Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ 3b559640

History | View | Annotate | Download (231.7 kB)

1
#
2
#
3

    
4
# Copyright (C) 2006, 2007, 2008 Google Inc.
5
#
6
# This program is free software; you can redistribute it and/or modify
7
# it under the terms of the GNU General Public License as published by
8
# the Free Software Foundation; either version 2 of the License, or
9
# (at your option) any later version.
10
#
11
# This program is distributed in the hope that it will be useful, but
12
# WITHOUT ANY WARRANTY; without even the implied warranty of
13
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14
# General Public License for more details.
15
#
16
# You should have received a copy of the GNU General Public License
17
# along with this program; if not, write to the Free Software
18
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
19
# 02110-1301, USA.
20

    
21

    
22
"""Module implementing the master-side code."""
23

    
24
# pylint: disable-msg=W0613,W0201
25

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

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

    
48

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

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

61
  Note that all commands require root permissions.
62

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

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

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

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

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

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

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

    
109
  ssh = property(fget=__GetSSH)
110

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

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

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

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

126
    """
127
    pass
128

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

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

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

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

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

150
    Examples::
151

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

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

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

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

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

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

189
    """
190

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

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

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

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

205
    """
206
    raise NotImplementedError
207

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

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

215
    """
216
    raise NotImplementedError
217

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

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

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

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

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

237
    """
238
    raise NotImplementedError
239

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

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

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

258
    """
259
    return lu_result
260

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
326
    del self.recalculate_locks[locking.LEVEL_NODE]
327

    
328

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

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

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

    
339

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

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

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

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

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

    
366
  return utils.NiceSort(wanted)
367

    
368

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

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

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

    
385
  if instances:
386
    wanted = []
387

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

    
394
  else:
395
    wanted = lu.cfg.GetInstanceList()
396
  return utils.NiceSort(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:
654
      feedback_fn("  - ERROR: connection to %s failed" % (node))
655
      return True
656

    
657
    if local_version != remote_version:
658
      feedback_fn("  - ERROR: sw version mismatch: master %s, node(%s) %s" %
659
                      (local_version, node, remote_version))
660
      return True
661

    
662
    # checks vg existance and size > 20G
663

    
664
    bad = False
665
    vglist = node_result.get(constants.NV_VGLIST, None)
666
    if not vglist:
667
      feedback_fn("  - ERROR: unable to check volume groups on node %s." %
668
                      (node,))
669
      bad = True
670
    else:
671
      vgstatus = utils.CheckVolumeGroupSize(vglist, self.cfg.GetVGName(),
672
                                            constants.MIN_VG_SIZE)
673
      if vgstatus:
674
        feedback_fn("  - ERROR: %s on node %s" % (vgstatus, node))
675
        bad = True
676

    
677
    # checks config file checksum
678

    
679
    remote_cksum = node_result.get(constants.NV_FILELIST, None)
680
    if not isinstance(remote_cksum, dict):
681
      bad = True
682
      feedback_fn("  - ERROR: node hasn't returned file checksum data")
683
    else:
684
      for file_name in file_list:
685
        node_is_mc = nodeinfo.master_candidate
686
        must_have_file = file_name not in master_files
687
        if file_name not in remote_cksum:
688
          if node_is_mc or must_have_file:
689
            bad = True
690
            feedback_fn("  - ERROR: file '%s' missing" % file_name)
691
        elif remote_cksum[file_name] != local_cksum[file_name]:
692
          if node_is_mc or must_have_file:
693
            bad = True
694
            feedback_fn("  - ERROR: file '%s' has wrong checksum" % file_name)
695
          else:
696
            # not candidate and this is not a must-have file
697
            bad = True
698
            feedback_fn("  - ERROR: non master-candidate has old/wrong file"
699
                        " '%s'" % file_name)
700
        else:
701
          # all good, except non-master/non-must have combination
702
          if not node_is_mc and not must_have_file:
703
            feedback_fn("  - ERROR: file '%s' should not exist on non master"
704
                        " candidates" % file_name)
705

    
706
    # checks ssh to any
707

    
708
    if constants.NV_NODELIST not in node_result:
709
      bad = True
710
      feedback_fn("  - ERROR: node hasn't returned node ssh connectivity data")
711
    else:
712
      if node_result[constants.NV_NODELIST]:
713
        bad = True
714
        for node in node_result[constants.NV_NODELIST]:
715
          feedback_fn("  - ERROR: ssh communication with node '%s': %s" %
716
                          (node, node_result[constants.NV_NODELIST][node]))
717

    
718
    if constants.NV_NODENETTEST not in node_result:
719
      bad = True
720
      feedback_fn("  - ERROR: node hasn't returned node tcp connectivity data")
721
    else:
722
      if node_result[constants.NV_NODENETTEST]:
723
        bad = True
724
        nlist = utils.NiceSort(node_result[constants.NV_NODENETTEST].keys())
725
        for node in nlist:
726
          feedback_fn("  - ERROR: tcp communication with node '%s': %s" %
727
                          (node, node_result[constants.NV_NODENETTEST][node]))
728

    
729
    hyp_result = node_result.get(constants.NV_HYPERVISOR, None)
730
    if isinstance(hyp_result, dict):
731
      for hv_name, hv_result in hyp_result.iteritems():
732
        if hv_result is not None:
733
          feedback_fn("  - ERROR: hypervisor %s verify failure: '%s'" %
734
                      (hv_name, hv_result))
735

    
736
    # check used drbd list
737
    used_minors = node_result.get(constants.NV_DRBDLIST, [])
738
    for minor, (iname, must_exist) in drbd_map.items():
739
      if minor not in used_minors and must_exist:
740
        feedback_fn("  - ERROR: drbd minor %d of instance %s is not active" %
741
                    (minor, iname))
742
        bad = True
743
    for minor in used_minors:
744
      if minor not in drbd_map:
745
        feedback_fn("  - ERROR: unallocated drbd minor %d is in use" % minor)
746
        bad = True
747

    
748
    return bad
749

    
750
  def _VerifyInstance(self, instance, instanceconfig, node_vol_is,
751
                      node_instance, feedback_fn, n_offline):
752
    """Verify an instance.
753

754
    This function checks to see if the required block devices are
755
    available on the instance's node.
756

757
    """
758
    bad = False
759

    
760
    node_current = instanceconfig.primary_node
761

    
762
    node_vol_should = {}
763
    instanceconfig.MapLVsByNode(node_vol_should)
764

    
765
    for node in node_vol_should:
766
      if node in n_offline:
767
        # ignore missing volumes on offline nodes
768
        continue
769
      for volume in node_vol_should[node]:
770
        if node not in node_vol_is or volume not in node_vol_is[node]:
771
          feedback_fn("  - ERROR: volume %s missing on node %s" %
772
                          (volume, node))
773
          bad = True
774

    
775
    if instanceconfig.admin_up:
776
      if ((node_current not in node_instance or
777
          not instance in node_instance[node_current]) and
778
          node_current not in n_offline):
779
        feedback_fn("  - ERROR: instance %s not running on node %s" %
780
                        (instance, node_current))
781
        bad = True
782

    
783
    for node in node_instance:
784
      if (not node == node_current):
785
        if instance in node_instance[node]:
786
          feedback_fn("  - ERROR: instance %s should not run on node %s" %
787
                          (instance, node))
788
          bad = True
789

    
790
    return bad
791

    
792
  def _VerifyOrphanVolumes(self, node_vol_should, node_vol_is, feedback_fn):
793
    """Verify if there are any unknown volumes in the cluster.
794

795
    The .os, .swap and backup volumes are ignored. All other volumes are
796
    reported as unknown.
797

798
    """
799
    bad = False
800

    
801
    for node in node_vol_is:
802
      for volume in node_vol_is[node]:
803
        if node not in node_vol_should or volume not in node_vol_should[node]:
804
          feedback_fn("  - ERROR: volume %s on node %s should not exist" %
805
                      (volume, node))
806
          bad = True
807
    return bad
808

    
809
  def _VerifyOrphanInstances(self, instancelist, node_instance, feedback_fn):
810
    """Verify the list of running instances.
811

812
    This checks what instances are running but unknown to the cluster.
813

814
    """
815
    bad = False
816
    for node in node_instance:
817
      for runninginstance in node_instance[node]:
818
        if runninginstance not in instancelist:
819
          feedback_fn("  - ERROR: instance %s on node %s should not exist" %
820
                          (runninginstance, node))
821
          bad = True
822
    return bad
823

    
824
  def _VerifyNPlusOneMemory(self, node_info, instance_cfg, feedback_fn):
825
    """Verify N+1 Memory Resilience.
826

827
    Check that if one single node dies we can still start all the instances it
828
    was primary for.
829

830
    """
831
    bad = False
832

    
833
    for node, nodeinfo in node_info.iteritems():
834
      # This code checks that every node which is now listed as secondary has
835
      # enough memory to host all instances it is supposed to should a single
836
      # other node in the cluster fail.
837
      # FIXME: not ready for failover to an arbitrary node
838
      # FIXME: does not support file-backed instances
839
      # WARNING: we currently take into account down instances as well as up
840
      # ones, considering that even if they're down someone might want to start
841
      # them even in the event of a node failure.
842
      for prinode, instances in nodeinfo['sinst-by-pnode'].iteritems():
843
        needed_mem = 0
844
        for instance in instances:
845
          bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
846
          if bep[constants.BE_AUTO_BALANCE]:
847
            needed_mem += bep[constants.BE_MEMORY]
848
        if nodeinfo['mfree'] < needed_mem:
849
          feedback_fn("  - ERROR: not enough memory on node %s to accomodate"
850
                      " failovers should node %s fail" % (node, prinode))
851
          bad = True
852
    return bad
853

    
854
  def CheckPrereq(self):
855
    """Check prerequisites.
856

857
    Transform the list of checks we're going to skip into a set and check that
858
    all its members are valid.
859

860
    """
861
    self.skip_set = frozenset(self.op.skip_checks)
862
    if not constants.VERIFY_OPTIONAL_CHECKS.issuperset(self.skip_set):
863
      raise errors.OpPrereqError("Invalid checks to be skipped specified")
864

    
865
  def BuildHooksEnv(self):
866
    """Build hooks env.
867

868
    Cluster-Verify hooks just rone in the post phase and their failure makes
869
    the output be logged in the verify output and the verification to fail.
870

871
    """
872
    all_nodes = self.cfg.GetNodeList()
873
    # TODO: populate the environment with useful information for verify hooks
874
    env = {}
875
    return env, [], all_nodes
876

    
877
  def Exec(self, feedback_fn):
878
    """Verify integrity of cluster, performing various test on nodes.
879

880
    """
881
    bad = False
882
    feedback_fn("* Verifying global settings")
883
    for msg in self.cfg.VerifyConfig():
884
      feedback_fn("  - ERROR: %s" % msg)
885

    
886
    vg_name = self.cfg.GetVGName()
887
    hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
888
    nodelist = utils.NiceSort(self.cfg.GetNodeList())
889
    nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
890
    instancelist = utils.NiceSort(self.cfg.GetInstanceList())
891
    instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
892
                        for iname in instancelist)
893
    i_non_redundant = [] # Non redundant instances
894
    i_non_a_balanced = [] # Non auto-balanced instances
895
    n_offline = [] # List of offline nodes
896
    node_volume = {}
897
    node_instance = {}
898
    node_info = {}
899
    instance_cfg = {}
900

    
901
    # FIXME: verify OS list
902
    # do local checksums
903
    master_files = [constants.CLUSTER_CONF_FILE]
904

    
905
    file_names = ssconf.SimpleStore().GetFileList()
906
    file_names.append(constants.SSL_CERT_FILE)
907
    file_names.append(constants.RAPI_CERT_FILE)
908
    file_names.extend(master_files)
909

    
910
    local_checksums = utils.FingerprintFiles(file_names)
911

    
912
    feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
913
    node_verify_param = {
914
      constants.NV_FILELIST: file_names,
915
      constants.NV_NODELIST: [node.name for node in nodeinfo
916
                              if not node.offline],
917
      constants.NV_HYPERVISOR: hypervisors,
918
      constants.NV_NODENETTEST: [(node.name, node.primary_ip,
919
                                  node.secondary_ip) for node in nodeinfo
920
                                 if not node.offline],
921
      constants.NV_LVLIST: vg_name,
922
      constants.NV_INSTANCELIST: hypervisors,
923
      constants.NV_VGLIST: None,
924
      constants.NV_VERSION: None,
925
      constants.NV_HVINFO: self.cfg.GetHypervisorType(),
926
      constants.NV_DRBDLIST: None,
927
      }
928
    all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
929
                                           self.cfg.GetClusterName())
930

    
931
    cluster = self.cfg.GetClusterInfo()
932
    master_node = self.cfg.GetMasterNode()
933
    all_drbd_map = self.cfg.ComputeDRBDMap()
934

    
935
    for node_i in nodeinfo:
936
      node = node_i.name
937
      nresult = all_nvinfo[node].data
938

    
939
      if node_i.offline:
940
        feedback_fn("* Skipping offline node %s" % (node,))
941
        n_offline.append(node)
942
        continue
943

    
944
      if node == master_node:
945
        ntype = "master"
946
      elif node_i.master_candidate:
947
        ntype = "master candidate"
948
      else:
949
        ntype = "regular"
950
      feedback_fn("* Verifying node %s (%s)" % (node, ntype))
951

    
952
      if all_nvinfo[node].failed or not isinstance(nresult, dict):
953
        feedback_fn("  - ERROR: connection to %s failed" % (node,))
954
        bad = True
955
        continue
956

    
957
      node_drbd = {}
958
      for minor, instance in all_drbd_map[node].items():
959
        instance = instanceinfo[instance]
960
        node_drbd[minor] = (instance.name, instance.admin_up)
961
      result = self._VerifyNode(node_i, file_names, local_checksums,
962
                                nresult, feedback_fn, master_files,
963
                                node_drbd)
964
      bad = bad or result
965

    
966
      lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
967
      if isinstance(lvdata, basestring):
968
        feedback_fn("  - ERROR: LVM problem on node %s: %s" %
969
                    (node, lvdata.encode('string_escape')))
970
        bad = True
971
        node_volume[node] = {}
972
      elif not isinstance(lvdata, dict):
973
        feedback_fn("  - ERROR: connection to %s failed (lvlist)" % (node,))
974
        bad = True
975
        continue
976
      else:
977
        node_volume[node] = lvdata
978

    
979
      # node_instance
980
      idata = nresult.get(constants.NV_INSTANCELIST, None)
981
      if not isinstance(idata, list):
982
        feedback_fn("  - ERROR: connection to %s failed (instancelist)" %
983
                    (node,))
984
        bad = True
985
        continue
986

    
987
      node_instance[node] = idata
988

    
989
      # node_info
990
      nodeinfo = nresult.get(constants.NV_HVINFO, None)
991
      if not isinstance(nodeinfo, dict):
992
        feedback_fn("  - ERROR: connection to %s failed (hvinfo)" % (node,))
993
        bad = True
994
        continue
995

    
996
      try:
997
        node_info[node] = {
998
          "mfree": int(nodeinfo['memory_free']),
999
          "dfree": int(nresult[constants.NV_VGLIST][vg_name]),
1000
          "pinst": [],
1001
          "sinst": [],
1002
          # dictionary holding all instances this node is secondary for,
1003
          # grouped by their primary node. Each key is a cluster node, and each
1004
          # value is a list of instances which have the key as primary and the
1005
          # current node as secondary.  this is handy to calculate N+1 memory
1006
          # availability if you can only failover from a primary to its
1007
          # secondary.
1008
          "sinst-by-pnode": {},
1009
        }
1010
      except ValueError:
1011
        feedback_fn("  - ERROR: invalid value returned from node %s" % (node,))
1012
        bad = True
1013
        continue
1014

    
1015
    node_vol_should = {}
1016

    
1017
    for instance in instancelist:
1018
      feedback_fn("* Verifying instance %s" % instance)
1019
      inst_config = instanceinfo[instance]
1020
      result =  self._VerifyInstance(instance, inst_config, node_volume,
1021
                                     node_instance, feedback_fn, n_offline)
1022
      bad = bad or result
1023
      inst_nodes_offline = []
1024

    
1025
      inst_config.MapLVsByNode(node_vol_should)
1026

    
1027
      instance_cfg[instance] = inst_config
1028

    
1029
      pnode = inst_config.primary_node
1030
      if pnode in node_info:
1031
        node_info[pnode]['pinst'].append(instance)
1032
      elif pnode not in n_offline:
1033
        feedback_fn("  - ERROR: instance %s, connection to primary node"
1034
                    " %s failed" % (instance, pnode))
1035
        bad = True
1036

    
1037
      if pnode in n_offline:
1038
        inst_nodes_offline.append(pnode)
1039

    
1040
      # If the instance is non-redundant we cannot survive losing its primary
1041
      # node, so we are not N+1 compliant. On the other hand we have no disk
1042
      # templates with more than one secondary so that situation is not well
1043
      # supported either.
1044
      # FIXME: does not support file-backed instances
1045
      if len(inst_config.secondary_nodes) == 0:
1046
        i_non_redundant.append(instance)
1047
      elif len(inst_config.secondary_nodes) > 1:
1048
        feedback_fn("  - WARNING: multiple secondaries for instance %s"
1049
                    % instance)
1050

    
1051
      if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
1052
        i_non_a_balanced.append(instance)
1053

    
1054
      for snode in inst_config.secondary_nodes:
1055
        if snode in node_info:
1056
          node_info[snode]['sinst'].append(instance)
1057
          if pnode not in node_info[snode]['sinst-by-pnode']:
1058
            node_info[snode]['sinst-by-pnode'][pnode] = []
1059
          node_info[snode]['sinst-by-pnode'][pnode].append(instance)
1060
        elif snode not in n_offline:
1061
          feedback_fn("  - ERROR: instance %s, connection to secondary node"
1062
                      " %s failed" % (instance, snode))
1063
          bad = True
1064
        if snode in n_offline:
1065
          inst_nodes_offline.append(snode)
1066

    
1067
      if inst_nodes_offline:
1068
        # warn that the instance lives on offline nodes, and set bad=True
1069
        feedback_fn("  - ERROR: instance lives on offline node(s) %s" %
1070
                    ", ".join(inst_nodes_offline))
1071
        bad = True
1072

    
1073
    feedback_fn("* Verifying orphan volumes")
1074
    result = self._VerifyOrphanVolumes(node_vol_should, node_volume,
1075
                                       feedback_fn)
1076
    bad = bad or result
1077

    
1078
    feedback_fn("* Verifying remaining instances")
1079
    result = self._VerifyOrphanInstances(instancelist, node_instance,
1080
                                         feedback_fn)
1081
    bad = bad or result
1082

    
1083
    if constants.VERIFY_NPLUSONE_MEM not in self.skip_set:
1084
      feedback_fn("* Verifying N+1 Memory redundancy")
1085
      result = self._VerifyNPlusOneMemory(node_info, instance_cfg, feedback_fn)
1086
      bad = bad or result
1087

    
1088
    feedback_fn("* Other Notes")
1089
    if i_non_redundant:
1090
      feedback_fn("  - NOTICE: %d non-redundant instance(s) found."
1091
                  % len(i_non_redundant))
1092

    
1093
    if i_non_a_balanced:
1094
      feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
1095
                  % len(i_non_a_balanced))
1096

    
1097
    if n_offline:
1098
      feedback_fn("  - NOTICE: %d offline node(s) found." % len(n_offline))
1099

    
1100
    return not bad
1101

    
1102
  def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
1103
    """Analize the post-hooks' result
1104

1105
    This method analyses the hook result, handles it, and sends some
1106
    nicely-formatted feedback back to the user.
1107

1108
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
1109
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
1110
    @param hooks_results: the results of the multi-node hooks rpc call
1111
    @param feedback_fn: function used send feedback back to the caller
1112
    @param lu_result: previous Exec result
1113
    @return: the new Exec result, based on the previous result
1114
        and hook results
1115

1116
    """
1117
    # We only really run POST phase hooks, and are only interested in
1118
    # their results
1119
    if phase == constants.HOOKS_PHASE_POST:
1120
      # Used to change hooks' output to proper indentation
1121
      indent_re = re.compile('^', re.M)
1122
      feedback_fn("* Hooks Results")
1123
      if not hooks_results:
1124
        feedback_fn("  - ERROR: general communication failure")
1125
        lu_result = 1
1126
      else:
1127
        for node_name in hooks_results:
1128
          show_node_header = True
1129
          res = hooks_results[node_name]
1130
          if res.failed or res.data is False or not isinstance(res.data, list):
1131
            if res.offline:
1132
              # no need to warn or set fail return value
1133
              continue
1134
            feedback_fn("    Communication failure in hooks execution")
1135
            lu_result = 1
1136
            continue
1137
          for script, hkr, output in res.data:
1138
            if hkr == constants.HKR_FAIL:
1139
              # The node header is only shown once, if there are
1140
              # failing hooks on that node
1141
              if show_node_header:
1142
                feedback_fn("  Node %s:" % node_name)
1143
                show_node_header = False
1144
              feedback_fn("    ERROR: Script %s failed, output:" % script)
1145
              output = indent_re.sub('      ', output)
1146
              feedback_fn("%s" % output)
1147
              lu_result = 1
1148

    
1149
      return lu_result
1150

    
1151

    
1152
class LUVerifyDisks(NoHooksLU):
1153
  """Verifies the cluster disks status.
1154

1155
  """
1156
  _OP_REQP = []
1157
  REQ_BGL = False
1158

    
1159
  def ExpandNames(self):
1160
    self.needed_locks = {
1161
      locking.LEVEL_NODE: locking.ALL_SET,
1162
      locking.LEVEL_INSTANCE: locking.ALL_SET,
1163
    }
1164
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
1165

    
1166
  def CheckPrereq(self):
1167
    """Check prerequisites.
1168

1169
    This has no prerequisites.
1170

1171
    """
1172
    pass
1173

    
1174
  def Exec(self, feedback_fn):
1175
    """Verify integrity of cluster disks.
1176

1177
    """
1178
    result = res_nodes, res_nlvm, res_instances, res_missing = [], {}, [], {}
1179

    
1180
    vg_name = self.cfg.GetVGName()
1181
    nodes = utils.NiceSort(self.cfg.GetNodeList())
1182
    instances = [self.cfg.GetInstanceInfo(name)
1183
                 for name in self.cfg.GetInstanceList()]
1184

    
1185
    nv_dict = {}
1186
    for inst in instances:
1187
      inst_lvs = {}
1188
      if (not inst.admin_up or
1189
          inst.disk_template not in constants.DTS_NET_MIRROR):
1190
        continue
1191
      inst.MapLVsByNode(inst_lvs)
1192
      # transform { iname: {node: [vol,],},} to {(node, vol): iname}
1193
      for node, vol_list in inst_lvs.iteritems():
1194
        for vol in vol_list:
1195
          nv_dict[(node, vol)] = inst
1196

    
1197
    if not nv_dict:
1198
      return result
1199

    
1200
    node_lvs = self.rpc.call_volume_list(nodes, vg_name)
1201

    
1202
    to_act = set()
1203
    for node in nodes:
1204
      # node_volume
1205
      lvs = node_lvs[node]
1206
      if lvs.failed:
1207
        if not lvs.offline:
1208
          self.LogWarning("Connection to node %s failed: %s" %
1209
                          (node, lvs.data))
1210
        continue
1211
      lvs = lvs.data
1212
      if isinstance(lvs, basestring):
1213
        logging.warning("Error enumerating LVs on node %s: %s", node, lvs)
1214
        res_nlvm[node] = lvs
1215
      elif not isinstance(lvs, dict):
1216
        logging.warning("Connection to node %s failed or invalid data"
1217
                        " returned", node)
1218
        res_nodes.append(node)
1219
        continue
1220

    
1221
      for lv_name, (_, lv_inactive, lv_online) in lvs.iteritems():
1222
        inst = nv_dict.pop((node, lv_name), None)
1223
        if (not lv_online and inst is not None
1224
            and inst.name not in res_instances):
1225
          res_instances.append(inst.name)
1226

    
1227
    # any leftover items in nv_dict are missing LVs, let's arrange the
1228
    # data better
1229
    for key, inst in nv_dict.iteritems():
1230
      if inst.name not in res_missing:
1231
        res_missing[inst.name] = []
1232
      res_missing[inst.name].append(key)
1233

    
1234
    return result
1235

    
1236

    
1237
class LURenameCluster(LogicalUnit):
1238
  """Rename the cluster.
1239

1240
  """
1241
  HPATH = "cluster-rename"
1242
  HTYPE = constants.HTYPE_CLUSTER
1243
  _OP_REQP = ["name"]
1244

    
1245
  def BuildHooksEnv(self):
1246
    """Build hooks env.
1247

1248
    """
1249
    env = {
1250
      "OP_TARGET": self.cfg.GetClusterName(),
1251
      "NEW_NAME": self.op.name,
1252
      }
1253
    mn = self.cfg.GetMasterNode()
1254
    return env, [mn], [mn]
1255

    
1256
  def CheckPrereq(self):
1257
    """Verify that the passed name is a valid one.
1258

1259
    """
1260
    hostname = utils.HostInfo(self.op.name)
1261

    
1262
    new_name = hostname.name
1263
    self.ip = new_ip = hostname.ip
1264
    old_name = self.cfg.GetClusterName()
1265
    old_ip = self.cfg.GetMasterIP()
1266
    if new_name == old_name and new_ip == old_ip:
1267
      raise errors.OpPrereqError("Neither the name nor the IP address of the"
1268
                                 " cluster has changed")
1269
    if new_ip != old_ip:
1270
      if utils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
1271
        raise errors.OpPrereqError("The given cluster IP address (%s) is"
1272
                                   " reachable on the network. Aborting." %
1273
                                   new_ip)
1274

    
1275
    self.op.name = new_name
1276

    
1277
  def Exec(self, feedback_fn):
1278
    """Rename the cluster.
1279

1280
    """
1281
    clustername = self.op.name
1282
    ip = self.ip
1283

    
1284
    # shutdown the master IP
1285
    master = self.cfg.GetMasterNode()
1286
    result = self.rpc.call_node_stop_master(master, False)
1287
    if result.failed or not result.data:
1288
      raise errors.OpExecError("Could not disable the master role")
1289

    
1290
    try:
1291
      cluster = self.cfg.GetClusterInfo()
1292
      cluster.cluster_name = clustername
1293
      cluster.master_ip = ip
1294
      self.cfg.Update(cluster)
1295

    
1296
      # update the known hosts file
1297
      ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
1298
      node_list = self.cfg.GetNodeList()
1299
      try:
1300
        node_list.remove(master)
1301
      except ValueError:
1302
        pass
1303
      result = self.rpc.call_upload_file(node_list,
1304
                                         constants.SSH_KNOWN_HOSTS_FILE)
1305
      for to_node, to_result in result.iteritems():
1306
        if to_result.failed or not to_result.data:
1307
          logging.error("Copy of file %s to node %s failed",
1308
                        constants.SSH_KNOWN_HOSTS_FILE, to_node)
1309

    
1310
    finally:
1311
      result = self.rpc.call_node_start_master(master, False)
1312
      if result.failed or not result.data:
1313
        self.LogWarning("Could not re-enable the master role on"
1314
                        " the master, please restart manually.")
1315

    
1316

    
1317
def _RecursiveCheckIfLVMBased(disk):
1318
  """Check if the given disk or its children are lvm-based.
1319

1320
  @type disk: L{objects.Disk}
1321
  @param disk: the disk to check
1322
  @rtype: booleean
1323
  @return: boolean indicating whether a LD_LV dev_type was found or not
1324

1325
  """
1326
  if disk.children:
1327
    for chdisk in disk.children:
1328
      if _RecursiveCheckIfLVMBased(chdisk):
1329
        return True
1330
  return disk.dev_type == constants.LD_LV
1331

    
1332

    
1333
class LUSetClusterParams(LogicalUnit):
1334
  """Change the parameters of the cluster.
1335

1336
  """
1337
  HPATH = "cluster-modify"
1338
  HTYPE = constants.HTYPE_CLUSTER
1339
  _OP_REQP = []
1340
  REQ_BGL = False
1341

    
1342
  def CheckParameters(self):
1343
    """Check parameters
1344

1345
    """
1346
    if not hasattr(self.op, "candidate_pool_size"):
1347
      self.op.candidate_pool_size = None
1348
    if self.op.candidate_pool_size is not None:
1349
      try:
1350
        self.op.candidate_pool_size = int(self.op.candidate_pool_size)
1351
      except ValueError, err:
1352
        raise errors.OpPrereqError("Invalid candidate_pool_size value: %s" %
1353
                                   str(err))
1354
      if self.op.candidate_pool_size < 1:
1355
        raise errors.OpPrereqError("At least one master candidate needed")
1356

    
1357
  def ExpandNames(self):
1358
    # FIXME: in the future maybe other cluster params won't require checking on
1359
    # all nodes to be modified.
1360
    self.needed_locks = {
1361
      locking.LEVEL_NODE: locking.ALL_SET,
1362
    }
1363
    self.share_locks[locking.LEVEL_NODE] = 1
1364

    
1365
  def BuildHooksEnv(self):
1366
    """Build hooks env.
1367

1368
    """
1369
    env = {
1370
      "OP_TARGET": self.cfg.GetClusterName(),
1371
      "NEW_VG_NAME": self.op.vg_name,
1372
      }
1373
    mn = self.cfg.GetMasterNode()
1374
    return env, [mn], [mn]
1375

    
1376
  def CheckPrereq(self):
1377
    """Check prerequisites.
1378

1379
    This checks whether the given params don't conflict and
1380
    if the given volume group is valid.
1381

1382
    """
1383
    # FIXME: This only works because there is only one parameter that can be
1384
    # changed or removed.
1385
    if self.op.vg_name is not None and not self.op.vg_name:
1386
      instances = self.cfg.GetAllInstancesInfo().values()
1387
      for inst in instances:
1388
        for disk in inst.disks:
1389
          if _RecursiveCheckIfLVMBased(disk):
1390
            raise errors.OpPrereqError("Cannot disable lvm storage while"
1391
                                       " lvm-based instances exist")
1392

    
1393
    node_list = self.acquired_locks[locking.LEVEL_NODE]
1394

    
1395
    # if vg_name not None, checks given volume group on all nodes
1396
    if self.op.vg_name:
1397
      vglist = self.rpc.call_vg_list(node_list)
1398
      for node in node_list:
1399
        if vglist[node].failed:
1400
          # ignoring down node
1401
          self.LogWarning("Node %s unreachable/error, ignoring" % node)
1402
          continue
1403
        vgstatus = utils.CheckVolumeGroupSize(vglist[node].data,
1404
                                              self.op.vg_name,
1405
                                              constants.MIN_VG_SIZE)
1406
        if vgstatus:
1407
          raise errors.OpPrereqError("Error on node '%s': %s" %
1408
                                     (node, vgstatus))
1409

    
1410
    self.cluster = cluster = self.cfg.GetClusterInfo()
1411
    # validate beparams changes
1412
    if self.op.beparams:
1413
      utils.CheckBEParams(self.op.beparams)
1414
      self.new_beparams = cluster.FillDict(
1415
        cluster.beparams[constants.BEGR_DEFAULT], self.op.beparams)
1416

    
1417
    # hypervisor list/parameters
1418
    self.new_hvparams = cluster.FillDict(cluster.hvparams, {})
1419
    if self.op.hvparams:
1420
      if not isinstance(self.op.hvparams, dict):
1421
        raise errors.OpPrereqError("Invalid 'hvparams' parameter on input")
1422
      for hv_name, hv_dict in self.op.hvparams.items():
1423
        if hv_name not in self.new_hvparams:
1424
          self.new_hvparams[hv_name] = hv_dict
1425
        else:
1426
          self.new_hvparams[hv_name].update(hv_dict)
1427

    
1428
    if self.op.enabled_hypervisors is not None:
1429
      self.hv_list = self.op.enabled_hypervisors
1430
    else:
1431
      self.hv_list = cluster.enabled_hypervisors
1432

    
1433
    if self.op.hvparams or self.op.enabled_hypervisors is not None:
1434
      # either the enabled list has changed, or the parameters have, validate
1435
      for hv_name, hv_params in self.new_hvparams.items():
1436
        if ((self.op.hvparams and hv_name in self.op.hvparams) or
1437
            (self.op.enabled_hypervisors and
1438
             hv_name in self.op.enabled_hypervisors)):
1439
          # either this is a new hypervisor, or its parameters have changed
1440
          hv_class = hypervisor.GetHypervisor(hv_name)
1441
          hv_class.CheckParameterSyntax(hv_params)
1442
          _CheckHVParams(self, node_list, hv_name, hv_params)
1443

    
1444
  def Exec(self, feedback_fn):
1445
    """Change the parameters of the cluster.
1446

1447
    """
1448
    if self.op.vg_name is not None:
1449
      if self.op.vg_name != self.cfg.GetVGName():
1450
        self.cfg.SetVGName(self.op.vg_name)
1451
      else:
1452
        feedback_fn("Cluster LVM configuration already in desired"
1453
                    " state, not changing")
1454
    if self.op.hvparams:
1455
      self.cluster.hvparams = self.new_hvparams
1456
    if self.op.enabled_hypervisors is not None:
1457
      self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
1458
    if self.op.beparams:
1459
      self.cluster.beparams[constants.BEGR_DEFAULT] = self.new_beparams
1460
    if self.op.candidate_pool_size is not None:
1461
      self.cluster.candidate_pool_size = self.op.candidate_pool_size
1462

    
1463
    self.cfg.Update(self.cluster)
1464

    
1465
    # we want to update nodes after the cluster so that if any errors
1466
    # happen, we have recorded and saved the cluster info
1467
    if self.op.candidate_pool_size is not None:
1468
      _AdjustCandidatePool(self)
1469

    
1470

    
1471
class LURedistributeConfig(NoHooksLU):
1472
  """Force the redistribution of cluster configuration.
1473

1474
  This is a very simple LU.
1475

1476
  """
1477
  _OP_REQP = []
1478
  REQ_BGL = False
1479

    
1480
  def ExpandNames(self):
1481
    self.needed_locks = {
1482
      locking.LEVEL_NODE: locking.ALL_SET,
1483
    }
1484
    self.share_locks[locking.LEVEL_NODE] = 1
1485

    
1486
  def CheckPrereq(self):
1487
    """Check prerequisites.
1488

1489
    """
1490

    
1491
  def Exec(self, feedback_fn):
1492
    """Redistribute the configuration.
1493

1494
    """
1495
    self.cfg.Update(self.cfg.GetClusterInfo())
1496

    
1497

    
1498
def _WaitForSync(lu, instance, oneshot=False, unlock=False):
1499
  """Sleep and poll for an instance's disk to sync.
1500

1501
  """
1502
  if not instance.disks:
1503
    return True
1504

    
1505
  if not oneshot:
1506
    lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
1507

    
1508
  node = instance.primary_node
1509

    
1510
  for dev in instance.disks:
1511
    lu.cfg.SetDiskID(dev, node)
1512

    
1513
  retries = 0
1514
  while True:
1515
    max_time = 0
1516
    done = True
1517
    cumul_degraded = False
1518
    rstats = lu.rpc.call_blockdev_getmirrorstatus(node, instance.disks)
1519
    if rstats.failed or not rstats.data:
1520
      lu.LogWarning("Can't get any data from node %s", node)
1521
      retries += 1
1522
      if retries >= 10:
1523
        raise errors.RemoteError("Can't contact node %s for mirror data,"
1524
                                 " aborting." % node)
1525
      time.sleep(6)
1526
      continue
1527
    rstats = rstats.data
1528
    retries = 0
1529
    for i, mstat in enumerate(rstats):
1530
      if mstat is None:
1531
        lu.LogWarning("Can't compute data for node %s/%s",
1532
                           node, instance.disks[i].iv_name)
1533
        continue
1534
      # we ignore the ldisk parameter
1535
      perc_done, est_time, is_degraded, _ = mstat
1536
      cumul_degraded = cumul_degraded or (is_degraded and perc_done is None)
1537
      if perc_done is not None:
1538
        done = False
1539
        if est_time is not None:
1540
          rem_time = "%d estimated seconds remaining" % est_time
1541
          max_time = est_time
1542
        else:
1543
          rem_time = "no time estimate"
1544
        lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
1545
                        (instance.disks[i].iv_name, perc_done, rem_time))
1546
    if done or oneshot:
1547
      break
1548

    
1549
    time.sleep(min(60, max_time))
1550

    
1551
  if done:
1552
    lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
1553
  return not cumul_degraded
1554

    
1555

    
1556
def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
1557
  """Check that mirrors are not degraded.
1558

1559
  The ldisk parameter, if True, will change the test from the
1560
  is_degraded attribute (which represents overall non-ok status for
1561
  the device(s)) to the ldisk (representing the local storage status).
1562

1563
  """
1564
  lu.cfg.SetDiskID(dev, node)
1565
  if ldisk:
1566
    idx = 6
1567
  else:
1568
    idx = 5
1569

    
1570
  result = True
1571
  if on_primary or dev.AssembleOnSecondary():
1572
    rstats = lu.rpc.call_blockdev_find(node, dev)
1573
    if rstats.failed or not rstats.data:
1574
      logging.warning("Node %s: disk degraded, not found or node down", node)
1575
      result = False
1576
    else:
1577
      result = result and (not rstats.data[idx])
1578
  if dev.children:
1579
    for child in dev.children:
1580
      result = result and _CheckDiskConsistency(lu, child, node, on_primary)
1581

    
1582
  return result
1583

    
1584

    
1585
class LUDiagnoseOS(NoHooksLU):
1586
  """Logical unit for OS diagnose/query.
1587

1588
  """
1589
  _OP_REQP = ["output_fields", "names"]
1590
  REQ_BGL = False
1591
  _FIELDS_STATIC = utils.FieldSet()
1592
  _FIELDS_DYNAMIC = utils.FieldSet("name", "valid", "node_status")
1593

    
1594
  def ExpandNames(self):
1595
    if self.op.names:
1596
      raise errors.OpPrereqError("Selective OS query not supported")
1597

    
1598
    _CheckOutputFields(static=self._FIELDS_STATIC,
1599
                       dynamic=self._FIELDS_DYNAMIC,
1600
                       selected=self.op.output_fields)
1601

    
1602
    # Lock all nodes, in shared mode
1603
    self.needed_locks = {}
1604
    self.share_locks[locking.LEVEL_NODE] = 1
1605
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
1606

    
1607
  def CheckPrereq(self):
1608
    """Check prerequisites.
1609

1610
    """
1611

    
1612
  @staticmethod
1613
  def _DiagnoseByOS(node_list, rlist):
1614
    """Remaps a per-node return list into an a per-os per-node dictionary
1615

1616
    @param node_list: a list with the names of all nodes
1617
    @param rlist: a map with node names as keys and OS objects as values
1618

1619
    @rtype: dict
1620
    @returns: a dictionary with osnames as keys and as value another map, with
1621
        nodes as keys and list of OS objects as values, eg::
1622

1623
          {"debian-etch": {"node1": [<object>,...],
1624
                           "node2": [<object>,]}
1625
          }
1626

1627
    """
1628
    all_os = {}
1629
    for node_name, nr in rlist.iteritems():
1630
      if nr.failed or not nr.data:
1631
        continue
1632
      for os_obj in nr.data:
1633
        if os_obj.name not in all_os:
1634
          # build a list of nodes for this os containing empty lists
1635
          # for each node in node_list
1636
          all_os[os_obj.name] = {}
1637
          for nname in node_list:
1638
            all_os[os_obj.name][nname] = []
1639
        all_os[os_obj.name][node_name].append(os_obj)
1640
    return all_os
1641

    
1642
  def Exec(self, feedback_fn):
1643
    """Compute the list of OSes.
1644

1645
    """
1646
    node_list = self.acquired_locks[locking.LEVEL_NODE]
1647
    valid_nodes = [node for node in self.cfg.GetOnlineNodeList()
1648
                   if node in node_list]
1649
    node_data = self.rpc.call_os_diagnose(valid_nodes)
1650
    if node_data == False:
1651
      raise errors.OpExecError("Can't gather the list of OSes")
1652
    pol = self._DiagnoseByOS(valid_nodes, node_data)
1653
    output = []
1654
    for os_name, os_data in pol.iteritems():
1655
      row = []
1656
      for field in self.op.output_fields:
1657
        if field == "name":
1658
          val = os_name
1659
        elif field == "valid":
1660
          val = utils.all([osl and osl[0] for osl in os_data.values()])
1661
        elif field == "node_status":
1662
          val = {}
1663
          for node_name, nos_list in os_data.iteritems():
1664
            val[node_name] = [(v.status, v.path) for v in nos_list]
1665
        else:
1666
          raise errors.ParameterError(field)
1667
        row.append(val)
1668
      output.append(row)
1669

    
1670
    return output
1671

    
1672

    
1673
class LURemoveNode(LogicalUnit):
1674
  """Logical unit for removing a node.
1675

1676
  """
1677
  HPATH = "node-remove"
1678
  HTYPE = constants.HTYPE_NODE
1679
  _OP_REQP = ["node_name"]
1680

    
1681
  def BuildHooksEnv(self):
1682
    """Build hooks env.
1683

1684
    This doesn't run on the target node in the pre phase as a failed
1685
    node would then be impossible to remove.
1686

1687
    """
1688
    env = {
1689
      "OP_TARGET": self.op.node_name,
1690
      "NODE_NAME": self.op.node_name,
1691
      }
1692
    all_nodes = self.cfg.GetNodeList()
1693
    all_nodes.remove(self.op.node_name)
1694
    return env, all_nodes, all_nodes
1695

    
1696
  def CheckPrereq(self):
1697
    """Check prerequisites.
1698

1699
    This checks:
1700
     - the node exists in the configuration
1701
     - it does not have primary or secondary instances
1702
     - it's not the master
1703

1704
    Any errors are signalled by raising errors.OpPrereqError.
1705

1706
    """
1707
    node = self.cfg.GetNodeInfo(self.cfg.ExpandNodeName(self.op.node_name))
1708
    if node is None:
1709
      raise errors.OpPrereqError, ("Node '%s' is unknown." % self.op.node_name)
1710

    
1711
    instance_list = self.cfg.GetInstanceList()
1712

    
1713
    masternode = self.cfg.GetMasterNode()
1714
    if node.name == masternode:
1715
      raise errors.OpPrereqError("Node is the master node,"
1716
                                 " you need to failover first.")
1717

    
1718
    for instance_name in instance_list:
1719
      instance = self.cfg.GetInstanceInfo(instance_name)
1720
      if node.name in instance.all_nodes:
1721
        raise errors.OpPrereqError("Instance %s is still running on the node,"
1722
                                   " please remove first." % instance_name)
1723
    self.op.node_name = node.name
1724
    self.node = node
1725

    
1726
  def Exec(self, feedback_fn):
1727
    """Removes the node from the cluster.
1728

1729
    """
1730
    node = self.node
1731
    logging.info("Stopping the node daemon and removing configs from node %s",
1732
                 node.name)
1733

    
1734
    self.context.RemoveNode(node.name)
1735

    
1736
    self.rpc.call_node_leave_cluster(node.name)
1737

    
1738
    # Promote nodes to master candidate as needed
1739
    _AdjustCandidatePool(self)
1740

    
1741

    
1742
class LUQueryNodes(NoHooksLU):
1743
  """Logical unit for querying nodes.
1744

1745
  """
1746
  _OP_REQP = ["output_fields", "names"]
1747
  REQ_BGL = False
1748
  _FIELDS_DYNAMIC = utils.FieldSet(
1749
    "dtotal", "dfree",
1750
    "mtotal", "mnode", "mfree",
1751
    "bootid",
1752
    "ctotal",
1753
    )
1754

    
1755
  _FIELDS_STATIC = utils.FieldSet(
1756
    "name", "pinst_cnt", "sinst_cnt",
1757
    "pinst_list", "sinst_list",
1758
    "pip", "sip", "tags",
1759
    "serial_no",
1760
    "master_candidate",
1761
    "master",
1762
    "offline",
1763
    )
1764

    
1765
  def ExpandNames(self):
1766
    _CheckOutputFields(static=self._FIELDS_STATIC,
1767
                       dynamic=self._FIELDS_DYNAMIC,
1768
                       selected=self.op.output_fields)
1769

    
1770
    self.needed_locks = {}
1771
    self.share_locks[locking.LEVEL_NODE] = 1
1772

    
1773
    if self.op.names:
1774
      self.wanted = _GetWantedNodes(self, self.op.names)
1775
    else:
1776
      self.wanted = locking.ALL_SET
1777

    
1778
    self.do_locking = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
1779
    if self.do_locking:
1780
      # if we don't request only static fields, we need to lock the nodes
1781
      self.needed_locks[locking.LEVEL_NODE] = self.wanted
1782

    
1783

    
1784
  def CheckPrereq(self):
1785
    """Check prerequisites.
1786

1787
    """
1788
    # The validation of the node list is done in the _GetWantedNodes,
1789
    # if non empty, and if empty, there's no validation to do
1790
    pass
1791

    
1792
  def Exec(self, feedback_fn):
1793
    """Computes the list of nodes and their attributes.
1794

1795
    """
1796
    all_info = self.cfg.GetAllNodesInfo()
1797
    if self.do_locking:
1798
      nodenames = self.acquired_locks[locking.LEVEL_NODE]
1799
    elif self.wanted != locking.ALL_SET:
1800
      nodenames = self.wanted
1801
      missing = set(nodenames).difference(all_info.keys())
1802
      if missing:
1803
        raise errors.OpExecError(
1804
          "Some nodes were removed before retrieving their data: %s" % missing)
1805
    else:
1806
      nodenames = all_info.keys()
1807

    
1808
    nodenames = utils.NiceSort(nodenames)
1809
    nodelist = [all_info[name] for name in nodenames]
1810

    
1811
    # begin data gathering
1812

    
1813
    if self.do_locking:
1814
      live_data = {}
1815
      node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
1816
                                          self.cfg.GetHypervisorType())
1817
      for name in nodenames:
1818
        nodeinfo = node_data[name]
1819
        if not nodeinfo.failed and nodeinfo.data:
1820
          nodeinfo = nodeinfo.data
1821
          fn = utils.TryConvert
1822
          live_data[name] = {
1823
            "mtotal": fn(int, nodeinfo.get('memory_total', None)),
1824
            "mnode": fn(int, nodeinfo.get('memory_dom0', None)),
1825
            "mfree": fn(int, nodeinfo.get('memory_free', None)),
1826
            "dtotal": fn(int, nodeinfo.get('vg_size', None)),
1827
            "dfree": fn(int, nodeinfo.get('vg_free', None)),
1828
            "ctotal": fn(int, nodeinfo.get('cpu_total', None)),
1829
            "bootid": nodeinfo.get('bootid', None),
1830
            }
1831
        else:
1832
          live_data[name] = {}
1833
    else:
1834
      live_data = dict.fromkeys(nodenames, {})
1835

    
1836
    node_to_primary = dict([(name, set()) for name in nodenames])
1837
    node_to_secondary = dict([(name, set()) for name in nodenames])
1838

    
1839
    inst_fields = frozenset(("pinst_cnt", "pinst_list",
1840
                             "sinst_cnt", "sinst_list"))
1841
    if inst_fields & frozenset(self.op.output_fields):
1842
      instancelist = self.cfg.GetInstanceList()
1843

    
1844
      for instance_name in instancelist:
1845
        inst = self.cfg.GetInstanceInfo(instance_name)
1846
        if inst.primary_node in node_to_primary:
1847
          node_to_primary[inst.primary_node].add(inst.name)
1848
        for secnode in inst.secondary_nodes:
1849
          if secnode in node_to_secondary:
1850
            node_to_secondary[secnode].add(inst.name)
1851

    
1852
    master_node = self.cfg.GetMasterNode()
1853

    
1854
    # end data gathering
1855

    
1856
    output = []
1857
    for node in nodelist:
1858
      node_output = []
1859
      for field in self.op.output_fields:
1860
        if field == "name":
1861
          val = node.name
1862
        elif field == "pinst_list":
1863
          val = list(node_to_primary[node.name])
1864
        elif field == "sinst_list":
1865
          val = list(node_to_secondary[node.name])
1866
        elif field == "pinst_cnt":
1867
          val = len(node_to_primary[node.name])
1868
        elif field == "sinst_cnt":
1869
          val = len(node_to_secondary[node.name])
1870
        elif field == "pip":
1871
          val = node.primary_ip
1872
        elif field == "sip":
1873
          val = node.secondary_ip
1874
        elif field == "tags":
1875
          val = list(node.GetTags())
1876
        elif field == "serial_no":
1877
          val = node.serial_no
1878
        elif field == "master_candidate":
1879
          val = node.master_candidate
1880
        elif field == "master":
1881
          val = node.name == master_node
1882
        elif field == "offline":
1883
          val = node.offline
1884
        elif self._FIELDS_DYNAMIC.Matches(field):
1885
          val = live_data[node.name].get(field, None)
1886
        else:
1887
          raise errors.ParameterError(field)
1888
        node_output.append(val)
1889
      output.append(node_output)
1890

    
1891
    return output
1892

    
1893

    
1894
class LUQueryNodeVolumes(NoHooksLU):
1895
  """Logical unit for getting volumes on node(s).
1896

1897
  """
1898
  _OP_REQP = ["nodes", "output_fields"]
1899
  REQ_BGL = False
1900
  _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
1901
  _FIELDS_STATIC = utils.FieldSet("node")
1902

    
1903
  def ExpandNames(self):
1904
    _CheckOutputFields(static=self._FIELDS_STATIC,
1905
                       dynamic=self._FIELDS_DYNAMIC,
1906
                       selected=self.op.output_fields)
1907

    
1908
    self.needed_locks = {}
1909
    self.share_locks[locking.LEVEL_NODE] = 1
1910
    if not self.op.nodes:
1911
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
1912
    else:
1913
      self.needed_locks[locking.LEVEL_NODE] = \
1914
        _GetWantedNodes(self, self.op.nodes)
1915

    
1916
  def CheckPrereq(self):
1917
    """Check prerequisites.
1918

1919
    This checks that the fields required are valid output fields.
1920

1921
    """
1922
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
1923

    
1924
  def Exec(self, feedback_fn):
1925
    """Computes the list of nodes and their attributes.
1926

1927
    """
1928
    nodenames = self.nodes
1929
    volumes = self.rpc.call_node_volumes(nodenames)
1930

    
1931
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
1932
             in self.cfg.GetInstanceList()]
1933

    
1934
    lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
1935

    
1936
    output = []
1937
    for node in nodenames:
1938
      if node not in volumes or volumes[node].failed or not volumes[node].data:
1939
        continue
1940

    
1941
      node_vols = volumes[node].data[:]
1942
      node_vols.sort(key=lambda vol: vol['dev'])
1943

    
1944
      for vol in node_vols:
1945
        node_output = []
1946
        for field in self.op.output_fields:
1947
          if field == "node":
1948
            val = node
1949
          elif field == "phys":
1950
            val = vol['dev']
1951
          elif field == "vg":
1952
            val = vol['vg']
1953
          elif field == "name":
1954
            val = vol['name']
1955
          elif field == "size":
1956
            val = int(float(vol['size']))
1957
          elif field == "instance":
1958
            for inst in ilist:
1959
              if node not in lv_by_node[inst]:
1960
                continue
1961
              if vol['name'] in lv_by_node[inst][node]:
1962
                val = inst.name
1963
                break
1964
            else:
1965
              val = '-'
1966
          else:
1967
            raise errors.ParameterError(field)
1968
          node_output.append(str(val))
1969

    
1970
        output.append(node_output)
1971

    
1972
    return output
1973

    
1974

    
1975
class LUAddNode(LogicalUnit):
1976
  """Logical unit for adding node to the cluster.
1977

1978
  """
1979
  HPATH = "node-add"
1980
  HTYPE = constants.HTYPE_NODE
1981
  _OP_REQP = ["node_name"]
1982

    
1983
  def BuildHooksEnv(self):
1984
    """Build hooks env.
1985

1986
    This will run on all nodes before, and on all nodes + the new node after.
1987

1988
    """
1989
    env = {
1990
      "OP_TARGET": self.op.node_name,
1991
      "NODE_NAME": self.op.node_name,
1992
      "NODE_PIP": self.op.primary_ip,
1993
      "NODE_SIP": self.op.secondary_ip,
1994
      }
1995
    nodes_0 = self.cfg.GetNodeList()
1996
    nodes_1 = nodes_0 + [self.op.node_name, ]
1997
    return env, nodes_0, nodes_1
1998

    
1999
  def CheckPrereq(self):
2000
    """Check prerequisites.
2001

2002
    This checks:
2003
     - the new node is not already in the config
2004
     - it is resolvable
2005
     - its parameters (single/dual homed) matches the cluster
2006

2007
    Any errors are signalled by raising errors.OpPrereqError.
2008

2009
    """
2010
    node_name = self.op.node_name
2011
    cfg = self.cfg
2012

    
2013
    dns_data = utils.HostInfo(node_name)
2014

    
2015
    node = dns_data.name
2016
    primary_ip = self.op.primary_ip = dns_data.ip
2017
    secondary_ip = getattr(self.op, "secondary_ip", None)
2018
    if secondary_ip is None:
2019
      secondary_ip = primary_ip
2020
    if not utils.IsValidIP(secondary_ip):
2021
      raise errors.OpPrereqError("Invalid secondary IP given")
2022
    self.op.secondary_ip = secondary_ip
2023

    
2024
    node_list = cfg.GetNodeList()
2025
    if not self.op.readd and node in node_list:
2026
      raise errors.OpPrereqError("Node %s is already in the configuration" %
2027
                                 node)
2028
    elif self.op.readd and node not in node_list:
2029
      raise errors.OpPrereqError("Node %s is not in the configuration" % node)
2030

    
2031
    for existing_node_name in node_list:
2032
      existing_node = cfg.GetNodeInfo(existing_node_name)
2033

    
2034
      if self.op.readd and node == existing_node_name:
2035
        if (existing_node.primary_ip != primary_ip or
2036
            existing_node.secondary_ip != secondary_ip):
2037
          raise errors.OpPrereqError("Readded node doesn't have the same IP"
2038
                                     " address configuration as before")
2039
        continue
2040

    
2041
      if (existing_node.primary_ip == primary_ip or
2042
          existing_node.secondary_ip == primary_ip or
2043
          existing_node.primary_ip == secondary_ip or
2044
          existing_node.secondary_ip == secondary_ip):
2045
        raise errors.OpPrereqError("New node ip address(es) conflict with"
2046
                                   " existing node %s" % existing_node.name)
2047

    
2048
    # check that the type of the node (single versus dual homed) is the
2049
    # same as for the master
2050
    myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
2051
    master_singlehomed = myself.secondary_ip == myself.primary_ip
2052
    newbie_singlehomed = secondary_ip == primary_ip
2053
    if master_singlehomed != newbie_singlehomed:
2054
      if master_singlehomed:
2055
        raise errors.OpPrereqError("The master has no private ip but the"
2056
                                   " new node has one")
2057
      else:
2058
        raise errors.OpPrereqError("The master has a private ip but the"
2059
                                   " new node doesn't have one")
2060

    
2061
    # checks reachablity
2062
    if not utils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
2063
      raise errors.OpPrereqError("Node not reachable by ping")
2064

    
2065
    if not newbie_singlehomed:
2066
      # check reachability from my secondary ip to newbie's secondary ip
2067
      if not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
2068
                           source=myself.secondary_ip):
2069
        raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
2070
                                   " based ping to noded port")
2071

    
2072
    cp_size = self.cfg.GetClusterInfo().candidate_pool_size
2073
    mc_now, _ = self.cfg.GetMasterCandidateStats()
2074
    master_candidate = mc_now < cp_size
2075

    
2076
    self.new_node = objects.Node(name=node,
2077
                                 primary_ip=primary_ip,
2078
                                 secondary_ip=secondary_ip,
2079
                                 master_candidate=master_candidate,
2080
                                 offline=False)
2081

    
2082
  def Exec(self, feedback_fn):
2083
    """Adds the new node to the cluster.
2084

2085
    """
2086
    new_node = self.new_node
2087
    node = new_node.name
2088

    
2089
    # check connectivity
2090
    result = self.rpc.call_version([node])[node]
2091
    result.Raise()
2092
    if result.data:
2093
      if constants.PROTOCOL_VERSION == result.data:
2094
        logging.info("Communication to node %s fine, sw version %s match",
2095
                     node, result.data)
2096
      else:
2097
        raise errors.OpExecError("Version mismatch master version %s,"
2098
                                 " node version %s" %
2099
                                 (constants.PROTOCOL_VERSION, result.data))
2100
    else:
2101
      raise errors.OpExecError("Cannot get version from the new node")
2102

    
2103
    # setup ssh on node
2104
    logging.info("Copy ssh key to node %s", node)
2105
    priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
2106
    keyarray = []
2107
    keyfiles = [constants.SSH_HOST_DSA_PRIV, constants.SSH_HOST_DSA_PUB,
2108
                constants.SSH_HOST_RSA_PRIV, constants.SSH_HOST_RSA_PUB,
2109
                priv_key, pub_key]
2110

    
2111
    for i in keyfiles:
2112
      f = open(i, 'r')
2113
      try:
2114
        keyarray.append(f.read())
2115
      finally:
2116
        f.close()
2117

    
2118
    result = self.rpc.call_node_add(node, keyarray[0], keyarray[1],
2119
                                    keyarray[2],
2120
                                    keyarray[3], keyarray[4], keyarray[5])
2121

    
2122
    if result.failed or not result.data:
2123
      raise errors.OpExecError("Cannot transfer ssh keys to the new node")
2124

    
2125
    # Add node to our /etc/hosts, and add key to known_hosts
2126
    utils.AddHostToEtcHosts(new_node.name)
2127

    
2128
    if new_node.secondary_ip != new_node.primary_ip:
2129
      result = self.rpc.call_node_has_ip_address(new_node.name,
2130
                                                 new_node.secondary_ip)
2131
      if result.failed or not result.data:
2132
        raise errors.OpExecError("Node claims it doesn't have the secondary ip"
2133
                                 " you gave (%s). Please fix and re-run this"
2134
                                 " command." % new_node.secondary_ip)
2135

    
2136
    node_verify_list = [self.cfg.GetMasterNode()]
2137
    node_verify_param = {
2138
      'nodelist': [node],
2139
      # TODO: do a node-net-test as well?
2140
    }
2141

    
2142
    result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
2143
                                       self.cfg.GetClusterName())
2144
    for verifier in node_verify_list:
2145
      if result[verifier].failed or not result[verifier].data:
2146
        raise errors.OpExecError("Cannot communicate with %s's node daemon"
2147
                                 " for remote verification" % verifier)
2148
      if result[verifier].data['nodelist']:
2149
        for failed in result[verifier].data['nodelist']:
2150
          feedback_fn("ssh/hostname verification failed %s -> %s" %
2151
                      (verifier, result[verifier]['nodelist'][failed]))
2152
        raise errors.OpExecError("ssh/hostname verification failed.")
2153

    
2154
    # Distribute updated /etc/hosts and known_hosts to all nodes,
2155
    # including the node just added
2156
    myself = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
2157
    dist_nodes = self.cfg.GetNodeList()
2158
    if not self.op.readd:
2159
      dist_nodes.append(node)
2160
    if myself.name in dist_nodes:
2161
      dist_nodes.remove(myself.name)
2162

    
2163
    logging.debug("Copying hosts and known_hosts to all nodes")
2164
    for fname in (constants.ETC_HOSTS, constants.SSH_KNOWN_HOSTS_FILE):
2165
      result = self.rpc.call_upload_file(dist_nodes, fname)
2166
      for to_node, to_result in result.iteritems():
2167
        if to_result.failed or not to_result.data:
2168
          logging.error("Copy of file %s to node %s failed", fname, to_node)
2169

    
2170
    to_copy = []
2171
    if constants.HT_XEN_HVM in self.cfg.GetClusterInfo().enabled_hypervisors:
2172
      to_copy.append(constants.VNC_PASSWORD_FILE)
2173
    for fname in to_copy:
2174
      result = self.rpc.call_upload_file([node], fname)
2175
      if result[node].failed or not result[node]:
2176
        logging.error("Could not copy file %s to node %s", fname, node)
2177

    
2178
    if self.op.readd:
2179
      self.context.ReaddNode(new_node)
2180
    else:
2181
      self.context.AddNode(new_node)
2182

    
2183

    
2184
class LUSetNodeParams(LogicalUnit):
2185
  """Modifies the parameters of a node.
2186

2187
  """
2188
  HPATH = "node-modify"
2189
  HTYPE = constants.HTYPE_NODE
2190
  _OP_REQP = ["node_name"]
2191
  REQ_BGL = False
2192

    
2193
  def CheckArguments(self):
2194
    node_name = self.cfg.ExpandNodeName(self.op.node_name)
2195
    if node_name is None:
2196
      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name)
2197
    self.op.node_name = node_name
2198
    _CheckBooleanOpField(self.op, 'master_candidate')
2199
    _CheckBooleanOpField(self.op, 'offline')
2200
    if self.op.master_candidate is None and self.op.offline is None:
2201
      raise errors.OpPrereqError("Please pass at least one modification")
2202
    if self.op.offline == True and self.op.master_candidate == True:
2203
      raise errors.OpPrereqError("Can't set the node into offline and"
2204
                                 " master_candidate at the same time")
2205

    
2206
  def ExpandNames(self):
2207
    self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
2208

    
2209
  def BuildHooksEnv(self):
2210
    """Build hooks env.
2211

2212
    This runs on the master node.
2213

2214
    """
2215
    env = {
2216
      "OP_TARGET": self.op.node_name,
2217
      "MASTER_CANDIDATE": str(self.op.master_candidate),
2218
      "OFFLINE": str(self.op.offline),
2219
      }
2220
    nl = [self.cfg.GetMasterNode(),
2221
          self.op.node_name]
2222
    return env, nl, nl
2223

    
2224
  def CheckPrereq(self):
2225
    """Check prerequisites.
2226

2227
    This only checks the instance list against the existing names.
2228

2229
    """
2230
    node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
2231

    
2232
    if ((self.op.master_candidate == False or self.op.offline == True)
2233
        and node.master_candidate):
2234
      # we will demote the node from master_candidate
2235
      if self.op.node_name == self.cfg.GetMasterNode():
2236
        raise errors.OpPrereqError("The master node has to be a"
2237
                                   " master candidate and online")
2238
      cp_size = self.cfg.GetClusterInfo().candidate_pool_size
2239
      num_candidates, _ = self.cfg.GetMasterCandidateStats()
2240
      if num_candidates <= cp_size:
2241
        msg = ("Not enough master candidates (desired"
2242
               " %d, new value will be %d)" % (cp_size, num_candidates-1))
2243
        if self.op.force:
2244
          self.LogWarning(msg)
2245
        else:
2246
          raise errors.OpPrereqError(msg)
2247

    
2248
    if (self.op.master_candidate == True and node.offline and
2249
        not self.op.offline == False):
2250
      raise errors.OpPrereqError("Can't set an offline node to"
2251
                                 " master_candidate")
2252

    
2253
    return
2254

    
2255
  def Exec(self, feedback_fn):
2256
    """Modifies a node.
2257

2258
    """
2259
    node = self.node
2260

    
2261
    result = []
2262

    
2263
    if self.op.offline is not None:
2264
      node.offline = self.op.offline
2265
      result.append(("offline", str(self.op.offline)))
2266
      if self.op.offline == True and node.master_candidate:
2267
        node.master_candidate = False
2268
        result.append(("master_candidate", "auto-demotion due to offline"))
2269

    
2270
    if self.op.master_candidate is not None:
2271
      node.master_candidate = self.op.master_candidate
2272
      result.append(("master_candidate", str(self.op.master_candidate)))
2273
      if self.op.master_candidate == False:
2274
        rrc = self.rpc.call_node_demote_from_mc(node.name)
2275
        if (rrc.failed or not isinstance(rrc.data, (tuple, list))
2276
            or len(rrc.data) != 2):
2277
          self.LogWarning("Node rpc error: %s" % rrc.error)
2278
        elif not rrc.data[0]:
2279
          self.LogWarning("Node failed to demote itself: %s" % rrc.data[1])
2280

    
2281
    # this will trigger configuration file update, if needed
2282
    self.cfg.Update(node)
2283
    # this will trigger job queue propagation or cleanup
2284
    if self.op.node_name != self.cfg.GetMasterNode():
2285
      self.context.ReaddNode(node)
2286

    
2287
    return result
2288

    
2289

    
2290
class LUQueryClusterInfo(NoHooksLU):
2291
  """Query cluster configuration.
2292

2293
  """
2294
  _OP_REQP = []
2295
  REQ_BGL = False
2296

    
2297
  def ExpandNames(self):
2298
    self.needed_locks = {}
2299

    
2300
  def CheckPrereq(self):
2301
    """No prerequsites needed for this LU.
2302

2303
    """
2304
    pass
2305

    
2306
  def Exec(self, feedback_fn):
2307
    """Return cluster config.
2308

2309
    """
2310
    cluster = self.cfg.GetClusterInfo()
2311
    result = {
2312
      "software_version": constants.RELEASE_VERSION,
2313
      "protocol_version": constants.PROTOCOL_VERSION,
2314
      "config_version": constants.CONFIG_VERSION,
2315
      "os_api_version": constants.OS_API_VERSION,
2316
      "export_version": constants.EXPORT_VERSION,
2317
      "architecture": (platform.architecture()[0], platform.machine()),
2318
      "name": cluster.cluster_name,
2319
      "master": cluster.master_node,
2320
      "default_hypervisor": cluster.default_hypervisor,
2321
      "enabled_hypervisors": cluster.enabled_hypervisors,
2322
      "hvparams": cluster.hvparams,
2323
      "beparams": cluster.beparams,
2324
      "candidate_pool_size": cluster.candidate_pool_size,
2325
      }
2326

    
2327
    return result
2328

    
2329

    
2330
class LUQueryConfigValues(NoHooksLU):
2331
  """Return configuration values.
2332

2333
  """
2334
  _OP_REQP = []
2335
  REQ_BGL = False
2336
  _FIELDS_DYNAMIC = utils.FieldSet()
2337
  _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag")
2338

    
2339
  def ExpandNames(self):
2340
    self.needed_locks = {}
2341

    
2342
    _CheckOutputFields(static=self._FIELDS_STATIC,
2343
                       dynamic=self._FIELDS_DYNAMIC,
2344
                       selected=self.op.output_fields)
2345

    
2346
  def CheckPrereq(self):
2347
    """No prerequisites.
2348

2349
    """
2350
    pass
2351

    
2352
  def Exec(self, feedback_fn):
2353
    """Dump a representation of the cluster config to the standard output.
2354

2355
    """
2356
    values = []
2357
    for field in self.op.output_fields:
2358
      if field == "cluster_name":
2359
        entry = self.cfg.GetClusterName()
2360
      elif field == "master_node":
2361
        entry = self.cfg.GetMasterNode()
2362
      elif field == "drain_flag":
2363
        entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
2364
      else:
2365
        raise errors.ParameterError(field)
2366
      values.append(entry)
2367
    return values
2368

    
2369

    
2370
class LUActivateInstanceDisks(NoHooksLU):
2371
  """Bring up an instance's disks.
2372

2373
  """
2374
  _OP_REQP = ["instance_name"]
2375
  REQ_BGL = False
2376

    
2377
  def ExpandNames(self):
2378
    self._ExpandAndLockInstance()
2379
    self.needed_locks[locking.LEVEL_NODE] = []
2380
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2381

    
2382
  def DeclareLocks(self, level):
2383
    if level == locking.LEVEL_NODE:
2384
      self._LockInstancesNodes()
2385

    
2386
  def CheckPrereq(self):
2387
    """Check prerequisites.
2388

2389
    This checks that the instance is in the cluster.
2390

2391
    """
2392
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2393
    assert self.instance is not None, \
2394
      "Cannot retrieve locked instance %s" % self.op.instance_name
2395
    _CheckNodeOnline(self, self.instance.primary_node)
2396

    
2397
  def Exec(self, feedback_fn):
2398
    """Activate the disks.
2399

2400
    """
2401
    disks_ok, disks_info = _AssembleInstanceDisks(self, self.instance)
2402
    if not disks_ok:
2403
      raise errors.OpExecError("Cannot activate block devices")
2404

    
2405
    return disks_info
2406

    
2407

    
2408
def _AssembleInstanceDisks(lu, instance, ignore_secondaries=False):
2409
  """Prepare the block devices for an instance.
2410

2411
  This sets up the block devices on all nodes.
2412

2413
  @type lu: L{LogicalUnit}
2414
  @param lu: the logical unit on whose behalf we execute
2415
  @type instance: L{objects.Instance}
2416
  @param instance: the instance for whose disks we assemble
2417
  @type ignore_secondaries: boolean
2418
  @param ignore_secondaries: if true, errors on secondary nodes
2419
      won't result in an error return from the function
2420
  @return: False if the operation failed, otherwise a list of
2421
      (host, instance_visible_name, node_visible_name)
2422
      with the mapping from node devices to instance devices
2423

2424
  """
2425
  device_info = []
2426
  disks_ok = True
2427
  iname = instance.name
2428
  # With the two passes mechanism we try to reduce the window of
2429
  # opportunity for the race condition of switching DRBD to primary
2430
  # before handshaking occured, but we do not eliminate it
2431

    
2432
  # The proper fix would be to wait (with some limits) until the
2433
  # connection has been made and drbd transitions from WFConnection
2434
  # into any other network-connected state (Connected, SyncTarget,
2435
  # SyncSource, etc.)
2436

    
2437
  # 1st pass, assemble on all nodes in secondary mode
2438
  for inst_disk in instance.disks:
2439
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
2440
      lu.cfg.SetDiskID(node_disk, node)
2441
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
2442
      if result.failed or not result:
2443
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
2444
                           " (is_primary=False, pass=1)",
2445
                           inst_disk.iv_name, node)
2446
        if not ignore_secondaries:
2447
          disks_ok = False
2448

    
2449
  # FIXME: race condition on drbd migration to primary
2450

    
2451
  # 2nd pass, do only the primary node
2452
  for inst_disk in instance.disks:
2453
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
2454
      if node != instance.primary_node:
2455
        continue
2456
      lu.cfg.SetDiskID(node_disk, node)
2457
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
2458
      if result.failed or not result:
2459
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
2460
                           " (is_primary=True, pass=2)",
2461
                           inst_disk.iv_name, node)
2462
        disks_ok = False
2463
    device_info.append((instance.primary_node, inst_disk.iv_name, result.data))
2464

    
2465
  # leave the disks configured for the primary node
2466
  # this is a workaround that would be fixed better by
2467
  # improving the logical/physical id handling
2468
  for disk in instance.disks:
2469
    lu.cfg.SetDiskID(disk, instance.primary_node)
2470

    
2471
  return disks_ok, device_info
2472

    
2473

    
2474
def _StartInstanceDisks(lu, instance, force):
2475
  """Start the disks of an instance.
2476

2477
  """
2478
  disks_ok, dummy = _AssembleInstanceDisks(lu, instance,
2479
                                           ignore_secondaries=force)
2480
  if not disks_ok:
2481
    _ShutdownInstanceDisks(lu, instance)
2482
    if force is not None and not force:
2483
      lu.proc.LogWarning("", hint="If the message above refers to a"
2484
                         " secondary node,"
2485
                         " you can retry the operation using '--force'.")
2486
    raise errors.OpExecError("Disk consistency error")
2487

    
2488

    
2489
class LUDeactivateInstanceDisks(NoHooksLU):
2490
  """Shutdown an instance's disks.
2491

2492
  """
2493
  _OP_REQP = ["instance_name"]
2494
  REQ_BGL = False
2495

    
2496
  def ExpandNames(self):
2497
    self._ExpandAndLockInstance()
2498
    self.needed_locks[locking.LEVEL_NODE] = []
2499
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2500

    
2501
  def DeclareLocks(self, level):
2502
    if level == locking.LEVEL_NODE:
2503
      self._LockInstancesNodes()
2504

    
2505
  def CheckPrereq(self):
2506
    """Check prerequisites.
2507

2508
    This checks that the instance is in the cluster.
2509

2510
    """
2511
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2512
    assert self.instance is not None, \
2513
      "Cannot retrieve locked instance %s" % self.op.instance_name
2514

    
2515
  def Exec(self, feedback_fn):
2516
    """Deactivate the disks
2517

2518
    """
2519
    instance = self.instance
2520
    _SafeShutdownInstanceDisks(self, instance)
2521

    
2522

    
2523
def _SafeShutdownInstanceDisks(lu, instance):
2524
  """Shutdown block devices of an instance.
2525

2526
  This function checks if an instance is running, before calling
2527
  _ShutdownInstanceDisks.
2528

2529
  """
2530
  ins_l = lu.rpc.call_instance_list([instance.primary_node],
2531
                                      [instance.hypervisor])
2532
  ins_l = ins_l[instance.primary_node]
2533
  if ins_l.failed or not isinstance(ins_l.data, list):
2534
    raise errors.OpExecError("Can't contact node '%s'" %
2535
                             instance.primary_node)
2536

    
2537
  if instance.name in ins_l.data:
2538
    raise errors.OpExecError("Instance is running, can't shutdown"
2539
                             " block devices.")
2540

    
2541
  _ShutdownInstanceDisks(lu, instance)
2542

    
2543

    
2544
def _ShutdownInstanceDisks(lu, instance, ignore_primary=False):
2545
  """Shutdown block devices of an instance.
2546

2547
  This does the shutdown on all nodes of the instance.
2548

2549
  If the ignore_primary is false, errors on the primary node are
2550
  ignored.
2551

2552
  """
2553
  result = True
2554
  for disk in instance.disks:
2555
    for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
2556
      lu.cfg.SetDiskID(top_disk, node)
2557
      result = lu.rpc.call_blockdev_shutdown(node, top_disk)
2558
      if result.failed or not result.data:
2559
        logging.error("Could not shutdown block device %s on node %s",
2560
                      disk.iv_name, node)
2561
        if not ignore_primary or node != instance.primary_node:
2562
          result = False
2563
  return result
2564

    
2565

    
2566
def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
2567
  """Checks if a node has enough free memory.
2568

2569
  This function check if a given node has the needed amount of free
2570
  memory. In case the node has less memory or we cannot get the
2571
  information from the node, this function raise an OpPrereqError
2572
  exception.
2573

2574
  @type lu: C{LogicalUnit}
2575
  @param lu: a logical unit from which we get configuration data
2576
  @type node: C{str}
2577
  @param node: the node to check
2578
  @type reason: C{str}
2579
  @param reason: string to use in the error message
2580
  @type requested: C{int}
2581
  @param requested: the amount of memory in MiB to check for
2582
  @type hypervisor_name: C{str}
2583
  @param hypervisor_name: the hypervisor to ask for memory stats
2584
  @raise errors.OpPrereqError: if the node doesn't have enough memory, or
2585
      we cannot check the node
2586

2587
  """
2588
  nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
2589
  nodeinfo[node].Raise()
2590
  free_mem = nodeinfo[node].data.get('memory_free')
2591
  if not isinstance(free_mem, int):
2592
    raise errors.OpPrereqError("Can't compute free memory on node %s, result"
2593
                             " was '%s'" % (node, free_mem))
2594
  if requested > free_mem:
2595
    raise errors.OpPrereqError("Not enough memory on node %s for %s:"
2596
                             " needed %s MiB, available %s MiB" %
2597
                             (node, reason, requested, free_mem))
2598

    
2599

    
2600
class LUStartupInstance(LogicalUnit):
2601
  """Starts an instance.
2602

2603
  """
2604
  HPATH = "instance-start"
2605
  HTYPE = constants.HTYPE_INSTANCE
2606
  _OP_REQP = ["instance_name", "force"]
2607
  REQ_BGL = False
2608

    
2609
  def ExpandNames(self):
2610
    self._ExpandAndLockInstance()
2611

    
2612
  def BuildHooksEnv(self):
2613
    """Build hooks env.
2614

2615
    This runs on master, primary and secondary nodes of the instance.
2616

2617
    """
2618
    env = {
2619
      "FORCE": self.op.force,
2620
      }
2621
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2622
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2623
    return env, nl, nl
2624

    
2625
  def CheckPrereq(self):
2626
    """Check prerequisites.
2627

2628
    This checks that the instance is in the cluster.
2629

2630
    """
2631
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2632
    assert self.instance is not None, \
2633
      "Cannot retrieve locked instance %s" % self.op.instance_name
2634

    
2635
    _CheckNodeOnline(self, instance.primary_node)
2636

    
2637
    bep = self.cfg.GetClusterInfo().FillBE(instance)
2638
    # check bridges existance
2639
    _CheckInstanceBridgesExist(self, instance)
2640

    
2641
    _CheckNodeFreeMemory(self, instance.primary_node,
2642
                         "starting instance %s" % instance.name,
2643
                         bep[constants.BE_MEMORY], instance.hypervisor)
2644

    
2645
  def Exec(self, feedback_fn):
2646
    """Start the instance.
2647

2648
    """
2649
    instance = self.instance
2650
    force = self.op.force
2651
    extra_args = getattr(self.op, "extra_args", "")
2652

    
2653
    self.cfg.MarkInstanceUp(instance.name)
2654

    
2655
    node_current = instance.primary_node
2656

    
2657
    _StartInstanceDisks(self, instance, force)
2658

    
2659
    result = self.rpc.call_instance_start(node_current, instance, extra_args)
2660
    msg = result.RemoteFailMsg()
2661
    if msg:
2662
      _ShutdownInstanceDisks(self, instance)
2663
      raise errors.OpExecError("Could not start instance: %s" % msg)
2664

    
2665

    
2666
class LURebootInstance(LogicalUnit):
2667
  """Reboot an instance.
2668

2669
  """
2670
  HPATH = "instance-reboot"
2671
  HTYPE = constants.HTYPE_INSTANCE
2672
  _OP_REQP = ["instance_name", "ignore_secondaries", "reboot_type"]
2673
  REQ_BGL = False
2674

    
2675
  def ExpandNames(self):
2676
    if self.op.reboot_type not in [constants.INSTANCE_REBOOT_SOFT,
2677
                                   constants.INSTANCE_REBOOT_HARD,
2678
                                   constants.INSTANCE_REBOOT_FULL]:
2679
      raise errors.ParameterError("reboot type not in [%s, %s, %s]" %
2680
                                  (constants.INSTANCE_REBOOT_SOFT,
2681
                                   constants.INSTANCE_REBOOT_HARD,
2682
                                   constants.INSTANCE_REBOOT_FULL))
2683
    self._ExpandAndLockInstance()
2684

    
2685
  def BuildHooksEnv(self):
2686
    """Build hooks env.
2687

2688
    This runs on master, primary and secondary nodes of the instance.
2689

2690
    """
2691
    env = {
2692
      "IGNORE_SECONDARIES": self.op.ignore_secondaries,
2693
      }
2694
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2695
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2696
    return env, nl, nl
2697

    
2698
  def CheckPrereq(self):
2699
    """Check prerequisites.
2700

2701
    This checks that the instance is in the cluster.
2702

2703
    """
2704
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2705
    assert self.instance is not None, \
2706
      "Cannot retrieve locked instance %s" % self.op.instance_name
2707

    
2708
    _CheckNodeOnline(self, instance.primary_node)
2709

    
2710
    # check bridges existance
2711
    _CheckInstanceBridgesExist(self, instance)
2712

    
2713
  def Exec(self, feedback_fn):
2714
    """Reboot the instance.
2715

2716
    """
2717
    instance = self.instance
2718
    ignore_secondaries = self.op.ignore_secondaries
2719
    reboot_type = self.op.reboot_type
2720
    extra_args = getattr(self.op, "extra_args", "")
2721

    
2722
    node_current = instance.primary_node
2723

    
2724
    if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
2725
                       constants.INSTANCE_REBOOT_HARD]:
2726
      result = self.rpc.call_instance_reboot(node_current, instance,
2727
                                             reboot_type, extra_args)
2728
      if result.failed or not result.data:
2729
        raise errors.OpExecError("Could not reboot instance")
2730
    else:
2731
      if not self.rpc.call_instance_shutdown(node_current, instance):
2732
        raise errors.OpExecError("could not shutdown instance for full reboot")
2733
      _ShutdownInstanceDisks(self, instance)
2734
      _StartInstanceDisks(self, instance, ignore_secondaries)
2735
      result = self.rpc.call_instance_start(node_current, instance, extra_args)
2736
      msg = result.RemoteFailMsg()
2737
      if msg:
2738
        _ShutdownInstanceDisks(self, instance)
2739
        raise errors.OpExecError("Could not start instance for"
2740
                                 " full reboot: %s" % msg)
2741

    
2742
    self.cfg.MarkInstanceUp(instance.name)
2743

    
2744

    
2745
class LUShutdownInstance(LogicalUnit):
2746
  """Shutdown an instance.
2747

2748
  """
2749
  HPATH = "instance-stop"
2750
  HTYPE = constants.HTYPE_INSTANCE
2751
  _OP_REQP = ["instance_name"]
2752
  REQ_BGL = False
2753

    
2754
  def ExpandNames(self):
2755
    self._ExpandAndLockInstance()
2756

    
2757
  def BuildHooksEnv(self):
2758
    """Build hooks env.
2759

2760
    This runs on master, primary and secondary nodes of the instance.
2761

2762
    """
2763
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2764
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2765
    return env, nl, nl
2766

    
2767
  def CheckPrereq(self):
2768
    """Check prerequisites.
2769

2770
    This checks that the instance is in the cluster.
2771

2772
    """
2773
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2774
    assert self.instance is not None, \
2775
      "Cannot retrieve locked instance %s" % self.op.instance_name
2776
    _CheckNodeOnline(self, self.instance.primary_node)
2777

    
2778
  def Exec(self, feedback_fn):
2779
    """Shutdown the instance.
2780

2781
    """
2782
    instance = self.instance
2783
    node_current = instance.primary_node
2784
    self.cfg.MarkInstanceDown(instance.name)
2785
    result = self.rpc.call_instance_shutdown(node_current, instance)
2786
    if result.failed or not result.data:
2787
      self.proc.LogWarning("Could not shutdown instance")
2788

    
2789
    _ShutdownInstanceDisks(self, instance)
2790

    
2791

    
2792
class LUReinstallInstance(LogicalUnit):
2793
  """Reinstall an instance.
2794

2795
  """
2796
  HPATH = "instance-reinstall"
2797
  HTYPE = constants.HTYPE_INSTANCE
2798
  _OP_REQP = ["instance_name"]
2799
  REQ_BGL = False
2800

    
2801
  def ExpandNames(self):
2802
    self._ExpandAndLockInstance()
2803

    
2804
  def BuildHooksEnv(self):
2805
    """Build hooks env.
2806

2807
    This runs on master, primary and secondary nodes of the instance.
2808

2809
    """
2810
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2811
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2812
    return env, nl, nl
2813

    
2814
  def CheckPrereq(self):
2815
    """Check prerequisites.
2816

2817
    This checks that the instance is in the cluster and is not running.
2818

2819
    """
2820
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2821
    assert instance is not None, \
2822
      "Cannot retrieve locked instance %s" % self.op.instance_name
2823
    _CheckNodeOnline(self, instance.primary_node)
2824

    
2825
    if instance.disk_template == constants.DT_DISKLESS:
2826
      raise errors.OpPrereqError("Instance '%s' has no disks" %
2827
                                 self.op.instance_name)
2828
    if instance.admin_up:
2829
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
2830
                                 self.op.instance_name)
2831
    remote_info = self.rpc.call_instance_info(instance.primary_node,
2832
                                              instance.name,
2833
                                              instance.hypervisor)
2834
    if remote_info.failed or remote_info.data:
2835
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
2836
                                 (self.op.instance_name,
2837
                                  instance.primary_node))
2838

    
2839
    self.op.os_type = getattr(self.op, "os_type", None)
2840
    if self.op.os_type is not None:
2841
      # OS verification
2842
      pnode = self.cfg.GetNodeInfo(
2843
        self.cfg.ExpandNodeName(instance.primary_node))
2844
      if pnode is None:
2845
        raise errors.OpPrereqError("Primary node '%s' is unknown" %
2846
                                   self.op.pnode)
2847
      result = self.rpc.call_os_get(pnode.name, self.op.os_type)
2848
      result.Raise()
2849
      if not isinstance(result.data, objects.OS):
2850
        raise errors.OpPrereqError("OS '%s' not in supported OS list for"
2851
                                   " primary node"  % self.op.os_type)
2852

    
2853
    self.instance = instance
2854

    
2855
  def Exec(self, feedback_fn):
2856
    """Reinstall the instance.
2857

2858
    """
2859
    inst = self.instance
2860

    
2861
    if self.op.os_type is not None:
2862
      feedback_fn("Changing OS to '%s'..." % self.op.os_type)
2863
      inst.os = self.op.os_type
2864
      self.cfg.Update(inst)
2865

    
2866
    _StartInstanceDisks(self, inst, None)
2867
    try:
2868
      feedback_fn("Running the instance OS create scripts...")
2869
      result = self.rpc.call_instance_os_add(inst.primary_node, inst)
2870
      msg = result.RemoteFailMsg()
2871
      if msg:
2872
        raise errors.OpExecError("Could not install OS for instance %s"
2873
                                 " on node %s: %s" %
2874
                                 (inst.name, inst.primary_node, msg))
2875
    finally:
2876
      _ShutdownInstanceDisks(self, inst)
2877

    
2878

    
2879
class LURenameInstance(LogicalUnit):
2880
  """Rename an instance.
2881

2882
  """
2883
  HPATH = "instance-rename"
2884
  HTYPE = constants.HTYPE_INSTANCE
2885
  _OP_REQP = ["instance_name", "new_name"]
2886

    
2887
  def BuildHooksEnv(self):
2888
    """Build hooks env.
2889

2890
    This runs on master, primary and secondary nodes of the instance.
2891

2892
    """
2893
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2894
    env["INSTANCE_NEW_NAME"] = self.op.new_name
2895
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2896
    return env, nl, nl
2897

    
2898
  def CheckPrereq(self):
2899
    """Check prerequisites.
2900

2901
    This checks that the instance is in the cluster and is not running.
2902

2903
    """
2904
    instance = self.cfg.GetInstanceInfo(
2905
      self.cfg.ExpandInstanceName(self.op.instance_name))
2906
    if instance is None:
2907
      raise errors.OpPrereqError("Instance '%s' not known" %
2908
                                 self.op.instance_name)
2909
    _CheckNodeOnline(self, instance.primary_node)
2910

    
2911
    if instance.admin_up:
2912
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
2913
                                 self.op.instance_name)
2914
    remote_info = self.rpc.call_instance_info(instance.primary_node,
2915
                                              instance.name,
2916
                                              instance.hypervisor)
2917
    remote_info.Raise()
2918
    if remote_info.data:
2919
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
2920
                                 (self.op.instance_name,
2921
                                  instance.primary_node))
2922
    self.instance = instance
2923

    
2924
    # new name verification
2925
    name_info = utils.HostInfo(self.op.new_name)
2926

    
2927
    self.op.new_name = new_name = name_info.name
2928
    instance_list = self.cfg.GetInstanceList()
2929
    if new_name in instance_list:
2930
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
2931
                                 new_name)
2932

    
2933
    if not getattr(self.op, "ignore_ip", False):
2934
      if utils.TcpPing(name_info.ip, constants.DEFAULT_NODED_PORT):
2935
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
2936
                                   (name_info.ip, new_name))
2937

    
2938

    
2939
  def Exec(self, feedback_fn):
2940
    """Reinstall the instance.
2941

2942
    """
2943
    inst = self.instance
2944
    old_name = inst.name
2945

    
2946
    if inst.disk_template == constants.DT_FILE:
2947
      old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
2948

    
2949
    self.cfg.RenameInstance(inst.name, self.op.new_name)
2950
    # Change the instance lock. This is definitely safe while we hold the BGL
2951
    self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
2952
    self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
2953

    
2954
    # re-read the instance from the configuration after rename
2955
    inst = self.cfg.GetInstanceInfo(self.op.new_name)
2956

    
2957
    if inst.disk_template == constants.DT_FILE:
2958
      new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
2959
      result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
2960
                                                     old_file_storage_dir,
2961
                                                     new_file_storage_dir)
2962
      result.Raise()
2963
      if not result.data:
2964
        raise errors.OpExecError("Could not connect to node '%s' to rename"
2965
                                 " directory '%s' to '%s' (but the instance"
2966
                                 " has been renamed in Ganeti)" % (
2967
                                 inst.primary_node, old_file_storage_dir,
2968
                                 new_file_storage_dir))
2969

    
2970
      if not result.data[0]:
2971
        raise errors.OpExecError("Could not rename directory '%s' to '%s'"
2972
                                 " (but the instance has been renamed in"
2973
                                 " Ganeti)" % (old_file_storage_dir,
2974
                                               new_file_storage_dir))
2975

    
2976
    _StartInstanceDisks(self, inst, None)
2977
    try:
2978
      result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
2979
                                                 old_name)
2980
      msg = result.RemoteFailMsg()
2981
      if msg:
2982
        msg = ("Could not run OS rename script for instance %s on node %s"
2983
               " (but the instance has been renamed in Ganeti): %s" %
2984
               (inst.name, inst.primary_node, msg))
2985
        self.proc.LogWarning(msg)
2986
    finally:
2987
      _ShutdownInstanceDisks(self, inst)
2988

    
2989

    
2990
class LURemoveInstance(LogicalUnit):
2991
  """Remove an instance.
2992

2993
  """
2994
  HPATH = "instance-remove"
2995
  HTYPE = constants.HTYPE_INSTANCE
2996
  _OP_REQP = ["instance_name", "ignore_failures"]
2997
  REQ_BGL = False
2998

    
2999
  def ExpandNames(self):
3000
    self._ExpandAndLockInstance()
3001
    self.needed_locks[locking.LEVEL_NODE] = []
3002
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3003

    
3004
  def DeclareLocks(self, level):
3005
    if level == locking.LEVEL_NODE:
3006
      self._LockInstancesNodes()
3007

    
3008
  def BuildHooksEnv(self):
3009
    """Build hooks env.
3010

3011
    This runs on master, primary and secondary nodes of the instance.
3012

3013
    """
3014
    env = _BuildInstanceHookEnvByObject(self, self.instance)
3015
    nl = [self.cfg.GetMasterNode()]
3016
    return env, nl, nl
3017

    
3018
  def CheckPrereq(self):
3019
    """Check prerequisites.
3020

3021
    This checks that the instance is in the cluster.
3022

3023
    """
3024
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3025
    assert self.instance is not None, \
3026
      "Cannot retrieve locked instance %s" % self.op.instance_name
3027

    
3028
  def Exec(self, feedback_fn):
3029
    """Remove the instance.
3030

3031
    """
3032
    instance = self.instance
3033
    logging.info("Shutting down instance %s on node %s",
3034
                 instance.name, instance.primary_node)
3035

    
3036
    result = self.rpc.call_instance_shutdown(instance.primary_node, instance)
3037
    if result.failed or not result.data:
3038
      if self.op.ignore_failures:
3039
        feedback_fn("Warning: can't shutdown instance")
3040
      else:
3041
        raise errors.OpExecError("Could not shutdown instance %s on node %s" %
3042
                                 (instance.name, instance.primary_node))
3043

    
3044
    logging.info("Removing block devices for instance %s", instance.name)
3045

    
3046
    if not _RemoveDisks(self, instance):
3047
      if self.op.ignore_failures:
3048
        feedback_fn("Warning: can't remove instance's disks")
3049
      else:
3050
        raise errors.OpExecError("Can't remove instance's disks")
3051

    
3052
    logging.info("Removing instance %s out of cluster config", instance.name)
3053

    
3054
    self.cfg.RemoveInstance(instance.name)
3055
    self.remove_locks[locking.LEVEL_INSTANCE] = instance.name
3056

    
3057

    
3058
class LUQueryInstances(NoHooksLU):
3059
  """Logical unit for querying instances.
3060

3061
  """
3062
  _OP_REQP = ["output_fields", "names"]
3063
  REQ_BGL = False
3064
  _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
3065
                                    "admin_state", "admin_ram",
3066
                                    "disk_template", "ip", "mac", "bridge",
3067
                                    "sda_size", "sdb_size", "vcpus", "tags",
3068
                                    "network_port", "beparams",
3069
                                    "(disk).(size)/([0-9]+)",
3070
                                    "(disk).(sizes)",
3071
                                    "(nic).(mac|ip|bridge)/([0-9]+)",
3072
                                    "(nic).(macs|ips|bridges)",
3073
                                    "(disk|nic).(count)",
3074
                                    "serial_no", "hypervisor", "hvparams",] +
3075
                                  ["hv/%s" % name
3076
                                   for name in constants.HVS_PARAMETERS] +
3077
                                  ["be/%s" % name
3078
                                   for name in constants.BES_PARAMETERS])
3079
  _FIELDS_DYNAMIC = utils.FieldSet("oper_state", "oper_ram", "status")
3080

    
3081

    
3082
  def ExpandNames(self):
3083
    _CheckOutputFields(static=self._FIELDS_STATIC,
3084
                       dynamic=self._FIELDS_DYNAMIC,
3085
                       selected=self.op.output_fields)
3086

    
3087
    self.needed_locks = {}
3088
    self.share_locks[locking.LEVEL_INSTANCE] = 1
3089
    self.share_locks[locking.LEVEL_NODE] = 1
3090

    
3091
    if self.op.names:
3092
      self.wanted = _GetWantedInstances(self, self.op.names)
3093
    else:
3094
      self.wanted = locking.ALL_SET
3095

    
3096
    self.do_locking = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
3097
    if self.do_locking:
3098
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
3099
      self.needed_locks[locking.LEVEL_NODE] = []
3100
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3101

    
3102
  def DeclareLocks(self, level):
3103
    if level == locking.LEVEL_NODE and self.do_locking:
3104
      self._LockInstancesNodes()
3105

    
3106
  def CheckPrereq(self):
3107
    """Check prerequisites.
3108

3109
    """
3110
    pass
3111

    
3112
  def Exec(self, feedback_fn):
3113
    """Computes the list of nodes and their attributes.
3114

3115
    """
3116
    all_info = self.cfg.GetAllInstancesInfo()
3117
    if self.do_locking:
3118
      instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
3119
    elif self.wanted != locking.ALL_SET:
3120
      instance_names = self.wanted
3121
      missing = set(instance_names).difference(all_info.keys())
3122
      if missing:
3123
        raise errors.OpExecError(
3124
          "Some instances were removed before retrieving their data: %s"
3125
          % missing)
3126
    else:
3127
      instance_names = all_info.keys()
3128

    
3129
    instance_names = utils.NiceSort(instance_names)
3130
    instance_list = [all_info[iname] for iname in instance_names]
3131

    
3132
    # begin data gathering
3133

    
3134
    nodes = frozenset([inst.primary_node for inst in instance_list])
3135
    hv_list = list(set([inst.hypervisor for inst in instance_list]))
3136

    
3137
    bad_nodes = []
3138
    off_nodes = []
3139
    if self.do_locking:
3140
      live_data = {}
3141
      node_data = self.rpc.call_all_instances_info(nodes, hv_list)
3142
      for name in nodes:
3143
        result = node_data[name]
3144
        if result.offline:
3145
          # offline nodes will be in both lists
3146
          off_nodes.append(name)
3147
        if result.failed:
3148
          bad_nodes.append(name)
3149
        else:
3150
          if result.data:
3151
            live_data.update(result.data)
3152
            # else no instance is alive
3153
    else:
3154
      live_data = dict([(name, {}) for name in instance_names])
3155

    
3156
    # end data gathering
3157

    
3158
    HVPREFIX = "hv/"
3159
    BEPREFIX = "be/"
3160
    output = []
3161
    for instance in instance_list:
3162
      iout = []
3163
      i_hv = self.cfg.GetClusterInfo().FillHV(instance)
3164
      i_be = self.cfg.GetClusterInfo().FillBE(instance)
3165
      for field in self.op.output_fields:
3166
        st_match = self._FIELDS_STATIC.Matches(field)
3167
        if field == "name":
3168
          val = instance.name
3169
        elif field == "os":
3170
          val = instance.os
3171
        elif field == "pnode":
3172
          val = instance.primary_node
3173
        elif field == "snodes":
3174
          val = list(instance.secondary_nodes)
3175
        elif field == "admin_state":
3176
          val = instance.admin_up
3177
        elif field == "oper_state":
3178
          if instance.primary_node in bad_nodes:
3179
            val = None
3180
          else:
3181
            val = bool(live_data.get(instance.name))
3182
        elif field == "status":
3183
          if instance.primary_node in off_nodes:
3184
            val = "ERROR_nodeoffline"
3185
          elif instance.primary_node in bad_nodes:
3186
            val = "ERROR_nodedown"
3187
          else:
3188
            running = bool(live_data.get(instance.name))
3189
            if running:
3190
              if instance.admin_up:
3191
                val = "running"
3192
              else:
3193
                val = "ERROR_up"
3194
            else:
3195
              if instance.admin_up:
3196
                val = "ERROR_down"
3197
              else:
3198
                val = "ADMIN_down"
3199
        elif field == "oper_ram":
3200
          if instance.primary_node in bad_nodes:
3201
            val = None
3202
          elif instance.name in live_data:
3203
            val = live_data[instance.name].get("memory", "?")
3204
          else:
3205
            val = "-"
3206
        elif field == "disk_template":
3207
          val = instance.disk_template
3208
        elif field == "ip":
3209
          val = instance.nics[0].ip
3210
        elif field == "bridge":
3211
          val = instance.nics[0].bridge
3212
        elif field == "mac":
3213
          val = instance.nics[0].mac
3214
        elif field == "sda_size" or field == "sdb_size":
3215
          idx = ord(field[2]) - ord('a')
3216
          try:
3217
            val = instance.FindDisk(idx).size
3218
          except errors.OpPrereqError:
3219
            val = None
3220
        elif field == "tags":
3221
          val = list(instance.GetTags())
3222
        elif field == "serial_no":
3223
          val = instance.serial_no
3224
        elif field == "network_port":
3225
          val = instance.network_port
3226
        elif field == "hypervisor":
3227
          val = instance.hypervisor
3228
        elif field == "hvparams":
3229
          val = i_hv
3230
        elif (field.startswith(HVPREFIX) and
3231
              field[len(HVPREFIX):] in constants.HVS_PARAMETERS):
3232
          val = i_hv.get(field[len(HVPREFIX):], None)
3233
        elif field == "beparams":
3234
          val = i_be
3235
        elif (field.startswith(BEPREFIX) and
3236
              field[len(BEPREFIX):] in constants.BES_PARAMETERS):
3237
          val = i_be.get(field[len(BEPREFIX):], None)
3238
        elif st_match and st_match.groups():
3239
          # matches a variable list
3240
          st_groups = st_match.groups()
3241
          if st_groups and st_groups[0] == "disk":
3242
            if st_groups[1] == "count":
3243
              val = len(instance.disks)
3244
            elif st_groups[1] == "sizes":
3245
              val = [disk.size for disk in instance.disks]
3246
            elif st_groups[1] == "size":
3247
              try:
3248
                val = instance.FindDisk(st_groups[2]).size
3249
              except errors.OpPrereqError:
3250
                val = None
3251
            else:
3252
              assert False, "Unhandled disk parameter"
3253
          elif st_groups[0] == "nic":
3254
            if st_groups[1] == "count":
3255
              val = len(instance.nics)
3256
            elif st_groups[1] == "macs":
3257
              val = [nic.mac for nic in instance.nics]
3258
            elif st_groups[1] == "ips":
3259
              val = [nic.ip for nic in instance.nics]
3260
            elif st_groups[1] == "bridges":
3261
              val = [nic.bridge for nic in instance.nics]
3262
            else:
3263
              # index-based item
3264
              nic_idx = int(st_groups[2])
3265
              if nic_idx >= len(instance.nics):
3266
                val = None
3267
              else:
3268
                if st_groups[1] == "mac":
3269
                  val = instance.nics[nic_idx].mac
3270
                elif st_groups[1] == "ip":
3271
                  val = instance.nics[nic_idx].ip
3272
                elif st_groups[1] == "bridge":
3273
                  val = instance.nics[nic_idx].bridge
3274
                else:
3275
                  assert False, "Unhandled NIC parameter"
3276
          else:
3277
            assert False, "Unhandled variable parameter"
3278
        else:
3279
          raise errors.ParameterError(field)
3280
        iout.append(val)
3281
      output.append(iout)
3282

    
3283
    return output
3284

    
3285

    
3286
class LUFailoverInstance(LogicalUnit):
3287
  """Failover an instance.
3288

3289
  """
3290
  HPATH = "instance-failover"
3291
  HTYPE = constants.HTYPE_INSTANCE
3292
  _OP_REQP = ["instance_name", "ignore_consistency"]
3293
  REQ_BGL = False
3294

    
3295
  def ExpandNames(self):
3296
    self._ExpandAndLockInstance()
3297
    self.needed_locks[locking.LEVEL_NODE] = []
3298
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3299

    
3300
  def DeclareLocks(self, level):
3301
    if level == locking.LEVEL_NODE:
3302
      self._LockInstancesNodes()
3303

    
3304
  def BuildHooksEnv(self):
3305
    """Build hooks env.
3306

3307
    This runs on master, primary and secondary nodes of the instance.
3308

3309
    """
3310
    env = {
3311
      "IGNORE_CONSISTENCY": self.op.ignore_consistency,
3312
      }
3313
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
3314
    nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
3315
    return env, nl, nl
3316

    
3317
  def CheckPrereq(self):
3318
    """Check prerequisites.
3319

3320
    This checks that the instance is in the cluster.
3321

3322
    """
3323
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3324
    assert self.instance is not None, \
3325
      "Cannot retrieve locked instance %s" % self.op.instance_name
3326

    
3327
    bep = self.cfg.GetClusterInfo().FillBE(instance)
3328
    if instance.disk_template not in constants.DTS_NET_MIRROR:
3329
      raise errors.OpPrereqError("Instance's disk layout is not"
3330
                                 " network mirrored, cannot failover.")
3331

    
3332
    secondary_nodes = instance.secondary_nodes
3333
    if not secondary_nodes:
3334
      raise errors.ProgrammerError("no secondary node but using "
3335
                                   "a mirrored disk template")
3336

    
3337
    target_node = secondary_nodes[0]
3338
    _CheckNodeOnline(self, target_node)
3339
    # check memory requirements on the secondary node
3340
    _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
3341
                         instance.name, bep[constants.BE_MEMORY],
3342
                         instance.hypervisor)
3343

    
3344
    # check bridge existance
3345
    brlist = [nic.bridge for nic in instance.nics]
3346
    result = self.rpc.call_bridges_exist(target_node, brlist)
3347
    result.Raise()
3348
    if not result.data:
3349
      raise errors.OpPrereqError("One or more target bridges %s does not"
3350
                                 " exist on destination node '%s'" %
3351
                                 (brlist, target_node))
3352

    
3353
  def Exec(self, feedback_fn):
3354
    """Failover an instance.
3355

3356
    The failover is done by shutting it down on its present node and
3357
    starting it on the secondary.
3358

3359
    """
3360
    instance = self.instance
3361

    
3362
    source_node = instance.primary_node
3363
    target_node = instance.secondary_nodes[0]
3364

    
3365
    feedback_fn("* checking disk consistency between source and target")
3366
    for dev in instance.disks:
3367
      # for drbd, these are drbd over lvm
3368
      if not _CheckDiskConsistency(self, dev, target_node, False):
3369
        if instance.admin_up and not self.op.ignore_consistency:
3370
          raise errors.OpExecError("Disk %s is degraded on target node,"
3371
                                   " aborting failover." % dev.iv_name)
3372

    
3373
    feedback_fn("* shutting down instance on source node")
3374
    logging.info("Shutting down instance %s on node %s",
3375
                 instance.name, source_node)
3376

    
3377
    result = self.rpc.call_instance_shutdown(source_node, instance)
3378
    if result.failed or not result.data:
3379
      if self.op.ignore_consistency:
3380
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
3381
                             " Proceeding"
3382
                             " anyway. Please make sure node %s is down",
3383
                             instance.name, source_node, source_node)
3384
      else:
3385
        raise errors.OpExecError("Could not shutdown instance %s on node %s" %
3386
                                 (instance.name, source_node))
3387

    
3388
    feedback_fn("* deactivating the instance's disks on source node")
3389
    if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
3390
      raise errors.OpExecError("Can't shut down the instance's disks.")
3391

    
3392
    instance.primary_node = target_node
3393
    # distribute new instance config to the other nodes
3394
    self.cfg.Update(instance)
3395

    
3396
    # Only start the instance if it's marked as up
3397
    if instance.admin_up:
3398
      feedback_fn("* activating the instance's disks on target node")
3399
      logging.info("Starting instance %s on node %s",
3400
                   instance.name, target_node)
3401

    
3402
      disks_ok, dummy = _AssembleInstanceDisks(self, instance,
3403
                                               ignore_secondaries=True)
3404
      if not disks_ok:
3405
        _ShutdownInstanceDisks(self, instance)
3406
        raise errors.OpExecError("Can't activate the instance's disks")
3407

    
3408
      feedback_fn("* starting the instance on the target node")
3409
      result = self.rpc.call_instance_start(target_node, instance, None)
3410
      msg = result.RemoteFailMsg()
3411
      if msg:
3412
        _ShutdownInstanceDisks(self, instance)
3413
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
3414
                                 (instance.name, target_node, msg))
3415

    
3416

    
3417
class LUMigrateInstance(LogicalUnit):
3418
  """Migrate an instance.
3419

3420
  This is migration without shutting down, compared to the failover,
3421
  which is done with shutdown.
3422

3423
  """
3424
  HPATH = "instance-migrate"
3425
  HTYPE = constants.HTYPE_INSTANCE
3426
  _OP_REQP = ["instance_name", "live", "cleanup"]
3427

    
3428
  REQ_BGL = False
3429

    
3430
  def ExpandNames(self):
3431
    self._ExpandAndLockInstance()
3432
    self.needed_locks[locking.LEVEL_NODE] = []
3433
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3434

    
3435
  def DeclareLocks(self, level):
3436
    if level == locking.LEVEL_NODE:
3437
      self._LockInstancesNodes()
3438

    
3439
  def BuildHooksEnv(self):
3440
    """Build hooks env.
3441

3442
    This runs on master, primary and secondary nodes of the instance.
3443

3444
    """
3445
    env = _BuildInstanceHookEnvByObject(self, self.instance)
3446
    nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
3447
    return env, nl, nl
3448

    
3449
  def CheckPrereq(self):
3450
    """Check prerequisites.
3451

3452
    This checks that the instance is in the cluster.
3453

3454
    """
3455
    instance = self.cfg.GetInstanceInfo(
3456
      self.cfg.ExpandInstanceName(self.op.instance_name))
3457
    if instance is None:
3458
      raise errors.OpPrereqError("Instance '%s' not known" %
3459
                                 self.op.instance_name)
3460

    
3461
    if instance.disk_template != constants.DT_DRBD8:
3462
      raise errors.OpPrereqError("Instance's disk layout is not"
3463
                                 " drbd8, cannot migrate.")
3464

    
3465
    secondary_nodes = instance.secondary_nodes
3466
    if not secondary_nodes:
3467
      raise errors.ProgrammerError("no secondary node but using "
3468
                                   "drbd8 disk template")
3469

    
3470
    i_be = self.cfg.GetClusterInfo().FillBE(instance)
3471

    
3472
    target_node = secondary_nodes[0]
3473
    # check memory requirements on the secondary node
3474
    _CheckNodeFreeMemory(self, target_node, "migrating instance %s" %
3475
                         instance.name, i_be[constants.BE_MEMORY],
3476
                         instance.hypervisor)
3477

    
3478
    # check bridge existance
3479
    brlist = [nic.bridge for nic in instance.nics]
3480
    result = self.rpc.call_bridges_exist(target_node, brlist)
3481
    if result.failed or not result.data:
3482
      raise errors.OpPrereqError("One or more target bridges %s does not"
3483
                                 " exist on destination node '%s'" %
3484
                                 (brlist, target_node))
3485

    
3486
    if not self.op.cleanup:
3487
      result = self.rpc.call_instance_migratable(instance.primary_node,
3488
                                                 instance)
3489
      msg = result.RemoteFailMsg()
3490
      if msg:
3491
        raise errors.OpPrereqError("Can't migrate: %s - please use failover" %
3492
                                   msg)
3493

    
3494
    self.instance = instance
3495

    
3496
  def _WaitUntilSync(self):
3497
    """Poll with custom rpc for disk sync.
3498

3499
    This uses our own step-based rpc call.
3500

3501
    """
3502
    self.feedback_fn("* wait until resync is done")
3503
    all_done = False
3504
    while not all_done:
3505
      all_done = True
3506
      result = self.rpc.call_drbd_wait_sync(self.all_nodes,
3507
                                            self.nodes_ip,
3508
                                            self.instance.disks)
3509
      min_percent = 100
3510
      for node, nres in result.items():
3511
        msg = nres.RemoteFailMsg()
3512
        if msg:
3513
          raise errors.OpExecError("Cannot resync disks on node %s: %s" %
3514
                                   (node, msg))
3515
        node_done, node_percent = nres.data[1]
3516
        all_done = all_done and node_done
3517
        if node_percent is not None:
3518
          min_percent = min(min_percent, node_percent)
3519
      if not all_done:
3520
        if min_percent < 100:
3521
          self.feedback_fn("   - progress: %.1f%%" % min_percent)
3522
        time.sleep(2)
3523

    
3524
  def _EnsureSecondary(self, node):
3525
    """Demote a node to secondary.
3526

3527
    """
3528
    self.feedback_fn("* switching node %s to secondary mode" % node)
3529

    
3530
    for dev in self.instance.disks:
3531
      self.cfg.SetDiskID(dev, node)
3532

    
3533
    result = self.rpc.call_blockdev_close(node, self.instance.name,
3534
                                          self.instance.disks)
3535
    msg = result.RemoteFailMsg()
3536
    if msg:
3537
      raise errors.OpExecError("Cannot change disk to secondary on node %s,"
3538
                               " error %s" % (node, msg))
3539

    
3540
  def _GoStandalone(self):
3541
    """Disconnect from the network.
3542

3543
    """
3544
    self.feedback_fn("* changing into standalone mode")
3545
    result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
3546
                                               self.instance.disks)
3547
    for node, nres in result.items():
3548
      msg = nres.RemoteFailMsg()
3549
      if msg:
3550
        raise errors.OpExecError("Cannot disconnect disks node %s,"
3551
                                 " error %s" % (node, msg))
3552

    
3553
  def _GoReconnect(self, multimaster):
3554
    """Reconnect to the network.
3555

3556
    """
3557
    if multimaster:
3558
      msg = "dual-master"
3559
    else:
3560
      msg = "single-master"
3561
    self.feedback_fn("* changing disks into %s mode" % msg)
3562
    result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
3563
                                           self.instance.disks,
3564
                                           self.instance.name, multimaster)
3565
    for node, nres in result.items():
3566
      msg = nres.RemoteFailMsg()
3567
      if msg:
3568
        raise errors.OpExecError("Cannot change disks config on node %s,"
3569
                                 " error: %s" % (node, msg))
3570

    
3571
  def _ExecCleanup(self):
3572
    """Try to cleanup after a failed migration.
3573

3574
    The cleanup is done by:
3575
      - check that the instance is running only on one node
3576
        (and update the config if needed)
3577
      - change disks on its secondary node to secondary
3578
      - wait until disks are fully synchronized
3579
      - disconnect from the network
3580
      - change disks into single-master mode
3581
      - wait again until disks are fully synchronized
3582

3583
    """
3584
    instance = self.instance
3585
    target_node = self.target_node
3586
    source_node = self.source_node
3587

    
3588
    # check running on only one node
3589
    self.feedback_fn("* checking where the instance actually runs"
3590
                     " (if this hangs, the hypervisor might be in"
3591
                     " a bad state)")
3592
    ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
3593
    for node, result in ins_l.items():
3594
      result.Raise()
3595
      if not isinstance(result.data, list):
3596
        raise errors.OpExecError("Can't contact node '%s'" % node)
3597

    
3598
    runningon_source = instance.name in ins_l[source_node].data
3599
    runningon_target = instance.name in ins_l[target_node].data
3600

    
3601
    if runningon_source and runningon_target:
3602
      raise errors.OpExecError("Instance seems to be running on two nodes,"
3603
                               " or the hypervisor is confused. You will have"
3604
                               " to ensure manually that it runs only on one"
3605
                               " and restart this operation.")
3606

    
3607
    if not (runningon_source or runningon_target):
3608
      raise errors.OpExecError("Instance does not seem to be running at all."
3609
                               " In this case, it's safer to repair by"
3610
                               " running 'gnt-instance stop' to ensure disk"
3611
                               " shutdown, and then restarting it.")
3612

    
3613
    if runningon_target:
3614
      # the migration has actually succeeded, we need to update the config
3615
      self.feedback_fn("* instance running on secondary node (%s),"
3616
                       " updating config" % target_node)
3617
      instance.primary_node = target_node
3618
      self.cfg.Update(instance)
3619
      demoted_node = source_node
3620
    else:
3621
      self.feedback_fn("* instance confirmed to be running on its"
3622
                       " primary node (%s)" % source_node)
3623
      demoted_node = target_node
3624

    
3625
    self._EnsureSecondary(demoted_node)
3626
    try:
3627
      self._WaitUntilSync()
3628
    except errors.OpExecError:
3629
      # we ignore here errors, since if the device is standalone, it
3630
      # won't be able to sync
3631
      pass
3632
    self._GoStandalone()
3633
    self._GoReconnect(False)
3634
    self._WaitUntilSync()
3635

    
3636
    self.feedback_fn("* done")
3637

    
3638
  def _RevertDiskStatus(self):
3639
    """Try to revert the disk status after a failed migration.
3640

3641
    """
3642
    target_node = self.target_node
3643
    try:
3644
      self._EnsureSecondary(target_node)
3645
      self._GoStandalone()
3646
      self._GoReconnect(False)
3647
      self._WaitUntilSync()
3648
    except errors.OpExecError, err:
3649
      self.LogWarning("Migration failed and I can't reconnect the"
3650
                      " drives: error '%s'\n"
3651
                      "Please look and recover the instance status" %
3652
                      str(err))
3653

    
3654
  def _AbortMigration(self):
3655
    """Call the hypervisor code to abort a started migration.
3656

3657
    """
3658
    instance = self.instance
3659
    target_node = self.target_node
3660
    migration_info = self.migration_info
3661

    
3662
    abort_result = self.rpc.call_finalize_migration(target_node,
3663
                                                    instance,
3664
                                                    migration_info,
3665
                                                    False)
3666
    abort_msg = abort_result.RemoteFailMsg()
3667
    if abort_msg:
3668
      logging.error("Aborting migration failed on target node %s: %s" %
3669
                    (target_node, abort_msg))
3670
      # Don't raise an exception here, as we stil have to try to revert the
3671
      # disk status, even if this step failed.
3672

    
3673
  def _ExecMigration(self):
3674
    """Migrate an instance.
3675

3676
    The migrate is done by:
3677
      - change the disks into dual-master mode
3678
      - wait until disks are fully synchronized again
3679
      - migrate the instance
3680
      - change disks on the new secondary node (the old primary) to secondary
3681
      - wait until disks are fully synchronized
3682
      - change disks into single-master mode
3683

3684
    """
3685
    instance = self.instance
3686
    target_node = self.target_node
3687
    source_node = self.source_node
3688

    
3689
    self.feedback_fn("* checking disk consistency between source and target")
3690
    for dev in instance.disks:
3691
      if not _CheckDiskConsistency(self, dev, target_node, False):
3692
        raise errors.OpExecError("Disk %s is degraded or not fully"
3693
                                 " synchronized on target node,"
3694
                                 " aborting migrate." % dev.iv_name)
3695

    
3696
    # First get the migration information from the remote node
3697
    result = self.rpc.call_migration_info(source_node, instance)
3698
    msg = result.RemoteFailMsg()
3699
    if msg:
3700
      log_err = ("Failed fetching source migration information from %s: %s" %
3701
                  (source_node, msg))
3702
      logging.error(log_err)
3703
      raise errors.OpExecError(log_err)
3704

    
3705
    self.migration_info = migration_info = result.data[1]
3706

    
3707
    # Then switch the disks to master/master mode
3708
    self._EnsureSecondary(target_node)
3709
    self._GoStandalone()
3710
    self._GoReconnect(True)
3711
    self._WaitUntilSync()
3712

    
3713
    self.feedback_fn("* preparing %s to accept the instance" % target_node)
3714
    result = self.rpc.call_accept_instance(target_node,
3715
                                           instance,
3716
                                           migration_info,
3717
                                           self.nodes_ip[target_node])
3718

    
3719
    msg = result.RemoteFailMsg()
3720
    if msg:
3721
      logging.error("Instance pre-migration failed, trying to revert"
3722
                    " disk status: %s", msg)
3723
      self._AbortMigration()
3724
      self._RevertDiskStatus()
3725
      raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
3726
                               (instance.name, msg))
3727

    
3728
    self.feedback_fn("* migrating instance to %s" % target_node)
3729
    time.sleep(10)
3730
    result = self.rpc.call_instance_migrate(source_node, instance,
3731
                                            self.nodes_ip[target_node],
3732
                                            self.op.live)
3733
    msg = result.RemoteFailMsg()
3734
    if msg:
3735
      logging.error("Instance migration failed, trying to revert"
3736
                    " disk status: %s", msg)
3737
      self._AbortMigration()
3738
      self._RevertDiskStatus()
3739
      raise errors.OpExecError("Could not migrate instance %s: %s" %
3740
                               (instance.name, msg))
3741
    time.sleep(10)
3742

    
3743
    instance.primary_node = target_node
3744
    # distribute new instance config to the other nodes
3745
    self.cfg.Update(instance)
3746

    
3747
    result = self.rpc.call_finalize_migration(target_node,
3748
                                              instance,
3749
                                              migration_info,
3750
                                              True)
3751
    msg = result.RemoteFailMsg()
3752
    if msg:
3753
      logging.error("Instance migration succeeded, but finalization failed:"
3754
                    " %s" % msg)
3755
      raise errors.OpExecError("Could not finalize instance migration: %s" %
3756
                               msg)
3757

    
3758
    self._EnsureSecondary(source_node)
3759
    self._WaitUntilSync()
3760
    self._GoStandalone()
3761
    self._GoReconnect(False)
3762
    self._WaitUntilSync()
3763

    
3764
    self.feedback_fn("* done")
3765

    
3766
  def Exec(self, feedback_fn):
3767
    """Perform the migration.
3768

3769
    """
3770
    self.feedback_fn = feedback_fn
3771

    
3772
    self.source_node = self.instance.primary_node
3773
    self.target_node = self.instance.secondary_nodes[0]
3774
    self.all_nodes = [self.source_node, self.target_node]
3775
    self.nodes_ip = {
3776
      self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
3777
      self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
3778
      }
3779
    if self.op.cleanup:
3780
      return self._ExecCleanup()
3781
    else:
3782
      return self._ExecMigration()
3783

    
3784

    
3785
def _CreateBlockDev(lu, node, instance, device, force_create,
3786
                    info, force_open):
3787
  """Create a tree of block devices on a given node.
3788

3789
  If this device type has to be created on secondaries, create it and
3790
  all its children.
3791

3792
  If not, just recurse to children keeping the same 'force' value.
3793

3794
  @param lu: the lu on whose behalf we execute
3795
  @param node: the node on which to create the device
3796
  @type instance: L{objects.Instance}
3797
  @param instance: the instance which owns the device
3798
  @type device: L{objects.Disk}
3799
  @param device: the device to create
3800
  @type force_create: boolean
3801
  @param force_create: whether to force creation of this device; this
3802
      will be change to True whenever we find a device which has
3803
      CreateOnSecondary() attribute
3804
  @param info: the extra 'metadata' we should attach to the device
3805
      (this will be represented as a LVM tag)
3806
  @type force_open: boolean
3807
  @param force_open: this parameter will be passes to the
3808
      L{backend.CreateBlockDevice} function where it specifies
3809
      whether we run on primary or not, and it affects both
3810
      the child assembly and the device own Open() execution
3811

3812
  """
3813
  if device.CreateOnSecondary():
3814
    force_create = True
3815

    
3816
  if device.children:
3817
    for child in device.children:
3818
      _CreateBlockDev(lu, node, instance, child, force_create,
3819
                      info, force_open)
3820

    
3821
  if not force_create:
3822
    return
3823

    
3824
  _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
3825

    
3826

    
3827
def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
3828
  """Create a single block device on a given node.
3829

3830
  This will not recurse over children of the device, so they must be
3831
  created in advance.
3832

3833
  @param lu: the lu on whose behalf we execute
3834
  @param node: the node on which to create the device
3835
  @type instance: L{objects.Instance}
3836
  @param instance: the instance which owns the device
3837
  @type device: L{objects.Disk}
3838
  @param device: the device to create
3839
  @param info: the extra 'metadata' we should attach to the device
3840
      (this will be represented as a LVM tag)
3841
  @type force_open: boolean
3842
  @param force_open: this parameter will be passes to the
3843
      L{backend.CreateBlockDevice} function where it specifies
3844
      whether we run on primary or not, and it affects both
3845
      the child assembly and the device own Open() execution
3846

3847
  """
3848
  lu.cfg.SetDiskID(device, node)
3849
  result = lu.rpc.call_blockdev_create(node, device, device.size,
3850
                                       instance.name, force_open, info)
3851
  msg = result.RemoteFailMsg()
3852
  if msg:
3853
    raise errors.OpExecError("Can't create block device %s on"
3854
                             " node %s for instance %s: %s" %
3855
                             (device, node, instance.name, msg))
3856
  if device.physical_id is None:
3857
    device.physical_id = result.data[1]
3858

    
3859

    
3860
def _GenerateUniqueNames(lu, exts):
3861
  """Generate a suitable LV name.
3862

3863
  This will generate a logical volume name for the given instance.
3864

3865
  """
3866
  results = []
3867
  for val in exts:
3868
    new_id = lu.cfg.GenerateUniqueID()
3869
    results.append("%s%s" % (new_id, val))
3870
  return results
3871

    
3872

    
3873
def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
3874
                         p_minor, s_minor):
3875
  """Generate a drbd8 device complete with its children.
3876

3877
  """
3878
  port = lu.cfg.AllocatePort()
3879
  vgname = lu.cfg.GetVGName()
3880
  shared_secret = lu.cfg.GenerateDRBDSecret()
3881
  dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
3882
                          logical_id=(vgname, names[0]))
3883
  dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
3884
                          logical_id=(vgname, names[1]))
3885
  drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
3886
                          logical_id=(primary, secondary, port,
3887
                                      p_minor, s_minor,
3888
                                      shared_secret),
3889
                          children=[dev_data, dev_meta],
3890
                          iv_name=iv_name)
3891
  return drbd_dev
3892

    
3893

    
3894
def _GenerateDiskTemplate(lu, template_name,
3895
                          instance_name, primary_node,
3896
                          secondary_nodes, disk_info,
3897
                          file_storage_dir, file_driver,
3898
                          base_index):
3899
  """Generate the entire disk layout for a given template type.
3900

3901
  """
3902
  #TODO: compute space requirements
3903

    
3904
  vgname = lu.cfg.GetVGName()
3905
  disk_count = len(disk_info)
3906
  disks = []
3907
  if template_name == constants.DT_DISKLESS:
3908
    pass
3909
  elif template_name == constants.DT_PLAIN:
3910
    if len(secondary_nodes) != 0:
3911
      raise errors.ProgrammerError("Wrong template configuration")
3912

    
3913
    names = _GenerateUniqueNames(lu, [".disk%d" % i
3914
                                      for i in range(disk_count)])
3915
    for idx, disk in enumerate(disk_info):
3916
      disk_index = idx + base_index
3917
      disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
3918
                              logical_id=(vgname, names[idx]),
3919
                              iv_name="disk/%d" % disk_index)
3920
      disks.append(disk_dev)
3921
  elif template_name == constants.DT_DRBD8:
3922
    if len(secondary_nodes) != 1:
3923
      raise errors.ProgrammerError("Wrong template configuration")
3924
    remote_node = secondary_nodes[0]
3925
    minors = lu.cfg.AllocateDRBDMinor(
3926
      [primary_node, remote_node] * len(disk_info), instance_name)
3927

    
3928
    names = []
3929
    for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % i
3930
                                               for i in range(disk_count)]):
3931
      names.append(lv_prefix + "_data")
3932
      names.append(lv_prefix + "_meta")
3933
    for idx, disk in enumerate(disk_info):
3934
      disk_index = idx + base_index
3935
      disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
3936
                                      disk["size"], names[idx*2:idx*2+2],
3937
                                      "disk/%d" % disk_index,
3938
                                      minors[idx*2], minors[idx*2+1])
3939
      disks.append(disk_dev)
3940
  elif template_name == constants.DT_FILE:
3941
    if len(secondary_nodes) != 0:
3942
      raise errors.ProgrammerError("Wrong template configuration")
3943

    
3944
    for idx, disk in enumerate(disk_info):
3945
      disk_index = idx + base_index
3946
      disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
3947
                              iv_name="disk/%d" % disk_index,
3948
                              logical_id=(file_driver,
3949
                                          "%s/disk%d" % (file_storage_dir,
3950
                                                         idx)))
3951
      disks.append(disk_dev)
3952
  else:
3953
    raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
3954
  return disks
3955

    
3956

    
3957
def _GetInstanceInfoText(instance):
3958
  """Compute that text that should be added to the disk's metadata.
3959

3960
  """
3961
  return "originstname+%s" % instance.name
3962

    
3963

    
3964
def _CreateDisks(lu, instance):
3965
  """Create all disks for an instance.
3966

3967
  This abstracts away some work from AddInstance.
3968

3969
  @type lu: L{LogicalUnit}
3970
  @param lu: the logical unit on whose behalf we execute
3971
  @type instance: L{objects.Instance}
3972
  @param instance: the instance whose disks we should create
3973
  @rtype: boolean
3974
  @return: the success of the creation
3975

3976
  """
3977
  info = _GetInstanceInfoText(instance)
3978
  pnode = instance.primary_node
3979

    
3980
  if instance.disk_template == constants.DT_FILE:
3981
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
3982
    result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
3983

    
3984
    if result.failed or not result.data:
3985
      raise errors.OpExecError("Could not connect to node '%s'" % pnode)
3986

    
3987
    if not result.data[0]:
3988
      raise errors.OpExecError("Failed to create directory '%s'" %
3989
                               file_storage_dir)
3990

    
3991
  # Note: this needs to be kept in sync with adding of disks in
3992
  # LUSetInstanceParams
3993
  for device in instance.disks:
3994
    logging.info("Creating volume %s for instance %s",
3995
                 device.iv_name, instance.name)
3996
    #HARDCODE
3997
    for node in instance.all_nodes:
3998
      f_create = node == pnode
3999
      _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
4000

    
4001

    
4002
def _RemoveDisks(lu, instance):
4003
  """Remove all disks for an instance.
4004

4005
  This abstracts away some work from `AddInstance()` and
4006
  `RemoveInstance()`. Note that in case some of the devices couldn't
4007
  be removed, the removal will continue with the other ones (compare
4008
  with `_CreateDisks()`).
4009

4010
  @type lu: L{LogicalUnit}
4011
  @param lu: the logical unit on whose behalf we execute
4012
  @type instance: L{objects.Instance}
4013
  @param instance: the instance whose disks we should remove
4014
  @rtype: boolean
4015
  @return: the success of the removal
4016

4017
  """
4018
  logging.info("Removing block devices for instance %s", instance.name)
4019

    
4020
  result = True
4021
  for device in instance.disks:
4022
    for node, disk in device.ComputeNodeTree(instance.primary_node):
4023
      lu.cfg.SetDiskID(disk, node)
4024
      result = lu.rpc.call_blockdev_remove(node, disk)
4025
      if result.failed or not result.data:
4026
        lu.proc.LogWarning("Could not remove block device %s on node %s,"
4027
                           " continuing anyway", device.iv_name, node)
4028
        result = False
4029

    
4030
  if instance.disk_template == constants.DT_FILE:
4031
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
4032
    result = lu.rpc.call_file_storage_dir_remove(instance.primary_node,
4033
                                                 file_storage_dir)
4034
    if result.failed or not result.data:
4035
      logging.error("Could not remove directory '%s'", file_storage_dir)
4036
      result = False
4037

    
4038
  return result
4039

    
4040

    
4041
def _ComputeDiskSize(disk_template, disks):
4042
  """Compute disk size requirements in the volume group
4043

4044
  """
4045
  # Required free disk space as a function of disk and swap space
4046
  req_size_dict = {
4047
    constants.DT_DISKLESS: None,
4048
    constants.DT_PLAIN: sum(d["size"] for d in disks),
4049
    # 128 MB are added for drbd metadata for each disk
4050
    constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
4051
    constants.DT_FILE: None,
4052
  }
4053

    
4054
  if disk_template not in req_size_dict:
4055
    raise errors.ProgrammerError("Disk template '%s' size requirement"
4056
                                 " is unknown" %  disk_template)
4057

    
4058
  return req_size_dict[disk_template]
4059

    
4060

    
4061
def _CheckHVParams(lu, nodenames, hvname, hvparams):
4062
  """Hypervisor parameter validation.
4063

4064
  This function abstract the hypervisor parameter validation to be
4065
  used in both instance create and instance modify.
4066

4067
  @type lu: L{LogicalUnit}
4068
  @param lu: the logical unit for which we check
4069
  @type nodenames: list
4070
  @param nodenames: the list of nodes on which we should check
4071
  @type hvname: string
4072
  @param hvname: the name of the hypervisor we should use
4073
  @type hvparams: dict
4074
  @param hvparams: the parameters which we need to check
4075
  @raise errors.OpPrereqError: if the parameters are not valid
4076

4077
  """
4078
  hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
4079
                                                  hvname,
4080
                                                  hvparams)
4081
  for node in nodenames:
4082
    info = hvinfo[node]
4083
    info.Raise()
4084
    if not info.data or not isinstance(info.data, (tuple, list)):
4085
      raise errors.OpPrereqError("Cannot get current information"
4086
                                 " from node '%s' (%s)" % (node, info.data))
4087
    if not info.data[0]:
4088
      raise errors.OpPrereqError("Hypervisor parameter validation failed:"
4089
                                 " %s" % info.data[1])
4090

    
4091

    
4092
class LUCreateInstance(LogicalUnit):
4093
  """Create an instance.
4094

4095
  """
4096
  HPATH = "instance-add"
4097
  HTYPE = constants.HTYPE_INSTANCE
4098
  _OP_REQP = ["instance_name", "disks", "disk_template",
4099
              "mode", "start",
4100
              "wait_for_sync", "ip_check", "nics",
4101
              "hvparams", "beparams"]
4102
  REQ_BGL = False
4103

    
4104
  def _ExpandNode(self, node):
4105
    """Expands and checks one node name.
4106

4107
    """
4108
    node_full = self.cfg.ExpandNodeName(node)
4109
    if node_full is None:
4110
      raise errors.OpPrereqError("Unknown node %s" % node)
4111
    return node_full
4112

    
4113
  def ExpandNames(self):
4114
    """ExpandNames for CreateInstance.
4115

4116
    Figure out the right locks for instance creation.
4117

4118
    """
4119
    self.needed_locks = {}
4120

    
4121
    # set optional parameters to none if they don't exist
4122
    for attr in ["pnode", "snode", "iallocator", "hypervisor"]:
4123
      if not hasattr(self.op, attr):
4124
        setattr(self.op, attr, None)
4125

    
4126
    # cheap checks, mostly valid constants given
4127

    
4128
    # verify creation mode
4129
    if self.op.mode not in (constants.INSTANCE_CREATE,
4130
                            constants.INSTANCE_IMPORT):
4131
      raise errors.OpPrereqError("Invalid instance creation mode '%s'" %
4132
                                 self.op.mode)
4133

    
4134
    # disk template and mirror node verification
4135
    if self.op.disk_template not in constants.DISK_TEMPLATES:
4136
      raise errors.OpPrereqError("Invalid disk template name")
4137

    
4138
    if self.op.hypervisor is None:
4139
      self.op.hypervisor = self.cfg.GetHypervisorType()
4140

    
4141
    cluster = self.cfg.GetClusterInfo()
4142
    enabled_hvs = cluster.enabled_hypervisors
4143
    if self.op.hypervisor not in enabled_hvs:
4144
      raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
4145
                                 " cluster (%s)" % (self.op.hypervisor,
4146
                                  ",".join(enabled_hvs)))
4147

    
4148
    # check hypervisor parameter syntax (locally)
4149

    
4150
    filled_hvp = cluster.FillDict(cluster.hvparams[self.op.hypervisor],
4151
                                  self.op.hvparams)
4152
    hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
4153
    hv_type.CheckParameterSyntax(filled_hvp)
4154

    
4155
    # fill and remember the beparams dict
4156
    utils.CheckBEParams(self.op.beparams)
4157
    self.be_full = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
4158
                                    self.op.beparams)
4159

    
4160
    #### instance parameters check
4161

    
4162
    # instance name verification
4163
    hostname1 = utils.HostInfo(self.op.instance_name)
4164
    self.op.instance_name = instance_name = hostname1.name
4165

    
4166
    # this is just a preventive check, but someone might still add this
4167
    # instance in the meantime, and creation will fail at lock-add time
4168
    if instance_name in self.cfg.GetInstanceList():
4169
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
4170
                                 instance_name)
4171

    
4172
    self.add_locks[locking.LEVEL_INSTANCE] = instance_name
4173

    
4174
    # NIC buildup
4175
    self.nics = []
4176
    for nic in self.op.nics:
4177
      # ip validity checks
4178
      ip = nic.get("ip", None)
4179
      if ip is None or ip.lower() == "none":
4180
        nic_ip = None
4181
      elif ip.lower() == constants.VALUE_AUTO:
4182
        nic_ip = hostname1.ip
4183
      else:
4184
        if not utils.IsValidIP(ip):
4185
          raise errors.OpPrereqError("Given IP address '%s' doesn't look"
4186
                                     " like a valid IP" % ip)
4187
        nic_ip = ip
4188

    
4189
      # MAC address verification
4190
      mac = nic.get("mac", constants.VALUE_AUTO)
4191
      if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
4192
        if not utils.IsValidMac(mac.lower()):
4193
          raise errors.OpPrereqError("Invalid MAC address specified: %s" %
4194
                                     mac)
4195
      # bridge verification
4196
      bridge = nic.get("bridge", self.cfg.GetDefBridge())
4197
      self.nics.append(objects.NIC(mac=mac, ip=nic_ip, bridge=bridge))
4198

    
4199
    # disk checks/pre-build
4200
    self.disks = []
4201
    for disk in self.op.disks:
4202
      mode = disk.get("mode", constants.DISK_RDWR)
4203
      if mode not in constants.DISK_ACCESS_SET:
4204
        raise errors.OpPrereqError("Invalid disk access mode '%s'" %
4205
                                   mode)
4206
      size = disk.get("size", None)
4207
      if size is None:
4208
        raise errors.OpPrereqError("Missing disk size")
4209
      try:
4210
        size = int(size)
4211
      except ValueError:
4212
        raise errors.OpPrereqError("Invalid disk size '%s'" % size)
4213
      self.disks.append({"size": size, "mode": mode})
4214

    
4215
    # used in CheckPrereq for ip ping check
4216
    self.check_ip = hostname1.ip
4217

    
4218
    # file storage checks
4219
    if (self.op.file_driver and
4220
        not self.op.file_driver in constants.FILE_DRIVER):
4221
      raise errors.OpPrereqError("Invalid file driver name '%s'" %
4222
                                 self.op.file_driver)
4223

    
4224
    if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
4225
      raise errors.OpPrereqError("File storage directory path not absolute")
4226

    
4227
    ### Node/iallocator related checks
4228
    if [self.op.iallocator, self.op.pnode].count(None) != 1:
4229
      raise errors.OpPrereqError("One and only one of iallocator and primary"
4230
                                 " node must be given")
4231

    
4232
    if self.op.iallocator:
4233
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4234
    else:
4235
      self.op.pnode = self._ExpandNode(self.op.pnode)
4236
      nodelist = [self.op.pnode]
4237
      if self.op.snode is not None:
4238
        self.op.snode = self._ExpandNode(self.op.snode)
4239
        nodelist.append(self.op.snode)
4240
      self.needed_locks[locking.LEVEL_NODE] = nodelist
4241

    
4242
    # in case of import lock the source node too
4243
    if self.op.mode == constants.INSTANCE_IMPORT:
4244
      src_node = getattr(self.op, "src_node", None)
4245
      src_path = getattr(self.op, "src_path", None)
4246

    
4247
      if src_path is None:
4248
        self.op.src_path = src_path = self.op.instance_name
4249

    
4250
      if src_node is None:
4251
        self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4252
        self.op.src_node = None
4253
        if os.path.isabs(src_path):
4254
          raise errors.OpPrereqError("Importing an instance from an absolute"
4255
                                     " path requires a source node option.")
4256
      else:
4257
        self.op.src_node = src_node = self._ExpandNode(src_node)
4258
        if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
4259
          self.needed_locks[locking.LEVEL_NODE].append(src_node)
4260
        if not os.path.isabs(src_path):
4261
          self.op.src_path = src_path = \
4262
            os.path.join(constants.EXPORT_DIR, src_path)
4263

    
4264
    else: # INSTANCE_CREATE
4265
      if getattr(self.op, "os_type", None) is None:
4266
        raise errors.OpPrereqError("No guest OS specified")
4267

    
4268
  def _RunAllocator(self):
4269
    """Run the allocator based on input opcode.
4270

4271
    """
4272
    nics = [n.ToDict() for n in self.nics]
4273
    ial = IAllocator(self,
4274
                     mode=constants.IALLOCATOR_MODE_ALLOC,
4275
                     name=self.op.instance_name,
4276
                     disk_template=self.op.disk_template,
4277
                     tags=[],
4278
                     os=self.op.os_type,
4279
                     vcpus=self.be_full[constants.BE_VCPUS],
4280
                     mem_size=self.be_full[constants.BE_MEMORY],
4281
                     disks=self.disks,
4282
                     nics=nics,
4283
                     hypervisor=self.op.hypervisor,
4284
                     )
4285

    
4286
    ial.Run(self.op.iallocator)
4287

    
4288
    if not ial.success:
4289
      raise errors.OpPrereqError("Can't compute nodes using"
4290
                                 " iallocator '%s': %s" % (self.op.iallocator,
4291
                                                           ial.info))
4292
    if len(ial.nodes) != ial.required_nodes:
4293
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
4294
                                 " of nodes (%s), required %s" %
4295
                                 (self.op.iallocator, len(ial.nodes),
4296
                                  ial.required_nodes))
4297
    self.op.pnode = ial.nodes[0]
4298
    self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
4299
                 self.op.instance_name, self.op.iallocator,
4300
                 ", ".join(ial.nodes))
4301
    if ial.required_nodes == 2:
4302
      self.op.snode = ial.nodes[1]
4303

    
4304
  def BuildHooksEnv(self):
4305
    """Build hooks env.
4306

4307
    This runs on master, primary and secondary nodes of the instance.
4308

4309
    """
4310
    env = {
4311
      "INSTANCE_DISK_TEMPLATE": self.op.disk_template,
4312
      "INSTANCE_DISK_SIZE": ",".join(str(d["size"]) for d in self.disks),
4313
      "INSTANCE_ADD_MODE": self.op.mode,
4314
      }
4315
    if self.op.mode == constants.INSTANCE_IMPORT:
4316
      env["INSTANCE_SRC_NODE"] = self.op.src_node
4317
      env["INSTANCE_SRC_PATH"] = self.op.src_path
4318
      env["INSTANCE_SRC_IMAGES"] = self.src_images
4319

    
4320
    env.update(_BuildInstanceHookEnv(name=self.op.instance_name,
4321
      primary_node=self.op.pnode,
4322
      secondary_nodes=self.secondaries,
4323
      status=self.instance_status,
4324
      os_type=self.op.os_type,
4325
      memory=self.be_full[constants.BE_MEMORY],
4326
      vcpus=self.be_full[constants.BE_VCPUS],
4327
      nics=[(n.ip, n.bridge, n.mac) for n in self.nics],
4328
    ))
4329

    
4330
    nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
4331
          self.secondaries)
4332
    return env, nl, nl
4333

    
4334

    
4335
  def CheckPrereq(self):
4336
    """Check prerequisites.
4337

4338
    """
4339
    if (not self.cfg.GetVGName() and
4340
        self.op.disk_template not in constants.DTS_NOT_LVM):
4341
      raise errors.OpPrereqError("Cluster does not support lvm-based"
4342
                                 " instances")
4343

    
4344

    
4345
    if self.op.mode == constants.INSTANCE_IMPORT:
4346
      src_node = self.op.src_node
4347
      src_path = self.op.src_path
4348

    
4349
      if src_node is None:
4350
        exp_list = self.rpc.call_export_list(
4351
          self.acquired_locks[locking.LEVEL_NODE])
4352
        found = False
4353
        for node in exp_list:
4354
          if not exp_list[node].failed and src_path in exp_list[node].data:
4355
            found = True
4356
            self.op.src_node = src_node = node
4357
            self.op.src_path = src_path = os.path.join(constants.EXPORT_DIR,
4358
                                                       src_path)
4359
            break
4360
        if not found:
4361
          raise errors.OpPrereqError("No export found for relative path %s" %
4362
                                      src_path)
4363

    
4364
      _CheckNodeOnline(self, src_node)
4365
      result = self.rpc.call_export_info(src_node, src_path)
4366
      result.Raise()
4367
      if not result.data:
4368
        raise errors.OpPrereqError("No export found in dir %s" % src_path)
4369

    
4370
      export_info = result.data
4371
      if not export_info.has_section(constants.INISECT_EXP):
4372
        raise errors.ProgrammerError("Corrupted export config")
4373

    
4374
      ei_version = export_info.get(constants.INISECT_EXP, 'version')
4375
      if (int(ei_version) != constants.EXPORT_VERSION):
4376
        raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
4377
                                   (ei_version, constants.EXPORT_VERSION))
4378

    
4379
      # Check that the new instance doesn't have less disks than the export
4380
      instance_disks = len(self.disks)
4381
      export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
4382
      if instance_disks < export_disks:
4383
        raise errors.OpPrereqError("Not enough disks to import."
4384
                                   " (instance: %d, export: %d)" %
4385
                                   (instance_disks, export_disks))
4386

    
4387
      self.op.os_type = export_info.get(constants.INISECT_EXP, 'os')
4388
      disk_images = []
4389
      for idx in range(export_disks):
4390
        option = 'disk%d_dump' % idx
4391
        if export_info.has_option(constants.INISECT_INS, option):
4392
          # FIXME: are the old os-es, disk sizes, etc. useful?
4393
          export_name = export_info.get(constants.INISECT_INS, option)
4394
          image = os.path.join(src_path, export_name)
4395
          disk_images.append(image)
4396
        else:
4397
          disk_images.append(False)
4398

    
4399
      self.src_images = disk_images
4400

    
4401
      old_name = export_info.get(constants.INISECT_INS, 'name')
4402
      # FIXME: int() here could throw a ValueError on broken exports
4403
      exp_nic_count = int(export_info.get(constants.INISECT_INS, 'nic_count'))
4404
      if self.op.instance_name == old_name:
4405
        for idx, nic in enumerate(self.nics):
4406
          if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
4407
            nic_mac_ini = 'nic%d_mac' % idx
4408
            nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
4409

    
4410
    # ip ping checks (we use the same ip that was resolved in ExpandNames)
4411
    if self.op.start and not self.op.ip_check:
4412
      raise errors.OpPrereqError("Cannot ignore IP address conflicts when"
4413
                                 " adding an instance in start mode")
4414

    
4415
    if self.op.ip_check:
4416
      if utils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
4417
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
4418
                                   (self.check_ip, self.op.instance_name))
4419

    
4420
    #### allocator run
4421

    
4422
    if self.op.iallocator is not None:
4423
      self._RunAllocator()
4424

    
4425
    #### node related checks
4426

    
4427
    # check primary node
4428
    self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
4429
    assert self.pnode is not None, \
4430
      "Cannot retrieve locked node %s" % self.op.pnode
4431
    if pnode.offline:
4432
      raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
4433
                                 pnode.name)
4434

    
4435
    self.secondaries = []
4436

    
4437
    # mirror node verification
4438
    if self.op.disk_template in constants.DTS_NET_MIRROR:
4439
      if self.op.snode is None:
4440
        raise errors.OpPrereqError("The networked disk templates need"
4441
                                   " a mirror node")
4442
      if self.op.snode == pnode.name:
4443
        raise errors.OpPrereqError("The secondary node cannot be"
4444
                                   " the primary node.")
4445
      self.secondaries.append(self.op.snode)
4446
      _CheckNodeOnline(self, self.op.snode)
4447

    
4448
    nodenames = [pnode.name] + self.secondaries
4449

    
4450
    req_size = _ComputeDiskSize(self.op.disk_template,
4451
                                self.disks)
4452

    
4453
    # Check lv size requirements
4454
    if req_size is not None:
4455
      nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
4456
                                         self.op.hypervisor)
4457
      for node in nodenames:
4458
        info = nodeinfo[node]
4459
        info.Raise()
4460
        info = info.data
4461
        if not info:
4462
          raise errors.OpPrereqError("Cannot get current information"
4463
                                     " from node '%s'" % node)
4464
        vg_free = info.get('vg_free', None)
4465
        if not isinstance(vg_free, int):
4466
          raise errors.OpPrereqError("Can't compute free disk space on"
4467
                                     " node %s" % node)
4468
        if req_size > info['vg_free']:
4469
          raise errors.OpPrereqError("Not enough disk space on target node %s."
4470
                                     " %d MB available, %d MB required" %
4471
                                     (node, info['vg_free'], req_size))
4472

    
4473
    _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
4474

    
4475
    # os verification
4476
    result = self.rpc.call_os_get(pnode.name, self.op.os_type)
4477
    result.Raise()
4478
    if not isinstance(result.data, objects.OS):
4479
      raise errors.OpPrereqError("OS '%s' not in supported os list for"
4480
                                 " primary node"  % self.op.os_type)
4481

    
4482
    # bridge check on primary node
4483
    bridges = [n.bridge for n in self.nics]
4484
    result = self.rpc.call_bridges_exist(self.pnode.name, bridges)
4485
    result.Raise()
4486
    if not result.data:
4487
      raise errors.OpPrereqError("One of the target bridges '%s' does not"
4488
                                 " exist on destination node '%s'" %
4489
                                 (",".join(bridges), pnode.name))
4490

    
4491
    # memory check on primary node
4492
    if self.op.start:
4493
      _CheckNodeFreeMemory(self, self.pnode.name,
4494
                           "creating instance %s" % self.op.instance_name,
4495
                           self.be_full[constants.BE_MEMORY],
4496
                           self.op.hypervisor)
4497

    
4498
    self.instance_status = self.op.start
4499

    
4500
  def Exec(self, feedback_fn):
4501
    """Create and add the instance to the cluster.
4502

4503
    """
4504
    instance = self.op.instance_name
4505
    pnode_name = self.pnode.name
4506

    
4507
    for nic in self.nics:
4508
      if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
4509
        nic.mac = self.cfg.GenerateMAC()
4510

    
4511
    ht_kind = self.op.hypervisor
4512
    if ht_kind in constants.HTS_REQ_PORT:
4513
      network_port = self.cfg.AllocatePort()
4514
    else:
4515
      network_port = None
4516

    
4517
    ##if self.op.vnc_bind_address is None:
4518
    ##  self.op.vnc_bind_address = constants.VNC_DEFAULT_BIND_ADDRESS
4519

    
4520
    # this is needed because os.path.join does not accept None arguments
4521
    if self.op.file_storage_dir is None:
4522
      string_file_storage_dir = ""
4523
    else:
4524
      string_file_storage_dir = self.op.file_storage_dir
4525

    
4526
    # build the full file storage dir path
4527
    file_storage_dir = os.path.normpath(os.path.join(
4528
                                        self.cfg.GetFileStorageDir(),
4529
                                        string_file_storage_dir, instance))
4530

    
4531

    
4532
    disks = _GenerateDiskTemplate(self,
4533
                                  self.op.disk_template,
4534
                                  instance, pnode_name,
4535
                                  self.secondaries,
4536
                                  self.disks,
4537
                                  file_storage_dir,
4538
                                  self.op.file_driver,
4539
                                  0)
4540

    
4541
    iobj = objects.Instance(name=instance, os=self.op.os_type,
4542
                            primary_node=pnode_name,
4543
                            nics=self.nics, disks=disks,
4544
                            disk_template=self.op.disk_template,
4545
                            admin_up=self.instance_status,
4546
                            network_port=network_port,
4547
                            beparams=self.op.beparams,
4548
                            hvparams=self.op.hvparams,
4549
                            hypervisor=self.op.hypervisor,
4550
                            )
4551

    
4552
    feedback_fn("* creating instance disks...")
4553
    try:
4554
      _CreateDisks(self, iobj)
4555
    except errors.OpExecError:
4556
      self.LogWarning("Device creation failed, reverting...")
4557
      try:
4558
        _RemoveDisks(self, iobj)
4559
      finally:
4560
        self.cfg.ReleaseDRBDMinors(instance)
4561
        raise
4562

    
4563
    feedback_fn("adding instance %s to cluster config" % instance)
4564

    
4565
    self.cfg.AddInstance(iobj)
4566
    # Declare that we don't want to remove the instance lock anymore, as we've
4567
    # added the instance to the config
4568
    del self.remove_locks[locking.LEVEL_INSTANCE]
4569
    # Unlock all the nodes
4570
    if self.op.mode == constants.INSTANCE_IMPORT:
4571
      nodes_keep = [self.op.src_node]
4572
      nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
4573
                       if node != self.op.src_node]
4574
      self.context.glm.release(locking.LEVEL_NODE, nodes_release)
4575
      self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
4576
    else:
4577
      self.context.glm.release(locking.LEVEL_NODE)
4578
      del self.acquired_locks[locking.LEVEL_NODE]
4579

    
4580
    if self.op.wait_for_sync:
4581
      disk_abort = not _WaitForSync(self, iobj)
4582
    elif iobj.disk_template in constants.DTS_NET_MIRROR:
4583
      # make sure the disks are not degraded (still sync-ing is ok)
4584
      time.sleep(15)
4585
      feedback_fn("* checking mirrors status")
4586
      disk_abort = not _WaitForSync(self, iobj, oneshot=True)
4587
    else:
4588
      disk_abort = False
4589

    
4590
    if disk_abort:
4591
      _RemoveDisks(self, iobj)
4592
      self.cfg.RemoveInstance(iobj.name)
4593
      # Make sure the instance lock gets removed
4594
      self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
4595
      raise errors.OpExecError("There are some degraded disks for"
4596
                               " this instance")
4597

    
4598
    feedback_fn("creating os for instance %s on node %s" %
4599
                (instance, pnode_name))
4600

    
4601
    if iobj.disk_template != constants.DT_DISKLESS:
4602
      if self.op.mode == constants.INSTANCE_CREATE:
4603
        feedback_fn("* running the instance OS create scripts...")
4604
        result = self.rpc.call_instance_os_add(pnode_name, iobj)
4605
        msg = result.RemoteFailMsg()
4606
        if msg:
4607
          raise errors.OpExecError("Could not add os for instance %s"
4608
                                   " on node %s: %s" %
4609
                                   (instance, pnode_name, msg))
4610

    
4611
      elif self.op.mode == constants.INSTANCE_IMPORT:
4612
        feedback_fn("* running the instance OS import scripts...")
4613
        src_node = self.op.src_node
4614
        src_images = self.src_images
4615
        cluster_name = self.cfg.GetClusterName()
4616
        import_result = self.rpc.call_instance_os_import(pnode_name, iobj,
4617
                                                         src_node, src_images,
4618
                                                         cluster_name)
4619
        import_result.Raise()
4620
        for idx, result in enumerate(import_result.data):
4621
          if not result:
4622
            self.LogWarning("Could not import the image %s for instance"
4623
                            " %s, disk %d, on node %s" %
4624
                            (src_images[idx], instance, idx, pnode_name))
4625
      else:
4626
        # also checked in the prereq part
4627
        raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
4628
                                     % self.op.mode)
4629

    
4630
    if self.op.start:
4631
      logging.info("Starting instance %s on node %s", instance, pnode_name)
4632
      feedback_fn("* starting instance...")
4633
      result = self.rpc.call_instance_start(pnode_name, iobj, None)
4634
      msg = result.RemoteFailMsg()
4635
      if msg:
4636
        raise errors.OpExecError("Could not start instance: %s" % msg)
4637

    
4638

    
4639
class LUConnectConsole(NoHooksLU):
4640
  """Connect to an instance's console.
4641

4642
  This is somewhat special in that it returns the command line that
4643
  you need to run on the master node in order to connect to the
4644
  console.
4645

4646
  """
4647
  _OP_REQP = ["instance_name"]
4648
  REQ_BGL = False
4649

    
4650
  def ExpandNames(self):
4651
    self._ExpandAndLockInstance()
4652

    
4653
  def CheckPrereq(self):
4654
    """Check prerequisites.
4655

4656
    This checks that the instance is in the cluster.
4657

4658
    """
4659
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4660
    assert self.instance is not None, \
4661
      "Cannot retrieve locked instance %s" % self.op.instance_name
4662
    _CheckNodeOnline(self, self.instance.primary_node)
4663

    
4664
  def Exec(self, feedback_fn):
4665
    """Connect to the console of an instance
4666

4667
    """
4668
    instance = self.instance
4669
    node = instance.primary_node
4670

    
4671
    node_insts = self.rpc.call_instance_list([node],
4672
                                             [instance.hypervisor])[node]
4673
    node_insts.Raise()
4674

    
4675
    if instance.name not in node_insts.data:
4676
      raise errors.OpExecError("Instance %s is not running." % instance.name)
4677

    
4678
    logging.debug("Connecting to console of %s on %s", instance.name, node)
4679

    
4680
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
4681
    console_cmd = hyper.GetShellCommandForConsole(instance)
4682

    
4683
    # build ssh cmdline
4684
    return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
4685

    
4686

    
4687
class LUReplaceDisks(LogicalUnit):
4688
  """Replace the disks of an instance.
4689

4690
  """
4691
  HPATH = "mirrors-replace"
4692
  HTYPE = constants.HTYPE_INSTANCE
4693
  _OP_REQP = ["instance_name", "mode", "disks"]
4694
  REQ_BGL = False
4695

    
4696
  def CheckArguments(self):
4697
    if not hasattr(self.op, "remote_node"):
4698
      self.op.remote_node = None
4699
    if not hasattr(self.op, "iallocator"):
4700
      self.op.iallocator = None
4701

    
4702
    # check for valid parameter combination
4703
    cnt = [self.op.remote_node, self.op.iallocator].count(None)
4704
    if self.op.mode == constants.REPLACE_DISK_CHG:
4705
      if cnt == 2:
4706
        raise errors.OpPrereqError("When changing the secondary either an"
4707
                                   " iallocator script must be used or the"
4708
                                   " new node given")
4709
      elif cnt == 0:
4710
        raise errors.OpPrereqError("Give either the iallocator or the new"
4711
                                   " secondary, not both")
4712
    else: # not replacing the secondary
4713
      if cnt != 2:
4714
        raise errors.OpPrereqError("The iallocator and new node options can"
4715
                                   " be used only when changing the"
4716
                                   " secondary node")
4717

    
4718
  def ExpandNames(self):
4719
    self._ExpandAndLockInstance()
4720

    
4721
    if self.op.iallocator is not None:
4722
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4723
    elif self.op.remote_node is not None:
4724
      remote_node = self.cfg.ExpandNodeName(self.op.remote_node)
4725
      if remote_node is None:
4726
        raise errors.OpPrereqError("Node '%s' not known" %
4727
                                   self.op.remote_node)
4728
      self.op.remote_node = remote_node
4729
      # Warning: do not remove the locking of the new secondary here
4730
      # unless DRBD8.AddChildren is changed to work in parallel;
4731
      # currently it doesn't since parallel invocations of
4732
      # FindUnusedMinor will conflict
4733
      self.needed_locks[locking.LEVEL_NODE] = [remote_node]
4734
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
4735
    else:
4736
      self.needed_locks[locking.LEVEL_NODE] = []
4737
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4738

    
4739
  def DeclareLocks(self, level):
4740
    # If we're not already locking all nodes in the set we have to declare the
4741
    # instance's primary/secondary nodes.
4742
    if (level == locking.LEVEL_NODE and
4743
        self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
4744
      self._LockInstancesNodes()
4745

    
4746
  def _RunAllocator(self):
4747
    """Compute a new secondary node using an IAllocator.
4748

4749
    """
4750
    ial = IAllocator(self,
4751
                     mode=constants.IALLOCATOR_MODE_RELOC,
4752
                     name=self.op.instance_name,
4753
                     relocate_from=[self.sec_node])
4754

    
4755
    ial.Run(self.op.iallocator)
4756

    
4757
    if not ial.success:
4758
      raise errors.OpPrereqError("Can't compute nodes using"
4759
                                 " iallocator '%s': %s" % (self.op.iallocator,
4760
                                                           ial.info))
4761
    if len(ial.nodes) != ial.required_nodes:
4762
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
4763
                                 " of nodes (%s), required %s" %
4764
                                 (len(ial.nodes), ial.required_nodes))
4765
    self.op.remote_node = ial.nodes[0]
4766
    self.LogInfo("Selected new secondary for the instance: %s",
4767
                 self.op.remote_node)
4768

    
4769
  def BuildHooksEnv(self):
4770
    """Build hooks env.
4771

4772
    This runs on the master, the primary and all the secondaries.
4773

4774
    """
4775
    env = {
4776
      "MODE": self.op.mode,
4777
      "NEW_SECONDARY": self.op.remote_node,
4778
      "OLD_SECONDARY": self.instance.secondary_nodes[0],
4779
      }
4780
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4781
    nl = [
4782
      self.cfg.GetMasterNode(),
4783
      self.instance.primary_node,
4784
      ]
4785
    if self.op.remote_node is not None:
4786
      nl.append(self.op.remote_node)
4787
    return env, nl, nl
4788

    
4789
  def CheckPrereq(self):
4790
    """Check prerequisites.
4791

4792
    This checks that the instance is in the cluster.
4793

4794
    """
4795
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4796
    assert instance is not None, \
4797
      "Cannot retrieve locked instance %s" % self.op.instance_name
4798
    self.instance = instance
4799

    
4800
    if instance.disk_template != constants.DT_DRBD8:
4801
      raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
4802
                                 " instances")
4803

    
4804
    if len(instance.secondary_nodes) != 1:
4805
      raise errors.OpPrereqError("The instance has a strange layout,"
4806
                                 " expected one secondary but found %d" %
4807
                                 len(instance.secondary_nodes))
4808

    
4809
    self.sec_node = instance.secondary_nodes[0]
4810

    
4811
    if self.op.iallocator is not None:
4812
      self._RunAllocator()
4813

    
4814
    remote_node = self.op.remote_node
4815
    if remote_node is not None:
4816
      self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
4817
      assert self.remote_node_info is not None, \
4818
        "Cannot retrieve locked node %s" % remote_node
4819
    else:
4820
      self.remote_node_info = None
4821
    if remote_node == instance.primary_node:
4822
      raise errors.OpPrereqError("The specified node is the primary node of"
4823
                                 " the instance.")
4824
    elif remote_node == self.sec_node:
4825
      raise errors.OpPrereqError("The specified node is already the"
4826
                                 " secondary node of the instance.")
4827

    
4828
    if self.op.mode == constants.REPLACE_DISK_PRI:
4829
      n1 = self.tgt_node = instance.primary_node
4830
      n2 = self.oth_node = self.sec_node
4831
    elif self.op.mode == constants.REPLACE_DISK_SEC:
4832
      n1 = self.tgt_node = self.sec_node
4833
      n2 = self.oth_node = instance.primary_node
4834
    elif self.op.mode == constants.REPLACE_DISK_CHG:
4835
      n1 = self.new_node = remote_node
4836
      n2 = self.oth_node = instance.primary_node
4837
      self.tgt_node = self.sec_node
4838
    else:
4839
      raise errors.ProgrammerError("Unhandled disk replace mode")
4840

    
4841
    _CheckNodeOnline(self, n1)
4842
    _CheckNodeOnline(self, n2)
4843

    
4844
    if not self.op.disks:
4845
      self.op.disks = range(len(instance.disks))
4846

    
4847
    for disk_idx in self.op.disks:
4848
      instance.FindDisk(disk_idx)
4849

    
4850
  def _ExecD8DiskOnly(self, feedback_fn):
4851
    """Replace a disk on the primary or secondary for dbrd8.
4852

4853
    The algorithm for replace is quite complicated:
4854

4855
      1. for each disk to be replaced:
4856

4857
        1. create new LVs on the target node with unique names
4858
        1. detach old LVs from the drbd device
4859
        1. rename old LVs to name_replaced.<time_t>
4860
        1. rename new LVs to old LVs
4861
        1. attach the new LVs (with the old names now) to the drbd device
4862

4863
      1. wait for sync across all devices
4864

4865
      1. for each modified disk:
4866

4867
        1. remove old LVs (which have the name name_replaces.<time_t>)
4868

4869
    Failures are not very well handled.
4870

4871
    """
4872
    steps_total = 6
4873
    warning, info = (self.proc.LogWarning, self.proc.LogInfo)
4874
    instance = self.instance
4875
    iv_names = {}
4876
    vgname = self.cfg.GetVGName()
4877
    # start of work
4878
    cfg = self.cfg
4879
    tgt_node = self.tgt_node
4880
    oth_node = self.oth_node
4881

    
4882
    # Step: check device activation
4883
    self.proc.LogStep(1, steps_total, "check device existence")
4884
    info("checking volume groups")
4885
    my_vg = cfg.GetVGName()
4886
    results = self.rpc.call_vg_list([oth_node, tgt_node])
4887
    if not results:
4888
      raise errors.OpExecError("Can't list volume groups on the nodes")
4889
    for node in oth_node, tgt_node:
4890
      res = results[node]
4891
      if res.failed or not res.data or my_vg not in res.data:
4892
        raise errors.OpExecError("Volume group '%s' not found on %s" %
4893
                                 (my_vg, node))
4894
    for idx, dev in enumerate(instance.disks):
4895
      if idx not in self.op.disks:
4896
        continue
4897
      for node in tgt_node, oth_node:
4898
        info("checking disk/%d on %s" % (idx, node))
4899
        cfg.SetDiskID(dev, node)
4900
        if not self.rpc.call_blockdev_find(node, dev):
4901
          raise errors.OpExecError("Can't find disk/%d on node %s" %
4902
                                   (idx, node))
4903

    
4904
    # Step: check other node consistency
4905
    self.proc.LogStep(2, steps_total, "check peer consistency")
4906
    for idx, dev in enumerate(instance.disks):
4907
      if idx not in self.op.disks:
4908
        continue
4909
      info("checking disk/%d consistency on %s" % (idx, oth_node))
4910
      if not _CheckDiskConsistency(self, dev, oth_node,
4911
                                   oth_node==instance.primary_node):
4912
        raise errors.OpExecError("Peer node (%s) has degraded storage, unsafe"
4913
                                 " to replace disks on this node (%s)" %
4914
                                 (oth_node, tgt_node))
4915

    
4916
    # Step: create new storage
4917
    self.proc.LogStep(3, steps_total, "allocate new storage")
4918
    for idx, dev in enumerate(instance.disks):
4919
      if idx not in self.op.disks:
4920
        continue
4921
      size = dev.size
4922
      cfg.SetDiskID(dev, tgt_node)
4923
      lv_names = [".disk%d_%s" % (idx, suf)
4924
                  for suf in ["data", "meta"]]
4925
      names = _GenerateUniqueNames(self, lv_names)
4926
      lv_data = objects.Disk(dev_type=constants.LD_LV, size=size,
4927
                             logical_id=(vgname, names[0]))
4928
      lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
4929
                             logical_id=(vgname, names[1]))
4930
      new_lvs = [lv_data, lv_meta]
4931
      old_lvs = dev.children
4932
      iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
4933
      info("creating new local storage on %s for %s" %
4934
           (tgt_node, dev.iv_name))
4935
      # we pass force_create=True to force the LVM creation
4936
      for new_lv in new_lvs:
4937
        _CreateBlockDev(self, tgt_node, instance, new_lv, True,
4938
                        _GetInstanceInfoText(instance), False)
4939

    
4940
    # Step: for each lv, detach+rename*2+attach
4941
    self.proc.LogStep(4, steps_total, "change drbd configuration")
4942
    for dev, old_lvs, new_lvs in iv_names.itervalues():
4943
      info("detaching %s drbd from local storage" % dev.iv_name)
4944
      result = self.rpc.call_blockdev_removechildren(tgt_node, dev, old_lvs)
4945
      result.Raise()
4946
      if not result.data:
4947
        raise errors.OpExecError("Can't detach drbd from local storage on node"
4948
                                 " %s for device %s" % (tgt_node, dev.iv_name))
4949
      #dev.children = []
4950
      #cfg.Update(instance)
4951

    
4952
      # ok, we created the new LVs, so now we know we have the needed
4953
      # storage; as such, we proceed on the target node to rename
4954
      # old_lv to _old, and new_lv to old_lv; note that we rename LVs
4955
      # using the assumption that logical_id == physical_id (which in
4956
      # turn is the unique_id on that node)
4957

    
4958
      # FIXME(iustin): use a better name for the replaced LVs
4959
      temp_suffix = int(time.time())
4960
      ren_fn = lambda d, suff: (d.physical_id[0],
4961
                                d.physical_id[1] + "_replaced-%s" % suff)
4962
      # build the rename list based on what LVs exist on the node
4963
      rlist = []
4964
      for to_ren in old_lvs:
4965
        find_res = self.rpc.call_blockdev_find(tgt_node, to_ren)
4966
        if not find_res.failed and find_res.data is not None: # device exists
4967
          rlist.append((to_ren, ren_fn(to_ren, temp_suffix)))
4968

    
4969
      info("renaming the old LVs on the target node")
4970
      result = self.rpc.call_blockdev_rename(tgt_node, rlist)
4971
      result.Raise()
4972
      if not result.data:
4973
        raise errors.OpExecError("Can't rename old LVs on node %s" % tgt_node)
4974
      # now we rename the new LVs to the old LVs
4975
      info("renaming the new LVs on the target node")
4976
      rlist = [(new, old.physical_id) for old, new in zip(old_lvs, new_lvs)]
4977
      result = self.rpc.call_blockdev_rename(tgt_node, rlist)
4978
      result.Raise()
4979
      if not result.data:
4980
        raise errors.OpExecError("Can't rename new LVs on node %s" % tgt_node)
4981

    
4982
      for old, new in zip(old_lvs, new_lvs):
4983
        new.logical_id = old.logical_id
4984
        cfg.SetDiskID(new, tgt_node)
4985

    
4986
      for disk in old_lvs:
4987
        disk.logical_id = ren_fn(disk, temp_suffix)
4988
        cfg.SetDiskID(disk, tgt_node)
4989

    
4990
      # now that the new lvs have the old name, we can add them to the device
4991
      info("adding new mirror component on %s" % tgt_node)
4992
      result = self.rpc.call_blockdev_addchildren(tgt_node, dev, new_lvs)
4993
      if result.failed or not result.data:
4994
        for new_lv in new_lvs:
4995
          result = self.rpc.call_blockdev_remove(tgt_node, new_lv)
4996
          if result.failed or not result.data:
4997
            warning("Can't rollback device %s", hint="manually cleanup unused"
4998
                    " logical volumes")
4999
        raise errors.OpExecError("Can't add local storage to drbd")
5000

    
5001
      dev.children = new_lvs
5002
      cfg.Update(instance)
5003

    
5004
    # Step: wait for sync
5005

    
5006
    # this can fail as the old devices are degraded and _WaitForSync
5007
    # does a combined result over all disks, so we don't check its
5008
    # return value
5009
    self.proc.LogStep(5, steps_total, "sync devices")
5010
    _WaitForSync(self, instance, unlock=True)
5011

    
5012
    # so check manually all the devices
5013
    for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
5014
      cfg.SetDiskID(dev, instance.primary_node)
5015
      result = self.rpc.call_blockdev_find(instance.primary_node, dev)
5016
      if result.failed or result.data[5]:
5017
        raise errors.OpExecError("DRBD device %s is degraded!" % name)
5018

    
5019
    # Step: remove old storage
5020
    self.proc.LogStep(6, steps_total, "removing old storage")
5021
    for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
5022
      info("remove logical volumes for %s" % name)
5023
      for lv in old_lvs:
5024
        cfg.SetDiskID(lv, tgt_node)
5025
        result = self.rpc.call_blockdev_remove(tgt_node, lv)
5026
        if result.failed or not result.data:
5027
          warning("Can't remove old LV", hint="manually remove unused LVs")
5028
          continue
5029

    
5030
  def _ExecD8Secondary(self, feedback_fn):
5031
    """Replace the secondary node for drbd8.
5032

5033
    The algorithm for replace is quite complicated:
5034
      - for all disks of the instance:
5035
        - create new LVs on the new node with same names
5036
        - shutdown the drbd device on the old secondary
5037
        - disconnect the drbd network on the primary
5038
        - create the drbd device on the new secondary
5039
        - network attach the drbd on the primary, using an artifice:
5040
          the drbd code for Attach() will connect to the network if it
5041
          finds a device which is connected to the good local disks but
5042
          not network enabled
5043
      - wait for sync across all devices
5044
      - remove all disks from the old secondary
5045

5046
    Failures are not very well handled.
5047

5048
    """
5049
    steps_total = 6
5050
    warning, info = (self.proc.LogWarning, self.proc.LogInfo)
5051
    instance = self.instance
5052
    iv_names = {}
5053
    # start of work
5054
    cfg = self.cfg
5055
    old_node = self.tgt_node
5056
    new_node = self.new_node
5057
    pri_node = instance.primary_node
5058
    nodes_ip = {
5059
      old_node: self.cfg.GetNodeInfo(old_node).secondary_ip,
5060
      new_node: self.cfg.GetNodeInfo(new_node).secondary_ip,
5061
      pri_node: self.cfg.GetNodeInfo(pri_node).secondary_ip,
5062
      }
5063

    
5064
    # Step: check device activation
5065
    self.proc.LogStep(1, steps_total, "check device existence")
5066
    info("checking volume groups")
5067
    my_vg = cfg.GetVGName()
5068
    results = self.rpc.call_vg_list([pri_node, new_node])
5069
    for node in pri_node, new_node:
5070
      res = results[node]
5071
      if res.failed or not res.data or my_vg not in res.data:
5072
        raise errors.OpExecError("Volume group '%s' not found on %s" %
5073
                                 (my_vg, node))
5074
    for idx, dev in enumerate(instance.disks):
5075
      if idx not in self.op.disks:
5076
        continue
5077
      info("checking disk/%d on %s" % (idx, pri_node))
5078
      cfg.SetDiskID(dev, pri_node)
5079
      result = self.rpc.call_blockdev_find(pri_node, dev)
5080
      result.Raise()
5081
      if not result.data:
5082
        raise errors.OpExecError("Can't find disk/%d on node %s" %
5083
                                 (idx, pri_node))
5084

    
5085
    # Step: check other node consistency
5086
    self.proc.LogStep(2, steps_total, "check peer consistency")
5087
    for idx, dev in enumerate(instance.disks):
5088
      if idx not in self.op.disks:
5089
        continue
5090
      info("checking disk/%d consistency on %s" % (idx, pri_node))
5091
      if not _CheckDiskConsistency(self, dev, pri_node, True, ldisk=True):
5092
        raise errors.OpExecError("Primary node (%s) has degraded storage,"
5093
                                 " unsafe to replace the secondary" %
5094
                                 pri_node)
5095

    
5096
    # Step: create new storage
5097
    self.proc.LogStep(3, steps_total, "allocate new storage")
5098
    for idx, dev in enumerate(instance.disks):
5099
      info("adding new local storage on %s for disk/%d" %
5100
           (new_node, idx))
5101
      # we pass force_create=True to force LVM creation
5102
      for new_lv in dev.children:
5103
        _CreateBlockDev(self, new_node, instance, new_lv, True,
5104
                        _GetInstanceInfoText(instance), False)
5105

    
5106
    # Step 4: dbrd minors and drbd setups changes
5107
    # after this, we must manually remove the drbd minors on both the
5108
    # error and the success paths
5109
    minors = cfg.AllocateDRBDMinor([new_node for dev in instance.disks],
5110
                                   instance.name)
5111
    logging.debug("Allocated minors %s" % (minors,))
5112
    self.proc.LogStep(4, steps_total, "changing drbd configuration")
5113
    for idx, (dev, new_minor) in enumerate(zip(instance.disks, minors)):
5114
      size = dev.size
5115
      info("activating a new drbd on %s for disk/%d" % (new_node, idx))
5116
      # create new devices on new_node; note that we create two IDs:
5117
      # one without port, so the drbd will be activated without
5118
      # networking information on the new node at this stage, and one
5119
      # with network, for the latter activation in step 4
5120
      (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
5121
      if pri_node == o_node1:
5122
        p_minor = o_minor1
5123
      else:
5124
        p_minor = o_minor2
5125

    
5126
      new_alone_id = (pri_node, new_node, None, p_minor, new_minor, o_secret)
5127
      new_net_id = (pri_node, new_node, o_port, p_minor, new_minor, o_secret)
5128

    
5129
      iv_names[idx] = (dev, dev.children, new_net_id)
5130
      logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
5131
                    new_net_id)
5132
      new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
5133
                              logical_id=new_alone_id,
5134
                              children=dev.children)
5135
      try:
5136
        _CreateSingleBlockDev(self, new_node, instance, new_drbd,
5137
                              _GetInstanceInfoText(instance), False)
5138
      except errors.BlockDeviceError:
5139
        self.cfg.ReleaseDRBDMinors(instance.name)
5140
        raise
5141

    
5142
    for idx, dev in enumerate(instance.disks):
5143
      # we have new devices, shutdown the drbd on the old secondary
5144
      info("shutting down drbd for disk/%d on old node" % idx)
5145
      cfg.SetDiskID(dev, old_node)
5146
      result = self.rpc.call_blockdev_shutdown(old_node, dev)
5147
      if result.failed or not result.data:
5148
        warning("Failed to shutdown drbd for disk/%d on old node" % idx,
5149
                hint="Please cleanup this device manually as soon as possible")
5150

    
5151
    info("detaching primary drbds from the network (=> standalone)")
5152
    result = self.rpc.call_drbd_disconnect_net([pri_node], nodes_ip,
5153
                                               instance.disks)[pri_node]
5154

    
5155
    msg = result.RemoteFailMsg()
5156
    if msg:
5157
      # detaches didn't succeed (unlikely)
5158
      self.cfg.ReleaseDRBDMinors(instance.name)
5159
      raise errors.OpExecError("Can't detach the disks from the network on"
5160
                               " old node: %s" % (msg,))
5161

    
5162
    # if we managed to detach at least one, we update all the disks of
5163
    # the instance to point to the new secondary
5164
    info("updating instance configuration")
5165
    for dev, _, new_logical_id in iv_names.itervalues():
5166
      dev.logical_id = new_logical_id
5167
      cfg.SetDiskID(dev, pri_node)
5168
    cfg.Update(instance)
5169

    
5170
    # and now perform the drbd attach
5171
    info("attaching primary drbds to new secondary (standalone => connected)")
5172
    result = self.rpc.call_drbd_attach_net([pri_node, new_node], nodes_ip,
5173
                                           instance.disks, instance.name,
5174
                                           False)
5175
    for to_node, to_result in result.items():
5176
      msg = to_result.RemoteFailMsg()
5177
      if msg:
5178
        warning("can't attach drbd disks on node %s: %s", to_node, msg,
5179
                hint="please do a gnt-instance info to see the"
5180
                " status of disks")
5181

    
5182
    # this can fail as the old devices are degraded and _WaitForSync
5183
    # does a combined result over all disks, so we don't check its
5184
    # return value
5185
    self.proc.LogStep(5, steps_total, "sync devices")
5186
    _WaitForSync(self, instance, unlock=True)
5187

    
5188
    # so check manually all the devices
5189
    for idx, (dev, old_lvs, _) in iv_names.iteritems():
5190
      cfg.SetDiskID(dev, pri_node)
5191
      result = self.rpc.call_blockdev_find(pri_node, dev)
5192
      result.Raise()
5193
      if result.data[5]:
5194
        raise errors.OpExecError("DRBD device disk/%d is degraded!" % idx)
5195

    
5196
    self.proc.LogStep(6, steps_total, "removing old storage")
5197
    for idx, (dev, old_lvs, _) in iv_names.iteritems():
5198
      info("remove logical volumes for disk/%d" % idx)
5199
      for lv in old_lvs:
5200
        cfg.SetDiskID(lv, old_node)
5201
        result = self.rpc.call_blockdev_remove(old_node, lv)
5202
        if result.failed or not result.data:
5203
          warning("Can't remove LV on old secondary",
5204
                  hint="Cleanup stale volumes by hand")
5205

    
5206
  def Exec(self, feedback_fn):
5207
    """Execute disk replacement.
5208

5209
    This dispatches the disk replacement to the appropriate handler.
5210

5211
    """
5212
    instance = self.instance
5213

    
5214
    # Activate the instance disks if we're replacing them on a down instance
5215
    if not instance.admin_up:
5216
      _StartInstanceDisks(self, instance, True)
5217

    
5218
    if self.op.mode == constants.REPLACE_DISK_CHG:
5219
      fn = self._ExecD8Secondary
5220
    else:
5221
      fn = self._ExecD8DiskOnly
5222

    
5223
    ret = fn(feedback_fn)
5224

    
5225
    # Deactivate the instance disks if we're replacing them on a down instance
5226
    if not instance.admin_up:
5227
      _SafeShutdownInstanceDisks(self, instance)
5228

    
5229
    return ret
5230

    
5231

    
5232
class LUGrowDisk(LogicalUnit):
5233
  """Grow a disk of an instance.
5234

5235
  """
5236
  HPATH = "disk-grow"
5237
  HTYPE = constants.HTYPE_INSTANCE
5238
  _OP_REQP = ["instance_name", "disk", "amount", "wait_for_sync"]
5239
  REQ_BGL = False
5240

    
5241
  def ExpandNames(self):
5242
    self._ExpandAndLockInstance()
5243
    self.needed_locks[locking.LEVEL_NODE] = []
5244
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5245

    
5246
  def DeclareLocks(self, level):
5247
    if level == locking.LEVEL_NODE:
5248
      self._LockInstancesNodes()
5249

    
5250
  def BuildHooksEnv(self):
5251
    """Build hooks env.
5252

5253
    This runs on the master, the primary and all the secondaries.
5254

5255
    """
5256
    env = {
5257
      "DISK": self.op.disk,
5258
      "AMOUNT": self.op.amount,
5259
      }
5260
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5261
    nl = [
5262
      self.cfg.GetMasterNode(),
5263
      self.instance.primary_node,
5264
      ]
5265
    return env, nl, nl
5266

    
5267
  def CheckPrereq(self):
5268
    """Check prerequisites.
5269

5270
    This checks that the instance is in the cluster.
5271

5272
    """
5273
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5274
    assert instance is not None, \
5275
      "Cannot retrieve locked instance %s" % self.op.instance_name
5276
    nodenames = list(instance.all_nodes)
5277
    for node in nodenames:
5278
      _CheckNodeOnline(self, node)
5279

    
5280

    
5281
    self.instance = instance
5282

    
5283
    if instance.disk_template not in (constants.DT_PLAIN, constants.DT_DRBD8):
5284
      raise errors.OpPrereqError("Instance's disk layout does not support"
5285
                                 " growing.")
5286

    
5287
    self.disk = instance.FindDisk(self.op.disk)
5288

    
5289
    nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
5290
                                       instance.hypervisor)
5291
    for node in nodenames:
5292
      info = nodeinfo[node]
5293
      if info.failed or not info.data:
5294
        raise errors.OpPrereqError("Cannot get current information"
5295
                                   " from node '%s'" % node)
5296
      vg_free = info.data.get('vg_free', None)
5297
      if not isinstance(vg_free, int):
5298
        raise errors.OpPrereqError("Can't compute free disk space on"
5299
                                   " node %s" % node)
5300
      if self.op.amount > vg_free:
5301
        raise errors.OpPrereqError("Not enough disk space on target node %s:"
5302
                                   " %d MiB available, %d MiB required" %
5303
                                   (node, vg_free, self.op.amount))
5304

    
5305
  def Exec(self, feedback_fn):
5306
    """Execute disk grow.
5307

5308
    """
5309
    instance = self.instance
5310
    disk = self.disk
5311
    for node in instance.all_nodes:
5312
      self.cfg.SetDiskID(disk, node)
5313
      result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
5314
      result.Raise()
5315
      if (not result.data or not isinstance(result.data, (list, tuple)) or
5316
          len(result.data) != 2):
5317
        raise errors.OpExecError("Grow request failed to node %s" % node)
5318
      elif not result.data[0]:
5319
        raise errors.OpExecError("Grow request failed to node %s: %s" %
5320
                                 (node, result.data[1]))
5321
    disk.RecordGrow(self.op.amount)
5322
    self.cfg.Update(instance)
5323
    if self.op.wait_for_sync:
5324
      disk_abort = not _WaitForSync(self, instance)
5325
      if disk_abort:
5326
        self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
5327
                             " status.\nPlease check the instance.")
5328

    
5329

    
5330
class LUQueryInstanceData(NoHooksLU):
5331
  """Query runtime instance data.
5332

5333
  """
5334
  _OP_REQP = ["instances", "static"]
5335
  REQ_BGL = False
5336

    
5337
  def ExpandNames(self):
5338
    self.needed_locks = {}
5339
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
5340

    
5341
    if not isinstance(self.op.instances, list):
5342
      raise errors.OpPrereqError("Invalid argument type 'instances'")
5343

    
5344
    if self.op.instances:
5345
      self.wanted_names = []
5346
      for name in self.op.instances:
5347
        full_name = self.cfg.ExpandInstanceName(name)
5348
        if full_name is None:
5349
          raise errors.OpPrereqError("Instance '%s' not known" % name)
5350
        self.wanted_names.append(full_name)
5351
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
5352
    else:
5353
      self.wanted_names = None
5354
      self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
5355

    
5356
    self.needed_locks[locking.LEVEL_NODE] = []
5357
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5358

    
5359
  def DeclareLocks(self, level):
5360
    if level == locking.LEVEL_NODE:
5361
      self._LockInstancesNodes()
5362

    
5363
  def CheckPrereq(self):
5364
    """Check prerequisites.
5365

5366
    This only checks the optional instance list against the existing names.
5367

5368
    """
5369
    if self.wanted_names is None:
5370
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
5371

    
5372
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
5373
                             in self.wanted_names]
5374
    return
5375

    
5376
  def _ComputeDiskStatus(self, instance, snode, dev):
5377
    """Compute block device status.
5378

5379
    """
5380
    static = self.op.static
5381
    if not static:
5382
      self.cfg.SetDiskID(dev, instance.primary_node)
5383
      dev_pstatus = self.rpc.call_blockdev_find(instance.primary_node, dev)
5384
      dev_pstatus.Raise()
5385
      dev_pstatus = dev_pstatus.data
5386
    else:
5387
      dev_pstatus = None
5388

    
5389
    if dev.dev_type in constants.LDS_DRBD:
5390
      # we change the snode then (otherwise we use the one passed in)
5391
      if dev.logical_id[0] == instance.primary_node:
5392
        snode = dev.logical_id[1]
5393
      else:
5394
        snode = dev.logical_id[0]
5395

    
5396
    if snode and not static:
5397
      self.cfg.SetDiskID(dev, snode)
5398
      dev_sstatus = self.rpc.call_blockdev_find(snode, dev)
5399
      dev_sstatus.Raise()
5400
      dev_sstatus = dev_sstatus.data
5401
    else:
5402
      dev_sstatus = None
5403

    
5404
    if dev.children:
5405
      dev_children = [self._ComputeDiskStatus(instance, snode, child)
5406
                      for child in dev.children]
5407
    else:
5408
      dev_children = []
5409

    
5410
    data = {
5411
      "iv_name": dev.iv_name,
5412
      "dev_type": dev.dev_type,
5413
      "logical_id": dev.logical_id,
5414
      "physical_id": dev.physical_id,
5415
      "pstatus": dev_pstatus,
5416
      "sstatus": dev_sstatus,
5417
      "children": dev_children,
5418
      "mode": dev.mode,
5419
      }
5420

    
5421
    return data
5422

    
5423
  def Exec(self, feedback_fn):
5424
    """Gather and return data"""
5425
    result = {}
5426

    
5427
    cluster = self.cfg.GetClusterInfo()
5428

    
5429
    for instance in self.wanted_instances:
5430
      if not self.op.static:
5431
        remote_info = self.rpc.call_instance_info(instance.primary_node,
5432
                                                  instance.name,
5433
                                                  instance.hypervisor)
5434
        remote_info.Raise()
5435
        remote_info = remote_info.data
5436
        if remote_info and "state" in remote_info:
5437
          remote_state = "up"
5438
        else:
5439
          remote_state = "down"
5440
      else:
5441
        remote_state = None
5442
      if instance.admin_up:
5443
        config_state = "up"
5444
      else:
5445
        config_state = "down"
5446

    
5447
      disks = [self._ComputeDiskStatus(instance, None, device)
5448
               for device in instance.disks]
5449

    
5450
      idict = {
5451
        "name": instance.name,
5452
        "config_state": config_state,
5453
        "run_state": remote_state,
5454
        "pnode": instance.primary_node,
5455
        "snodes": instance.secondary_nodes,
5456
        "os": instance.os,
5457
        "nics": [(nic.mac, nic.ip, nic.bridge) for nic in instance.nics],
5458
        "disks": disks,
5459
        "hypervisor": instance.hypervisor,
5460
        "network_port": instance.network_port,
5461
        "hv_instance": instance.hvparams,
5462
        "hv_actual": cluster.FillHV(instance),
5463
        "be_instance": instance.beparams,
5464
        "be_actual": cluster.FillBE(instance),
5465
        }
5466

    
5467
      result[instance.name] = idict
5468

    
5469
    return result
5470

    
5471

    
5472
class LUSetInstanceParams(LogicalUnit):
5473
  """Modifies an instances's parameters.
5474

5475
  """
5476
  HPATH = "instance-modify"
5477
  HTYPE = constants.HTYPE_INSTANCE
5478
  _OP_REQP = ["instance_name"]
5479
  REQ_BGL = False
5480

    
5481
  def CheckArguments(self):
5482
    if not hasattr(self.op, 'nics'):
5483
      self.op.nics = []
5484
    if not hasattr(self.op, 'disks'):
5485
      self.op.disks = []
5486
    if not hasattr(self.op, 'beparams'):
5487
      self.op.beparams = {}
5488
    if not hasattr(self.op, 'hvparams'):
5489
      self.op.hvparams = {}
5490
    self.op.force = getattr(self.op, "force", False)
5491
    if not (self.op.nics or self.op.disks or
5492
            self.op.hvparams or self.op.beparams):
5493
      raise errors.OpPrereqError("No changes submitted")
5494

    
5495
    utils.CheckBEParams(self.op.beparams)
5496

    
5497
    # Disk validation
5498
    disk_addremove = 0
5499
    for disk_op, disk_dict in self.op.disks:
5500
      if disk_op == constants.DDM_REMOVE:
5501
        disk_addremove += 1
5502
        continue
5503
      elif disk_op == constants.DDM_ADD:
5504
        disk_addremove += 1
5505
      else:
5506
        if not isinstance(disk_op, int):
5507
          raise errors.OpPrereqError("Invalid disk index")
5508
      if disk_op == constants.DDM_ADD:
5509
        mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
5510
        if mode not in (constants.DISK_RDONLY, constants.DISK_RDWR):
5511
          raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode)
5512
        size = disk_dict.get('size', None)
5513
        if size is None:
5514
          raise errors.OpPrereqError("Required disk parameter size missing")
5515
        try:
5516
          size = int(size)
5517
        except ValueError, err:
5518
          raise errors.OpPrereqError("Invalid disk size parameter: %s" %
5519
                                     str(err))
5520
        disk_dict['size'] = size
5521
      else:
5522
        # modification of disk
5523
        if 'size' in disk_dict:
5524
          raise errors.OpPrereqError("Disk size change not possible, use"
5525
                                     " grow-disk")
5526

    
5527
    if disk_addremove > 1:
5528
      raise errors.OpPrereqError("Only one disk add or remove operation"
5529
                                 " supported at a time")
5530

    
5531
    # NIC validation
5532
    nic_addremove = 0
5533
    for nic_op, nic_dict in self.op.nics:
5534
      if nic_op == constants.DDM_REMOVE:
5535
        nic_addremove += 1
5536
        continue
5537
      elif nic_op == constants.DDM_ADD:
5538
        nic_addremove += 1
5539
      else:
5540
        if not isinstance(nic_op, int):
5541
          raise errors.OpPrereqError("Invalid nic index")
5542

    
5543
      # nic_dict should be a dict
5544
      nic_ip = nic_dict.get('ip', None)
5545
      if nic_ip is not None:
5546
        if nic_ip.lower() == "none":
5547
          nic_dict['ip'] = None
5548
        else:
5549
          if not utils.IsValidIP(nic_ip):
5550
            raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip)
5551
      # we can only check None bridges and assign the default one
5552
      nic_bridge = nic_dict.get('bridge', None)
5553
      if nic_bridge is None:
5554
        nic_dict['bridge'] = self.cfg.GetDefBridge()
5555
      # but we can validate MACs
5556
      nic_mac = nic_dict.get('mac', None)
5557
      if nic_mac is not None:
5558
        if self.cfg.IsMacInUse(nic_mac):
5559
          raise errors.OpPrereqError("MAC address %s already in use"
5560
                                     " in cluster" % nic_mac)
5561
        if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
5562
          if not utils.IsValidMac(nic_mac):
5563
            raise errors.OpPrereqError("Invalid MAC address %s" % nic_mac)
5564
    if nic_addremove > 1:
5565
      raise errors.OpPrereqError("Only one NIC add or remove operation"
5566
                                 " supported at a time")
5567

    
5568
  def ExpandNames(self):
5569
    self._ExpandAndLockInstance()
5570
    self.needed_locks[locking.LEVEL_NODE] = []
5571
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5572

    
5573
  def DeclareLocks(self, level):
5574
    if level == locking.LEVEL_NODE:
5575
      self._LockInstancesNodes()
5576

    
5577
  def BuildHooksEnv(self):
5578
    """Build hooks env.
5579

5580
    This runs on the master, primary and secondaries.
5581

5582
    """
5583
    args = dict()
5584
    if constants.BE_MEMORY in self.be_new:
5585
      args['memory'] = self.be_new[constants.BE_MEMORY]
5586
    if constants.BE_VCPUS in self.be_new:
5587
      args['vcpus'] = self.be_new[constants.BE_VCPUS]
5588
    # FIXME: readd disk/nic changes
5589
    env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
5590
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5591
    return env, nl, nl
5592

    
5593
  def CheckPrereq(self):
5594
    """Check prerequisites.
5595

5596
    This only checks the instance list against the existing names.
5597

5598
    """
5599
    force = self.force = self.op.force
5600

    
5601
    # checking the new params on the primary/secondary nodes
5602

    
5603
    instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5604
    assert self.instance is not None, \
5605
      "Cannot retrieve locked instance %s" % self.op.instance_name
5606
    pnode = instance.primary_node
5607
    nodelist = list(instance.all_nodes)
5608

    
5609
    # hvparams processing
5610
    if self.op.hvparams:
5611
      i_hvdict = copy.deepcopy(instance.hvparams)
5612
      for key, val in self.op.hvparams.iteritems():
5613
        if val == constants.VALUE_DEFAULT:
5614
          try:
5615
            del i_hvdict[key]
5616
          except KeyError:
5617
            pass
5618
        elif val == constants.VALUE_NONE:
5619
          i_hvdict[key] = None
5620
        else:
5621
          i_hvdict[key] = val
5622
      cluster = self.cfg.GetClusterInfo()
5623
      hv_new = cluster.FillDict(cluster.hvparams[instance.hypervisor],
5624
                                i_hvdict)
5625
      # local check
5626
      hypervisor.GetHypervisor(
5627
        instance.hypervisor).CheckParameterSyntax(hv_new)
5628
      _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
5629
      self.hv_new = hv_new # the new actual values
5630
      self.hv_inst = i_hvdict # the new dict (without defaults)
5631
    else:
5632
      self.hv_new = self.hv_inst = {}
5633

    
5634
    # beparams processing
5635
    if self.op.beparams:
5636
      i_bedict = copy.deepcopy(instance.beparams)
5637
      for key, val in self.op.beparams.iteritems():
5638
        if val == constants.VALUE_DEFAULT:
5639
          try:
5640
            del i_bedict[key]
5641
          except KeyError:
5642
            pass
5643
        else:
5644
          i_bedict[key] = val
5645
      cluster = self.cfg.GetClusterInfo()
5646
      be_new = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
5647
                                i_bedict)
5648
      self.be_new = be_new # the new actual values
5649
      self.be_inst = i_bedict # the new dict (without defaults)
5650
    else:
5651
      self.be_new = self.be_inst = {}
5652

    
5653
    self.warn = []
5654

    
5655
    if constants.BE_MEMORY in self.op.beparams and not self.force:
5656
      mem_check_list = [pnode]
5657
      if be_new[constants.BE_AUTO_BALANCE]:
5658
        # either we changed auto_balance to yes or it was from before
5659
        mem_check_list.extend(instance.secondary_nodes)
5660
      instance_info = self.rpc.call_instance_info(pnode, instance.name,
5661
                                                  instance.hypervisor)
5662
      nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
5663
                                         instance.hypervisor)
5664
      if nodeinfo[pnode].failed or not isinstance(nodeinfo[pnode].data, dict):
5665
        # Assume the primary node is unreachable and go ahead
5666
        self.warn.append("Can't get info from primary node %s" % pnode)
5667
      else:
5668
        if not instance_info.failed and instance_info.data:
5669
          current_mem = instance_info.data['memory']
5670
        else:
5671
          # Assume instance not running
5672
          # (there is a slight race condition here, but it's not very probable,
5673
          # and we have no other way to check)
5674
          current_mem = 0
5675
        miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
5676
                    nodeinfo[pnode].data['memory_free'])
5677
        if miss_mem > 0:
5678
          raise errors.OpPrereqError("This change will prevent the instance"
5679
                                     " from starting, due to %d MB of memory"
5680
                                     " missing on its primary node" % miss_mem)
5681

    
5682
      if be_new[constants.BE_AUTO_BALANCE]:
5683
        for node, nres in nodeinfo.iteritems():
5684
          if node not in instance.secondary_nodes:
5685
            continue
5686
          if nres.failed or not isinstance(nres.data, dict):
5687
            self.warn.append("Can't get info from secondary node %s" % node)
5688
          elif be_new[constants.BE_MEMORY] > nres.data['memory_free']:
5689
            self.warn.append("Not enough memory to failover instance to"
5690
                             " secondary node %s" % node)
5691

    
5692
    # NIC processing
5693
    for nic_op, nic_dict in self.op.nics:
5694
      if nic_op == constants.DDM_REMOVE:
5695
        if not instance.nics:
5696
          raise errors.OpPrereqError("Instance has no NICs, cannot remove")
5697
        continue
5698
      if nic_op != constants.DDM_ADD:
5699
        # an existing nic
5700
        if nic_op < 0 or nic_op >= len(instance.nics):
5701
          raise errors.OpPrereqError("Invalid NIC index %s, valid values"
5702
                                     " are 0 to %d" %
5703
                                     (nic_op, len(instance.nics)))
5704
      nic_bridge = nic_dict.get('bridge', None)
5705
      if nic_bridge is not None:
5706
        if not self.rpc.call_bridges_exist(pnode, [nic_bridge]):
5707
          msg = ("Bridge '%s' doesn't exist on one of"
5708
                 " the instance nodes" % nic_bridge)
5709
          if self.force:
5710
            self.warn.append(msg)
5711
          else:
5712
            raise errors.OpPrereqError(msg)
5713

    
5714
    # DISK processing
5715
    if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
5716
      raise errors.OpPrereqError("Disk operations not supported for"
5717
                                 " diskless instances")
5718
    for disk_op, disk_dict in self.op.disks:
5719
      if disk_op == constants.DDM_REMOVE:
5720
        if len(instance.disks) == 1:
5721
          raise errors.OpPrereqError("Cannot remove the last disk of"
5722
                                     " an instance")
5723
        ins_l = self.rpc.call_instance_list([pnode], [instance.hypervisor])
5724
        ins_l = ins_l[pnode]
5725
        if ins_l.failed or not isinstance(ins_l.data, list):
5726
          raise errors.OpPrereqError("Can't contact node '%s'" % pnode)
5727
        if instance.name in ins_l.data:
5728
          raise errors.OpPrereqError("Instance is running, can't remove"
5729
                                     " disks.")
5730

    
5731
      if (disk_op == constants.DDM_ADD and
5732
          len(instance.nics) >= constants.MAX_DISKS):
5733
        raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
5734
                                   " add more" % constants.MAX_DISKS)
5735
      if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
5736
        # an existing disk
5737
        if disk_op < 0 or disk_op >= len(instance.disks):
5738
          raise errors.OpPrereqError("Invalid disk index %s, valid values"
5739
                                     " are 0 to %d" %
5740
                                     (disk_op, len(instance.disks)))
5741

    
5742
    return
5743

    
5744
  def Exec(self, feedback_fn):
5745
    """Modifies an instance.
5746

5747
    All parameters take effect only at the next restart of the instance.
5748

5749
    """
5750
    # Process here the warnings from CheckPrereq, as we don't have a
5751
    # feedback_fn there.
5752
    for warn in self.warn:
5753
      feedback_fn("WARNING: %s" % warn)
5754

    
5755
    result = []
5756
    instance = self.instance
5757
    # disk changes
5758
    for disk_op, disk_dict in self.op.disks:
5759
      if disk_op == constants.DDM_REMOVE:
5760
        # remove the last disk
5761
        device = instance.disks.pop()
5762
        device_idx = len(instance.disks)
5763
        for node, disk in device.ComputeNodeTree(instance.primary_node):
5764
          self.cfg.SetDiskID(disk, node)
5765
          rpc_result = self.rpc.call_blockdev_remove(node, disk)
5766
          if rpc_result.failed or not rpc_result.data:
5767
            self.proc.LogWarning("Could not remove disk/%d on node %s,"
5768
                                 " continuing anyway", device_idx, node)
5769
        result.append(("disk/%d" % device_idx, "remove"))
5770
      elif disk_op == constants.DDM_ADD:
5771
        # add a new disk
5772
        if instance.disk_template == constants.DT_FILE:
5773
          file_driver, file_path = instance.disks[0].logical_id
5774
          file_path = os.path.dirname(file_path)
5775
        else:
5776
          file_driver = file_path = None
5777
        disk_idx_base = len(instance.disks)
5778
        new_disk = _GenerateDiskTemplate(self,
5779
                                         instance.disk_template,
5780
                                         instance.name, instance.primary_node,
5781
                                         instance.secondary_nodes,
5782
                                         [disk_dict],
5783
                                         file_path,
5784
                                         file_driver,
5785
                                         disk_idx_base)[0]
5786
        new_disk.mode = disk_dict['mode']
5787
        instance.disks.append(new_disk)
5788
        info = _GetInstanceInfoText(instance)
5789

    
5790
        logging.info("Creating volume %s for instance %s",
5791
                     new_disk.iv_name, instance.name)
5792
        # Note: this needs to be kept in sync with _CreateDisks
5793
        #HARDCODE
5794
        for node in instance.all_nodes:
5795
          f_create = node == instance.primary_node
5796
          try:
5797
            _CreateBlockDev(self, node, instance, new_disk,
5798
                            f_create, info, f_create)
5799
          except errors.OpExecError, err:
5800
            self.LogWarning("Failed to create volume %s (%s) on"
5801
                            " node %s: %s",
5802
                            new_disk.iv_name, new_disk, node, err)
5803
        result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
5804
                       (new_disk.size, new_disk.mode)))
5805
      else:
5806
        # change a given disk
5807
        instance.disks[disk_op].mode = disk_dict['mode']
5808
        result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
5809
    # NIC changes
5810
    for nic_op, nic_dict in self.op.nics:
5811
      if nic_op == constants.DDM_REMOVE:
5812
        # remove the last nic
5813
        del instance.nics[-1]
5814
        result.append(("nic.%d" % len(instance.nics), "remove"))
5815
      elif nic_op == constants.DDM_ADD:
5816
        # add a new nic
5817
        if 'mac' not in nic_dict:
5818
          mac = constants.VALUE_GENERATE
5819
        else:
5820
          mac = nic_dict['mac']
5821
        if mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
5822
          mac = self.cfg.GenerateMAC()
5823
        new_nic = objects.NIC(mac=mac, ip=nic_dict.get('ip', None),
5824
                              bridge=nic_dict.get('bridge', None))
5825
        instance.nics.append(new_nic)
5826
        result.append(("nic.%d" % (len(instance.nics) - 1),
5827
                       "add:mac=%s,ip=%s,bridge=%s" %
5828
                       (new_nic.mac, new_nic.ip, new_nic.bridge)))
5829
      else:
5830
        # change a given nic
5831
        for key in 'mac', 'ip', 'bridge':
5832
          if key in nic_dict:
5833
            setattr(instance.nics[nic_op], key, nic_dict[key])
5834
            result.append(("nic.%s/%d" % (key, nic_op), nic_dict[key]))
5835

    
5836
    # hvparams changes
5837
    if self.op.hvparams:
5838
      instance.hvparams = self.hv_new
5839
      for key, val in self.op.hvparams.iteritems():
5840
        result.append(("hv/%s" % key, val))
5841

    
5842
    # beparams changes
5843
    if self.op.beparams:
5844
      instance.beparams = self.be_inst
5845
      for key, val in self.op.beparams.iteritems():
5846
        result.append(("be/%s" % key, val))
5847

    
5848
    self.cfg.Update(instance)
5849

    
5850
    return result
5851

    
5852

    
5853
class LUQueryExports(NoHooksLU):
5854
  """Query the exports list
5855

5856
  """
5857
  _OP_REQP = ['nodes']
5858
  REQ_BGL = False
5859

    
5860
  def ExpandNames(self):
5861
    self.needed_locks = {}
5862
    self.share_locks[locking.LEVEL_NODE] = 1
5863
    if not self.op.nodes:
5864
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5865
    else:
5866
      self.needed_locks[locking.LEVEL_NODE] = \
5867
        _GetWantedNodes(self, self.op.nodes)
5868

    
5869
  def CheckPrereq(self):
5870
    """Check prerequisites.
5871

5872
    """
5873
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
5874

    
5875
  def Exec(self, feedback_fn):
5876
    """Compute the list of all the exported system images.
5877

5878
    @rtype: dict
5879
    @return: a dictionary with the structure node->(export-list)
5880
        where export-list is a list of the instances exported on
5881
        that node.
5882

5883
    """
5884
    rpcresult = self.rpc.call_export_list(self.nodes)
5885
    result = {}
5886
    for node in rpcresult:
5887
      if rpcresult[node].failed:
5888
        result[node] = False
5889
      else:
5890
        result[node] = rpcresult[node].data
5891

    
5892
    return result
5893

    
5894

    
5895
class LUExportInstance(LogicalUnit):
5896
  """Export an instance to an image in the cluster.
5897

5898
  """
5899
  HPATH = "instance-export"
5900
  HTYPE = constants.HTYPE_INSTANCE
5901
  _OP_REQP = ["instance_name", "target_node", "shutdown"]
5902
  REQ_BGL = False
5903

    
5904
  def ExpandNames(self):
5905
    self._ExpandAndLockInstance()
5906
    # FIXME: lock only instance primary and destination node
5907
    #
5908
    # Sad but true, for now we have do lock all nodes, as we don't know where
5909
    # the previous export might be, and and in this LU we search for it and
5910
    # remove it from its current node. In the future we could fix this by:
5911
    #  - making a tasklet to search (share-lock all), then create the new one,
5912
    #    then one to remove, after
5913
    #  - removing the removal operation altoghether
5914
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5915

    
5916
  def DeclareLocks(self, level):
5917
    """Last minute lock declaration."""
5918
    # All nodes are locked anyway, so nothing to do here.
5919

    
5920
  def BuildHooksEnv(self):
5921
    """Build hooks env.
5922

5923
    This will run on the master, primary node and target node.
5924

5925
    """
5926
    env = {
5927
      "EXPORT_NODE": self.op.target_node,
5928
      "EXPORT_DO_SHUTDOWN": self.op.shutdown,
5929
      }
5930
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5931
    nl = [self.cfg.GetMasterNode(), self.instance.primary_node,
5932
          self.op.target_node]
5933
    return env, nl, nl
5934

    
5935
  def CheckPrereq(self):
5936
    """Check prerequisites.
5937

5938
    This checks that the instance and node names are valid.
5939

5940
    """
5941
    instance_name = self.op.instance_name
5942
    self.instance = self.cfg.GetInstanceInfo(instance_name)
5943
    assert self.instance is not None, \
5944
          "Cannot retrieve locked instance %s" % self.op.instance_name
5945
    _CheckNodeOnline(self, self.instance.primary_node)
5946

    
5947
    self.dst_node = self.cfg.GetNodeInfo(
5948
      self.cfg.ExpandNodeName(self.op.target_node))
5949

    
5950
    if self.dst_node is None:
5951
      # This is wrong node name, not a non-locked node
5952
      raise errors.OpPrereqError("Wrong node name %s" % self.op.target_node)
5953
    _CheckNodeOnline(self, self.dst_node.name)
5954

    
5955
    # instance disk type verification
5956
    for disk in self.instance.disks:
5957
      if disk.dev_type == constants.LD_FILE:
5958
        raise errors.OpPrereqError("Export not supported for instances with"
5959
                                   " file-based disks")
5960

    
5961
  def Exec(self, feedback_fn):
5962
    """Export an instance to an image in the cluster.
5963

5964
    """
5965
    instance = self.instance
5966
    dst_node = self.dst_node
5967
    src_node = instance.primary_node
5968
    if self.op.shutdown:
5969
      # shutdown the instance, but not the disks
5970
      result = self.rpc.call_instance_shutdown(src_node, instance)
5971
      result.Raise()
5972
      if not result.data:
5973
        raise errors.OpExecError("Could not shutdown instance %s on node %s" %
5974
                                 (instance.name, src_node))
5975

    
5976
    vgname = self.cfg.GetVGName()
5977

    
5978
    snap_disks = []
5979

    
5980
    # set the disks ID correctly since call_instance_start needs the
5981
    # correct drbd minor to create the symlinks
5982
    for disk in instance.disks:
5983
      self.cfg.SetDiskID(disk, src_node)
5984

    
5985
    try:
5986
      for disk in instance.disks:
5987
        # new_dev_name will be a snapshot of an lvm leaf of the one we passed
5988
        new_dev_name = self.rpc.call_blockdev_snapshot(src_node, disk)
5989
        if new_dev_name.failed or not new_dev_name.data:
5990
          self.LogWarning("Could not snapshot block device %s on node %s",
5991
                          disk.logical_id[1], src_node)
5992
          snap_disks.append(False)
5993
        else:
5994
          new_dev = objects.Disk(dev_type=constants.LD_LV, size=disk.size,
5995
                                 logical_id=(vgname, new_dev_name.data),
5996
                                 physical_id=(vgname, new_dev_name.data),
5997
                                 iv_name=disk.iv_name)
5998
          snap_disks.append(new_dev)
5999

    
6000
    finally:
6001
      if self.op.shutdown and instance.admin_up:
6002
        result = self.rpc.call_instance_start(src_node, instance, None)
6003
        msg = result.RemoteFailMsg()
6004
        if msg:
6005
          _ShutdownInstanceDisks(self, instance)
6006
          raise errors.OpExecError("Could not start instance: %s" % msg)
6007

    
6008
    # TODO: check for size
6009

    
6010
    cluster_name = self.cfg.GetClusterName()
6011
    for idx, dev in enumerate(snap_disks):
6012
      if dev:
6013
        result = self.rpc.call_snapshot_export(src_node, dev, dst_node.name,
6014
                                               instance, cluster_name, idx)
6015
        if result.failed or not result.data:
6016
          self.LogWarning("Could not export block device %s from node %s to"
6017
                          " node %s", dev.logical_id[1], src_node,
6018
                          dst_node.name)
6019
        result = self.rpc.call_blockdev_remove(src_node, dev)
6020
        if result.failed or not result.data:
6021
          self.LogWarning("Could not remove snapshot block device %s from node"
6022
                          " %s", dev.logical_id[1], src_node)
6023

    
6024
    result = self.rpc.call_finalize_export(dst_node.name, instance, snap_disks)
6025
    if result.failed or not result.data:
6026
      self.LogWarning("Could not finalize export for instance %s on node %s",
6027
                      instance.name, dst_node.name)
6028

    
6029
    nodelist = self.cfg.GetNodeList()
6030
    nodelist.remove(dst_node.name)
6031

    
6032
    # on one-node clusters nodelist will be empty after the removal
6033
    # if we proceed the backup would be removed because OpQueryExports
6034
    # substitutes an empty list with the full cluster node list.
6035
    if nodelist:
6036
      exportlist = self.rpc.call_export_list(nodelist)
6037
      for node in exportlist:
6038
        if exportlist[node].failed:
6039
          continue
6040
        if instance.name in exportlist[node].data:
6041
          if not self.rpc.call_export_remove(node, instance.name):
6042
            self.LogWarning("Could not remove older export for instance %s"
6043
                            " on node %s", instance.name, node)
6044

    
6045

    
6046
class LURemoveExport(NoHooksLU):
6047
  """Remove exports related to the named instance.
6048

6049
  """
6050
  _OP_REQP = ["instance_name"]
6051
  REQ_BGL = False
6052

    
6053
  def ExpandNames(self):
6054
    self.needed_locks = {}
6055
    # We need all nodes to be locked in order for RemoveExport to work, but we
6056
    # don't need to lock the instance itself, as nothing will happen to it (and
6057
    # we can remove exports also for a removed instance)
6058
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6059

    
6060
  def CheckPrereq(self):
6061
    """Check prerequisites.
6062
    """
6063
    pass
6064

    
6065
  def Exec(self, feedback_fn):
6066
    """Remove any export.
6067

6068
    """
6069
    instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
6070
    # If the instance was not found we'll try with the name that was passed in.
6071
    # This will only work if it was an FQDN, though.
6072
    fqdn_warn = False
6073
    if not instance_name:
6074
      fqdn_warn = True
6075
      instance_name = self.op.instance_name
6076

    
6077
    exportlist = self.rpc.call_export_list(self.acquired_locks[
6078
      locking.LEVEL_NODE])
6079
    found = False
6080
    for node in exportlist:
6081
      if exportlist[node].failed:
6082
        self.LogWarning("Failed to query node %s, continuing" % node)
6083
        continue
6084
      if instance_name in exportlist[node].data:
6085
        found = True
6086
        result = self.rpc.call_export_remove(node, instance_name)
6087
        if result.failed or not result.data:
6088
          logging.error("Could not remove export for instance %s"
6089
                        " on node %s", instance_name, node)
6090

    
6091
    if fqdn_warn and not found:
6092
      feedback_fn("Export not found. If trying to remove an export belonging"
6093
                  " to a deleted instance please use its Fully Qualified"
6094
                  " Domain Name.")
6095

    
6096

    
6097
class TagsLU(NoHooksLU):
6098
  """Generic tags LU.
6099

6100
  This is an abstract class which is the parent of all the other tags LUs.
6101

6102
  """
6103

    
6104
  def ExpandNames(self):
6105
    self.needed_locks = {}
6106
    if self.op.kind == constants.TAG_NODE:
6107
      name = self.cfg.ExpandNodeName(self.op.name)
6108
      if name is None:
6109
        raise errors.OpPrereqError("Invalid node name (%s)" %
6110
                                   (self.op.name,))
6111
      self.op.name = name
6112
      self.needed_locks[locking.LEVEL_NODE] = name
6113
    elif self.op.kind == constants.TAG_INSTANCE:
6114
      name = self.cfg.ExpandInstanceName(self.op.name)
6115
      if name is None:
6116
        raise errors.OpPrereqError("Invalid instance name (%s)" %
6117
                                   (self.op.name,))
6118
      self.op.name = name
6119
      self.needed_locks[locking.LEVEL_INSTANCE] = name
6120

    
6121
  def CheckPrereq(self):
6122
    """Check prerequisites.
6123

6124
    """
6125
    if self.op.kind == constants.TAG_CLUSTER:
6126
      self.target = self.cfg.GetClusterInfo()
6127
    elif self.op.kind == constants.TAG_NODE:
6128
      self.target = self.cfg.GetNodeInfo(self.op.name)
6129
    elif self.op.kind == constants.TAG_INSTANCE:
6130
      self.target = self.cfg.GetInstanceInfo(self.op.name)
6131
    else:
6132
      raise errors.OpPrereqError("Wrong tag type requested (%s)" %
6133
                                 str(self.op.kind))
6134

    
6135

    
6136
class LUGetTags(TagsLU):
6137
  """Returns the tags of a given object.
6138

6139
  """
6140
  _OP_REQP = ["kind", "name"]
6141
  REQ_BGL = False
6142

    
6143
  def Exec(self, feedback_fn):
6144
    """Returns the tag list.
6145

6146
    """
6147
    return list(self.target.GetTags())
6148

    
6149

    
6150
class LUSearchTags(NoHooksLU):
6151
  """Searches the tags for a given pattern.
6152

6153
  """
6154
  _OP_REQP = ["pattern"]
6155
  REQ_BGL = False
6156

    
6157
  def ExpandNames(self):
6158
    self.needed_locks = {}
6159

    
6160
  def CheckPrereq(self):
6161
    """Check prerequisites.
6162

6163
    This checks the pattern passed for validity by compiling it.
6164

6165
    """
6166
    try:
6167
      self.re = re.compile(self.op.pattern)
6168
    except re.error, err:
6169
      raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
6170
                                 (self.op.pattern, err))
6171

    
6172
  def Exec(self, feedback_fn):
6173
    """Returns the tag list.
6174

6175
    """
6176
    cfg = self.cfg
6177
    tgts = [("/cluster", cfg.GetClusterInfo())]
6178
    ilist = cfg.GetAllInstancesInfo().values()
6179
    tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
6180
    nlist = cfg.GetAllNodesInfo().values()
6181
    tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
6182
    results = []
6183
    for path, target in tgts:
6184
      for tag in target.GetTags():
6185
        if self.re.search(tag):
6186
          results.append((path, tag))
6187
    return results
6188

    
6189

    
6190
class LUAddTags(TagsLU):
6191
  """Sets a tag on a given object.
6192

6193
  """
6194
  _OP_REQP = ["kind", "name", "tags"]
6195
  REQ_BGL = False
6196

    
6197
  def CheckPrereq(self):
6198
    """Check prerequisites.
6199

6200
    This checks the type and length of the tag name and value.
6201

6202
    """
6203
    TagsLU.CheckPrereq(self)
6204
    for tag in self.op.tags:
6205
      objects.TaggableObject.ValidateTag(tag)
6206

    
6207
  def Exec(self, feedback_fn):
6208
    """Sets the tag.
6209

6210
    """
6211
    try:
6212
      for tag in self.op.tags:
6213
        self.target.AddTag(tag)
6214
    except errors.TagError, err:
6215
      raise errors.OpExecError("Error while setting tag: %s" % str(err))
6216
    try:
6217
      self.cfg.Update(self.target)
6218
    except errors.ConfigurationError:
6219
      raise errors.OpRetryError("There has been a modification to the"
6220
                                " config file and the operation has been"
6221
                                " aborted. Please retry.")
6222

    
6223

    
6224
class LUDelTags(TagsLU):
6225
  """Delete a list of tags from a given object.
6226

6227
  """
6228
  _OP_REQP = ["kind", "name", "tags"]
6229
  REQ_BGL = False
6230

    
6231
  def CheckPrereq(self):
6232
    """Check prerequisites.
6233

6234
    This checks that we have the given tag.
6235

6236
    """
6237
    TagsLU.CheckPrereq(self)
6238
    for tag in self.op.tags:
6239
      objects.TaggableObject.ValidateTag(tag)
6240
    del_tags = frozenset(self.op.tags)
6241
    cur_tags = self.target.GetTags()
6242
    if not del_tags <= cur_tags:
6243
      diff_tags = del_tags - cur_tags
6244
      diff_names = ["'%s'" % tag for tag in diff_tags]
6245
      diff_names.sort()
6246
      raise errors.OpPrereqError("Tag(s) %s not found" %
6247
                                 (",".join(diff_names)))
6248

    
6249
  def Exec(self, feedback_fn):
6250
    """Remove the tag from the object.
6251

6252
    """
6253
    for tag in self.op.tags:
6254
      self.target.RemoveTag(tag)
6255
    try:
6256
      self.cfg.Update(self.target)
6257
    except errors.ConfigurationError:
6258
      raise errors.OpRetryError("There has been a modification to the"
6259
                                " config file and the operation has been"
6260
                                " aborted. Please retry.")
6261

    
6262

    
6263
class LUTestDelay(NoHooksLU):
6264
  """Sleep for a specified amount of time.
6265

6266
  This LU sleeps on the master and/or nodes for a specified amount of
6267
  time.
6268

6269
  """
6270
  _OP_REQP = ["duration", "on_master", "on_nodes"]
6271
  REQ_BGL = False
6272

    
6273
  def ExpandNames(self):
6274
    """Expand names and set required locks.
6275

6276
    This expands the node list, if any.
6277

6278
    """
6279
    self.needed_locks = {}
6280
    if self.op.on_nodes:
6281
      # _GetWantedNodes can be used here, but is not always appropriate to use
6282
      # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
6283
      # more information.
6284
      self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
6285
      self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
6286

    
6287
  def CheckPrereq(self):
6288
    """Check prerequisites.
6289

6290
    """
6291

    
6292
  def Exec(self, feedback_fn):
6293
    """Do the actual sleep.
6294

6295
    """
6296
    if self.op.on_master:
6297
      if not utils.TestDelay(self.op.duration):
6298
        raise errors.OpExecError("Error during master delay test")
6299
    if self.op.on_nodes:
6300
      result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
6301
      if not result:
6302
        raise errors.OpExecError("Complete failure from rpc call")
6303
      for node, node_result in result.items():
6304
        node_result.Raise()
6305
        if not node_result.data:
6306
          raise errors.OpExecError("Failure during rpc call to node %s,"
6307
                                   " result: %s" % (node, node_result.data))
6308

    
6309

    
6310
class IAllocator(object):
6311
  """IAllocator framework.
6312

6313
  An IAllocator instance has three sets of attributes:
6314
    - cfg that is needed to query the cluster
6315
    - input data (all members of the _KEYS class attribute are required)
6316
    - four buffer attributes (in|out_data|text), that represent the
6317
      input (to the external script) in text and data structure format,
6318
      and the output from it, again in two formats
6319
    - the result variables from the script (success, info, nodes) for
6320
      easy usage
6321

6322
  """
6323
  _ALLO_KEYS = [
6324
    "mem_size", "disks", "disk_template",
6325
    "os", "tags", "nics", "vcpus", "hypervisor",
6326
    ]
6327
  _RELO_KEYS = [
6328
    "relocate_from",
6329
    ]
6330

    
6331
  def __init__(self, lu, mode, name, **kwargs):
6332
    self.lu = lu
6333
    # init buffer variables
6334
    self.in_text = self.out_text = self.in_data = self.out_data = None
6335
    # init all input fields so that pylint is happy
6336
    self.mode = mode
6337
    self.name = name
6338
    self.mem_size = self.disks = self.disk_template = None
6339
    self.os = self.tags = self.nics = self.vcpus = None
6340
    self.hypervisor = None
6341
    self.relocate_from = None
6342
    # computed fields
6343
    self.required_nodes = None
6344
    # init result fields
6345
    self.success = self.info = self.nodes = None
6346
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6347
      keyset = self._ALLO_KEYS
6348
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
6349
      keyset = self._RELO_KEYS
6350
    else:
6351
      raise errors.ProgrammerError("Unknown mode '%s' passed to the"
6352
                                   " IAllocator" % self.mode)
6353
    for key in kwargs:
6354
      if key not in keyset:
6355
        raise errors.ProgrammerError("Invalid input parameter '%s' to"
6356
                                     " IAllocator" % key)
6357
      setattr(self, key, kwargs[key])
6358
    for key in keyset:
6359
      if key not in kwargs:
6360
        raise errors.ProgrammerError("Missing input parameter '%s' to"
6361
                                     " IAllocator" % key)
6362
    self._BuildInputData()
6363

    
6364
  def _ComputeClusterData(self):
6365
    """Compute the generic allocator input data.
6366

6367
    This is the data that is independent of the actual operation.
6368

6369
    """
6370
    cfg = self.lu.cfg
6371
    cluster_info = cfg.GetClusterInfo()
6372
    # cluster data
6373
    data = {
6374
      "version": 1,
6375
      "cluster_name": cfg.GetClusterName(),
6376
      "cluster_tags": list(cluster_info.GetTags()),
6377
      "enable_hypervisors": list(cluster_info.enabled_hypervisors),
6378
      # we don't have job IDs
6379
      }
6380
    iinfo = cfg.GetAllInstancesInfo().values()
6381
    i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
6382

    
6383
    # node data
6384
    node_results = {}
6385
    node_list = cfg.GetNodeList()
6386

    
6387
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6388
      hypervisor_name = self.hypervisor
6389
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
6390
      hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
6391

    
6392
    node_data = self.lu.rpc.call_node_info(node_list, cfg.GetVGName(),
6393
                                           hypervisor_name)
6394
    node_iinfo = self.lu.rpc.call_all_instances_info(node_list,
6395
                       cluster_info.enabled_hypervisors)
6396
    for nname in node_list:
6397
      ninfo = cfg.GetNodeInfo(nname)
6398
      node_data[nname].Raise()
6399
      if not isinstance(node_data[nname].data, dict):
6400
        raise errors.OpExecError("Can't get data for node %s" % nname)
6401
      remote_info = node_data[nname].data
6402
      for attr in ['memory_total', 'memory_free', 'memory_dom0',
6403
                   'vg_size', 'vg_free', 'cpu_total']:
6404
        if attr not in remote_info:
6405
          raise errors.OpExecError("Node '%s' didn't return attribute '%s'" %
6406
                                   (nname, attr))
6407
        try:
6408
          remote_info[attr] = int(remote_info[attr])
6409
        except ValueError, err:
6410
          raise errors.OpExecError("Node '%s' returned invalid value for '%s':"
6411
                                   " %s" % (nname, attr, str(err)))
6412
      # compute memory used by primary instances
6413
      i_p_mem = i_p_up_mem = 0
6414
      for iinfo, beinfo in i_list:
6415
        if iinfo.primary_node == nname:
6416
          i_p_mem += beinfo[constants.BE_MEMORY]
6417
          if iinfo.name not in node_iinfo[nname]:
6418
            i_used_mem = 0
6419
          else:
6420
            i_used_mem = int(node_iinfo[nname][iinfo.name]['memory'])
6421
          i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
6422
          remote_info['memory_free'] -= max(0, i_mem_diff)
6423

    
6424
          if iinfo.admin_up:
6425
            i_p_up_mem += beinfo[constants.BE_MEMORY]
6426

    
6427
      # compute memory used by instances
6428
      pnr = {
6429
        "tags": list(ninfo.GetTags()),
6430
        "total_memory": remote_info['memory_total'],
6431
        "reserved_memory": remote_info['memory_dom0'],
6432
        "free_memory": remote_info['memory_free'],
6433
        "i_pri_memory": i_p_mem,
6434
        "i_pri_up_memory": i_p_up_mem,
6435
        "total_disk": remote_info['vg_size'],
6436
        "free_disk": remote_info['vg_free'],
6437
        "primary_ip": ninfo.primary_ip,
6438
        "secondary_ip": ninfo.secondary_ip,
6439
        "total_cpus": remote_info['cpu_total'],
6440
        "offline": ninfo.offline,
6441
        }
6442
      node_results[nname] = pnr
6443
    data["nodes"] = node_results
6444

    
6445
    # instance data
6446
    instance_data = {}
6447
    for iinfo, beinfo in i_list:
6448
      nic_data = [{"mac": n.mac, "ip": n.ip, "bridge": n.bridge}
6449
                  for n in iinfo.nics]
6450
      pir = {
6451
        "tags": list(iinfo.GetTags()),
6452
        "should_run": iinfo.admin_up,
6453
        "vcpus": beinfo[constants.BE_VCPUS],
6454
        "memory": beinfo[constants.BE_MEMORY],
6455
        "os": iinfo.os,
6456
        "nodes": list(iinfo.all_nodes),
6457
        "nics": nic_data,
6458
        "disks": [{"size": dsk.size, "mode": "w"} for dsk in iinfo.disks],
6459
        "disk_template": iinfo.disk_template,
6460
        "hypervisor": iinfo.hypervisor,
6461
        }
6462
      instance_data[iinfo.name] = pir
6463

    
6464
    data["instances"] = instance_data
6465

    
6466
    self.in_data = data
6467

    
6468
  def _AddNewInstance(self):
6469
    """Add new instance data to allocator structure.
6470

6471
    This in combination with _AllocatorGetClusterData will create the
6472
    correct structure needed as input for the allocator.
6473

6474
    The checks for the completeness of the opcode must have already been
6475
    done.
6476

6477
    """
6478
    data = self.in_data
6479

    
6480
    disk_space = _ComputeDiskSize(self.disk_template, self.disks)
6481

    
6482
    if self.disk_template in constants.DTS_NET_MIRROR:
6483
      self.required_nodes = 2
6484
    else:
6485
      self.required_nodes = 1
6486
    request = {
6487
      "type": "allocate",
6488
      "name": self.name,
6489
      "disk_template": self.disk_template,
6490
      "tags": self.tags,
6491
      "os": self.os,
6492
      "vcpus": self.vcpus,
6493
      "memory": self.mem_size,
6494
      "disks": self.disks,
6495
      "disk_space_total": disk_space,
6496
      "nics": self.nics,
6497
      "required_nodes": self.required_nodes,
6498
      }
6499
    data["request"] = request
6500

    
6501
  def _AddRelocateInstance(self):
6502
    """Add relocate instance data to allocator structure.
6503

6504
    This in combination with _IAllocatorGetClusterData will create the
6505
    correct structure needed as input for the allocator.
6506

6507
    The checks for the completeness of the opcode must have already been
6508
    done.
6509

6510
    """
6511
    instance = self.lu.cfg.GetInstanceInfo(self.name)
6512
    if instance is None:
6513
      raise errors.ProgrammerError("Unknown instance '%s' passed to"
6514
                                   " IAllocator" % self.name)
6515

    
6516
    if instance.disk_template not in constants.DTS_NET_MIRROR:
6517
      raise errors.OpPrereqError("Can't relocate non-mirrored instances")
6518

    
6519
    if len(instance.secondary_nodes) != 1:
6520
      raise errors.OpPrereqError("Instance has not exactly one secondary node")
6521

    
6522
    self.required_nodes = 1
6523
    disk_sizes = [{'size': disk.size} for disk in instance.disks]
6524
    disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
6525

    
6526
    request = {
6527
      "type": "relocate",
6528
      "name": self.name,
6529
      "disk_space_total": disk_space,
6530
      "required_nodes": self.required_nodes,
6531
      "relocate_from": self.relocate_from,
6532
      }
6533
    self.in_data["request"] = request
6534

    
6535
  def _BuildInputData(self):
6536
    """Build input data structures.
6537

6538
    """
6539
    self._ComputeClusterData()
6540

    
6541
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6542
      self._AddNewInstance()
6543
    else:
6544
      self._AddRelocateInstance()
6545

    
6546
    self.in_text = serializer.Dump(self.in_data)
6547

    
6548
  def Run(self, name, validate=True, call_fn=None):
6549
    """Run an instance allocator and return the results.
6550

6551
    """
6552
    if call_fn is None:
6553
      call_fn = self.lu.rpc.call_iallocator_runner
6554
    data = self.in_text
6555

    
6556
    result = call_fn(self.lu.cfg.GetMasterNode(), name, self.in_text)
6557
    result.Raise()
6558

    
6559
    if not isinstance(result.data, (list, tuple)) or len(result.data) != 4:
6560
      raise errors.OpExecError("Invalid result from master iallocator runner")
6561

    
6562
    rcode, stdout, stderr, fail = result.data
6563

    
6564
    if rcode == constants.IARUN_NOTFOUND:
6565
      raise errors.OpExecError("Can't find allocator '%s'" % name)
6566
    elif rcode == constants.IARUN_FAILURE:
6567
      raise errors.OpExecError("Instance allocator call failed: %s,"
6568
                               " output: %s" % (fail, stdout+stderr))
6569
    self.out_text = stdout
6570
    if validate:
6571
      self._ValidateResult()
6572

    
6573
  def _ValidateResult(self):
6574
    """Process the allocator results.
6575

6576
    This will process and if successful save the result in
6577
    self.out_data and the other parameters.
6578

6579
    """
6580
    try:
6581
      rdict = serializer.Load(self.out_text)
6582
    except Exception, err:
6583
      raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
6584

    
6585
    if not isinstance(rdict, dict):
6586
      raise errors.OpExecError("Can't parse iallocator results: not a dict")
6587

    
6588
    for key in "success", "info", "nodes":
6589
      if key not in rdict:
6590
        raise errors.OpExecError("Can't parse iallocator results:"
6591
                                 " missing key '%s'" % key)
6592
      setattr(self, key, rdict[key])
6593

    
6594
    if not isinstance(rdict["nodes"], list):
6595
      raise errors.OpExecError("Can't parse iallocator results: 'nodes' key"
6596
                               " is not a list")
6597
    self.out_data = rdict
6598

    
6599

    
6600
class LUTestAllocator(NoHooksLU):
6601
  """Run allocator tests.
6602

6603
  This LU runs the allocator tests
6604

6605
  """
6606
  _OP_REQP = ["direction", "mode", "name"]
6607

    
6608
  def CheckPrereq(self):
6609
    """Check prerequisites.
6610

6611
    This checks the opcode parameters depending on the director and mode test.
6612

6613
    """
6614
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
6615
      for attr in ["name", "mem_size", "disks", "disk_template",
6616
                   "os", "tags", "nics", "vcpus"]:
6617
        if not hasattr(self.op, attr):
6618
          raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
6619
                                     attr)
6620
      iname = self.cfg.ExpandInstanceName(self.op.name)
6621
      if iname is not None:
6622
        raise errors.OpPrereqError("Instance '%s' already in the cluster" %
6623
                                   iname)
6624
      if not isinstance(self.op.nics, list):
6625
        raise errors.OpPrereqError("Invalid parameter 'nics'")
6626
      for row in self.op.nics:
6627
        if (not isinstance(row, dict) or
6628
            "mac" not in row or
6629
            "ip" not in row or
6630
            "bridge" not in row):
6631
          raise errors.OpPrereqError("Invalid contents of the"
6632
                                     " 'nics' parameter")
6633
      if not isinstance(self.op.disks, list):
6634
        raise errors.OpPrereqError("Invalid parameter 'disks'")
6635
      for row in self.op.disks:
6636
        if (not isinstance(row, dict) or
6637
            "size" not in row or
6638
            not isinstance(row["size"], int) or
6639
            "mode" not in row or
6640
            row["mode"] not in ['r', 'w']):
6641
          raise errors.OpPrereqError("Invalid contents of the"
6642
                                     " 'disks' parameter")
6643
      if not hasattr(self.op, "hypervisor") or self.op.hypervisor is None:
6644
        self.op.hypervisor = self.cfg.GetHypervisorType()
6645
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
6646
      if not hasattr(self.op, "name"):
6647
        raise errors.OpPrereqError("Missing attribute 'name' on opcode input")
6648
      fname = self.cfg.ExpandInstanceName(self.op.name)
6649
      if fname is None:
6650
        raise errors.OpPrereqError("Instance '%s' not found for relocation" %
6651
                                   self.op.name)
6652
      self.op.name = fname
6653
      self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
6654
    else:
6655
      raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
6656
                                 self.op.mode)
6657

    
6658
    if self.op.direction == constants.IALLOCATOR_DIR_OUT:
6659
      if not hasattr(self.op, "allocator") or self.op.allocator is None:
6660
        raise errors.OpPrereqError("Missing allocator name")
6661
    elif self.op.direction != constants.IALLOCATOR_DIR_IN:
6662
      raise errors.OpPrereqError("Wrong allocator test '%s'" %
6663
                                 self.op.direction)
6664

    
6665
  def Exec(self, feedback_fn):
6666
    """Run the allocator test.
6667

6668
    """
6669
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
6670
      ial = IAllocator(self,
6671
                       mode=self.op.mode,
6672
                       name=self.op.name,
6673
                       mem_size=self.op.mem_size,
6674
                       disks=self.op.disks,
6675
                       disk_template=self.op.disk_template,
6676
                       os=self.op.os,
6677
                       tags=self.op.tags,
6678
                       nics=self.op.nics,
6679
                       vcpus=self.op.vcpus,
6680
                       hypervisor=self.op.hypervisor,
6681
                       )
6682
    else:
6683
      ial = IAllocator(self,
6684
                       mode=self.op.mode,
6685
                       name=self.op.name,
6686
                       relocate_from=list(self.relocate_from),
6687
                       )
6688

    
6689
    if self.op.direction == constants.IALLOCATOR_DIR_IN:
6690
      result = ial.in_text
6691
    else:
6692
      ial.Run(self.op.allocator, validate=False)
6693
      result = ial.out_text
6694
    return result