Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ 53c14ef1

History | View | Annotate | Download (234.4 kB)

1
#
2
#
3

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

    
21

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

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

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

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

    
48

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

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

61
  Note that all commands require root permissions.
62

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

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

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

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

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

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

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

    
109
  ssh = property(fget=__GetSSH)
110

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

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

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

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

126
    """
127
    pass
128

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

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

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

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

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

150
    Examples::
151

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

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

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

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

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

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

189
    """
190

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

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

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

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

205
    """
206
    raise NotImplementedError
207

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

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

215
    """
216
    raise NotImplementedError
217

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

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

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

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

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

237
    """
238
    raise NotImplementedError
239

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

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

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

258
    """
259
    return lu_result
260

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
326
    del self.recalculate_locks[locking.LEVEL_NODE]
327

    
328

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

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

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

    
339

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

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

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

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

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

    
366
  return utils.NiceSort(wanted)
367

    
368

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

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

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

    
385
  if instances:
386
    wanted = []
387

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

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

    
398

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

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

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

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

    
417

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

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

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

    
431

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

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

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

    
443

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

448
  This builds the hook environment from individual variables.
449

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

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

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

    
497
  env["INSTANCE_NIC_COUNT"] = nic_count
498

    
499
  return env
500

    
501

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

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

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

    
532

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

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

    
548

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

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

    
562

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

566
  """
567
  _OP_REQP = []
568

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

572
    This checks whether the cluster is empty.
573

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

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

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

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

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

    
602

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

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

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

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

624
    Test list:
625

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

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

642
    """
643
    node = nodeinfo.name
644

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

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

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

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

    
665
    bad = False
666

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

    
673
    # checks vg existence and size > 20G
674

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

    
687
    # checks config file checksum
688

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

    
716
    # checks ssh to any
717

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

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

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

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

    
758
    return bad
759

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

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

767
    """
768
    bad = False
769

    
770
    node_current = instanceconfig.primary_node
771

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

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

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

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

    
800
    return bad
801

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

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

808
    """
809
    bad = False
810

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

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

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

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

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

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

840
    """
841
    bad = False
842

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

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

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

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

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

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

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

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

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

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

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

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

    
920
    local_checksums = utils.FingerprintFiles(file_names)
921

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

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

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

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

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

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

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

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

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

    
997
      node_instance[node] = idata
998

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

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

    
1025
    node_vol_should = {}
1026

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

    
1035
      inst_config.MapLVsByNode(node_vol_should)
1036

    
1037
      instance_cfg[instance] = inst_config
1038

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

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

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

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

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

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

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

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

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

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

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

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

    
1110
    return not bad
1111

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

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

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

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

    
1159
      return lu_result
1160

    
1161

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

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

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

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

1179
    This has no prerequisites.
1180

1181
    """
1182
    pass
1183

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

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

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

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

    
1207
    if not nv_dict:
1208
      return result
1209

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

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

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

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

    
1244
    return result
1245

    
1246

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

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

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

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

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

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

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

    
1285
    self.op.name = new_name
1286

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

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

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

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

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

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

    
1326

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

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

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

    
1342

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
1480

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

1484
  This is a very simple LU.
1485

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

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

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

1499
    """
1500

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

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

    
1507

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

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

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

    
1518
  node = instance.primary_node
1519

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

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

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

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

    
1565

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

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

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

    
1580
  result = True
1581
  if on_primary or dev.AssembleOnSecondary():
1582
    rstats = lu.rpc.call_blockdev_find(node, dev)
1583
    msg = rstats.RemoteFailMsg()
1584
    if msg:
1585
      lu.LogWarning("Can't find disk on node %s: %s", node, msg)
1586
      result = False
1587
    elif not rstats.payload:
1588
      lu.LogWarning("Can't find disk on node %s", node)
1589
      result = False
1590
    else:
1591
      result = result and (not rstats.payload[idx])
1592
  if dev.children:
1593
    for child in dev.children:
1594
      result = result and _CheckDiskConsistency(lu, child, node, on_primary)
1595

    
1596
  return result
1597

    
1598

    
1599
class LUDiagnoseOS(NoHooksLU):
1600
  """Logical unit for OS diagnose/query.
1601

1602
  """
1603
  _OP_REQP = ["output_fields", "names"]
1604
  REQ_BGL = False
1605
  _FIELDS_STATIC = utils.FieldSet()
1606
  _FIELDS_DYNAMIC = utils.FieldSet("name", "valid", "node_status")
1607

    
1608
  def ExpandNames(self):
1609
    if self.op.names:
1610
      raise errors.OpPrereqError("Selective OS query not supported")
1611

    
1612
    _CheckOutputFields(static=self._FIELDS_STATIC,
1613
                       dynamic=self._FIELDS_DYNAMIC,
1614
                       selected=self.op.output_fields)
1615

    
1616
    # Lock all nodes, in shared mode
1617
    self.needed_locks = {}
1618
    self.share_locks[locking.LEVEL_NODE] = 1
1619
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
1620

    
1621
  def CheckPrereq(self):
1622
    """Check prerequisites.
1623

1624
    """
1625

    
1626
  @staticmethod
1627
  def _DiagnoseByOS(node_list, rlist):
1628
    """Remaps a per-node return list into an a per-os per-node dictionary
1629

1630
    @param node_list: a list with the names of all nodes
1631
    @param rlist: a map with node names as keys and OS objects as values
1632

1633
    @rtype: dict
1634
    @returns: a dictionary with osnames as keys and as value another map, with
1635
        nodes as keys and list of OS objects as values, eg::
1636

1637
          {"debian-etch": {"node1": [<object>,...],
1638
                           "node2": [<object>,]}
1639
          }
1640

1641
    """
1642
    all_os = {}
1643
    for node_name, nr in rlist.iteritems():
1644
      if nr.failed or not nr.data:
1645
        continue
1646
      for os_obj in nr.data:
1647
        if os_obj.name not in all_os:
1648
          # build a list of nodes for this os containing empty lists
1649
          # for each node in node_list
1650
          all_os[os_obj.name] = {}
1651
          for nname in node_list:
1652
            all_os[os_obj.name][nname] = []
1653
        all_os[os_obj.name][node_name].append(os_obj)
1654
    return all_os
1655

    
1656
  def Exec(self, feedback_fn):
1657
    """Compute the list of OSes.
1658

1659
    """
1660
    node_list = self.acquired_locks[locking.LEVEL_NODE]
1661
    valid_nodes = [node for node in self.cfg.GetOnlineNodeList()
1662
                   if node in node_list]
1663
    node_data = self.rpc.call_os_diagnose(valid_nodes)
1664
    if node_data == False:
1665
      raise errors.OpExecError("Can't gather the list of OSes")
1666
    pol = self._DiagnoseByOS(valid_nodes, node_data)
1667
    output = []
1668
    for os_name, os_data in pol.iteritems():
1669
      row = []
1670
      for field in self.op.output_fields:
1671
        if field == "name":
1672
          val = os_name
1673
        elif field == "valid":
1674
          val = utils.all([osl and osl[0] for osl in os_data.values()])
1675
        elif field == "node_status":
1676
          val = {}
1677
          for node_name, nos_list in os_data.iteritems():
1678
            val[node_name] = [(v.status, v.path) for v in nos_list]
1679
        else:
1680
          raise errors.ParameterError(field)
1681
        row.append(val)
1682
      output.append(row)
1683

    
1684
    return output
1685

    
1686

    
1687
class LURemoveNode(LogicalUnit):
1688
  """Logical unit for removing a node.
1689

1690
  """
1691
  HPATH = "node-remove"
1692
  HTYPE = constants.HTYPE_NODE
1693
  _OP_REQP = ["node_name"]
1694

    
1695
  def BuildHooksEnv(self):
1696
    """Build hooks env.
1697

1698
    This doesn't run on the target node in the pre phase as a failed
1699
    node would then be impossible to remove.
1700

1701
    """
1702
    env = {
1703
      "OP_TARGET": self.op.node_name,
1704
      "NODE_NAME": self.op.node_name,
1705
      }
1706
    all_nodes = self.cfg.GetNodeList()
1707
    all_nodes.remove(self.op.node_name)
1708
    return env, all_nodes, all_nodes
1709

    
1710
  def CheckPrereq(self):
1711
    """Check prerequisites.
1712

1713
    This checks:
1714
     - the node exists in the configuration
1715
     - it does not have primary or secondary instances
1716
     - it's not the master
1717

1718
    Any errors are signalled by raising errors.OpPrereqError.
1719

1720
    """
1721
    node = self.cfg.GetNodeInfo(self.cfg.ExpandNodeName(self.op.node_name))
1722
    if node is None:
1723
      raise errors.OpPrereqError, ("Node '%s' is unknown." % self.op.node_name)
1724

    
1725
    instance_list = self.cfg.GetInstanceList()
1726

    
1727
    masternode = self.cfg.GetMasterNode()
1728
    if node.name == masternode:
1729
      raise errors.OpPrereqError("Node is the master node,"
1730
                                 " you need to failover first.")
1731

    
1732
    for instance_name in instance_list:
1733
      instance = self.cfg.GetInstanceInfo(instance_name)
1734
      if node.name in instance.all_nodes:
1735
        raise errors.OpPrereqError("Instance %s is still running on the node,"
1736
                                   " please remove first." % instance_name)
1737
    self.op.node_name = node.name
1738
    self.node = node
1739

    
1740
  def Exec(self, feedback_fn):
1741
    """Removes the node from the cluster.
1742

1743
    """
1744
    node = self.node
1745
    logging.info("Stopping the node daemon and removing configs from node %s",
1746
                 node.name)
1747

    
1748
    self.context.RemoveNode(node.name)
1749

    
1750
    self.rpc.call_node_leave_cluster(node.name)
1751

    
1752
    # Promote nodes to master candidate as needed
1753
    _AdjustCandidatePool(self)
1754

    
1755

    
1756
class LUQueryNodes(NoHooksLU):
1757
  """Logical unit for querying nodes.
1758

1759
  """
1760
  _OP_REQP = ["output_fields", "names", "use_locking"]
1761
  REQ_BGL = False
1762
  _FIELDS_DYNAMIC = utils.FieldSet(
1763
    "dtotal", "dfree",
1764
    "mtotal", "mnode", "mfree",
1765
    "bootid",
1766
    "ctotal", "cnodes", "csockets",
1767
    )
1768

    
1769
  _FIELDS_STATIC = utils.FieldSet(
1770
    "name", "pinst_cnt", "sinst_cnt",
1771
    "pinst_list", "sinst_list",
1772
    "pip", "sip", "tags",
1773
    "serial_no",
1774
    "master_candidate",
1775
    "master",
1776
    "offline",
1777
    )
1778

    
1779
  def ExpandNames(self):
1780
    _CheckOutputFields(static=self._FIELDS_STATIC,
1781
                       dynamic=self._FIELDS_DYNAMIC,
1782
                       selected=self.op.output_fields)
1783

    
1784
    self.needed_locks = {}
1785
    self.share_locks[locking.LEVEL_NODE] = 1
1786

    
1787
    if self.op.names:
1788
      self.wanted = _GetWantedNodes(self, self.op.names)
1789
    else:
1790
      self.wanted = locking.ALL_SET
1791

    
1792
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
1793
    self.do_locking = self.do_node_query and self.op.use_locking
1794
    if self.do_locking:
1795
      # if we don't request only static fields, we need to lock the nodes
1796
      self.needed_locks[locking.LEVEL_NODE] = self.wanted
1797

    
1798

    
1799
  def CheckPrereq(self):
1800
    """Check prerequisites.
1801

1802
    """
1803
    # The validation of the node list is done in the _GetWantedNodes,
1804
    # if non empty, and if empty, there's no validation to do
1805
    pass
1806

    
1807
  def Exec(self, feedback_fn):
1808
    """Computes the list of nodes and their attributes.
1809

1810
    """
1811
    all_info = self.cfg.GetAllNodesInfo()
1812
    if self.do_locking:
1813
      nodenames = self.acquired_locks[locking.LEVEL_NODE]
1814
    elif self.wanted != locking.ALL_SET:
1815
      nodenames = self.wanted
1816
      missing = set(nodenames).difference(all_info.keys())
1817
      if missing:
1818
        raise errors.OpExecError(
1819
          "Some nodes were removed before retrieving their data: %s" % missing)
1820
    else:
1821
      nodenames = all_info.keys()
1822

    
1823
    nodenames = utils.NiceSort(nodenames)
1824
    nodelist = [all_info[name] for name in nodenames]
1825

    
1826
    # begin data gathering
1827

    
1828
    if self.do_node_query:
1829
      live_data = {}
1830
      node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
1831
                                          self.cfg.GetHypervisorType())
1832
      for name in nodenames:
1833
        nodeinfo = node_data[name]
1834
        if not nodeinfo.failed and nodeinfo.data:
1835
          nodeinfo = nodeinfo.data
1836
          fn = utils.TryConvert
1837
          live_data[name] = {
1838
            "mtotal": fn(int, nodeinfo.get('memory_total', None)),
1839
            "mnode": fn(int, nodeinfo.get('memory_dom0', None)),
1840
            "mfree": fn(int, nodeinfo.get('memory_free', None)),
1841
            "dtotal": fn(int, nodeinfo.get('vg_size', None)),
1842
            "dfree": fn(int, nodeinfo.get('vg_free', None)),
1843
            "ctotal": fn(int, nodeinfo.get('cpu_total', None)),
1844
            "bootid": nodeinfo.get('bootid', None),
1845
            "cnodes": fn(int, nodeinfo.get('cpu_nodes', None)),
1846
            "csockets": fn(int, nodeinfo.get('cpu_sockets', None)),
1847
            }
1848
        else:
1849
          live_data[name] = {}
1850
    else:
1851
      live_data = dict.fromkeys(nodenames, {})
1852

    
1853
    node_to_primary = dict([(name, set()) for name in nodenames])
1854
    node_to_secondary = dict([(name, set()) for name in nodenames])
1855

    
1856
    inst_fields = frozenset(("pinst_cnt", "pinst_list",
1857
                             "sinst_cnt", "sinst_list"))
1858
    if inst_fields & frozenset(self.op.output_fields):
1859
      instancelist = self.cfg.GetInstanceList()
1860

    
1861
      for instance_name in instancelist:
1862
        inst = self.cfg.GetInstanceInfo(instance_name)
1863
        if inst.primary_node in node_to_primary:
1864
          node_to_primary[inst.primary_node].add(inst.name)
1865
        for secnode in inst.secondary_nodes:
1866
          if secnode in node_to_secondary:
1867
            node_to_secondary[secnode].add(inst.name)
1868

    
1869
    master_node = self.cfg.GetMasterNode()
1870

    
1871
    # end data gathering
1872

    
1873
    output = []
1874
    for node in nodelist:
1875
      node_output = []
1876
      for field in self.op.output_fields:
1877
        if field == "name":
1878
          val = node.name
1879
        elif field == "pinst_list":
1880
          val = list(node_to_primary[node.name])
1881
        elif field == "sinst_list":
1882
          val = list(node_to_secondary[node.name])
1883
        elif field == "pinst_cnt":
1884
          val = len(node_to_primary[node.name])
1885
        elif field == "sinst_cnt":
1886
          val = len(node_to_secondary[node.name])
1887
        elif field == "pip":
1888
          val = node.primary_ip
1889
        elif field == "sip":
1890
          val = node.secondary_ip
1891
        elif field == "tags":
1892
          val = list(node.GetTags())
1893
        elif field == "serial_no":
1894
          val = node.serial_no
1895
        elif field == "master_candidate":
1896
          val = node.master_candidate
1897
        elif field == "master":
1898
          val = node.name == master_node
1899
        elif field == "offline":
1900
          val = node.offline
1901
        elif self._FIELDS_DYNAMIC.Matches(field):
1902
          val = live_data[node.name].get(field, None)
1903
        else:
1904
          raise errors.ParameterError(field)
1905
        node_output.append(val)
1906
      output.append(node_output)
1907

    
1908
    return output
1909

    
1910

    
1911
class LUQueryNodeVolumes(NoHooksLU):
1912
  """Logical unit for getting volumes on node(s).
1913

1914
  """
1915
  _OP_REQP = ["nodes", "output_fields"]
1916
  REQ_BGL = False
1917
  _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
1918
  _FIELDS_STATIC = utils.FieldSet("node")
1919

    
1920
  def ExpandNames(self):
1921
    _CheckOutputFields(static=self._FIELDS_STATIC,
1922
                       dynamic=self._FIELDS_DYNAMIC,
1923
                       selected=self.op.output_fields)
1924

    
1925
    self.needed_locks = {}
1926
    self.share_locks[locking.LEVEL_NODE] = 1
1927
    if not self.op.nodes:
1928
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
1929
    else:
1930
      self.needed_locks[locking.LEVEL_NODE] = \
1931
        _GetWantedNodes(self, self.op.nodes)
1932

    
1933
  def CheckPrereq(self):
1934
    """Check prerequisites.
1935

1936
    This checks that the fields required are valid output fields.
1937

1938
    """
1939
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
1940

    
1941
  def Exec(self, feedback_fn):
1942
    """Computes the list of nodes and their attributes.
1943

1944
    """
1945
    nodenames = self.nodes
1946
    volumes = self.rpc.call_node_volumes(nodenames)
1947

    
1948
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
1949
             in self.cfg.GetInstanceList()]
1950

    
1951
    lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
1952

    
1953
    output = []
1954
    for node in nodenames:
1955
      if node not in volumes or volumes[node].failed or not volumes[node].data:
1956
        continue
1957

    
1958
      node_vols = volumes[node].data[:]
1959
      node_vols.sort(key=lambda vol: vol['dev'])
1960

    
1961
      for vol in node_vols:
1962
        node_output = []
1963
        for field in self.op.output_fields:
1964
          if field == "node":
1965
            val = node
1966
          elif field == "phys":
1967
            val = vol['dev']
1968
          elif field == "vg":
1969
            val = vol['vg']
1970
          elif field == "name":
1971
            val = vol['name']
1972
          elif field == "size":
1973
            val = int(float(vol['size']))
1974
          elif field == "instance":
1975
            for inst in ilist:
1976
              if node not in lv_by_node[inst]:
1977
                continue
1978
              if vol['name'] in lv_by_node[inst][node]:
1979
                val = inst.name
1980
                break
1981
            else:
1982
              val = '-'
1983
          else:
1984
            raise errors.ParameterError(field)
1985
          node_output.append(str(val))
1986

    
1987
        output.append(node_output)
1988

    
1989
    return output
1990

    
1991

    
1992
class LUAddNode(LogicalUnit):
1993
  """Logical unit for adding node to the cluster.
1994

1995
  """
1996
  HPATH = "node-add"
1997
  HTYPE = constants.HTYPE_NODE
1998
  _OP_REQP = ["node_name"]
1999

    
2000
  def BuildHooksEnv(self):
2001
    """Build hooks env.
2002

2003
    This will run on all nodes before, and on all nodes + the new node after.
2004

2005
    """
2006
    env = {
2007
      "OP_TARGET": self.op.node_name,
2008
      "NODE_NAME": self.op.node_name,
2009
      "NODE_PIP": self.op.primary_ip,
2010
      "NODE_SIP": self.op.secondary_ip,
2011
      }
2012
    nodes_0 = self.cfg.GetNodeList()
2013
    nodes_1 = nodes_0 + [self.op.node_name, ]
2014
    return env, nodes_0, nodes_1
2015

    
2016
  def CheckPrereq(self):
2017
    """Check prerequisites.
2018

2019
    This checks:
2020
     - the new node is not already in the config
2021
     - it is resolvable
2022
     - its parameters (single/dual homed) matches the cluster
2023

2024
    Any errors are signalled by raising errors.OpPrereqError.
2025

2026
    """
2027
    node_name = self.op.node_name
2028
    cfg = self.cfg
2029

    
2030
    dns_data = utils.HostInfo(node_name)
2031

    
2032
    node = dns_data.name
2033
    primary_ip = self.op.primary_ip = dns_data.ip
2034
    secondary_ip = getattr(self.op, "secondary_ip", None)
2035
    if secondary_ip is None:
2036
      secondary_ip = primary_ip
2037
    if not utils.IsValidIP(secondary_ip):
2038
      raise errors.OpPrereqError("Invalid secondary IP given")
2039
    self.op.secondary_ip = secondary_ip
2040

    
2041
    node_list = cfg.GetNodeList()
2042
    if not self.op.readd and node in node_list:
2043
      raise errors.OpPrereqError("Node %s is already in the configuration" %
2044
                                 node)
2045
    elif self.op.readd and node not in node_list:
2046
      raise errors.OpPrereqError("Node %s is not in the configuration" % node)
2047

    
2048
    for existing_node_name in node_list:
2049
      existing_node = cfg.GetNodeInfo(existing_node_name)
2050

    
2051
      if self.op.readd and node == existing_node_name:
2052
        if (existing_node.primary_ip != primary_ip or
2053
            existing_node.secondary_ip != secondary_ip):
2054
          raise errors.OpPrereqError("Readded node doesn't have the same IP"
2055
                                     " address configuration as before")
2056
        continue
2057

    
2058
      if (existing_node.primary_ip == primary_ip or
2059
          existing_node.secondary_ip == primary_ip or
2060
          existing_node.primary_ip == secondary_ip or
2061
          existing_node.secondary_ip == secondary_ip):
2062
        raise errors.OpPrereqError("New node ip address(es) conflict with"
2063
                                   " existing node %s" % existing_node.name)
2064

    
2065
    # check that the type of the node (single versus dual homed) is the
2066
    # same as for the master
2067
    myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
2068
    master_singlehomed = myself.secondary_ip == myself.primary_ip
2069
    newbie_singlehomed = secondary_ip == primary_ip
2070
    if master_singlehomed != newbie_singlehomed:
2071
      if master_singlehomed:
2072
        raise errors.OpPrereqError("The master has no private ip but the"
2073
                                   " new node has one")
2074
      else:
2075
        raise errors.OpPrereqError("The master has a private ip but the"
2076
                                   " new node doesn't have one")
2077

    
2078
    # checks reachablity
2079
    if not utils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
2080
      raise errors.OpPrereqError("Node not reachable by ping")
2081

    
2082
    if not newbie_singlehomed:
2083
      # check reachability from my secondary ip to newbie's secondary ip
2084
      if not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
2085
                           source=myself.secondary_ip):
2086
        raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
2087
                                   " based ping to noded port")
2088

    
2089
    cp_size = self.cfg.GetClusterInfo().candidate_pool_size
2090
    mc_now, _ = self.cfg.GetMasterCandidateStats()
2091
    master_candidate = mc_now < cp_size
2092

    
2093
    self.new_node = objects.Node(name=node,
2094
                                 primary_ip=primary_ip,
2095
                                 secondary_ip=secondary_ip,
2096
                                 master_candidate=master_candidate,
2097
                                 offline=False)
2098

    
2099
  def Exec(self, feedback_fn):
2100
    """Adds the new node to the cluster.
2101

2102
    """
2103
    new_node = self.new_node
2104
    node = new_node.name
2105

    
2106
    # check connectivity
2107
    result = self.rpc.call_version([node])[node]
2108
    result.Raise()
2109
    if result.data:
2110
      if constants.PROTOCOL_VERSION == result.data:
2111
        logging.info("Communication to node %s fine, sw version %s match",
2112
                     node, result.data)
2113
      else:
2114
        raise errors.OpExecError("Version mismatch master version %s,"
2115
                                 " node version %s" %
2116
                                 (constants.PROTOCOL_VERSION, result.data))
2117
    else:
2118
      raise errors.OpExecError("Cannot get version from the new node")
2119

    
2120
    # setup ssh on node
2121
    logging.info("Copy ssh key to node %s", node)
2122
    priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
2123
    keyarray = []
2124
    keyfiles = [constants.SSH_HOST_DSA_PRIV, constants.SSH_HOST_DSA_PUB,
2125
                constants.SSH_HOST_RSA_PRIV, constants.SSH_HOST_RSA_PUB,
2126
                priv_key, pub_key]
2127

    
2128
    for i in keyfiles:
2129
      f = open(i, 'r')
2130
      try:
2131
        keyarray.append(f.read())
2132
      finally:
2133
        f.close()
2134

    
2135
    result = self.rpc.call_node_add(node, keyarray[0], keyarray[1],
2136
                                    keyarray[2],
2137
                                    keyarray[3], keyarray[4], keyarray[5])
2138

    
2139
    msg = result.RemoteFailMsg()
2140
    if msg:
2141
      raise errors.OpExecError("Cannot transfer ssh keys to the"
2142
                               " new node: %s" % msg)
2143

    
2144
    # Add node to our /etc/hosts, and add key to known_hosts
2145
    utils.AddHostToEtcHosts(new_node.name)
2146

    
2147
    if new_node.secondary_ip != new_node.primary_ip:
2148
      result = self.rpc.call_node_has_ip_address(new_node.name,
2149
                                                 new_node.secondary_ip)
2150
      if result.failed or not result.data:
2151
        raise errors.OpExecError("Node claims it doesn't have the secondary ip"
2152
                                 " you gave (%s). Please fix and re-run this"
2153
                                 " command." % new_node.secondary_ip)
2154

    
2155
    node_verify_list = [self.cfg.GetMasterNode()]
2156
    node_verify_param = {
2157
      'nodelist': [node],
2158
      # TODO: do a node-net-test as well?
2159
    }
2160

    
2161
    result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
2162
                                       self.cfg.GetClusterName())
2163
    for verifier in node_verify_list:
2164
      if result[verifier].failed or not result[verifier].data:
2165
        raise errors.OpExecError("Cannot communicate with %s's node daemon"
2166
                                 " for remote verification" % verifier)
2167
      if result[verifier].data['nodelist']:
2168
        for failed in result[verifier].data['nodelist']:
2169
          feedback_fn("ssh/hostname verification failed %s -> %s" %
2170
                      (verifier, result[verifier].data['nodelist'][failed]))
2171
        raise errors.OpExecError("ssh/hostname verification failed.")
2172

    
2173
    # Distribute updated /etc/hosts and known_hosts to all nodes,
2174
    # including the node just added
2175
    myself = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
2176
    dist_nodes = self.cfg.GetNodeList()
2177
    if not self.op.readd:
2178
      dist_nodes.append(node)
2179
    if myself.name in dist_nodes:
2180
      dist_nodes.remove(myself.name)
2181

    
2182
    logging.debug("Copying hosts and known_hosts to all nodes")
2183
    for fname in (constants.ETC_HOSTS, constants.SSH_KNOWN_HOSTS_FILE):
2184
      result = self.rpc.call_upload_file(dist_nodes, fname)
2185
      for to_node, to_result in result.iteritems():
2186
        if to_result.failed or not to_result.data:
2187
          logging.error("Copy of file %s to node %s failed", fname, to_node)
2188

    
2189
    to_copy = []
2190
    enabled_hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
2191
    if constants.HTS_USE_VNC.intersection(enabled_hypervisors):
2192
      to_copy.append(constants.VNC_PASSWORD_FILE)
2193

    
2194
    for fname in to_copy:
2195
      result = self.rpc.call_upload_file([node], fname)
2196
      if result[node].failed or not result[node]:
2197
        logging.error("Could not copy file %s to node %s", fname, node)
2198

    
2199
    if self.op.readd:
2200
      self.context.ReaddNode(new_node)
2201
    else:
2202
      self.context.AddNode(new_node)
2203

    
2204

    
2205
class LUSetNodeParams(LogicalUnit):
2206
  """Modifies the parameters of a node.
2207

2208
  """
2209
  HPATH = "node-modify"
2210
  HTYPE = constants.HTYPE_NODE
2211
  _OP_REQP = ["node_name"]
2212
  REQ_BGL = False
2213

    
2214
  def CheckArguments(self):
2215
    node_name = self.cfg.ExpandNodeName(self.op.node_name)
2216
    if node_name is None:
2217
      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name)
2218
    self.op.node_name = node_name
2219
    _CheckBooleanOpField(self.op, 'master_candidate')
2220
    _CheckBooleanOpField(self.op, 'offline')
2221
    if self.op.master_candidate is None and self.op.offline is None:
2222
      raise errors.OpPrereqError("Please pass at least one modification")
2223
    if self.op.offline == True and self.op.master_candidate == True:
2224
      raise errors.OpPrereqError("Can't set the node into offline and"
2225
                                 " master_candidate at the same time")
2226

    
2227
  def ExpandNames(self):
2228
    self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
2229

    
2230
  def BuildHooksEnv(self):
2231
    """Build hooks env.
2232

2233
    This runs on the master node.
2234

2235
    """
2236
    env = {
2237
      "OP_TARGET": self.op.node_name,
2238
      "MASTER_CANDIDATE": str(self.op.master_candidate),
2239
      "OFFLINE": str(self.op.offline),
2240
      }
2241
    nl = [self.cfg.GetMasterNode(),
2242
          self.op.node_name]
2243
    return env, nl, nl
2244

    
2245
  def CheckPrereq(self):
2246
    """Check prerequisites.
2247

2248
    This only checks the instance list against the existing names.
2249

2250
    """
2251
    node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
2252

    
2253
    if ((self.op.master_candidate == False or self.op.offline == True)
2254
        and node.master_candidate):
2255
      # we will demote the node from master_candidate
2256
      if self.op.node_name == self.cfg.GetMasterNode():
2257
        raise errors.OpPrereqError("The master node has to be a"
2258
                                   " master candidate and online")
2259
      cp_size = self.cfg.GetClusterInfo().candidate_pool_size
2260
      num_candidates, _ = self.cfg.GetMasterCandidateStats()
2261
      if num_candidates <= cp_size:
2262
        msg = ("Not enough master candidates (desired"
2263
               " %d, new value will be %d)" % (cp_size, num_candidates-1))
2264
        if self.op.force:
2265
          self.LogWarning(msg)
2266
        else:
2267
          raise errors.OpPrereqError(msg)
2268

    
2269
    if (self.op.master_candidate == True and node.offline and
2270
        not self.op.offline == False):
2271
      raise errors.OpPrereqError("Can't set an offline node to"
2272
                                 " master_candidate")
2273

    
2274
    return
2275

    
2276
  def Exec(self, feedback_fn):
2277
    """Modifies a node.
2278

2279
    """
2280
    node = self.node
2281

    
2282
    result = []
2283

    
2284
    if self.op.offline is not None:
2285
      node.offline = self.op.offline
2286
      result.append(("offline", str(self.op.offline)))
2287
      if self.op.offline == True and node.master_candidate:
2288
        node.master_candidate = False
2289
        result.append(("master_candidate", "auto-demotion due to offline"))
2290

    
2291
    if self.op.master_candidate is not None:
2292
      node.master_candidate = self.op.master_candidate
2293
      result.append(("master_candidate", str(self.op.master_candidate)))
2294
      if self.op.master_candidate == False:
2295
        rrc = self.rpc.call_node_demote_from_mc(node.name)
2296
        msg = rrc.RemoteFailMsg()
2297
        if msg:
2298
          self.LogWarning("Node failed to demote itself: %s" % msg)
2299

    
2300
    # this will trigger configuration file update, if needed
2301
    self.cfg.Update(node)
2302
    # this will trigger job queue propagation or cleanup
2303
    if self.op.node_name != self.cfg.GetMasterNode():
2304
      self.context.ReaddNode(node)
2305

    
2306
    return result
2307

    
2308

    
2309
class LUQueryClusterInfo(NoHooksLU):
2310
  """Query cluster configuration.
2311

2312
  """
2313
  _OP_REQP = []
2314
  REQ_BGL = False
2315

    
2316
  def ExpandNames(self):
2317
    self.needed_locks = {}
2318

    
2319
  def CheckPrereq(self):
2320
    """No prerequsites needed for this LU.
2321

2322
    """
2323
    pass
2324

    
2325
  def Exec(self, feedback_fn):
2326
    """Return cluster config.
2327

2328
    """
2329
    cluster = self.cfg.GetClusterInfo()
2330
    result = {
2331
      "software_version": constants.RELEASE_VERSION,
2332
      "protocol_version": constants.PROTOCOL_VERSION,
2333
      "config_version": constants.CONFIG_VERSION,
2334
      "os_api_version": constants.OS_API_VERSION,
2335
      "export_version": constants.EXPORT_VERSION,
2336
      "architecture": (platform.architecture()[0], platform.machine()),
2337
      "name": cluster.cluster_name,
2338
      "master": cluster.master_node,
2339
      "default_hypervisor": cluster.default_hypervisor,
2340
      "enabled_hypervisors": cluster.enabled_hypervisors,
2341
      "hvparams": dict([(hypervisor, cluster.hvparams[hypervisor])
2342
                        for hypervisor in cluster.enabled_hypervisors]),
2343
      "beparams": cluster.beparams,
2344
      "candidate_pool_size": cluster.candidate_pool_size,
2345
      }
2346

    
2347
    return result
2348

    
2349

    
2350
class LUQueryConfigValues(NoHooksLU):
2351
  """Return configuration values.
2352

2353
  """
2354
  _OP_REQP = []
2355
  REQ_BGL = False
2356
  _FIELDS_DYNAMIC = utils.FieldSet()
2357
  _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag")
2358

    
2359
  def ExpandNames(self):
2360
    self.needed_locks = {}
2361

    
2362
    _CheckOutputFields(static=self._FIELDS_STATIC,
2363
                       dynamic=self._FIELDS_DYNAMIC,
2364
                       selected=self.op.output_fields)
2365

    
2366
  def CheckPrereq(self):
2367
    """No prerequisites.
2368

2369
    """
2370
    pass
2371

    
2372
  def Exec(self, feedback_fn):
2373
    """Dump a representation of the cluster config to the standard output.
2374

2375
    """
2376
    values = []
2377
    for field in self.op.output_fields:
2378
      if field == "cluster_name":
2379
        entry = self.cfg.GetClusterName()
2380
      elif field == "master_node":
2381
        entry = self.cfg.GetMasterNode()
2382
      elif field == "drain_flag":
2383
        entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
2384
      else:
2385
        raise errors.ParameterError(field)
2386
      values.append(entry)
2387
    return values
2388

    
2389

    
2390
class LUActivateInstanceDisks(NoHooksLU):
2391
  """Bring up an instance's disks.
2392

2393
  """
2394
  _OP_REQP = ["instance_name"]
2395
  REQ_BGL = False
2396

    
2397
  def ExpandNames(self):
2398
    self._ExpandAndLockInstance()
2399
    self.needed_locks[locking.LEVEL_NODE] = []
2400
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2401

    
2402
  def DeclareLocks(self, level):
2403
    if level == locking.LEVEL_NODE:
2404
      self._LockInstancesNodes()
2405

    
2406
  def CheckPrereq(self):
2407
    """Check prerequisites.
2408

2409
    This checks that the instance is in the cluster.
2410

2411
    """
2412
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2413
    assert self.instance is not None, \
2414
      "Cannot retrieve locked instance %s" % self.op.instance_name
2415
    _CheckNodeOnline(self, self.instance.primary_node)
2416

    
2417
  def Exec(self, feedback_fn):
2418
    """Activate the disks.
2419

2420
    """
2421
    disks_ok, disks_info = _AssembleInstanceDisks(self, self.instance)
2422
    if not disks_ok:
2423
      raise errors.OpExecError("Cannot activate block devices")
2424

    
2425
    return disks_info
2426

    
2427

    
2428
def _AssembleInstanceDisks(lu, instance, ignore_secondaries=False):
2429
  """Prepare the block devices for an instance.
2430

2431
  This sets up the block devices on all nodes.
2432

2433
  @type lu: L{LogicalUnit}
2434
  @param lu: the logical unit on whose behalf we execute
2435
  @type instance: L{objects.Instance}
2436
  @param instance: the instance for whose disks we assemble
2437
  @type ignore_secondaries: boolean
2438
  @param ignore_secondaries: if true, errors on secondary nodes
2439
      won't result in an error return from the function
2440
  @return: False if the operation failed, otherwise a list of
2441
      (host, instance_visible_name, node_visible_name)
2442
      with the mapping from node devices to instance devices
2443

2444
  """
2445
  device_info = []
2446
  disks_ok = True
2447
  iname = instance.name
2448
  # With the two passes mechanism we try to reduce the window of
2449
  # opportunity for the race condition of switching DRBD to primary
2450
  # before handshaking occured, but we do not eliminate it
2451

    
2452
  # The proper fix would be to wait (with some limits) until the
2453
  # connection has been made and drbd transitions from WFConnection
2454
  # into any other network-connected state (Connected, SyncTarget,
2455
  # SyncSource, etc.)
2456

    
2457
  # 1st pass, assemble on all nodes in secondary mode
2458
  for inst_disk in instance.disks:
2459
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
2460
      lu.cfg.SetDiskID(node_disk, node)
2461
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
2462
      msg = result.RemoteFailMsg()
2463
      if msg:
2464
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
2465
                           " (is_primary=False, pass=1): %s",
2466
                           inst_disk.iv_name, node, msg)
2467
        if not ignore_secondaries:
2468
          disks_ok = False
2469

    
2470
  # FIXME: race condition on drbd migration to primary
2471

    
2472
  # 2nd pass, do only the primary node
2473
  for inst_disk in instance.disks:
2474
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
2475
      if node != instance.primary_node:
2476
        continue
2477
      lu.cfg.SetDiskID(node_disk, node)
2478
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
2479
      msg = result.RemoteFailMsg()
2480
      if msg:
2481
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
2482
                           " (is_primary=True, pass=2): %s",
2483
                           inst_disk.iv_name, node, msg)
2484
        disks_ok = False
2485
    device_info.append((instance.primary_node, inst_disk.iv_name, result.data))
2486

    
2487
  # leave the disks configured for the primary node
2488
  # this is a workaround that would be fixed better by
2489
  # improving the logical/physical id handling
2490
  for disk in instance.disks:
2491
    lu.cfg.SetDiskID(disk, instance.primary_node)
2492

    
2493
  return disks_ok, device_info
2494

    
2495

    
2496
def _StartInstanceDisks(lu, instance, force):
2497
  """Start the disks of an instance.
2498

2499
  """
2500
  disks_ok, dummy = _AssembleInstanceDisks(lu, instance,
2501
                                           ignore_secondaries=force)
2502
  if not disks_ok:
2503
    _ShutdownInstanceDisks(lu, instance)
2504
    if force is not None and not force:
2505
      lu.proc.LogWarning("", hint="If the message above refers to a"
2506
                         " secondary node,"
2507
                         " you can retry the operation using '--force'.")
2508
    raise errors.OpExecError("Disk consistency error")
2509

    
2510

    
2511
class LUDeactivateInstanceDisks(NoHooksLU):
2512
  """Shutdown an instance's disks.
2513

2514
  """
2515
  _OP_REQP = ["instance_name"]
2516
  REQ_BGL = False
2517

    
2518
  def ExpandNames(self):
2519
    self._ExpandAndLockInstance()
2520
    self.needed_locks[locking.LEVEL_NODE] = []
2521
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2522

    
2523
  def DeclareLocks(self, level):
2524
    if level == locking.LEVEL_NODE:
2525
      self._LockInstancesNodes()
2526

    
2527
  def CheckPrereq(self):
2528
    """Check prerequisites.
2529

2530
    This checks that the instance is in the cluster.
2531

2532
    """
2533
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2534
    assert self.instance is not None, \
2535
      "Cannot retrieve locked instance %s" % self.op.instance_name
2536

    
2537
  def Exec(self, feedback_fn):
2538
    """Deactivate the disks
2539

2540
    """
2541
    instance = self.instance
2542
    _SafeShutdownInstanceDisks(self, instance)
2543

    
2544

    
2545
def _SafeShutdownInstanceDisks(lu, instance):
2546
  """Shutdown block devices of an instance.
2547

2548
  This function checks if an instance is running, before calling
2549
  _ShutdownInstanceDisks.
2550

2551
  """
2552
  ins_l = lu.rpc.call_instance_list([instance.primary_node],
2553
                                      [instance.hypervisor])
2554
  ins_l = ins_l[instance.primary_node]
2555
  if ins_l.failed or not isinstance(ins_l.data, list):
2556
    raise errors.OpExecError("Can't contact node '%s'" %
2557
                             instance.primary_node)
2558

    
2559
  if instance.name in ins_l.data:
2560
    raise errors.OpExecError("Instance is running, can't shutdown"
2561
                             " block devices.")
2562

    
2563
  _ShutdownInstanceDisks(lu, instance)
2564

    
2565

    
2566
def _ShutdownInstanceDisks(lu, instance, ignore_primary=False):
2567
  """Shutdown block devices of an instance.
2568

2569
  This does the shutdown on all nodes of the instance.
2570

2571
  If the ignore_primary is false, errors on the primary node are
2572
  ignored.
2573

2574
  """
2575
  result = True
2576
  for disk in instance.disks:
2577
    for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
2578
      lu.cfg.SetDiskID(top_disk, node)
2579
      result = lu.rpc.call_blockdev_shutdown(node, top_disk)
2580
      if result.failed or not result.data:
2581
        logging.error("Could not shutdown block device %s on node %s",
2582
                      disk.iv_name, node)
2583
        if not ignore_primary or node != instance.primary_node:
2584
          result = False
2585
  return result
2586

    
2587

    
2588
def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
2589
  """Checks if a node has enough free memory.
2590

2591
  This function check if a given node has the needed amount of free
2592
  memory. In case the node has less memory or we cannot get the
2593
  information from the node, this function raise an OpPrereqError
2594
  exception.
2595

2596
  @type lu: C{LogicalUnit}
2597
  @param lu: a logical unit from which we get configuration data
2598
  @type node: C{str}
2599
  @param node: the node to check
2600
  @type reason: C{str}
2601
  @param reason: string to use in the error message
2602
  @type requested: C{int}
2603
  @param requested: the amount of memory in MiB to check for
2604
  @type hypervisor_name: C{str}
2605
  @param hypervisor_name: the hypervisor to ask for memory stats
2606
  @raise errors.OpPrereqError: if the node doesn't have enough memory, or
2607
      we cannot check the node
2608

2609
  """
2610
  nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
2611
  nodeinfo[node].Raise()
2612
  free_mem = nodeinfo[node].data.get('memory_free')
2613
  if not isinstance(free_mem, int):
2614
    raise errors.OpPrereqError("Can't compute free memory on node %s, result"
2615
                             " was '%s'" % (node, free_mem))
2616
  if requested > free_mem:
2617
    raise errors.OpPrereqError("Not enough memory on node %s for %s:"
2618
                             " needed %s MiB, available %s MiB" %
2619
                             (node, reason, requested, free_mem))
2620

    
2621

    
2622
class LUStartupInstance(LogicalUnit):
2623
  """Starts an instance.
2624

2625
  """
2626
  HPATH = "instance-start"
2627
  HTYPE = constants.HTYPE_INSTANCE
2628
  _OP_REQP = ["instance_name", "force"]
2629
  REQ_BGL = False
2630

    
2631
  def ExpandNames(self):
2632
    self._ExpandAndLockInstance()
2633

    
2634
  def BuildHooksEnv(self):
2635
    """Build hooks env.
2636

2637
    This runs on master, primary and secondary nodes of the instance.
2638

2639
    """
2640
    env = {
2641
      "FORCE": self.op.force,
2642
      }
2643
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2644
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2645
    return env, nl, nl
2646

    
2647
  def CheckPrereq(self):
2648
    """Check prerequisites.
2649

2650
    This checks that the instance is in the cluster.
2651

2652
    """
2653
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2654
    assert self.instance is not None, \
2655
      "Cannot retrieve locked instance %s" % self.op.instance_name
2656

    
2657
    _CheckNodeOnline(self, instance.primary_node)
2658

    
2659
    bep = self.cfg.GetClusterInfo().FillBE(instance)
2660
    # check bridges existance
2661
    _CheckInstanceBridgesExist(self, instance)
2662

    
2663
    _CheckNodeFreeMemory(self, instance.primary_node,
2664
                         "starting instance %s" % instance.name,
2665
                         bep[constants.BE_MEMORY], instance.hypervisor)
2666

    
2667
  def Exec(self, feedback_fn):
2668
    """Start the instance.
2669

2670
    """
2671
    instance = self.instance
2672
    force = self.op.force
2673
    extra_args = getattr(self.op, "extra_args", "")
2674

    
2675
    self.cfg.MarkInstanceUp(instance.name)
2676

    
2677
    node_current = instance.primary_node
2678

    
2679
    _StartInstanceDisks(self, instance, force)
2680

    
2681
    result = self.rpc.call_instance_start(node_current, instance, extra_args)
2682
    msg = result.RemoteFailMsg()
2683
    if msg:
2684
      _ShutdownInstanceDisks(self, instance)
2685
      raise errors.OpExecError("Could not start instance: %s" % msg)
2686

    
2687

    
2688
class LURebootInstance(LogicalUnit):
2689
  """Reboot an instance.
2690

2691
  """
2692
  HPATH = "instance-reboot"
2693
  HTYPE = constants.HTYPE_INSTANCE
2694
  _OP_REQP = ["instance_name", "ignore_secondaries", "reboot_type"]
2695
  REQ_BGL = False
2696

    
2697
  def ExpandNames(self):
2698
    if self.op.reboot_type not in [constants.INSTANCE_REBOOT_SOFT,
2699
                                   constants.INSTANCE_REBOOT_HARD,
2700
                                   constants.INSTANCE_REBOOT_FULL]:
2701
      raise errors.ParameterError("reboot type not in [%s, %s, %s]" %
2702
                                  (constants.INSTANCE_REBOOT_SOFT,
2703
                                   constants.INSTANCE_REBOOT_HARD,
2704
                                   constants.INSTANCE_REBOOT_FULL))
2705
    self._ExpandAndLockInstance()
2706

    
2707
  def BuildHooksEnv(self):
2708
    """Build hooks env.
2709

2710
    This runs on master, primary and secondary nodes of the instance.
2711

2712
    """
2713
    env = {
2714
      "IGNORE_SECONDARIES": self.op.ignore_secondaries,
2715
      }
2716
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2717
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2718
    return env, nl, nl
2719

    
2720
  def CheckPrereq(self):
2721
    """Check prerequisites.
2722

2723
    This checks that the instance is in the cluster.
2724

2725
    """
2726
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2727
    assert self.instance is not None, \
2728
      "Cannot retrieve locked instance %s" % self.op.instance_name
2729

    
2730
    _CheckNodeOnline(self, instance.primary_node)
2731

    
2732
    # check bridges existance
2733
    _CheckInstanceBridgesExist(self, instance)
2734

    
2735
  def Exec(self, feedback_fn):
2736
    """Reboot the instance.
2737

2738
    """
2739
    instance = self.instance
2740
    ignore_secondaries = self.op.ignore_secondaries
2741
    reboot_type = self.op.reboot_type
2742
    extra_args = getattr(self.op, "extra_args", "")
2743

    
2744
    node_current = instance.primary_node
2745

    
2746
    if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
2747
                       constants.INSTANCE_REBOOT_HARD]:
2748
      result = self.rpc.call_instance_reboot(node_current, instance,
2749
                                             reboot_type, extra_args)
2750
      if result.failed or not result.data:
2751
        raise errors.OpExecError("Could not reboot instance")
2752
    else:
2753
      if not self.rpc.call_instance_shutdown(node_current, instance):
2754
        raise errors.OpExecError("could not shutdown instance for full reboot")
2755
      _ShutdownInstanceDisks(self, instance)
2756
      _StartInstanceDisks(self, instance, ignore_secondaries)
2757
      result = self.rpc.call_instance_start(node_current, instance, extra_args)
2758
      msg = result.RemoteFailMsg()
2759
      if msg:
2760
        _ShutdownInstanceDisks(self, instance)
2761
        raise errors.OpExecError("Could not start instance for"
2762
                                 " full reboot: %s" % msg)
2763

    
2764
    self.cfg.MarkInstanceUp(instance.name)
2765

    
2766

    
2767
class LUShutdownInstance(LogicalUnit):
2768
  """Shutdown an instance.
2769

2770
  """
2771
  HPATH = "instance-stop"
2772
  HTYPE = constants.HTYPE_INSTANCE
2773
  _OP_REQP = ["instance_name"]
2774
  REQ_BGL = False
2775

    
2776
  def ExpandNames(self):
2777
    self._ExpandAndLockInstance()
2778

    
2779
  def BuildHooksEnv(self):
2780
    """Build hooks env.
2781

2782
    This runs on master, primary and secondary nodes of the instance.
2783

2784
    """
2785
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2786
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2787
    return env, nl, nl
2788

    
2789
  def CheckPrereq(self):
2790
    """Check prerequisites.
2791

2792
    This checks that the instance is in the cluster.
2793

2794
    """
2795
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2796
    assert self.instance is not None, \
2797
      "Cannot retrieve locked instance %s" % self.op.instance_name
2798
    _CheckNodeOnline(self, self.instance.primary_node)
2799

    
2800
  def Exec(self, feedback_fn):
2801
    """Shutdown the instance.
2802

2803
    """
2804
    instance = self.instance
2805
    node_current = instance.primary_node
2806
    self.cfg.MarkInstanceDown(instance.name)
2807
    result = self.rpc.call_instance_shutdown(node_current, instance)
2808
    if result.failed or not result.data:
2809
      self.proc.LogWarning("Could not shutdown instance")
2810

    
2811
    _ShutdownInstanceDisks(self, instance)
2812

    
2813

    
2814
class LUReinstallInstance(LogicalUnit):
2815
  """Reinstall an instance.
2816

2817
  """
2818
  HPATH = "instance-reinstall"
2819
  HTYPE = constants.HTYPE_INSTANCE
2820
  _OP_REQP = ["instance_name"]
2821
  REQ_BGL = False
2822

    
2823
  def ExpandNames(self):
2824
    self._ExpandAndLockInstance()
2825

    
2826
  def BuildHooksEnv(self):
2827
    """Build hooks env.
2828

2829
    This runs on master, primary and secondary nodes of the instance.
2830

2831
    """
2832
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2833
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2834
    return env, nl, nl
2835

    
2836
  def CheckPrereq(self):
2837
    """Check prerequisites.
2838

2839
    This checks that the instance is in the cluster and is not running.
2840

2841
    """
2842
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2843
    assert instance is not None, \
2844
      "Cannot retrieve locked instance %s" % self.op.instance_name
2845
    _CheckNodeOnline(self, instance.primary_node)
2846

    
2847
    if instance.disk_template == constants.DT_DISKLESS:
2848
      raise errors.OpPrereqError("Instance '%s' has no disks" %
2849
                                 self.op.instance_name)
2850
    if instance.admin_up:
2851
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
2852
                                 self.op.instance_name)
2853
    remote_info = self.rpc.call_instance_info(instance.primary_node,
2854
                                              instance.name,
2855
                                              instance.hypervisor)
2856
    if remote_info.failed or remote_info.data:
2857
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
2858
                                 (self.op.instance_name,
2859
                                  instance.primary_node))
2860

    
2861
    self.op.os_type = getattr(self.op, "os_type", None)
2862
    if self.op.os_type is not None:
2863
      # OS verification
2864
      pnode = self.cfg.GetNodeInfo(
2865
        self.cfg.ExpandNodeName(instance.primary_node))
2866
      if pnode is None:
2867
        raise errors.OpPrereqError("Primary node '%s' is unknown" %
2868
                                   self.op.pnode)
2869
      result = self.rpc.call_os_get(pnode.name, self.op.os_type)
2870
      result.Raise()
2871
      if not isinstance(result.data, objects.OS):
2872
        raise errors.OpPrereqError("OS '%s' not in supported OS list for"
2873
                                   " primary node"  % self.op.os_type)
2874

    
2875
    self.instance = instance
2876

    
2877
  def Exec(self, feedback_fn):
2878
    """Reinstall the instance.
2879

2880
    """
2881
    inst = self.instance
2882

    
2883
    if self.op.os_type is not None:
2884
      feedback_fn("Changing OS to '%s'..." % self.op.os_type)
2885
      inst.os = self.op.os_type
2886
      self.cfg.Update(inst)
2887

    
2888
    _StartInstanceDisks(self, inst, None)
2889
    try:
2890
      feedback_fn("Running the instance OS create scripts...")
2891
      result = self.rpc.call_instance_os_add(inst.primary_node, inst)
2892
      msg = result.RemoteFailMsg()
2893
      if msg:
2894
        raise errors.OpExecError("Could not install OS for instance %s"
2895
                                 " on node %s: %s" %
2896
                                 (inst.name, inst.primary_node, msg))
2897
    finally:
2898
      _ShutdownInstanceDisks(self, inst)
2899

    
2900

    
2901
class LURenameInstance(LogicalUnit):
2902
  """Rename an instance.
2903

2904
  """
2905
  HPATH = "instance-rename"
2906
  HTYPE = constants.HTYPE_INSTANCE
2907
  _OP_REQP = ["instance_name", "new_name"]
2908

    
2909
  def BuildHooksEnv(self):
2910
    """Build hooks env.
2911

2912
    This runs on master, primary and secondary nodes of the instance.
2913

2914
    """
2915
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2916
    env["INSTANCE_NEW_NAME"] = self.op.new_name
2917
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2918
    return env, nl, nl
2919

    
2920
  def CheckPrereq(self):
2921
    """Check prerequisites.
2922

2923
    This checks that the instance is in the cluster and is not running.
2924

2925
    """
2926
    instance = self.cfg.GetInstanceInfo(
2927
      self.cfg.ExpandInstanceName(self.op.instance_name))
2928
    if instance is None:
2929
      raise errors.OpPrereqError("Instance '%s' not known" %
2930
                                 self.op.instance_name)
2931
    _CheckNodeOnline(self, instance.primary_node)
2932

    
2933
    if instance.admin_up:
2934
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
2935
                                 self.op.instance_name)
2936
    remote_info = self.rpc.call_instance_info(instance.primary_node,
2937
                                              instance.name,
2938
                                              instance.hypervisor)
2939
    remote_info.Raise()
2940
    if remote_info.data:
2941
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
2942
                                 (self.op.instance_name,
2943
                                  instance.primary_node))
2944
    self.instance = instance
2945

    
2946
    # new name verification
2947
    name_info = utils.HostInfo(self.op.new_name)
2948

    
2949
    self.op.new_name = new_name = name_info.name
2950
    instance_list = self.cfg.GetInstanceList()
2951
    if new_name in instance_list:
2952
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
2953
                                 new_name)
2954

    
2955
    if not getattr(self.op, "ignore_ip", False):
2956
      if utils.TcpPing(name_info.ip, constants.DEFAULT_NODED_PORT):
2957
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
2958
                                   (name_info.ip, new_name))
2959

    
2960

    
2961
  def Exec(self, feedback_fn):
2962
    """Reinstall the instance.
2963

2964
    """
2965
    inst = self.instance
2966
    old_name = inst.name
2967

    
2968
    if inst.disk_template == constants.DT_FILE:
2969
      old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
2970

    
2971
    self.cfg.RenameInstance(inst.name, self.op.new_name)
2972
    # Change the instance lock. This is definitely safe while we hold the BGL
2973
    self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
2974
    self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
2975

    
2976
    # re-read the instance from the configuration after rename
2977
    inst = self.cfg.GetInstanceInfo(self.op.new_name)
2978

    
2979
    if inst.disk_template == constants.DT_FILE:
2980
      new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
2981
      result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
2982
                                                     old_file_storage_dir,
2983
                                                     new_file_storage_dir)
2984
      result.Raise()
2985
      if not result.data:
2986
        raise errors.OpExecError("Could not connect to node '%s' to rename"
2987
                                 " directory '%s' to '%s' (but the instance"
2988
                                 " has been renamed in Ganeti)" % (
2989
                                 inst.primary_node, old_file_storage_dir,
2990
                                 new_file_storage_dir))
2991

    
2992
      if not result.data[0]:
2993
        raise errors.OpExecError("Could not rename directory '%s' to '%s'"
2994
                                 " (but the instance has been renamed in"
2995
                                 " Ganeti)" % (old_file_storage_dir,
2996
                                               new_file_storage_dir))
2997

    
2998
    _StartInstanceDisks(self, inst, None)
2999
    try:
3000
      result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
3001
                                                 old_name)
3002
      msg = result.RemoteFailMsg()
3003
      if msg:
3004
        msg = ("Could not run OS rename script for instance %s on node %s"
3005
               " (but the instance has been renamed in Ganeti): %s" %
3006
               (inst.name, inst.primary_node, msg))
3007
        self.proc.LogWarning(msg)
3008
    finally:
3009
      _ShutdownInstanceDisks(self, inst)
3010

    
3011

    
3012
class LURemoveInstance(LogicalUnit):
3013
  """Remove an instance.
3014

3015
  """
3016
  HPATH = "instance-remove"
3017
  HTYPE = constants.HTYPE_INSTANCE
3018
  _OP_REQP = ["instance_name", "ignore_failures"]
3019
  REQ_BGL = False
3020

    
3021
  def ExpandNames(self):
3022
    self._ExpandAndLockInstance()
3023
    self.needed_locks[locking.LEVEL_NODE] = []
3024
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3025

    
3026
  def DeclareLocks(self, level):
3027
    if level == locking.LEVEL_NODE:
3028
      self._LockInstancesNodes()
3029

    
3030
  def BuildHooksEnv(self):
3031
    """Build hooks env.
3032

3033
    This runs on master, primary and secondary nodes of the instance.
3034

3035
    """
3036
    env = _BuildInstanceHookEnvByObject(self, self.instance)
3037
    nl = [self.cfg.GetMasterNode()]
3038
    return env, nl, nl
3039

    
3040
  def CheckPrereq(self):
3041
    """Check prerequisites.
3042

3043
    This checks that the instance is in the cluster.
3044

3045
    """
3046
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3047
    assert self.instance is not None, \
3048
      "Cannot retrieve locked instance %s" % self.op.instance_name
3049

    
3050
  def Exec(self, feedback_fn):
3051
    """Remove the instance.
3052

3053
    """
3054
    instance = self.instance
3055
    logging.info("Shutting down instance %s on node %s",
3056
                 instance.name, instance.primary_node)
3057

    
3058
    result = self.rpc.call_instance_shutdown(instance.primary_node, instance)
3059
    if result.failed or not result.data:
3060
      if self.op.ignore_failures:
3061
        feedback_fn("Warning: can't shutdown instance")
3062
      else:
3063
        raise errors.OpExecError("Could not shutdown instance %s on node %s" %
3064
                                 (instance.name, instance.primary_node))
3065

    
3066
    logging.info("Removing block devices for instance %s", instance.name)
3067

    
3068
    if not _RemoveDisks(self, instance):
3069
      if self.op.ignore_failures:
3070
        feedback_fn("Warning: can't remove instance's disks")
3071
      else:
3072
        raise errors.OpExecError("Can't remove instance's disks")
3073

    
3074
    logging.info("Removing instance %s out of cluster config", instance.name)
3075

    
3076
    self.cfg.RemoveInstance(instance.name)
3077
    self.remove_locks[locking.LEVEL_INSTANCE] = instance.name
3078

    
3079

    
3080
class LUQueryInstances(NoHooksLU):
3081
  """Logical unit for querying instances.
3082

3083
  """
3084
  _OP_REQP = ["output_fields", "names", "use_locking"]
3085
  REQ_BGL = False
3086
  _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
3087
                                    "admin_state", "admin_ram",
3088
                                    "disk_template", "ip", "mac", "bridge",
3089
                                    "sda_size", "sdb_size", "vcpus", "tags",
3090
                                    "network_port", "beparams",
3091
                                    "(disk).(size)/([0-9]+)",
3092
                                    "(disk).(sizes)", "disk_usage",
3093
                                    "(nic).(mac|ip|bridge)/([0-9]+)",
3094
                                    "(nic).(macs|ips|bridges)",
3095
                                    "(disk|nic).(count)",
3096
                                    "serial_no", "hypervisor", "hvparams",] +
3097
                                  ["hv/%s" % name
3098
                                   for name in constants.HVS_PARAMETERS] +
3099
                                  ["be/%s" % name
3100
                                   for name in constants.BES_PARAMETERS])
3101
  _FIELDS_DYNAMIC = utils.FieldSet("oper_state", "oper_ram", "status")
3102

    
3103

    
3104
  def ExpandNames(self):
3105
    _CheckOutputFields(static=self._FIELDS_STATIC,
3106
                       dynamic=self._FIELDS_DYNAMIC,
3107
                       selected=self.op.output_fields)
3108

    
3109
    self.needed_locks = {}
3110
    self.share_locks[locking.LEVEL_INSTANCE] = 1
3111
    self.share_locks[locking.LEVEL_NODE] = 1
3112

    
3113
    if self.op.names:
3114
      self.wanted = _GetWantedInstances(self, self.op.names)
3115
    else:
3116
      self.wanted = locking.ALL_SET
3117

    
3118
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
3119
    self.do_locking = self.do_node_query and self.op.use_locking
3120
    if self.do_locking:
3121
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
3122
      self.needed_locks[locking.LEVEL_NODE] = []
3123
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3124

    
3125
  def DeclareLocks(self, level):
3126
    if level == locking.LEVEL_NODE and self.do_locking:
3127
      self._LockInstancesNodes()
3128

    
3129
  def CheckPrereq(self):
3130
    """Check prerequisites.
3131

3132
    """
3133
    pass
3134

    
3135
  def Exec(self, feedback_fn):
3136
    """Computes the list of nodes and their attributes.
3137

3138
    """
3139
    all_info = self.cfg.GetAllInstancesInfo()
3140
    if self.wanted == locking.ALL_SET:
3141
      # caller didn't specify instance names, so ordering is not important
3142
      if self.do_locking:
3143
        instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
3144
      else:
3145
        instance_names = all_info.keys()
3146
      instance_names = utils.NiceSort(instance_names)
3147
    else:
3148
      # caller did specify names, so we must keep the ordering
3149
      if self.do_locking:
3150
        tgt_set = self.acquired_locks[locking.LEVEL_INSTANCE]
3151
      else:
3152
        tgt_set = all_info.keys()
3153
      missing = set(self.wanted).difference(tgt_set)
3154
      if missing:
3155
        raise errors.OpExecError("Some instances were removed before"
3156
                                 " retrieving their data: %s" % missing)
3157
      instance_names = self.wanted
3158

    
3159
    instance_list = [all_info[iname] for iname in instance_names]
3160

    
3161
    # begin data gathering
3162

    
3163
    nodes = frozenset([inst.primary_node for inst in instance_list])
3164
    hv_list = list(set([inst.hypervisor for inst in instance_list]))
3165

    
3166
    bad_nodes = []
3167
    off_nodes = []
3168
    if self.do_node_query:
3169
      live_data = {}
3170
      node_data = self.rpc.call_all_instances_info(nodes, hv_list)
3171
      for name in nodes:
3172
        result = node_data[name]
3173
        if result.offline:
3174
          # offline nodes will be in both lists
3175
          off_nodes.append(name)
3176
        if result.failed:
3177
          bad_nodes.append(name)
3178
        else:
3179
          if result.data:
3180
            live_data.update(result.data)
3181
            # else no instance is alive
3182
    else:
3183
      live_data = dict([(name, {}) for name in instance_names])
3184

    
3185
    # end data gathering
3186

    
3187
    HVPREFIX = "hv/"
3188
    BEPREFIX = "be/"
3189
    output = []
3190
    for instance in instance_list:
3191
      iout = []
3192
      i_hv = self.cfg.GetClusterInfo().FillHV(instance)
3193
      i_be = self.cfg.GetClusterInfo().FillBE(instance)
3194
      for field in self.op.output_fields:
3195
        st_match = self._FIELDS_STATIC.Matches(field)
3196
        if field == "name":
3197
          val = instance.name
3198
        elif field == "os":
3199
          val = instance.os
3200
        elif field == "pnode":
3201
          val = instance.primary_node
3202
        elif field == "snodes":
3203
          val = list(instance.secondary_nodes)
3204
        elif field == "admin_state":
3205
          val = instance.admin_up
3206
        elif field == "oper_state":
3207
          if instance.primary_node in bad_nodes:
3208
            val = None
3209
          else:
3210
            val = bool(live_data.get(instance.name))
3211
        elif field == "status":
3212
          if instance.primary_node in off_nodes:
3213
            val = "ERROR_nodeoffline"
3214
          elif instance.primary_node in bad_nodes:
3215
            val = "ERROR_nodedown"
3216
          else:
3217
            running = bool(live_data.get(instance.name))
3218
            if running:
3219
              if instance.admin_up:
3220
                val = "running"
3221
              else:
3222
                val = "ERROR_up"
3223
            else:
3224
              if instance.admin_up:
3225
                val = "ERROR_down"
3226
              else:
3227
                val = "ADMIN_down"
3228
        elif field == "oper_ram":
3229
          if instance.primary_node in bad_nodes:
3230
            val = None
3231
          elif instance.name in live_data:
3232
            val = live_data[instance.name].get("memory", "?")
3233
          else:
3234
            val = "-"
3235
        elif field == "disk_template":
3236
          val = instance.disk_template
3237
        elif field == "ip":
3238
          val = instance.nics[0].ip
3239
        elif field == "bridge":
3240
          val = instance.nics[0].bridge
3241
        elif field == "mac":
3242
          val = instance.nics[0].mac
3243
        elif field == "sda_size" or field == "sdb_size":
3244
          idx = ord(field[2]) - ord('a')
3245
          try:
3246
            val = instance.FindDisk(idx).size
3247
          except errors.OpPrereqError:
3248
            val = None
3249
        elif field == "disk_usage": # total disk usage per node
3250
          disk_sizes = [{'size': disk.size} for disk in instance.disks]
3251
          val = _ComputeDiskSize(instance.disk_template, disk_sizes)
3252
        elif field == "tags":
3253
          val = list(instance.GetTags())
3254
        elif field == "serial_no":
3255
          val = instance.serial_no
3256
        elif field == "network_port":
3257
          val = instance.network_port
3258
        elif field == "hypervisor":
3259
          val = instance.hypervisor
3260
        elif field == "hvparams":
3261
          val = i_hv
3262
        elif (field.startswith(HVPREFIX) and
3263
              field[len(HVPREFIX):] in constants.HVS_PARAMETERS):
3264
          val = i_hv.get(field[len(HVPREFIX):], None)
3265
        elif field == "beparams":
3266
          val = i_be
3267
        elif (field.startswith(BEPREFIX) and
3268
              field[len(BEPREFIX):] in constants.BES_PARAMETERS):
3269
          val = i_be.get(field[len(BEPREFIX):], None)
3270
        elif st_match and st_match.groups():
3271
          # matches a variable list
3272
          st_groups = st_match.groups()
3273
          if st_groups and st_groups[0] == "disk":
3274
            if st_groups[1] == "count":
3275
              val = len(instance.disks)
3276
            elif st_groups[1] == "sizes":
3277
              val = [disk.size for disk in instance.disks]
3278
            elif st_groups[1] == "size":
3279
              try:
3280
                val = instance.FindDisk(st_groups[2]).size
3281
              except errors.OpPrereqError:
3282
                val = None
3283
            else:
3284
              assert False, "Unhandled disk parameter"
3285
          elif st_groups[0] == "nic":
3286
            if st_groups[1] == "count":
3287
              val = len(instance.nics)
3288
            elif st_groups[1] == "macs":
3289
              val = [nic.mac for nic in instance.nics]
3290
            elif st_groups[1] == "ips":
3291
              val = [nic.ip for nic in instance.nics]
3292
            elif st_groups[1] == "bridges":
3293
              val = [nic.bridge for nic in instance.nics]
3294
            else:
3295
              # index-based item
3296
              nic_idx = int(st_groups[2])
3297
              if nic_idx >= len(instance.nics):
3298
                val = None
3299
              else:
3300
                if st_groups[1] == "mac":
3301
                  val = instance.nics[nic_idx].mac
3302
                elif st_groups[1] == "ip":
3303
                  val = instance.nics[nic_idx].ip
3304
                elif st_groups[1] == "bridge":
3305
                  val = instance.nics[nic_idx].bridge
3306
                else:
3307
                  assert False, "Unhandled NIC parameter"
3308
          else:
3309
            assert False, "Unhandled variable parameter"
3310
        else:
3311
          raise errors.ParameterError(field)
3312
        iout.append(val)
3313
      output.append(iout)
3314

    
3315
    return output
3316

    
3317

    
3318
class LUFailoverInstance(LogicalUnit):
3319
  """Failover an instance.
3320

3321
  """
3322
  HPATH = "instance-failover"
3323
  HTYPE = constants.HTYPE_INSTANCE
3324
  _OP_REQP = ["instance_name", "ignore_consistency"]
3325
  REQ_BGL = False
3326

    
3327
  def ExpandNames(self):
3328
    self._ExpandAndLockInstance()
3329
    self.needed_locks[locking.LEVEL_NODE] = []
3330
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3331

    
3332
  def DeclareLocks(self, level):
3333
    if level == locking.LEVEL_NODE:
3334
      self._LockInstancesNodes()
3335

    
3336
  def BuildHooksEnv(self):
3337
    """Build hooks env.
3338

3339
    This runs on master, primary and secondary nodes of the instance.
3340

3341
    """
3342
    env = {
3343
      "IGNORE_CONSISTENCY": self.op.ignore_consistency,
3344
      }
3345
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
3346
    nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
3347
    return env, nl, nl
3348

    
3349
  def CheckPrereq(self):
3350
    """Check prerequisites.
3351

3352
    This checks that the instance is in the cluster.
3353

3354
    """
3355
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3356
    assert self.instance is not None, \
3357
      "Cannot retrieve locked instance %s" % self.op.instance_name
3358

    
3359
    bep = self.cfg.GetClusterInfo().FillBE(instance)
3360
    if instance.disk_template not in constants.DTS_NET_MIRROR:
3361
      raise errors.OpPrereqError("Instance's disk layout is not"
3362
                                 " network mirrored, cannot failover.")
3363

    
3364
    secondary_nodes = instance.secondary_nodes
3365
    if not secondary_nodes:
3366
      raise errors.ProgrammerError("no secondary node but using "
3367
                                   "a mirrored disk template")
3368

    
3369
    target_node = secondary_nodes[0]
3370
    _CheckNodeOnline(self, target_node)
3371
    # check memory requirements on the secondary node
3372
    _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
3373
                         instance.name, bep[constants.BE_MEMORY],
3374
                         instance.hypervisor)
3375

    
3376
    # check bridge existance
3377
    brlist = [nic.bridge for nic in instance.nics]
3378
    result = self.rpc.call_bridges_exist(target_node, brlist)
3379
    result.Raise()
3380
    if not result.data:
3381
      raise errors.OpPrereqError("One or more target bridges %s does not"
3382
                                 " exist on destination node '%s'" %
3383
                                 (brlist, target_node))
3384

    
3385
  def Exec(self, feedback_fn):
3386
    """Failover an instance.
3387

3388
    The failover is done by shutting it down on its present node and
3389
    starting it on the secondary.
3390

3391
    """
3392
    instance = self.instance
3393

    
3394
    source_node = instance.primary_node
3395
    target_node = instance.secondary_nodes[0]
3396

    
3397
    feedback_fn("* checking disk consistency between source and target")
3398
    for dev in instance.disks:
3399
      # for drbd, these are drbd over lvm
3400
      if not _CheckDiskConsistency(self, dev, target_node, False):
3401
        if instance.admin_up and not self.op.ignore_consistency:
3402
          raise errors.OpExecError("Disk %s is degraded on target node,"
3403
                                   " aborting failover." % dev.iv_name)
3404

    
3405
    feedback_fn("* shutting down instance on source node")
3406
    logging.info("Shutting down instance %s on node %s",
3407
                 instance.name, source_node)
3408

    
3409
    result = self.rpc.call_instance_shutdown(source_node, instance)
3410
    if result.failed or not result.data:
3411
      if self.op.ignore_consistency:
3412
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
3413
                             " Proceeding"
3414
                             " anyway. Please make sure node %s is down",
3415
                             instance.name, source_node, source_node)
3416
      else:
3417
        raise errors.OpExecError("Could not shutdown instance %s on node %s" %
3418
                                 (instance.name, source_node))
3419

    
3420
    feedback_fn("* deactivating the instance's disks on source node")
3421
    if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
3422
      raise errors.OpExecError("Can't shut down the instance's disks.")
3423

    
3424
    instance.primary_node = target_node
3425
    # distribute new instance config to the other nodes
3426
    self.cfg.Update(instance)
3427

    
3428
    # Only start the instance if it's marked as up
3429
    if instance.admin_up:
3430
      feedback_fn("* activating the instance's disks on target node")
3431
      logging.info("Starting instance %s on node %s",
3432
                   instance.name, target_node)
3433

    
3434
      disks_ok, dummy = _AssembleInstanceDisks(self, instance,
3435
                                               ignore_secondaries=True)
3436
      if not disks_ok:
3437
        _ShutdownInstanceDisks(self, instance)
3438
        raise errors.OpExecError("Can't activate the instance's disks")
3439

    
3440
      feedback_fn("* starting the instance on the target node")
3441
      result = self.rpc.call_instance_start(target_node, instance, None)
3442
      msg = result.RemoteFailMsg()
3443
      if msg:
3444
        _ShutdownInstanceDisks(self, instance)
3445
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
3446
                                 (instance.name, target_node, msg))
3447

    
3448

    
3449
class LUMigrateInstance(LogicalUnit):
3450
  """Migrate an instance.
3451

3452
  This is migration without shutting down, compared to the failover,
3453
  which is done with shutdown.
3454

3455
  """
3456
  HPATH = "instance-migrate"
3457
  HTYPE = constants.HTYPE_INSTANCE
3458
  _OP_REQP = ["instance_name", "live", "cleanup"]
3459

    
3460
  REQ_BGL = False
3461

    
3462
  def ExpandNames(self):
3463
    self._ExpandAndLockInstance()
3464
    self.needed_locks[locking.LEVEL_NODE] = []
3465
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3466

    
3467
  def DeclareLocks(self, level):
3468
    if level == locking.LEVEL_NODE:
3469
      self._LockInstancesNodes()
3470

    
3471
  def BuildHooksEnv(self):
3472
    """Build hooks env.
3473

3474
    This runs on master, primary and secondary nodes of the instance.
3475

3476
    """
3477
    env = _BuildInstanceHookEnvByObject(self, self.instance)
3478
    nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
3479
    return env, nl, nl
3480

    
3481
  def CheckPrereq(self):
3482
    """Check prerequisites.
3483

3484
    This checks that the instance is in the cluster.
3485

3486
    """
3487
    instance = self.cfg.GetInstanceInfo(
3488
      self.cfg.ExpandInstanceName(self.op.instance_name))
3489
    if instance is None:
3490
      raise errors.OpPrereqError("Instance '%s' not known" %
3491
                                 self.op.instance_name)
3492

    
3493
    if instance.disk_template != constants.DT_DRBD8:
3494
      raise errors.OpPrereqError("Instance's disk layout is not"
3495
                                 " drbd8, cannot migrate.")
3496

    
3497
    secondary_nodes = instance.secondary_nodes
3498
    if not secondary_nodes:
3499
      raise errors.ProgrammerError("no secondary node but using "
3500
                                   "drbd8 disk template")
3501

    
3502
    i_be = self.cfg.GetClusterInfo().FillBE(instance)
3503

    
3504
    target_node = secondary_nodes[0]
3505
    # check memory requirements on the secondary node
3506
    _CheckNodeFreeMemory(self, target_node, "migrating instance %s" %
3507
                         instance.name, i_be[constants.BE_MEMORY],
3508
                         instance.hypervisor)
3509

    
3510
    # check bridge existance
3511
    brlist = [nic.bridge for nic in instance.nics]
3512
    result = self.rpc.call_bridges_exist(target_node, brlist)
3513
    if result.failed or not result.data:
3514
      raise errors.OpPrereqError("One or more target bridges %s does not"
3515
                                 " exist on destination node '%s'" %
3516
                                 (brlist, target_node))
3517

    
3518
    if not self.op.cleanup:
3519
      result = self.rpc.call_instance_migratable(instance.primary_node,
3520
                                                 instance)
3521
      msg = result.RemoteFailMsg()
3522
      if msg:
3523
        raise errors.OpPrereqError("Can't migrate: %s - please use failover" %
3524
                                   msg)
3525

    
3526
    self.instance = instance
3527

    
3528
  def _WaitUntilSync(self):
3529
    """Poll with custom rpc for disk sync.
3530

3531
    This uses our own step-based rpc call.
3532

3533
    """
3534
    self.feedback_fn("* wait until resync is done")
3535
    all_done = False
3536
    while not all_done:
3537
      all_done = True
3538
      result = self.rpc.call_drbd_wait_sync(self.all_nodes,
3539
                                            self.nodes_ip,
3540
                                            self.instance.disks)
3541
      min_percent = 100
3542
      for node, nres in result.items():
3543
        msg = nres.RemoteFailMsg()
3544
        if msg:
3545
          raise errors.OpExecError("Cannot resync disks on node %s: %s" %
3546
                                   (node, msg))
3547
        node_done, node_percent = nres.payload
3548
        all_done = all_done and node_done
3549
        if node_percent is not None:
3550
          min_percent = min(min_percent, node_percent)
3551
      if not all_done:
3552
        if min_percent < 100:
3553
          self.feedback_fn("   - progress: %.1f%%" % min_percent)
3554
        time.sleep(2)
3555

    
3556
  def _EnsureSecondary(self, node):
3557
    """Demote a node to secondary.
3558

3559
    """
3560
    self.feedback_fn("* switching node %s to secondary mode" % node)
3561

    
3562
    for dev in self.instance.disks:
3563
      self.cfg.SetDiskID(dev, node)
3564

    
3565
    result = self.rpc.call_blockdev_close(node, self.instance.name,
3566
                                          self.instance.disks)
3567
    msg = result.RemoteFailMsg()
3568
    if msg:
3569
      raise errors.OpExecError("Cannot change disk to secondary on node %s,"
3570
                               " error %s" % (node, msg))
3571

    
3572
  def _GoStandalone(self):
3573
    """Disconnect from the network.
3574

3575
    """
3576
    self.feedback_fn("* changing into standalone mode")
3577
    result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
3578
                                               self.instance.disks)
3579
    for node, nres in result.items():
3580
      msg = nres.RemoteFailMsg()
3581
      if msg:
3582
        raise errors.OpExecError("Cannot disconnect disks node %s,"
3583
                                 " error %s" % (node, msg))
3584

    
3585
  def _GoReconnect(self, multimaster):
3586
    """Reconnect to the network.
3587

3588
    """
3589
    if multimaster:
3590
      msg = "dual-master"
3591
    else:
3592
      msg = "single-master"
3593
    self.feedback_fn("* changing disks into %s mode" % msg)
3594
    result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
3595
                                           self.instance.disks,
3596
                                           self.instance.name, multimaster)
3597
    for node, nres in result.items():
3598
      msg = nres.RemoteFailMsg()
3599
      if msg:
3600
        raise errors.OpExecError("Cannot change disks config on node %s,"
3601
                                 " error: %s" % (node, msg))
3602

    
3603
  def _ExecCleanup(self):
3604
    """Try to cleanup after a failed migration.
3605

3606
    The cleanup is done by:
3607
      - check that the instance is running only on one node
3608
        (and update the config if needed)
3609
      - change disks on its secondary node to secondary
3610
      - wait until disks are fully synchronized
3611
      - disconnect from the network
3612
      - change disks into single-master mode
3613
      - wait again until disks are fully synchronized
3614

3615
    """
3616
    instance = self.instance
3617
    target_node = self.target_node
3618
    source_node = self.source_node
3619

    
3620
    # check running on only one node
3621
    self.feedback_fn("* checking where the instance actually runs"
3622
                     " (if this hangs, the hypervisor might be in"
3623
                     " a bad state)")
3624
    ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
3625
    for node, result in ins_l.items():
3626
      result.Raise()
3627
      if not isinstance(result.data, list):
3628
        raise errors.OpExecError("Can't contact node '%s'" % node)
3629

    
3630
    runningon_source = instance.name in ins_l[source_node].data
3631
    runningon_target = instance.name in ins_l[target_node].data
3632

    
3633
    if runningon_source and runningon_target:
3634
      raise errors.OpExecError("Instance seems to be running on two nodes,"
3635
                               " or the hypervisor is confused. You will have"
3636
                               " to ensure manually that it runs only on one"
3637
                               " and restart this operation.")
3638

    
3639
    if not (runningon_source or runningon_target):
3640
      raise errors.OpExecError("Instance does not seem to be running at all."
3641
                               " In this case, it's safer to repair by"
3642
                               " running 'gnt-instance stop' to ensure disk"
3643
                               " shutdown, and then restarting it.")
3644

    
3645
    if runningon_target:
3646
      # the migration has actually succeeded, we need to update the config
3647
      self.feedback_fn("* instance running on secondary node (%s),"
3648
                       " updating config" % target_node)
3649
      instance.primary_node = target_node
3650
      self.cfg.Update(instance)
3651
      demoted_node = source_node
3652
    else:
3653
      self.feedback_fn("* instance confirmed to be running on its"
3654
                       " primary node (%s)" % source_node)
3655
      demoted_node = target_node
3656

    
3657
    self._EnsureSecondary(demoted_node)
3658
    try:
3659
      self._WaitUntilSync()
3660
    except errors.OpExecError:
3661
      # we ignore here errors, since if the device is standalone, it
3662
      # won't be able to sync
3663
      pass
3664
    self._GoStandalone()
3665
    self._GoReconnect(False)
3666
    self._WaitUntilSync()
3667

    
3668
    self.feedback_fn("* done")
3669

    
3670
  def _RevertDiskStatus(self):
3671
    """Try to revert the disk status after a failed migration.
3672

3673
    """
3674
    target_node = self.target_node
3675
    try:
3676
      self._EnsureSecondary(target_node)
3677
      self._GoStandalone()
3678
      self._GoReconnect(False)
3679
      self._WaitUntilSync()
3680
    except errors.OpExecError, err:
3681
      self.LogWarning("Migration failed and I can't reconnect the"
3682
                      " drives: error '%s'\n"
3683
                      "Please look and recover the instance status" %
3684
                      str(err))
3685

    
3686
  def _AbortMigration(self):
3687
    """Call the hypervisor code to abort a started migration.
3688

3689
    """
3690
    instance = self.instance
3691
    target_node = self.target_node
3692
    migration_info = self.migration_info
3693

    
3694
    abort_result = self.rpc.call_finalize_migration(target_node,
3695
                                                    instance,
3696
                                                    migration_info,
3697
                                                    False)
3698
    abort_msg = abort_result.RemoteFailMsg()
3699
    if abort_msg:
3700
      logging.error("Aborting migration failed on target node %s: %s" %
3701
                    (target_node, abort_msg))
3702
      # Don't raise an exception here, as we stil have to try to revert the
3703
      # disk status, even if this step failed.
3704

    
3705
  def _ExecMigration(self):
3706
    """Migrate an instance.
3707

3708
    The migrate is done by:
3709
      - change the disks into dual-master mode
3710
      - wait until disks are fully synchronized again
3711
      - migrate the instance
3712
      - change disks on the new secondary node (the old primary) to secondary
3713
      - wait until disks are fully synchronized
3714
      - change disks into single-master mode
3715

3716
    """
3717
    instance = self.instance
3718
    target_node = self.target_node
3719
    source_node = self.source_node
3720

    
3721
    self.feedback_fn("* checking disk consistency between source and target")
3722
    for dev in instance.disks:
3723
      if not _CheckDiskConsistency(self, dev, target_node, False):
3724
        raise errors.OpExecError("Disk %s is degraded or not fully"
3725
                                 " synchronized on target node,"
3726
                                 " aborting migrate." % dev.iv_name)
3727

    
3728
    # First get the migration information from the remote node
3729
    result = self.rpc.call_migration_info(source_node, instance)
3730
    msg = result.RemoteFailMsg()
3731
    if msg:
3732
      log_err = ("Failed fetching source migration information from %s: %s" %
3733
                 (source_node, msg))
3734
      logging.error(log_err)
3735
      raise errors.OpExecError(log_err)
3736

    
3737
    self.migration_info = migration_info = result.payload
3738

    
3739
    # Then switch the disks to master/master mode
3740
    self._EnsureSecondary(target_node)
3741
    self._GoStandalone()
3742
    self._GoReconnect(True)
3743
    self._WaitUntilSync()
3744

    
3745
    self.feedback_fn("* preparing %s to accept the instance" % target_node)
3746
    result = self.rpc.call_accept_instance(target_node,
3747
                                           instance,
3748
                                           migration_info,
3749
                                           self.nodes_ip[target_node])
3750

    
3751
    msg = result.RemoteFailMsg()
3752
    if msg:
3753
      logging.error("Instance pre-migration failed, trying to revert"
3754
                    " disk status: %s", msg)
3755
      self._AbortMigration()
3756
      self._RevertDiskStatus()
3757
      raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
3758
                               (instance.name, msg))
3759

    
3760
    self.feedback_fn("* migrating instance to %s" % target_node)
3761
    time.sleep(10)
3762
    result = self.rpc.call_instance_migrate(source_node, instance,
3763
                                            self.nodes_ip[target_node],
3764
                                            self.op.live)
3765
    msg = result.RemoteFailMsg()
3766
    if msg:
3767
      logging.error("Instance migration failed, trying to revert"
3768
                    " disk status: %s", msg)
3769
      self._AbortMigration()
3770
      self._RevertDiskStatus()
3771
      raise errors.OpExecError("Could not migrate instance %s: %s" %
3772
                               (instance.name, msg))
3773
    time.sleep(10)
3774

    
3775
    instance.primary_node = target_node
3776
    # distribute new instance config to the other nodes
3777
    self.cfg.Update(instance)
3778

    
3779
    result = self.rpc.call_finalize_migration(target_node,
3780
                                              instance,
3781
                                              migration_info,
3782
                                              True)
3783
    msg = result.RemoteFailMsg()
3784
    if msg:
3785
      logging.error("Instance migration succeeded, but finalization failed:"
3786
                    " %s" % msg)
3787
      raise errors.OpExecError("Could not finalize instance migration: %s" %
3788
                               msg)
3789

    
3790
    self._EnsureSecondary(source_node)
3791
    self._WaitUntilSync()
3792
    self._GoStandalone()
3793
    self._GoReconnect(False)
3794
    self._WaitUntilSync()
3795

    
3796
    self.feedback_fn("* done")
3797

    
3798
  def Exec(self, feedback_fn):
3799
    """Perform the migration.
3800

3801
    """
3802
    self.feedback_fn = feedback_fn
3803

    
3804
    self.source_node = self.instance.primary_node
3805
    self.target_node = self.instance.secondary_nodes[0]
3806
    self.all_nodes = [self.source_node, self.target_node]
3807
    self.nodes_ip = {
3808
      self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
3809
      self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
3810
      }
3811
    if self.op.cleanup:
3812
      return self._ExecCleanup()
3813
    else:
3814
      return self._ExecMigration()
3815

    
3816

    
3817
def _CreateBlockDev(lu, node, instance, device, force_create,
3818
                    info, force_open):
3819
  """Create a tree of block devices on a given node.
3820

3821
  If this device type has to be created on secondaries, create it and
3822
  all its children.
3823

3824
  If not, just recurse to children keeping the same 'force' value.
3825

3826
  @param lu: the lu on whose behalf we execute
3827
  @param node: the node on which to create the device
3828
  @type instance: L{objects.Instance}
3829
  @param instance: the instance which owns the device
3830
  @type device: L{objects.Disk}
3831
  @param device: the device to create
3832
  @type force_create: boolean
3833
  @param force_create: whether to force creation of this device; this
3834
      will be change to True whenever we find a device which has
3835
      CreateOnSecondary() attribute
3836
  @param info: the extra 'metadata' we should attach to the device
3837
      (this will be represented as a LVM tag)
3838
  @type force_open: boolean
3839
  @param force_open: this parameter will be passes to the
3840
      L{backend.BlockdevCreate} function where it specifies
3841
      whether we run on primary or not, and it affects both
3842
      the child assembly and the device own Open() execution
3843

3844
  """
3845
  if device.CreateOnSecondary():
3846
    force_create = True
3847

    
3848
  if device.children:
3849
    for child in device.children:
3850
      _CreateBlockDev(lu, node, instance, child, force_create,
3851
                      info, force_open)
3852

    
3853
  if not force_create:
3854
    return
3855

    
3856
  _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
3857

    
3858

    
3859
def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
3860
  """Create a single block device on a given node.
3861

3862
  This will not recurse over children of the device, so they must be
3863
  created in advance.
3864

3865
  @param lu: the lu on whose behalf we execute
3866
  @param node: the node on which to create the device
3867
  @type instance: L{objects.Instance}
3868
  @param instance: the instance which owns the device
3869
  @type device: L{objects.Disk}
3870
  @param device: the device to create
3871
  @param info: the extra 'metadata' we should attach to the device
3872
      (this will be represented as a LVM tag)
3873
  @type force_open: boolean
3874
  @param force_open: this parameter will be passes to the
3875
      L{backend.BlockdevCreate} function where it specifies
3876
      whether we run on primary or not, and it affects both
3877
      the child assembly and the device own Open() execution
3878

3879
  """
3880
  lu.cfg.SetDiskID(device, node)
3881
  result = lu.rpc.call_blockdev_create(node, device, device.size,
3882
                                       instance.name, force_open, info)
3883
  msg = result.RemoteFailMsg()
3884
  if msg:
3885
    raise errors.OpExecError("Can't create block device %s on"
3886
                             " node %s for instance %s: %s" %
3887
                             (device, node, instance.name, msg))
3888
  if device.physical_id is None:
3889
    device.physical_id = result.payload
3890

    
3891

    
3892
def _GenerateUniqueNames(lu, exts):
3893
  """Generate a suitable LV name.
3894

3895
  This will generate a logical volume name for the given instance.
3896

3897
  """
3898
  results = []
3899
  for val in exts:
3900
    new_id = lu.cfg.GenerateUniqueID()
3901
    results.append("%s%s" % (new_id, val))
3902
  return results
3903

    
3904

    
3905
def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
3906
                         p_minor, s_minor):
3907
  """Generate a drbd8 device complete with its children.
3908

3909
  """
3910
  port = lu.cfg.AllocatePort()
3911
  vgname = lu.cfg.GetVGName()
3912
  shared_secret = lu.cfg.GenerateDRBDSecret()
3913
  dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
3914
                          logical_id=(vgname, names[0]))
3915
  dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
3916
                          logical_id=(vgname, names[1]))
3917
  drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
3918
                          logical_id=(primary, secondary, port,
3919
                                      p_minor, s_minor,
3920
                                      shared_secret),
3921
                          children=[dev_data, dev_meta],
3922
                          iv_name=iv_name)
3923
  return drbd_dev
3924

    
3925

    
3926
def _GenerateDiskTemplate(lu, template_name,
3927
                          instance_name, primary_node,
3928
                          secondary_nodes, disk_info,
3929
                          file_storage_dir, file_driver,
3930
                          base_index):
3931
  """Generate the entire disk layout for a given template type.
3932

3933
  """
3934
  #TODO: compute space requirements
3935

    
3936
  vgname = lu.cfg.GetVGName()
3937
  disk_count = len(disk_info)
3938
  disks = []
3939
  if template_name == constants.DT_DISKLESS:
3940
    pass
3941
  elif template_name == constants.DT_PLAIN:
3942
    if len(secondary_nodes) != 0:
3943
      raise errors.ProgrammerError("Wrong template configuration")
3944

    
3945
    names = _GenerateUniqueNames(lu, [".disk%d" % i
3946
                                      for i in range(disk_count)])
3947
    for idx, disk in enumerate(disk_info):
3948
      disk_index = idx + base_index
3949
      disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
3950
                              logical_id=(vgname, names[idx]),
3951
                              iv_name="disk/%d" % disk_index,
3952
                              mode=disk["mode"])
3953
      disks.append(disk_dev)
3954
  elif template_name == constants.DT_DRBD8:
3955
    if len(secondary_nodes) != 1:
3956
      raise errors.ProgrammerError("Wrong template configuration")
3957
    remote_node = secondary_nodes[0]
3958
    minors = lu.cfg.AllocateDRBDMinor(
3959
      [primary_node, remote_node] * len(disk_info), instance_name)
3960

    
3961
    names = []
3962
    for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % i
3963
                                               for i in range(disk_count)]):
3964
      names.append(lv_prefix + "_data")
3965
      names.append(lv_prefix + "_meta")
3966
    for idx, disk in enumerate(disk_info):
3967
      disk_index = idx + base_index
3968
      disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
3969
                                      disk["size"], names[idx*2:idx*2+2],
3970
                                      "disk/%d" % disk_index,
3971
                                      minors[idx*2], minors[idx*2+1])
3972
      disk_dev.mode = disk["mode"]
3973
      disks.append(disk_dev)
3974
  elif template_name == constants.DT_FILE:
3975
    if len(secondary_nodes) != 0:
3976
      raise errors.ProgrammerError("Wrong template configuration")
3977

    
3978
    for idx, disk in enumerate(disk_info):
3979
      disk_index = idx + base_index
3980
      disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
3981
                              iv_name="disk/%d" % disk_index,
3982
                              logical_id=(file_driver,
3983
                                          "%s/disk%d" % (file_storage_dir,
3984
                                                         idx)),
3985
                              mode=disk["mode"])
3986
      disks.append(disk_dev)
3987
  else:
3988
    raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
3989
  return disks
3990

    
3991

    
3992
def _GetInstanceInfoText(instance):
3993
  """Compute that text that should be added to the disk's metadata.
3994

3995
  """
3996
  return "originstname+%s" % instance.name
3997

    
3998

    
3999
def _CreateDisks(lu, instance):
4000
  """Create all disks for an instance.
4001

4002
  This abstracts away some work from AddInstance.
4003

4004
  @type lu: L{LogicalUnit}
4005
  @param lu: the logical unit on whose behalf we execute
4006
  @type instance: L{objects.Instance}
4007
  @param instance: the instance whose disks we should create
4008
  @rtype: boolean
4009
  @return: the success of the creation
4010

4011
  """
4012
  info = _GetInstanceInfoText(instance)
4013
  pnode = instance.primary_node
4014

    
4015
  if instance.disk_template == constants.DT_FILE:
4016
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
4017
    result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
4018

    
4019
    if result.failed or not result.data:
4020
      raise errors.OpExecError("Could not connect to node '%s'" % pnode)
4021

    
4022
    if not result.data[0]:
4023
      raise errors.OpExecError("Failed to create directory '%s'" %
4024
                               file_storage_dir)
4025

    
4026
  # Note: this needs to be kept in sync with adding of disks in
4027
  # LUSetInstanceParams
4028
  for device in instance.disks:
4029
    logging.info("Creating volume %s for instance %s",
4030
                 device.iv_name, instance.name)
4031
    #HARDCODE
4032
    for node in instance.all_nodes:
4033
      f_create = node == pnode
4034
      _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
4035

    
4036

    
4037
def _RemoveDisks(lu, instance):
4038
  """Remove all disks for an instance.
4039

4040
  This abstracts away some work from `AddInstance()` and
4041
  `RemoveInstance()`. Note that in case some of the devices couldn't
4042
  be removed, the removal will continue with the other ones (compare
4043
  with `_CreateDisks()`).
4044

4045
  @type lu: L{LogicalUnit}
4046
  @param lu: the logical unit on whose behalf we execute
4047
  @type instance: L{objects.Instance}
4048
  @param instance: the instance whose disks we should remove
4049
  @rtype: boolean
4050
  @return: the success of the removal
4051

4052
  """
4053
  logging.info("Removing block devices for instance %s", instance.name)
4054

    
4055
  result = True
4056
  for device in instance.disks:
4057
    for node, disk in device.ComputeNodeTree(instance.primary_node):
4058
      lu.cfg.SetDiskID(disk, node)
4059
      result = lu.rpc.call_blockdev_remove(node, disk)
4060
      if result.failed or not result.data:
4061
        lu.proc.LogWarning("Could not remove block device %s on node %s,"
4062
                           " continuing anyway", device.iv_name, node)
4063
        result = False
4064

    
4065
  if instance.disk_template == constants.DT_FILE:
4066
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
4067
    result = lu.rpc.call_file_storage_dir_remove(instance.primary_node,
4068
                                                 file_storage_dir)
4069
    if result.failed or not result.data:
4070
      logging.error("Could not remove directory '%s'", file_storage_dir)
4071
      result = False
4072

    
4073
  return result
4074

    
4075

    
4076
def _ComputeDiskSize(disk_template, disks):
4077
  """Compute disk size requirements in the volume group
4078

4079
  """
4080
  # Required free disk space as a function of disk and swap space
4081
  req_size_dict = {
4082
    constants.DT_DISKLESS: None,
4083
    constants.DT_PLAIN: sum(d["size"] for d in disks),
4084
    # 128 MB are added for drbd metadata for each disk
4085
    constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
4086
    constants.DT_FILE: None,
4087
  }
4088

    
4089
  if disk_template not in req_size_dict:
4090
    raise errors.ProgrammerError("Disk template '%s' size requirement"
4091
                                 " is unknown" %  disk_template)
4092

    
4093
  return req_size_dict[disk_template]
4094

    
4095

    
4096
def _CheckHVParams(lu, nodenames, hvname, hvparams):
4097
  """Hypervisor parameter validation.
4098

4099
  This function abstract the hypervisor parameter validation to be
4100
  used in both instance create and instance modify.
4101

4102
  @type lu: L{LogicalUnit}
4103
  @param lu: the logical unit for which we check
4104
  @type nodenames: list
4105
  @param nodenames: the list of nodes on which we should check
4106
  @type hvname: string
4107
  @param hvname: the name of the hypervisor we should use
4108
  @type hvparams: dict
4109
  @param hvparams: the parameters which we need to check
4110
  @raise errors.OpPrereqError: if the parameters are not valid
4111

4112
  """
4113
  hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
4114
                                                  hvname,
4115
                                                  hvparams)
4116
  for node in nodenames:
4117
    info = hvinfo[node]
4118
    if info.offline:
4119
      continue
4120
    msg = info.RemoteFailMsg()
4121
    if msg:
4122
      raise errors.OpPrereqError("Hypervisor parameter validation failed:"
4123
                                 " %s" % msg)
4124

    
4125

    
4126
class LUCreateInstance(LogicalUnit):
4127
  """Create an instance.
4128

4129
  """
4130
  HPATH = "instance-add"
4131
  HTYPE = constants.HTYPE_INSTANCE
4132
  _OP_REQP = ["instance_name", "disks", "disk_template",
4133
              "mode", "start",
4134
              "wait_for_sync", "ip_check", "nics",
4135
              "hvparams", "beparams"]
4136
  REQ_BGL = False
4137

    
4138
  def _ExpandNode(self, node):
4139
    """Expands and checks one node name.
4140

4141
    """
4142
    node_full = self.cfg.ExpandNodeName(node)
4143
    if node_full is None:
4144
      raise errors.OpPrereqError("Unknown node %s" % node)
4145
    return node_full
4146

    
4147
  def ExpandNames(self):
4148
    """ExpandNames for CreateInstance.
4149

4150
    Figure out the right locks for instance creation.
4151

4152
    """
4153
    self.needed_locks = {}
4154

    
4155
    # set optional parameters to none if they don't exist
4156
    for attr in ["pnode", "snode", "iallocator", "hypervisor"]:
4157
      if not hasattr(self.op, attr):
4158
        setattr(self.op, attr, None)
4159

    
4160
    # cheap checks, mostly valid constants given
4161

    
4162
    # verify creation mode
4163
    if self.op.mode not in (constants.INSTANCE_CREATE,
4164
                            constants.INSTANCE_IMPORT):
4165
      raise errors.OpPrereqError("Invalid instance creation mode '%s'" %
4166
                                 self.op.mode)
4167

    
4168
    # disk template and mirror node verification
4169
    if self.op.disk_template not in constants.DISK_TEMPLATES:
4170
      raise errors.OpPrereqError("Invalid disk template name")
4171

    
4172
    if self.op.hypervisor is None:
4173
      self.op.hypervisor = self.cfg.GetHypervisorType()
4174

    
4175
    cluster = self.cfg.GetClusterInfo()
4176
    enabled_hvs = cluster.enabled_hypervisors
4177
    if self.op.hypervisor not in enabled_hvs:
4178
      raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
4179
                                 " cluster (%s)" % (self.op.hypervisor,
4180
                                  ",".join(enabled_hvs)))
4181

    
4182
    # check hypervisor parameter syntax (locally)
4183

    
4184
    filled_hvp = cluster.FillDict(cluster.hvparams[self.op.hypervisor],
4185
                                  self.op.hvparams)
4186
    hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
4187
    hv_type.CheckParameterSyntax(filled_hvp)
4188

    
4189
    # fill and remember the beparams dict
4190
    utils.CheckBEParams(self.op.beparams)
4191
    self.be_full = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
4192
                                    self.op.beparams)
4193

    
4194
    #### instance parameters check
4195

    
4196
    # instance name verification
4197
    hostname1 = utils.HostInfo(self.op.instance_name)
4198
    self.op.instance_name = instance_name = hostname1.name
4199

    
4200
    # this is just a preventive check, but someone might still add this
4201
    # instance in the meantime, and creation will fail at lock-add time
4202
    if instance_name in self.cfg.GetInstanceList():
4203
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
4204
                                 instance_name)
4205

    
4206
    self.add_locks[locking.LEVEL_INSTANCE] = instance_name
4207

    
4208
    # NIC buildup
4209
    self.nics = []
4210
    for nic in self.op.nics:
4211
      # ip validity checks
4212
      ip = nic.get("ip", None)
4213
      if ip is None or ip.lower() == "none":
4214
        nic_ip = None
4215
      elif ip.lower() == constants.VALUE_AUTO:
4216
        nic_ip = hostname1.ip
4217
      else:
4218
        if not utils.IsValidIP(ip):
4219
          raise errors.OpPrereqError("Given IP address '%s' doesn't look"
4220
                                     " like a valid IP" % ip)
4221
        nic_ip = ip
4222

    
4223
      # MAC address verification
4224
      mac = nic.get("mac", constants.VALUE_AUTO)
4225
      if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
4226
        if not utils.IsValidMac(mac.lower()):
4227
          raise errors.OpPrereqError("Invalid MAC address specified: %s" %
4228
                                     mac)
4229
      # bridge verification
4230
      bridge = nic.get("bridge", None)
4231
      if bridge is None:
4232
        bridge = self.cfg.GetDefBridge()
4233
      self.nics.append(objects.NIC(mac=mac, ip=nic_ip, bridge=bridge))
4234

    
4235
    # disk checks/pre-build
4236
    self.disks = []
4237
    for disk in self.op.disks:
4238
      mode = disk.get("mode", constants.DISK_RDWR)
4239
      if mode not in constants.DISK_ACCESS_SET:
4240
        raise errors.OpPrereqError("Invalid disk access mode '%s'" %
4241
                                   mode)
4242
      size = disk.get("size", None)
4243
      if size is None:
4244
        raise errors.OpPrereqError("Missing disk size")
4245
      try:
4246
        size = int(size)
4247
      except ValueError:
4248
        raise errors.OpPrereqError("Invalid disk size '%s'" % size)
4249
      self.disks.append({"size": size, "mode": mode})
4250

    
4251
    # used in CheckPrereq for ip ping check
4252
    self.check_ip = hostname1.ip
4253

    
4254
    # file storage checks
4255
    if (self.op.file_driver and
4256
        not self.op.file_driver in constants.FILE_DRIVER):
4257
      raise errors.OpPrereqError("Invalid file driver name '%s'" %
4258
                                 self.op.file_driver)
4259

    
4260
    if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
4261
      raise errors.OpPrereqError("File storage directory path not absolute")
4262

    
4263
    ### Node/iallocator related checks
4264
    if [self.op.iallocator, self.op.pnode].count(None) != 1:
4265
      raise errors.OpPrereqError("One and only one of iallocator and primary"
4266
                                 " node must be given")
4267

    
4268
    if self.op.iallocator:
4269
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4270
    else:
4271
      self.op.pnode = self._ExpandNode(self.op.pnode)
4272
      nodelist = [self.op.pnode]
4273
      if self.op.snode is not None:
4274
        self.op.snode = self._ExpandNode(self.op.snode)
4275
        nodelist.append(self.op.snode)
4276
      self.needed_locks[locking.LEVEL_NODE] = nodelist
4277

    
4278
    # in case of import lock the source node too
4279
    if self.op.mode == constants.INSTANCE_IMPORT:
4280
      src_node = getattr(self.op, "src_node", None)
4281
      src_path = getattr(self.op, "src_path", None)
4282

    
4283
      if src_path is None:
4284
        self.op.src_path = src_path = self.op.instance_name
4285

    
4286
      if src_node is None:
4287
        self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4288
        self.op.src_node = None
4289
        if os.path.isabs(src_path):
4290
          raise errors.OpPrereqError("Importing an instance from an absolute"
4291
                                     " path requires a source node option.")
4292
      else:
4293
        self.op.src_node = src_node = self._ExpandNode(src_node)
4294
        if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
4295
          self.needed_locks[locking.LEVEL_NODE].append(src_node)
4296
        if not os.path.isabs(src_path):
4297
          self.op.src_path = src_path = \
4298
            os.path.join(constants.EXPORT_DIR, src_path)
4299

    
4300
    else: # INSTANCE_CREATE
4301
      if getattr(self.op, "os_type", None) is None:
4302
        raise errors.OpPrereqError("No guest OS specified")
4303

    
4304
  def _RunAllocator(self):
4305
    """Run the allocator based on input opcode.
4306

4307
    """
4308
    nics = [n.ToDict() for n in self.nics]
4309
    ial = IAllocator(self,
4310
                     mode=constants.IALLOCATOR_MODE_ALLOC,
4311
                     name=self.op.instance_name,
4312
                     disk_template=self.op.disk_template,
4313
                     tags=[],
4314
                     os=self.op.os_type,
4315
                     vcpus=self.be_full[constants.BE_VCPUS],
4316
                     mem_size=self.be_full[constants.BE_MEMORY],
4317
                     disks=self.disks,
4318
                     nics=nics,
4319
                     hypervisor=self.op.hypervisor,
4320
                     )
4321

    
4322
    ial.Run(self.op.iallocator)
4323

    
4324
    if not ial.success:
4325
      raise errors.OpPrereqError("Can't compute nodes using"
4326
                                 " iallocator '%s': %s" % (self.op.iallocator,
4327
                                                           ial.info))
4328
    if len(ial.nodes) != ial.required_nodes:
4329
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
4330
                                 " of nodes (%s), required %s" %
4331
                                 (self.op.iallocator, len(ial.nodes),
4332
                                  ial.required_nodes))
4333
    self.op.pnode = ial.nodes[0]
4334
    self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
4335
                 self.op.instance_name, self.op.iallocator,
4336
                 ", ".join(ial.nodes))
4337
    if ial.required_nodes == 2:
4338
      self.op.snode = ial.nodes[1]
4339

    
4340
  def BuildHooksEnv(self):
4341
    """Build hooks env.
4342

4343
    This runs on master, primary and secondary nodes of the instance.
4344

4345
    """
4346
    env = {
4347
      "INSTANCE_DISK_TEMPLATE": self.op.disk_template,
4348
      "INSTANCE_DISK_SIZE": ",".join(str(d["size"]) for d in self.disks),
4349
      "INSTANCE_ADD_MODE": self.op.mode,
4350
      }
4351
    if self.op.mode == constants.INSTANCE_IMPORT:
4352
      env["INSTANCE_SRC_NODE"] = self.op.src_node
4353
      env["INSTANCE_SRC_PATH"] = self.op.src_path
4354
      env["INSTANCE_SRC_IMAGES"] = self.src_images
4355

    
4356
    env.update(_BuildInstanceHookEnv(name=self.op.instance_name,
4357
      primary_node=self.op.pnode,
4358
      secondary_nodes=self.secondaries,
4359
      status=self.op.start,
4360
      os_type=self.op.os_type,
4361
      memory=self.be_full[constants.BE_MEMORY],
4362
      vcpus=self.be_full[constants.BE_VCPUS],
4363
      nics=[(n.ip, n.bridge, n.mac) for n in self.nics],
4364
    ))
4365

    
4366
    nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
4367
          self.secondaries)
4368
    return env, nl, nl
4369

    
4370

    
4371
  def CheckPrereq(self):
4372
    """Check prerequisites.
4373

4374
    """
4375
    if (not self.cfg.GetVGName() and
4376
        self.op.disk_template not in constants.DTS_NOT_LVM):
4377
      raise errors.OpPrereqError("Cluster does not support lvm-based"
4378
                                 " instances")
4379

    
4380

    
4381
    if self.op.mode == constants.INSTANCE_IMPORT:
4382
      src_node = self.op.src_node
4383
      src_path = self.op.src_path
4384

    
4385
      if src_node is None:
4386
        exp_list = self.rpc.call_export_list(
4387
          self.acquired_locks[locking.LEVEL_NODE])
4388
        found = False
4389
        for node in exp_list:
4390
          if not exp_list[node].failed and src_path in exp_list[node].data:
4391
            found = True
4392
            self.op.src_node = src_node = node
4393
            self.op.src_path = src_path = os.path.join(constants.EXPORT_DIR,
4394
                                                       src_path)
4395
            break
4396
        if not found:
4397
          raise errors.OpPrereqError("No export found for relative path %s" %
4398
                                      src_path)
4399

    
4400
      _CheckNodeOnline(self, src_node)
4401
      result = self.rpc.call_export_info(src_node, src_path)
4402
      result.Raise()
4403
      if not result.data:
4404
        raise errors.OpPrereqError("No export found in dir %s" % src_path)
4405

    
4406
      export_info = result.data
4407
      if not export_info.has_section(constants.INISECT_EXP):
4408
        raise errors.ProgrammerError("Corrupted export config")
4409

    
4410
      ei_version = export_info.get(constants.INISECT_EXP, 'version')
4411
      if (int(ei_version) != constants.EXPORT_VERSION):
4412
        raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
4413
                                   (ei_version, constants.EXPORT_VERSION))
4414

    
4415
      # Check that the new instance doesn't have less disks than the export
4416
      instance_disks = len(self.disks)
4417
      export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
4418
      if instance_disks < export_disks:
4419
        raise errors.OpPrereqError("Not enough disks to import."
4420
                                   " (instance: %d, export: %d)" %
4421
                                   (instance_disks, export_disks))
4422

    
4423
      self.op.os_type = export_info.get(constants.INISECT_EXP, 'os')
4424
      disk_images = []
4425
      for idx in range(export_disks):
4426
        option = 'disk%d_dump' % idx
4427
        if export_info.has_option(constants.INISECT_INS, option):
4428
          # FIXME: are the old os-es, disk sizes, etc. useful?
4429
          export_name = export_info.get(constants.INISECT_INS, option)
4430
          image = os.path.join(src_path, export_name)
4431
          disk_images.append(image)
4432
        else:
4433
          disk_images.append(False)
4434

    
4435
      self.src_images = disk_images
4436

    
4437
      old_name = export_info.get(constants.INISECT_INS, 'name')
4438
      # FIXME: int() here could throw a ValueError on broken exports
4439
      exp_nic_count = int(export_info.get(constants.INISECT_INS, 'nic_count'))
4440
      if self.op.instance_name == old_name:
4441
        for idx, nic in enumerate(self.nics):
4442
          if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
4443
            nic_mac_ini = 'nic%d_mac' % idx
4444
            nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
4445

    
4446
    # ip ping checks (we use the same ip that was resolved in ExpandNames)
4447
    if self.op.start and not self.op.ip_check:
4448
      raise errors.OpPrereqError("Cannot ignore IP address conflicts when"
4449
                                 " adding an instance in start mode")
4450

    
4451
    if self.op.ip_check:
4452
      if utils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
4453
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
4454
                                   (self.check_ip, self.op.instance_name))
4455

    
4456
    #### allocator run
4457

    
4458
    if self.op.iallocator is not None:
4459
      self._RunAllocator()
4460

    
4461
    #### node related checks
4462

    
4463
    # check primary node
4464
    self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
4465
    assert self.pnode is not None, \
4466
      "Cannot retrieve locked node %s" % self.op.pnode
4467
    if pnode.offline:
4468
      raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
4469
                                 pnode.name)
4470

    
4471
    self.secondaries = []
4472

    
4473
    # mirror node verification
4474
    if self.op.disk_template in constants.DTS_NET_MIRROR:
4475
      if self.op.snode is None:
4476
        raise errors.OpPrereqError("The networked disk templates need"
4477
                                   " a mirror node")
4478
      if self.op.snode == pnode.name:
4479
        raise errors.OpPrereqError("The secondary node cannot be"
4480
                                   " the primary node.")
4481
      self.secondaries.append(self.op.snode)
4482
      _CheckNodeOnline(self, self.op.snode)
4483

    
4484
    nodenames = [pnode.name] + self.secondaries
4485

    
4486
    req_size = _ComputeDiskSize(self.op.disk_template,
4487
                                self.disks)
4488

    
4489
    # Check lv size requirements
4490
    if req_size is not None:
4491
      nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
4492
                                         self.op.hypervisor)
4493
      for node in nodenames:
4494
        info = nodeinfo[node]
4495
        info.Raise()
4496
        info = info.data
4497
        if not info:
4498
          raise errors.OpPrereqError("Cannot get current information"
4499
                                     " from node '%s'" % node)
4500
        vg_free = info.get('vg_free', None)
4501
        if not isinstance(vg_free, int):
4502
          raise errors.OpPrereqError("Can't compute free disk space on"
4503
                                     " node %s" % node)
4504
        if req_size > info['vg_free']:
4505
          raise errors.OpPrereqError("Not enough disk space on target node %s."
4506
                                     " %d MB available, %d MB required" %
4507
                                     (node, info['vg_free'], req_size))
4508

    
4509
    _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
4510

    
4511
    # os verification
4512
    result = self.rpc.call_os_get(pnode.name, self.op.os_type)
4513
    result.Raise()
4514
    if not isinstance(result.data, objects.OS):
4515
      raise errors.OpPrereqError("OS '%s' not in supported os list for"
4516
                                 " primary node"  % self.op.os_type)
4517

    
4518
    # bridge check on primary node
4519
    bridges = [n.bridge for n in self.nics]
4520
    result = self.rpc.call_bridges_exist(self.pnode.name, bridges)
4521
    result.Raise()
4522
    if not result.data:
4523
      raise errors.OpPrereqError("One of the target bridges '%s' does not"
4524
                                 " exist on destination node '%s'" %
4525
                                 (",".join(bridges), pnode.name))
4526

    
4527
    # memory check on primary node
4528
    if self.op.start:
4529
      _CheckNodeFreeMemory(self, self.pnode.name,
4530
                           "creating instance %s" % self.op.instance_name,
4531
                           self.be_full[constants.BE_MEMORY],
4532
                           self.op.hypervisor)
4533

    
4534
  def Exec(self, feedback_fn):
4535
    """Create and add the instance to the cluster.
4536

4537
    """
4538
    instance = self.op.instance_name
4539
    pnode_name = self.pnode.name
4540

    
4541
    for nic in self.nics:
4542
      if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
4543
        nic.mac = self.cfg.GenerateMAC()
4544

    
4545
    ht_kind = self.op.hypervisor
4546
    if ht_kind in constants.HTS_REQ_PORT:
4547
      network_port = self.cfg.AllocatePort()
4548
    else:
4549
      network_port = None
4550

    
4551
    ##if self.op.vnc_bind_address is None:
4552
    ##  self.op.vnc_bind_address = constants.VNC_DEFAULT_BIND_ADDRESS
4553

    
4554
    # this is needed because os.path.join does not accept None arguments
4555
    if self.op.file_storage_dir is None:
4556
      string_file_storage_dir = ""
4557
    else:
4558
      string_file_storage_dir = self.op.file_storage_dir
4559

    
4560
    # build the full file storage dir path
4561
    file_storage_dir = os.path.normpath(os.path.join(
4562
                                        self.cfg.GetFileStorageDir(),
4563
                                        string_file_storage_dir, instance))
4564

    
4565

    
4566
    disks = _GenerateDiskTemplate(self,
4567
                                  self.op.disk_template,
4568
                                  instance, pnode_name,
4569
                                  self.secondaries,
4570
                                  self.disks,
4571
                                  file_storage_dir,
4572
                                  self.op.file_driver,
4573
                                  0)
4574

    
4575
    iobj = objects.Instance(name=instance, os=self.op.os_type,
4576
                            primary_node=pnode_name,
4577
                            nics=self.nics, disks=disks,
4578
                            disk_template=self.op.disk_template,
4579
                            admin_up=False,
4580
                            network_port=network_port,
4581
                            beparams=self.op.beparams,
4582
                            hvparams=self.op.hvparams,
4583
                            hypervisor=self.op.hypervisor,
4584
                            )
4585

    
4586
    feedback_fn("* creating instance disks...")
4587
    try:
4588
      _CreateDisks(self, iobj)
4589
    except errors.OpExecError:
4590
      self.LogWarning("Device creation failed, reverting...")
4591
      try:
4592
        _RemoveDisks(self, iobj)
4593
      finally:
4594
        self.cfg.ReleaseDRBDMinors(instance)
4595
        raise
4596

    
4597
    feedback_fn("adding instance %s to cluster config" % instance)
4598

    
4599
    self.cfg.AddInstance(iobj)
4600
    # Declare that we don't want to remove the instance lock anymore, as we've
4601
    # added the instance to the config
4602
    del self.remove_locks[locking.LEVEL_INSTANCE]
4603
    # Unlock all the nodes
4604
    if self.op.mode == constants.INSTANCE_IMPORT:
4605
      nodes_keep = [self.op.src_node]
4606
      nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
4607
                       if node != self.op.src_node]
4608
      self.context.glm.release(locking.LEVEL_NODE, nodes_release)
4609
      self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
4610
    else:
4611
      self.context.glm.release(locking.LEVEL_NODE)
4612
      del self.acquired_locks[locking.LEVEL_NODE]
4613

    
4614
    if self.op.wait_for_sync:
4615
      disk_abort = not _WaitForSync(self, iobj)
4616
    elif iobj.disk_template in constants.DTS_NET_MIRROR:
4617
      # make sure the disks are not degraded (still sync-ing is ok)
4618
      time.sleep(15)
4619
      feedback_fn("* checking mirrors status")
4620
      disk_abort = not _WaitForSync(self, iobj, oneshot=True)
4621
    else:
4622
      disk_abort = False
4623

    
4624
    if disk_abort:
4625
      _RemoveDisks(self, iobj)
4626
      self.cfg.RemoveInstance(iobj.name)
4627
      # Make sure the instance lock gets removed
4628
      self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
4629
      raise errors.OpExecError("There are some degraded disks for"
4630
                               " this instance")
4631

    
4632
    feedback_fn("creating os for instance %s on node %s" %
4633
                (instance, pnode_name))
4634

    
4635
    if iobj.disk_template != constants.DT_DISKLESS:
4636
      if self.op.mode == constants.INSTANCE_CREATE:
4637
        feedback_fn("* running the instance OS create scripts...")
4638
        result = self.rpc.call_instance_os_add(pnode_name, iobj)
4639
        msg = result.RemoteFailMsg()
4640
        if msg:
4641
          raise errors.OpExecError("Could not add os for instance %s"
4642
                                   " on node %s: %s" %
4643
                                   (instance, pnode_name, msg))
4644

    
4645
      elif self.op.mode == constants.INSTANCE_IMPORT:
4646
        feedback_fn("* running the instance OS import scripts...")
4647
        src_node = self.op.src_node
4648
        src_images = self.src_images
4649
        cluster_name = self.cfg.GetClusterName()
4650
        import_result = self.rpc.call_instance_os_import(pnode_name, iobj,
4651
                                                         src_node, src_images,
4652
                                                         cluster_name)
4653
        import_result.Raise()
4654
        for idx, result in enumerate(import_result.data):
4655
          if not result:
4656
            self.LogWarning("Could not import the image %s for instance"
4657
                            " %s, disk %d, on node %s" %
4658
                            (src_images[idx], instance, idx, pnode_name))
4659
      else:
4660
        # also checked in the prereq part
4661
        raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
4662
                                     % self.op.mode)
4663

    
4664
    if self.op.start:
4665
      iobj.admin_up = True
4666
      self.cfg.Update(iobj)
4667
      logging.info("Starting instance %s on node %s", instance, pnode_name)
4668
      feedback_fn("* starting instance...")
4669
      result = self.rpc.call_instance_start(pnode_name, iobj, None)
4670
      msg = result.RemoteFailMsg()
4671
      if msg:
4672
        raise errors.OpExecError("Could not start instance: %s" % msg)
4673

    
4674

    
4675
class LUConnectConsole(NoHooksLU):
4676
  """Connect to an instance's console.
4677

4678
  This is somewhat special in that it returns the command line that
4679
  you need to run on the master node in order to connect to the
4680
  console.
4681

4682
  """
4683
  _OP_REQP = ["instance_name"]
4684
  REQ_BGL = False
4685

    
4686
  def ExpandNames(self):
4687
    self._ExpandAndLockInstance()
4688

    
4689
  def CheckPrereq(self):
4690
    """Check prerequisites.
4691

4692
    This checks that the instance is in the cluster.
4693

4694
    """
4695
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4696
    assert self.instance is not None, \
4697
      "Cannot retrieve locked instance %s" % self.op.instance_name
4698
    _CheckNodeOnline(self, self.instance.primary_node)
4699

    
4700
  def Exec(self, feedback_fn):
4701
    """Connect to the console of an instance
4702

4703
    """
4704
    instance = self.instance
4705
    node = instance.primary_node
4706

    
4707
    node_insts = self.rpc.call_instance_list([node],
4708
                                             [instance.hypervisor])[node]
4709
    node_insts.Raise()
4710

    
4711
    if instance.name not in node_insts.data:
4712
      raise errors.OpExecError("Instance %s is not running." % instance.name)
4713

    
4714
    logging.debug("Connecting to console of %s on %s", instance.name, node)
4715

    
4716
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
4717
    cluster = self.cfg.GetClusterInfo()
4718
    # beparams and hvparams are passed separately, to avoid editing the
4719
    # instance and then saving the defaults in the instance itself.
4720
    hvparams = cluster.FillHV(instance)
4721
    beparams = cluster.FillBE(instance)
4722
    console_cmd = hyper.GetShellCommandForConsole(instance, hvparams, beparams)
4723

    
4724
    # build ssh cmdline
4725
    return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
4726

    
4727

    
4728
class LUReplaceDisks(LogicalUnit):
4729
  """Replace the disks of an instance.
4730

4731
  """
4732
  HPATH = "mirrors-replace"
4733
  HTYPE = constants.HTYPE_INSTANCE
4734
  _OP_REQP = ["instance_name", "mode", "disks"]
4735
  REQ_BGL = False
4736

    
4737
  def CheckArguments(self):
4738
    if not hasattr(self.op, "remote_node"):
4739
      self.op.remote_node = None
4740
    if not hasattr(self.op, "iallocator"):
4741
      self.op.iallocator = None
4742

    
4743
    # check for valid parameter combination
4744
    cnt = [self.op.remote_node, self.op.iallocator].count(None)
4745
    if self.op.mode == constants.REPLACE_DISK_CHG:
4746
      if cnt == 2:
4747
        raise errors.OpPrereqError("When changing the secondary either an"
4748
                                   " iallocator script must be used or the"
4749
                                   " new node given")
4750
      elif cnt == 0:
4751
        raise errors.OpPrereqError("Give either the iallocator or the new"
4752
                                   " secondary, not both")
4753
    else: # not replacing the secondary
4754
      if cnt != 2:
4755
        raise errors.OpPrereqError("The iallocator and new node options can"
4756
                                   " be used only when changing the"
4757
                                   " secondary node")
4758

    
4759
  def ExpandNames(self):
4760
    self._ExpandAndLockInstance()
4761

    
4762
    if self.op.iallocator is not None:
4763
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4764
    elif self.op.remote_node is not None:
4765
      remote_node = self.cfg.ExpandNodeName(self.op.remote_node)
4766
      if remote_node is None:
4767
        raise errors.OpPrereqError("Node '%s' not known" %
4768
                                   self.op.remote_node)
4769
      self.op.remote_node = remote_node
4770
      # Warning: do not remove the locking of the new secondary here
4771
      # unless DRBD8.AddChildren is changed to work in parallel;
4772
      # currently it doesn't since parallel invocations of
4773
      # FindUnusedMinor will conflict
4774
      self.needed_locks[locking.LEVEL_NODE] = [remote_node]
4775
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
4776
    else:
4777
      self.needed_locks[locking.LEVEL_NODE] = []
4778
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4779

    
4780
  def DeclareLocks(self, level):
4781
    # If we're not already locking all nodes in the set we have to declare the
4782
    # instance's primary/secondary nodes.
4783
    if (level == locking.LEVEL_NODE and
4784
        self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
4785
      self._LockInstancesNodes()
4786

    
4787
  def _RunAllocator(self):
4788
    """Compute a new secondary node using an IAllocator.
4789

4790
    """
4791
    ial = IAllocator(self,
4792
                     mode=constants.IALLOCATOR_MODE_RELOC,
4793
                     name=self.op.instance_name,
4794
                     relocate_from=[self.sec_node])
4795

    
4796
    ial.Run(self.op.iallocator)
4797

    
4798
    if not ial.success:
4799
      raise errors.OpPrereqError("Can't compute nodes using"
4800
                                 " iallocator '%s': %s" % (self.op.iallocator,
4801
                                                           ial.info))
4802
    if len(ial.nodes) != ial.required_nodes:
4803
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
4804
                                 " of nodes (%s), required %s" %
4805
                                 (len(ial.nodes), ial.required_nodes))
4806
    self.op.remote_node = ial.nodes[0]
4807
    self.LogInfo("Selected new secondary for the instance: %s",
4808
                 self.op.remote_node)
4809

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

4813
    This runs on the master, the primary and all the secondaries.
4814

4815
    """
4816
    env = {
4817
      "MODE": self.op.mode,
4818
      "NEW_SECONDARY": self.op.remote_node,
4819
      "OLD_SECONDARY": self.instance.secondary_nodes[0],
4820
      }
4821
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4822
    nl = [
4823
      self.cfg.GetMasterNode(),
4824
      self.instance.primary_node,
4825
      ]
4826
    if self.op.remote_node is not None:
4827
      nl.append(self.op.remote_node)
4828
    return env, nl, nl
4829

    
4830
  def CheckPrereq(self):
4831
    """Check prerequisites.
4832

4833
    This checks that the instance is in the cluster.
4834

4835
    """
4836
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4837
    assert instance is not None, \
4838
      "Cannot retrieve locked instance %s" % self.op.instance_name
4839
    self.instance = instance
4840

    
4841
    if instance.disk_template != constants.DT_DRBD8:
4842
      raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
4843
                                 " instances")
4844

    
4845
    if len(instance.secondary_nodes) != 1:
4846
      raise errors.OpPrereqError("The instance has a strange layout,"
4847
                                 " expected one secondary but found %d" %
4848
                                 len(instance.secondary_nodes))
4849

    
4850
    self.sec_node = instance.secondary_nodes[0]
4851

    
4852
    if self.op.iallocator is not None:
4853
      self._RunAllocator()
4854

    
4855
    remote_node = self.op.remote_node
4856
    if remote_node is not None:
4857
      self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
4858
      assert self.remote_node_info is not None, \
4859
        "Cannot retrieve locked node %s" % remote_node
4860
    else:
4861
      self.remote_node_info = None
4862
    if remote_node == instance.primary_node:
4863
      raise errors.OpPrereqError("The specified node is the primary node of"
4864
                                 " the instance.")
4865
    elif remote_node == self.sec_node:
4866
      raise errors.OpPrereqError("The specified node is already the"
4867
                                 " secondary node of the instance.")
4868

    
4869
    if self.op.mode == constants.REPLACE_DISK_PRI:
4870
      n1 = self.tgt_node = instance.primary_node
4871
      n2 = self.oth_node = self.sec_node
4872
    elif self.op.mode == constants.REPLACE_DISK_SEC:
4873
      n1 = self.tgt_node = self.sec_node
4874
      n2 = self.oth_node = instance.primary_node
4875
    elif self.op.mode == constants.REPLACE_DISK_CHG:
4876
      n1 = self.new_node = remote_node
4877
      n2 = self.oth_node = instance.primary_node
4878
      self.tgt_node = self.sec_node
4879
    else:
4880
      raise errors.ProgrammerError("Unhandled disk replace mode")
4881

    
4882
    _CheckNodeOnline(self, n1)
4883
    _CheckNodeOnline(self, n2)
4884

    
4885
    if not self.op.disks:
4886
      self.op.disks = range(len(instance.disks))
4887

    
4888
    for disk_idx in self.op.disks:
4889
      instance.FindDisk(disk_idx)
4890

    
4891
  def _ExecD8DiskOnly(self, feedback_fn):
4892
    """Replace a disk on the primary or secondary for dbrd8.
4893

4894
    The algorithm for replace is quite complicated:
4895

4896
      1. for each disk to be replaced:
4897

4898
        1. create new LVs on the target node with unique names
4899
        1. detach old LVs from the drbd device
4900
        1. rename old LVs to name_replaced.<time_t>
4901
        1. rename new LVs to old LVs
4902
        1. attach the new LVs (with the old names now) to the drbd device
4903

4904
      1. wait for sync across all devices
4905

4906
      1. for each modified disk:
4907

4908
        1. remove old LVs (which have the name name_replaces.<time_t>)
4909

4910
    Failures are not very well handled.
4911

4912
    """
4913
    steps_total = 6
4914
    warning, info = (self.proc.LogWarning, self.proc.LogInfo)
4915
    instance = self.instance
4916
    iv_names = {}
4917
    vgname = self.cfg.GetVGName()
4918
    # start of work
4919
    cfg = self.cfg
4920
    tgt_node = self.tgt_node
4921
    oth_node = self.oth_node
4922

    
4923
    # Step: check device activation
4924
    self.proc.LogStep(1, steps_total, "check device existence")
4925
    info("checking volume groups")
4926
    my_vg = cfg.GetVGName()
4927
    results = self.rpc.call_vg_list([oth_node, tgt_node])
4928
    if not results:
4929
      raise errors.OpExecError("Can't list volume groups on the nodes")
4930
    for node in oth_node, tgt_node:
4931
      res = results[node]
4932
      if res.failed or not res.data or my_vg not in res.data:
4933
        raise errors.OpExecError("Volume group '%s' not found on %s" %
4934
                                 (my_vg, node))
4935
    for idx, dev in enumerate(instance.disks):
4936
      if idx not in self.op.disks:
4937
        continue
4938
      for node in tgt_node, oth_node:
4939
        info("checking disk/%d on %s" % (idx, node))
4940
        cfg.SetDiskID(dev, node)
4941
        result = self.rpc.call_blockdev_find(node, dev)
4942
        msg = result.RemoteFailMsg()
4943
        if not msg and not result.payload:
4944
          msg = "disk not found"
4945
        if msg:
4946
          raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
4947
                                   (idx, node, msg))
4948

    
4949
    # Step: check other node consistency
4950
    self.proc.LogStep(2, steps_total, "check peer consistency")
4951
    for idx, dev in enumerate(instance.disks):
4952
      if idx not in self.op.disks:
4953
        continue
4954
      info("checking disk/%d consistency on %s" % (idx, oth_node))
4955
      if not _CheckDiskConsistency(self, dev, oth_node,
4956
                                   oth_node==instance.primary_node):
4957
        raise errors.OpExecError("Peer node (%s) has degraded storage, unsafe"
4958
                                 " to replace disks on this node (%s)" %
4959
                                 (oth_node, tgt_node))
4960

    
4961
    # Step: create new storage
4962
    self.proc.LogStep(3, steps_total, "allocate new storage")
4963
    for idx, dev in enumerate(instance.disks):
4964
      if idx not in self.op.disks:
4965
        continue
4966
      size = dev.size
4967
      cfg.SetDiskID(dev, tgt_node)
4968
      lv_names = [".disk%d_%s" % (idx, suf)
4969
                  for suf in ["data", "meta"]]
4970
      names = _GenerateUniqueNames(self, lv_names)
4971
      lv_data = objects.Disk(dev_type=constants.LD_LV, size=size,
4972
                             logical_id=(vgname, names[0]))
4973
      lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
4974
                             logical_id=(vgname, names[1]))
4975
      new_lvs = [lv_data, lv_meta]
4976
      old_lvs = dev.children
4977
      iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
4978
      info("creating new local storage on %s for %s" %
4979
           (tgt_node, dev.iv_name))
4980
      # we pass force_create=True to force the LVM creation
4981
      for new_lv in new_lvs:
4982
        _CreateBlockDev(self, tgt_node, instance, new_lv, True,
4983
                        _GetInstanceInfoText(instance), False)
4984

    
4985
    # Step: for each lv, detach+rename*2+attach
4986
    self.proc.LogStep(4, steps_total, "change drbd configuration")
4987
    for dev, old_lvs, new_lvs in iv_names.itervalues():
4988
      info("detaching %s drbd from local storage" % dev.iv_name)
4989
      result = self.rpc.call_blockdev_removechildren(tgt_node, dev, old_lvs)
4990
      result.Raise()
4991
      if not result.data:
4992
        raise errors.OpExecError("Can't detach drbd from local storage on node"
4993
                                 " %s for device %s" % (tgt_node, dev.iv_name))
4994
      #dev.children = []
4995
      #cfg.Update(instance)
4996

    
4997
      # ok, we created the new LVs, so now we know we have the needed
4998
      # storage; as such, we proceed on the target node to rename
4999
      # old_lv to _old, and new_lv to old_lv; note that we rename LVs
5000
      # using the assumption that logical_id == physical_id (which in
5001
      # turn is the unique_id on that node)
5002

    
5003
      # FIXME(iustin): use a better name for the replaced LVs
5004
      temp_suffix = int(time.time())
5005
      ren_fn = lambda d, suff: (d.physical_id[0],
5006
                                d.physical_id[1] + "_replaced-%s" % suff)
5007
      # build the rename list based on what LVs exist on the node
5008
      rlist = []
5009
      for to_ren in old_lvs:
5010
        result = self.rpc.call_blockdev_find(tgt_node, to_ren)
5011
        if not result.RemoteFailMsg() and result.payload:
5012
          # device exists
5013
          rlist.append((to_ren, ren_fn(to_ren, temp_suffix)))
5014

    
5015
      info("renaming the old LVs on the target node")
5016
      result = self.rpc.call_blockdev_rename(tgt_node, rlist)
5017
      result.Raise()
5018
      if not result.data:
5019
        raise errors.OpExecError("Can't rename old LVs on node %s" % tgt_node)
5020
      # now we rename the new LVs to the old LVs
5021
      info("renaming the new LVs on the target node")
5022
      rlist = [(new, old.physical_id) for old, new in zip(old_lvs, new_lvs)]
5023
      result = self.rpc.call_blockdev_rename(tgt_node, rlist)
5024
      result.Raise()
5025
      if not result.data:
5026
        raise errors.OpExecError("Can't rename new LVs on node %s" % tgt_node)
5027

    
5028
      for old, new in zip(old_lvs, new_lvs):
5029
        new.logical_id = old.logical_id
5030
        cfg.SetDiskID(new, tgt_node)
5031

    
5032
      for disk in old_lvs:
5033
        disk.logical_id = ren_fn(disk, temp_suffix)
5034
        cfg.SetDiskID(disk, tgt_node)
5035

    
5036
      # now that the new lvs have the old name, we can add them to the device
5037
      info("adding new mirror component on %s" % tgt_node)
5038
      result = self.rpc.call_blockdev_addchildren(tgt_node, dev, new_lvs)
5039
      if result.failed or not result.data:
5040
        for new_lv in new_lvs:
5041
          result = self.rpc.call_blockdev_remove(tgt_node, new_lv)
5042
          if result.failed or not result.data:
5043
            warning("Can't rollback device %s", hint="manually cleanup unused"
5044
                    " logical volumes")
5045
        raise errors.OpExecError("Can't add local storage to drbd")
5046

    
5047
      dev.children = new_lvs
5048
      cfg.Update(instance)
5049

    
5050
    # Step: wait for sync
5051

    
5052
    # this can fail as the old devices are degraded and _WaitForSync
5053
    # does a combined result over all disks, so we don't check its
5054
    # return value
5055
    self.proc.LogStep(5, steps_total, "sync devices")
5056
    _WaitForSync(self, instance, unlock=True)
5057

    
5058
    # so check manually all the devices
5059
    for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
5060
      cfg.SetDiskID(dev, instance.primary_node)
5061
      result = self.rpc.call_blockdev_find(instance.primary_node, dev)
5062
      msg = result.RemoteFailMsg()
5063
      if not msg and not result.payload:
5064
        msg = "disk not found"
5065
      if msg:
5066
        raise errors.OpExecError("Can't find DRBD device %s: %s" %
5067
                                 (name, msg))
5068
      if result.payload[5]:
5069
        raise errors.OpExecError("DRBD device %s is degraded!" % name)
5070

    
5071
    # Step: remove old storage
5072
    self.proc.LogStep(6, steps_total, "removing old storage")
5073
    for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
5074
      info("remove logical volumes for %s" % name)
5075
      for lv in old_lvs:
5076
        cfg.SetDiskID(lv, tgt_node)
5077
        result = self.rpc.call_blockdev_remove(tgt_node, lv)
5078
        if result.failed or not result.data:
5079
          warning("Can't remove old LV", hint="manually remove unused LVs")
5080
          continue
5081

    
5082
  def _ExecD8Secondary(self, feedback_fn):
5083
    """Replace the secondary node for drbd8.
5084

5085
    The algorithm for replace is quite complicated:
5086
      - for all disks of the instance:
5087
        - create new LVs on the new node with same names
5088
        - shutdown the drbd device on the old secondary
5089
        - disconnect the drbd network on the primary
5090
        - create the drbd device on the new secondary
5091
        - network attach the drbd on the primary, using an artifice:
5092
          the drbd code for Attach() will connect to the network if it
5093
          finds a device which is connected to the good local disks but
5094
          not network enabled
5095
      - wait for sync across all devices
5096
      - remove all disks from the old secondary
5097

5098
    Failures are not very well handled.
5099

5100
    """
5101
    steps_total = 6
5102
    warning, info = (self.proc.LogWarning, self.proc.LogInfo)
5103
    instance = self.instance
5104
    iv_names = {}
5105
    # start of work
5106
    cfg = self.cfg
5107
    old_node = self.tgt_node
5108
    new_node = self.new_node
5109
    pri_node = instance.primary_node
5110
    nodes_ip = {
5111
      old_node: self.cfg.GetNodeInfo(old_node).secondary_ip,
5112
      new_node: self.cfg.GetNodeInfo(new_node).secondary_ip,
5113
      pri_node: self.cfg.GetNodeInfo(pri_node).secondary_ip,
5114
      }
5115

    
5116
    # Step: check device activation
5117
    self.proc.LogStep(1, steps_total, "check device existence")
5118
    info("checking volume groups")
5119
    my_vg = cfg.GetVGName()
5120
    results = self.rpc.call_vg_list([pri_node, new_node])
5121
    for node in pri_node, new_node:
5122
      res = results[node]
5123
      if res.failed or not res.data or my_vg not in res.data:
5124
        raise errors.OpExecError("Volume group '%s' not found on %s" %
5125
                                 (my_vg, node))
5126
    for idx, dev in enumerate(instance.disks):
5127
      if idx not in self.op.disks:
5128
        continue
5129
      info("checking disk/%d on %s" % (idx, pri_node))
5130
      cfg.SetDiskID(dev, pri_node)
5131
      result = self.rpc.call_blockdev_find(pri_node, dev)
5132
      msg = result.RemoteFailMsg()
5133
      if not msg and not result.payload:
5134
        msg = "disk not found"
5135
      if msg:
5136
        raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
5137
                                 (idx, pri_node, msg))
5138

    
5139
    # Step: check other node consistency
5140
    self.proc.LogStep(2, steps_total, "check peer consistency")
5141
    for idx, dev in enumerate(instance.disks):
5142
      if idx not in self.op.disks:
5143
        continue
5144
      info("checking disk/%d consistency on %s" % (idx, pri_node))
5145
      if not _CheckDiskConsistency(self, dev, pri_node, True, ldisk=True):
5146
        raise errors.OpExecError("Primary node (%s) has degraded storage,"
5147
                                 " unsafe to replace the secondary" %
5148
                                 pri_node)
5149

    
5150
    # Step: create new storage
5151
    self.proc.LogStep(3, steps_total, "allocate new storage")
5152
    for idx, dev in enumerate(instance.disks):
5153
      info("adding new local storage on %s for disk/%d" %
5154
           (new_node, idx))
5155
      # we pass force_create=True to force LVM creation
5156
      for new_lv in dev.children:
5157
        _CreateBlockDev(self, new_node, instance, new_lv, True,
5158
                        _GetInstanceInfoText(instance), False)
5159

    
5160
    # Step 4: dbrd minors and drbd setups changes
5161
    # after this, we must manually remove the drbd minors on both the
5162
    # error and the success paths
5163
    minors = cfg.AllocateDRBDMinor([new_node for dev in instance.disks],
5164
                                   instance.name)
5165
    logging.debug("Allocated minors %s" % (minors,))
5166
    self.proc.LogStep(4, steps_total, "changing drbd configuration")
5167
    for idx, (dev, new_minor) in enumerate(zip(instance.disks, minors)):
5168
      size = dev.size
5169
      info("activating a new drbd on %s for disk/%d" % (new_node, idx))
5170
      # create new devices on new_node; note that we create two IDs:
5171
      # one without port, so the drbd will be activated without
5172
      # networking information on the new node at this stage, and one
5173
      # with network, for the latter activation in step 4
5174
      (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
5175
      if pri_node == o_node1:
5176
        p_minor = o_minor1
5177
      else:
5178
        p_minor = o_minor2
5179

    
5180
      new_alone_id = (pri_node, new_node, None, p_minor, new_minor, o_secret)
5181
      new_net_id = (pri_node, new_node, o_port, p_minor, new_minor, o_secret)
5182

    
5183
      iv_names[idx] = (dev, dev.children, new_net_id)
5184
      logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
5185
                    new_net_id)
5186
      new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
5187
                              logical_id=new_alone_id,
5188
                              children=dev.children)
5189
      try:
5190
        _CreateSingleBlockDev(self, new_node, instance, new_drbd,
5191
                              _GetInstanceInfoText(instance), False)
5192
      except errors.BlockDeviceError:
5193
        self.cfg.ReleaseDRBDMinors(instance.name)
5194
        raise
5195

    
5196
    for idx, dev in enumerate(instance.disks):
5197
      # we have new devices, shutdown the drbd on the old secondary
5198
      info("shutting down drbd for disk/%d on old node" % idx)
5199
      cfg.SetDiskID(dev, old_node)
5200
      result = self.rpc.call_blockdev_shutdown(old_node, dev)
5201
      if result.failed or not result.data:
5202
        warning("Failed to shutdown drbd for disk/%d on old node" % idx,
5203
                hint="Please cleanup this device manually as soon as possible")
5204

    
5205
    info("detaching primary drbds from the network (=> standalone)")
5206
    result = self.rpc.call_drbd_disconnect_net([pri_node], nodes_ip,
5207
                                               instance.disks)[pri_node]
5208

    
5209
    msg = result.RemoteFailMsg()
5210
    if msg:
5211
      # detaches didn't succeed (unlikely)
5212
      self.cfg.ReleaseDRBDMinors(instance.name)
5213
      raise errors.OpExecError("Can't detach the disks from the network on"
5214
                               " old node: %s" % (msg,))
5215

    
5216
    # if we managed to detach at least one, we update all the disks of
5217
    # the instance to point to the new secondary
5218
    info("updating instance configuration")
5219
    for dev, _, new_logical_id in iv_names.itervalues():
5220
      dev.logical_id = new_logical_id
5221
      cfg.SetDiskID(dev, pri_node)
5222
    cfg.Update(instance)
5223

    
5224
    # and now perform the drbd attach
5225
    info("attaching primary drbds to new secondary (standalone => connected)")
5226
    result = self.rpc.call_drbd_attach_net([pri_node, new_node], nodes_ip,
5227
                                           instance.disks, instance.name,
5228
                                           False)
5229
    for to_node, to_result in result.items():
5230
      msg = to_result.RemoteFailMsg()
5231
      if msg:
5232
        warning("can't attach drbd disks on node %s: %s", to_node, msg,
5233
                hint="please do a gnt-instance info to see the"
5234
                " status of disks")
5235

    
5236
    # this can fail as the old devices are degraded and _WaitForSync
5237
    # does a combined result over all disks, so we don't check its
5238
    # return value
5239
    self.proc.LogStep(5, steps_total, "sync devices")
5240
    _WaitForSync(self, instance, unlock=True)
5241

    
5242
    # so check manually all the devices
5243
    for idx, (dev, old_lvs, _) in iv_names.iteritems():
5244
      cfg.SetDiskID(dev, pri_node)
5245
      result = self.rpc.call_blockdev_find(pri_node, dev)
5246
      msg = result.RemoteFailMsg()
5247
      if not msg and not result.payload:
5248
        msg = "disk not found"
5249
      if msg:
5250
        raise errors.OpExecError("Can't find DRBD device disk/%d: %s" %
5251
                                 (idx, msg))
5252
      if result.payload[5]:
5253
        raise errors.OpExecError("DRBD device disk/%d is degraded!" % idx)
5254

    
5255
    self.proc.LogStep(6, steps_total, "removing old storage")
5256
    for idx, (dev, old_lvs, _) in iv_names.iteritems():
5257
      info("remove logical volumes for disk/%d" % idx)
5258
      for lv in old_lvs:
5259
        cfg.SetDiskID(lv, old_node)
5260
        result = self.rpc.call_blockdev_remove(old_node, lv)
5261
        if result.failed or not result.data:
5262
          warning("Can't remove LV on old secondary",
5263
                  hint="Cleanup stale volumes by hand")
5264

    
5265
  def Exec(self, feedback_fn):
5266
    """Execute disk replacement.
5267

5268
    This dispatches the disk replacement to the appropriate handler.
5269

5270
    """
5271
    instance = self.instance
5272

    
5273
    # Activate the instance disks if we're replacing them on a down instance
5274
    if not instance.admin_up:
5275
      _StartInstanceDisks(self, instance, True)
5276

    
5277
    if self.op.mode == constants.REPLACE_DISK_CHG:
5278
      fn = self._ExecD8Secondary
5279
    else:
5280
      fn = self._ExecD8DiskOnly
5281

    
5282
    ret = fn(feedback_fn)
5283

    
5284
    # Deactivate the instance disks if we're replacing them on a down instance
5285
    if not instance.admin_up:
5286
      _SafeShutdownInstanceDisks(self, instance)
5287

    
5288
    return ret
5289

    
5290

    
5291
class LUGrowDisk(LogicalUnit):
5292
  """Grow a disk of an instance.
5293

5294
  """
5295
  HPATH = "disk-grow"
5296
  HTYPE = constants.HTYPE_INSTANCE
5297
  _OP_REQP = ["instance_name", "disk", "amount", "wait_for_sync"]
5298
  REQ_BGL = False
5299

    
5300
  def ExpandNames(self):
5301
    self._ExpandAndLockInstance()
5302
    self.needed_locks[locking.LEVEL_NODE] = []
5303
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5304

    
5305
  def DeclareLocks(self, level):
5306
    if level == locking.LEVEL_NODE:
5307
      self._LockInstancesNodes()
5308

    
5309
  def BuildHooksEnv(self):
5310
    """Build hooks env.
5311

5312
    This runs on the master, the primary and all the secondaries.
5313

5314
    """
5315
    env = {
5316
      "DISK": self.op.disk,
5317
      "AMOUNT": self.op.amount,
5318
      }
5319
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5320
    nl = [
5321
      self.cfg.GetMasterNode(),
5322
      self.instance.primary_node,
5323
      ]
5324
    return env, nl, nl
5325

    
5326
  def CheckPrereq(self):
5327
    """Check prerequisites.
5328

5329
    This checks that the instance is in the cluster.
5330

5331
    """
5332
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5333
    assert instance is not None, \
5334
      "Cannot retrieve locked instance %s" % self.op.instance_name
5335
    nodenames = list(instance.all_nodes)
5336
    for node in nodenames:
5337
      _CheckNodeOnline(self, node)
5338

    
5339

    
5340
    self.instance = instance
5341

    
5342
    if instance.disk_template not in (constants.DT_PLAIN, constants.DT_DRBD8):
5343
      raise errors.OpPrereqError("Instance's disk layout does not support"
5344
                                 " growing.")
5345

    
5346
    self.disk = instance.FindDisk(self.op.disk)
5347

    
5348
    nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
5349
                                       instance.hypervisor)
5350
    for node in nodenames:
5351
      info = nodeinfo[node]
5352
      if info.failed or not info.data:
5353
        raise errors.OpPrereqError("Cannot get current information"
5354
                                   " from node '%s'" % node)
5355
      vg_free = info.data.get('vg_free', None)
5356
      if not isinstance(vg_free, int):
5357
        raise errors.OpPrereqError("Can't compute free disk space on"
5358
                                   " node %s" % node)
5359
      if self.op.amount > vg_free:
5360
        raise errors.OpPrereqError("Not enough disk space on target node %s:"
5361
                                   " %d MiB available, %d MiB required" %
5362
                                   (node, vg_free, self.op.amount))
5363

    
5364
  def Exec(self, feedback_fn):
5365
    """Execute disk grow.
5366

5367
    """
5368
    instance = self.instance
5369
    disk = self.disk
5370
    for node in instance.all_nodes:
5371
      self.cfg.SetDiskID(disk, node)
5372
      result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
5373
      msg = result.RemoteFailMsg()
5374
      if msg:
5375
        raise errors.OpExecError("Grow request failed to node %s: %s" %
5376
                                 (node, msg))
5377
    disk.RecordGrow(self.op.amount)
5378
    self.cfg.Update(instance)
5379
    if self.op.wait_for_sync:
5380
      disk_abort = not _WaitForSync(self, instance)
5381
      if disk_abort:
5382
        self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
5383
                             " status.\nPlease check the instance.")
5384

    
5385

    
5386
class LUQueryInstanceData(NoHooksLU):
5387
  """Query runtime instance data.
5388

5389
  """
5390
  _OP_REQP = ["instances", "static"]
5391
  REQ_BGL = False
5392

    
5393
  def ExpandNames(self):
5394
    self.needed_locks = {}
5395
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
5396

    
5397
    if not isinstance(self.op.instances, list):
5398
      raise errors.OpPrereqError("Invalid argument type 'instances'")
5399

    
5400
    if self.op.instances:
5401
      self.wanted_names = []
5402
      for name in self.op.instances:
5403
        full_name = self.cfg.ExpandInstanceName(name)
5404
        if full_name is None:
5405
          raise errors.OpPrereqError("Instance '%s' not known" % name)
5406
        self.wanted_names.append(full_name)
5407
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
5408
    else:
5409
      self.wanted_names = None
5410
      self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
5411

    
5412
    self.needed_locks[locking.LEVEL_NODE] = []
5413
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5414

    
5415
  def DeclareLocks(self, level):
5416
    if level == locking.LEVEL_NODE:
5417
      self._LockInstancesNodes()
5418

    
5419
  def CheckPrereq(self):
5420
    """Check prerequisites.
5421

5422
    This only checks the optional instance list against the existing names.
5423

5424
    """
5425
    if self.wanted_names is None:
5426
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
5427

    
5428
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
5429
                             in self.wanted_names]
5430
    return
5431

    
5432
  def _ComputeDiskStatus(self, instance, snode, dev):
5433
    """Compute block device status.
5434

5435
    """
5436
    static = self.op.static
5437
    if not static:
5438
      self.cfg.SetDiskID(dev, instance.primary_node)
5439
      dev_pstatus = self.rpc.call_blockdev_find(instance.primary_node, dev)
5440
      msg = dev_pstatus.RemoteFailMsg()
5441
      if msg:
5442
        raise errors.OpExecError("Can't compute disk status for %s: %s" %
5443
                                 (instance.name, msg))
5444
      dev_pstatus = dev_pstatus.payload
5445
    else:
5446
      dev_pstatus = None
5447

    
5448
    if dev.dev_type in constants.LDS_DRBD:
5449
      # we change the snode then (otherwise we use the one passed in)
5450
      if dev.logical_id[0] == instance.primary_node:
5451
        snode = dev.logical_id[1]
5452
      else:
5453
        snode = dev.logical_id[0]
5454

    
5455
    if snode and not static:
5456
      self.cfg.SetDiskID(dev, snode)
5457
      dev_sstatus = self.rpc.call_blockdev_find(snode, dev)
5458
      msg = dev_sstatus.RemoteFailMsg()
5459
      if msg:
5460
        raise errors.OpExecError("Can't compute disk status for %s: %s" %
5461
                                 (instance.name, msg))
5462
      dev_sstatus = dev_sstatus.payload
5463
    else:
5464
      dev_sstatus = None
5465

    
5466
    if dev.children:
5467
      dev_children = [self._ComputeDiskStatus(instance, snode, child)
5468
                      for child in dev.children]
5469
    else:
5470
      dev_children = []
5471

    
5472
    data = {
5473
      "iv_name": dev.iv_name,
5474
      "dev_type": dev.dev_type,
5475
      "logical_id": dev.logical_id,
5476
      "physical_id": dev.physical_id,
5477
      "pstatus": dev_pstatus,
5478
      "sstatus": dev_sstatus,
5479
      "children": dev_children,
5480
      "mode": dev.mode,
5481
      }
5482

    
5483
    return data
5484

    
5485
  def Exec(self, feedback_fn):
5486
    """Gather and return data"""
5487
    result = {}
5488

    
5489
    cluster = self.cfg.GetClusterInfo()
5490

    
5491
    for instance in self.wanted_instances:
5492
      if not self.op.static:
5493
        remote_info = self.rpc.call_instance_info(instance.primary_node,
5494
                                                  instance.name,
5495
                                                  instance.hypervisor)
5496
        remote_info.Raise()
5497
        remote_info = remote_info.data
5498
        if remote_info and "state" in remote_info:
5499
          remote_state = "up"
5500
        else:
5501
          remote_state = "down"
5502
      else:
5503
        remote_state = None
5504
      if instance.admin_up:
5505
        config_state = "up"
5506
      else:
5507
        config_state = "down"
5508

    
5509
      disks = [self._ComputeDiskStatus(instance, None, device)
5510
               for device in instance.disks]
5511

    
5512
      idict = {
5513
        "name": instance.name,
5514
        "config_state": config_state,
5515
        "run_state": remote_state,
5516
        "pnode": instance.primary_node,
5517
        "snodes": instance.secondary_nodes,
5518
        "os": instance.os,
5519
        "nics": [(nic.mac, nic.ip, nic.bridge) for nic in instance.nics],
5520
        "disks": disks,
5521
        "hypervisor": instance.hypervisor,
5522
        "network_port": instance.network_port,
5523
        "hv_instance": instance.hvparams,
5524
        "hv_actual": cluster.FillHV(instance),
5525
        "be_instance": instance.beparams,
5526
        "be_actual": cluster.FillBE(instance),
5527
        }
5528

    
5529
      result[instance.name] = idict
5530

    
5531
    return result
5532

    
5533

    
5534
class LUSetInstanceParams(LogicalUnit):
5535
  """Modifies an instances's parameters.
5536

5537
  """
5538
  HPATH = "instance-modify"
5539
  HTYPE = constants.HTYPE_INSTANCE
5540
  _OP_REQP = ["instance_name"]
5541
  REQ_BGL = False
5542

    
5543
  def CheckArguments(self):
5544
    if not hasattr(self.op, 'nics'):
5545
      self.op.nics = []
5546
    if not hasattr(self.op, 'disks'):
5547
      self.op.disks = []
5548
    if not hasattr(self.op, 'beparams'):
5549
      self.op.beparams = {}
5550
    if not hasattr(self.op, 'hvparams'):
5551
      self.op.hvparams = {}
5552
    self.op.force = getattr(self.op, "force", False)
5553
    if not (self.op.nics or self.op.disks or
5554
            self.op.hvparams or self.op.beparams):
5555
      raise errors.OpPrereqError("No changes submitted")
5556

    
5557
    utils.CheckBEParams(self.op.beparams)
5558

    
5559
    # Disk validation
5560
    disk_addremove = 0
5561
    for disk_op, disk_dict in self.op.disks:
5562
      if disk_op == constants.DDM_REMOVE:
5563
        disk_addremove += 1
5564
        continue
5565
      elif disk_op == constants.DDM_ADD:
5566
        disk_addremove += 1
5567
      else:
5568
        if not isinstance(disk_op, int):
5569
          raise errors.OpPrereqError("Invalid disk index")
5570
      if disk_op == constants.DDM_ADD:
5571
        mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
5572
        if mode not in constants.DISK_ACCESS_SET:
5573
          raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode)
5574
        size = disk_dict.get('size', None)
5575
        if size is None:
5576
          raise errors.OpPrereqError("Required disk parameter size missing")
5577
        try:
5578
          size = int(size)
5579
        except ValueError, err:
5580
          raise errors.OpPrereqError("Invalid disk size parameter: %s" %
5581
                                     str(err))
5582
        disk_dict['size'] = size
5583
      else:
5584
        # modification of disk
5585
        if 'size' in disk_dict:
5586
          raise errors.OpPrereqError("Disk size change not possible, use"
5587
                                     " grow-disk")
5588

    
5589
    if disk_addremove > 1:
5590
      raise errors.OpPrereqError("Only one disk add or remove operation"
5591
                                 " supported at a time")
5592

    
5593
    # NIC validation
5594
    nic_addremove = 0
5595
    for nic_op, nic_dict in self.op.nics:
5596
      if nic_op == constants.DDM_REMOVE:
5597
        nic_addremove += 1
5598
        continue
5599
      elif nic_op == constants.DDM_ADD:
5600
        nic_addremove += 1
5601
      else:
5602
        if not isinstance(nic_op, int):
5603
          raise errors.OpPrereqError("Invalid nic index")
5604

    
5605
      # nic_dict should be a dict
5606
      nic_ip = nic_dict.get('ip', None)
5607
      if nic_ip is not None:
5608
        if nic_ip.lower() == "none":
5609
          nic_dict['ip'] = None
5610
        else:
5611
          if not utils.IsValidIP(nic_ip):
5612
            raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip)
5613
      # we can only check None bridges and assign the default one
5614
      nic_bridge = nic_dict.get('bridge', None)
5615
      if nic_bridge is None:
5616
        nic_dict['bridge'] = self.cfg.GetDefBridge()
5617
      # but we can validate MACs
5618
      nic_mac = nic_dict.get('mac', None)
5619
      if nic_mac is not None:
5620
        if self.cfg.IsMacInUse(nic_mac):
5621
          raise errors.OpPrereqError("MAC address %s already in use"
5622
                                     " in cluster" % nic_mac)
5623
        if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
5624
          if not utils.IsValidMac(nic_mac):
5625
            raise errors.OpPrereqError("Invalid MAC address %s" % nic_mac)
5626
    if nic_addremove > 1:
5627
      raise errors.OpPrereqError("Only one NIC add or remove operation"
5628
                                 " supported at a time")
5629

    
5630
  def ExpandNames(self):
5631
    self._ExpandAndLockInstance()
5632
    self.needed_locks[locking.LEVEL_NODE] = []
5633
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5634

    
5635
  def DeclareLocks(self, level):
5636
    if level == locking.LEVEL_NODE:
5637
      self._LockInstancesNodes()
5638

    
5639
  def BuildHooksEnv(self):
5640
    """Build hooks env.
5641

5642
    This runs on the master, primary and secondaries.
5643

5644
    """
5645
    args = dict()
5646
    if constants.BE_MEMORY in self.be_new:
5647
      args['memory'] = self.be_new[constants.BE_MEMORY]
5648
    if constants.BE_VCPUS in self.be_new:
5649
      args['vcpus'] = self.be_new[constants.BE_VCPUS]
5650
    # FIXME: readd disk/nic changes
5651
    env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
5652
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5653
    return env, nl, nl
5654

    
5655
  def CheckPrereq(self):
5656
    """Check prerequisites.
5657

5658
    This only checks the instance list against the existing names.
5659

5660
    """
5661
    force = self.force = self.op.force
5662

    
5663
    # checking the new params on the primary/secondary nodes
5664

    
5665
    instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5666
    assert self.instance is not None, \
5667
      "Cannot retrieve locked instance %s" % self.op.instance_name
5668
    pnode = instance.primary_node
5669
    nodelist = list(instance.all_nodes)
5670

    
5671
    # hvparams processing
5672
    if self.op.hvparams:
5673
      i_hvdict = copy.deepcopy(instance.hvparams)
5674
      for key, val in self.op.hvparams.iteritems():
5675
        if val == constants.VALUE_DEFAULT:
5676
          try:
5677
            del i_hvdict[key]
5678
          except KeyError:
5679
            pass
5680
        elif val == constants.VALUE_NONE:
5681
          i_hvdict[key] = None
5682
        else:
5683
          i_hvdict[key] = val
5684
      cluster = self.cfg.GetClusterInfo()
5685
      hv_new = cluster.FillDict(cluster.hvparams[instance.hypervisor],
5686
                                i_hvdict)
5687
      # local check
5688
      hypervisor.GetHypervisor(
5689
        instance.hypervisor).CheckParameterSyntax(hv_new)
5690
      _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
5691
      self.hv_new = hv_new # the new actual values
5692
      self.hv_inst = i_hvdict # the new dict (without defaults)
5693
    else:
5694
      self.hv_new = self.hv_inst = {}
5695

    
5696
    # beparams processing
5697
    if self.op.beparams:
5698
      i_bedict = copy.deepcopy(instance.beparams)
5699
      for key, val in self.op.beparams.iteritems():
5700
        if val == constants.VALUE_DEFAULT:
5701
          try:
5702
            del i_bedict[key]
5703
          except KeyError:
5704
            pass
5705
        else:
5706
          i_bedict[key] = val
5707
      cluster = self.cfg.GetClusterInfo()
5708
      be_new = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
5709
                                i_bedict)
5710
      self.be_new = be_new # the new actual values
5711
      self.be_inst = i_bedict # the new dict (without defaults)
5712
    else:
5713
      self.be_new = self.be_inst = {}
5714

    
5715
    self.warn = []
5716

    
5717
    if constants.BE_MEMORY in self.op.beparams and not self.force:
5718
      mem_check_list = [pnode]
5719
      if be_new[constants.BE_AUTO_BALANCE]:
5720
        # either we changed auto_balance to yes or it was from before
5721
        mem_check_list.extend(instance.secondary_nodes)
5722
      instance_info = self.rpc.call_instance_info(pnode, instance.name,
5723
                                                  instance.hypervisor)
5724
      nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
5725
                                         instance.hypervisor)
5726
      if nodeinfo[pnode].failed or not isinstance(nodeinfo[pnode].data, dict):
5727
        # Assume the primary node is unreachable and go ahead
5728
        self.warn.append("Can't get info from primary node %s" % pnode)
5729
      else:
5730
        if not instance_info.failed and instance_info.data:
5731
          current_mem = instance_info.data['memory']
5732
        else:
5733
          # Assume instance not running
5734
          # (there is a slight race condition here, but it's not very probable,
5735
          # and we have no other way to check)
5736
          current_mem = 0
5737
        miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
5738
                    nodeinfo[pnode].data['memory_free'])
5739
        if miss_mem > 0:
5740
          raise errors.OpPrereqError("This change will prevent the instance"
5741
                                     " from starting, due to %d MB of memory"
5742
                                     " missing on its primary node" % miss_mem)
5743

    
5744
      if be_new[constants.BE_AUTO_BALANCE]:
5745
        for node, nres in nodeinfo.iteritems():
5746
          if node not in instance.secondary_nodes:
5747
            continue
5748
          if nres.failed or not isinstance(nres.data, dict):
5749
            self.warn.append("Can't get info from secondary node %s" % node)
5750
          elif be_new[constants.BE_MEMORY] > nres.data['memory_free']:
5751
            self.warn.append("Not enough memory to failover instance to"
5752
                             " secondary node %s" % node)
5753

    
5754
    # NIC processing
5755
    for nic_op, nic_dict in self.op.nics:
5756
      if nic_op == constants.DDM_REMOVE:
5757
        if not instance.nics:
5758
          raise errors.OpPrereqError("Instance has no NICs, cannot remove")
5759
        continue
5760
      if nic_op != constants.DDM_ADD:
5761
        # an existing nic
5762
        if nic_op < 0 or nic_op >= len(instance.nics):
5763
          raise errors.OpPrereqError("Invalid NIC index %s, valid values"
5764
                                     " are 0 to %d" %
5765
                                     (nic_op, len(instance.nics)))
5766
      nic_bridge = nic_dict.get('bridge', None)
5767
      if nic_bridge is not None:
5768
        if not self.rpc.call_bridges_exist(pnode, [nic_bridge]):
5769
          msg = ("Bridge '%s' doesn't exist on one of"
5770
                 " the instance nodes" % nic_bridge)
5771
          if self.force:
5772
            self.warn.append(msg)
5773
          else:
5774
            raise errors.OpPrereqError(msg)
5775

    
5776
    # DISK processing
5777
    if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
5778
      raise errors.OpPrereqError("Disk operations not supported for"
5779
                                 " diskless instances")
5780
    for disk_op, disk_dict in self.op.disks:
5781
      if disk_op == constants.DDM_REMOVE:
5782
        if len(instance.disks) == 1:
5783
          raise errors.OpPrereqError("Cannot remove the last disk of"
5784
                                     " an instance")
5785
        ins_l = self.rpc.call_instance_list([pnode], [instance.hypervisor])
5786
        ins_l = ins_l[pnode]
5787
        if ins_l.failed or not isinstance(ins_l.data, list):
5788
          raise errors.OpPrereqError("Can't contact node '%s'" % pnode)
5789
        if instance.name in ins_l.data:
5790
          raise errors.OpPrereqError("Instance is running, can't remove"
5791
                                     " disks.")
5792

    
5793
      if (disk_op == constants.DDM_ADD and
5794
          len(instance.nics) >= constants.MAX_DISKS):
5795
        raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
5796
                                   " add more" % constants.MAX_DISKS)
5797
      if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
5798
        # an existing disk
5799
        if disk_op < 0 or disk_op >= len(instance.disks):
5800
          raise errors.OpPrereqError("Invalid disk index %s, valid values"
5801
                                     " are 0 to %d" %
5802
                                     (disk_op, len(instance.disks)))
5803

    
5804
    return
5805

    
5806
  def Exec(self, feedback_fn):
5807
    """Modifies an instance.
5808

5809
    All parameters take effect only at the next restart of the instance.
5810

5811
    """
5812
    # Process here the warnings from CheckPrereq, as we don't have a
5813
    # feedback_fn there.
5814
    for warn in self.warn:
5815
      feedback_fn("WARNING: %s" % warn)
5816

    
5817
    result = []
5818
    instance = self.instance
5819
    # disk changes
5820
    for disk_op, disk_dict in self.op.disks:
5821
      if disk_op == constants.DDM_REMOVE:
5822
        # remove the last disk
5823
        device = instance.disks.pop()
5824
        device_idx = len(instance.disks)
5825
        for node, disk in device.ComputeNodeTree(instance.primary_node):
5826
          self.cfg.SetDiskID(disk, node)
5827
          rpc_result = self.rpc.call_blockdev_remove(node, disk)
5828
          if rpc_result.failed or not rpc_result.data:
5829
            self.proc.LogWarning("Could not remove disk/%d on node %s,"
5830
                                 " continuing anyway", device_idx, node)
5831
        result.append(("disk/%d" % device_idx, "remove"))
5832
      elif disk_op == constants.DDM_ADD:
5833
        # add a new disk
5834
        if instance.disk_template == constants.DT_FILE:
5835
          file_driver, file_path = instance.disks[0].logical_id
5836
          file_path = os.path.dirname(file_path)
5837
        else:
5838
          file_driver = file_path = None
5839
        disk_idx_base = len(instance.disks)
5840
        new_disk = _GenerateDiskTemplate(self,
5841
                                         instance.disk_template,
5842
                                         instance.name, instance.primary_node,
5843
                                         instance.secondary_nodes,
5844
                                         [disk_dict],
5845
                                         file_path,
5846
                                         file_driver,
5847
                                         disk_idx_base)[0]
5848
        instance.disks.append(new_disk)
5849
        info = _GetInstanceInfoText(instance)
5850

    
5851
        logging.info("Creating volume %s for instance %s",
5852
                     new_disk.iv_name, instance.name)
5853
        # Note: this needs to be kept in sync with _CreateDisks
5854
        #HARDCODE
5855
        for node in instance.all_nodes:
5856
          f_create = node == instance.primary_node
5857
          try:
5858
            _CreateBlockDev(self, node, instance, new_disk,
5859
                            f_create, info, f_create)
5860
          except errors.OpExecError, err:
5861
            self.LogWarning("Failed to create volume %s (%s) on"
5862
                            " node %s: %s",
5863
                            new_disk.iv_name, new_disk, node, err)
5864
        result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
5865
                       (new_disk.size, new_disk.mode)))
5866
      else:
5867
        # change a given disk
5868
        instance.disks[disk_op].mode = disk_dict['mode']
5869
        result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
5870
    # NIC changes
5871
    for nic_op, nic_dict in self.op.nics:
5872
      if nic_op == constants.DDM_REMOVE:
5873
        # remove the last nic
5874
        del instance.nics[-1]
5875
        result.append(("nic.%d" % len(instance.nics), "remove"))
5876
      elif nic_op == constants.DDM_ADD:
5877
        # add a new nic
5878
        if 'mac' not in nic_dict:
5879
          mac = constants.VALUE_GENERATE
5880
        else:
5881
          mac = nic_dict['mac']
5882
        if mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
5883
          mac = self.cfg.GenerateMAC()
5884
        new_nic = objects.NIC(mac=mac, ip=nic_dict.get('ip', None),
5885
                              bridge=nic_dict.get('bridge', None))
5886
        instance.nics.append(new_nic)
5887
        result.append(("nic.%d" % (len(instance.nics) - 1),
5888
                       "add:mac=%s,ip=%s,bridge=%s" %
5889
                       (new_nic.mac, new_nic.ip, new_nic.bridge)))
5890
      else:
5891
        # change a given nic
5892
        for key in 'mac', 'ip', 'bridge':
5893
          if key in nic_dict:
5894
            setattr(instance.nics[nic_op], key, nic_dict[key])
5895
            result.append(("nic.%s/%d" % (key, nic_op), nic_dict[key]))
5896

    
5897
    # hvparams changes
5898
    if self.op.hvparams:
5899
      instance.hvparams = self.hv_inst
5900
      for key, val in self.op.hvparams.iteritems():
5901
        result.append(("hv/%s" % key, val))
5902

    
5903
    # beparams changes
5904
    if self.op.beparams:
5905
      instance.beparams = self.be_inst
5906
      for key, val in self.op.beparams.iteritems():
5907
        result.append(("be/%s" % key, val))
5908

    
5909
    self.cfg.Update(instance)
5910

    
5911
    return result
5912

    
5913

    
5914
class LUQueryExports(NoHooksLU):
5915
  """Query the exports list
5916

5917
  """
5918
  _OP_REQP = ['nodes']
5919
  REQ_BGL = False
5920

    
5921
  def ExpandNames(self):
5922
    self.needed_locks = {}
5923
    self.share_locks[locking.LEVEL_NODE] = 1
5924
    if not self.op.nodes:
5925
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5926
    else:
5927
      self.needed_locks[locking.LEVEL_NODE] = \
5928
        _GetWantedNodes(self, self.op.nodes)
5929

    
5930
  def CheckPrereq(self):
5931
    """Check prerequisites.
5932

5933
    """
5934
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
5935

    
5936
  def Exec(self, feedback_fn):
5937
    """Compute the list of all the exported system images.
5938

5939
    @rtype: dict
5940
    @return: a dictionary with the structure node->(export-list)
5941
        where export-list is a list of the instances exported on
5942
        that node.
5943

5944
    """
5945
    rpcresult = self.rpc.call_export_list(self.nodes)
5946
    result = {}
5947
    for node in rpcresult:
5948
      if rpcresult[node].failed:
5949
        result[node] = False
5950
      else:
5951
        result[node] = rpcresult[node].data
5952

    
5953
    return result
5954

    
5955

    
5956
class LUExportInstance(LogicalUnit):
5957
  """Export an instance to an image in the cluster.
5958

5959
  """
5960
  HPATH = "instance-export"
5961
  HTYPE = constants.HTYPE_INSTANCE
5962
  _OP_REQP = ["instance_name", "target_node", "shutdown"]
5963
  REQ_BGL = False
5964

    
5965
  def ExpandNames(self):
5966
    self._ExpandAndLockInstance()
5967
    # FIXME: lock only instance primary and destination node
5968
    #
5969
    # Sad but true, for now we have do lock all nodes, as we don't know where
5970
    # the previous export might be, and and in this LU we search for it and
5971
    # remove it from its current node. In the future we could fix this by:
5972
    #  - making a tasklet to search (share-lock all), then create the new one,
5973
    #    then one to remove, after
5974
    #  - removing the removal operation altoghether
5975
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5976

    
5977
  def DeclareLocks(self, level):
5978
    """Last minute lock declaration."""
5979
    # All nodes are locked anyway, so nothing to do here.
5980

    
5981
  def BuildHooksEnv(self):
5982
    """Build hooks env.
5983

5984
    This will run on the master, primary node and target node.
5985

5986
    """
5987
    env = {
5988
      "EXPORT_NODE": self.op.target_node,
5989
      "EXPORT_DO_SHUTDOWN": self.op.shutdown,
5990
      }
5991
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5992
    nl = [self.cfg.GetMasterNode(), self.instance.primary_node,
5993
          self.op.target_node]
5994
    return env, nl, nl
5995

    
5996
  def CheckPrereq(self):
5997
    """Check prerequisites.
5998

5999
    This checks that the instance and node names are valid.
6000

6001
    """
6002
    instance_name = self.op.instance_name
6003
    self.instance = self.cfg.GetInstanceInfo(instance_name)
6004
    assert self.instance is not None, \
6005
          "Cannot retrieve locked instance %s" % self.op.instance_name
6006
    _CheckNodeOnline(self, self.instance.primary_node)
6007

    
6008
    self.dst_node = self.cfg.GetNodeInfo(
6009
      self.cfg.ExpandNodeName(self.op.target_node))
6010

    
6011
    if self.dst_node is None:
6012
      # This is wrong node name, not a non-locked node
6013
      raise errors.OpPrereqError("Wrong node name %s" % self.op.target_node)
6014
    _CheckNodeOnline(self, self.dst_node.name)
6015

    
6016
    # instance disk type verification
6017
    for disk in self.instance.disks:
6018
      if disk.dev_type == constants.LD_FILE:
6019
        raise errors.OpPrereqError("Export not supported for instances with"
6020
                                   " file-based disks")
6021

    
6022
  def Exec(self, feedback_fn):
6023
    """Export an instance to an image in the cluster.
6024

6025
    """
6026
    instance = self.instance
6027
    dst_node = self.dst_node
6028
    src_node = instance.primary_node
6029
    if self.op.shutdown:
6030
      # shutdown the instance, but not the disks
6031
      result = self.rpc.call_instance_shutdown(src_node, instance)
6032
      result.Raise()
6033
      if not result.data:
6034
        raise errors.OpExecError("Could not shutdown instance %s on node %s" %
6035
                                 (instance.name, src_node))
6036

    
6037
    vgname = self.cfg.GetVGName()
6038

    
6039
    snap_disks = []
6040

    
6041
    # set the disks ID correctly since call_instance_start needs the
6042
    # correct drbd minor to create the symlinks
6043
    for disk in instance.disks:
6044
      self.cfg.SetDiskID(disk, src_node)
6045

    
6046
    try:
6047
      for disk in instance.disks:
6048
        # new_dev_name will be a snapshot of an lvm leaf of the one we passed
6049
        new_dev_name = self.rpc.call_blockdev_snapshot(src_node, disk)
6050
        if new_dev_name.failed or not new_dev_name.data:
6051
          self.LogWarning("Could not snapshot block device %s on node %s",
6052
                          disk.logical_id[1], src_node)
6053
          snap_disks.append(False)
6054
        else:
6055
          new_dev = objects.Disk(dev_type=constants.LD_LV, size=disk.size,
6056
                                 logical_id=(vgname, new_dev_name.data),
6057
                                 physical_id=(vgname, new_dev_name.data),
6058
                                 iv_name=disk.iv_name)
6059
          snap_disks.append(new_dev)
6060

    
6061
    finally:
6062
      if self.op.shutdown and instance.admin_up:
6063
        result = self.rpc.call_instance_start(src_node, instance, None)
6064
        msg = result.RemoteFailMsg()
6065
        if msg:
6066
          _ShutdownInstanceDisks(self, instance)
6067
          raise errors.OpExecError("Could not start instance: %s" % msg)
6068

    
6069
    # TODO: check for size
6070

    
6071
    cluster_name = self.cfg.GetClusterName()
6072
    for idx, dev in enumerate(snap_disks):
6073
      if dev:
6074
        result = self.rpc.call_snapshot_export(src_node, dev, dst_node.name,
6075
                                               instance, cluster_name, idx)
6076
        if result.failed or not result.data:
6077
          self.LogWarning("Could not export block device %s from node %s to"
6078
                          " node %s", dev.logical_id[1], src_node,
6079
                          dst_node.name)
6080
        result = self.rpc.call_blockdev_remove(src_node, dev)
6081
        if result.failed or not result.data:
6082
          self.LogWarning("Could not remove snapshot block device %s from node"
6083
                          " %s", dev.logical_id[1], src_node)
6084

    
6085
    result = self.rpc.call_finalize_export(dst_node.name, instance, snap_disks)
6086
    if result.failed or not result.data:
6087
      self.LogWarning("Could not finalize export for instance %s on node %s",
6088
                      instance.name, dst_node.name)
6089

    
6090
    nodelist = self.cfg.GetNodeList()
6091
    nodelist.remove(dst_node.name)
6092

    
6093
    # on one-node clusters nodelist will be empty after the removal
6094
    # if we proceed the backup would be removed because OpQueryExports
6095
    # substitutes an empty list with the full cluster node list.
6096
    if nodelist:
6097
      exportlist = self.rpc.call_export_list(nodelist)
6098
      for node in exportlist:
6099
        if exportlist[node].failed:
6100
          continue
6101
        if instance.name in exportlist[node].data:
6102
          if not self.rpc.call_export_remove(node, instance.name):
6103
            self.LogWarning("Could not remove older export for instance %s"
6104
                            " on node %s", instance.name, node)
6105

    
6106

    
6107
class LURemoveExport(NoHooksLU):
6108
  """Remove exports related to the named instance.
6109

6110
  """
6111
  _OP_REQP = ["instance_name"]
6112
  REQ_BGL = False
6113

    
6114
  def ExpandNames(self):
6115
    self.needed_locks = {}
6116
    # We need all nodes to be locked in order for RemoveExport to work, but we
6117
    # don't need to lock the instance itself, as nothing will happen to it (and
6118
    # we can remove exports also for a removed instance)
6119
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6120

    
6121
  def CheckPrereq(self):
6122
    """Check prerequisites.
6123
    """
6124
    pass
6125

    
6126
  def Exec(self, feedback_fn):
6127
    """Remove any export.
6128

6129
    """
6130
    instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
6131
    # If the instance was not found we'll try with the name that was passed in.
6132
    # This will only work if it was an FQDN, though.
6133
    fqdn_warn = False
6134
    if not instance_name:
6135
      fqdn_warn = True
6136
      instance_name = self.op.instance_name
6137

    
6138
    exportlist = self.rpc.call_export_list(self.acquired_locks[
6139
      locking.LEVEL_NODE])
6140
    found = False
6141
    for node in exportlist:
6142
      if exportlist[node].failed:
6143
        self.LogWarning("Failed to query node %s, continuing" % node)
6144
        continue
6145
      if instance_name in exportlist[node].data:
6146
        found = True
6147
        result = self.rpc.call_export_remove(node, instance_name)
6148
        if result.failed or not result.data:
6149
          logging.error("Could not remove export for instance %s"
6150
                        " on node %s", instance_name, node)
6151

    
6152
    if fqdn_warn and not found:
6153
      feedback_fn("Export not found. If trying to remove an export belonging"
6154
                  " to a deleted instance please use its Fully Qualified"
6155
                  " Domain Name.")
6156

    
6157

    
6158
class TagsLU(NoHooksLU):
6159
  """Generic tags LU.
6160

6161
  This is an abstract class which is the parent of all the other tags LUs.
6162

6163
  """
6164

    
6165
  def ExpandNames(self):
6166
    self.needed_locks = {}
6167
    if self.op.kind == constants.TAG_NODE:
6168
      name = self.cfg.ExpandNodeName(self.op.name)
6169
      if name is None:
6170
        raise errors.OpPrereqError("Invalid node name (%s)" %
6171
                                   (self.op.name,))
6172
      self.op.name = name
6173
      self.needed_locks[locking.LEVEL_NODE] = name
6174
    elif self.op.kind == constants.TAG_INSTANCE:
6175
      name = self.cfg.ExpandInstanceName(self.op.name)
6176
      if name is None:
6177
        raise errors.OpPrereqError("Invalid instance name (%s)" %
6178
                                   (self.op.name,))
6179
      self.op.name = name
6180
      self.needed_locks[locking.LEVEL_INSTANCE] = name
6181

    
6182
  def CheckPrereq(self):
6183
    """Check prerequisites.
6184

6185
    """
6186
    if self.op.kind == constants.TAG_CLUSTER:
6187
      self.target = self.cfg.GetClusterInfo()
6188
    elif self.op.kind == constants.TAG_NODE:
6189
      self.target = self.cfg.GetNodeInfo(self.op.name)
6190
    elif self.op.kind == constants.TAG_INSTANCE:
6191
      self.target = self.cfg.GetInstanceInfo(self.op.name)
6192
    else:
6193
      raise errors.OpPrereqError("Wrong tag type requested (%s)" %
6194
                                 str(self.op.kind))
6195

    
6196

    
6197
class LUGetTags(TagsLU):
6198
  """Returns the tags of a given object.
6199

6200
  """
6201
  _OP_REQP = ["kind", "name"]
6202
  REQ_BGL = False
6203

    
6204
  def Exec(self, feedback_fn):
6205
    """Returns the tag list.
6206

6207
    """
6208
    return list(self.target.GetTags())
6209

    
6210

    
6211
class LUSearchTags(NoHooksLU):
6212
  """Searches the tags for a given pattern.
6213

6214
  """
6215
  _OP_REQP = ["pattern"]
6216
  REQ_BGL = False
6217

    
6218
  def ExpandNames(self):
6219
    self.needed_locks = {}
6220

    
6221
  def CheckPrereq(self):
6222
    """Check prerequisites.
6223

6224
    This checks the pattern passed for validity by compiling it.
6225

6226
    """
6227
    try:
6228
      self.re = re.compile(self.op.pattern)
6229
    except re.error, err:
6230
      raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
6231
                                 (self.op.pattern, err))
6232

    
6233
  def Exec(self, feedback_fn):
6234
    """Returns the tag list.
6235

6236
    """
6237
    cfg = self.cfg
6238
    tgts = [("/cluster", cfg.GetClusterInfo())]
6239
    ilist = cfg.GetAllInstancesInfo().values()
6240
    tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
6241
    nlist = cfg.GetAllNodesInfo().values()
6242
    tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
6243
    results = []
6244
    for path, target in tgts:
6245
      for tag in target.GetTags():
6246
        if self.re.search(tag):
6247
          results.append((path, tag))
6248
    return results
6249

    
6250

    
6251
class LUAddTags(TagsLU):
6252
  """Sets a tag on a given object.
6253

6254
  """
6255
  _OP_REQP = ["kind", "name", "tags"]
6256
  REQ_BGL = False
6257

    
6258
  def CheckPrereq(self):
6259
    """Check prerequisites.
6260

6261
    This checks the type and length of the tag name and value.
6262

6263
    """
6264
    TagsLU.CheckPrereq(self)
6265
    for tag in self.op.tags:
6266
      objects.TaggableObject.ValidateTag(tag)
6267

    
6268
  def Exec(self, feedback_fn):
6269
    """Sets the tag.
6270

6271
    """
6272
    try:
6273
      for tag in self.op.tags:
6274
        self.target.AddTag(tag)
6275
    except errors.TagError, err:
6276
      raise errors.OpExecError("Error while setting tag: %s" % str(err))
6277
    try:
6278
      self.cfg.Update(self.target)
6279
    except errors.ConfigurationError:
6280
      raise errors.OpRetryError("There has been a modification to the"
6281
                                " config file and the operation has been"
6282
                                " aborted. Please retry.")
6283

    
6284

    
6285
class LUDelTags(TagsLU):
6286
  """Delete a list of tags from a given object.
6287

6288
  """
6289
  _OP_REQP = ["kind", "name", "tags"]
6290
  REQ_BGL = False
6291

    
6292
  def CheckPrereq(self):
6293
    """Check prerequisites.
6294

6295
    This checks that we have the given tag.
6296

6297
    """
6298
    TagsLU.CheckPrereq(self)
6299
    for tag in self.op.tags:
6300
      objects.TaggableObject.ValidateTag(tag)
6301
    del_tags = frozenset(self.op.tags)
6302
    cur_tags = self.target.GetTags()
6303
    if not del_tags <= cur_tags:
6304
      diff_tags = del_tags - cur_tags
6305
      diff_names = ["'%s'" % tag for tag in diff_tags]
6306
      diff_names.sort()
6307
      raise errors.OpPrereqError("Tag(s) %s not found" %
6308
                                 (",".join(diff_names)))
6309

    
6310
  def Exec(self, feedback_fn):
6311
    """Remove the tag from the object.
6312

6313
    """
6314
    for tag in self.op.tags:
6315
      self.target.RemoveTag(tag)
6316
    try:
6317
      self.cfg.Update(self.target)
6318
    except errors.ConfigurationError:
6319
      raise errors.OpRetryError("There has been a modification to the"
6320
                                " config file and the operation has been"
6321
                                " aborted. Please retry.")
6322

    
6323

    
6324
class LUTestDelay(NoHooksLU):
6325
  """Sleep for a specified amount of time.
6326

6327
  This LU sleeps on the master and/or nodes for a specified amount of
6328
  time.
6329

6330
  """
6331
  _OP_REQP = ["duration", "on_master", "on_nodes"]
6332
  REQ_BGL = False
6333

    
6334
  def ExpandNames(self):
6335
    """Expand names and set required locks.
6336

6337
    This expands the node list, if any.
6338

6339
    """
6340
    self.needed_locks = {}
6341
    if self.op.on_nodes:
6342
      # _GetWantedNodes can be used here, but is not always appropriate to use
6343
      # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
6344
      # more information.
6345
      self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
6346
      self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
6347

    
6348
  def CheckPrereq(self):
6349
    """Check prerequisites.
6350

6351
    """
6352

    
6353
  def Exec(self, feedback_fn):
6354
    """Do the actual sleep.
6355

6356
    """
6357
    if self.op.on_master:
6358
      if not utils.TestDelay(self.op.duration):
6359
        raise errors.OpExecError("Error during master delay test")
6360
    if self.op.on_nodes:
6361
      result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
6362
      if not result:
6363
        raise errors.OpExecError("Complete failure from rpc call")
6364
      for node, node_result in result.items():
6365
        node_result.Raise()
6366
        if not node_result.data:
6367
          raise errors.OpExecError("Failure during rpc call to node %s,"
6368
                                   " result: %s" % (node, node_result.data))
6369

    
6370

    
6371
class IAllocator(object):
6372
  """IAllocator framework.
6373

6374
  An IAllocator instance has three sets of attributes:
6375
    - cfg that is needed to query the cluster
6376
    - input data (all members of the _KEYS class attribute are required)
6377
    - four buffer attributes (in|out_data|text), that represent the
6378
      input (to the external script) in text and data structure format,
6379
      and the output from it, again in two formats
6380
    - the result variables from the script (success, info, nodes) for
6381
      easy usage
6382

6383
  """
6384
  _ALLO_KEYS = [
6385
    "mem_size", "disks", "disk_template",
6386
    "os", "tags", "nics", "vcpus", "hypervisor",
6387
    ]
6388
  _RELO_KEYS = [
6389
    "relocate_from",
6390
    ]
6391

    
6392
  def __init__(self, lu, mode, name, **kwargs):
6393
    self.lu = lu
6394
    # init buffer variables
6395
    self.in_text = self.out_text = self.in_data = self.out_data = None
6396
    # init all input fields so that pylint is happy
6397
    self.mode = mode
6398
    self.name = name
6399
    self.mem_size = self.disks = self.disk_template = None
6400
    self.os = self.tags = self.nics = self.vcpus = None
6401
    self.hypervisor = None
6402
    self.relocate_from = None
6403
    # computed fields
6404
    self.required_nodes = None
6405
    # init result fields
6406
    self.success = self.info = self.nodes = None
6407
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6408
      keyset = self._ALLO_KEYS
6409
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
6410
      keyset = self._RELO_KEYS
6411
    else:
6412
      raise errors.ProgrammerError("Unknown mode '%s' passed to the"
6413
                                   " IAllocator" % self.mode)
6414
    for key in kwargs:
6415
      if key not in keyset:
6416
        raise errors.ProgrammerError("Invalid input parameter '%s' to"
6417
                                     " IAllocator" % key)
6418
      setattr(self, key, kwargs[key])
6419
    for key in keyset:
6420
      if key not in kwargs:
6421
        raise errors.ProgrammerError("Missing input parameter '%s' to"
6422
                                     " IAllocator" % key)
6423
    self._BuildInputData()
6424

    
6425
  def _ComputeClusterData(self):
6426
    """Compute the generic allocator input data.
6427

6428
    This is the data that is independent of the actual operation.
6429

6430
    """
6431
    cfg = self.lu.cfg
6432
    cluster_info = cfg.GetClusterInfo()
6433
    # cluster data
6434
    data = {
6435
      "version": 1,
6436
      "cluster_name": cfg.GetClusterName(),
6437
      "cluster_tags": list(cluster_info.GetTags()),
6438
      "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
6439
      # we don't have job IDs
6440
      }
6441
    iinfo = cfg.GetAllInstancesInfo().values()
6442
    i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
6443

    
6444
    # node data
6445
    node_results = {}
6446
    node_list = cfg.GetNodeList()
6447

    
6448
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6449
      hypervisor_name = self.hypervisor
6450
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
6451
      hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
6452

    
6453
    node_data = self.lu.rpc.call_node_info(node_list, cfg.GetVGName(),
6454
                                           hypervisor_name)
6455
    node_iinfo = self.lu.rpc.call_all_instances_info(node_list,
6456
                       cluster_info.enabled_hypervisors)
6457
    for nname, nresult in node_data.items():
6458
      # first fill in static (config-based) values
6459
      ninfo = cfg.GetNodeInfo(nname)
6460
      pnr = {
6461
        "tags": list(ninfo.GetTags()),
6462
        "primary_ip": ninfo.primary_ip,
6463
        "secondary_ip": ninfo.secondary_ip,
6464
        "offline": ninfo.offline,
6465
        "master_candidate": ninfo.master_candidate,
6466
        }
6467

    
6468
      if not ninfo.offline:
6469
        nresult.Raise()
6470
        if not isinstance(nresult.data, dict):
6471
          raise errors.OpExecError("Can't get data for node %s" % nname)
6472
        remote_info = nresult.data
6473
        for attr in ['memory_total', 'memory_free', 'memory_dom0',
6474
                     'vg_size', 'vg_free', 'cpu_total']:
6475
          if attr not in remote_info:
6476
            raise errors.OpExecError("Node '%s' didn't return attribute"
6477
                                     " '%s'" % (nname, attr))
6478
          try:
6479
            remote_info[attr] = int(remote_info[attr])
6480
          except ValueError, err:
6481
            raise errors.OpExecError("Node '%s' returned invalid value"
6482
                                     " for '%s': %s" % (nname, attr, err))
6483
        # compute memory used by primary instances
6484
        i_p_mem = i_p_up_mem = 0
6485
        for iinfo, beinfo in i_list:
6486
          if iinfo.primary_node == nname:
6487
            i_p_mem += beinfo[constants.BE_MEMORY]
6488
            if iinfo.name not in node_iinfo[nname].data:
6489
              i_used_mem = 0
6490
            else:
6491
              i_used_mem = int(node_iinfo[nname].data[iinfo.name]['memory'])
6492
            i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
6493
            remote_info['memory_free'] -= max(0, i_mem_diff)
6494

    
6495
            if iinfo.admin_up:
6496
              i_p_up_mem += beinfo[constants.BE_MEMORY]
6497

    
6498
        # compute memory used by instances
6499
        pnr_dyn = {
6500
          "total_memory": remote_info['memory_total'],
6501
          "reserved_memory": remote_info['memory_dom0'],
6502
          "free_memory": remote_info['memory_free'],
6503
          "total_disk": remote_info['vg_size'],
6504
          "free_disk": remote_info['vg_free'],
6505
          "total_cpus": remote_info['cpu_total'],
6506
          "i_pri_memory": i_p_mem,
6507
          "i_pri_up_memory": i_p_up_mem,
6508
          }
6509
        pnr.update(pnr_dyn)
6510

    
6511
      node_results[nname] = pnr
6512
    data["nodes"] = node_results
6513

    
6514
    # instance data
6515
    instance_data = {}
6516
    for iinfo, beinfo in i_list:
6517
      nic_data = [{"mac": n.mac, "ip": n.ip, "bridge": n.bridge}
6518
                  for n in iinfo.nics]
6519
      pir = {
6520
        "tags": list(iinfo.GetTags()),
6521
        "admin_up": iinfo.admin_up,
6522
        "vcpus": beinfo[constants.BE_VCPUS],
6523
        "memory": beinfo[constants.BE_MEMORY],
6524
        "os": iinfo.os,
6525
        "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
6526
        "nics": nic_data,
6527
        "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
6528
        "disk_template": iinfo.disk_template,
6529
        "hypervisor": iinfo.hypervisor,
6530
        }
6531
      instance_data[iinfo.name] = pir
6532

    
6533
    data["instances"] = instance_data
6534

    
6535
    self.in_data = data
6536

    
6537
  def _AddNewInstance(self):
6538
    """Add new instance data to allocator structure.
6539

6540
    This in combination with _AllocatorGetClusterData will create the
6541
    correct structure needed as input for the allocator.
6542

6543
    The checks for the completeness of the opcode must have already been
6544
    done.
6545

6546
    """
6547
    data = self.in_data
6548

    
6549
    disk_space = _ComputeDiskSize(self.disk_template, self.disks)
6550

    
6551
    if self.disk_template in constants.DTS_NET_MIRROR:
6552
      self.required_nodes = 2
6553
    else:
6554
      self.required_nodes = 1
6555
    request = {
6556
      "type": "allocate",
6557
      "name": self.name,
6558
      "disk_template": self.disk_template,
6559
      "tags": self.tags,
6560
      "os": self.os,
6561
      "vcpus": self.vcpus,
6562
      "memory": self.mem_size,
6563
      "disks": self.disks,
6564
      "disk_space_total": disk_space,
6565
      "nics": self.nics,
6566
      "required_nodes": self.required_nodes,
6567
      }
6568
    data["request"] = request
6569

    
6570
  def _AddRelocateInstance(self):
6571
    """Add relocate instance data to allocator structure.
6572

6573
    This in combination with _IAllocatorGetClusterData will create the
6574
    correct structure needed as input for the allocator.
6575

6576
    The checks for the completeness of the opcode must have already been
6577
    done.
6578

6579
    """
6580
    instance = self.lu.cfg.GetInstanceInfo(self.name)
6581
    if instance is None:
6582
      raise errors.ProgrammerError("Unknown instance '%s' passed to"
6583
                                   " IAllocator" % self.name)
6584

    
6585
    if instance.disk_template not in constants.DTS_NET_MIRROR:
6586
      raise errors.OpPrereqError("Can't relocate non-mirrored instances")
6587

    
6588
    if len(instance.secondary_nodes) != 1:
6589
      raise errors.OpPrereqError("Instance has not exactly one secondary node")
6590

    
6591
    self.required_nodes = 1
6592
    disk_sizes = [{'size': disk.size} for disk in instance.disks]
6593
    disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
6594

    
6595
    request = {
6596
      "type": "relocate",
6597
      "name": self.name,
6598
      "disk_space_total": disk_space,
6599
      "required_nodes": self.required_nodes,
6600
      "relocate_from": self.relocate_from,
6601
      }
6602
    self.in_data["request"] = request
6603

    
6604
  def _BuildInputData(self):
6605
    """Build input data structures.
6606

6607
    """
6608
    self._ComputeClusterData()
6609

    
6610
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6611
      self._AddNewInstance()
6612
    else:
6613
      self._AddRelocateInstance()
6614

    
6615
    self.in_text = serializer.Dump(self.in_data)
6616

    
6617
  def Run(self, name, validate=True, call_fn=None):
6618
    """Run an instance allocator and return the results.
6619

6620
    """
6621
    if call_fn is None:
6622
      call_fn = self.lu.rpc.call_iallocator_runner
6623
    data = self.in_text
6624

    
6625
    result = call_fn(self.lu.cfg.GetMasterNode(), name, self.in_text)
6626
    result.Raise()
6627

    
6628
    if not isinstance(result.data, (list, tuple)) or len(result.data) != 4:
6629
      raise errors.OpExecError("Invalid result from master iallocator runner")
6630

    
6631
    rcode, stdout, stderr, fail = result.data
6632

    
6633
    if rcode == constants.IARUN_NOTFOUND:
6634
      raise errors.OpExecError("Can't find allocator '%s'" % name)
6635
    elif rcode == constants.IARUN_FAILURE:
6636
      raise errors.OpExecError("Instance allocator call failed: %s,"
6637
                               " output: %s" % (fail, stdout+stderr))
6638
    self.out_text = stdout
6639
    if validate:
6640
      self._ValidateResult()
6641

    
6642
  def _ValidateResult(self):
6643
    """Process the allocator results.
6644

6645
    This will process and if successful save the result in
6646
    self.out_data and the other parameters.
6647

6648
    """
6649
    try:
6650
      rdict = serializer.Load(self.out_text)
6651
    except Exception, err:
6652
      raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
6653

    
6654
    if not isinstance(rdict, dict):
6655
      raise errors.OpExecError("Can't parse iallocator results: not a dict")
6656

    
6657
    for key in "success", "info", "nodes":
6658
      if key not in rdict:
6659
        raise errors.OpExecError("Can't parse iallocator results:"
6660
                                 " missing key '%s'" % key)
6661
      setattr(self, key, rdict[key])
6662

    
6663
    if not isinstance(rdict["nodes"], list):
6664
      raise errors.OpExecError("Can't parse iallocator results: 'nodes' key"
6665
                               " is not a list")
6666
    self.out_data = rdict
6667

    
6668

    
6669
class LUTestAllocator(NoHooksLU):
6670
  """Run allocator tests.
6671

6672
  This LU runs the allocator tests
6673

6674
  """
6675
  _OP_REQP = ["direction", "mode", "name"]
6676

    
6677
  def CheckPrereq(self):
6678
    """Check prerequisites.
6679

6680
    This checks the opcode parameters depending on the director and mode test.
6681

6682
    """
6683
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
6684
      for attr in ["name", "mem_size", "disks", "disk_template",
6685
                   "os", "tags", "nics", "vcpus"]:
6686
        if not hasattr(self.op, attr):
6687
          raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
6688
                                     attr)
6689
      iname = self.cfg.ExpandInstanceName(self.op.name)
6690
      if iname is not None:
6691
        raise errors.OpPrereqError("Instance '%s' already in the cluster" %
6692
                                   iname)
6693
      if not isinstance(self.op.nics, list):
6694
        raise errors.OpPrereqError("Invalid parameter 'nics'")
6695
      for row in self.op.nics:
6696
        if (not isinstance(row, dict) or
6697
            "mac" not in row or
6698
            "ip" not in row or
6699
            "bridge" not in row):
6700
          raise errors.OpPrereqError("Invalid contents of the"
6701
                                     " 'nics' parameter")
6702
      if not isinstance(self.op.disks, list):
6703
        raise errors.OpPrereqError("Invalid parameter 'disks'")
6704
      for row in self.op.disks:
6705
        if (not isinstance(row, dict) or
6706
            "size" not in row or
6707
            not isinstance(row["size"], int) or
6708
            "mode" not in row or
6709
            row["mode"] not in ['r', 'w']):
6710
          raise errors.OpPrereqError("Invalid contents of the"
6711
                                     " 'disks' parameter")
6712
      if not hasattr(self.op, "hypervisor") or self.op.hypervisor is None:
6713
        self.op.hypervisor = self.cfg.GetHypervisorType()
6714
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
6715
      if not hasattr(self.op, "name"):
6716
        raise errors.OpPrereqError("Missing attribute 'name' on opcode input")
6717
      fname = self.cfg.ExpandInstanceName(self.op.name)
6718
      if fname is None:
6719
        raise errors.OpPrereqError("Instance '%s' not found for relocation" %
6720
                                   self.op.name)
6721
      self.op.name = fname
6722
      self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
6723
    else:
6724
      raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
6725
                                 self.op.mode)
6726

    
6727
    if self.op.direction == constants.IALLOCATOR_DIR_OUT:
6728
      if not hasattr(self.op, "allocator") or self.op.allocator is None:
6729
        raise errors.OpPrereqError("Missing allocator name")
6730
    elif self.op.direction != constants.IALLOCATOR_DIR_IN:
6731
      raise errors.OpPrereqError("Wrong allocator test '%s'" %
6732
                                 self.op.direction)
6733

    
6734
  def Exec(self, feedback_fn):
6735
    """Run the allocator test.
6736

6737
    """
6738
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
6739
      ial = IAllocator(self,
6740
                       mode=self.op.mode,
6741
                       name=self.op.name,
6742
                       mem_size=self.op.mem_size,
6743
                       disks=self.op.disks,
6744
                       disk_template=self.op.disk_template,
6745
                       os=self.op.os,
6746
                       tags=self.op.tags,
6747
                       nics=self.op.nics,
6748
                       vcpus=self.op.vcpus,
6749
                       hypervisor=self.op.hypervisor,
6750
                       )
6751
    else:
6752
      ial = IAllocator(self,
6753
                       mode=self.op.mode,
6754
                       name=self.op.name,
6755
                       relocate_from=list(self.relocate_from),
6756
                       )
6757

    
6758
    if self.op.direction == constants.IALLOCATOR_DIR_IN:
6759
      result = ial.in_text
6760
    else:
6761
      ial.Run(self.op.allocator, validate=False)
6762
      result = ial.out_text
6763
    return result