], src_node)
6051

    
6052
    result = self.rpc.call_finalize_export(dst_node.name, instance, snap_disks)
6053
    if result.failed or not result.data:
6054
      self.LogWarning("Could not finalize export for instance %s on node %s",
6055
                      instance.name, dst_node.name)
6056

    
6057
    nodelist = self.cfg.GetNodeList()
6058
    nodelist.remove(dst_node.name)
6059

    
6060
    # on one-node clusters nodelist will be empty after the removal
6061
    # if we proceed the backup would be removed because OpQueryExports
6062
    # substitutes an empty list with the full cluster node list.
6063
    if nodelist:
6064
      exportlist = self.rpc.call_export_list(nodelist)
6065
      for node in exportlist:
6066
        if exportlist[node].failed:
6067
          continue
6068
        if instance.name in exportlist[node].data:
6069
          if not self.rpc.call_export_remove(node, instance.name):
6070
            self.LogWarning("Could not remove older export for instance %s"
6071
                            " on node %s", instance.name, node)
6072

    
6073

    
6074
class LURemoveExport(NoHooksLU):
6075
  """Remove exports related to the named instance.
6076

6077
  """
6078
  _OP_REQP = ["instance_name"]
6079
  REQ_BGL = False
6080

    
6081
  def ExpandNames(self):
6082
    self.needed_locks = {}
6083
    # We need all nodes to be locked in order for RemoveExport to work, but we
6084
    # don't need to lock the instance itself, as nothing will happen to it (and
6085
    # we can remove exports also for a removed instance)
6086
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6087

    
6088
  def CheckPrereq(self):
6089
    """Check prerequisites.
6090
    """
6091
    pass
6092

    
6093
  def Exec(self, feedback_fn):
6094
    """Remove any export.
6095

6096
    """
6097
    instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
6098
    # If the instance was not found we'll try with the name that was passed in.
6099
    # This will only work if it was an FQDN, though.
6100
    fqdn_warn = False
6101
    if not instance_name:
6102
      fqdn_warn = True
6103
      instance_name = self.op.instance_name
6104

    
6105
    exportlist = self.rpc.call_export_list(self.acquired_locks[
6106
      locking.LEVEL_NODE])
6107
    found = False
6108
    for node in exportlist:
6109
      if exportlist[node].failed:
6110
        self.LogWarning("Failed to query node %s, continuing" % node)
6111
        continue
6112
      if instance_name in exportlist[node].data:
6113
        found = True
6114
        result = self.rpc.call_export_remove(node, instance_name)
6115
        if result.failed or not result.data:
6116
          logging.error("Could not remove export for instance %s"
6117
                        " on node %s", instance_name, node)
6118

    
6119
    if fqdn_warn and not found:
6120
      feedback_fn("Export not found. If trying to remove an export belonging"
6121
                  " to a deleted instance please use its Fully Qualified"
6122
                  " Domain Name.")
6123

    
6124

    
6125
class TagsLU(NoHooksLU):
6126
  """Generic tags LU.
6127

6128
  This is an abstract class which is the parent of all the other tags LUs.
6129

6130
  """
6131

    
6132
  def ExpandNames(self):
6133
    self.needed_locks = {}
6134
    if self.op.kind == constants.TAG_NODE:
6135
      name = self.cfg.ExpandNodeName(self.op.name)
6136
      if name is None:
6137
        raise errors.OpPrereqError("Invalid node name (%s)" %
6138
                                   (self.op.name,))
6139
      self.op.name = name
6140
      self.needed_locks[locking.LEVEL_NODE] = name
6141
    elif self.op.kind == constants.TAG_INSTANCE:
6142
      name = self.cfg.ExpandInstanceName(self.op.name)
6143
      if name is None:
6144
        raise errors.OpPrereqError("Invalid instance name (%s)" %
6145
                                   (self.op.name,))
6146
      self.op.name = name
6147
      self.needed_locks[locking.LEVEL_INSTANCE] = name
6148

    
6149
  def CheckPrereq(self):
6150
    """Check prerequisites.
6151

6152
    """
6153
    if self.op.kind == constants.TAG_CLUSTER:
6154
      self.target = self.cfg.GetClusterInfo()
6155
    elif self.op.kind == constants.TAG_NODE:
6156
      self.target = self.cfg.GetNodeInfo(self.op.name)
6157
    elif self.op.kind == constants.TAG_INSTANCE:
6158
      self.target = self.cfg.GetInstanceInfo(self.op.name)
6159
    else:
6160
      raise errors.OpPrereqError("Wrong tag type requested (%s)" %
6161
                                 str(self.op.kind))
6162

    
6163

    
6164
class LUGetTags(TagsLU):
6165
  """Returns the tags of a given object.
6166

6167
  """
6168
  _OP_REQP = ["kind", "name"]
6169
  REQ_BGL = False
6170

    
6171
  def Exec(self, feedback_fn):
6172
    """Returns the tag list.
6173

6174
    """
6175
    return list(self.target.GetTags())
6176

    
6177

    
6178
class LUSearchTags(NoHooksLU):
6179
  """Searches the tags for a given pattern.
6180

6181
  """
6182
  _OP_REQP = ["pattern"]
6183
  REQ_BGL = False
6184

    
6185
  def ExpandNames(self):
6186
    self.needed_locks = {}
6187

    
6188
  def CheckPrereq(self):
6189
    """Check prerequisites.
6190

6191
    This checks the pattern passed for validity by compiling it.
6192

6193
    """
6194
    try:
6195
      self.re = re.compile(self.op.pattern)
6196
    except re.error, err:
6197
      raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
6198
                                 (self.op.pattern, err))
6199

    
6200
  def Exec(self, feedback_fn):
6201
    """Returns the tag list.
6202

6203
    """
6204
    cfg = self.cfg
6205
    tgts = [("/cluster", cfg.GetClusterInfo())]
6206
    ilist = cfg.GetAllInstancesInfo().values()
6207
    tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
6208
    nlist = cfg.GetAllNodesInfo().values()
6209
    tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
6210
    results = []
6211
    for path, target in tgts:
6212
      for tag in target.GetTags():
6213
        if self.re.search(tag):
6214
          results.append((path, tag))
6215
    return results
6216

    
6217

    
6218
class LUAddTags(TagsLU):
6219
  """Sets a tag on a given object.
6220

6221
  """
6222
  _OP_REQP = ["kind", "name", "tags"]
6223
  REQ_BGL = False
6224

    
6225
  def CheckPrereq(self):
6226
    """Check prerequisites.
6227

6228
    This checks the type and length of the tag name and value.
6229

6230
    """
6231
    TagsLU.CheckPrereq(self)
6232
    for tag in self.op.tags:
6233
      objects.TaggableObject.ValidateTag(tag)
6234

    
6235
  def Exec(self, feedback_fn):
6236
    """Sets the tag.
6237

6238
    """
6239
    try:
6240
      for tag in self.op.tags:
6241
        self.target.AddTag(tag)
6242
    except errors.TagError, err:
6243
      raise errors.OpExecError("Error while setting tag: %s" % str(err))
6244
    try:
6245
      self.cfg.Update(self.target)
6246
    except errors.ConfigurationError:
6247
      raise errors.OpRetryError("There has been a modification to the"
6248
                                " config file and the operation has been"
6249
                                " aborted. Please retry.")
6250

    
6251

    
6252
class LUDelTags(TagsLU):
6253
  """Delete a list of tags from a given object.
6254

6255
  """
6256
  _OP_REQP = ["kind", "name", "tags"]
6257
  REQ_BGL = False
6258

    
6259
  def CheckPrereq(self):
6260
    """Check prerequisites.
6261

6262
    This checks that we have the given tag.
6263

6264
    """
6265
    TagsLU.CheckPrereq(self)
6266
    for tag in self.op.tags:
6267
      objects.TaggableObject.ValidateTag(tag)
6268
    del_tags = frozenset(self.op.tags)
6269
    cur_tags = self.target.GetTags()
6270
    if not del_tags <= cur_tags:
6271
      diff_tags = del_tags - cur_tags
6272
      diff_names = ["'%s'" % tag for tag in diff_tags]
6273
      diff_names.sort()
6274
      raise errors.OpPrereqError("Tag(s) %s not found" %
6275
                                 (",".join(diff_names)))
6276

    
6277
  def Exec(self, feedback_fn):
6278
    """Remove the tag from the object.
6279

6280
    """
6281
    for tag in self.op.tags:
6282
      self.target.RemoveTag(tag)
6283
    try:
6284
      self.cfg.Update(self.target)
6285
    except errors.ConfigurationError:
6286
      raise errors.OpRetryError("There has been a modification to the"
6287
                                " config file and the operation has been"
6288
                                " aborted. Please retry.")
6289

    
6290

    
6291
class LUTestDelay(NoHooksLU):
6292
  """Sleep for a specified amount of time.
6293

6294
  This LU sleeps on the master and/or nodes for a specified amount of
6295
  time.
6296

6297
  """
6298
  _OP_REQP = ["duration", "on_master", "on_nodes"]
6299
  REQ_BGL = False
6300

    
6301
  def ExpandNames(self):
6302
    """Expand names and set required locks.
6303

6304
    This expands the node list, if any.
6305

6306
    """
6307
    self.needed_locks = {}
6308
    if self.op.on_nodes:
6309
      # _GetWantedNodes can be used here, but is not always appropriate to use
6310
      # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
6311
      # more information.
6312
      self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
6313
      self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
6314

    
6315
  def CheckPrereq(self):
6316
    """Check prerequisites.
6317

6318
    """
6319

    
6320
  def Exec(self, feedback_fn):
6321
    """Do the actual sleep.
6322

6323
    """
6324
    if self.op.on_master:
6325
      if not utils.TestDelay(self.op.duration):
6326
        raise errors.OpExecError("Error during master delay test")
6327
    if self.op.on_nodes:
6328
      result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
6329
      if not result:
6330
        raise errors.OpExecError("Complete failure from rpc call")
6331
      for node, node_result in result.items():
6332
        node_result.Raise()
6333
        if not node_result.data:
6334
          raise errors.OpExecError("Failure during rpc call to node %s,"
6335
                                   " result: %s" % (node, node_result.data))
6336

    
6337

    
6338
class IAllocator(object):
6339
  """IAllocator framework.
6340

6341
  An IAllocator instance has three sets of attributes:
6342
    - cfg that is needed to query the cluster
6343
    - input data (all members of the _KEYS class attribute are required)
6344
    - four buffer attributes (in|out_data|text), that represent the
6345
      input (to the external script) in text and data structure format,
6346
      and the output from it, again in two formats
6347
    - the result variables from the script (success, info, nodes) for
6348
      easy usage
6349

6350
  """
6351
  _ALLO_KEYS = [
6352
    "mem_size", "disks", "disk_template",
6353
    "os", "tags", "nics", "vcpus", "hypervisor",
6354
    ]
6355
  _RELO_KEYS = [
6356
    "relocate_from",
6357
    ]
6358

    
6359
  def __init__(self, lu, mode, name, **kwargs):
6360
    self.lu = lu
6361
    # init buffer variables
6362
    self.in_text = self.out_text = self.in_data = self.out_data = None
6363
    # init all input fields so that pylint is happy
6364
    self.mode = mode
6365
    self.name = name
6366
    self.mem_size = self.disks = self.disk_template = None
6367
    self.os = self.tags = self.nics = self.vcpus = None
6368
    self.hypervisor = None
6369
    self.relocate_from = None
6370
    # computed fields
6371
    self.required_nodes = None
6372
    # init result fields
6373
    self.success = self.info = self.nodes = None
6374
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6375
      keyset = self._ALLO_KEYS
6376
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
6377
      keyset = self._RELO_KEYS
6378
    else:
6379
      raise errors.ProgrammerError("Unknown mode '%s' passed to the"
6380
                                   " IAllocator" % self.mode)
6381
    for key in kwargs:
6382
      if key not in keyset:
6383
        raise errors.ProgrammerError("Invalid input parameter '%s' to"
6384
                                     " IAllocator" % key)
6385
      setattr(self, key, kwargs[key])
6386
    for key in keyset:
6387
      if key not in kwargs:
6388
        raise errors.ProgrammerError("Missing input parameter '%s' to"
6389
                                     " IAllocator" % key)
6390
    self._BuildInputData()
6391

    
6392
  def _ComputeClusterData(self):
6393
    """Compute the generic allocator input data.
6394

6395
    This is the data that is independent of the actual operation.
6396

6397
    """
6398
    cfg = self.lu.cfg
6399
    cluster_info = cfg.GetClusterInfo()
6400
    # cluster data
6401
    data = {
6402
      "version": 1,
6403
      "cluster_name": cfg.GetClusterName(),
6404
      "cluster_tags": list(cluster_info.GetTags()),
6405
      "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
6406
      # we don't have job IDs
6407
      }
6408
    iinfo = cfg.GetAllInstancesInfo().values()
6409
    i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
6410

    
6411
    # node data
6412
    node_results = {}
6413
    node_list = cfg.GetNodeList()
6414

    
6415
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6416
      hypervisor_name = self.hypervisor
6417
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
6418
      hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
6419

    
6420
    node_data = self.lu.rpc.call_node_info(node_list, cfg.GetVGName(),
6421
                                           hypervisor_name)
6422
    node_iinfo = self.lu.rpc.call_all_instances_info(node_list,
6423
                       cluster_info.enabled_hypervisors)
6424
    for nname, nresult in node_data.items():
6425
      # first fill in static (config-based) values
6426
      ninfo = cfg.GetNodeInfo(nname)
6427
      pnr = {
6428
        "tags": list(ninfo.GetTags()),
6429
        "primary_ip": ninfo.primary_ip,
6430
        "secondary_ip": ninfo.secondary_ip,
6431
        "offline": ninfo.offline,
6432
        "master_candidate": ninfo.master_candidate,
6433
        }
6434

    
6435
      if not ninfo.offline:
6436
        nresult.Raise()
6437
        if not isinstance(nresult.data, dict):
6438
          raise errors.OpExecError("Can't get data for node %s" % nname)
6439
        remote_info = nresult.data
6440
        for attr in ['memory_total', 'memory_free', 'memory_dom0',
6441
                     'vg_size', 'vg_free', 'cpu_total']:
6442
          if attr not in remote_info:
6443
            raise errors.OpExecError("Node '%s' didn't return attribute"
6444
                                     " '%s'" % (nname, attr))
6445
          try:
6446
            remote_info[attr] = int(remote_info[attr])
6447
          except ValueError, err:
6448
            raise errors.OpExecError("Node '%s' returned invalid value"
6449
                                     " for '%s': %s" % (nname, attr, err))
6450
        # compute memory used by primary instances
6451
        i_p_mem = i_p_up_mem = 0
6452
        for iinfo, beinfo in i_list:
6453
          if iinfo.primary_node == nname:
6454
            i_p_mem += beinfo[constants.BE_MEMORY]
6455
            if iinfo.name not in node_iinfo[nname].data:
6456
              i_used_mem = 0
6457
            else:
6458
              i_used_mem = int(node_iinfo[nname].data[iinfo.name]['memory'])
6459
            i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
6460
            remote_info['memory_free'] -= max(0, i_mem_diff)
6461

    
6462
            if iinfo.admin_up:
6463
              i_p_up_mem += beinfo[constants.BE_MEMORY]
6464

    
6465
        # compute memory used by instances
6466
        pnr_dyn = {
6467
          "total_memory": remote_info['memory_total'],
6468
          "reserved_memory": remote_info['memory_dom0'],
6469
          "free_memory": remote_info['memory_free'],
6470
          "total_disk": remote_info['vg_size'],
6471
          "free_disk": remote_info['vg_free'],
6472
          "total_cpus": remote_info['cpu_total'],
6473
          "i_pri_memory": i_p_mem,
6474
          "i_pri_up_memory": i_p_up_mem,
6475
          }
6476
        pnr.update(pnr_dyn)
6477

    
6478
      node_results[nname] = pnr
6479
    data["nodes"] = node_results
6480

    
6481
    # instance data
6482
    instance_data = {}
6483
    for iinfo, beinfo in i_list:
6484
      nic_data = [{"mac": n.mac, "ip": n.ip, "bridge": n.bridge}
6485
                  for n in iinfo.nics]
6486
      pir = {
6487
        "tags": list(iinfo.GetTags()),
6488
        "admin_up": iinfo.admin_up,
6489
        "vcpus": beinfo[constants.BE_VCPUS],
6490
        "memory": beinfo[constants.BE_MEMORY],
6491
        "os": iinfo.os,
6492
        "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
6493
        "nics": nic_data,
6494
        "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
6495
        "disk_template": iinfo.disk_template,
6496
        "hypervisor": iinfo.hypervisor,
6497
        }
6498
      instance_data[iinfo.name] = pir
6499

    
6500
    data["instances"] = instance_data
6501

    
6502
    self.in_data = data
6503

    
6504
  def _AddNewInstance(self):
6505
    """Add new instance data to allocator structure.
6506

6507
    This in combination with _AllocatorGetClusterData will create the
6508
    correct structure needed as input for the allocator.
6509

6510
    The checks for the completeness of the opcode must have already been
6511
    done.
6512

6513
    """
6514
    data = self.in_data
6515

    
6516
    disk_space = _ComputeDiskSize(self.disk_template, self.disks)
6517

    
6518
    if self.disk_template in constants.DTS_NET_MIRROR:
6519
      self.required_nodes = 2
6520
    else:
6521
      self.required_nodes = 1
6522
    request = {
6523
      "type": "allocate",
6524
      "name": self.name,
6525
      "disk_template": self.disk_template,
6526
      "tags": self.tags,
6527
      "os": self.os,
6528
      "vcpus": self.vcpus,
6529
      "memory": self.mem_size,
6530
      "disks": self.disks,
6531
      "disk_space_total": disk_space,
6532
      "nics": self.nics,
6533
      "required_nodes": self.required_nodes,
6534
      }
6535
    data["request"] = request
6536

    
6537
  def _AddRelocateInstance(self):
6538
    """Add relocate instance data to allocator structure.
6539

6540
    This in combination with _IAllocatorGetClusterData will create the
6541
    correct structure needed as input for the allocator.
6542

6543
    The checks for the completeness of the opcode must have already been
6544
    done.
6545

6546
    """
6547
    instance = self.lu.cfg.GetInstanceInfo(self.name)
6548
    if instance is None:
6549
      raise errors.ProgrammerError("Unknown instance '%s' passed to"
6550
                                   " IAllocator" % self.name)
6551

    
6552
    if instance.disk_template not in constants.DTS_NET_MIRROR:
6553
      raise errors.OpPrereqError("Can't relocate non-mirrored instances")
6554

    
6555
    if len(instance.secondary_nodes) != 1:
6556
      raise errors.OpPrereqError("Instance has not exactly one secondary node")
6557

    
6558
    self.required_nodes = 1
6559
    disk_sizes = [{'size': disk.size} for disk in instance.disks]
6560
    disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
6561

    
6562
    request = {
6563
      "type": "relocate",
6564
      "name": self.name,
6565
      "disk_space_total": disk_space,
6566
      "required_nodes": self.required_nodes,
6567
      "relocate_from": self.relocate_from,
6568
      }
6569
    self.in_data["request"] = request
6570

    
6571
  def _BuildInputData(self):
6572
    """Build input data structures.
6573

6574
    """
6575
    self._ComputeClusterData()
6576

    
6577
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6578
      self._AddNewInstance()
6579
    else:
6580
      self._AddRelocateInstance()
6581

    
6582
    self.in_text = serializer.Dump(self.in_data)
6583

    
6584
  def Run(self, name, validate=True, call_fn=None):
6585
    """Run an instance allocator and return the results.
6586

6587
    """
6588
    if call_fn is None:
6589
      call_fn = self.lu.rpc.call_iallocator_runner
6590
    data = self.in_text
6591

    
6592
    result = call_fn(self.lu.cfg.GetMasterNode(), name, self.in_text)
6593
    result.Raise()
6594

    
6595
    if not isinstance(result.data, (list, tuple)) or len(result.data) != 4:
6596
      raise errors.OpExecError("Invalid result from master iallocator runner")
6597

    
6598
    rcode, stdout, stderr, fail = result.data
6599

    
6600
    if rcode == constants.IARUN_NOTFOUND:
6601
      raise errors.OpExecError("Can't find allocator '%s'" % name)
6602
    elif rcode == constants.IARUN_FAILURE:
6603
      raise errors.OpExecError("Instance allocator call failed: %s,"
6604
                               " output: %s" % (fail, stdout+stderr))
6605
    self.out_text = stdout
6606
    if validate:
6607
      self._ValidateResult()
6608

    
6609
  def _ValidateResult(self):
6610
    """Process the allocator results.
6611

6612
    This will process and if successful save the result in
6613
    self.out_data and the other parameters.
6614

6615
    """
6616
    try:
6617
      rdict = serializer.Load(self.out_text)
6618
    except Exception, err:
6619
      raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
6620

    
6621
    if not isinstance(rdict, dict):
6622
      raise errors.OpExecError("Can't parse iallocator results: not a dict")
6623

    
6624
    for key in "success", "info", "nodes":
6625
      if key not in rdict:
6626
        raise errors.OpExecError("Can't parse iallocator results:"
6627
                                 " missing key '%s'" % key)
6628
      setattr(self, key, rdict[key])
6629

    
6630
    if not isinstance(rdict["nodes"], list):
6631
      raise errors.OpExecError("Can't parse iallocator results: 'nodes' key"
6632
                               " is not a list")
6633
    self.out_data = rdict
6634

    
6635

    
6636
class LUTestAllocator(NoHooksLU):
6637
  """Run allocator tests.
6638

6639
  This LU runs the allocator tests
6640

6641
  """
6642
  _OP_REQP = ["direction", "mode", "name"]
6643

    
6644
  def CheckPrereq(self):
6645
    """Check prerequisites.
6646

6647
    This checks the opcode parameters depending on the director and mode test.
6648

6649
    """
6650
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
6651
      for attr in ["name", "mem_size", "disks", "disk_template",
6652
                   "os", "tags", "nics", "vcpus"]:
6653
        if not hasattr(self.op, attr):
6654
          raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
6655
                                     attr)
6656
      iname = self.cfg.ExpandInstanceName(self.op.name)
6657
      if iname is not None:
6658
        raise errors.OpPrereqError("Instance '%s' already in the cluster" %
6659
                                   iname)
6660
      if not isinstance(self.op.nics, list):
6661
        raise errors.OpPrereqError("Invalid parameter 'nics'")
6662
      for row in self.op.nics:
6663
        if (not isinstance(row, dict) or
6664
            "mac" not in row or
6665
            "ip" not in row or
6666
            "bridge" not in row):
6667
          raise errors.OpPrereqError("Invalid contents of the"
6668
                                     " 'nics' parameter")
6669
      if not isinstance(self.op.disks, list):
6670
        raise errors.OpPrereqError("Invalid parameter 'disks'")
6671
      for row in self.op.disks:
6672
        if (not isinstance(row, dict) or
6673
            "size" not in row or
6674
            not isinstance(row["size"], int) or
6675
            "mode" not in row or
6676
            row["mode"] not in ['r', 'w']):
6677
          raise errors.OpPrereqError("Invalid contents of the"
6678
                                     " 'disks' parameter")
6679
      if not hasattr(self.op, "hypervisor") or self.op.hypervisor is None:
6680
        self.op.hypervisor = self.cfg.GetHypervisorType()
6681
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
6682
      if not hasattr(self.op, "name"):
6683
        raise errors.OpPrereqError("Missing attribute 'name' on opcode input")
6684
      fname = self.cfg.ExpandInstanceName(self.op.name)
6685
      if fname is None:
6686
        raise errors.OpPrereqError("Instance '%s' not found for relocation" %
6687
                                   self.op.name)
6688
      self.op.name = fname
6689
      self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
6690
    else:
6691
      raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
6692
                                 self.op.mode)
6693

    
6694
    if self.op.direction == constants.IALLOCATOR_DIR_OUT:
6695
      if not hasattr(self.op, "allocator") or self.op.allocator is None:
6696
        raise errors.OpPrereqError("Missing allocator name")
6697
    elif self.op.direction != constants.IALLOCATOR_DIR_IN:
6698
      raise errors.OpPrereqError("Wrong allocator test '%s'" %
6699
                                 self.op.direction)
6700

    
6701
  def Exec(self, feedback_fn):
6702
    """Run the allocator test.
6703

6704
    """
6705
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
6706
      ial = IAllocator(self,
6707
                       mode=self.op.mode,
6708
                       name=self.op.name,
6709
                       mem_size=self.op.mem_size,
6710
                       disks=self.op.disks,
6711
                       disk_template=self.op.disk_template,
6712
                       os=self.op.os,
6713
                       tags=self.op.tags,
6714
                       nics=self.op.nics,
6715
                       vcpus=self.op.vcpus,
6716
                       hypervisor=self.op.hypervisor,
6717
                       )
6718
    else:
6719
      ial = IAllocator(self,
6720
                       mode=self.op.mode,
6721
                       name=self.op.name,
6722
                       relocate_from=list(self.relocate_from),
6723
                       )
6724

    
6725
    if self.op.direction == constants.IALLOCATOR_DIR_IN:
6726
      result = ial.in_text
6727
    else:
6728
      ial.Run(self.op.allocator, validate=False)
6729
      result = ial.out_text
6730
    return result