Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ 23829f6f

History | View | Annotate | Download (234.1 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
      if result.failed or not result:
2463
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
2464
                           " (is_primary=False, pass=1)",
2465
                           inst_disk.iv_name, node)
2466
        if not ignore_secondaries:
2467
          disks_ok = False
2468

    
2469
  # FIXME: race condition on drbd migration to primary
2470

    
2471
  # 2nd pass, do only the primary node
2472
  for inst_disk in instance.disks:
2473
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
2474
      if node != instance.primary_node:
2475
        continue
2476
      lu.cfg.SetDiskID(node_disk, node)
2477
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
2478
      if result.failed or not result:
2479
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
2480
                           " (is_primary=True, pass=2)",
2481
                           inst_disk.iv_name, node)
2482
        disks_ok = False
2483
    device_info.append((instance.primary_node, inst_disk.iv_name, result.data))
2484

    
2485
  # leave the disks configured for the primary node
2486
  # this is a workaround that would be fixed better by
2487
  # improving the logical/physical id handling
2488
  for disk in instance.disks:
2489
    lu.cfg.SetDiskID(disk, instance.primary_node)
2490

    
2491
  return disks_ok, device_info
2492

    
2493

    
2494
def _StartInstanceDisks(lu, instance, force):
2495
  """Start the disks of an instance.
2496

2497
  """
2498
  disks_ok, dummy = _AssembleInstanceDisks(lu, instance,
2499
                                           ignore_secondaries=force)
2500
  if not disks_ok:
2501
    _ShutdownInstanceDisks(lu, instance)
2502
    if force is not None and not force:
2503
      lu.proc.LogWarning("", hint="If the message above refers to a"
2504
                         " secondary node,"
2505
                         " you can retry the operation using '--force'.")
2506
    raise errors.OpExecError("Disk consistency error")
2507

    
2508

    
2509
class LUDeactivateInstanceDisks(NoHooksLU):
2510
  """Shutdown an instance's disks.
2511

2512
  """
2513
  _OP_REQP = ["instance_name"]
2514
  REQ_BGL = False
2515

    
2516
  def ExpandNames(self):
2517
    self._ExpandAndLockInstance()
2518
    self.needed_locks[locking.LEVEL_NODE] = []
2519
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2520

    
2521
  def DeclareLocks(self, level):
2522
    if level == locking.LEVEL_NODE:
2523
      self._LockInstancesNodes()
2524

    
2525
  def CheckPrereq(self):
2526
    """Check prerequisites.
2527

2528
    This checks that the instance is in the cluster.
2529

2530
    """
2531
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2532
    assert self.instance is not None, \
2533
      "Cannot retrieve locked instance %s" % self.op.instance_name
2534

    
2535
  def Exec(self, feedback_fn):
2536
    """Deactivate the disks
2537

2538
    """
2539
    instance = self.instance
2540
    _SafeShutdownInstanceDisks(self, instance)
2541

    
2542

    
2543
def _SafeShutdownInstanceDisks(lu, instance):
2544
  """Shutdown block devices of an instance.
2545

2546
  This function checks if an instance is running, before calling
2547
  _ShutdownInstanceDisks.
2548

2549
  """
2550
  ins_l = lu.rpc.call_instance_list([instance.primary_node],
2551
                                      [instance.hypervisor])
2552
  ins_l = ins_l[instance.primary_node]
2553
  if ins_l.failed or not isinstance(ins_l.data, list):
2554
    raise errors.OpExecError("Can't contact node '%s'" %
2555
                             instance.primary_node)
2556

    
2557
  if instance.name in ins_l.data:
2558
    raise errors.OpExecError("Instance is running, can't shutdown"
2559
                             " block devices.")
2560

    
2561
  _ShutdownInstanceDisks(lu, instance)
2562

    
2563

    
2564
def _ShutdownInstanceDisks(lu, instance, ignore_primary=False):
2565
  """Shutdown block devices of an instance.
2566

2567
  This does the shutdown on all nodes of the instance.
2568

2569
  If the ignore_primary is false, errors on the primary node are
2570
  ignored.
2571

2572
  """
2573
  result = True
2574
  for disk in instance.disks:
2575
    for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
2576
      lu.cfg.SetDiskID(top_disk, node)
2577
      result = lu.rpc.call_blockdev_shutdown(node, top_disk)
2578
      if result.failed or not result.data:
2579
        logging.error("Could not shutdown block device %s on node %s",
2580
                      disk.iv_name, node)
2581
        if not ignore_primary or node != instance.primary_node:
2582
          result = False
2583
  return result
2584

    
2585

    
2586
def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
2587
  """Checks if a node has enough free memory.
2588

2589
  This function check if a given node has the needed amount of free
2590
  memory. In case the node has less memory or we cannot get the
2591
  information from the node, this function raise an OpPrereqError
2592
  exception.
2593

2594
  @type lu: C{LogicalUnit}
2595
  @param lu: a logical unit from which we get configuration data
2596
  @type node: C{str}
2597
  @param node: the node to check
2598
  @type reason: C{str}
2599
  @param reason: string to use in the error message
2600
  @type requested: C{int}
2601
  @param requested: the amount of memory in MiB to check for
2602
  @type hypervisor_name: C{str}
2603
  @param hypervisor_name: the hypervisor to ask for memory stats
2604
  @raise errors.OpPrereqError: if the node doesn't have enough memory, or
2605
      we cannot check the node
2606

2607
  """
2608
  nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
2609
  nodeinfo[node].Raise()
2610
  free_mem = nodeinfo[node].data.get('memory_free')
2611
  if not isinstance(free_mem, int):
2612
    raise errors.OpPrereqError("Can't compute free memory on node %s, result"
2613
                             " was '%s'" % (node, free_mem))
2614
  if requested > free_mem:
2615
    raise errors.OpPrereqError("Not enough memory on node %s for %s:"
2616
                             " needed %s MiB, available %s MiB" %
2617
                             (node, reason, requested, free_mem))
2618

    
2619

    
2620
class LUStartupInstance(LogicalUnit):
2621
  """Starts an instance.
2622

2623
  """
2624
  HPATH = "instance-start"
2625
  HTYPE = constants.HTYPE_INSTANCE
2626
  _OP_REQP = ["instance_name", "force"]
2627
  REQ_BGL = False
2628

    
2629
  def ExpandNames(self):
2630
    self._ExpandAndLockInstance()
2631

    
2632
  def BuildHooksEnv(self):
2633
    """Build hooks env.
2634

2635
    This runs on master, primary and secondary nodes of the instance.
2636

2637
    """
2638
    env = {
2639
      "FORCE": self.op.force,
2640
      }
2641
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2642
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2643
    return env, nl, nl
2644

    
2645
  def CheckPrereq(self):
2646
    """Check prerequisites.
2647

2648
    This checks that the instance is in the cluster.
2649

2650
    """
2651
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2652
    assert self.instance is not None, \
2653
      "Cannot retrieve locked instance %s" % self.op.instance_name
2654

    
2655
    _CheckNodeOnline(self, instance.primary_node)
2656

    
2657
    bep = self.cfg.GetClusterInfo().FillBE(instance)
2658
    # check bridges existance
2659
    _CheckInstanceBridgesExist(self, instance)
2660

    
2661
    _CheckNodeFreeMemory(self, instance.primary_node,
2662
                         "starting instance %s" % instance.name,
2663
                         bep[constants.BE_MEMORY], instance.hypervisor)
2664

    
2665
  def Exec(self, feedback_fn):
2666
    """Start the instance.
2667

2668
    """
2669
    instance = self.instance
2670
    force = self.op.force
2671
    extra_args = getattr(self.op, "extra_args", "")
2672

    
2673
    self.cfg.MarkInstanceUp(instance.name)
2674

    
2675
    node_current = instance.primary_node
2676

    
2677
    _StartInstanceDisks(self, instance, force)
2678

    
2679
    result = self.rpc.call_instance_start(node_current, instance, extra_args)
2680
    msg = result.RemoteFailMsg()
2681
    if msg:
2682
      _ShutdownInstanceDisks(self, instance)
2683
      raise errors.OpExecError("Could not start instance: %s" % msg)
2684

    
2685

    
2686
class LURebootInstance(LogicalUnit):
2687
  """Reboot an instance.
2688

2689
  """
2690
  HPATH = "instance-reboot"
2691
  HTYPE = constants.HTYPE_INSTANCE
2692
  _OP_REQP = ["instance_name", "ignore_secondaries", "reboot_type"]
2693
  REQ_BGL = False
2694

    
2695
  def ExpandNames(self):
2696
    if self.op.reboot_type not in [constants.INSTANCE_REBOOT_SOFT,
2697
                                   constants.INSTANCE_REBOOT_HARD,
2698
                                   constants.INSTANCE_REBOOT_FULL]:
2699
      raise errors.ParameterError("reboot type not in [%s, %s, %s]" %
2700
                                  (constants.INSTANCE_REBOOT_SOFT,
2701
                                   constants.INSTANCE_REBOOT_HARD,
2702
                                   constants.INSTANCE_REBOOT_FULL))
2703
    self._ExpandAndLockInstance()
2704

    
2705
  def BuildHooksEnv(self):
2706
    """Build hooks env.
2707

2708
    This runs on master, primary and secondary nodes of the instance.
2709

2710
    """
2711
    env = {
2712
      "IGNORE_SECONDARIES": self.op.ignore_secondaries,
2713
      }
2714
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2715
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2716
    return env, nl, nl
2717

    
2718
  def CheckPrereq(self):
2719
    """Check prerequisites.
2720

2721
    This checks that the instance is in the cluster.
2722

2723
    """
2724
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2725
    assert self.instance is not None, \
2726
      "Cannot retrieve locked instance %s" % self.op.instance_name
2727

    
2728
    _CheckNodeOnline(self, instance.primary_node)
2729

    
2730
    # check bridges existance
2731
    _CheckInstanceBridgesExist(self, instance)
2732

    
2733
  def Exec(self, feedback_fn):
2734
    """Reboot the instance.
2735

2736
    """
2737
    instance = self.instance
2738
    ignore_secondaries = self.op.ignore_secondaries
2739
    reboot_type = self.op.reboot_type
2740
    extra_args = getattr(self.op, "extra_args", "")
2741

    
2742
    node_current = instance.primary_node
2743

    
2744
    if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
2745
                       constants.INSTANCE_REBOOT_HARD]:
2746
      result = self.rpc.call_instance_reboot(node_current, instance,
2747
                                             reboot_type, extra_args)
2748
      if result.failed or not result.data:
2749
        raise errors.OpExecError("Could not reboot instance")
2750
    else:
2751
      if not self.rpc.call_instance_shutdown(node_current, instance):
2752
        raise errors.OpExecError("could not shutdown instance for full reboot")
2753
      _ShutdownInstanceDisks(self, instance)
2754
      _StartInstanceDisks(self, instance, ignore_secondaries)
2755
      result = self.rpc.call_instance_start(node_current, instance, extra_args)
2756
      msg = result.RemoteFailMsg()
2757
      if msg:
2758
        _ShutdownInstanceDisks(self, instance)
2759
        raise errors.OpExecError("Could not start instance for"
2760
                                 " full reboot: %s" % msg)
2761

    
2762
    self.cfg.MarkInstanceUp(instance.name)
2763

    
2764

    
2765
class LUShutdownInstance(LogicalUnit):
2766
  """Shutdown an instance.
2767

2768
  """
2769
  HPATH = "instance-stop"
2770
  HTYPE = constants.HTYPE_INSTANCE
2771
  _OP_REQP = ["instance_name"]
2772
  REQ_BGL = False
2773

    
2774
  def ExpandNames(self):
2775
    self._ExpandAndLockInstance()
2776

    
2777
  def BuildHooksEnv(self):
2778
    """Build hooks env.
2779

2780
    This runs on master, primary and secondary nodes of the instance.
2781

2782
    """
2783
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2784
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2785
    return env, nl, nl
2786

    
2787
  def CheckPrereq(self):
2788
    """Check prerequisites.
2789

2790
    This checks that the instance is in the cluster.
2791

2792
    """
2793
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2794
    assert self.instance is not None, \
2795
      "Cannot retrieve locked instance %s" % self.op.instance_name
2796
    _CheckNodeOnline(self, self.instance.primary_node)
2797

    
2798
  def Exec(self, feedback_fn):
2799
    """Shutdown the instance.
2800

2801
    """
2802
    instance = self.instance
2803
    node_current = instance.primary_node
2804
    self.cfg.MarkInstanceDown(instance.name)
2805
    result = self.rpc.call_instance_shutdown(node_current, instance)
2806
    if result.failed or not result.data:
2807
      self.proc.LogWarning("Could not shutdown instance")
2808

    
2809
    _ShutdownInstanceDisks(self, instance)
2810

    
2811

    
2812
class LUReinstallInstance(LogicalUnit):
2813
  """Reinstall an instance.
2814

2815
  """
2816
  HPATH = "instance-reinstall"
2817
  HTYPE = constants.HTYPE_INSTANCE
2818
  _OP_REQP = ["instance_name"]
2819
  REQ_BGL = False
2820

    
2821
  def ExpandNames(self):
2822
    self._ExpandAndLockInstance()
2823

    
2824
  def BuildHooksEnv(self):
2825
    """Build hooks env.
2826

2827
    This runs on master, primary and secondary nodes of the instance.
2828

2829
    """
2830
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2831
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2832
    return env, nl, nl
2833

    
2834
  def CheckPrereq(self):
2835
    """Check prerequisites.
2836

2837
    This checks that the instance is in the cluster and is not running.
2838

2839
    """
2840
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2841
    assert instance is not None, \
2842
      "Cannot retrieve locked instance %s" % self.op.instance_name
2843
    _CheckNodeOnline(self, instance.primary_node)
2844

    
2845
    if instance.disk_template == constants.DT_DISKLESS:
2846
      raise errors.OpPrereqError("Instance '%s' has no disks" %
2847
                                 self.op.instance_name)
2848
    if instance.admin_up:
2849
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
2850
                                 self.op.instance_name)
2851
    remote_info = self.rpc.call_instance_info(instance.primary_node,
2852
                                              instance.name,
2853
                                              instance.hypervisor)
2854
    if remote_info.failed or remote_info.data:
2855
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
2856
                                 (self.op.instance_name,
2857
                                  instance.primary_node))
2858

    
2859
    self.op.os_type = getattr(self.op, "os_type", None)
2860
    if self.op.os_type is not None:
2861
      # OS verification
2862
      pnode = self.cfg.GetNodeInfo(
2863
        self.cfg.ExpandNodeName(instance.primary_node))
2864
      if pnode is None:
2865
        raise errors.OpPrereqError("Primary node '%s' is unknown" %
2866
                                   self.op.pnode)
2867
      result = self.rpc.call_os_get(pnode.name, self.op.os_type)
2868
      result.Raise()
2869
      if not isinstance(result.data, objects.OS):
2870
        raise errors.OpPrereqError("OS '%s' not in supported OS list for"
2871
                                   " primary node"  % self.op.os_type)
2872

    
2873
    self.instance = instance
2874

    
2875
  def Exec(self, feedback_fn):
2876
    """Reinstall the instance.
2877

2878
    """
2879
    inst = self.instance
2880

    
2881
    if self.op.os_type is not None:
2882
      feedback_fn("Changing OS to '%s'..." % self.op.os_type)
2883
      inst.os = self.op.os_type
2884
      self.cfg.Update(inst)
2885

    
2886
    _StartInstanceDisks(self, inst, None)
2887
    try:
2888
      feedback_fn("Running the instance OS create scripts...")
2889
      result = self.rpc.call_instance_os_add(inst.primary_node, inst)
2890
      msg = result.RemoteFailMsg()
2891
      if msg:
2892
        raise errors.OpExecError("Could not install OS for instance %s"
2893
                                 " on node %s: %s" %
2894
                                 (inst.name, inst.primary_node, msg))
2895
    finally:
2896
      _ShutdownInstanceDisks(self, inst)
2897

    
2898

    
2899
class LURenameInstance(LogicalUnit):
2900
  """Rename an instance.
2901

2902
  """
2903
  HPATH = "instance-rename"
2904
  HTYPE = constants.HTYPE_INSTANCE
2905
  _OP_REQP = ["instance_name", "new_name"]
2906

    
2907
  def BuildHooksEnv(self):
2908
    """Build hooks env.
2909

2910
    This runs on master, primary and secondary nodes of the instance.
2911

2912
    """
2913
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2914
    env["INSTANCE_NEW_NAME"] = self.op.new_name
2915
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
2916
    return env, nl, nl
2917

    
2918
  def CheckPrereq(self):
2919
    """Check prerequisites.
2920

2921
    This checks that the instance is in the cluster and is not running.
2922

2923
    """
2924
    instance = self.cfg.GetInstanceInfo(
2925
      self.cfg.ExpandInstanceName(self.op.instance_name))
2926
    if instance is None:
2927
      raise errors.OpPrereqError("Instance '%s' not known" %
2928
                                 self.op.instance_name)
2929
    _CheckNodeOnline(self, instance.primary_node)
2930

    
2931
    if instance.admin_up:
2932
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
2933
                                 self.op.instance_name)
2934
    remote_info = self.rpc.call_instance_info(instance.primary_node,
2935
                                              instance.name,
2936
                                              instance.hypervisor)
2937
    remote_info.Raise()
2938
    if remote_info.data:
2939
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
2940
                                 (self.op.instance_name,
2941
                                  instance.primary_node))
2942
    self.instance = instance
2943

    
2944
    # new name verification
2945
    name_info = utils.HostInfo(self.op.new_name)
2946

    
2947
    self.op.new_name = new_name = name_info.name
2948
    instance_list = self.cfg.GetInstanceList()
2949
    if new_name in instance_list:
2950
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
2951
                                 new_name)
2952

    
2953
    if not getattr(self.op, "ignore_ip", False):
2954
      if utils.TcpPing(name_info.ip, constants.DEFAULT_NODED_PORT):
2955
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
2956
                                   (name_info.ip, new_name))
2957

    
2958

    
2959
  def Exec(self, feedback_fn):
2960
    """Reinstall the instance.
2961

2962
    """
2963
    inst = self.instance
2964
    old_name = inst.name
2965

    
2966
    if inst.disk_template == constants.DT_FILE:
2967
      old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
2968

    
2969
    self.cfg.RenameInstance(inst.name, self.op.new_name)
2970
    # Change the instance lock. This is definitely safe while we hold the BGL
2971
    self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
2972
    self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
2973

    
2974
    # re-read the instance from the configuration after rename
2975
    inst = self.cfg.GetInstanceInfo(self.op.new_name)
2976

    
2977
    if inst.disk_template == constants.DT_FILE:
2978
      new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
2979
      result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
2980
                                                     old_file_storage_dir,
2981
                                                     new_file_storage_dir)
2982
      result.Raise()
2983
      if not result.data:
2984
        raise errors.OpExecError("Could not connect to node '%s' to rename"
2985
                                 " directory '%s' to '%s' (but the instance"
2986
                                 " has been renamed in Ganeti)" % (
2987
                                 inst.primary_node, old_file_storage_dir,
2988
                                 new_file_storage_dir))
2989

    
2990
      if not result.data[0]:
2991
        raise errors.OpExecError("Could not rename directory '%s' to '%s'"
2992
                                 " (but the instance has been renamed in"
2993
                                 " Ganeti)" % (old_file_storage_dir,
2994
                                               new_file_storage_dir))
2995

    
2996
    _StartInstanceDisks(self, inst, None)
2997
    try:
2998
      result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
2999
                                                 old_name)
3000
      msg = result.RemoteFailMsg()
3001
      if msg:
3002
        msg = ("Could not run OS rename script for instance %s on node %s"
3003
               " (but the instance has been renamed in Ganeti): %s" %
3004
               (inst.name, inst.primary_node, msg))
3005
        self.proc.LogWarning(msg)
3006
    finally:
3007
      _ShutdownInstanceDisks(self, inst)
3008

    
3009

    
3010
class LURemoveInstance(LogicalUnit):
3011
  """Remove an instance.
3012

3013
  """
3014
  HPATH = "instance-remove"
3015
  HTYPE = constants.HTYPE_INSTANCE
3016
  _OP_REQP = ["instance_name", "ignore_failures"]
3017
  REQ_BGL = False
3018

    
3019
  def ExpandNames(self):
3020
    self._ExpandAndLockInstance()
3021
    self.needed_locks[locking.LEVEL_NODE] = []
3022
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3023

    
3024
  def DeclareLocks(self, level):
3025
    if level == locking.LEVEL_NODE:
3026
      self._LockInstancesNodes()
3027

    
3028
  def BuildHooksEnv(self):
3029
    """Build hooks env.
3030

3031
    This runs on master, primary and secondary nodes of the instance.
3032

3033
    """
3034
    env = _BuildInstanceHookEnvByObject(self, self.instance)
3035
    nl = [self.cfg.GetMasterNode()]
3036
    return env, nl, nl
3037

    
3038
  def CheckPrereq(self):
3039
    """Check prerequisites.
3040

3041
    This checks that the instance is in the cluster.
3042

3043
    """
3044
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3045
    assert self.instance is not None, \
3046
      "Cannot retrieve locked instance %s" % self.op.instance_name
3047

    
3048
  def Exec(self, feedback_fn):
3049
    """Remove the instance.
3050

3051
    """
3052
    instance = self.instance
3053
    logging.info("Shutting down instance %s on node %s",
3054
                 instance.name, instance.primary_node)
3055

    
3056
    result = self.rpc.call_instance_shutdown(instance.primary_node, instance)
3057
    if result.failed or not result.data:
3058
      if self.op.ignore_failures:
3059
        feedback_fn("Warning: can't shutdown instance")
3060
      else:
3061
        raise errors.OpExecError("Could not shutdown instance %s on node %s" %
3062
                                 (instance.name, instance.primary_node))
3063

    
3064
    logging.info("Removing block devices for instance %s", instance.name)
3065

    
3066
    if not _RemoveDisks(self, instance):
3067
      if self.op.ignore_failures:
3068
        feedback_fn("Warning: can't remove instance's disks")
3069
      else:
3070
        raise errors.OpExecError("Can't remove instance's disks")
3071

    
3072
    logging.info("Removing instance %s out of cluster config", instance.name)
3073

    
3074
    self.cfg.RemoveInstance(instance.name)
3075
    self.remove_locks[locking.LEVEL_INSTANCE] = instance.name
3076

    
3077

    
3078
class LUQueryInstances(NoHooksLU):
3079
  """Logical unit for querying instances.
3080

3081
  """
3082
  _OP_REQP = ["output_fields", "names", "use_locking"]
3083
  REQ_BGL = False
3084
  _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
3085
                                    "admin_state", "admin_ram",
3086
                                    "disk_template", "ip", "mac", "bridge",
3087
                                    "sda_size", "sdb_size", "vcpus", "tags",
3088
                                    "network_port", "beparams",
3089
                                    "(disk).(size)/([0-9]+)",
3090
                                    "(disk).(sizes)",
3091
                                    "(nic).(mac|ip|bridge)/([0-9]+)",
3092
                                    "(nic).(macs|ips|bridges)",
3093
                                    "(disk|nic).(count)",
3094
                                    "serial_no", "hypervisor", "hvparams",] +
3095
                                  ["hv/%s" % name
3096
                                   for name in constants.HVS_PARAMETERS] +
3097
                                  ["be/%s" % name
3098
                                   for name in constants.BES_PARAMETERS])
3099
  _FIELDS_DYNAMIC = utils.FieldSet("oper_state", "oper_ram", "status")
3100

    
3101

    
3102
  def ExpandNames(self):
3103
    _CheckOutputFields(static=self._FIELDS_STATIC,
3104
                       dynamic=self._FIELDS_DYNAMIC,
3105
                       selected=self.op.output_fields)
3106

    
3107
    self.needed_locks = {}
3108
    self.share_locks[locking.LEVEL_INSTANCE] = 1
3109
    self.share_locks[locking.LEVEL_NODE] = 1
3110

    
3111
    if self.op.names:
3112
      self.wanted = _GetWantedInstances(self, self.op.names)
3113
    else:
3114
      self.wanted = locking.ALL_SET
3115

    
3116
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
3117
    self.do_locking = self.do_node_query and self.op.use_locking
3118
    if self.do_locking:
3119
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
3120
      self.needed_locks[locking.LEVEL_NODE] = []
3121
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3122

    
3123
  def DeclareLocks(self, level):
3124
    if level == locking.LEVEL_NODE and self.do_locking:
3125
      self._LockInstancesNodes()
3126

    
3127
  def CheckPrereq(self):
3128
    """Check prerequisites.
3129

3130
    """
3131
    pass
3132

    
3133
  def Exec(self, feedback_fn):
3134
    """Computes the list of nodes and their attributes.
3135

3136
    """
3137
    all_info = self.cfg.GetAllInstancesInfo()
3138
    if self.wanted == locking.ALL_SET:
3139
      # caller didn't specify instance names, so ordering is not important
3140
      if self.do_locking:
3141
        instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
3142
      else:
3143
        instance_names = all_info.keys()
3144
      instance_names = utils.NiceSort(instance_names)
3145
    else:
3146
      # caller did specify names, so we must keep the ordering
3147
      if self.do_locking:
3148
        tgt_set = self.acquired_locks[locking.LEVEL_INSTANCE]
3149
      else:
3150
        tgt_set = all_info.keys()
3151
      missing = set(self.wanted).difference(tgt_set)
3152
      if missing:
3153
        raise errors.OpExecError("Some instances were removed before"
3154
                                 " retrieving their data: %s" % missing)
3155
      instance_names = self.wanted
3156

    
3157
    instance_list = [all_info[iname] for iname in instance_names]
3158

    
3159
    # begin data gathering
3160

    
3161
    nodes = frozenset([inst.primary_node for inst in instance_list])
3162
    hv_list = list(set([inst.hypervisor for inst in instance_list]))
3163

    
3164
    bad_nodes = []
3165
    off_nodes = []
3166
    if self.do_node_query:
3167
      live_data = {}
3168
      node_data = self.rpc.call_all_instances_info(nodes, hv_list)
3169
      for name in nodes:
3170
        result = node_data[name]
3171
        if result.offline:
3172
          # offline nodes will be in both lists
3173
          off_nodes.append(name)
3174
        if result.failed:
3175
          bad_nodes.append(name)
3176
        else:
3177
          if result.data:
3178
            live_data.update(result.data)
3179
            # else no instance is alive
3180
    else:
3181
      live_data = dict([(name, {}) for name in instance_names])
3182

    
3183
    # end data gathering
3184

    
3185
    HVPREFIX = "hv/"
3186
    BEPREFIX = "be/"
3187
    output = []
3188
    for instance in instance_list:
3189
      iout = []
3190
      i_hv = self.cfg.GetClusterInfo().FillHV(instance)
3191
      i_be = self.cfg.GetClusterInfo().FillBE(instance)
3192
      for field in self.op.output_fields:
3193
        st_match = self._FIELDS_STATIC.Matches(field)
3194
        if field == "name":
3195
          val = instance.name
3196
        elif field == "os":
3197
          val = instance.os
3198
        elif field == "pnode":
3199
          val = instance.primary_node
3200
        elif field == "snodes":
3201
          val = list(instance.secondary_nodes)
3202
        elif field == "admin_state":
3203
          val = instance.admin_up
3204
        elif field == "oper_state":
3205
          if instance.primary_node in bad_nodes:
3206
            val = None
3207
          else:
3208
            val = bool(live_data.get(instance.name))
3209
        elif field == "status":
3210
          if instance.primary_node in off_nodes:
3211
            val = "ERROR_nodeoffline"
3212
          elif instance.primary_node in bad_nodes:
3213
            val = "ERROR_nodedown"
3214
          else:
3215
            running = bool(live_data.get(instance.name))
3216
            if running:
3217
              if instance.admin_up:
3218
                val = "running"
3219
              else:
3220
                val = "ERROR_up"
3221
            else:
3222
              if instance.admin_up:
3223
                val = "ERROR_down"
3224
              else:
3225
                val = "ADMIN_down"
3226
        elif field == "oper_ram":
3227
          if instance.primary_node in bad_nodes:
3228
            val = None
3229
          elif instance.name in live_data:
3230
            val = live_data[instance.name].get("memory", "?")
3231
          else:
3232
            val = "-"
3233
        elif field == "disk_template":
3234
          val = instance.disk_template
3235
        elif field == "ip":
3236
          val = instance.nics[0].ip
3237
        elif field == "bridge":
3238
          val = instance.nics[0].bridge
3239
        elif field == "mac":
3240
          val = instance.nics[0].mac
3241
        elif field == "sda_size" or field == "sdb_size":
3242
          idx = ord(field[2]) - ord('a')
3243
          try:
3244
            val = instance.FindDisk(idx).size
3245
          except errors.OpPrereqError:
3246
            val = None
3247
        elif field == "tags":
3248
          val = list(instance.GetTags())
3249
        elif field == "serial_no":
3250
          val = instance.serial_no
3251
        elif field == "network_port":
3252
          val = instance.network_port
3253
        elif field == "hypervisor":
3254
          val = instance.hypervisor
3255
        elif field == "hvparams":
3256
          val = i_hv
3257
        elif (field.startswith(HVPREFIX) and
3258
              field[len(HVPREFIX):] in constants.HVS_PARAMETERS):
3259
          val = i_hv.get(field[len(HVPREFIX):], None)
3260
        elif field == "beparams":
3261
          val = i_be
3262
        elif (field.startswith(BEPREFIX) and
3263
              field[len(BEPREFIX):] in constants.BES_PARAMETERS):
3264
          val = i_be.get(field[len(BEPREFIX):], None)
3265
        elif st_match and st_match.groups():
3266
          # matches a variable list
3267
          st_groups = st_match.groups()
3268
          if st_groups and st_groups[0] == "disk":
3269
            if st_groups[1] == "count":
3270
              val = len(instance.disks)
3271
            elif st_groups[1] == "sizes":
3272
              val = [disk.size for disk in instance.disks]
3273
            elif st_groups[1] == "size":
3274
              try:
3275
                val = instance.FindDisk(st_groups[2]).size
3276
              except errors.OpPrereqError:
3277
                val = None
3278
            else:
3279
              assert False, "Unhandled disk parameter"
3280
          elif st_groups[0] == "nic":
3281
            if st_groups[1] == "count":
3282
              val = len(instance.nics)
3283
            elif st_groups[1] == "macs":
3284
              val = [nic.mac for nic in instance.nics]
3285
            elif st_groups[1] == "ips":
3286
              val = [nic.ip for nic in instance.nics]
3287
            elif st_groups[1] == "bridges":
3288
              val = [nic.bridge for nic in instance.nics]
3289
            else:
3290
              # index-based item
3291
              nic_idx = int(st_groups[2])
3292
              if nic_idx >= len(instance.nics):
3293
                val = None
3294
              else:
3295
                if st_groups[1] == "mac":
3296
                  val = instance.nics[nic_idx].mac
3297
                elif st_groups[1] == "ip":
3298
                  val = instance.nics[nic_idx].ip
3299
                elif st_groups[1] == "bridge":
3300
                  val = instance.nics[nic_idx].bridge
3301
                else:
3302
                  assert False, "Unhandled NIC parameter"
3303
          else:
3304
            assert False, "Unhandled variable parameter"
3305
        else:
3306
          raise errors.ParameterError(field)
3307
        iout.append(val)
3308
      output.append(iout)
3309

    
3310
    return output
3311

    
3312

    
3313
class LUFailoverInstance(LogicalUnit):
3314
  """Failover an instance.
3315

3316
  """
3317
  HPATH = "instance-failover"
3318
  HTYPE = constants.HTYPE_INSTANCE
3319
  _OP_REQP = ["instance_name", "ignore_consistency"]
3320
  REQ_BGL = False
3321

    
3322
  def ExpandNames(self):
3323
    self._ExpandAndLockInstance()
3324
    self.needed_locks[locking.LEVEL_NODE] = []
3325
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3326

    
3327
  def DeclareLocks(self, level):
3328
    if level == locking.LEVEL_NODE:
3329
      self._LockInstancesNodes()
3330

    
3331
  def BuildHooksEnv(self):
3332
    """Build hooks env.
3333

3334
    This runs on master, primary and secondary nodes of the instance.
3335

3336
    """
3337
    env = {
3338
      "IGNORE_CONSISTENCY": self.op.ignore_consistency,
3339
      }
3340
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
3341
    nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
3342
    return env, nl, nl
3343

    
3344
  def CheckPrereq(self):
3345
    """Check prerequisites.
3346

3347
    This checks that the instance is in the cluster.
3348

3349
    """
3350
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3351
    assert self.instance is not None, \
3352
      "Cannot retrieve locked instance %s" % self.op.instance_name
3353

    
3354
    bep = self.cfg.GetClusterInfo().FillBE(instance)
3355
    if instance.disk_template not in constants.DTS_NET_MIRROR:
3356
      raise errors.OpPrereqError("Instance's disk layout is not"
3357
                                 " network mirrored, cannot failover.")
3358

    
3359
    secondary_nodes = instance.secondary_nodes
3360
    if not secondary_nodes:
3361
      raise errors.ProgrammerError("no secondary node but using "
3362
                                   "a mirrored disk template")
3363

    
3364
    target_node = secondary_nodes[0]
3365
    _CheckNodeOnline(self, target_node)
3366
    # check memory requirements on the secondary node
3367
    _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
3368
                         instance.name, bep[constants.BE_MEMORY],
3369
                         instance.hypervisor)
3370

    
3371
    # check bridge existance
3372
    brlist = [nic.bridge for nic in instance.nics]
3373
    result = self.rpc.call_bridges_exist(target_node, brlist)
3374
    result.Raise()
3375
    if not result.data:
3376
      raise errors.OpPrereqError("One or more target bridges %s does not"
3377
                                 " exist on destination node '%s'" %
3378
                                 (brlist, target_node))
3379

    
3380
  def Exec(self, feedback_fn):
3381
    """Failover an instance.
3382

3383
    The failover is done by shutting it down on its present node and
3384
    starting it on the secondary.
3385

3386
    """
3387
    instance = self.instance
3388

    
3389
    source_node = instance.primary_node
3390
    target_node = instance.secondary_nodes[0]
3391

    
3392
    feedback_fn("* checking disk consistency between source and target")
3393
    for dev in instance.disks:
3394
      # for drbd, these are drbd over lvm
3395
      if not _CheckDiskConsistency(self, dev, target_node, False):
3396
        if instance.admin_up and not self.op.ignore_consistency:
3397
          raise errors.OpExecError("Disk %s is degraded on target node,"
3398
                                   " aborting failover." % dev.iv_name)
3399

    
3400
    feedback_fn("* shutting down instance on source node")
3401
    logging.info("Shutting down instance %s on node %s",
3402
                 instance.name, source_node)
3403

    
3404
    result = self.rpc.call_instance_shutdown(source_node, instance)
3405
    if result.failed or not result.data:
3406
      if self.op.ignore_consistency:
3407
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
3408
                             " Proceeding"
3409
                             " anyway. Please make sure node %s is down",
3410
                             instance.name, source_node, source_node)
3411
      else:
3412
        raise errors.OpExecError("Could not shutdown instance %s on node %s" %
3413
                                 (instance.name, source_node))
3414

    
3415
    feedback_fn("* deactivating the instance's disks on source node")
3416
    if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
3417
      raise errors.OpExecError("Can't shut down the instance's disks.")
3418

    
3419
    instance.primary_node = target_node
3420
    # distribute new instance config to the other nodes
3421
    self.cfg.Update(instance)
3422

    
3423
    # Only start the instance if it's marked as up
3424
    if instance.admin_up:
3425
      feedback_fn("* activating the instance's disks on target node")
3426
      logging.info("Starting instance %s on node %s",
3427
                   instance.name, target_node)
3428

    
3429
      disks_ok, dummy = _AssembleInstanceDisks(self, instance,
3430
                                               ignore_secondaries=True)
3431
      if not disks_ok:
3432
        _ShutdownInstanceDisks(self, instance)
3433
        raise errors.OpExecError("Can't activate the instance's disks")
3434

    
3435
      feedback_fn("* starting the instance on the target node")
3436
      result = self.rpc.call_instance_start(target_node, instance, None)
3437
      msg = result.RemoteFailMsg()
3438
      if msg:
3439
        _ShutdownInstanceDisks(self, instance)
3440
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
3441
                                 (instance.name, target_node, msg))
3442

    
3443

    
3444
class LUMigrateInstance(LogicalUnit):
3445
  """Migrate an instance.
3446

3447
  This is migration without shutting down, compared to the failover,
3448
  which is done with shutdown.
3449

3450
  """
3451
  HPATH = "instance-migrate"
3452
  HTYPE = constants.HTYPE_INSTANCE
3453
  _OP_REQP = ["instance_name", "live", "cleanup"]
3454

    
3455
  REQ_BGL = False
3456

    
3457
  def ExpandNames(self):
3458
    self._ExpandAndLockInstance()
3459
    self.needed_locks[locking.LEVEL_NODE] = []
3460
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3461

    
3462
  def DeclareLocks(self, level):
3463
    if level == locking.LEVEL_NODE:
3464
      self._LockInstancesNodes()
3465

    
3466
  def BuildHooksEnv(self):
3467
    """Build hooks env.
3468

3469
    This runs on master, primary and secondary nodes of the instance.
3470

3471
    """
3472
    env = _BuildInstanceHookEnvByObject(self, self.instance)
3473
    nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
3474
    return env, nl, nl
3475

    
3476
  def CheckPrereq(self):
3477
    """Check prerequisites.
3478

3479
    This checks that the instance is in the cluster.
3480

3481
    """
3482
    instance = self.cfg.GetInstanceInfo(
3483
      self.cfg.ExpandInstanceName(self.op.instance_name))
3484
    if instance is None:
3485
      raise errors.OpPrereqError("Instance '%s' not known" %
3486
                                 self.op.instance_name)
3487

    
3488
    if instance.disk_template != constants.DT_DRBD8:
3489
      raise errors.OpPrereqError("Instance's disk layout is not"
3490
                                 " drbd8, cannot migrate.")
3491

    
3492
    secondary_nodes = instance.secondary_nodes
3493
    if not secondary_nodes:
3494
      raise errors.ProgrammerError("no secondary node but using "
3495
                                   "drbd8 disk template")
3496

    
3497
    i_be = self.cfg.GetClusterInfo().FillBE(instance)
3498

    
3499
    target_node = secondary_nodes[0]
3500
    # check memory requirements on the secondary node
3501
    _CheckNodeFreeMemory(self, target_node, "migrating instance %s" %
3502
                         instance.name, i_be[constants.BE_MEMORY],
3503
                         instance.hypervisor)
3504

    
3505
    # check bridge existance
3506
    brlist = [nic.bridge for nic in instance.nics]
3507
    result = self.rpc.call_bridges_exist(target_node, brlist)
3508
    if result.failed or not result.data:
3509
      raise errors.OpPrereqError("One or more target bridges %s does not"
3510
                                 " exist on destination node '%s'" %
3511
                                 (brlist, target_node))
3512

    
3513
    if not self.op.cleanup:
3514
      result = self.rpc.call_instance_migratable(instance.primary_node,
3515
                                                 instance)
3516
      msg = result.RemoteFailMsg()
3517
      if msg:
3518
        raise errors.OpPrereqError("Can't migrate: %s - please use failover" %
3519
                                   msg)
3520

    
3521
    self.instance = instance
3522

    
3523
  def _WaitUntilSync(self):
3524
    """Poll with custom rpc for disk sync.
3525

3526
    This uses our own step-based rpc call.
3527

3528
    """
3529
    self.feedback_fn("* wait until resync is done")
3530
    all_done = False
3531
    while not all_done:
3532
      all_done = True
3533
      result = self.rpc.call_drbd_wait_sync(self.all_nodes,
3534
                                            self.nodes_ip,
3535
                                            self.instance.disks)
3536
      min_percent = 100
3537
      for node, nres in result.items():
3538
        msg = nres.RemoteFailMsg()
3539
        if msg:
3540
          raise errors.OpExecError("Cannot resync disks on node %s: %s" %
3541
                                   (node, msg))
3542
        node_done, node_percent = nres.payload
3543
        all_done = all_done and node_done
3544
        if node_percent is not None:
3545
          min_percent = min(min_percent, node_percent)
3546
      if not all_done:
3547
        if min_percent < 100:
3548
          self.feedback_fn("   - progress: %.1f%%" % min_percent)
3549
        time.sleep(2)
3550

    
3551
  def _EnsureSecondary(self, node):
3552
    """Demote a node to secondary.
3553

3554
    """
3555
    self.feedback_fn("* switching node %s to secondary mode" % node)
3556

    
3557
    for dev in self.instance.disks:
3558
      self.cfg.SetDiskID(dev, node)
3559

    
3560
    result = self.rpc.call_blockdev_close(node, self.instance.name,
3561
                                          self.instance.disks)
3562
    msg = result.RemoteFailMsg()
3563
    if msg:
3564
      raise errors.OpExecError("Cannot change disk to secondary on node %s,"
3565
                               " error %s" % (node, msg))
3566

    
3567
  def _GoStandalone(self):
3568
    """Disconnect from the network.
3569

3570
    """
3571
    self.feedback_fn("* changing into standalone mode")
3572
    result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
3573
                                               self.instance.disks)
3574
    for node, nres in result.items():
3575
      msg = nres.RemoteFailMsg()
3576
      if msg:
3577
        raise errors.OpExecError("Cannot disconnect disks node %s,"
3578
                                 " error %s" % (node, msg))
3579

    
3580
  def _GoReconnect(self, multimaster):
3581
    """Reconnect to the network.
3582

3583
    """
3584
    if multimaster:
3585
      msg = "dual-master"
3586
    else:
3587
      msg = "single-master"
3588
    self.feedback_fn("* changing disks into %s mode" % msg)
3589
    result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
3590
                                           self.instance.disks,
3591
                                           self.instance.name, multimaster)
3592
    for node, nres in result.items():
3593
      msg = nres.RemoteFailMsg()
3594
      if msg:
3595
        raise errors.OpExecError("Cannot change disks config on node %s,"
3596
                                 " error: %s" % (node, msg))
3597

    
3598
  def _ExecCleanup(self):
3599
    """Try to cleanup after a failed migration.
3600

3601
    The cleanup is done by:
3602
      - check that the instance is running only on one node
3603
        (and update the config if needed)
3604
      - change disks on its secondary node to secondary
3605
      - wait until disks are fully synchronized
3606
      - disconnect from the network
3607
      - change disks into single-master mode
3608
      - wait again until disks are fully synchronized
3609

3610
    """
3611
    instance = self.instance
3612
    target_node = self.target_node
3613
    source_node = self.source_node
3614

    
3615
    # check running on only one node
3616
    self.feedback_fn("* checking where the instance actually runs"
3617
                     " (if this hangs, the hypervisor might be in"
3618
                     " a bad state)")
3619
    ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
3620
    for node, result in ins_l.items():
3621
      result.Raise()
3622
      if not isinstance(result.data, list):
3623
        raise errors.OpExecError("Can't contact node '%s'" % node)
3624

    
3625
    runningon_source = instance.name in ins_l[source_node].data
3626
    runningon_target = instance.name in ins_l[target_node].data
3627

    
3628
    if runningon_source and runningon_target:
3629
      raise errors.OpExecError("Instance seems to be running on two nodes,"
3630
                               " or the hypervisor is confused. You will have"
3631
                               " to ensure manually that it runs only on one"
3632
                               " and restart this operation.")
3633

    
3634
    if not (runningon_source or runningon_target):
3635
      raise errors.OpExecError("Instance does not seem to be running at all."
3636
                               " In this case, it's safer to repair by"
3637
                               " running 'gnt-instance stop' to ensure disk"
3638
                               " shutdown, and then restarting it.")
3639

    
3640
    if runningon_target:
3641
      # the migration has actually succeeded, we need to update the config
3642
      self.feedback_fn("* instance running on secondary node (%s),"
3643
                       " updating config" % target_node)
3644
      instance.primary_node = target_node
3645
      self.cfg.Update(instance)
3646
      demoted_node = source_node
3647
    else:
3648
      self.feedback_fn("* instance confirmed to be running on its"
3649
                       " primary node (%s)" % source_node)
3650
      demoted_node = target_node
3651

    
3652
    self._EnsureSecondary(demoted_node)
3653
    try:
3654
      self._WaitUntilSync()
3655
    except errors.OpExecError:
3656
      # we ignore here errors, since if the device is standalone, it
3657
      # won't be able to sync
3658
      pass
3659
    self._GoStandalone()
3660
    self._GoReconnect(False)
3661
    self._WaitUntilSync()
3662

    
3663
    self.feedback_fn("* done")
3664

    
3665
  def _RevertDiskStatus(self):
3666
    """Try to revert the disk status after a failed migration.
3667

3668
    """
3669
    target_node = self.target_node
3670
    try:
3671
      self._EnsureSecondary(target_node)
3672
      self._GoStandalone()
3673
      self._GoReconnect(False)
3674
      self._WaitUntilSync()
3675
    except errors.OpExecError, err:
3676
      self.LogWarning("Migration failed and I can't reconnect the"
3677
                      " drives: error '%s'\n"
3678
                      "Please look and recover the instance status" %
3679
                      str(err))
3680

    
3681
  def _AbortMigration(self):
3682
    """Call the hypervisor code to abort a started migration.
3683

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

    
3689
    abort_result = self.rpc.call_finalize_migration(target_node,
3690
                                                    instance,
3691
                                                    migration_info,
3692
                                                    False)
3693
    abort_msg = abort_result.RemoteFailMsg()
3694
    if abort_msg:
3695
      logging.error("Aborting migration failed on target node %s: %s" %
3696
                    (target_node, abort_msg))
3697
      # Don't raise an exception here, as we stil have to try to revert the
3698
      # disk status, even if this step failed.
3699

    
3700
  def _ExecMigration(self):
3701
    """Migrate an instance.
3702

3703
    The migrate is done by:
3704
      - change the disks into dual-master mode
3705
      - wait until disks are fully synchronized again
3706
      - migrate the instance
3707
      - change disks on the new secondary node (the old primary) to secondary
3708
      - wait until disks are fully synchronized
3709
      - change disks into single-master mode
3710

3711
    """
3712
    instance = self.instance
3713
    target_node = self.target_node
3714
    source_node = self.source_node
3715

    
3716
    self.feedback_fn("* checking disk consistency between source and target")
3717
    for dev in instance.disks:
3718
      if not _CheckDiskConsistency(self, dev, target_node, False):
3719
        raise errors.OpExecError("Disk %s is degraded or not fully"
3720
                                 " synchronized on target node,"
3721
                                 " aborting migrate." % dev.iv_name)
3722

    
3723
    # First get the migration information from the remote node
3724
    result = self.rpc.call_migration_info(source_node, instance)
3725
    msg = result.RemoteFailMsg()
3726
    if msg:
3727
      log_err = ("Failed fetching source migration information from %s: %s" %
3728
                 (source_node, msg))
3729
      logging.error(log_err)
3730
      raise errors.OpExecError(log_err)
3731

    
3732
    self.migration_info = migration_info = result.payload
3733

    
3734
    # Then switch the disks to master/master mode
3735
    self._EnsureSecondary(target_node)
3736
    self._GoStandalone()
3737
    self._GoReconnect(True)
3738
    self._WaitUntilSync()
3739

    
3740
    self.feedback_fn("* preparing %s to accept the instance" % target_node)
3741
    result = self.rpc.call_accept_instance(target_node,
3742
                                           instance,
3743
                                           migration_info,
3744
                                           self.nodes_ip[target_node])
3745

    
3746
    msg = result.RemoteFailMsg()
3747
    if msg:
3748
      logging.error("Instance pre-migration failed, trying to revert"
3749
                    " disk status: %s", msg)
3750
      self._AbortMigration()
3751
      self._RevertDiskStatus()
3752
      raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
3753
                               (instance.name, msg))
3754

    
3755
    self.feedback_fn("* migrating instance to %s" % target_node)
3756
    time.sleep(10)
3757
    result = self.rpc.call_instance_migrate(source_node, instance,
3758
                                            self.nodes_ip[target_node],
3759
                                            self.op.live)
3760
    msg = result.RemoteFailMsg()
3761
    if msg:
3762
      logging.error("Instance migration failed, trying to revert"
3763
                    " disk status: %s", msg)
3764
      self._AbortMigration()
3765
      self._RevertDiskStatus()
3766
      raise errors.OpExecError("Could not migrate instance %s: %s" %
3767
                               (instance.name, msg))
3768
    time.sleep(10)
3769

    
3770
    instance.primary_node = target_node
3771
    # distribute new instance config to the other nodes
3772
    self.cfg.Update(instance)
3773

    
3774
    result = self.rpc.call_finalize_migration(target_node,
3775
                                              instance,
3776
                                              migration_info,
3777
                                              True)
3778
    msg = result.RemoteFailMsg()
3779
    if msg:
3780
      logging.error("Instance migration succeeded, but finalization failed:"
3781
                    " %s" % msg)
3782
      raise errors.OpExecError("Could not finalize instance migration: %s" %
3783
                               msg)
3784

    
3785
    self._EnsureSecondary(source_node)
3786
    self._WaitUntilSync()
3787
    self._GoStandalone()
3788
    self._GoReconnect(False)
3789
    self._WaitUntilSync()
3790

    
3791
    self.feedback_fn("* done")
3792

    
3793
  def Exec(self, feedback_fn):
3794
    """Perform the migration.
3795

3796
    """
3797
    self.feedback_fn = feedback_fn
3798

    
3799
    self.source_node = self.instance.primary_node
3800
    self.target_node = self.instance.secondary_nodes[0]
3801
    self.all_nodes = [self.source_node, self.target_node]
3802
    self.nodes_ip = {
3803
      self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
3804
      self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
3805
      }
3806
    if self.op.cleanup:
3807
      return self._ExecCleanup()
3808
    else:
3809
      return self._ExecMigration()
3810

    
3811

    
3812
def _CreateBlockDev(lu, node, instance, device, force_create,
3813
                    info, force_open):
3814
  """Create a tree of block devices on a given node.
3815

3816
  If this device type has to be created on secondaries, create it and
3817
  all its children.
3818

3819
  If not, just recurse to children keeping the same 'force' value.
3820

3821
  @param lu: the lu on whose behalf we execute
3822
  @param node: the node on which to create the device
3823
  @type instance: L{objects.Instance}
3824
  @param instance: the instance which owns the device
3825
  @type device: L{objects.Disk}
3826
  @param device: the device to create
3827
  @type force_create: boolean
3828
  @param force_create: whether to force creation of this device; this
3829
      will be change to True whenever we find a device which has
3830
      CreateOnSecondary() attribute
3831
  @param info: the extra 'metadata' we should attach to the device
3832
      (this will be represented as a LVM tag)
3833
  @type force_open: boolean
3834
  @param force_open: this parameter will be passes to the
3835
      L{backend.CreateBlockDevice} function where it specifies
3836
      whether we run on primary or not, and it affects both
3837
      the child assembly and the device own Open() execution
3838

3839
  """
3840
  if device.CreateOnSecondary():
3841
    force_create = True
3842

    
3843
  if device.children:
3844
    for child in device.children:
3845
      _CreateBlockDev(lu, node, instance, child, force_create,
3846
                      info, force_open)
3847

    
3848
  if not force_create:
3849
    return
3850

    
3851
  _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
3852

    
3853

    
3854
def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
3855
  """Create a single block device on a given node.
3856

3857
  This will not recurse over children of the device, so they must be
3858
  created in advance.
3859

3860
  @param lu: the lu on whose behalf we execute
3861
  @param node: the node on which to create the device
3862
  @type instance: L{objects.Instance}
3863
  @param instance: the instance which owns the device
3864
  @type device: L{objects.Disk}
3865
  @param device: the device to create
3866
  @param info: the extra 'metadata' we should attach to the device
3867
      (this will be represented as a LVM tag)
3868
  @type force_open: boolean
3869
  @param force_open: this parameter will be passes to the
3870
      L{backend.CreateBlockDevice} function where it specifies
3871
      whether we run on primary or not, and it affects both
3872
      the child assembly and the device own Open() execution
3873

3874
  """
3875
  lu.cfg.SetDiskID(device, node)
3876
  result = lu.rpc.call_blockdev_create(node, device, device.size,
3877
                                       instance.name, force_open, info)
3878
  msg = result.RemoteFailMsg()
3879
  if msg:
3880
    raise errors.OpExecError("Can't create block device %s on"
3881
                             " node %s for instance %s: %s" %
3882
                             (device, node, instance.name, msg))
3883
  if device.physical_id is None:
3884
    device.physical_id = result.payload
3885

    
3886

    
3887
def _GenerateUniqueNames(lu, exts):
3888
  """Generate a suitable LV name.
3889

3890
  This will generate a logical volume name for the given instance.
3891

3892
  """
3893
  results = []
3894
  for val in exts:
3895
    new_id = lu.cfg.GenerateUniqueID()
3896
    results.append("%s%s" % (new_id, val))
3897
  return results
3898

    
3899

    
3900
def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
3901
                         p_minor, s_minor):
3902
  """Generate a drbd8 device complete with its children.
3903

3904
  """
3905
  port = lu.cfg.AllocatePort()
3906
  vgname = lu.cfg.GetVGName()
3907
  shared_secret = lu.cfg.GenerateDRBDSecret()
3908
  dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
3909
                          logical_id=(vgname, names[0]))
3910
  dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
3911
                          logical_id=(vgname, names[1]))
3912
  drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
3913
                          logical_id=(primary, secondary, port,
3914
                                      p_minor, s_minor,
3915
                                      shared_secret),
3916
                          children=[dev_data, dev_meta],
3917
                          iv_name=iv_name)
3918
  return drbd_dev
3919

    
3920

    
3921
def _GenerateDiskTemplate(lu, template_name,
3922
                          instance_name, primary_node,
3923
                          secondary_nodes, disk_info,
3924
                          file_storage_dir, file_driver,
3925
                          base_index):
3926
  """Generate the entire disk layout for a given template type.
3927

3928
  """
3929
  #TODO: compute space requirements
3930

    
3931
  vgname = lu.cfg.GetVGName()
3932
  disk_count = len(disk_info)
3933
  disks = []
3934
  if template_name == constants.DT_DISKLESS:
3935
    pass
3936
  elif template_name == constants.DT_PLAIN:
3937
    if len(secondary_nodes) != 0:
3938
      raise errors.ProgrammerError("Wrong template configuration")
3939

    
3940
    names = _GenerateUniqueNames(lu, [".disk%d" % i
3941
                                      for i in range(disk_count)])
3942
    for idx, disk in enumerate(disk_info):
3943
      disk_index = idx + base_index
3944
      disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
3945
                              logical_id=(vgname, names[idx]),
3946
                              iv_name="disk/%d" % disk_index,
3947
                              mode=disk["mode"])
3948
      disks.append(disk_dev)
3949
  elif template_name == constants.DT_DRBD8:
3950
    if len(secondary_nodes) != 1:
3951
      raise errors.ProgrammerError("Wrong template configuration")
3952
    remote_node = secondary_nodes[0]
3953
    minors = lu.cfg.AllocateDRBDMinor(
3954
      [primary_node, remote_node] * len(disk_info), instance_name)
3955

    
3956
    names = []
3957
    for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % i
3958
                                               for i in range(disk_count)]):
3959
      names.append(lv_prefix + "_data")
3960
      names.append(lv_prefix + "_meta")
3961
    for idx, disk in enumerate(disk_info):
3962
      disk_index = idx + base_index
3963
      disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
3964
                                      disk["size"], names[idx*2:idx*2+2],
3965
                                      "disk/%d" % disk_index,
3966
                                      minors[idx*2], minors[idx*2+1])
3967
      disk_dev.mode = disk["mode"]
3968
      disks.append(disk_dev)
3969
  elif template_name == constants.DT_FILE:
3970
    if len(secondary_nodes) != 0:
3971
      raise errors.ProgrammerError("Wrong template configuration")
3972

    
3973
    for idx, disk in enumerate(disk_info):
3974
      disk_index = idx + base_index
3975
      disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
3976
                              iv_name="disk/%d" % disk_index,
3977
                              logical_id=(file_driver,
3978
                                          "%s/disk%d" % (file_storage_dir,
3979
                                                         idx)),
3980
                              mode=disk["mode"])
3981
      disks.append(disk_dev)
3982
  else:
3983
    raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
3984
  return disks
3985

    
3986

    
3987
def _GetInstanceInfoText(instance):
3988
  """Compute that text that should be added to the disk's metadata.
3989

3990
  """
3991
  return "originstname+%s" % instance.name
3992

    
3993

    
3994
def _CreateDisks(lu, instance):
3995
  """Create all disks for an instance.
3996

3997
  This abstracts away some work from AddInstance.
3998

3999
  @type lu: L{LogicalUnit}
4000
  @param lu: the logical unit on whose behalf we execute
4001
  @type instance: L{objects.Instance}
4002
  @param instance: the instance whose disks we should create
4003
  @rtype: boolean
4004
  @return: the success of the creation
4005

4006
  """
4007
  info = _GetInstanceInfoText(instance)
4008
  pnode = instance.primary_node
4009

    
4010
  if instance.disk_template == constants.DT_FILE:
4011
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
4012
    result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
4013

    
4014
    if result.failed or not result.data:
4015
      raise errors.OpExecError("Could not connect to node '%s'" % pnode)
4016

    
4017
    if not result.data[0]:
4018
      raise errors.OpExecError("Failed to create directory '%s'" %
4019
                               file_storage_dir)
4020

    
4021
  # Note: this needs to be kept in sync with adding of disks in
4022
  # LUSetInstanceParams
4023
  for device in instance.disks:
4024
    logging.info("Creating volume %s for instance %s",
4025
                 device.iv_name, instance.name)
4026
    #HARDCODE
4027
    for node in instance.all_nodes:
4028
      f_create = node == pnode
4029
      _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
4030

    
4031

    
4032
def _RemoveDisks(lu, instance):
4033
  """Remove all disks for an instance.
4034

4035
  This abstracts away some work from `AddInstance()` and
4036
  `RemoveInstance()`. Note that in case some of the devices couldn't
4037
  be removed, the removal will continue with the other ones (compare
4038
  with `_CreateDisks()`).
4039

4040
  @type lu: L{LogicalUnit}
4041
  @param lu: the logical unit on whose behalf we execute
4042
  @type instance: L{objects.Instance}
4043
  @param instance: the instance whose disks we should remove
4044
  @rtype: boolean
4045
  @return: the success of the removal
4046

4047
  """
4048
  logging.info("Removing block devices for instance %s", instance.name)
4049

    
4050
  result = True
4051
  for device in instance.disks:
4052
    for node, disk in device.ComputeNodeTree(instance.primary_node):
4053
      lu.cfg.SetDiskID(disk, node)
4054
      result = lu.rpc.call_blockdev_remove(node, disk)
4055
      if result.failed or not result.data:
4056
        lu.proc.LogWarning("Could not remove block device %s on node %s,"
4057
                           " continuing anyway", device.iv_name, node)
4058
        result = False
4059

    
4060
  if instance.disk_template == constants.DT_FILE:
4061
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
4062
    result = lu.rpc.call_file_storage_dir_remove(instance.primary_node,
4063
                                                 file_storage_dir)
4064
    if result.failed or not result.data:
4065
      logging.error("Could not remove directory '%s'", file_storage_dir)
4066
      result = False
4067

    
4068
  return result
4069

    
4070

    
4071
def _ComputeDiskSize(disk_template, disks):
4072
  """Compute disk size requirements in the volume group
4073

4074
  """
4075
  # Required free disk space as a function of disk and swap space
4076
  req_size_dict = {
4077
    constants.DT_DISKLESS: None,
4078
    constants.DT_PLAIN: sum(d["size"] for d in disks),
4079
    # 128 MB are added for drbd metadata for each disk
4080
    constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
4081
    constants.DT_FILE: None,
4082
  }
4083

    
4084
  if disk_template not in req_size_dict:
4085
    raise errors.ProgrammerError("Disk template '%s' size requirement"
4086
                                 " is unknown" %  disk_template)
4087

    
4088
  return req_size_dict[disk_template]
4089

    
4090

    
4091
def _CheckHVParams(lu, nodenames, hvname, hvparams):
4092
  """Hypervisor parameter validation.
4093

4094
  This function abstract the hypervisor parameter validation to be
4095
  used in both instance create and instance modify.
4096

4097
  @type lu: L{LogicalUnit}
4098
  @param lu: the logical unit for which we check
4099
  @type nodenames: list
4100
  @param nodenames: the list of nodes on which we should check
4101
  @type hvname: string
4102
  @param hvname: the name of the hypervisor we should use
4103
  @type hvparams: dict
4104
  @param hvparams: the parameters which we need to check
4105
  @raise errors.OpPrereqError: if the parameters are not valid
4106

4107
  """
4108
  hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
4109
                                                  hvname,
4110
                                                  hvparams)
4111
  for node in nodenames:
4112
    info = hvinfo[node]
4113
    if info.offline:
4114
      continue
4115
    msg = info.RemoteFailMsg()
4116
    if msg:
4117
      raise errors.OpPrereqError("Hypervisor parameter validation failed:"
4118
                                 " %s" % msg)
4119

    
4120

    
4121
class LUCreateInstance(LogicalUnit):
4122
  """Create an instance.
4123

4124
  """
4125
  HPATH = "instance-add"
4126
  HTYPE = constants.HTYPE_INSTANCE
4127
  _OP_REQP = ["instance_name", "disks", "disk_template",
4128
              "mode", "start",
4129
              "wait_for_sync", "ip_check", "nics",
4130
              "hvparams", "beparams"]
4131
  REQ_BGL = False
4132

    
4133
  def _ExpandNode(self, node):
4134
    """Expands and checks one node name.
4135

4136
    """
4137
    node_full = self.cfg.ExpandNodeName(node)
4138
    if node_full is None:
4139
      raise errors.OpPrereqError("Unknown node %s" % node)
4140
    return node_full
4141

    
4142
  def ExpandNames(self):
4143
    """ExpandNames for CreateInstance.
4144

4145
    Figure out the right locks for instance creation.
4146

4147
    """
4148
    self.needed_locks = {}
4149

    
4150
    # set optional parameters to none if they don't exist
4151
    for attr in ["pnode", "snode", "iallocator", "hypervisor"]:
4152
      if not hasattr(self.op, attr):
4153
        setattr(self.op, attr, None)
4154

    
4155
    # cheap checks, mostly valid constants given
4156

    
4157
    # verify creation mode
4158
    if self.op.mode not in (constants.INSTANCE_CREATE,
4159
                            constants.INSTANCE_IMPORT):
4160
      raise errors.OpPrereqError("Invalid instance creation mode '%s'" %
4161
                                 self.op.mode)
4162

    
4163
    # disk template and mirror node verification
4164
    if self.op.disk_template not in constants.DISK_TEMPLATES:
4165
      raise errors.OpPrereqError("Invalid disk template name")
4166

    
4167
    if self.op.hypervisor is None:
4168
      self.op.hypervisor = self.cfg.GetHypervisorType()
4169

    
4170
    cluster = self.cfg.GetClusterInfo()
4171
    enabled_hvs = cluster.enabled_hypervisors
4172
    if self.op.hypervisor not in enabled_hvs:
4173
      raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
4174
                                 " cluster (%s)" % (self.op.hypervisor,
4175
                                  ",".join(enabled_hvs)))
4176

    
4177
    # check hypervisor parameter syntax (locally)
4178

    
4179
    filled_hvp = cluster.FillDict(cluster.hvparams[self.op.hypervisor],
4180
                                  self.op.hvparams)
4181
    hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
4182
    hv_type.CheckParameterSyntax(filled_hvp)
4183

    
4184
    # fill and remember the beparams dict
4185
    utils.CheckBEParams(self.op.beparams)
4186
    self.be_full = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
4187
                                    self.op.beparams)
4188

    
4189
    #### instance parameters check
4190

    
4191
    # instance name verification
4192
    hostname1 = utils.HostInfo(self.op.instance_name)
4193
    self.op.instance_name = instance_name = hostname1.name
4194

    
4195
    # this is just a preventive check, but someone might still add this
4196
    # instance in the meantime, and creation will fail at lock-add time
4197
    if instance_name in self.cfg.GetInstanceList():
4198
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
4199
                                 instance_name)
4200

    
4201
    self.add_locks[locking.LEVEL_INSTANCE] = instance_name
4202

    
4203
    # NIC buildup
4204
    self.nics = []
4205
    for nic in self.op.nics:
4206
      # ip validity checks
4207
      ip = nic.get("ip", None)
4208
      if ip is None or ip.lower() == "none":
4209
        nic_ip = None
4210
      elif ip.lower() == constants.VALUE_AUTO:
4211
        nic_ip = hostname1.ip
4212
      else:
4213
        if not utils.IsValidIP(ip):
4214
          raise errors.OpPrereqError("Given IP address '%s' doesn't look"
4215
                                     " like a valid IP" % ip)
4216
        nic_ip = ip
4217

    
4218
      # MAC address verification
4219
      mac = nic.get("mac", constants.VALUE_AUTO)
4220
      if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
4221
        if not utils.IsValidMac(mac.lower()):
4222
          raise errors.OpPrereqError("Invalid MAC address specified: %s" %
4223
                                     mac)
4224
      # bridge verification
4225
      bridge = nic.get("bridge", None)
4226
      if bridge is None:
4227
        bridge = self.cfg.GetDefBridge()
4228
      self.nics.append(objects.NIC(mac=mac, ip=nic_ip, bridge=bridge))
4229

    
4230
    # disk checks/pre-build
4231
    self.disks = []
4232
    for disk in self.op.disks:
4233
      mode = disk.get("mode", constants.DISK_RDWR)
4234
      if mode not in constants.DISK_ACCESS_SET:
4235
        raise errors.OpPrereqError("Invalid disk access mode '%s'" %
4236
                                   mode)
4237
      size = disk.get("size", None)
4238
      if size is None:
4239
        raise errors.OpPrereqError("Missing disk size")
4240
      try:
4241
        size = int(size)
4242
      except ValueError:
4243
        raise errors.OpPrereqError("Invalid disk size '%s'" % size)
4244
      self.disks.append({"size": size, "mode": mode})
4245

    
4246
    # used in CheckPrereq for ip ping check
4247
    self.check_ip = hostname1.ip
4248

    
4249
    # file storage checks
4250
    if (self.op.file_driver and
4251
        not self.op.file_driver in constants.FILE_DRIVER):
4252
      raise errors.OpPrereqError("Invalid file driver name '%s'" %
4253
                                 self.op.file_driver)
4254

    
4255
    if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
4256
      raise errors.OpPrereqError("File storage directory path not absolute")
4257

    
4258
    ### Node/iallocator related checks
4259
    if [self.op.iallocator, self.op.pnode].count(None) != 1:
4260
      raise errors.OpPrereqError("One and only one of iallocator and primary"
4261
                                 " node must be given")
4262

    
4263
    if self.op.iallocator:
4264
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4265
    else:
4266
      self.op.pnode = self._ExpandNode(self.op.pnode)
4267
      nodelist = [self.op.pnode]
4268
      if self.op.snode is not None:
4269
        self.op.snode = self._ExpandNode(self.op.snode)
4270
        nodelist.append(self.op.snode)
4271
      self.needed_locks[locking.LEVEL_NODE] = nodelist
4272

    
4273
    # in case of import lock the source node too
4274
    if self.op.mode == constants.INSTANCE_IMPORT:
4275
      src_node = getattr(self.op, "src_node", None)
4276
      src_path = getattr(self.op, "src_path", None)
4277

    
4278
      if src_path is None:
4279
        self.op.src_path = src_path = self.op.instance_name
4280

    
4281
      if src_node is None:
4282
        self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4283
        self.op.src_node = None
4284
        if os.path.isabs(src_path):
4285
          raise errors.OpPrereqError("Importing an instance from an absolute"
4286
                                     " path requires a source node option.")
4287
      else:
4288
        self.op.src_node = src_node = self._ExpandNode(src_node)
4289
        if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
4290
          self.needed_locks[locking.LEVEL_NODE].append(src_node)
4291
        if not os.path.isabs(src_path):
4292
          self.op.src_path = src_path = \
4293
            os.path.join(constants.EXPORT_DIR, src_path)
4294

    
4295
    else: # INSTANCE_CREATE
4296
      if getattr(self.op, "os_type", None) is None:
4297
        raise errors.OpPrereqError("No guest OS specified")
4298

    
4299
  def _RunAllocator(self):
4300
    """Run the allocator based on input opcode.
4301

4302
    """
4303
    nics = [n.ToDict() for n in self.nics]
4304
    ial = IAllocator(self,
4305
                     mode=constants.IALLOCATOR_MODE_ALLOC,
4306
                     name=self.op.instance_name,
4307
                     disk_template=self.op.disk_template,
4308
                     tags=[],
4309
                     os=self.op.os_type,
4310
                     vcpus=self.be_full[constants.BE_VCPUS],
4311
                     mem_size=self.be_full[constants.BE_MEMORY],
4312
                     disks=self.disks,
4313
                     nics=nics,
4314
                     hypervisor=self.op.hypervisor,
4315
                     )
4316

    
4317
    ial.Run(self.op.iallocator)
4318

    
4319
    if not ial.success:
4320
      raise errors.OpPrereqError("Can't compute nodes using"
4321
                                 " iallocator '%s': %s" % (self.op.iallocator,
4322
                                                           ial.info))
4323
    if len(ial.nodes) != ial.required_nodes:
4324
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
4325
                                 " of nodes (%s), required %s" %
4326
                                 (self.op.iallocator, len(ial.nodes),
4327
                                  ial.required_nodes))
4328
    self.op.pnode = ial.nodes[0]
4329
    self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
4330
                 self.op.instance_name, self.op.iallocator,
4331
                 ", ".join(ial.nodes))
4332
    if ial.required_nodes == 2:
4333
      self.op.snode = ial.nodes[1]
4334

    
4335
  def BuildHooksEnv(self):
4336
    """Build hooks env.
4337

4338
    This runs on master, primary and secondary nodes of the instance.
4339

4340
    """
4341
    env = {
4342
      "INSTANCE_DISK_TEMPLATE": self.op.disk_template,
4343
      "INSTANCE_DISK_SIZE": ",".join(str(d["size"]) for d in self.disks),
4344
      "INSTANCE_ADD_MODE": self.op.mode,
4345
      }
4346
    if self.op.mode == constants.INSTANCE_IMPORT:
4347
      env["INSTANCE_SRC_NODE"] = self.op.src_node
4348
      env["INSTANCE_SRC_PATH"] = self.op.src_path
4349
      env["INSTANCE_SRC_IMAGES"] = self.src_images
4350

    
4351
    env.update(_BuildInstanceHookEnv(name=self.op.instance_name,
4352
      primary_node=self.op.pnode,
4353
      secondary_nodes=self.secondaries,
4354
      status=self.op.start,
4355
      os_type=self.op.os_type,
4356
      memory=self.be_full[constants.BE_MEMORY],
4357
      vcpus=self.be_full[constants.BE_VCPUS],
4358
      nics=[(n.ip, n.bridge, n.mac) for n in self.nics],
4359
    ))
4360

    
4361
    nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
4362
          self.secondaries)
4363
    return env, nl, nl
4364

    
4365

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

4369
    """
4370
    if (not self.cfg.GetVGName() and
4371
        self.op.disk_template not in constants.DTS_NOT_LVM):
4372
      raise errors.OpPrereqError("Cluster does not support lvm-based"
4373
                                 " instances")
4374

    
4375

    
4376
    if self.op.mode == constants.INSTANCE_IMPORT:
4377
      src_node = self.op.src_node
4378
      src_path = self.op.src_path
4379

    
4380
      if src_node is None:
4381
        exp_list = self.rpc.call_export_list(
4382
          self.acquired_locks[locking.LEVEL_NODE])
4383
        found = False
4384
        for node in exp_list:
4385
          if not exp_list[node].failed and src_path in exp_list[node].data:
4386
            found = True
4387
            self.op.src_node = src_node = node
4388
            self.op.src_path = src_path = os.path.join(constants.EXPORT_DIR,
4389
                                                       src_path)
4390
            break
4391
        if not found:
4392
          raise errors.OpPrereqError("No export found for relative path %s" %
4393
                                      src_path)
4394

    
4395
      _CheckNodeOnline(self, src_node)
4396
      result = self.rpc.call_export_info(src_node, src_path)
4397
      result.Raise()
4398
      if not result.data:
4399
        raise errors.OpPrereqError("No export found in dir %s" % src_path)
4400

    
4401
      export_info = result.data
4402
      if not export_info.has_section(constants.INISECT_EXP):
4403
        raise errors.ProgrammerError("Corrupted export config")
4404

    
4405
      ei_version = export_info.get(constants.INISECT_EXP, 'version')
4406
      if (int(ei_version) != constants.EXPORT_VERSION):
4407
        raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
4408
                                   (ei_version, constants.EXPORT_VERSION))
4409

    
4410
      # Check that the new instance doesn't have less disks than the export
4411
      instance_disks = len(self.disks)
4412
      export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
4413
      if instance_disks < export_disks:
4414
        raise errors.OpPrereqError("Not enough disks to import."
4415
                                   " (instance: %d, export: %d)" %
4416
                                   (instance_disks, export_disks))
4417

    
4418
      self.op.os_type = export_info.get(constants.INISECT_EXP, 'os')
4419
      disk_images = []
4420
      for idx in range(export_disks):
4421
        option = 'disk%d_dump' % idx
4422
        if export_info.has_option(constants.INISECT_INS, option):
4423
          # FIXME: are the old os-es, disk sizes, etc. useful?
4424
          export_name = export_info.get(constants.INISECT_INS, option)
4425
          image = os.path.join(src_path, export_name)
4426
          disk_images.append(image)
4427
        else:
4428
          disk_images.append(False)
4429

    
4430
      self.src_images = disk_images
4431

    
4432
      old_name = export_info.get(constants.INISECT_INS, 'name')
4433
      # FIXME: int() here could throw a ValueError on broken exports
4434
      exp_nic_count = int(export_info.get(constants.INISECT_INS, 'nic_count'))
4435
      if self.op.instance_name == old_name:
4436
        for idx, nic in enumerate(self.nics):
4437
          if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
4438
            nic_mac_ini = 'nic%d_mac' % idx
4439
            nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
4440

    
4441
    # ip ping checks (we use the same ip that was resolved in ExpandNames)
4442
    if self.op.start and not self.op.ip_check:
4443
      raise errors.OpPrereqError("Cannot ignore IP address conflicts when"
4444
                                 " adding an instance in start mode")
4445

    
4446
    if self.op.ip_check:
4447
      if utils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
4448
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
4449
                                   (self.check_ip, self.op.instance_name))
4450

    
4451
    #### allocator run
4452

    
4453
    if self.op.iallocator is not None:
4454
      self._RunAllocator()
4455

    
4456
    #### node related checks
4457

    
4458
    # check primary node
4459
    self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
4460
    assert self.pnode is not None, \
4461
      "Cannot retrieve locked node %s" % self.op.pnode
4462
    if pnode.offline:
4463
      raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
4464
                                 pnode.name)
4465

    
4466
    self.secondaries = []
4467

    
4468
    # mirror node verification
4469
    if self.op.disk_template in constants.DTS_NET_MIRROR:
4470
      if self.op.snode is None:
4471
        raise errors.OpPrereqError("The networked disk templates need"
4472
                                   " a mirror node")
4473
      if self.op.snode == pnode.name:
4474
        raise errors.OpPrereqError("The secondary node cannot be"
4475
                                   " the primary node.")
4476
      self.secondaries.append(self.op.snode)
4477
      _CheckNodeOnline(self, self.op.snode)
4478

    
4479
    nodenames = [pnode.name] + self.secondaries
4480

    
4481
    req_size = _ComputeDiskSize(self.op.disk_template,
4482
                                self.disks)
4483

    
4484
    # Check lv size requirements
4485
    if req_size is not None:
4486
      nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
4487
                                         self.op.hypervisor)
4488
      for node in nodenames:
4489
        info = nodeinfo[node]
4490
        info.Raise()
4491
        info = info.data
4492
        if not info:
4493
          raise errors.OpPrereqError("Cannot get current information"
4494
                                     " from node '%s'" % node)
4495
        vg_free = info.get('vg_free', None)
4496
        if not isinstance(vg_free, int):
4497
          raise errors.OpPrereqError("Can't compute free disk space on"
4498
                                     " node %s" % node)
4499
        if req_size > info['vg_free']:
4500
          raise errors.OpPrereqError("Not enough disk space on target node %s."
4501
                                     " %d MB available, %d MB required" %
4502
                                     (node, info['vg_free'], req_size))
4503

    
4504
    _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
4505

    
4506
    # os verification
4507
    result = self.rpc.call_os_get(pnode.name, self.op.os_type)
4508
    result.Raise()
4509
    if not isinstance(result.data, objects.OS):
4510
      raise errors.OpPrereqError("OS '%s' not in supported os list for"
4511
                                 " primary node"  % self.op.os_type)
4512

    
4513
    # bridge check on primary node
4514
    bridges = [n.bridge for n in self.nics]
4515
    result = self.rpc.call_bridges_exist(self.pnode.name, bridges)
4516
    result.Raise()
4517
    if not result.data:
4518
      raise errors.OpPrereqError("One of the target bridges '%s' does not"
4519
                                 " exist on destination node '%s'" %
4520
                                 (",".join(bridges), pnode.name))
4521

    
4522
    # memory check on primary node
4523
    if self.op.start:
4524
      _CheckNodeFreeMemory(self, self.pnode.name,
4525
                           "creating instance %s" % self.op.instance_name,
4526
                           self.be_full[constants.BE_MEMORY],
4527
                           self.op.hypervisor)
4528

    
4529
  def Exec(self, feedback_fn):
4530
    """Create and add the instance to the cluster.
4531

4532
    """
4533
    instance = self.op.instance_name
4534
    pnode_name = self.pnode.name
4535

    
4536
    for nic in self.nics:
4537
      if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
4538
        nic.mac = self.cfg.GenerateMAC()
4539

    
4540
    ht_kind = self.op.hypervisor
4541
    if ht_kind in constants.HTS_REQ_PORT:
4542
      network_port = self.cfg.AllocatePort()
4543
    else:
4544
      network_port = None
4545

    
4546
    ##if self.op.vnc_bind_address is None:
4547
    ##  self.op.vnc_bind_address = constants.VNC_DEFAULT_BIND_ADDRESS
4548

    
4549
    # this is needed because os.path.join does not accept None arguments
4550
    if self.op.file_storage_dir is None:
4551
      string_file_storage_dir = ""
4552
    else:
4553
      string_file_storage_dir = self.op.file_storage_dir
4554

    
4555
    # build the full file storage dir path
4556
    file_storage_dir = os.path.normpath(os.path.join(
4557
                                        self.cfg.GetFileStorageDir(),
4558
                                        string_file_storage_dir, instance))
4559

    
4560

    
4561
    disks = _GenerateDiskTemplate(self,
4562
                                  self.op.disk_template,
4563
                                  instance, pnode_name,
4564
                                  self.secondaries,
4565
                                  self.disks,
4566
                                  file_storage_dir,
4567
                                  self.op.file_driver,
4568
                                  0)
4569

    
4570
    iobj = objects.Instance(name=instance, os=self.op.os_type,
4571
                            primary_node=pnode_name,
4572
                            nics=self.nics, disks=disks,
4573
                            disk_template=self.op.disk_template,
4574
                            admin_up=False,
4575
                            network_port=network_port,
4576
                            beparams=self.op.beparams,
4577
                            hvparams=self.op.hvparams,
4578
                            hypervisor=self.op.hypervisor,
4579
                            )
4580

    
4581
    feedback_fn("* creating instance disks...")
4582
    try:
4583
      _CreateDisks(self, iobj)
4584
    except errors.OpExecError:
4585
      self.LogWarning("Device creation failed, reverting...")
4586
      try:
4587
        _RemoveDisks(self, iobj)
4588
      finally:
4589
        self.cfg.ReleaseDRBDMinors(instance)
4590
        raise
4591

    
4592
    feedback_fn("adding instance %s to cluster config" % instance)
4593

    
4594
    self.cfg.AddInstance(iobj)
4595
    # Declare that we don't want to remove the instance lock anymore, as we've
4596
    # added the instance to the config
4597
    del self.remove_locks[locking.LEVEL_INSTANCE]
4598
    # Unlock all the nodes
4599
    if self.op.mode == constants.INSTANCE_IMPORT:
4600
      nodes_keep = [self.op.src_node]
4601
      nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
4602
                       if node != self.op.src_node]
4603
      self.context.glm.release(locking.LEVEL_NODE, nodes_release)
4604
      self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
4605
    else:
4606
      self.context.glm.release(locking.LEVEL_NODE)
4607
      del self.acquired_locks[locking.LEVEL_NODE]
4608

    
4609
    if self.op.wait_for_sync:
4610
      disk_abort = not _WaitForSync(self, iobj)
4611
    elif iobj.disk_template in constants.DTS_NET_MIRROR:
4612
      # make sure the disks are not degraded (still sync-ing is ok)
4613
      time.sleep(15)
4614
      feedback_fn("* checking mirrors status")
4615
      disk_abort = not _WaitForSync(self, iobj, oneshot=True)
4616
    else:
4617
      disk_abort = False
4618

    
4619
    if disk_abort:
4620
      _RemoveDisks(self, iobj)
4621
      self.cfg.RemoveInstance(iobj.name)
4622
      # Make sure the instance lock gets removed
4623
      self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
4624
      raise errors.OpExecError("There are some degraded disks for"
4625
                               " this instance")
4626

    
4627
    feedback_fn("creating os for instance %s on node %s" %
4628
                (instance, pnode_name))
4629

    
4630
    if iobj.disk_template != constants.DT_DISKLESS:
4631
      if self.op.mode == constants.INSTANCE_CREATE:
4632
        feedback_fn("* running the instance OS create scripts...")
4633
        result = self.rpc.call_instance_os_add(pnode_name, iobj)
4634
        msg = result.RemoteFailMsg()
4635
        if msg:
4636
          raise errors.OpExecError("Could not add os for instance %s"
4637
                                   " on node %s: %s" %
4638
                                   (instance, pnode_name, msg))
4639

    
4640
      elif self.op.mode == constants.INSTANCE_IMPORT:
4641
        feedback_fn("* running the instance OS import scripts...")
4642
        src_node = self.op.src_node
4643
        src_images = self.src_images
4644
        cluster_name = self.cfg.GetClusterName()
4645
        import_result = self.rpc.call_instance_os_import(pnode_name, iobj,
4646
                                                         src_node, src_images,
4647
                                                         cluster_name)
4648
        import_result.Raise()
4649
        for idx, result in enumerate(import_result.data):
4650
          if not result:
4651
            self.LogWarning("Could not import the image %s for instance"
4652
                            " %s, disk %d, on node %s" %
4653
                            (src_images[idx], instance, idx, pnode_name))
4654
      else:
4655
        # also checked in the prereq part
4656
        raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
4657
                                     % self.op.mode)
4658

    
4659
    if self.op.start:
4660
      iobj.admin_up = True
4661
      self.cfg.Update(iobj)
4662
      logging.info("Starting instance %s on node %s", instance, pnode_name)
4663
      feedback_fn("* starting instance...")
4664
      result = self.rpc.call_instance_start(pnode_name, iobj, None)
4665
      msg = result.RemoteFailMsg()
4666
      if msg:
4667
        raise errors.OpExecError("Could not start instance: %s" % msg)
4668

    
4669

    
4670
class LUConnectConsole(NoHooksLU):
4671
  """Connect to an instance's console.
4672

4673
  This is somewhat special in that it returns the command line that
4674
  you need to run on the master node in order to connect to the
4675
  console.
4676

4677
  """
4678
  _OP_REQP = ["instance_name"]
4679
  REQ_BGL = False
4680

    
4681
  def ExpandNames(self):
4682
    self._ExpandAndLockInstance()
4683

    
4684
  def CheckPrereq(self):
4685
    """Check prerequisites.
4686

4687
    This checks that the instance is in the cluster.
4688

4689
    """
4690
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4691
    assert self.instance is not None, \
4692
      "Cannot retrieve locked instance %s" % self.op.instance_name
4693
    _CheckNodeOnline(self, self.instance.primary_node)
4694

    
4695
  def Exec(self, feedback_fn):
4696
    """Connect to the console of an instance
4697

4698
    """
4699
    instance = self.instance
4700
    node = instance.primary_node
4701

    
4702
    node_insts = self.rpc.call_instance_list([node],
4703
                                             [instance.hypervisor])[node]
4704
    node_insts.Raise()
4705

    
4706
    if instance.name not in node_insts.data:
4707
      raise errors.OpExecError("Instance %s is not running." % instance.name)
4708

    
4709
    logging.debug("Connecting to console of %s on %s", instance.name, node)
4710

    
4711
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
4712
    cluster = self.cfg.GetClusterInfo()
4713
    # beparams and hvparams are passed separately, to avoid editing the
4714
    # instance and then saving the defaults in the instance itself.
4715
    hvparams = cluster.FillHV(instance)
4716
    beparams = cluster.FillBE(instance)
4717
    console_cmd = hyper.GetShellCommandForConsole(instance, hvparams, beparams)
4718

    
4719
    # build ssh cmdline
4720
    return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
4721

    
4722

    
4723
class LUReplaceDisks(LogicalUnit):
4724
  """Replace the disks of an instance.
4725

4726
  """
4727
  HPATH = "mirrors-replace"
4728
  HTYPE = constants.HTYPE_INSTANCE
4729
  _OP_REQP = ["instance_name", "mode", "disks"]
4730
  REQ_BGL = False
4731

    
4732
  def CheckArguments(self):
4733
    if not hasattr(self.op, "remote_node"):
4734
      self.op.remote_node = None
4735
    if not hasattr(self.op, "iallocator"):
4736
      self.op.iallocator = None
4737

    
4738
    # check for valid parameter combination
4739
    cnt = [self.op.remote_node, self.op.iallocator].count(None)
4740
    if self.op.mode == constants.REPLACE_DISK_CHG:
4741
      if cnt == 2:
4742
        raise errors.OpPrereqError("When changing the secondary either an"
4743
                                   " iallocator script must be used or the"
4744
                                   " new node given")
4745
      elif cnt == 0:
4746
        raise errors.OpPrereqError("Give either the iallocator or the new"
4747
                                   " secondary, not both")
4748
    else: # not replacing the secondary
4749
      if cnt != 2:
4750
        raise errors.OpPrereqError("The iallocator and new node options can"
4751
                                   " be used only when changing the"
4752
                                   " secondary node")
4753

    
4754
  def ExpandNames(self):
4755
    self._ExpandAndLockInstance()
4756

    
4757
    if self.op.iallocator is not None:
4758
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4759
    elif self.op.remote_node is not None:
4760
      remote_node = self.cfg.ExpandNodeName(self.op.remote_node)
4761
      if remote_node is None:
4762
        raise errors.OpPrereqError("Node '%s' not known" %
4763
                                   self.op.remote_node)
4764
      self.op.remote_node = remote_node
4765
      # Warning: do not remove the locking of the new secondary here
4766
      # unless DRBD8.AddChildren is changed to work in parallel;
4767
      # currently it doesn't since parallel invocations of
4768
      # FindUnusedMinor will conflict
4769
      self.needed_locks[locking.LEVEL_NODE] = [remote_node]
4770
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
4771
    else:
4772
      self.needed_locks[locking.LEVEL_NODE] = []
4773
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4774

    
4775
  def DeclareLocks(self, level):
4776
    # If we're not already locking all nodes in the set we have to declare the
4777
    # instance's primary/secondary nodes.
4778
    if (level == locking.LEVEL_NODE and
4779
        self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
4780
      self._LockInstancesNodes()
4781

    
4782
  def _RunAllocator(self):
4783
    """Compute a new secondary node using an IAllocator.
4784

4785
    """
4786
    ial = IAllocator(self,
4787
                     mode=constants.IALLOCATOR_MODE_RELOC,
4788
                     name=self.op.instance_name,
4789
                     relocate_from=[self.sec_node])
4790

    
4791
    ial.Run(self.op.iallocator)
4792

    
4793
    if not ial.success:
4794
      raise errors.OpPrereqError("Can't compute nodes using"
4795
                                 " iallocator '%s': %s" % (self.op.iallocator,
4796
                                                           ial.info))
4797
    if len(ial.nodes) != ial.required_nodes:
4798
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
4799
                                 " of nodes (%s), required %s" %
4800
                                 (len(ial.nodes), ial.required_nodes))
4801
    self.op.remote_node = ial.nodes[0]
4802
    self.LogInfo("Selected new secondary for the instance: %s",
4803
                 self.op.remote_node)
4804

    
4805
  def BuildHooksEnv(self):
4806
    """Build hooks env.
4807

4808
    This runs on the master, the primary and all the secondaries.
4809

4810
    """
4811
    env = {
4812
      "MODE": self.op.mode,
4813
      "NEW_SECONDARY": self.op.remote_node,
4814
      "OLD_SECONDARY": self.instance.secondary_nodes[0],
4815
      }
4816
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4817
    nl = [
4818
      self.cfg.GetMasterNode(),
4819
      self.instance.primary_node,
4820
      ]
4821
    if self.op.remote_node is not None:
4822
      nl.append(self.op.remote_node)
4823
    return env, nl, nl
4824

    
4825
  def CheckPrereq(self):
4826
    """Check prerequisites.
4827

4828
    This checks that the instance is in the cluster.
4829

4830
    """
4831
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4832
    assert instance is not None, \
4833
      "Cannot retrieve locked instance %s" % self.op.instance_name
4834
    self.instance = instance
4835

    
4836
    if instance.disk_template != constants.DT_DRBD8:
4837
      raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
4838
                                 " instances")
4839

    
4840
    if len(instance.secondary_nodes) != 1:
4841
      raise errors.OpPrereqError("The instance has a strange layout,"
4842
                                 " expected one secondary but found %d" %
4843
                                 len(instance.secondary_nodes))
4844

    
4845
    self.sec_node = instance.secondary_nodes[0]
4846

    
4847
    if self.op.iallocator is not None:
4848
      self._RunAllocator()
4849

    
4850
    remote_node = self.op.remote_node
4851
    if remote_node is not None:
4852
      self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
4853
      assert self.remote_node_info is not None, \
4854
        "Cannot retrieve locked node %s" % remote_node
4855
    else:
4856
      self.remote_node_info = None
4857
    if remote_node == instance.primary_node:
4858
      raise errors.OpPrereqError("The specified node is the primary node of"
4859
                                 " the instance.")
4860
    elif remote_node == self.sec_node:
4861
      raise errors.OpPrereqError("The specified node is already the"
4862
                                 " secondary node of the instance.")
4863

    
4864
    if self.op.mode == constants.REPLACE_DISK_PRI:
4865
      n1 = self.tgt_node = instance.primary_node
4866
      n2 = self.oth_node = self.sec_node
4867
    elif self.op.mode == constants.REPLACE_DISK_SEC:
4868
      n1 = self.tgt_node = self.sec_node
4869
      n2 = self.oth_node = instance.primary_node
4870
    elif self.op.mode == constants.REPLACE_DISK_CHG:
4871
      n1 = self.new_node = remote_node
4872
      n2 = self.oth_node = instance.primary_node
4873
      self.tgt_node = self.sec_node
4874
    else:
4875
      raise errors.ProgrammerError("Unhandled disk replace mode")
4876

    
4877
    _CheckNodeOnline(self, n1)
4878
    _CheckNodeOnline(self, n2)
4879

    
4880
    if not self.op.disks:
4881
      self.op.disks = range(len(instance.disks))
4882

    
4883
    for disk_idx in self.op.disks:
4884
      instance.FindDisk(disk_idx)
4885

    
4886
  def _ExecD8DiskOnly(self, feedback_fn):
4887
    """Replace a disk on the primary or secondary for dbrd8.
4888

4889
    The algorithm for replace is quite complicated:
4890

4891
      1. for each disk to be replaced:
4892

4893
        1. create new LVs on the target node with unique names
4894
        1. detach old LVs from the drbd device
4895
        1. rename old LVs to name_replaced.<time_t>
4896
        1. rename new LVs to old LVs
4897
        1. attach the new LVs (with the old names now) to the drbd device
4898

4899
      1. wait for sync across all devices
4900

4901
      1. for each modified disk:
4902

4903
        1. remove old LVs (which have the name name_replaces.<time_t>)
4904

4905
    Failures are not very well handled.
4906

4907
    """
4908
    steps_total = 6
4909
    warning, info = (self.proc.LogWarning, self.proc.LogInfo)
4910
    instance = self.instance
4911
    iv_names = {}
4912
    vgname = self.cfg.GetVGName()
4913
    # start of work
4914
    cfg = self.cfg
4915
    tgt_node = self.tgt_node
4916
    oth_node = self.oth_node
4917

    
4918
    # Step: check device activation
4919
    self.proc.LogStep(1, steps_total, "check device existence")
4920
    info("checking volume groups")
4921
    my_vg = cfg.GetVGName()
4922
    results = self.rpc.call_vg_list([oth_node, tgt_node])
4923
    if not results:
4924
      raise errors.OpExecError("Can't list volume groups on the nodes")
4925
    for node in oth_node, tgt_node:
4926
      res = results[node]
4927
      if res.failed or not res.data or my_vg not in res.data:
4928
        raise errors.OpExecError("Volume group '%s' not found on %s" %
4929
                                 (my_vg, node))
4930
    for idx, dev in enumerate(instance.disks):
4931
      if idx not in self.op.disks:
4932
        continue
4933
      for node in tgt_node, oth_node:
4934
        info("checking disk/%d on %s" % (idx, node))
4935
        cfg.SetDiskID(dev, node)
4936
        result = self.rpc.call_blockdev_find(node, dev)
4937
        msg = result.RemoteFailMsg()
4938
        if not msg and not result.payload:
4939
          msg = "disk not found"
4940
        if msg:
4941
          raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
4942
                                   (idx, node, msg))
4943

    
4944
    # Step: check other node consistency
4945
    self.proc.LogStep(2, steps_total, "check peer consistency")
4946
    for idx, dev in enumerate(instance.disks):
4947
      if idx not in self.op.disks:
4948
        continue
4949
      info("checking disk/%d consistency on %s" % (idx, oth_node))
4950
      if not _CheckDiskConsistency(self, dev, oth_node,
4951
                                   oth_node==instance.primary_node):
4952
        raise errors.OpExecError("Peer node (%s) has degraded storage, unsafe"
4953
                                 " to replace disks on this node (%s)" %
4954
                                 (oth_node, tgt_node))
4955

    
4956
    # Step: create new storage
4957
    self.proc.LogStep(3, steps_total, "allocate new storage")
4958
    for idx, dev in enumerate(instance.disks):
4959
      if idx not in self.op.disks:
4960
        continue
4961
      size = dev.size
4962
      cfg.SetDiskID(dev, tgt_node)
4963
      lv_names = [".disk%d_%s" % (idx, suf)
4964
                  for suf in ["data", "meta"]]
4965
      names = _GenerateUniqueNames(self, lv_names)
4966
      lv_data = objects.Disk(dev_type=constants.LD_LV, size=size,
4967
                             logical_id=(vgname, names[0]))
4968
      lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
4969
                             logical_id=(vgname, names[1]))
4970
      new_lvs = [lv_data, lv_meta]
4971
      old_lvs = dev.children
4972
      iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
4973
      info("creating new local storage on %s for %s" %
4974
           (tgt_node, dev.iv_name))
4975
      # we pass force_create=True to force the LVM creation
4976
      for new_lv in new_lvs:
4977
        _CreateBlockDev(self, tgt_node, instance, new_lv, True,
4978
                        _GetInstanceInfoText(instance), False)
4979

    
4980
    # Step: for each lv, detach+rename*2+attach
4981
    self.proc.LogStep(4, steps_total, "change drbd configuration")
4982
    for dev, old_lvs, new_lvs in iv_names.itervalues():
4983
      info("detaching %s drbd from local storage" % dev.iv_name)
4984
      result = self.rpc.call_blockdev_removechildren(tgt_node, dev, old_lvs)
4985
      result.Raise()
4986
      if not result.data:
4987
        raise errors.OpExecError("Can't detach drbd from local storage on node"
4988
                                 " %s for device %s" % (tgt_node, dev.iv_name))
4989
      #dev.children = []
4990
      #cfg.Update(instance)
4991

    
4992
      # ok, we created the new LVs, so now we know we have the needed
4993
      # storage; as such, we proceed on the target node to rename
4994
      # old_lv to _old, and new_lv to old_lv; note that we rename LVs
4995
      # using the assumption that logical_id == physical_id (which in
4996
      # turn is the unique_id on that node)
4997

    
4998
      # FIXME(iustin): use a better name for the replaced LVs
4999
      temp_suffix = int(time.time())
5000
      ren_fn = lambda d, suff: (d.physical_id[0],
5001
                                d.physical_id[1] + "_replaced-%s" % suff)
5002
      # build the rename list based on what LVs exist on the node
5003
      rlist = []
5004
      for to_ren in old_lvs:
5005
        result = self.rpc.call_blockdev_find(tgt_node, to_ren)
5006
        if not result.RemoteFailMsg() and result.payload:
5007
          # device exists
5008
          rlist.append((to_ren, ren_fn(to_ren, temp_suffix)))
5009

    
5010
      info("renaming the old LVs on the target node")
5011
      result = self.rpc.call_blockdev_rename(tgt_node, rlist)
5012
      result.Raise()
5013
      if not result.data:
5014
        raise errors.OpExecError("Can't rename old LVs on node %s" % tgt_node)
5015
      # now we rename the new LVs to the old LVs
5016
      info("renaming the new LVs on the target node")
5017
      rlist = [(new, old.physical_id) for old, new in zip(old_lvs, new_lvs)]
5018
      result = self.rpc.call_blockdev_rename(tgt_node, rlist)
5019
      result.Raise()
5020
      if not result.data:
5021
        raise errors.OpExecError("Can't rename new LVs on node %s" % tgt_node)
5022

    
5023
      for old, new in zip(old_lvs, new_lvs):
5024
        new.logical_id = old.logical_id
5025
        cfg.SetDiskID(new, tgt_node)
5026

    
5027
      for disk in old_lvs:
5028
        disk.logical_id = ren_fn(disk, temp_suffix)
5029
        cfg.SetDiskID(disk, tgt_node)
5030

    
5031
      # now that the new lvs have the old name, we can add them to the device
5032
      info("adding new mirror component on %s" % tgt_node)
5033
      result = self.rpc.call_blockdev_addchildren(tgt_node, dev, new_lvs)
5034
      if result.failed or not result.data:
5035
        for new_lv in new_lvs:
5036
          result = self.rpc.call_blockdev_remove(tgt_node, new_lv)
5037
          if result.failed or not result.data:
5038
            warning("Can't rollback device %s", hint="manually cleanup unused"
5039
                    " logical volumes")
5040
        raise errors.OpExecError("Can't add local storage to drbd")
5041

    
5042
      dev.children = new_lvs
5043
      cfg.Update(instance)
5044

    
5045
    # Step: wait for sync
5046

    
5047
    # this can fail as the old devices are degraded and _WaitForSync
5048
    # does a combined result over all disks, so we don't check its
5049
    # return value
5050
    self.proc.LogStep(5, steps_total, "sync devices")
5051
    _WaitForSync(self, instance, unlock=True)
5052

    
5053
    # so check manually all the devices
5054
    for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
5055
      cfg.SetDiskID(dev, instance.primary_node)
5056
      result = self.rpc.call_blockdev_find(instance.primary_node, dev)
5057
      msg = result.RemoteFailMsg()
5058
      if not msg and not result.payload:
5059
        msg = "disk not found"
5060
      if msg:
5061
        raise errors.OpExecError("Can't find DRBD device %s: %s" %
5062
                                 (name, msg))
5063
      if result.payload[5]:
5064
        raise errors.OpExecError("DRBD device %s is degraded!" % name)
5065

    
5066
    # Step: remove old storage
5067
    self.proc.LogStep(6, steps_total, "removing old storage")
5068
    for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
5069
      info("remove logical volumes for %s" % name)
5070
      for lv in old_lvs:
5071
        cfg.SetDiskID(lv, tgt_node)
5072
        result = self.rpc.call_blockdev_remove(tgt_node, lv)
5073
        if result.failed or not result.data:
5074
          warning("Can't remove old LV", hint="manually remove unused LVs")
5075
          continue
5076

    
5077
  def _ExecD8Secondary(self, feedback_fn):
5078
    """Replace the secondary node for drbd8.
5079

5080
    The algorithm for replace is quite complicated:
5081
      - for all disks of the instance:
5082
        - create new LVs on the new node with same names
5083
        - shutdown the drbd device on the old secondary
5084
        - disconnect the drbd network on the primary
5085
        - create the drbd device on the new secondary
5086
        - network attach the drbd on the primary, using an artifice:
5087
          the drbd code for Attach() will connect to the network if it
5088
          finds a device which is connected to the good local disks but
5089
          not network enabled
5090
      - wait for sync across all devices
5091
      - remove all disks from the old secondary
5092

5093
    Failures are not very well handled.
5094

5095
    """
5096
    steps_total = 6
5097
    warning, info = (self.proc.LogWarning, self.proc.LogInfo)
5098
    instance = self.instance
5099
    iv_names = {}
5100
    # start of work
5101
    cfg = self.cfg
5102
    old_node = self.tgt_node
5103
    new_node = self.new_node
5104
    pri_node = instance.primary_node
5105
    nodes_ip = {
5106
      old_node: self.cfg.GetNodeInfo(old_node).secondary_ip,
5107
      new_node: self.cfg.GetNodeInfo(new_node).secondary_ip,
5108
      pri_node: self.cfg.GetNodeInfo(pri_node).secondary_ip,
5109
      }
5110

    
5111
    # Step: check device activation
5112
    self.proc.LogStep(1, steps_total, "check device existence")
5113
    info("checking volume groups")
5114
    my_vg = cfg.GetVGName()
5115
    results = self.rpc.call_vg_list([pri_node, new_node])
5116
    for node in pri_node, new_node:
5117
      res = results[node]
5118
      if res.failed or not res.data or my_vg not in res.data:
5119
        raise errors.OpExecError("Volume group '%s' not found on %s" %
5120
                                 (my_vg, node))
5121
    for idx, dev in enumerate(instance.disks):
5122
      if idx not in self.op.disks:
5123
        continue
5124
      info("checking disk/%d on %s" % (idx, pri_node))
5125
      cfg.SetDiskID(dev, pri_node)
5126
      result = self.rpc.call_blockdev_find(pri_node, dev)
5127
      msg = result.RemoteFailMsg()
5128
      if not msg and not result.payload:
5129
        msg = "disk not found"
5130
      if msg:
5131
        raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
5132
                                 (idx, pri_node, msg))
5133

    
5134
    # Step: check other node consistency
5135
    self.proc.LogStep(2, steps_total, "check peer consistency")
5136
    for idx, dev in enumerate(instance.disks):
5137
      if idx not in self.op.disks:
5138
        continue
5139
      info("checking disk/%d consistency on %s" % (idx, pri_node))
5140
      if not _CheckDiskConsistency(self, dev, pri_node, True, ldisk=True):
5141
        raise errors.OpExecError("Primary node (%s) has degraded storage,"
5142
                                 " unsafe to replace the secondary" %
5143
                                 pri_node)
5144

    
5145
    # Step: create new storage
5146
    self.proc.LogStep(3, steps_total, "allocate new storage")
5147
    for idx, dev in enumerate(instance.disks):
5148
      info("adding new local storage on %s for disk/%d" %
5149
           (new_node, idx))
5150
      # we pass force_create=True to force LVM creation
5151
      for new_lv in dev.children:
5152
        _CreateBlockDev(self, new_node, instance, new_lv, True,
5153
                        _GetInstanceInfoText(instance), False)
5154

    
5155
    # Step 4: dbrd minors and drbd setups changes
5156
    # after this, we must manually remove the drbd minors on both the
5157
    # error and the success paths
5158
    minors = cfg.AllocateDRBDMinor([new_node for dev in instance.disks],
5159
                                   instance.name)
5160
    logging.debug("Allocated minors %s" % (minors,))
5161
    self.proc.LogStep(4, steps_total, "changing drbd configuration")
5162
    for idx, (dev, new_minor) in enumerate(zip(instance.disks, minors)):
5163
      size = dev.size
5164
      info("activating a new drbd on %s for disk/%d" % (new_node, idx))
5165
      # create new devices on new_node; note that we create two IDs:
5166
      # one without port, so the drbd will be activated without
5167
      # networking information on the new node at this stage, and one
5168
      # with network, for the latter activation in step 4
5169
      (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
5170
      if pri_node == o_node1:
5171
        p_minor = o_minor1
5172
      else:
5173
        p_minor = o_minor2
5174

    
5175
      new_alone_id = (pri_node, new_node, None, p_minor, new_minor, o_secret)
5176
      new_net_id = (pri_node, new_node, o_port, p_minor, new_minor, o_secret)
5177

    
5178
      iv_names[idx] = (dev, dev.children, new_net_id)
5179
      logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
5180
                    new_net_id)
5181
      new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
5182
                              logical_id=new_alone_id,
5183
                              children=dev.children)
5184
      try:
5185
        _CreateSingleBlockDev(self, new_node, instance, new_drbd,
5186
                              _GetInstanceInfoText(instance), False)
5187
      except errors.BlockDeviceError:
5188
        self.cfg.ReleaseDRBDMinors(instance.name)
5189
        raise
5190

    
5191
    for idx, dev in enumerate(instance.disks):
5192
      # we have new devices, shutdown the drbd on the old secondary
5193
      info("shutting down drbd for disk/%d on old node" % idx)
5194
      cfg.SetDiskID(dev, old_node)
5195
      result = self.rpc.call_blockdev_shutdown(old_node, dev)
5196
      if result.failed or not result.data:
5197
        warning("Failed to shutdown drbd for disk/%d on old node" % idx,
5198
                hint="Please cleanup this device manually as soon as possible")
5199

    
5200
    info("detaching primary drbds from the network (=> standalone)")
5201
    result = self.rpc.call_drbd_disconnect_net([pri_node], nodes_ip,
5202
                                               instance.disks)[pri_node]
5203

    
5204
    msg = result.RemoteFailMsg()
5205
    if msg:
5206
      # detaches didn't succeed (unlikely)
5207
      self.cfg.ReleaseDRBDMinors(instance.name)
5208
      raise errors.OpExecError("Can't detach the disks from the network on"
5209
                               " old node: %s" % (msg,))
5210

    
5211
    # if we managed to detach at least one, we update all the disks of
5212
    # the instance to point to the new secondary
5213
    info("updating instance configuration")
5214
    for dev, _, new_logical_id in iv_names.itervalues():
5215
      dev.logical_id = new_logical_id
5216
      cfg.SetDiskID(dev, pri_node)
5217
    cfg.Update(instance)
5218

    
5219
    # and now perform the drbd attach
5220
    info("attaching primary drbds to new secondary (standalone => connected)")
5221
    result = self.rpc.call_drbd_attach_net([pri_node, new_node], nodes_ip,
5222
                                           instance.disks, instance.name,
5223
                                           False)
5224
    for to_node, to_result in result.items():
5225
      msg = to_result.RemoteFailMsg()
5226
      if msg:
5227
        warning("can't attach drbd disks on node %s: %s", to_node, msg,
5228
                hint="please do a gnt-instance info to see the"
5229
                " status of disks")
5230

    
5231
    # this can fail as the old devices are degraded and _WaitForSync
5232
    # does a combined result over all disks, so we don't check its
5233
    # return value
5234
    self.proc.LogStep(5, steps_total, "sync devices")
5235
    _WaitForSync(self, instance, unlock=True)
5236

    
5237
    # so check manually all the devices
5238
    for idx, (dev, old_lvs, _) in iv_names.iteritems():
5239
      cfg.SetDiskID(dev, pri_node)
5240
      result = self.rpc.call_blockdev_find(pri_node, dev)
5241
      msg = result.RemoteFailMsg()
5242
      if not msg and not result.payload:
5243
        msg = "disk not found"
5244
      if msg:
5245
        raise errors.OpExecError("Can't find DRBD device disk/%d: %s" %
5246
                                 (idx, msg))
5247
      if result.payload[5]:
5248
        raise errors.OpExecError("DRBD device disk/%d is degraded!" % idx)
5249

    
5250
    self.proc.LogStep(6, steps_total, "removing old storage")
5251
    for idx, (dev, old_lvs, _) in iv_names.iteritems():
5252
      info("remove logical volumes for disk/%d" % idx)
5253
      for lv in old_lvs:
5254
        cfg.SetDiskID(lv, old_node)
5255
        result = self.rpc.call_blockdev_remove(old_node, lv)
5256
        if result.failed or not result.data:
5257
          warning("Can't remove LV on old secondary",
5258
                  hint="Cleanup stale volumes by hand")
5259

    
5260
  def Exec(self, feedback_fn):
5261
    """Execute disk replacement.
5262

5263
    This dispatches the disk replacement to the appropriate handler.
5264

5265
    """
5266
    instance = self.instance
5267

    
5268
    # Activate the instance disks if we're replacing them on a down instance
5269
    if not instance.admin_up:
5270
      _StartInstanceDisks(self, instance, True)
5271

    
5272
    if self.op.mode == constants.REPLACE_DISK_CHG:
5273
      fn = self._ExecD8Secondary
5274
    else:
5275
      fn = self._ExecD8DiskOnly
5276

    
5277
    ret = fn(feedback_fn)
5278

    
5279
    # Deactivate the instance disks if we're replacing them on a down instance
5280
    if not instance.admin_up:
5281
      _SafeShutdownInstanceDisks(self, instance)
5282

    
5283
    return ret
5284

    
5285

    
5286
class LUGrowDisk(LogicalUnit):
5287
  """Grow a disk of an instance.
5288

5289
  """
5290
  HPATH = "disk-grow"
5291
  HTYPE = constants.HTYPE_INSTANCE
5292
  _OP_REQP = ["instance_name", "disk", "amount", "wait_for_sync"]
5293
  REQ_BGL = False
5294

    
5295
  def ExpandNames(self):
5296
    self._ExpandAndLockInstance()
5297
    self.needed_locks[locking.LEVEL_NODE] = []
5298
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5299

    
5300
  def DeclareLocks(self, level):
5301
    if level == locking.LEVEL_NODE:
5302
      self._LockInstancesNodes()
5303

    
5304
  def BuildHooksEnv(self):
5305
    """Build hooks env.
5306

5307
    This runs on the master, the primary and all the secondaries.
5308

5309
    """
5310
    env = {
5311
      "DISK": self.op.disk,
5312
      "AMOUNT": self.op.amount,
5313
      }
5314
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5315
    nl = [
5316
      self.cfg.GetMasterNode(),
5317
      self.instance.primary_node,
5318
      ]
5319
    return env, nl, nl
5320

    
5321
  def CheckPrereq(self):
5322
    """Check prerequisites.
5323

5324
    This checks that the instance is in the cluster.
5325

5326
    """
5327
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5328
    assert instance is not None, \
5329
      "Cannot retrieve locked instance %s" % self.op.instance_name
5330
    nodenames = list(instance.all_nodes)
5331
    for node in nodenames:
5332
      _CheckNodeOnline(self, node)
5333

    
5334

    
5335
    self.instance = instance
5336

    
5337
    if instance.disk_template not in (constants.DT_PLAIN, constants.DT_DRBD8):
5338
      raise errors.OpPrereqError("Instance's disk layout does not support"
5339
                                 " growing.")
5340

    
5341
    self.disk = instance.FindDisk(self.op.disk)
5342

    
5343
    nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
5344
                                       instance.hypervisor)
5345
    for node in nodenames:
5346
      info = nodeinfo[node]
5347
      if info.failed or not info.data:
5348
        raise errors.OpPrereqError("Cannot get current information"
5349
                                   " from node '%s'" % node)
5350
      vg_free = info.data.get('vg_free', None)
5351
      if not isinstance(vg_free, int):
5352
        raise errors.OpPrereqError("Can't compute free disk space on"
5353
                                   " node %s" % node)
5354
      if self.op.amount > vg_free:
5355
        raise errors.OpPrereqError("Not enough disk space on target node %s:"
5356
                                   " %d MiB available, %d MiB required" %
5357
                                   (node, vg_free, self.op.amount))
5358

    
5359
  def Exec(self, feedback_fn):
5360
    """Execute disk grow.
5361

5362
    """
5363
    instance = self.instance
5364
    disk = self.disk
5365
    for node in instance.all_nodes:
5366
      self.cfg.SetDiskID(disk, node)
5367
      result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
5368
      msg = result.RemoteFailMsg()
5369
      if msg:
5370
        raise errors.OpExecError("Grow request failed to node %s: %s" %
5371
                                 (node, msg))
5372
    disk.RecordGrow(self.op.amount)
5373
    self.cfg.Update(instance)
5374
    if self.op.wait_for_sync:
5375
      disk_abort = not _WaitForSync(self, instance)
5376
      if disk_abort:
5377
        self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
5378
                             " status.\nPlease check the instance.")
5379

    
5380

    
5381
class LUQueryInstanceData(NoHooksLU):
5382
  """Query runtime instance data.
5383

5384
  """
5385
  _OP_REQP = ["instances", "static"]
5386
  REQ_BGL = False
5387

    
5388
  def ExpandNames(self):
5389
    self.needed_locks = {}
5390
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
5391

    
5392
    if not isinstance(self.op.instances, list):
5393
      raise errors.OpPrereqError("Invalid argument type 'instances'")
5394

    
5395
    if self.op.instances:
5396
      self.wanted_names = []
5397
      for name in self.op.instances:
5398
        full_name = self.cfg.ExpandInstanceName(name)
5399
        if full_name is None:
5400
          raise errors.OpPrereqError("Instance '%s' not known" % name)
5401
        self.wanted_names.append(full_name)
5402
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
5403
    else:
5404
      self.wanted_names = None
5405
      self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
5406

    
5407
    self.needed_locks[locking.LEVEL_NODE] = []
5408
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5409

    
5410
  def DeclareLocks(self, level):
5411
    if level == locking.LEVEL_NODE:
5412
      self._LockInstancesNodes()
5413

    
5414
  def CheckPrereq(self):
5415
    """Check prerequisites.
5416

5417
    This only checks the optional instance list against the existing names.
5418

5419
    """
5420
    if self.wanted_names is None:
5421
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
5422

    
5423
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
5424
                             in self.wanted_names]
5425
    return
5426

    
5427
  def _ComputeDiskStatus(self, instance, snode, dev):
5428
    """Compute block device status.
5429

5430
    """
5431
    static = self.op.static
5432
    if not static:
5433
      self.cfg.SetDiskID(dev, instance.primary_node)
5434
      dev_pstatus = self.rpc.call_blockdev_find(instance.primary_node, dev)
5435
      msg = dev_pstatus.RemoteFailMsg()
5436
      if msg:
5437
        raise errors.OpExecError("Can't compute disk status for %s: %s" %
5438
                                 (instance.name, msg))
5439
      dev_pstatus = dev_pstatus.payload
5440
    else:
5441
      dev_pstatus = None
5442

    
5443
    if dev.dev_type in constants.LDS_DRBD:
5444
      # we change the snode then (otherwise we use the one passed in)
5445
      if dev.logical_id[0] == instance.primary_node:
5446
        snode = dev.logical_id[1]
5447
      else:
5448
        snode = dev.logical_id[0]
5449

    
5450
    if snode and not static:
5451
      self.cfg.SetDiskID(dev, snode)
5452
      dev_sstatus = self.rpc.call_blockdev_find(snode, dev)
5453
      msg = dev_sstatus.RemoteFailMsg()
5454
      if msg:
5455
        raise errors.OpExecError("Can't compute disk status for %s: %s" %
5456
                                 (instance.name, msg))
5457
      dev_sstatus = dev_sstatus.payload
5458
    else:
5459
      dev_sstatus = None
5460

    
5461
    if dev.children:
5462
      dev_children = [self._ComputeDiskStatus(instance, snode, child)
5463
                      for child in dev.children]
5464
    else:
5465
      dev_children = []
5466

    
5467
    data = {
5468
      "iv_name": dev.iv_name,
5469
      "dev_type": dev.dev_type,
5470
      "logical_id": dev.logical_id,
5471
      "physical_id": dev.physical_id,
5472
      "pstatus": dev_pstatus,
5473
      "sstatus": dev_sstatus,
5474
      "children": dev_children,
5475
      "mode": dev.mode,
5476
      }
5477

    
5478
    return data
5479

    
5480
  def Exec(self, feedback_fn):
5481
    """Gather and return data"""
5482
    result = {}
5483

    
5484
    cluster = self.cfg.GetClusterInfo()
5485

    
5486
    for instance in self.wanted_instances:
5487
      if not self.op.static:
5488
        remote_info = self.rpc.call_instance_info(instance.primary_node,
5489
                                                  instance.name,
5490
                                                  instance.hypervisor)
5491
        remote_info.Raise()
5492
        remote_info = remote_info.data
5493
        if remote_info and "state" in remote_info:
5494
          remote_state = "up"
5495
        else:
5496
          remote_state = "down"
5497
      else:
5498
        remote_state = None
5499
      if instance.admin_up:
5500
        config_state = "up"
5501
      else:
5502
        config_state = "down"
5503

    
5504
      disks = [self._ComputeDiskStatus(instance, None, device)
5505
               for device in instance.disks]
5506

    
5507
      idict = {
5508
        "name": instance.name,
5509
        "config_state": config_state,
5510
        "run_state": remote_state,
5511
        "pnode": instance.primary_node,
5512
        "snodes": instance.secondary_nodes,
5513
        "os": instance.os,
5514
        "nics": [(nic.mac, nic.ip, nic.bridge) for nic in instance.nics],
5515
        "disks": disks,
5516
        "hypervisor": instance.hypervisor,
5517
        "network_port": instance.network_port,
5518
        "hv_instance": instance.hvparams,
5519
        "hv_actual": cluster.FillHV(instance),
5520
        "be_instance": instance.beparams,
5521
        "be_actual": cluster.FillBE(instance),
5522
        }
5523

    
5524
      result[instance.name] = idict
5525

    
5526
    return result
5527

    
5528

    
5529
class LUSetInstanceParams(LogicalUnit):
5530
  """Modifies an instances's parameters.
5531

5532
  """
5533
  HPATH = "instance-modify"
5534
  HTYPE = constants.HTYPE_INSTANCE
5535
  _OP_REQP = ["instance_name"]
5536
  REQ_BGL = False
5537

    
5538
  def CheckArguments(self):
5539
    if not hasattr(self.op, 'nics'):
5540
      self.op.nics = []
5541
    if not hasattr(self.op, 'disks'):
5542
      self.op.disks = []
5543
    if not hasattr(self.op, 'beparams'):
5544
      self.op.beparams = {}
5545
    if not hasattr(self.op, 'hvparams'):
5546
      self.op.hvparams = {}
5547
    self.op.force = getattr(self.op, "force", False)
5548
    if not (self.op.nics or self.op.disks or
5549
            self.op.hvparams or self.op.beparams):
5550
      raise errors.OpPrereqError("No changes submitted")
5551

    
5552
    utils.CheckBEParams(self.op.beparams)
5553

    
5554
    # Disk validation
5555
    disk_addremove = 0
5556
    for disk_op, disk_dict in self.op.disks:
5557
      if disk_op == constants.DDM_REMOVE:
5558
        disk_addremove += 1
5559
        continue
5560
      elif disk_op == constants.DDM_ADD:
5561
        disk_addremove += 1
5562
      else:
5563
        if not isinstance(disk_op, int):
5564
          raise errors.OpPrereqError("Invalid disk index")
5565
      if disk_op == constants.DDM_ADD:
5566
        mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
5567
        if mode not in constants.DISK_ACCESS_SET:
5568
          raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode)
5569
        size = disk_dict.get('size', None)
5570
        if size is None:
5571
          raise errors.OpPrereqError("Required disk parameter size missing")
5572
        try:
5573
          size = int(size)
5574
        except ValueError, err:
5575
          raise errors.OpPrereqError("Invalid disk size parameter: %s" %
5576
                                     str(err))
5577
        disk_dict['size'] = size
5578
      else:
5579
        # modification of disk
5580
        if 'size' in disk_dict:
5581
          raise errors.OpPrereqError("Disk size change not possible, use"
5582
                                     " grow-disk")
5583

    
5584
    if disk_addremove > 1:
5585
      raise errors.OpPrereqError("Only one disk add or remove operation"
5586
                                 " supported at a time")
5587

    
5588
    # NIC validation
5589
    nic_addremove = 0
5590
    for nic_op, nic_dict in self.op.nics:
5591
      if nic_op == constants.DDM_REMOVE:
5592
        nic_addremove += 1
5593
        continue
5594
      elif nic_op == constants.DDM_ADD:
5595
        nic_addremove += 1
5596
      else:
5597
        if not isinstance(nic_op, int):
5598
          raise errors.OpPrereqError("Invalid nic index")
5599

    
5600
      # nic_dict should be a dict
5601
      nic_ip = nic_dict.get('ip', None)
5602
      if nic_ip is not None:
5603
        if nic_ip.lower() == "none":
5604
          nic_dict['ip'] = None
5605
        else:
5606
          if not utils.IsValidIP(nic_ip):
5607
            raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip)
5608
      # we can only check None bridges and assign the default one
5609
      nic_bridge = nic_dict.get('bridge', None)
5610
      if nic_bridge is None:
5611
        nic_dict['bridge'] = self.cfg.GetDefBridge()
5612
      # but we can validate MACs
5613
      nic_mac = nic_dict.get('mac', None)
5614
      if nic_mac is not None:
5615
        if self.cfg.IsMacInUse(nic_mac):
5616
          raise errors.OpPrereqError("MAC address %s already in use"
5617
                                     " in cluster" % nic_mac)
5618
        if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
5619
          if not utils.IsValidMac(nic_mac):
5620
            raise errors.OpPrereqError("Invalid MAC address %s" % nic_mac)
5621
    if nic_addremove > 1:
5622
      raise errors.OpPrereqError("Only one NIC add or remove operation"
5623
                                 " supported at a time")
5624

    
5625
  def ExpandNames(self):
5626
    self._ExpandAndLockInstance()
5627
    self.needed_locks[locking.LEVEL_NODE] = []
5628
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5629

    
5630
  def DeclareLocks(self, level):
5631
    if level == locking.LEVEL_NODE:
5632
      self._LockInstancesNodes()
5633

    
5634
  def BuildHooksEnv(self):
5635
    """Build hooks env.
5636

5637
    This runs on the master, primary and secondaries.
5638

5639
    """
5640
    args = dict()
5641
    if constants.BE_MEMORY in self.be_new:
5642
      args['memory'] = self.be_new[constants.BE_MEMORY]
5643
    if constants.BE_VCPUS in self.be_new:
5644
      args['vcpus'] = self.be_new[constants.BE_VCPUS]
5645
    # FIXME: readd disk/nic changes
5646
    env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
5647
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
5648
    return env, nl, nl
5649

    
5650
  def CheckPrereq(self):
5651
    """Check prerequisites.
5652

5653
    This only checks the instance list against the existing names.
5654

5655
    """
5656
    force = self.force = self.op.force
5657

    
5658
    # checking the new params on the primary/secondary nodes
5659

    
5660
    instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5661
    assert self.instance is not None, \
5662
      "Cannot retrieve locked instance %s" % self.op.instance_name
5663
    pnode = instance.primary_node
5664
    nodelist = list(instance.all_nodes)
5665

    
5666
    # hvparams processing
5667
    if self.op.hvparams:
5668
      i_hvdict = copy.deepcopy(instance.hvparams)
5669
      for key, val in self.op.hvparams.iteritems():
5670
        if val == constants.VALUE_DEFAULT:
5671
          try:
5672
            del i_hvdict[key]
5673
          except KeyError:
5674
            pass
5675
        elif val == constants.VALUE_NONE:
5676
          i_hvdict[key] = None
5677
        else:
5678
          i_hvdict[key] = val
5679
      cluster = self.cfg.GetClusterInfo()
5680
      hv_new = cluster.FillDict(cluster.hvparams[instance.hypervisor],
5681
                                i_hvdict)
5682
      # local check
5683
      hypervisor.GetHypervisor(
5684
        instance.hypervisor).CheckParameterSyntax(hv_new)
5685
      _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
5686
      self.hv_new = hv_new # the new actual values
5687
      self.hv_inst = i_hvdict # the new dict (without defaults)
5688
    else:
5689
      self.hv_new = self.hv_inst = {}
5690

    
5691
    # beparams processing
5692
    if self.op.beparams:
5693
      i_bedict = copy.deepcopy(instance.beparams)
5694
      for key, val in self.op.beparams.iteritems():
5695
        if val == constants.VALUE_DEFAULT:
5696
          try:
5697
            del i_bedict[key]
5698
          except KeyError:
5699
            pass
5700
        else:
5701
          i_bedict[key] = val
5702
      cluster = self.cfg.GetClusterInfo()
5703
      be_new = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
5704
                                i_bedict)
5705
      self.be_new = be_new # the new actual values
5706
      self.be_inst = i_bedict # the new dict (without defaults)
5707
    else:
5708
      self.be_new = self.be_inst = {}
5709

    
5710
    self.warn = []
5711

    
5712
    if constants.BE_MEMORY in self.op.beparams and not self.force:
5713
      mem_check_list = [pnode]
5714
      if be_new[constants.BE_AUTO_BALANCE]:
5715
        # either we changed auto_balance to yes or it was from before
5716
        mem_check_list.extend(instance.secondary_nodes)
5717
      instance_info = self.rpc.call_instance_info(pnode, instance.name,
5718
                                                  instance.hypervisor)
5719
      nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
5720
                                         instance.hypervisor)
5721
      if nodeinfo[pnode].failed or not isinstance(nodeinfo[pnode].data, dict):
5722
        # Assume the primary node is unreachable and go ahead
5723
        self.warn.append("Can't get info from primary node %s" % pnode)
5724
      else:
5725
        if not instance_info.failed and instance_info.data:
5726
          current_mem = instance_info.data['memory']
5727
        else:
5728
          # Assume instance not running
5729
          # (there is a slight race condition here, but it's not very probable,
5730
          # and we have no other way to check)
5731
          current_mem = 0
5732
        miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
5733
                    nodeinfo[pnode].data['memory_free'])
5734
        if miss_mem > 0:
5735
          raise errors.OpPrereqError("This change will prevent the instance"
5736
                                     " from starting, due to %d MB of memory"
5737
                                     " missing on its primary node" % miss_mem)
5738

    
5739
      if be_new[constants.BE_AUTO_BALANCE]:
5740
        for node, nres in nodeinfo.iteritems():
5741
          if node not in instance.secondary_nodes:
5742
            continue
5743
          if nres.failed or not isinstance(nres.data, dict):
5744
            self.warn.append("Can't get info from secondary node %s" % node)
5745
          elif be_new[constants.BE_MEMORY] > nres.data['memory_free']:
5746
            self.warn.append("Not enough memory to failover instance to"
5747
                             " secondary node %s" % node)
5748

    
5749
    # NIC processing
5750
    for nic_op, nic_dict in self.op.nics:
5751
      if nic_op == constants.DDM_REMOVE:
5752
        if not instance.nics:
5753
          raise errors.OpPrereqError("Instance has no NICs, cannot remove")
5754
        continue
5755
      if nic_op != constants.DDM_ADD:
5756
        # an existing nic
5757
        if nic_op < 0 or nic_op >= len(instance.nics):
5758
          raise errors.OpPrereqError("Invalid NIC index %s, valid values"
5759
                                     " are 0 to %d" %
5760
                                     (nic_op, len(instance.nics)))
5761
      nic_bridge = nic_dict.get('bridge', None)
5762
      if nic_bridge is not None:
5763
        if not self.rpc.call_bridges_exist(pnode, [nic_bridge]):
5764
          msg = ("Bridge '%s' doesn't exist on one of"
5765
                 " the instance nodes" % nic_bridge)
5766
          if self.force:
5767
            self.warn.append(msg)
5768
          else:
5769
            raise errors.OpPrereqError(msg)
5770

    
5771
    # DISK processing
5772
    if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
5773
      raise errors.OpPrereqError("Disk operations not supported for"
5774
                                 " diskless instances")
5775
    for disk_op, disk_dict in self.op.disks:
5776
      if disk_op == constants.DDM_REMOVE:
5777
        if len(instance.disks) == 1:
5778
          raise errors.OpPrereqError("Cannot remove the last disk of"
5779
                                     " an instance")
5780
        ins_l = self.rpc.call_instance_list([pnode], [instance.hypervisor])
5781
        ins_l = ins_l[pnode]
5782
        if ins_l.failed or not isinstance(ins_l.data, list):
5783
          raise errors.OpPrereqError("Can't contact node '%s'" % pnode)
5784
        if instance.name in ins_l.data:
5785
          raise errors.OpPrereqError("Instance is running, can't remove"
5786
                                     " disks.")
5787

    
5788
      if (disk_op == constants.DDM_ADD and
5789
          len(instance.nics) >= constants.MAX_DISKS):
5790
        raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
5791
                                   " add more" % constants.MAX_DISKS)
5792
      if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
5793
        # an existing disk
5794
        if disk_op < 0 or disk_op >= len(instance.disks):
5795
          raise errors.OpPrereqError("Invalid disk index %s, valid values"
5796
                                     " are 0 to %d" %
5797
                                     (disk_op, len(instance.disks)))
5798

    
5799
    return
5800

    
5801
  def Exec(self, feedback_fn):
5802
    """Modifies an instance.
5803

5804
    All parameters take effect only at the next restart of the instance.
5805

5806
    """
5807
    # Process here the warnings from CheckPrereq, as we don't have a
5808
    # feedback_fn there.
5809
    for warn in self.warn:
5810
      feedback_fn("WARNING: %s" % warn)
5811

    
5812
    result = []
5813
    instance = self.instance
5814
    # disk changes
5815
    for disk_op, disk_dict in self.op.disks:
5816
      if disk_op == constants.DDM_REMOVE:
5817
        # remove the last disk
5818
        device = instance.disks.pop()
5819
        device_idx = len(instance.disks)
5820
        for node, disk in device.ComputeNodeTree(instance.primary_node):
5821
          self.cfg.SetDiskID(disk, node)
5822
          rpc_result = self.rpc.call_blockdev_remove(node, disk)
5823
          if rpc_result.failed or not rpc_result.data:
5824
            self.proc.LogWarning("Could not remove disk/%d on node %s,"
5825
                                 " continuing anyway", device_idx, node)
5826
        result.append(("disk/%d" % device_idx, "remove"))
5827
      elif disk_op == constants.DDM_ADD:
5828
        # add a new disk
5829
        if instance.disk_template == constants.DT_FILE:
5830
          file_driver, file_path = instance.disks[0].logical_id
5831
          file_path = os.path.dirname(file_path)
5832
        else:
5833
          file_driver = file_path = None
5834
        disk_idx_base = len(instance.disks)
5835
        new_disk = _GenerateDiskTemplate(self,
5836
                                         instance.disk_template,
5837
                                         instance.name, instance.primary_node,
5838
                                         instance.secondary_nodes,
5839
                                         [disk_dict],
5840
                                         file_path,
5841
                                         file_driver,
5842
                                         disk_idx_base)[0]
5843
        instance.disks.append(new_disk)
5844
        info = _GetInstanceInfoText(instance)
5845

    
5846
        logging.info("Creating volume %s for instance %s",
5847
                     new_disk.iv_name, instance.name)
5848
        # Note: this needs to be kept in sync with _CreateDisks
5849
        #HARDCODE
5850
        for node in instance.all_nodes:
5851
          f_create = node == instance.primary_node
5852
          try:
5853
            _CreateBlockDev(self, node, instance, new_disk,
5854
                            f_create, info, f_create)
5855
          except errors.OpExecError, err:
5856
            self.LogWarning("Failed to create volume %s (%s) on"
5857
                            " node %s: %s",
5858
                            new_disk.iv_name, new_disk, node, err)
5859
        result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
5860
                       (new_disk.size, new_disk.mode)))
5861
      else:
5862
        # change a given disk
5863
        instance.disks[disk_op].mode = disk_dict['mode']
5864
        result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
5865
    # NIC changes
5866
    for nic_op, nic_dict in self.op.nics:
5867
      if nic_op == constants.DDM_REMOVE:
5868
        # remove the last nic
5869
        del instance.nics[-1]
5870
        result.append(("nic.%d" % len(instance.nics), "remove"))
5871
      elif nic_op == constants.DDM_ADD:
5872
        # add a new nic
5873
        if 'mac' not in nic_dict:
5874
          mac = constants.VALUE_GENERATE
5875
        else:
5876
          mac = nic_dict['mac']
5877
        if mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
5878
          mac = self.cfg.GenerateMAC()
5879
        new_nic = objects.NIC(mac=mac, ip=nic_dict.get('ip', None),
5880
                              bridge=nic_dict.get('bridge', None))
5881
        instance.nics.append(new_nic)
5882
        result.append(("nic.%d" % (len(instance.nics) - 1),
5883
                       "add:mac=%s,ip=%s,bridge=%s" %
5884
                       (new_nic.mac, new_nic.ip, new_nic.bridge)))
5885
      else:
5886
        # change a given nic
5887
        for key in 'mac', 'ip', 'bridge':
5888
          if key in nic_dict:
5889
            setattr(instance.nics[nic_op], key, nic_dict[key])
5890
            result.append(("nic.%s/%d" % (key, nic_op), nic_dict[key]))
5891

    
5892
    # hvparams changes
5893
    if self.op.hvparams:
5894
      instance.hvparams = self.hv_new
5895
      for key, val in self.op.hvparams.iteritems():
5896
        result.append(("hv/%s" % key, val))
5897

    
5898
    # beparams changes
5899
    if self.op.beparams:
5900
      instance.beparams = self.be_inst
5901
      for key, val in self.op.beparams.iteritems():
5902
        result.append(("be/%s" % key, val))
5903

    
5904
    self.cfg.Update(instance)
5905

    
5906
    return result
5907

    
5908

    
5909
class LUQueryExports(NoHooksLU):
5910
  """Query the exports list
5911

5912
  """
5913
  _OP_REQP = ['nodes']
5914
  REQ_BGL = False
5915

    
5916
  def ExpandNames(self):
5917
    self.needed_locks = {}
5918
    self.share_locks[locking.LEVEL_NODE] = 1
5919
    if not self.op.nodes:
5920
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5921
    else:
5922
      self.needed_locks[locking.LEVEL_NODE] = \
5923
        _GetWantedNodes(self, self.op.nodes)
5924

    
5925
  def CheckPrereq(self):
5926
    """Check prerequisites.
5927

5928
    """
5929
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
5930

    
5931
  def Exec(self, feedback_fn):
5932
    """Compute the list of all the exported system images.
5933

5934
    @rtype: dict
5935
    @return: a dictionary with the structure node->(export-list)
5936
        where export-list is a list of the instances exported on
5937
        that node.
5938

5939
    """
5940
    rpcresult = self.rpc.call_export_list(self.nodes)
5941
    result = {}
5942
    for node in rpcresult:
5943
      if rpcresult[node].failed:
5944
        result[node] = False
5945
      else:
5946
        result[node] = rpcresult[node].data
5947

    
5948
    return result
5949

    
5950

    
5951
class LUExportInstance(LogicalUnit):
5952
  """Export an instance to an image in the cluster.
5953

5954
  """
5955
  HPATH = "instance-export"
5956
  HTYPE = constants.HTYPE_INSTANCE
5957
  _OP_REQP = ["instance_name", "target_node", "shutdown"]
5958
  REQ_BGL = False
5959

    
5960
  def ExpandNames(self):
5961
    self._ExpandAndLockInstance()
5962
    # FIXME: lock only instance primary and destination node
5963
    #
5964
    # Sad but true, for now we have do lock all nodes, as we don't know where
5965
    # the previous export might be, and and in this LU we search for it and
5966
    # remove it from its current node. In the future we could fix this by:
5967
    #  - making a tasklet to search (share-lock all), then create the new one,
5968
    #    then one to remove, after
5969
    #  - removing the removal operation altoghether
5970
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5971

    
5972
  def DeclareLocks(self, level):
5973
    """Last minute lock declaration."""
5974
    # All nodes are locked anyway, so nothing to do here.
5975

    
5976
  def BuildHooksEnv(self):
5977
    """Build hooks env.
5978

5979
    This will run on the master, primary node and target node.
5980

5981
    """
5982
    env = {
5983
      "EXPORT_NODE": self.op.target_node,
5984
      "EXPORT_DO_SHUTDOWN": self.op.shutdown,
5985
      }
5986
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5987
    nl = [self.cfg.GetMasterNode(), self.instance.primary_node,
5988
          self.op.target_node]
5989
    return env, nl, nl
5990

    
5991
  def CheckPrereq(self):
5992
    """Check prerequisites.
5993

5994
    This checks that the instance and node names are valid.
5995

5996
    """
5997
    instance_name = self.op.instance_name
5998
    self.instance = self.cfg.GetInstanceInfo(instance_name)
5999
    assert self.instance is not None, \
6000
          "Cannot retrieve locked instance %s" % self.op.instance_name
6001
    _CheckNodeOnline(self, self.instance.primary_node)
6002

    
6003
    self.dst_node = self.cfg.GetNodeInfo(
6004
      self.cfg.ExpandNodeName(self.op.target_node))
6005

    
6006
    if self.dst_node is None:
6007
      # This is wrong node name, not a non-locked node
6008
      raise errors.OpPrereqError("Wrong node name %s" % self.op.target_node)
6009
    _CheckNodeOnline(self, self.dst_node.name)
6010

    
6011
    # instance disk type verification
6012
    for disk in self.instance.disks:
6013
      if disk.dev_type == constants.LD_FILE:
6014
        raise errors.OpPrereqError("Export not supported for instances with"
6015
                                   " file-based disks")
6016

    
6017
  def Exec(self, feedback_fn):
6018
    """Export an instance to an image in the cluster.
6019

6020
    """
6021
    instance = self.instance
6022
    dst_node = self.dst_node
6023
    src_node = instance.primary_node
6024
    if self.op.shutdown:
6025
      # shutdown the instance, but not the disks
6026
      result = self.rpc.call_instance_shutdown(src_node, instance)
6027
      result.Raise()
6028
      if not result.data:
6029
        raise errors.OpExecError("Could not shutdown instance %s on node %s" %
6030
                                 (instance.name, src_node))
6031

    
6032
    vgname = self.cfg.GetVGName()
6033

    
6034
    snap_disks = []
6035

    
6036
    # set the disks ID correctly since call_instance_start needs the
6037
    # correct drbd minor to create the symlinks
6038
    for disk in instance.disks:
6039
      self.cfg.SetDiskID(disk, src_node)
6040

    
6041
    try:
6042
      for disk in instance.disks:
6043
        # new_dev_name will be a snapshot of an lvm leaf of the one we passed
6044
        new_dev_name = self.rpc.call_blockdev_snapshot(src_node, disk)
6045
        if new_dev_name.failed or not new_dev_name.data:
6046
          self.LogWarning("Could not snapshot block device %s on node %s",
6047
                          disk.logical_id[1], src_node)
6048
          snap_disks.append(False)
6049
        else:
6050
          new_dev = objects.Disk(dev_type=constants.LD_LV, size=disk.size,
6051
                                 logical_id=(vgname, new_dev_name.data),
6052
                                 physical_id=(vgname, new_dev_name.data),
6053
                                 iv_name=disk.iv_name)
6054
          snap_disks.append(new_dev)
6055

    
6056
    finally:
6057
      if self.op.shutdown and instance.admin_up:
6058
        result = self.rpc.call_instance_start(src_node, instance, None)
6059
        msg = result.RemoteFailMsg()
6060
        if msg:
6061
          _ShutdownInstanceDisks(self, instance)
6062
          raise errors.OpExecError("Could not start instance: %s" % msg)
6063

    
6064
    # TODO: check for size
6065

    
6066
    cluster_name = self.cfg.GetClusterName()
6067
    for idx, dev in enumerate(snap_disks):
6068
      if dev:
6069
        result = self.rpc.call_snapshot_export(src_node, dev, dst_node.name,
6070
                                               instance, cluster_name, idx)
6071
        if result.failed or not result.data:
6072
          self.LogWarning("Could not export block device %s from node %s to"
6073
                          " node %s", dev.logical_id[1], src_node,
6074
                          dst_node.name)
6075
        result = self.rpc.call_blockdev_remove(src_node, dev)
6076
        if result.failed or not result.data:
6077
          self.LogWarning("Could not remove snapshot block device %s from node"
6078
                          " %s", dev.logical_id[1], src_node)
6079

    
6080
    result = self.rpc.call_finalize_export(dst_node.name, instance, snap_disks)
6081
    if result.failed or not result.data:
6082
      self.LogWarning("Could not finalize export for instance %s on node %s",
6083
                      instance.name, dst_node.name)
6084

    
6085
    nodelist = self.cfg.GetNodeList()
6086
    nodelist.remove(dst_node.name)
6087

    
6088
    # on one-node clusters nodelist will be empty after the removal
6089
    # if we proceed the backup would be removed because OpQueryExports
6090
    # substitutes an empty list with the full cluster node list.
6091
    if nodelist:
6092
      exportlist = self.rpc.call_export_list(nodelist)
6093
      for node in exportlist:
6094
        if exportlist[node].failed:
6095
          continue
6096
        if instance.name in exportlist[node].data:
6097
          if not self.rpc.call_export_remove(node, instance.name):
6098
            self.LogWarning("Could not remove older export for instance %s"
6099
                            " on node %s", instance.name, node)
6100

    
6101

    
6102
class LURemoveExport(NoHooksLU):
6103
  """Remove exports related to the named instance.
6104

6105
  """
6106
  _OP_REQP = ["instance_name"]
6107
  REQ_BGL = False
6108

    
6109
  def ExpandNames(self):
6110
    self.needed_locks = {}
6111
    # We need all nodes to be locked in order for RemoveExport to work, but we
6112
    # don't need to lock the instance itself, as nothing will happen to it (and
6113
    # we can remove exports also for a removed instance)
6114
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6115

    
6116
  def CheckPrereq(self):
6117
    """Check prerequisites.
6118
    """
6119
    pass
6120

    
6121
  def Exec(self, feedback_fn):
6122
    """Remove any export.
6123

6124
    """
6125
    instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
6126
    # If the instance was not found we'll try with the name that was passed in.
6127
    # This will only work if it was an FQDN, though.
6128
    fqdn_warn = False
6129
    if not instance_name:
6130
      fqdn_warn = True
6131
      instance_name = self.op.instance_name
6132

    
6133
    exportlist = self.rpc.call_export_list(self.acquired_locks[
6134
      locking.LEVEL_NODE])
6135
    found = False
6136
    for node in exportlist:
6137
      if exportlist[node].failed:
6138
        self.LogWarning("Failed to query node %s, continuing" % node)
6139
        continue
6140
      if instance_name in exportlist[node].data:
6141
        found = True
6142
        result = self.rpc.call_export_remove(node, instance_name)
6143
        if result.failed or not result.data:
6144
          logging.error("Could not remove export for instance %s"
6145
                        " on node %s", instance_name, node)
6146

    
6147
    if fqdn_warn and not found:
6148
      feedback_fn("Export not found. If trying to remove an export belonging"
6149
                  " to a deleted instance please use its Fully Qualified"
6150
                  " Domain Name.")
6151

    
6152

    
6153
class TagsLU(NoHooksLU):
6154
  """Generic tags LU.
6155

6156
  This is an abstract class which is the parent of all the other tags LUs.
6157

6158
  """
6159

    
6160
  def ExpandNames(self):
6161
    self.needed_locks = {}
6162
    if self.op.kind == constants.TAG_NODE:
6163
      name = self.cfg.ExpandNodeName(self.op.name)
6164
      if name is None:
6165
        raise errors.OpPrereqError("Invalid node name (%s)" %
6166
                                   (self.op.name,))
6167
      self.op.name = name
6168
      self.needed_locks[locking.LEVEL_NODE] = name
6169
    elif self.op.kind == constants.TAG_INSTANCE:
6170
      name = self.cfg.ExpandInstanceName(self.op.name)
6171
      if name is None:
6172
        raise errors.OpPrereqError("Invalid instance name (%s)" %
6173
                                   (self.op.name,))
6174
      self.op.name = name
6175
      self.needed_locks[locking.LEVEL_INSTANCE] = name
6176

    
6177
  def CheckPrereq(self):
6178
    """Check prerequisites.
6179

6180
    """
6181
    if self.op.kind == constants.TAG_CLUSTER:
6182
      self.target = self.cfg.GetClusterInfo()
6183
    elif self.op.kind == constants.TAG_NODE:
6184
      self.target = self.cfg.GetNodeInfo(self.op.name)
6185
    elif self.op.kind == constants.TAG_INSTANCE:
6186
      self.target = self.cfg.GetInstanceInfo(self.op.name)
6187
    else:
6188
      raise errors.OpPrereqError("Wrong tag type requested (%s)" %
6189
                                 str(self.op.kind))
6190

    
6191

    
6192
class LUGetTags(TagsLU):
6193
  """Returns the tags of a given object.
6194

6195
  """
6196
  _OP_REQP = ["kind", "name"]
6197
  REQ_BGL = False
6198

    
6199
  def Exec(self, feedback_fn):
6200
    """Returns the tag list.
6201

6202
    """
6203
    return list(self.target.GetTags())
6204

    
6205

    
6206
class LUSearchTags(NoHooksLU):
6207
  """Searches the tags for a given pattern.
6208

6209
  """
6210
  _OP_REQP = ["pattern"]
6211
  REQ_BGL = False
6212

    
6213
  def ExpandNames(self):
6214
    self.needed_locks = {}
6215

    
6216
  def CheckPrereq(self):
6217
    """Check prerequisites.
6218

6219
    This checks the pattern passed for validity by compiling it.
6220

6221
    """
6222
    try:
6223
      self.re = re.compile(self.op.pattern)
6224
    except re.error, err:
6225
      raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
6226
                                 (self.op.pattern, err))
6227

    
6228
  def Exec(self, feedback_fn):
6229
    """Returns the tag list.
6230

6231
    """
6232
    cfg = self.cfg
6233
    tgts = [("/cluster", cfg.GetClusterInfo())]
6234
    ilist = cfg.GetAllInstancesInfo().values()
6235
    tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
6236
    nlist = cfg.GetAllNodesInfo().values()
6237
    tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
6238
    results = []
6239
    for path, target in tgts:
6240
      for tag in target.GetTags():
6241
        if self.re.search(tag):
6242
          results.append((path, tag))
6243
    return results
6244

    
6245

    
6246
class LUAddTags(TagsLU):
6247
  """Sets a tag on a given object.
6248

6249
  """
6250
  _OP_REQP = ["kind", "name", "tags"]
6251
  REQ_BGL = False
6252

    
6253
  def CheckPrereq(self):
6254
    """Check prerequisites.
6255

6256
    This checks the type and length of the tag name and value.
6257

6258
    """
6259
    TagsLU.CheckPrereq(self)
6260
    for tag in self.op.tags:
6261
      objects.TaggableObject.ValidateTag(tag)
6262

    
6263
  def Exec(self, feedback_fn):
6264
    """Sets the tag.
6265

6266
    """
6267
    try:
6268
      for tag in self.op.tags:
6269
        self.target.AddTag(tag)
6270
    except errors.TagError, err:
6271
      raise errors.OpExecError("Error while setting tag: %s" % str(err))
6272
    try:
6273
      self.cfg.Update(self.target)
6274
    except errors.ConfigurationError:
6275
      raise errors.OpRetryError("There has been a modification to the"
6276
                                " config file and the operation has been"
6277
                                " aborted. Please retry.")
6278

    
6279

    
6280
class LUDelTags(TagsLU):
6281
  """Delete a list of tags from a given object.
6282

6283
  """
6284
  _OP_REQP = ["kind", "name", "tags"]
6285
  REQ_BGL = False
6286

    
6287
  def CheckPrereq(self):
6288
    """Check prerequisites.
6289

6290
    This checks that we have the given tag.
6291

6292
    """
6293
    TagsLU.CheckPrereq(self)
6294
    for tag in self.op.tags:
6295
      objects.TaggableObject.ValidateTag(tag)
6296
    del_tags = frozenset(self.op.tags)
6297
    cur_tags = self.target.GetTags()
6298
    if not del_tags <= cur_tags:
6299
      diff_tags = del_tags - cur_tags
6300
      diff_names = ["'%s'" % tag for tag in diff_tags]
6301
      diff_names.sort()
6302
      raise errors.OpPrereqError("Tag(s) %s not found" %
6303
                                 (",".join(diff_names)))
6304

    
6305
  def Exec(self, feedback_fn):
6306
    """Remove the tag from the object.
6307

6308
    """
6309
    for tag in self.op.tags:
6310
      self.target.RemoveTag(tag)
6311
    try:
6312
      self.cfg.Update(self.target)
6313
    except errors.ConfigurationError:
6314
      raise errors.OpRetryError("There has been a modification to the"
6315
                                " config file and the operation has been"
6316
                                " aborted. Please retry.")
6317

    
6318

    
6319
class LUTestDelay(NoHooksLU):
6320
  """Sleep for a specified amount of time.
6321

6322
  This LU sleeps on the master and/or nodes for a specified amount of
6323
  time.
6324

6325
  """
6326
  _OP_REQP = ["duration", "on_master", "on_nodes"]
6327
  REQ_BGL = False
6328

    
6329
  def ExpandNames(self):
6330
    """Expand names and set required locks.
6331

6332
    This expands the node list, if any.
6333

6334
    """
6335
    self.needed_locks = {}
6336
    if self.op.on_nodes:
6337
      # _GetWantedNodes can be used here, but is not always appropriate to use
6338
      # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
6339
      # more information.
6340
      self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
6341
      self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
6342

    
6343
  def CheckPrereq(self):
6344
    """Check prerequisites.
6345

6346
    """
6347

    
6348
  def Exec(self, feedback_fn):
6349
    """Do the actual sleep.
6350

6351
    """
6352
    if self.op.on_master:
6353
      if not utils.TestDelay(self.op.duration):
6354
        raise errors.OpExecError("Error during master delay test")
6355
    if self.op.on_nodes:
6356
      result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
6357
      if not result:
6358
        raise errors.OpExecError("Complete failure from rpc call")
6359
      for node, node_result in result.items():
6360
        node_result.Raise()
6361
        if not node_result.data:
6362
          raise errors.OpExecError("Failure during rpc call to node %s,"
6363
                                   " result: %s" % (node, node_result.data))
6364

    
6365

    
6366
class IAllocator(object):
6367
  """IAllocator framework.
6368

6369
  An IAllocator instance has three sets of attributes:
6370
    - cfg that is needed to query the cluster
6371
    - input data (all members of the _KEYS class attribute are required)
6372
    - four buffer attributes (in|out_data|text), that represent the
6373
      input (to the external script) in text and data structure format,
6374
      and the output from it, again in two formats
6375
    - the result variables from the script (success, info, nodes) for
6376
      easy usage
6377

6378
  """
6379
  _ALLO_KEYS = [
6380
    "mem_size", "disks", "disk_template",
6381
    "os", "tags", "nics", "vcpus", "hypervisor",
6382
    ]
6383
  _RELO_KEYS = [
6384
    "relocate_from",
6385
    ]
6386

    
6387
  def __init__(self, lu, mode, name, **kwargs):
6388
    self.lu = lu
6389
    # init buffer variables
6390
    self.in_text = self.out_text = self.in_data = self.out_data = None
6391
    # init all input fields so that pylint is happy
6392
    self.mode = mode
6393
    self.name = name
6394
    self.mem_size = self.disks = self.disk_template = None
6395
    self.os = self.tags = self.nics = self.vcpus = None
6396
    self.hypervisor = None
6397
    self.relocate_from = None
6398
    # computed fields
6399
    self.required_nodes = None
6400
    # init result fields
6401
    self.success = self.info = self.nodes = None
6402
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6403
      keyset = self._ALLO_KEYS
6404
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
6405
      keyset = self._RELO_KEYS
6406
    else:
6407
      raise errors.ProgrammerError("Unknown mode '%s' passed to the"
6408
                                   " IAllocator" % self.mode)
6409
    for key in kwargs:
6410
      if key not in keyset:
6411
        raise errors.ProgrammerError("Invalid input parameter '%s' to"
6412
                                     " IAllocator" % key)
6413
      setattr(self, key, kwargs[key])
6414
    for key in keyset:
6415
      if key not in kwargs:
6416
        raise errors.ProgrammerError("Missing input parameter '%s' to"
6417
                                     " IAllocator" % key)
6418
    self._BuildInputData()
6419

    
6420
  def _ComputeClusterData(self):
6421
    """Compute the generic allocator input data.
6422

6423
    This is the data that is independent of the actual operation.
6424

6425
    """
6426
    cfg = self.lu.cfg
6427
    cluster_info = cfg.GetClusterInfo()
6428
    # cluster data
6429
    data = {
6430
      "version": 1,
6431
      "cluster_name": cfg.GetClusterName(),
6432
      "cluster_tags": list(cluster_info.GetTags()),
6433
      "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
6434
      # we don't have job IDs
6435
      }
6436
    iinfo = cfg.GetAllInstancesInfo().values()
6437
    i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
6438

    
6439
    # node data
6440
    node_results = {}
6441
    node_list = cfg.GetNodeList()
6442

    
6443
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6444
      hypervisor_name = self.hypervisor
6445
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
6446
      hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
6447

    
6448
    node_data = self.lu.rpc.call_node_info(node_list, cfg.GetVGName(),
6449
                                           hypervisor_name)
6450
    node_iinfo = self.lu.rpc.call_all_instances_info(node_list,
6451
                       cluster_info.enabled_hypervisors)
6452
    for nname, nresult in node_data.items():
6453
      # first fill in static (config-based) values
6454
      ninfo = cfg.GetNodeInfo(nname)
6455
      pnr = {
6456
        "tags": list(ninfo.GetTags()),
6457
        "primary_ip": ninfo.primary_ip,
6458
        "secondary_ip": ninfo.secondary_ip,
6459
        "offline": ninfo.offline,
6460
        "master_candidate": ninfo.master_candidate,
6461
        }
6462

    
6463
      if not ninfo.offline:
6464
        nresult.Raise()
6465
        if not isinstance(nresult.data, dict):
6466
          raise errors.OpExecError("Can't get data for node %s" % nname)
6467
        remote_info = nresult.data
6468
        for attr in ['memory_total', 'memory_free', 'memory_dom0',
6469
                     'vg_size', 'vg_free', 'cpu_total']:
6470
          if attr not in remote_info:
6471
            raise errors.OpExecError("Node '%s' didn't return attribute"
6472
                                     " '%s'" % (nname, attr))
6473
          try:
6474
            remote_info[attr] = int(remote_info[attr])
6475
          except ValueError, err:
6476
            raise errors.OpExecError("Node '%s' returned invalid value"
6477
                                     " for '%s': %s" % (nname, attr, err))
6478
        # compute memory used by primary instances
6479
        i_p_mem = i_p_up_mem = 0
6480
        for iinfo, beinfo in i_list:
6481
          if iinfo.primary_node == nname:
6482
            i_p_mem += beinfo[constants.BE_MEMORY]
6483
            if iinfo.name not in node_iinfo[nname].data:
6484
              i_used_mem = 0
6485
            else:
6486
              i_used_mem = int(node_iinfo[nname].data[iinfo.name]['memory'])
6487
            i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
6488
            remote_info['memory_free'] -= max(0, i_mem_diff)
6489

    
6490
            if iinfo.admin_up:
6491
              i_p_up_mem += beinfo[constants.BE_MEMORY]
6492

    
6493
        # compute memory used by instances
6494
        pnr_dyn = {
6495
          "total_memory": remote_info['memory_total'],
6496
          "reserved_memory": remote_info['memory_dom0'],
6497
          "free_memory": remote_info['memory_free'],
6498
          "total_disk": remote_info['vg_size'],
6499
          "free_disk": remote_info['vg_free'],
6500
          "total_cpus": remote_info['cpu_total'],
6501
          "i_pri_memory": i_p_mem,
6502
          "i_pri_up_memory": i_p_up_mem,
6503
          }
6504
        pnr.update(pnr_dyn)
6505

    
6506
      node_results[nname] = pnr
6507
    data["nodes"] = node_results
6508

    
6509
    # instance data
6510
    instance_data = {}
6511
    for iinfo, beinfo in i_list:
6512
      nic_data = [{"mac": n.mac, "ip": n.ip, "bridge": n.bridge}
6513
                  for n in iinfo.nics]
6514
      pir = {
6515
        "tags": list(iinfo.GetTags()),
6516
        "admin_up": iinfo.admin_up,
6517
        "vcpus": beinfo[constants.BE_VCPUS],
6518
        "memory": beinfo[constants.BE_MEMORY],
6519
        "os": iinfo.os,
6520
        "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
6521
        "nics": nic_data,
6522
        "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
6523
        "disk_template": iinfo.disk_template,
6524
        "hypervisor": iinfo.hypervisor,
6525
        }
6526
      instance_data[iinfo.name] = pir
6527

    
6528
    data["instances"] = instance_data
6529

    
6530
    self.in_data = data
6531

    
6532
  def _AddNewInstance(self):
6533
    """Add new instance data to allocator structure.
6534

6535
    This in combination with _AllocatorGetClusterData will create the
6536
    correct structure needed as input for the allocator.
6537

6538
    The checks for the completeness of the opcode must have already been
6539
    done.
6540

6541
    """
6542
    data = self.in_data
6543

    
6544
    disk_space = _ComputeDiskSize(self.disk_template, self.disks)
6545

    
6546
    if self.disk_template in constants.DTS_NET_MIRROR:
6547
      self.required_nodes = 2
6548
    else:
6549
      self.required_nodes = 1
6550
    request = {
6551
      "type": "allocate",
6552
      "name": self.name,
6553
      "disk_template": self.disk_template,
6554
      "tags": self.tags,
6555
      "os": self.os,
6556
      "vcpus": self.vcpus,
6557
      "memory": self.mem_size,
6558
      "disks": self.disks,
6559
      "disk_space_total": disk_space,
6560
      "nics": self.nics,
6561
      "required_nodes": self.required_nodes,
6562
      }
6563
    data["request"] = request
6564

    
6565
  def _AddRelocateInstance(self):
6566
    """Add relocate instance data to allocator structure.
6567

6568
    This in combination with _IAllocatorGetClusterData will create the
6569
    correct structure needed as input for the allocator.
6570

6571
    The checks for the completeness of the opcode must have already been
6572
    done.
6573

6574
    """
6575
    instance = self.lu.cfg.GetInstanceInfo(self.name)
6576
    if instance is None:
6577
      raise errors.ProgrammerError("Unknown instance '%s' passed to"
6578
                                   " IAllocator" % self.name)
6579

    
6580
    if instance.disk_template not in constants.DTS_NET_MIRROR:
6581
      raise errors.OpPrereqError("Can't relocate non-mirrored instances")
6582

    
6583
    if len(instance.secondary_nodes) != 1:
6584
      raise errors.OpPrereqError("Instance has not exactly one secondary node")
6585

    
6586
    self.required_nodes = 1
6587
    disk_sizes = [{'size': disk.size} for disk in instance.disks]
6588
    disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
6589

    
6590
    request = {
6591
      "type": "relocate",
6592
      "name": self.name,
6593
      "disk_space_total": disk_space,
6594
      "required_nodes": self.required_nodes,
6595
      "relocate_from": self.relocate_from,
6596
      }
6597
    self.in_data["request"] = request
6598

    
6599
  def _BuildInputData(self):
6600
    """Build input data structures.
6601

6602
    """
6603
    self._ComputeClusterData()
6604

    
6605
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6606
      self._AddNewInstance()
6607
    else:
6608
      self._AddRelocateInstance()
6609

    
6610
    self.in_text = serializer.Dump(self.in_data)
6611

    
6612
  def Run(self, name, validate=True, call_fn=None):
6613
    """Run an instance allocator and return the results.
6614

6615
    """
6616
    if call_fn is None:
6617
      call_fn = self.lu.rpc.call_iallocator_runner
6618
    data = self.in_text
6619

    
6620
    result = call_fn(self.lu.cfg.GetMasterNode(), name, self.in_text)
6621
    result.Raise()
6622

    
6623
    if not isinstance(result.data, (list, tuple)) or len(result.data) != 4:
6624
      raise errors.OpExecError("Invalid result from master iallocator runner")
6625

    
6626
    rcode, stdout, stderr, fail = result.data
6627

    
6628
    if rcode == constants.IARUN_NOTFOUND:
6629
      raise errors.OpExecError("Can't find allocator '%s'" % name)
6630
    elif rcode == constants.IARUN_FAILURE:
6631
      raise errors.OpExecError("Instance allocator call failed: %s,"
6632
                               " output: %s" % (fail, stdout+stderr))
6633
    self.out_text = stdout
6634
    if validate:
6635
      self._ValidateResult()
6636

    
6637
  def _ValidateResult(self):
6638
    """Process the allocator results.
6639

6640
    This will process and if successful save the result in
6641
    self.out_data and the other parameters.
6642

6643
    """
6644
    try:
6645
      rdict = serializer.Load(self.out_text)
6646
    except Exception, err:
6647
      raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
6648

    
6649
    if not isinstance(rdict, dict):
6650
      raise errors.OpExecError("Can't parse iallocator results: not a dict")
6651

    
6652
    for key in "success", "info", "nodes":
6653
      if key not in rdict:
6654
        raise errors.OpExecError("Can't parse iallocator results:"
6655
                                 " missing key '%s'" % key)
6656
      setattr(self, key, rdict[key])
6657

    
6658
    if not isinstance(rdict["nodes"], list):
6659
      raise errors.OpExecError("Can't parse iallocator results: 'nodes' key"
6660
                               " is not a list")
6661
    self.out_data = rdict
6662

    
6663

    
6664
class LUTestAllocator(NoHooksLU):
6665
  """Run allocator tests.
6666

6667
  This LU runs the allocator tests
6668

6669
  """
6670
  _OP_REQP = ["direction", "mode", "name"]
6671

    
6672
  def CheckPrereq(self):
6673
    """Check prerequisites.
6674

6675
    This checks the opcode parameters depending on the director and mode test.
6676

6677
    """
6678
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
6679
      for attr in ["name", "mem_size", "disks", "disk_template",
6680
                   "os", "tags", "nics", "vcpus"]:
6681
        if not hasattr(self.op, attr):
6682
          raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
6683
                                     attr)
6684
      iname = self.cfg.ExpandInstanceName(self.op.name)
6685
      if iname is not None:
6686
        raise errors.OpPrereqError("Instance '%s' already in the cluster" %
6687
                                   iname)
6688
      if not isinstance(self.op.nics, list):
6689
        raise errors.OpPrereqError("Invalid parameter 'nics'")
6690
      for row in self.op.nics:
6691
        if (not isinstance(row, dict) or
6692
            "mac" not in row or
6693
            "ip" not in row or
6694
            "bridge" not in row):
6695
          raise errors.OpPrereqError("Invalid contents of the"
6696
                                     " 'nics' parameter")
6697
      if not isinstance(self.op.disks, list):
6698
        raise errors.OpPrereqError("Invalid parameter 'disks'")
6699
      for row in self.op.disks:
6700
        if (not isinstance(row, dict) or
6701
            "size" not in row or
6702
            not isinstance(row["size"], int) or
6703
            "mode" not in row or
6704
            row["mode"] not in ['r', 'w']):
6705
          raise errors.OpPrereqError("Invalid contents of the"
6706
                                     " 'disks' parameter")
6707
      if not hasattr(self.op, "hypervisor") or self.op.hypervisor is None:
6708
        self.op.hypervisor = self.cfg.GetHypervisorType()
6709
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
6710
      if not hasattr(self.op, "name"):
6711
        raise errors.OpPrereqError("Missing attribute 'name' on opcode input")
6712
      fname = self.cfg.ExpandInstanceName(self.op.name)
6713
      if fname is None:
6714
        raise errors.OpPrereqError("Instance '%s' not found for relocation" %
6715
                                   self.op.name)
6716
      self.op.name = fname
6717
      self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
6718
    else:
6719
      raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
6720
                                 self.op.mode)
6721

    
6722
    if self.op.direction == constants.IALLOCATOR_DIR_OUT:
6723
      if not hasattr(self.op, "allocator") or self.op.allocator is None:
6724
        raise errors.OpPrereqError("Missing allocator name")
6725
    elif self.op.direction != constants.IALLOCATOR_DIR_IN:
6726
      raise errors.OpPrereqError("Wrong allocator test '%s'" %
6727
                                 self.op.direction)
6728

    
6729
  def Exec(self, feedback_fn):
6730
    """Run the allocator test.
6731

6732
    """
6733
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
6734
      ial = IAllocator(self,
6735
                       mode=self.op.mode,
6736
                       name=self.op.name,
6737
                       mem_size=self.op.mem_size,
6738
                       disks=self.op.disks,
6739
                       disk_template=self.op.disk_template,
6740
                       os=self.op.os,
6741
                       tags=self.op.tags,
6742
                       nics=self.op.nics,
6743
                       vcpus=self.op.vcpus,
6744
                       hypervisor=self.op.hypervisor,
6745
                       )
6746
    else:
6747
      ial = IAllocator(self,
6748
                       mode=self.op.mode,
6749
                       name=self.op.name,
6750
                       relocate_from=list(self.relocate_from),
6751
                       )
6752

    
6753
    if self.op.direction == constants.IALLOCATOR_DIR_IN:
6754
      result = ial.in_text
6755
    else:
6756
      ial.Run(self.op.allocator, validate=False)
6757
      result = ial.out_text
6758
    return result