Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ 9ddb5e45

History | View | Annotate | Download (214.8 kB)

1
#
2
#
3

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

    
21

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

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

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

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

    
48

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

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

61
  Note that all commands require root permissions.
62

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

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

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

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

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

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

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

    
109
  ssh = property(fget=__GetSSH)
110

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

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

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

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

126
    """
127
    pass
128

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

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

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

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

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

150
    Examples::
151

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

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

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

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

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

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

189
    """
190

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

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

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

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

205
    """
206
    raise NotImplementedError
207

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

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

215
    """
216
    raise NotImplementedError
217

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

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

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

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

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

237
    """
238
    raise NotImplementedError
239

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

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

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

258
    """
259
    return lu_result
260

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
326
    del self.recalculate_locks[locking.LEVEL_NODE]
327

    
328

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

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

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

    
339

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

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

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

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

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

    
366
  return utils.NiceSort(wanted)
367

    
368

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

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

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

    
385
  if instances:
386
    wanted = []
387

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

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

    
398

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

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

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

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

    
417

    
418
def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
419
                          memory, vcpus, nics):
420
  """Builds instance related env variables for hooks
421

422
  This builds the hook environment from individual variables.
423

424
  @type name: string
425
  @param name: the name of the instance
426
  @type primary_node: string
427
  @param primary_node: the name of the instance's primary node
428
  @type secondary_nodes: list
429
  @param secondary_nodes: list of secondary nodes as strings
430
  @type os_type: string
431
  @param os_type: the name of the instance's OS
432
  @type status: string
433
  @param status: the desired status of the instances
434
  @type memory: string
435
  @param memory: the memory size of the instance
436
  @type vcpus: string
437
  @param vcpus: the count of VCPUs the instance has
438
  @type nics: list
439
  @param nics: list of tuples (ip, bridge, mac) representing
440
      the NICs the instance  has
441
  @rtype: dict
442
  @return: the hook environment for this instance
443

444
  """
445
  env = {
446
    "OP_TARGET": name,
447
    "INSTANCE_NAME": name,
448
    "INSTANCE_PRIMARY": primary_node,
449
    "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
450
    "INSTANCE_OS_TYPE": os_type,
451
    "INSTANCE_STATUS": status,
452
    "INSTANCE_MEMORY": memory,
453
    "INSTANCE_VCPUS": vcpus,
454
  }
455

    
456
  if nics:
457
    nic_count = len(nics)
458
    for idx, (ip, bridge, mac) in enumerate(nics):
459
      if ip is None:
460
        ip = ""
461
      env["INSTANCE_NIC%d_IP" % idx] = ip
462
      env["INSTANCE_NIC%d_BRIDGE" % idx] = bridge
463
      env["INSTANCE_NIC%d_HWADDR" % idx] = mac
464
  else:
465
    nic_count = 0
466

    
467
  env["INSTANCE_NIC_COUNT"] = nic_count
468

    
469
  return env
470

    
471

    
472
def _BuildInstanceHookEnvByObject(lu, instance, override=None):
473
  """Builds instance related env variables for hooks from an object.
474

475
  @type lu: L{LogicalUnit}
476
  @param lu: the logical unit on whose behalf we execute
477
  @type instance: L{objects.Instance}
478
  @param instance: the instance for which we should build the
479
      environment
480
  @type override: dict
481
  @param override: dictionary with key/values that will override
482
      our values
483
  @rtype: dict
484
  @return: the hook environment dictionary
485

486
  """
487
  bep = lu.cfg.GetClusterInfo().FillBE(instance)
488
  args = {
489
    'name': instance.name,
490
    'primary_node': instance.primary_node,
491
    'secondary_nodes': instance.secondary_nodes,
492
    'os_type': instance.os,
493
    'status': instance.os,
494
    'memory': bep[constants.BE_MEMORY],
495
    'vcpus': bep[constants.BE_VCPUS],
496
    'nics': [(nic.ip, nic.bridge, nic.mac) for nic in instance.nics],
497
  }
498
  if override:
499
    args.update(override)
500
  return _BuildInstanceHookEnv(**args)
501

    
502

    
503
def _CheckInstanceBridgesExist(lu, instance):
504
  """Check that the brigdes needed by an instance exist.
505

506
  """
507
  # check bridges existance
508
  brlist = [nic.bridge for nic in instance.nics]
509
  result = lu.rpc.call_bridges_exist(instance.primary_node, brlist)
510
  result.Raise()
511
  if not result.data:
512
    raise errors.OpPrereqError("One or more target bridges %s does not"
513
                               " exist on destination node '%s'" %
514
                               (brlist, instance.primary_node))
515

    
516

    
517
class LUDestroyCluster(NoHooksLU):
518
  """Logical unit for destroying the cluster.
519

520
  """
521
  _OP_REQP = []
522

    
523
  def CheckPrereq(self):
524
    """Check prerequisites.
525

526
    This checks whether the cluster is empty.
527

528
    Any errors are signalled by raising errors.OpPrereqError.
529

530
    """
531
    master = self.cfg.GetMasterNode()
532

    
533
    nodelist = self.cfg.GetNodeList()
534
    if len(nodelist) != 1 or nodelist[0] != master:
535
      raise errors.OpPrereqError("There are still %d node(s) in"
536
                                 " this cluster." % (len(nodelist) - 1))
537
    instancelist = self.cfg.GetInstanceList()
538
    if instancelist:
539
      raise errors.OpPrereqError("There are still %d instance(s) in"
540
                                 " this cluster." % len(instancelist))
541

    
542
  def Exec(self, feedback_fn):
543
    """Destroys the cluster.
544

545
    """
546
    master = self.cfg.GetMasterNode()
547
    result = self.rpc.call_node_stop_master(master, False)
548
    result.Raise()
549
    if not result.data:
550
      raise errors.OpExecError("Could not disable the master role")
551
    priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
552
    utils.CreateBackup(priv_key)
553
    utils.CreateBackup(pub_key)
554
    return master
555

    
556

    
557
class LUVerifyCluster(LogicalUnit):
558
  """Verifies the cluster status.
559

560
  """
561
  HPATH = "cluster-verify"
562
  HTYPE = constants.HTYPE_CLUSTER
563
  _OP_REQP = ["skip_checks"]
564
  REQ_BGL = False
565

    
566
  def ExpandNames(self):
567
    self.needed_locks = {
568
      locking.LEVEL_NODE: locking.ALL_SET,
569
      locking.LEVEL_INSTANCE: locking.ALL_SET,
570
    }
571
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
572

    
573
  def _VerifyNode(self, nodeinfo, file_list, local_cksum,
574
                  node_result, feedback_fn, master_files):
575
    """Run multiple tests against a node.
576

577
    Test list:
578

579
      - compares ganeti version
580
      - checks vg existance and size > 20G
581
      - checks config file checksum
582
      - checks ssh to other nodes
583

584
    @type nodeinfo: L{objects.Node}
585
    @param nodeinfo: the node to check
586
    @param file_list: required list of files
587
    @param local_cksum: dictionary of local files and their checksums
588
    @param node_result: the results from the node
589
    @param feedback_fn: function used to accumulate results
590
    @param master_files: list of files that only masters should have
591

592
    """
593
    node = nodeinfo.name
594

    
595
    # main result, node_result should be a non-empty dict
596
    if not node_result or not isinstance(node_result, dict):
597
      feedback_fn("  - ERROR: unable to verify node %s." % (node,))
598
      return True
599

    
600
    # compares ganeti version
601
    local_version = constants.PROTOCOL_VERSION
602
    remote_version = node_result.get('version', None)
603
    if not remote_version:
604
      feedback_fn("  - ERROR: connection to %s failed" % (node))
605
      return True
606

    
607
    if local_version != remote_version:
608
      feedback_fn("  - ERROR: sw version mismatch: master %s, node(%s) %s" %
609
                      (local_version, node, remote_version))
610
      return True
611

    
612
    # checks vg existance and size > 20G
613

    
614
    bad = False
615
    vglist = node_result.get(constants.NV_VGLIST, None)
616
    if not vglist:
617
      feedback_fn("  - ERROR: unable to check volume groups on node %s." %
618
                      (node,))
619
      bad = True
620
    else:
621
      vgstatus = utils.CheckVolumeGroupSize(vglist, self.cfg.GetVGName(),
622
                                            constants.MIN_VG_SIZE)
623
      if vgstatus:
624
        feedback_fn("  - ERROR: %s on node %s" % (vgstatus, node))
625
        bad = True
626

    
627
    # checks config file checksum
628

    
629
    remote_cksum = node_result.get(constants.NV_FILELIST, None)
630
    if not isinstance(remote_cksum, dict):
631
      bad = True
632
      feedback_fn("  - ERROR: node hasn't returned file checksum data")
633
    else:
634
      for file_name in file_list:
635
        node_is_mc = nodeinfo.master_candidate
636
        must_have_file = file_name not in master_files
637
        if file_name not in remote_cksum:
638
          if node_is_mc or must_have_file:
639
            bad = True
640
            feedback_fn("  - ERROR: file '%s' missing" % file_name)
641
        elif remote_cksum[file_name] != local_cksum[file_name]:
642
          if node_is_mc or must_have_file:
643
            bad = True
644
            feedback_fn("  - ERROR: file '%s' has wrong checksum" % file_name)
645
          else:
646
            # not candidate and this is not a must-have file
647
            bad = True
648
            feedback_fn("  - ERROR: non master-candidate has old/wrong file"
649
                        " '%s'" % file_name)
650
        else:
651
          # all good, except non-master/non-must have combination
652
          if not node_is_mc and not must_have_file:
653
            feedback_fn("  - ERROR: file '%s' should not exist on non master"
654
                        " candidates" % file_name)
655

    
656
    # checks ssh to any
657

    
658
    if constants.NV_NODELIST not in node_result:
659
      bad = True
660
      feedback_fn("  - ERROR: node hasn't returned node ssh connectivity data")
661
    else:
662
      if node_result[constants.NV_NODELIST]:
663
        bad = True
664
        for node in node_result[constants.NV_NODELIST]:
665
          feedback_fn("  - ERROR: ssh communication with node '%s': %s" %
666
                          (node, node_result[constants.NV_NODELIST][node]))
667

    
668
    if constants.NV_NODENETTEST not in node_result:
669
      bad = True
670
      feedback_fn("  - ERROR: node hasn't returned node tcp connectivity data")
671
    else:
672
      if node_result[constants.NV_NODENETTEST]:
673
        bad = True
674
        nlist = utils.NiceSort(node_result[constants.NV_NODENETTEST].keys())
675
        for node in nlist:
676
          feedback_fn("  - ERROR: tcp communication with node '%s': %s" %
677
                          (node, node_result[constants.NV_NODENETTEST][node]))
678

    
679
    hyp_result = node_result.get(constants.NV_HYPERVISOR, None)
680
    if isinstance(hyp_result, dict):
681
      for hv_name, hv_result in hyp_result.iteritems():
682
        if hv_result is not None:
683
          feedback_fn("  - ERROR: hypervisor %s verify failure: '%s'" %
684
                      (hv_name, hv_result))
685
    return bad
686

    
687
  def _VerifyInstance(self, instance, instanceconfig, node_vol_is,
688
                      node_instance, feedback_fn):
689
    """Verify an instance.
690

691
    This function checks to see if the required block devices are
692
    available on the instance's node.
693

694
    """
695
    bad = False
696

    
697
    node_current = instanceconfig.primary_node
698

    
699
    node_vol_should = {}
700
    instanceconfig.MapLVsByNode(node_vol_should)
701

    
702
    for node in node_vol_should:
703
      for volume in node_vol_should[node]:
704
        if node not in node_vol_is or volume not in node_vol_is[node]:
705
          feedback_fn("  - ERROR: volume %s missing on node %s" %
706
                          (volume, node))
707
          bad = True
708

    
709
    if not instanceconfig.status == 'down':
710
      if (node_current not in node_instance or
711
          not instance in node_instance[node_current]):
712
        feedback_fn("  - ERROR: instance %s not running on node %s" %
713
                        (instance, node_current))
714
        bad = True
715

    
716
    for node in node_instance:
717
      if (not node == node_current):
718
        if instance in node_instance[node]:
719
          feedback_fn("  - ERROR: instance %s should not run on node %s" %
720
                          (instance, node))
721
          bad = True
722

    
723
    return bad
724

    
725
  def _VerifyOrphanVolumes(self, node_vol_should, node_vol_is, feedback_fn):
726
    """Verify if there are any unknown volumes in the cluster.
727

728
    The .os, .swap and backup volumes are ignored. All other volumes are
729
    reported as unknown.
730

731
    """
732
    bad = False
733

    
734
    for node in node_vol_is:
735
      for volume in node_vol_is[node]:
736
        if node not in node_vol_should or volume not in node_vol_should[node]:
737
          feedback_fn("  - ERROR: volume %s on node %s should not exist" %
738
                      (volume, node))
739
          bad = True
740
    return bad
741

    
742
  def _VerifyOrphanInstances(self, instancelist, node_instance, feedback_fn):
743
    """Verify the list of running instances.
744

745
    This checks what instances are running but unknown to the cluster.
746

747
    """
748
    bad = False
749
    for node in node_instance:
750
      for runninginstance in node_instance[node]:
751
        if runninginstance not in instancelist:
752
          feedback_fn("  - ERROR: instance %s on node %s should not exist" %
753
                          (runninginstance, node))
754
          bad = True
755
    return bad
756

    
757
  def _VerifyNPlusOneMemory(self, node_info, instance_cfg, feedback_fn):
758
    """Verify N+1 Memory Resilience.
759

760
    Check that if one single node dies we can still start all the instances it
761
    was primary for.
762

763
    """
764
    bad = False
765

    
766
    for node, nodeinfo in node_info.iteritems():
767
      # This code checks that every node which is now listed as secondary has
768
      # enough memory to host all instances it is supposed to should a single
769
      # other node in the cluster fail.
770
      # FIXME: not ready for failover to an arbitrary node
771
      # FIXME: does not support file-backed instances
772
      # WARNING: we currently take into account down instances as well as up
773
      # ones, considering that even if they're down someone might want to start
774
      # them even in the event of a node failure.
775
      for prinode, instances in nodeinfo['sinst-by-pnode'].iteritems():
776
        needed_mem = 0
777
        for instance in instances:
778
          bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
779
          if bep[constants.BE_AUTO_BALANCE]:
780
            needed_mem += bep[constants.BE_MEMORY]
781
        if nodeinfo['mfree'] < needed_mem:
782
          feedback_fn("  - ERROR: not enough memory on node %s to accomodate"
783
                      " failovers should node %s fail" % (node, prinode))
784
          bad = True
785
    return bad
786

    
787
  def CheckPrereq(self):
788
    """Check prerequisites.
789

790
    Transform the list of checks we're going to skip into a set and check that
791
    all its members are valid.
792

793
    """
794
    self.skip_set = frozenset(self.op.skip_checks)
795
    if not constants.VERIFY_OPTIONAL_CHECKS.issuperset(self.skip_set):
796
      raise errors.OpPrereqError("Invalid checks to be skipped specified")
797

    
798
  def BuildHooksEnv(self):
799
    """Build hooks env.
800

801
    Cluster-Verify hooks just rone in the post phase and their failure makes
802
    the output be logged in the verify output and the verification to fail.
803

804
    """
805
    all_nodes = self.cfg.GetNodeList()
806
    # TODO: populate the environment with useful information for verify hooks
807
    env = {}
808
    return env, [], all_nodes
809

    
810
  def Exec(self, feedback_fn):
811
    """Verify integrity of cluster, performing various test on nodes.
812

813
    """
814
    bad = False
815
    feedback_fn("* Verifying global settings")
816
    for msg in self.cfg.VerifyConfig():
817
      feedback_fn("  - ERROR: %s" % msg)
818

    
819
    vg_name = self.cfg.GetVGName()
820
    hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
821
    nodelist = utils.NiceSort(self.cfg.GetNodeList())
822
    nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
823
    instancelist = utils.NiceSort(self.cfg.GetInstanceList())
824
    i_non_redundant = [] # Non redundant instances
825
    i_non_a_balanced = [] # Non auto-balanced instances
826
    node_volume = {}
827
    node_instance = {}
828
    node_info = {}
829
    instance_cfg = {}
830

    
831
    # FIXME: verify OS list
832
    # do local checksums
833
    master_files = [constants.CLUSTER_CONF_FILE]
834

    
835
    file_names = ssconf.SimpleStore().GetFileList()
836
    file_names.append(constants.SSL_CERT_FILE)
837
    file_names.extend(master_files)
838

    
839
    local_checksums = utils.FingerprintFiles(file_names)
840

    
841
    feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
842
    node_verify_param = {
843
      constants.NV_FILELIST: file_names,
844
      constants.NV_NODELIST: nodelist,
845
      constants.NV_HYPERVISOR: hypervisors,
846
      constants.NV_NODENETTEST: [(node.name, node.primary_ip,
847
                                  node.secondary_ip) for node in nodeinfo],
848
      constants.NV_LVLIST: vg_name,
849
      constants.NV_INSTANCELIST: hypervisors,
850
      constants.NV_VGLIST: None,
851
      constants.NV_VERSION: None,
852
      constants.NV_HVINFO: self.cfg.GetHypervisorType(),
853
      }
854
    all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
855
                                           self.cfg.GetClusterName())
856

    
857
    cluster = self.cfg.GetClusterInfo()
858
    master_node = self.cfg.GetMasterNode()
859
    for node_i in nodeinfo:
860
      node = node_i.name
861
      nresult = all_nvinfo[node].data
862

    
863
      if node == master_node:
864
        ntype = "master"
865
      elif node_i.master_candidate:
866
        ntype = "master candidate"
867
      else:
868
        ntype = "regular"
869
      feedback_fn("* Verifying node %s (%s)" % (node, ntype))
870

    
871
      if all_nvinfo[node].failed or not isinstance(nresult, dict):
872
        feedback_fn("  - ERROR: connection to %s failed" % (node,))
873
        bad = True
874
        continue
875

    
876
      result = self._VerifyNode(node_i, file_names, local_checksums,
877
                                nresult, feedback_fn, master_files)
878
      bad = bad or result
879

    
880
      lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
881
      if isinstance(lvdata, basestring):
882
        feedback_fn("  - ERROR: LVM problem on node %s: %s" %
883
                    (node, lvdata.encode('string_escape')))
884
        bad = True
885
        node_volume[node] = {}
886
      elif not isinstance(lvdata, dict):
887
        feedback_fn("  - ERROR: connection to %s failed (lvlist)" % (node,))
888
        bad = True
889
        continue
890
      else:
891
        node_volume[node] = lvdata
892

    
893
      # node_instance
894
      idata = nresult.get(constants.NV_INSTANCELIST, None)
895
      if not isinstance(idata, list):
896
        feedback_fn("  - ERROR: connection to %s failed (instancelist)" %
897
                    (node,))
898
        bad = True
899
        continue
900

    
901
      node_instance[node] = idata
902

    
903
      # node_info
904
      nodeinfo = nresult.get(constants.NV_HVINFO, None)
905
      if not isinstance(nodeinfo, dict):
906
        feedback_fn("  - ERROR: connection to %s failed (hvinfo)" % (node,))
907
        bad = True
908
        continue
909

    
910
      try:
911
        node_info[node] = {
912
          "mfree": int(nodeinfo['memory_free']),
913
          "dfree": int(nresult[constants.NV_VGLIST][vg_name]),
914
          "pinst": [],
915
          "sinst": [],
916
          # dictionary holding all instances this node is secondary for,
917
          # grouped by their primary node. Each key is a cluster node, and each
918
          # value is a list of instances which have the key as primary and the
919
          # current node as secondary.  this is handy to calculate N+1 memory
920
          # availability if you can only failover from a primary to its
921
          # secondary.
922
          "sinst-by-pnode": {},
923
        }
924
      except ValueError:
925
        feedback_fn("  - ERROR: invalid value returned from node %s" % (node,))
926
        bad = True
927
        continue
928

    
929
    node_vol_should = {}
930

    
931
    for instance in instancelist:
932
      feedback_fn("* Verifying instance %s" % instance)
933
      inst_config = self.cfg.GetInstanceInfo(instance)
934
      result =  self._VerifyInstance(instance, inst_config, node_volume,
935
                                     node_instance, feedback_fn)
936
      bad = bad or result
937

    
938
      inst_config.MapLVsByNode(node_vol_should)
939

    
940
      instance_cfg[instance] = inst_config
941

    
942
      pnode = inst_config.primary_node
943
      if pnode in node_info:
944
        node_info[pnode]['pinst'].append(instance)
945
      else:
946
        feedback_fn("  - ERROR: instance %s, connection to primary node"
947
                    " %s failed" % (instance, pnode))
948
        bad = True
949

    
950
      # If the instance is non-redundant we cannot survive losing its primary
951
      # node, so we are not N+1 compliant. On the other hand we have no disk
952
      # templates with more than one secondary so that situation is not well
953
      # supported either.
954
      # FIXME: does not support file-backed instances
955
      if len(inst_config.secondary_nodes) == 0:
956
        i_non_redundant.append(instance)
957
      elif len(inst_config.secondary_nodes) > 1:
958
        feedback_fn("  - WARNING: multiple secondaries for instance %s"
959
                    % instance)
960

    
961
      if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
962
        i_non_a_balanced.append(instance)
963

    
964
      for snode in inst_config.secondary_nodes:
965
        if snode in node_info:
966
          node_info[snode]['sinst'].append(instance)
967
          if pnode not in node_info[snode]['sinst-by-pnode']:
968
            node_info[snode]['sinst-by-pnode'][pnode] = []
969
          node_info[snode]['sinst-by-pnode'][pnode].append(instance)
970
        else:
971
          feedback_fn("  - ERROR: instance %s, connection to secondary node"
972
                      " %s failed" % (instance, snode))
973

    
974
    feedback_fn("* Verifying orphan volumes")
975
    result = self._VerifyOrphanVolumes(node_vol_should, node_volume,
976
                                       feedback_fn)
977
    bad = bad or result
978

    
979
    feedback_fn("* Verifying remaining instances")
980
    result = self._VerifyOrphanInstances(instancelist, node_instance,
981
                                         feedback_fn)
982
    bad = bad or result
983

    
984
    if constants.VERIFY_NPLUSONE_MEM not in self.skip_set:
985
      feedback_fn("* Verifying N+1 Memory redundancy")
986
      result = self._VerifyNPlusOneMemory(node_info, instance_cfg, feedback_fn)
987
      bad = bad or result
988

    
989
    feedback_fn("* Other Notes")
990
    if i_non_redundant:
991
      feedback_fn("  - NOTICE: %d non-redundant instance(s) found."
992
                  % len(i_non_redundant))
993

    
994
    if i_non_a_balanced:
995
      feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
996
                  % len(i_non_a_balanced))
997

    
998
    return not bad
999

    
1000
  def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
1001
    """Analize the post-hooks' result
1002

1003
    This method analyses the hook result, handles it, and sends some
1004
    nicely-formatted feedback back to the user.
1005

1006
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
1007
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
1008
    @param hooks_results: the results of the multi-node hooks rpc call
1009
    @param feedback_fn: function used send feedback back to the caller
1010
    @param lu_result: previous Exec result
1011
    @return: the new Exec result, based on the previous result
1012
        and hook results
1013

1014
    """
1015
    # We only really run POST phase hooks, and are only interested in
1016
    # their results
1017
    if phase == constants.HOOKS_PHASE_POST:
1018
      # Used to change hooks' output to proper indentation
1019
      indent_re = re.compile('^', re.M)
1020
      feedback_fn("* Hooks Results")
1021
      if not hooks_results:
1022
        feedback_fn("  - ERROR: general communication failure")
1023
        lu_result = 1
1024
      else:
1025
        for node_name in hooks_results:
1026
          show_node_header = True
1027
          res = hooks_results[node_name]
1028
          if res.failed or res.data is False or not isinstance(res.data, list):
1029
            feedback_fn("    Communication failure in hooks execution")
1030
            lu_result = 1
1031
            continue
1032
          for script, hkr, output in res.data:
1033
            if hkr == constants.HKR_FAIL:
1034
              # The node header is only shown once, if there are
1035
              # failing hooks on that node
1036
              if show_node_header:
1037
                feedback_fn("  Node %s:" % node_name)
1038
                show_node_header = False
1039
              feedback_fn("    ERROR: Script %s failed, output:" % script)
1040
              output = indent_re.sub('      ', output)
1041
              feedback_fn("%s" % output)
1042
              lu_result = 1
1043

    
1044
      return lu_result
1045

    
1046

    
1047
class LUVerifyDisks(NoHooksLU):
1048
  """Verifies the cluster disks status.
1049

1050
  """
1051
  _OP_REQP = []
1052
  REQ_BGL = False
1053

    
1054
  def ExpandNames(self):
1055
    self.needed_locks = {
1056
      locking.LEVEL_NODE: locking.ALL_SET,
1057
      locking.LEVEL_INSTANCE: locking.ALL_SET,
1058
    }
1059
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
1060

    
1061
  def CheckPrereq(self):
1062
    """Check prerequisites.
1063

1064
    This has no prerequisites.
1065

1066
    """
1067
    pass
1068

    
1069
  def Exec(self, feedback_fn):
1070
    """Verify integrity of cluster disks.
1071

1072
    """
1073
    result = res_nodes, res_nlvm, res_instances, res_missing = [], {}, [], {}
1074

    
1075
    vg_name = self.cfg.GetVGName()
1076
    nodes = utils.NiceSort(self.cfg.GetNodeList())
1077
    instances = [self.cfg.GetInstanceInfo(name)
1078
                 for name in self.cfg.GetInstanceList()]
1079

    
1080
    nv_dict = {}
1081
    for inst in instances:
1082
      inst_lvs = {}
1083
      if (inst.status != "up" or
1084
          inst.disk_template not in constants.DTS_NET_MIRROR):
1085
        continue
1086
      inst.MapLVsByNode(inst_lvs)
1087
      # transform { iname: {node: [vol,],},} to {(node, vol): iname}
1088
      for node, vol_list in inst_lvs.iteritems():
1089
        for vol in vol_list:
1090
          nv_dict[(node, vol)] = inst
1091

    
1092
    if not nv_dict:
1093
      return result
1094

    
1095
    node_lvs = self.rpc.call_volume_list(nodes, vg_name)
1096

    
1097
    to_act = set()
1098
    for node in nodes:
1099
      # node_volume
1100
      lvs = node_lvs[node]
1101
      if lvs.failed:
1102
        self.LogWarning("Connection to node %s failed: %s" %
1103
                        (node, lvs.data))
1104
        continue
1105
      lvs = lvs.data
1106
      if isinstance(lvs, basestring):
1107
        logging.warning("Error enumerating LVs on node %s: %s", node, lvs)
1108
        res_nlvm[node] = lvs
1109
      elif not isinstance(lvs, dict):
1110
        logging.warning("Connection to node %s failed or invalid data"
1111
                        " returned", node)
1112
        res_nodes.append(node)
1113
        continue
1114

    
1115
      for lv_name, (_, lv_inactive, lv_online) in lvs.iteritems():
1116
        inst = nv_dict.pop((node, lv_name), None)
1117
        if (not lv_online and inst is not None
1118
            and inst.name not in res_instances):
1119
          res_instances.append(inst.name)
1120

    
1121
    # any leftover items in nv_dict are missing LVs, let's arrange the
1122
    # data better
1123
    for key, inst in nv_dict.iteritems():
1124
      if inst.name not in res_missing:
1125
        res_missing[inst.name] = []
1126
      res_missing[inst.name].append(key)
1127

    
1128
    return result
1129

    
1130

    
1131
class LURenameCluster(LogicalUnit):
1132
  """Rename the cluster.
1133

1134
  """
1135
  HPATH = "cluster-rename"
1136
  HTYPE = constants.HTYPE_CLUSTER
1137
  _OP_REQP = ["name"]
1138

    
1139
  def BuildHooksEnv(self):
1140
    """Build hooks env.
1141

1142
    """
1143
    env = {
1144
      "OP_TARGET": self.cfg.GetClusterName(),
1145
      "NEW_NAME": self.op.name,
1146
      }
1147
    mn = self.cfg.GetMasterNode()
1148
    return env, [mn], [mn]
1149

    
1150
  def CheckPrereq(self):
1151
    """Verify that the passed name is a valid one.
1152

1153
    """
1154
    hostname = utils.HostInfo(self.op.name)
1155

    
1156
    new_name = hostname.name
1157
    self.ip = new_ip = hostname.ip
1158
    old_name = self.cfg.GetClusterName()
1159
    old_ip = self.cfg.GetMasterIP()
1160
    if new_name == old_name and new_ip == old_ip:
1161
      raise errors.OpPrereqError("Neither the name nor the IP address of the"
1162
                                 " cluster has changed")
1163
    if new_ip != old_ip:
1164
      if utils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
1165
        raise errors.OpPrereqError("The given cluster IP address (%s) is"
1166
                                   " reachable on the network. Aborting." %
1167
                                   new_ip)
1168

    
1169
    self.op.name = new_name
1170

    
1171
  def Exec(self, feedback_fn):
1172
    """Rename the cluster.
1173

1174
    """
1175
    clustername = self.op.name
1176
    ip = self.ip
1177

    
1178
    # shutdown the master IP
1179
    master = self.cfg.GetMasterNode()
1180
    result = self.rpc.call_node_stop_master(master, False)
1181
    if result.failed or not result.data:
1182
      raise errors.OpExecError("Could not disable the master role")
1183

    
1184
    try:
1185
      cluster = self.cfg.GetClusterInfo()
1186
      cluster.cluster_name = clustername
1187
      cluster.master_ip = ip
1188
      self.cfg.Update(cluster)
1189

    
1190
      # update the known hosts file
1191
      ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
1192
      node_list = self.cfg.GetNodeList()
1193
      try:
1194
        node_list.remove(master)
1195
      except ValueError:
1196
        pass
1197
      result = self.rpc.call_upload_file(node_list,
1198
                                         constants.SSH_KNOWN_HOSTS_FILE)
1199
      for to_node, to_result in result.iteritems():
1200
        if to_result.failed or not to_result.data:
1201
          logging.error("Copy of file %s to node %s failed", fname, to_node)
1202

    
1203
    finally:
1204
      result = self.rpc.call_node_start_master(master, False)
1205
      if result.failed or not result.data:
1206
        self.LogWarning("Could not re-enable the master role on"
1207
                        " the master, please restart manually.")
1208

    
1209

    
1210
def _RecursiveCheckIfLVMBased(disk):
1211
  """Check if the given disk or its children are lvm-based.
1212

1213
  @type disk: L{objects.Disk}
1214
  @param disk: the disk to check
1215
  @rtype: booleean
1216
  @return: boolean indicating whether a LD_LV dev_type was found or not
1217

1218
  """
1219
  if disk.children:
1220
    for chdisk in disk.children:
1221
      if _RecursiveCheckIfLVMBased(chdisk):
1222
        return True
1223
  return disk.dev_type == constants.LD_LV
1224

    
1225

    
1226
class LUSetClusterParams(LogicalUnit):
1227
  """Change the parameters of the cluster.
1228

1229
  """
1230
  HPATH = "cluster-modify"
1231
  HTYPE = constants.HTYPE_CLUSTER
1232
  _OP_REQP = []
1233
  REQ_BGL = False
1234

    
1235
  def CheckParameters(self):
1236
    """Check parameters
1237

1238
    """
1239
    if not hasattr(self.op, "candidate_pool_size"):
1240
      self.op.candidate_pool_size = None
1241
    if self.op.candidate_pool_size is not None:
1242
      try:
1243
        self.op.candidate_pool_size = int(self.op.candidate_pool_size)
1244
      except ValueError, err:
1245
        raise errors.OpPrereqError("Invalid candidate_pool_size value: %s" %
1246
                                   str(err))
1247
      if self.op.candidate_pool_size < 1:
1248
        raise errors.OpPrereqError("At least one master candidate needed")
1249

    
1250
  def ExpandNames(self):
1251
    # FIXME: in the future maybe other cluster params won't require checking on
1252
    # all nodes to be modified.
1253
    self.needed_locks = {
1254
      locking.LEVEL_NODE: locking.ALL_SET,
1255
    }
1256
    self.share_locks[locking.LEVEL_NODE] = 1
1257

    
1258
  def BuildHooksEnv(self):
1259
    """Build hooks env.
1260

1261
    """
1262
    env = {
1263
      "OP_TARGET": self.cfg.GetClusterName(),
1264
      "NEW_VG_NAME": self.op.vg_name,
1265
      }
1266
    mn = self.cfg.GetMasterNode()
1267
    return env, [mn], [mn]
1268

    
1269
  def CheckPrereq(self):
1270
    """Check prerequisites.
1271

1272
    This checks whether the given params don't conflict and
1273
    if the given volume group is valid.
1274

1275
    """
1276
    # FIXME: This only works because there is only one parameter that can be
1277
    # changed or removed.
1278
    if self.op.vg_name is not None and not self.op.vg_name:
1279
      instances = self.cfg.GetAllInstancesInfo().values()
1280
      for inst in instances:
1281
        for disk in inst.disks:
1282
          if _RecursiveCheckIfLVMBased(disk):
1283
            raise errors.OpPrereqError("Cannot disable lvm storage while"
1284
                                       " lvm-based instances exist")
1285

    
1286
    node_list = self.acquired_locks[locking.LEVEL_NODE]
1287

    
1288
    # if vg_name not None, checks given volume group on all nodes
1289
    if self.op.vg_name:
1290
      vglist = self.rpc.call_vg_list(node_list)
1291
      for node in node_list:
1292
        if vglist[node].failed:
1293
          # ignoring down node
1294
          self.LogWarning("Node %s unreachable/error, ignoring" % node)
1295
          continue
1296
        vgstatus = utils.CheckVolumeGroupSize(vglist[node].data,
1297
                                              self.op.vg_name,
1298
                                              constants.MIN_VG_SIZE)
1299
        if vgstatus:
1300
          raise errors.OpPrereqError("Error on node '%s': %s" %
1301
                                     (node, vgstatus))
1302

    
1303
    self.cluster = cluster = self.cfg.GetClusterInfo()
1304
    # validate beparams changes
1305
    if self.op.beparams:
1306
      utils.CheckBEParams(self.op.beparams)
1307
      self.new_beparams = cluster.FillDict(
1308
        cluster.beparams[constants.BEGR_DEFAULT], self.op.beparams)
1309

    
1310
    # hypervisor list/parameters
1311
    self.new_hvparams = cluster.FillDict(cluster.hvparams, {})
1312
    if self.op.hvparams:
1313
      if not isinstance(self.op.hvparams, dict):
1314
        raise errors.OpPrereqError("Invalid 'hvparams' parameter on input")
1315
      for hv_name, hv_dict in self.op.hvparams.items():
1316
        if hv_name not in self.new_hvparams:
1317
          self.new_hvparams[hv_name] = hv_dict
1318
        else:
1319
          self.new_hvparams[hv_name].update(hv_dict)
1320

    
1321
    if self.op.enabled_hypervisors is not None:
1322
      self.hv_list = self.op.enabled_hypervisors
1323
    else:
1324
      self.hv_list = cluster.enabled_hypervisors
1325

    
1326
    if self.op.hvparams or self.op.enabled_hypervisors is not None:
1327
      # either the enabled list has changed, or the parameters have, validate
1328
      for hv_name, hv_params in self.new_hvparams.items():
1329
        if ((self.op.hvparams and hv_name in self.op.hvparams) or
1330
            (self.op.enabled_hypervisors and
1331
             hv_name in self.op.enabled_hypervisors)):
1332
          # either this is a new hypervisor, or its parameters have changed
1333
          hv_class = hypervisor.GetHypervisor(hv_name)
1334
          hv_class.CheckParameterSyntax(hv_params)
1335
          _CheckHVParams(self, node_list, hv_name, hv_params)
1336

    
1337
  def Exec(self, feedback_fn):
1338
    """Change the parameters of the cluster.
1339

1340
    """
1341
    if self.op.vg_name is not None:
1342
      if self.op.vg_name != self.cfg.GetVGName():
1343
        self.cfg.SetVGName(self.op.vg_name)
1344
      else:
1345
        feedback_fn("Cluster LVM configuration already in desired"
1346
                    " state, not changing")
1347
    if self.op.hvparams:
1348
      self.cluster.hvparams = self.new_hvparams
1349
    if self.op.enabled_hypervisors is not None:
1350
      self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
1351
    if self.op.beparams:
1352
      self.cluster.beparams[constants.BEGR_DEFAULT] = self.new_beparams
1353
    if self.op.candidate_pool_size is not None:
1354
      self.cluster.candidate_pool_size = self.op.candidate_pool_size
1355

    
1356
    self.cfg.Update(self.cluster)
1357

    
1358
    # we want to update nodes after the cluster so that if any errors
1359
    # happen, we have recorded and saved the cluster info
1360
    if self.op.candidate_pool_size is not None:
1361
      node_info = self.cfg.GetAllNodesInfo().values()
1362
      num_candidates = len([node for node in node_info
1363
                            if node.master_candidate])
1364
      num_nodes = len(node_info)
1365
      if num_candidates < self.op.candidate_pool_size:
1366
        random.shuffle(node_info)
1367
        for node in node_info:
1368
          if num_candidates >= self.op.candidate_pool_size:
1369
            break
1370
          if node.master_candidate:
1371
            continue
1372
          node.master_candidate = True
1373
          self.LogInfo("Promoting node %s to master candidate", node.name)
1374
          self.cfg.Update(node)
1375
          self.context.ReaddNode(node)
1376
          num_candidates += 1
1377
      elif num_candidates > self.op.candidate_pool_size:
1378
        self.LogInfo("Note: more nodes are candidates (%d) than the new value"
1379
                     " of candidate_pool_size (%d)" %
1380
                     (num_candidates, self.op.candidate_pool_size))
1381

    
1382

    
1383
def _WaitForSync(lu, instance, oneshot=False, unlock=False):
1384
  """Sleep and poll for an instance's disk to sync.
1385

1386
  """
1387
  if not instance.disks:
1388
    return True
1389

    
1390
  if not oneshot:
1391
    lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
1392

    
1393
  node = instance.primary_node
1394

    
1395
  for dev in instance.disks:
1396
    lu.cfg.SetDiskID(dev, node)
1397

    
1398
  retries = 0
1399
  while True:
1400
    max_time = 0
1401
    done = True
1402
    cumul_degraded = False
1403
    rstats = lu.rpc.call_blockdev_getmirrorstatus(node, instance.disks)
1404
    if rstats.failed or not rstats.data:
1405
      lu.LogWarning("Can't get any data from node %s", node)
1406
      retries += 1
1407
      if retries >= 10:
1408
        raise errors.RemoteError("Can't contact node %s for mirror data,"
1409
                                 " aborting." % node)
1410
      time.sleep(6)
1411
      continue
1412
    rstats = rstats.data
1413
    retries = 0
1414
    for i in range(len(rstats)):
1415
      mstat = rstats[i]
1416
      if mstat is None:
1417
        lu.LogWarning("Can't compute data for node %s/%s",
1418
                           node, instance.disks[i].iv_name)
1419
        continue
1420
      # we ignore the ldisk parameter
1421
      perc_done, est_time, is_degraded, _ = mstat
1422
      cumul_degraded = cumul_degraded or (is_degraded and perc_done is None)
1423
      if perc_done is not None:
1424
        done = False
1425
        if est_time is not None:
1426
          rem_time = "%d estimated seconds remaining" % est_time
1427
          max_time = est_time
1428
        else:
1429
          rem_time = "no time estimate"
1430
        lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
1431
                        (instance.disks[i].iv_name, perc_done, rem_time))
1432
    if done or oneshot:
1433
      break
1434

    
1435
    time.sleep(min(60, max_time))
1436

    
1437
  if done:
1438
    lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
1439
  return not cumul_degraded
1440

    
1441

    
1442
def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
1443
  """Check that mirrors are not degraded.
1444

1445
  The ldisk parameter, if True, will change the test from the
1446
  is_degraded attribute (which represents overall non-ok status for
1447
  the device(s)) to the ldisk (representing the local storage status).
1448

1449
  """
1450
  lu.cfg.SetDiskID(dev, node)
1451
  if ldisk:
1452
    idx = 6
1453
  else:
1454
    idx = 5
1455

    
1456
  result = True
1457
  if on_primary or dev.AssembleOnSecondary():
1458
    rstats = lu.rpc.call_blockdev_find(node, dev)
1459
    if rstats.failed or not rstats.data:
1460
      logging.warning("Node %s: disk degraded, not found or node down", node)
1461
      result = False
1462
    else:
1463
      result = result and (not rstats.data[idx])
1464
  if dev.children:
1465
    for child in dev.children:
1466
      result = result and _CheckDiskConsistency(lu, child, node, on_primary)
1467

    
1468
  return result
1469

    
1470

    
1471
class LUDiagnoseOS(NoHooksLU):
1472
  """Logical unit for OS diagnose/query.
1473

1474
  """
1475
  _OP_REQP = ["output_fields", "names"]
1476
  REQ_BGL = False
1477
  _FIELDS_STATIC = utils.FieldSet()
1478
  _FIELDS_DYNAMIC = utils.FieldSet("name", "valid", "node_status")
1479

    
1480
  def ExpandNames(self):
1481
    if self.op.names:
1482
      raise errors.OpPrereqError("Selective OS query not supported")
1483

    
1484
    _CheckOutputFields(static=self._FIELDS_STATIC,
1485
                       dynamic=self._FIELDS_DYNAMIC,
1486
                       selected=self.op.output_fields)
1487

    
1488
    # Lock all nodes, in shared mode
1489
    self.needed_locks = {}
1490
    self.share_locks[locking.LEVEL_NODE] = 1
1491
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
1492

    
1493
  def CheckPrereq(self):
1494
    """Check prerequisites.
1495

1496
    """
1497

    
1498
  @staticmethod
1499
  def _DiagnoseByOS(node_list, rlist):
1500
    """Remaps a per-node return list into an a per-os per-node dictionary
1501

1502
    @param node_list: a list with the names of all nodes
1503
    @param rlist: a map with node names as keys and OS objects as values
1504

1505
    @rtype: dict
1506
    @returns: a dictionary with osnames as keys and as value another map, with
1507
        nodes as keys and list of OS objects as values, eg::
1508

1509
          {"debian-etch": {"node1": [<object>,...],
1510
                           "node2": [<object>,]}
1511
          }
1512

1513
    """
1514
    all_os = {}
1515
    for node_name, nr in rlist.iteritems():
1516
      if nr.failed or not nr.data:
1517
        continue
1518
      for os_obj in nr.data:
1519
        if os_obj.name not in all_os:
1520
          # build a list of nodes for this os containing empty lists
1521
          # for each node in node_list
1522
          all_os[os_obj.name] = {}
1523
          for nname in node_list:
1524
            all_os[os_obj.name][nname] = []
1525
        all_os[os_obj.name][node_name].append(os_obj)
1526
    return all_os
1527

    
1528
  def Exec(self, feedback_fn):
1529
    """Compute the list of OSes.
1530

1531
    """
1532
    node_list = self.acquired_locks[locking.LEVEL_NODE]
1533
    node_data = self.rpc.call_os_diagnose(node_list)
1534
    if node_data == False:
1535
      raise errors.OpExecError("Can't gather the list of OSes")
1536
    pol = self._DiagnoseByOS(node_list, node_data)
1537
    output = []
1538
    for os_name, os_data in pol.iteritems():
1539
      row = []
1540
      for field in self.op.output_fields:
1541
        if field == "name":
1542
          val = os_name
1543
        elif field == "valid":
1544
          val = utils.all([osl and osl[0] for osl in os_data.values()])
1545
        elif field == "node_status":
1546
          val = {}
1547
          for node_name, nos_list in os_data.iteritems():
1548
            val[node_name] = [(v.status, v.path) for v in nos_list]
1549
        else:
1550
          raise errors.ParameterError(field)
1551
        row.append(val)
1552
      output.append(row)
1553

    
1554
    return output
1555

    
1556

    
1557
class LURemoveNode(LogicalUnit):
1558
  """Logical unit for removing a node.
1559

1560
  """
1561
  HPATH = "node-remove"
1562
  HTYPE = constants.HTYPE_NODE
1563
  _OP_REQP = ["node_name"]
1564

    
1565
  def BuildHooksEnv(self):
1566
    """Build hooks env.
1567

1568
    This doesn't run on the target node in the pre phase as a failed
1569
    node would then be impossible to remove.
1570

1571
    """
1572
    env = {
1573
      "OP_TARGET": self.op.node_name,
1574
      "NODE_NAME": self.op.node_name,
1575
      }
1576
    all_nodes = self.cfg.GetNodeList()
1577
    all_nodes.remove(self.op.node_name)
1578
    return env, all_nodes, all_nodes
1579

    
1580
  def CheckPrereq(self):
1581
    """Check prerequisites.
1582

1583
    This checks:
1584
     - the node exists in the configuration
1585
     - it does not have primary or secondary instances
1586
     - it's not the master
1587

1588
    Any errors are signalled by raising errors.OpPrereqError.
1589

1590
    """
1591
    node = self.cfg.GetNodeInfo(self.cfg.ExpandNodeName(self.op.node_name))
1592
    if node is None:
1593
      raise errors.OpPrereqError, ("Node '%s' is unknown." % self.op.node_name)
1594

    
1595
    instance_list = self.cfg.GetInstanceList()
1596

    
1597
    masternode = self.cfg.GetMasterNode()
1598
    if node.name == masternode:
1599
      raise errors.OpPrereqError("Node is the master node,"
1600
                                 " you need to failover first.")
1601

    
1602
    for instance_name in instance_list:
1603
      instance = self.cfg.GetInstanceInfo(instance_name)
1604
      if node.name == instance.primary_node:
1605
        raise errors.OpPrereqError("Instance %s still running on the node,"
1606
                                   " please remove first." % instance_name)
1607
      if node.name in instance.secondary_nodes:
1608
        raise errors.OpPrereqError("Instance %s has node as a secondary,"
1609
                                   " please remove first." % instance_name)
1610
    self.op.node_name = node.name
1611
    self.node = node
1612

    
1613
  def Exec(self, feedback_fn):
1614
    """Removes the node from the cluster.
1615

1616
    """
1617
    node = self.node
1618
    logging.info("Stopping the node daemon and removing configs from node %s",
1619
                 node.name)
1620

    
1621
    self.context.RemoveNode(node.name)
1622

    
1623
    self.rpc.call_node_leave_cluster(node.name)
1624

    
1625
    # Promote nodes to master candidate as needed
1626
    cp_size = self.cfg.GetClusterInfo().candidate_pool_size
1627
    node_info = self.cfg.GetAllNodesInfo().values()
1628
    num_candidates = len([n for n in node_info
1629
                          if n.master_candidate])
1630
    num_nodes = len(node_info)
1631
    random.shuffle(node_info)
1632
    for node in node_info:
1633
      if num_candidates >= cp_size or num_candidates >= num_nodes:
1634
        break
1635
      if node.master_candidate:
1636
        continue
1637
      node.master_candidate = True
1638
      self.LogInfo("Promoting node %s to master candidate", node.name)
1639
      self.cfg.Update(node)
1640
      self.context.ReaddNode(node)
1641
      num_candidates += 1
1642

    
1643

    
1644
class LUQueryNodes(NoHooksLU):
1645
  """Logical unit for querying nodes.
1646

1647
  """
1648
  _OP_REQP = ["output_fields", "names"]
1649
  REQ_BGL = False
1650
  _FIELDS_DYNAMIC = utils.FieldSet(
1651
    "dtotal", "dfree",
1652
    "mtotal", "mnode", "mfree",
1653
    "bootid",
1654
    "ctotal",
1655
    )
1656

    
1657
  _FIELDS_STATIC = utils.FieldSet(
1658
    "name", "pinst_cnt", "sinst_cnt",
1659
    "pinst_list", "sinst_list",
1660
    "pip", "sip", "tags",
1661
    "serial_no",
1662
    "master_candidate",
1663
    "master",
1664
    "offline",
1665
    )
1666

    
1667
  def ExpandNames(self):
1668
    _CheckOutputFields(static=self._FIELDS_STATIC,
1669
                       dynamic=self._FIELDS_DYNAMIC,
1670
                       selected=self.op.output_fields)
1671

    
1672
    self.needed_locks = {}
1673
    self.share_locks[locking.LEVEL_NODE] = 1
1674

    
1675
    if self.op.names:
1676
      self.wanted = _GetWantedNodes(self, self.op.names)
1677
    else:
1678
      self.wanted = locking.ALL_SET
1679

    
1680
    self.do_locking = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
1681
    if self.do_locking:
1682
      # if we don't request only static fields, we need to lock the nodes
1683
      self.needed_locks[locking.LEVEL_NODE] = self.wanted
1684

    
1685

    
1686
  def CheckPrereq(self):
1687
    """Check prerequisites.
1688

1689
    """
1690
    # The validation of the node list is done in the _GetWantedNodes,
1691
    # if non empty, and if empty, there's no validation to do
1692
    pass
1693

    
1694
  def Exec(self, feedback_fn):
1695
    """Computes the list of nodes and their attributes.
1696

1697
    """
1698
    all_info = self.cfg.GetAllNodesInfo()
1699
    if self.do_locking:
1700
      nodenames = self.acquired_locks[locking.LEVEL_NODE]
1701
    elif self.wanted != locking.ALL_SET:
1702
      nodenames = self.wanted
1703
      missing = set(nodenames).difference(all_info.keys())
1704
      if missing:
1705
        raise errors.OpExecError(
1706
          "Some nodes were removed before retrieving their data: %s" % missing)
1707
    else:
1708
      nodenames = all_info.keys()
1709

    
1710
    nodenames = utils.NiceSort(nodenames)
1711
    nodelist = [all_info[name] for name in nodenames]
1712

    
1713
    # begin data gathering
1714

    
1715
    if self.do_locking:
1716
      live_data = {}
1717
      node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
1718
                                          self.cfg.GetHypervisorType())
1719
      for name in nodenames:
1720
        nodeinfo = node_data[name]
1721
        if not nodeinfo.failed and nodeinfo.data:
1722
          nodeinfo = nodeinfo.data
1723
          fn = utils.TryConvert
1724
          live_data[name] = {
1725
            "mtotal": fn(int, nodeinfo.get('memory_total', None)),
1726
            "mnode": fn(int, nodeinfo.get('memory_dom0', None)),
1727
            "mfree": fn(int, nodeinfo.get('memory_free', None)),
1728
            "dtotal": fn(int, nodeinfo.get('vg_size', None)),
1729
            "dfree": fn(int, nodeinfo.get('vg_free', None)),
1730
            "ctotal": fn(int, nodeinfo.get('cpu_total', None)),
1731
            "bootid": nodeinfo.get('bootid', None),
1732
            }
1733
        else:
1734
          live_data[name] = {}
1735
    else:
1736
      live_data = dict.fromkeys(nodenames, {})
1737

    
1738
    node_to_primary = dict([(name, set()) for name in nodenames])
1739
    node_to_secondary = dict([(name, set()) for name in nodenames])
1740

    
1741
    inst_fields = frozenset(("pinst_cnt", "pinst_list",
1742
                             "sinst_cnt", "sinst_list"))
1743
    if inst_fields & frozenset(self.op.output_fields):
1744
      instancelist = self.cfg.GetInstanceList()
1745

    
1746
      for instance_name in instancelist:
1747
        inst = self.cfg.GetInstanceInfo(instance_name)
1748
        if inst.primary_node in node_to_primary:
1749
          node_to_primary[inst.primary_node].add(inst.name)
1750
        for secnode in inst.secondary_nodes:
1751
          if secnode in node_to_secondary:
1752
            node_to_secondary[secnode].add(inst.name)
1753

    
1754
    master_node = self.cfg.GetMasterNode()
1755

    
1756
    # end data gathering
1757

    
1758
    output = []
1759
    for node in nodelist:
1760
      node_output = []
1761
      for field in self.op.output_fields:
1762
        if field == "name":
1763
          val = node.name
1764
        elif field == "pinst_list":
1765
          val = list(node_to_primary[node.name])
1766
        elif field == "sinst_list":
1767
          val = list(node_to_secondary[node.name])
1768
        elif field == "pinst_cnt":
1769
          val = len(node_to_primary[node.name])
1770
        elif field == "sinst_cnt":
1771
          val = len(node_to_secondary[node.name])
1772
        elif field == "pip":
1773
          val = node.primary_ip
1774
        elif field == "sip":
1775
          val = node.secondary_ip
1776
        elif field == "tags":
1777
          val = list(node.GetTags())
1778
        elif field == "serial_no":
1779
          val = node.serial_no
1780
        elif field == "master_candidate":
1781
          val = node.master_candidate
1782
        elif field == "master":
1783
          val = node.name == master_node
1784
        elif field == "offline":
1785
          val = node.offline
1786
        elif self._FIELDS_DYNAMIC.Matches(field):
1787
          val = live_data[node.name].get(field, None)
1788
        else:
1789
          raise errors.ParameterError(field)
1790
        node_output.append(val)
1791
      output.append(node_output)
1792

    
1793
    return output
1794

    
1795

    
1796
class LUQueryNodeVolumes(NoHooksLU):
1797
  """Logical unit for getting volumes on node(s).
1798

1799
  """
1800
  _OP_REQP = ["nodes", "output_fields"]
1801
  REQ_BGL = False
1802
  _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
1803
  _FIELDS_STATIC = utils.FieldSet("node")
1804

    
1805
  def ExpandNames(self):
1806
    _CheckOutputFields(static=self._FIELDS_STATIC,
1807
                       dynamic=self._FIELDS_DYNAMIC,
1808
                       selected=self.op.output_fields)
1809

    
1810
    self.needed_locks = {}
1811
    self.share_locks[locking.LEVEL_NODE] = 1
1812
    if not self.op.nodes:
1813
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
1814
    else:
1815
      self.needed_locks[locking.LEVEL_NODE] = \
1816
        _GetWantedNodes(self, self.op.nodes)
1817

    
1818
  def CheckPrereq(self):
1819
    """Check prerequisites.
1820

1821
    This checks that the fields required are valid output fields.
1822

1823
    """
1824
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
1825

    
1826
  def Exec(self, feedback_fn):
1827
    """Computes the list of nodes and their attributes.
1828

1829
    """
1830
    nodenames = self.nodes
1831
    volumes = self.rpc.call_node_volumes(nodenames)
1832

    
1833
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
1834
             in self.cfg.GetInstanceList()]
1835

    
1836
    lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
1837

    
1838
    output = []
1839
    for node in nodenames:
1840
      if node not in volumes or volumes[node].failed or not volumes[node].data:
1841
        continue
1842

    
1843
      node_vols = volumes[node].data[:]
1844
      node_vols.sort(key=lambda vol: vol['dev'])
1845

    
1846
      for vol in node_vols:
1847
        node_output = []
1848
        for field in self.op.output_fields:
1849
          if field == "node":
1850
            val = node
1851
          elif field == "phys":
1852
            val = vol['dev']
1853
          elif field == "vg":
1854
            val = vol['vg']
1855
          elif field == "name":
1856
            val = vol['name']
1857
          elif field == "size":
1858
            val = int(float(vol['size']))
1859
          elif field == "instance":
1860
            for inst in ilist:
1861
              if node not in lv_by_node[inst]:
1862
                continue
1863
              if vol['name'] in lv_by_node[inst][node]:
1864
                val = inst.name
1865
                break
1866
            else:
1867
              val = '-'
1868
          else:
1869
            raise errors.ParameterError(field)
1870
          node_output.append(str(val))
1871

    
1872
        output.append(node_output)
1873

    
1874
    return output
1875

    
1876

    
1877
class LUAddNode(LogicalUnit):
1878
  """Logical unit for adding node to the cluster.
1879

1880
  """
1881
  HPATH = "node-add"
1882
  HTYPE = constants.HTYPE_NODE
1883
  _OP_REQP = ["node_name"]
1884

    
1885
  def BuildHooksEnv(self):
1886
    """Build hooks env.
1887

1888
    This will run on all nodes before, and on all nodes + the new node after.
1889

1890
    """
1891
    env = {
1892
      "OP_TARGET": self.op.node_name,
1893
      "NODE_NAME": self.op.node_name,
1894
      "NODE_PIP": self.op.primary_ip,
1895
      "NODE_SIP": self.op.secondary_ip,
1896
      }
1897
    nodes_0 = self.cfg.GetNodeList()
1898
    nodes_1 = nodes_0 + [self.op.node_name, ]
1899
    return env, nodes_0, nodes_1
1900

    
1901
  def CheckPrereq(self):
1902
    """Check prerequisites.
1903

1904
    This checks:
1905
     - the new node is not already in the config
1906
     - it is resolvable
1907
     - its parameters (single/dual homed) matches the cluster
1908

1909
    Any errors are signalled by raising errors.OpPrereqError.
1910

1911
    """
1912
    node_name = self.op.node_name
1913
    cfg = self.cfg
1914

    
1915
    dns_data = utils.HostInfo(node_name)
1916

    
1917
    node = dns_data.name
1918
    primary_ip = self.op.primary_ip = dns_data.ip
1919
    secondary_ip = getattr(self.op, "secondary_ip", None)
1920
    if secondary_ip is None:
1921
      secondary_ip = primary_ip
1922
    if not utils.IsValidIP(secondary_ip):
1923
      raise errors.OpPrereqError("Invalid secondary IP given")
1924
    self.op.secondary_ip = secondary_ip
1925

    
1926
    node_list = cfg.GetNodeList()
1927
    if not self.op.readd and node in node_list:
1928
      raise errors.OpPrereqError("Node %s is already in the configuration" %
1929
                                 node)
1930
    elif self.op.readd and node not in node_list:
1931
      raise errors.OpPrereqError("Node %s is not in the configuration" % node)
1932

    
1933
    for existing_node_name in node_list:
1934
      existing_node = cfg.GetNodeInfo(existing_node_name)
1935

    
1936
      if self.op.readd and node == existing_node_name:
1937
        if (existing_node.primary_ip != primary_ip or
1938
            existing_node.secondary_ip != secondary_ip):
1939
          raise errors.OpPrereqError("Readded node doesn't have the same IP"
1940
                                     " address configuration as before")
1941
        continue
1942

    
1943
      if (existing_node.primary_ip == primary_ip or
1944
          existing_node.secondary_ip == primary_ip or
1945
          existing_node.primary_ip == secondary_ip or
1946
          existing_node.secondary_ip == secondary_ip):
1947
        raise errors.OpPrereqError("New node ip address(es) conflict with"
1948
                                   " existing node %s" % existing_node.name)
1949

    
1950
    # check that the type of the node (single versus dual homed) is the
1951
    # same as for the master
1952
    myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
1953
    master_singlehomed = myself.secondary_ip == myself.primary_ip
1954
    newbie_singlehomed = secondary_ip == primary_ip
1955
    if master_singlehomed != newbie_singlehomed:
1956
      if master_singlehomed:
1957
        raise errors.OpPrereqError("The master has no private ip but the"
1958
                                   " new node has one")
1959
      else:
1960
        raise errors.OpPrereqError("The master has a private ip but the"
1961
                                   " new node doesn't have one")
1962

    
1963
    # checks reachablity
1964
    if not utils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
1965
      raise errors.OpPrereqError("Node not reachable by ping")
1966

    
1967
    if not newbie_singlehomed:
1968
      # check reachability from my secondary ip to newbie's secondary ip
1969
      if not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
1970
                           source=myself.secondary_ip):
1971
        raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
1972
                                   " based ping to noded port")
1973

    
1974
    cp_size = self.cfg.GetClusterInfo().candidate_pool_size
1975
    node_info = self.cfg.GetAllNodesInfo().values()
1976
    num_candidates = len([n for n in node_info
1977
                          if n.master_candidate])
1978
    master_candidate = num_candidates < cp_size
1979

    
1980
    self.new_node = objects.Node(name=node,
1981
                                 primary_ip=primary_ip,
1982
                                 secondary_ip=secondary_ip,
1983
                                 master_candidate=master_candidate,
1984
                                 offline=False)
1985

    
1986
  def Exec(self, feedback_fn):
1987
    """Adds the new node to the cluster.
1988

1989
    """
1990
    new_node = self.new_node
1991
    node = new_node.name
1992

    
1993
    # check connectivity
1994
    result = self.rpc.call_version([node])[node]
1995
    result.Raise()
1996
    if result.data:
1997
      if constants.PROTOCOL_VERSION == result.data:
1998
        logging.info("Communication to node %s fine, sw version %s match",
1999
                     node, result.data)
2000
      else:
2001
        raise errors.OpExecError("Version mismatch master version %s,"
2002
                                 " node version %s" %
2003
                                 (constants.PROTOCOL_VERSION, result.data))
2004
    else:
2005
      raise errors.OpExecError("Cannot get version from the new node")
2006

    
2007
    # setup ssh on node
2008
    logging.info("Copy ssh key to node %s", node)
2009
    priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
2010
    keyarray = []
2011
    keyfiles = [constants.SSH_HOST_DSA_PRIV, constants.SSH_HOST_DSA_PUB,
2012
                constants.SSH_HOST_RSA_PRIV, constants.SSH_HOST_RSA_PUB,
2013
                priv_key, pub_key]
2014

    
2015
    for i in keyfiles:
2016
      f = open(i, 'r')
2017
      try:
2018
        keyarray.append(f.read())
2019
      finally:
2020
        f.close()
2021

    
2022
    result = self.rpc.call_node_add(node, keyarray[0], keyarray[1],
2023
                                    keyarray[2],
2024
                                    keyarray[3], keyarray[4], keyarray[5])
2025

    
2026
    if result.failed or not result.data:
2027
      raise errors.OpExecError("Cannot transfer ssh keys to the new node")
2028

    
2029
    # Add node to our /etc/hosts, and add key to known_hosts
2030
    utils.AddHostToEtcHosts(new_node.name)
2031

    
2032
    if new_node.secondary_ip != new_node.primary_ip:
2033
      result = self.rpc.call_node_has_ip_address(new_node.name,
2034
                                                 new_node.secondary_ip)
2035
      if result.failed or not result.data:
2036
        raise errors.OpExecError("Node claims it doesn't have the secondary ip"
2037
                                 " you gave (%s). Please fix and re-run this"
2038
                                 " command." % new_node.secondary_ip)
2039

    
2040
    node_verify_list = [self.cfg.GetMasterNode()]
2041
    node_verify_param = {
2042
      'nodelist': [node],
2043
      # TODO: do a node-net-test as well?
2044
    }
2045

    
2046
    result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
2047
                                       self.cfg.GetClusterName())
2048
    for verifier in node_verify_list:
2049
      if result[verifier].failed or not result[verifier].data:
2050
        raise errors.OpExecError("Cannot communicate with %s's node daemon"
2051
                                 " for remote verification" % verifier)
2052
      if result[verifier].data['nodelist']:
2053
        for failed in result[verifier].data['nodelist']:
2054
          feedback_fn("ssh/hostname verification failed %s -> %s" %
2055
                      (verifier, result[verifier]['nodelist'][failed]))
2056
        raise errors.OpExecError("ssh/hostname verification failed.")
2057

    
2058
    # Distribute updated /etc/hosts and known_hosts to all nodes,
2059
    # including the node just added
2060
    myself = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
2061
    dist_nodes = self.cfg.GetNodeList()
2062
    if not self.op.readd:
2063
      dist_nodes.append(node)
2064
    if myself.name in dist_nodes:
2065
      dist_nodes.remove(myself.name)
2066

    
2067
    logging.debug("Copying hosts and known_hosts to all nodes")
2068
    for fname in (constants.ETC_HOSTS, constants.SSH_KNOWN_HOSTS_FILE):
2069
      result = self.rpc.call_upload_file(dist_nodes, fname)
2070
      for to_node, to_result in result.iteritems():
2071
        if to_result.failed or not to_result.data:
2072
          logging.error("Copy of file %s to node %s failed", fname, to_node)
2073

    
2074
    to_copy = []
2075
    if constants.HT_XEN_HVM in self.cfg.GetClusterInfo().enabled_hypervisors:
2076
      to_copy.append(constants.VNC_PASSWORD_FILE)
2077
    for fname in to_copy:
2078
      result = self.rpc.call_upload_file([node], fname)
2079
      if result[node].failed or not result[node]:
2080
        logging.error("Could not copy file %s to node %s", fname, node)
2081

    
2082
    if self.op.readd:
2083
      self.context.ReaddNode(new_node)
2084
    else:
2085
      self.context.AddNode(new_node)
2086

    
2087

    
2088
class LUSetNodeParams(LogicalUnit):
2089
  """Modifies the parameters of a node.
2090

2091
  """
2092
  HPATH = "node-modify"
2093
  HTYPE = constants.HTYPE_NODE
2094
  _OP_REQP = ["node_name"]
2095
  REQ_BGL = False
2096

    
2097
  def CheckArguments(self):
2098
    node_name = self.cfg.ExpandNodeName(self.op.node_name)
2099
    if node_name is None:
2100
      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name)
2101
    self.op.node_name = node_name
2102
    if not hasattr(self.op, 'master_candidate'):
2103
      raise errors.OpPrereqError("Please pass at least one modification")
2104
    self.op.master_candidate = bool(self.op.master_candidate)
2105

    
2106
  def ExpandNames(self):
2107
    self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
2108

    
2109
  def BuildHooksEnv(self):
2110
    """Build hooks env.
2111

2112
    This runs on the master node.
2113

2114
    """
2115
    env = {
2116
      "OP_TARGET": self.op.node_name,
2117
      "MASTER_CANDIDATE": str(self.op.master_candidate),
2118
      }
2119
    nl = [self.cfg.GetMasterNode(),
2120
          self.op.node_name]
2121
    return env, nl, nl
2122

    
2123
  def CheckPrereq(self):
2124
    """Check prerequisites.
2125

2126
    This only checks the instance list against the existing names.
2127

2128
    """
2129
    force = self.force = self.op.force
2130

    
2131
    if self.op.master_candidate == False:
2132
      if self.op.node_name == self.cfg.GetMasterNode():
2133
        raise errors.OpPrereqError("The master node has to be a"
2134
                                   " master candidate")
2135
      cp_size = self.cfg.GetClusterInfo().candidate_pool_size
2136
      node_info = self.cfg.GetAllNodesInfo().values()
2137
      num_candidates = len([node for node in node_info
2138
                            if node.master_candidate])
2139
      if num_candidates <= cp_size:
2140
        msg = ("Not enough master candidates (desired"
2141
               " %d, new value will be %d)" % (cp_size, num_candidates-1))
2142
        if force:
2143
          self.LogWarning(msg)
2144
        else:
2145
          raise errors.OpPrereqError(msg)
2146

    
2147
    return
2148

    
2149
  def Exec(self, feedback_fn):
2150
    """Modifies a node.
2151

2152
    """
2153
    node = self.cfg.GetNodeInfo(self.op.node_name)
2154

    
2155
    result = []
2156

    
2157
    if self.op.master_candidate is not None:
2158
      node.master_candidate = self.op.master_candidate
2159
      result.append(("master_candidate", str(self.op.master_candidate)))
2160

    
2161
    # this will trigger configuration file update, if needed
2162
    self.cfg.Update(node)
2163
    # this will trigger job queue propagation or cleanup
2164
    if self.op.node_name != self.cfg.GetMasterNode():
2165
      self.context.ReaddNode(node)
2166

    
2167
    return result
2168

    
2169

    
2170
class LUQueryClusterInfo(NoHooksLU):
2171
  """Query cluster configuration.
2172

2173
  """
2174
  _OP_REQP = []
2175
  REQ_BGL = False
2176

    
2177
  def ExpandNames(self):
2178
    self.needed_locks = {}
2179

    
2180
  def CheckPrereq(self):
2181
    """No prerequsites needed for this LU.
2182

2183
    """
2184
    pass
2185

    
2186
  def Exec(self, feedback_fn):
2187
    """Return cluster config.
2188

2189
    """
2190
    cluster = self.cfg.GetClusterInfo()
2191
    result = {
2192
      "software_version": constants.RELEASE_VERSION,
2193
      "protocol_version": constants.PROTOCOL_VERSION,
2194
      "config_version": constants.CONFIG_VERSION,
2195
      "os_api_version": constants.OS_API_VERSION,
2196
      "export_version": constants.EXPORT_VERSION,
2197
      "architecture": (platform.architecture()[0], platform.machine()),
2198
      "name": cluster.cluster_name,
2199
      "master": cluster.master_node,
2200
      "default_hypervisor": cluster.default_hypervisor,
2201
      "enabled_hypervisors": cluster.enabled_hypervisors,
2202
      "hvparams": cluster.hvparams,
2203
      "beparams": cluster.beparams,
2204
      "candidate_pool_size": cluster.candidate_pool_size,
2205
      }
2206

    
2207
    return result
2208

    
2209

    
2210
class LUQueryConfigValues(NoHooksLU):
2211
  """Return configuration values.
2212

2213
  """
2214
  _OP_REQP = []
2215
  REQ_BGL = False
2216
  _FIELDS_DYNAMIC = utils.FieldSet()
2217
  _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag")
2218

    
2219
  def ExpandNames(self):
2220
    self.needed_locks = {}
2221

    
2222
    _CheckOutputFields(static=self._FIELDS_STATIC,
2223
                       dynamic=self._FIELDS_DYNAMIC,
2224
                       selected=self.op.output_fields)
2225

    
2226
  def CheckPrereq(self):
2227
    """No prerequisites.
2228

2229
    """
2230
    pass
2231

    
2232
  def Exec(self, feedback_fn):
2233
    """Dump a representation of the cluster config to the standard output.
2234

2235
    """
2236
    values = []
2237
    for field in self.op.output_fields:
2238
      if field == "cluster_name":
2239
        entry = self.cfg.GetClusterName()
2240
      elif field == "master_node":
2241
        entry = self.cfg.GetMasterNode()
2242
      elif field == "drain_flag":
2243
        entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
2244
      else:
2245
        raise errors.ParameterError(field)
2246
      values.append(entry)
2247
    return values
2248

    
2249

    
2250
class LUActivateInstanceDisks(NoHooksLU):
2251
  """Bring up an instance's disks.
2252

2253
  """
2254
  _OP_REQP = ["instance_name"]
2255
  REQ_BGL = False
2256

    
2257
  def ExpandNames(self):
2258
    self._ExpandAndLockInstance()
2259
    self.needed_locks[locking.LEVEL_NODE] = []
2260
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2261

    
2262
  def DeclareLocks(self, level):
2263
    if level == locking.LEVEL_NODE:
2264
      self._LockInstancesNodes()
2265

    
2266
  def CheckPrereq(self):
2267
    """Check prerequisites.
2268

2269
    This checks that the instance is in the cluster.
2270

2271
    """
2272
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2273
    assert self.instance is not None, \
2274
      "Cannot retrieve locked instance %s" % self.op.instance_name
2275

    
2276
  def Exec(self, feedback_fn):
2277
    """Activate the disks.
2278

2279
    """
2280
    disks_ok, disks_info = _AssembleInstanceDisks(self, self.instance)
2281
    if not disks_ok:
2282
      raise errors.OpExecError("Cannot activate block devices")
2283

    
2284
    return disks_info
2285

    
2286

    
2287
def _AssembleInstanceDisks(lu, instance, ignore_secondaries=False):
2288
  """Prepare the block devices for an instance.
2289

2290
  This sets up the block devices on all nodes.
2291

2292
  @type lu: L{LogicalUnit}
2293
  @param lu: the logical unit on whose behalf we execute
2294
  @type instance: L{objects.Instance}
2295
  @param instance: the instance for whose disks we assemble
2296
  @type ignore_secondaries: boolean
2297
  @param ignore_secondaries: if true, errors on secondary nodes
2298
      won't result in an error return from the function
2299
  @return: False if the operation failed, otherwise a list of
2300
      (host, instance_visible_name, node_visible_name)
2301
      with the mapping from node devices to instance devices
2302

2303
  """
2304
  device_info = []
2305
  disks_ok = True
2306
  iname = instance.name
2307
  # With the two passes mechanism we try to reduce the window of
2308
  # opportunity for the race condition of switching DRBD to primary
2309
  # before handshaking occured, but we do not eliminate it
2310

    
2311
  # The proper fix would be to wait (with some limits) until the
2312
  # connection has been made and drbd transitions from WFConnection
2313
  # into any other network-connected state (Connected, SyncTarget,
2314
  # SyncSource, etc.)
2315

    
2316
  # 1st pass, assemble on all nodes in secondary mode
2317
  for inst_disk in instance.disks:
2318
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
2319
      lu.cfg.SetDiskID(node_disk, node)
2320
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
2321
      if result.failed or not result:
2322
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
2323
                           " (is_primary=False, pass=1)",
2324
                           inst_disk.iv_name, node)
2325
        if not ignore_secondaries:
2326
          disks_ok = False
2327

    
2328
  # FIXME: race condition on drbd migration to primary
2329

    
2330
  # 2nd pass, do only the primary node
2331
  for inst_disk in instance.disks:
2332
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
2333
      if node != instance.primary_node:
2334
        continue
2335
      lu.cfg.SetDiskID(node_disk, node)
2336
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
2337
      if result.failed or not result:
2338
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
2339
                           " (is_primary=True, pass=2)",
2340
                           inst_disk.iv_name, node)
2341
        disks_ok = False
2342
    device_info.append((instance.primary_node, inst_disk.iv_name, result))
2343

    
2344
  # leave the disks configured for the primary node
2345
  # this is a workaround that would be fixed better by
2346
  # improving the logical/physical id handling
2347
  for disk in instance.disks:
2348
    lu.cfg.SetDiskID(disk, instance.primary_node)
2349

    
2350
  return disks_ok, device_info
2351

    
2352

    
2353
def _StartInstanceDisks(lu, instance, force):
2354
  """Start the disks of an instance.
2355

2356
  """
2357
  disks_ok, dummy = _AssembleInstanceDisks(lu, instance,
2358
                                           ignore_secondaries=force)
2359
  if not disks_ok:
2360
    _ShutdownInstanceDisks(lu, instance)
2361
    if force is not None and not force:
2362
      lu.proc.LogWarning("", hint="If the message above refers to a"
2363
                         " secondary node,"
2364
                         " you can retry the operation using '--force'.")
2365
    raise errors.OpExecError("Disk consistency error")
2366

    
2367

    
2368
class LUDeactivateInstanceDisks(NoHooksLU):
2369
  """Shutdown an instance's disks.
2370

2371
  """
2372
  _OP_REQP = ["instance_name"]
2373
  REQ_BGL = False
2374

    
2375
  def ExpandNames(self):
2376
    self._ExpandAndLockInstance()
2377
    self.needed_locks[locking.LEVEL_NODE] = []
2378
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2379

    
2380
  def DeclareLocks(self, level):
2381
    if level == locking.LEVEL_NODE:
2382
      self._LockInstancesNodes()
2383

    
2384
  def CheckPrereq(self):
2385
    """Check prerequisites.
2386

2387
    This checks that the instance is in the cluster.
2388

2389
    """
2390
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2391
    assert self.instance is not None, \
2392
      "Cannot retrieve locked instance %s" % self.op.instance_name
2393

    
2394
  def Exec(self, feedback_fn):
2395
    """Deactivate the disks
2396

2397
    """
2398
    instance = self.instance
2399
    _SafeShutdownInstanceDisks(self, instance)
2400

    
2401

    
2402
def _SafeShutdownInstanceDisks(lu, instance):
2403
  """Shutdown block devices of an instance.
2404

2405
  This function checks if an instance is running, before calling
2406
  _ShutdownInstanceDisks.
2407

2408
  """
2409
  ins_l = lu.rpc.call_instance_list([instance.primary_node],
2410
                                      [instance.hypervisor])
2411
  ins_l = ins_l[instance.primary_node]
2412
  if ins_l.failed or not isinstance(ins_l.data, list):
2413
    raise errors.OpExecError("Can't contact node '%s'" %
2414
                             instance.primary_node)
2415

    
2416
  if instance.name in ins_l.data:
2417
    raise errors.OpExecError("Instance is running, can't shutdown"
2418
                             " block devices.")
2419

    
2420
  _ShutdownInstanceDisks(lu, instance)
2421

    
2422

    
2423
def _ShutdownInstanceDisks(lu, instance, ignore_primary=False):
2424
  """Shutdown block devices of an instance.
2425

2426
  This does the shutdown on all nodes of the instance.
2427

2428
  If the ignore_primary is false, errors on the primary node are
2429
  ignored.
2430

2431
  """
2432
  result = True
2433
  for disk in instance.disks:
2434
    for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
2435
      lu.cfg.SetDiskID(top_disk, node)
2436
      result = lu.rpc.call_blockdev_shutdown(node, top_disk)
2437
      if result.failed or not result.data:
2438
        logging.error("Could not shutdown block device %s on node %s",
2439
                      disk.iv_name, node)
2440
        if not ignore_primary or node != instance.primary_node:
2441
          result = False
2442
  return result
2443

    
2444

    
2445
def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor):
2446
  """Checks if a node has enough free memory.
2447

2448
  This function check if a given node has the needed amount of free
2449
  memory. In case the node has less memory or we cannot get the
2450
  information from the node, this function raise an OpPrereqError
2451
  exception.
2452

2453
  @type lu: C{LogicalUnit}
2454
  @param lu: a logical unit from which we get configuration data
2455
  @type node: C{str}
2456
  @param node: the node to check
2457
  @type reason: C{str}
2458
  @param reason: string to use in the error message
2459
  @type requested: C{int}
2460
  @param requested: the amount of memory in MiB to check for
2461
  @type hypervisor: C{str}
2462
  @param hypervisor: the hypervisor to ask for memory stats
2463
  @raise errors.OpPrereqError: if the node doesn't have enough memory, or
2464
      we cannot check the node
2465

2466
  """
2467
  nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor)
2468
  nodeinfo[node].Raise()
2469
  free_mem = nodeinfo[node].data.get('memory_free')
2470
  if not isinstance(free_mem, int):
2471
    raise errors.OpPrereqError("Can't compute free memory on node %s, result"
2472
                             " was '%s'" % (node, free_mem))
2473
  if requested > free_mem:
2474
    raise errors.OpPrereqError("Not enough memory on node %s for %s:"
2475
                             " needed %s MiB, available %s MiB" %
2476
                             (node, reason, requested, free_mem))
2477

    
2478

    
2479
class LUStartupInstance(LogicalUnit):
2480
  """Starts an instance.
2481

2482
  """
2483
  HPATH = "instance-start"
2484
  HTYPE = constants.HTYPE_INSTANCE
2485
  _OP_REQP = ["instance_name", "force"]
2486
  REQ_BGL = False
2487

    
2488
  def ExpandNames(self):
2489
    self._ExpandAndLockInstance()
2490

    
2491
  def BuildHooksEnv(self):
2492
    """Build hooks env.
2493

2494
    This runs on master, primary and secondary nodes of the instance.
2495

2496
    """
2497
    env = {
2498
      "FORCE": self.op.force,
2499
      }
2500
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2501
    nl = ([self.cfg.GetMasterNode(), self.instance.primary_node] +
2502
          list(self.instance.secondary_nodes))
2503
    return env, nl, nl
2504

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

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

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

    
2515
    bep = self.cfg.GetClusterInfo().FillBE(instance)
2516
    # check bridges existance
2517
    _CheckInstanceBridgesExist(self, instance)
2518

    
2519
    _CheckNodeFreeMemory(self, instance.primary_node,
2520
                         "starting instance %s" % instance.name,
2521
                         bep[constants.BE_MEMORY], instance.hypervisor)
2522

    
2523
  def Exec(self, feedback_fn):
2524
    """Start the instance.
2525

2526
    """
2527
    instance = self.instance
2528
    force = self.op.force
2529
    extra_args = getattr(self.op, "extra_args", "")
2530

    
2531
    self.cfg.MarkInstanceUp(instance.name)
2532

    
2533
    node_current = instance.primary_node
2534

    
2535
    _StartInstanceDisks(self, instance, force)
2536

    
2537
    result = self.rpc.call_instance_start(node_current, instance, extra_args)
2538
    if result.failed or not result.data:
2539
      _ShutdownInstanceDisks(self, instance)
2540
      raise errors.OpExecError("Could not start instance")
2541

    
2542

    
2543
class LURebootInstance(LogicalUnit):
2544
  """Reboot an instance.
2545

2546
  """
2547
  HPATH = "instance-reboot"
2548
  HTYPE = constants.HTYPE_INSTANCE
2549
  _OP_REQP = ["instance_name", "ignore_secondaries", "reboot_type"]
2550
  REQ_BGL = False
2551

    
2552
  def ExpandNames(self):
2553
    if self.op.reboot_type not in [constants.INSTANCE_REBOOT_SOFT,
2554
                                   constants.INSTANCE_REBOOT_HARD,
2555
                                   constants.INSTANCE_REBOOT_FULL]:
2556
      raise errors.ParameterError("reboot type not in [%s, %s, %s]" %
2557
                                  (constants.INSTANCE_REBOOT_SOFT,
2558
                                   constants.INSTANCE_REBOOT_HARD,
2559
                                   constants.INSTANCE_REBOOT_FULL))
2560
    self._ExpandAndLockInstance()
2561

    
2562
  def BuildHooksEnv(self):
2563
    """Build hooks env.
2564

2565
    This runs on master, primary and secondary nodes of the instance.
2566

2567
    """
2568
    env = {
2569
      "IGNORE_SECONDARIES": self.op.ignore_secondaries,
2570
      }
2571
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2572
    nl = ([self.cfg.GetMasterNode(), self.instance.primary_node] +
2573
          list(self.instance.secondary_nodes))
2574
    return env, nl, nl
2575

    
2576
  def CheckPrereq(self):
2577
    """Check prerequisites.
2578

2579
    This checks that the instance is in the cluster.
2580

2581
    """
2582
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2583
    assert self.instance is not None, \
2584
      "Cannot retrieve locked instance %s" % self.op.instance_name
2585

    
2586
    # check bridges existance
2587
    _CheckInstanceBridgesExist(self, instance)
2588

    
2589
  def Exec(self, feedback_fn):
2590
    """Reboot the instance.
2591

2592
    """
2593
    instance = self.instance
2594
    ignore_secondaries = self.op.ignore_secondaries
2595
    reboot_type = self.op.reboot_type
2596
    extra_args = getattr(self.op, "extra_args", "")
2597

    
2598
    node_current = instance.primary_node
2599

    
2600
    if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
2601
                       constants.INSTANCE_REBOOT_HARD]:
2602
      result = self.rpc.call_instance_reboot(node_current, instance,
2603
                                             reboot_type, extra_args)
2604
      if result.failed or not result.data:
2605
        raise errors.OpExecError("Could not reboot instance")
2606
    else:
2607
      if not self.rpc.call_instance_shutdown(node_current, instance):
2608
        raise errors.OpExecError("could not shutdown instance for full reboot")
2609
      _ShutdownInstanceDisks(self, instance)
2610
      _StartInstanceDisks(self, instance, ignore_secondaries)
2611
      result = self.rpc.call_instance_start(node_current, instance, extra_args)
2612
      if result.failed or not result.data:
2613
        _ShutdownInstanceDisks(self, instance)
2614
        raise errors.OpExecError("Could not start instance for full reboot")
2615

    
2616
    self.cfg.MarkInstanceUp(instance.name)
2617

    
2618

    
2619
class LUShutdownInstance(LogicalUnit):
2620
  """Shutdown an instance.
2621

2622
  """
2623
  HPATH = "instance-stop"
2624
  HTYPE = constants.HTYPE_INSTANCE
2625
  _OP_REQP = ["instance_name"]
2626
  REQ_BGL = False
2627

    
2628
  def ExpandNames(self):
2629
    self._ExpandAndLockInstance()
2630

    
2631
  def BuildHooksEnv(self):
2632
    """Build hooks env.
2633

2634
    This runs on master, primary and secondary nodes of the instance.
2635

2636
    """
2637
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2638
    nl = ([self.cfg.GetMasterNode(), self.instance.primary_node] +
2639
          list(self.instance.secondary_nodes))
2640
    return env, nl, nl
2641

    
2642
  def CheckPrereq(self):
2643
    """Check prerequisites.
2644

2645
    This checks that the instance is in the cluster.
2646

2647
    """
2648
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2649
    assert self.instance is not None, \
2650
      "Cannot retrieve locked instance %s" % self.op.instance_name
2651

    
2652
  def Exec(self, feedback_fn):
2653
    """Shutdown the instance.
2654

2655
    """
2656
    instance = self.instance
2657
    node_current = instance.primary_node
2658
    self.cfg.MarkInstanceDown(instance.name)
2659
    result = self.rpc.call_instance_shutdown(node_current, instance)
2660
    if result.failed or not result.data:
2661
      self.proc.LogWarning("Could not shutdown instance")
2662

    
2663
    _ShutdownInstanceDisks(self, instance)
2664

    
2665

    
2666
class LUReinstallInstance(LogicalUnit):
2667
  """Reinstall an instance.
2668

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

    
2675
  def ExpandNames(self):
2676
    self._ExpandAndLockInstance()
2677

    
2678
  def BuildHooksEnv(self):
2679
    """Build hooks env.
2680

2681
    This runs on master, primary and secondary nodes of the instance.
2682

2683
    """
2684
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2685
    nl = ([self.cfg.GetMasterNode(), self.instance.primary_node] +
2686
          list(self.instance.secondary_nodes))
2687
    return env, nl, nl
2688

    
2689
  def CheckPrereq(self):
2690
    """Check prerequisites.
2691

2692
    This checks that the instance is in the cluster and is not running.
2693

2694
    """
2695
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2696
    assert instance is not None, \
2697
      "Cannot retrieve locked instance %s" % self.op.instance_name
2698

    
2699
    if instance.disk_template == constants.DT_DISKLESS:
2700
      raise errors.OpPrereqError("Instance '%s' has no disks" %
2701
                                 self.op.instance_name)
2702
    if instance.status != "down":
2703
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
2704
                                 self.op.instance_name)
2705
    remote_info = self.rpc.call_instance_info(instance.primary_node,
2706
                                              instance.name,
2707
                                              instance.hypervisor)
2708
    if remote_info.failed or remote_info.data:
2709
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
2710
                                 (self.op.instance_name,
2711
                                  instance.primary_node))
2712

    
2713
    self.op.os_type = getattr(self.op, "os_type", None)
2714
    if self.op.os_type is not None:
2715
      # OS verification
2716
      pnode = self.cfg.GetNodeInfo(
2717
        self.cfg.ExpandNodeName(instance.primary_node))
2718
      if pnode is None:
2719
        raise errors.OpPrereqError("Primary node '%s' is unknown" %
2720
                                   self.op.pnode)
2721
      result = self.rpc.call_os_get(pnode.name, self.op.os_type)
2722
      result.Raise()
2723
      if not isinstance(result.data, objects.OS):
2724
        raise errors.OpPrereqError("OS '%s' not in supported OS list for"
2725
                                   " primary node"  % self.op.os_type)
2726

    
2727
    self.instance = instance
2728

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

2732
    """
2733
    inst = self.instance
2734

    
2735
    if self.op.os_type is not None:
2736
      feedback_fn("Changing OS to '%s'..." % self.op.os_type)
2737
      inst.os = self.op.os_type
2738
      self.cfg.Update(inst)
2739

    
2740
    _StartInstanceDisks(self, inst, None)
2741
    try:
2742
      feedback_fn("Running the instance OS create scripts...")
2743
      result = self.rpc.call_instance_os_add(inst.primary_node, inst)
2744
      result.Raise()
2745
      if not result.data:
2746
        raise errors.OpExecError("Could not install OS for instance %s"
2747
                                 " on node %s" %
2748
                                 (inst.name, inst.primary_node))
2749
    finally:
2750
      _ShutdownInstanceDisks(self, inst)
2751

    
2752

    
2753
class LURenameInstance(LogicalUnit):
2754
  """Rename an instance.
2755

2756
  """
2757
  HPATH = "instance-rename"
2758
  HTYPE = constants.HTYPE_INSTANCE
2759
  _OP_REQP = ["instance_name", "new_name"]
2760

    
2761
  def BuildHooksEnv(self):
2762
    """Build hooks env.
2763

2764
    This runs on master, primary and secondary nodes of the instance.
2765

2766
    """
2767
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2768
    env["INSTANCE_NEW_NAME"] = self.op.new_name
2769
    nl = ([self.cfg.GetMasterNode(), self.instance.primary_node] +
2770
          list(self.instance.secondary_nodes))
2771
    return env, nl, nl
2772

    
2773
  def CheckPrereq(self):
2774
    """Check prerequisites.
2775

2776
    This checks that the instance is in the cluster and is not running.
2777

2778
    """
2779
    instance = self.cfg.GetInstanceInfo(
2780
      self.cfg.ExpandInstanceName(self.op.instance_name))
2781
    if instance is None:
2782
      raise errors.OpPrereqError("Instance '%s' not known" %
2783
                                 self.op.instance_name)
2784
    if instance.status != "down":
2785
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
2786
                                 self.op.instance_name)
2787
    remote_info = self.rpc.call_instance_info(instance.primary_node,
2788
                                              instance.name,
2789
                                              instance.hypervisor)
2790
    remote_info.Raise()
2791
    if remote_info.data:
2792
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
2793
                                 (self.op.instance_name,
2794
                                  instance.primary_node))
2795
    self.instance = instance
2796

    
2797
    # new name verification
2798
    name_info = utils.HostInfo(self.op.new_name)
2799

    
2800
    self.op.new_name = new_name = name_info.name
2801
    instance_list = self.cfg.GetInstanceList()
2802
    if new_name in instance_list:
2803
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
2804
                                 new_name)
2805

    
2806
    if not getattr(self.op, "ignore_ip", False):
2807
      if utils.TcpPing(name_info.ip, constants.DEFAULT_NODED_PORT):
2808
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
2809
                                   (name_info.ip, new_name))
2810

    
2811

    
2812
  def Exec(self, feedback_fn):
2813
    """Reinstall the instance.
2814

2815
    """
2816
    inst = self.instance
2817
    old_name = inst.name
2818

    
2819
    if inst.disk_template == constants.DT_FILE:
2820
      old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
2821

    
2822
    self.cfg.RenameInstance(inst.name, self.op.new_name)
2823
    # Change the instance lock. This is definitely safe while we hold the BGL
2824
    self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
2825
    self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
2826

    
2827
    # re-read the instance from the configuration after rename
2828
    inst = self.cfg.GetInstanceInfo(self.op.new_name)
2829

    
2830
    if inst.disk_template == constants.DT_FILE:
2831
      new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
2832
      result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
2833
                                                     old_file_storage_dir,
2834
                                                     new_file_storage_dir)
2835
      result.Raise()
2836
      if not result.data:
2837
        raise errors.OpExecError("Could not connect to node '%s' to rename"
2838
                                 " directory '%s' to '%s' (but the instance"
2839
                                 " has been renamed in Ganeti)" % (
2840
                                 inst.primary_node, old_file_storage_dir,
2841
                                 new_file_storage_dir))
2842

    
2843
      if not result.data[0]:
2844
        raise errors.OpExecError("Could not rename directory '%s' to '%s'"
2845
                                 " (but the instance has been renamed in"
2846
                                 " Ganeti)" % (old_file_storage_dir,
2847
                                               new_file_storage_dir))
2848

    
2849
    _StartInstanceDisks(self, inst, None)
2850
    try:
2851
      result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
2852
                                                 old_name)
2853
      if result.failed or not result.data:
2854
        msg = ("Could not run OS rename script for instance %s on node %s"
2855
               " (but the instance has been renamed in Ganeti)" %
2856
               (inst.name, inst.primary_node))
2857
        self.proc.LogWarning(msg)
2858
    finally:
2859
      _ShutdownInstanceDisks(self, inst)
2860

    
2861

    
2862
class LURemoveInstance(LogicalUnit):
2863
  """Remove an instance.
2864

2865
  """
2866
  HPATH = "instance-remove"
2867
  HTYPE = constants.HTYPE_INSTANCE
2868
  _OP_REQP = ["instance_name", "ignore_failures"]
2869
  REQ_BGL = False
2870

    
2871
  def ExpandNames(self):
2872
    self._ExpandAndLockInstance()
2873
    self.needed_locks[locking.LEVEL_NODE] = []
2874
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2875

    
2876
  def DeclareLocks(self, level):
2877
    if level == locking.LEVEL_NODE:
2878
      self._LockInstancesNodes()
2879

    
2880
  def BuildHooksEnv(self):
2881
    """Build hooks env.
2882

2883
    This runs on master, primary and secondary nodes of the instance.
2884

2885
    """
2886
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2887
    nl = [self.cfg.GetMasterNode()]
2888
    return env, nl, nl
2889

    
2890
  def CheckPrereq(self):
2891
    """Check prerequisites.
2892

2893
    This checks that the instance is in the cluster.
2894

2895
    """
2896
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2897
    assert self.instance is not None, \
2898
      "Cannot retrieve locked instance %s" % self.op.instance_name
2899

    
2900
  def Exec(self, feedback_fn):
2901
    """Remove the instance.
2902

2903
    """
2904
    instance = self.instance
2905
    logging.info("Shutting down instance %s on node %s",
2906
                 instance.name, instance.primary_node)
2907

    
2908
    result = self.rpc.call_instance_shutdown(instance.primary_node, instance)
2909
    if result.failed or not result.data:
2910
      if self.op.ignore_failures:
2911
        feedback_fn("Warning: can't shutdown instance")
2912
      else:
2913
        raise errors.OpExecError("Could not shutdown instance %s on node %s" %
2914
                                 (instance.name, instance.primary_node))
2915

    
2916
    logging.info("Removing block devices for instance %s", instance.name)
2917

    
2918
    if not _RemoveDisks(self, instance):
2919
      if self.op.ignore_failures:
2920
        feedback_fn("Warning: can't remove instance's disks")
2921
      else:
2922
        raise errors.OpExecError("Can't remove instance's disks")
2923

    
2924
    logging.info("Removing instance %s out of cluster config", instance.name)
2925

    
2926
    self.cfg.RemoveInstance(instance.name)
2927
    self.remove_locks[locking.LEVEL_INSTANCE] = instance.name
2928

    
2929

    
2930
class LUQueryInstances(NoHooksLU):
2931
  """Logical unit for querying instances.
2932

2933
  """
2934
  _OP_REQP = ["output_fields", "names"]
2935
  REQ_BGL = False
2936
  _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
2937
                                    "admin_state", "admin_ram",
2938
                                    "disk_template", "ip", "mac", "bridge",
2939
                                    "sda_size", "sdb_size", "vcpus", "tags",
2940
                                    "network_port", "beparams",
2941
                                    "(disk).(size)/([0-9]+)",
2942
                                    "(disk).(sizes)",
2943
                                    "(nic).(mac|ip|bridge)/([0-9]+)",
2944
                                    "(nic).(macs|ips|bridges)",
2945
                                    "(disk|nic).(count)",
2946
                                    "serial_no", "hypervisor", "hvparams",] +
2947
                                  ["hv/%s" % name
2948
                                   for name in constants.HVS_PARAMETERS] +
2949
                                  ["be/%s" % name
2950
                                   for name in constants.BES_PARAMETERS])
2951
  _FIELDS_DYNAMIC = utils.FieldSet("oper_state", "oper_ram", "status")
2952

    
2953

    
2954
  def ExpandNames(self):
2955
    _CheckOutputFields(static=self._FIELDS_STATIC,
2956
                       dynamic=self._FIELDS_DYNAMIC,
2957
                       selected=self.op.output_fields)
2958

    
2959
    self.needed_locks = {}
2960
    self.share_locks[locking.LEVEL_INSTANCE] = 1
2961
    self.share_locks[locking.LEVEL_NODE] = 1
2962

    
2963
    if self.op.names:
2964
      self.wanted = _GetWantedInstances(self, self.op.names)
2965
    else:
2966
      self.wanted = locking.ALL_SET
2967

    
2968
    self.do_locking = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
2969
    if self.do_locking:
2970
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
2971
      self.needed_locks[locking.LEVEL_NODE] = []
2972
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2973

    
2974
  def DeclareLocks(self, level):
2975
    if level == locking.LEVEL_NODE and self.do_locking:
2976
      self._LockInstancesNodes()
2977

    
2978
  def CheckPrereq(self):
2979
    """Check prerequisites.
2980

2981
    """
2982
    pass
2983

    
2984
  def Exec(self, feedback_fn):
2985
    """Computes the list of nodes and their attributes.
2986

2987
    """
2988
    all_info = self.cfg.GetAllInstancesInfo()
2989
    if self.do_locking:
2990
      instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
2991
    elif self.wanted != locking.ALL_SET:
2992
      instance_names = self.wanted
2993
      missing = set(instance_names).difference(all_info.keys())
2994
      if missing:
2995
        raise errors.OpExecError(
2996
          "Some instances were removed before retrieving their data: %s"
2997
          % missing)
2998
    else:
2999
      instance_names = all_info.keys()
3000

    
3001
    instance_names = utils.NiceSort(instance_names)
3002
    instance_list = [all_info[iname] for iname in instance_names]
3003

    
3004
    # begin data gathering
3005

    
3006
    nodes = frozenset([inst.primary_node for inst in instance_list])
3007
    hv_list = list(set([inst.hypervisor for inst in instance_list]))
3008

    
3009
    bad_nodes = []
3010
    if self.do_locking:
3011
      live_data = {}
3012
      node_data = self.rpc.call_all_instances_info(nodes, hv_list)
3013
      for name in nodes:
3014
        result = node_data[name]
3015
        if result.failed:
3016
          bad_nodes.append(name)
3017
        else:
3018
          if result.data:
3019
            live_data.update(result.data)
3020
            # else no instance is alive
3021
    else:
3022
      live_data = dict([(name, {}) for name in instance_names])
3023

    
3024
    # end data gathering
3025

    
3026
    HVPREFIX = "hv/"
3027
    BEPREFIX = "be/"
3028
    output = []
3029
    for instance in instance_list:
3030
      iout = []
3031
      i_hv = self.cfg.GetClusterInfo().FillHV(instance)
3032
      i_be = self.cfg.GetClusterInfo().FillBE(instance)
3033
      for field in self.op.output_fields:
3034
        st_match = self._FIELDS_STATIC.Matches(field)
3035
        if field == "name":
3036
          val = instance.name
3037
        elif field == "os":
3038
          val = instance.os
3039
        elif field == "pnode":
3040
          val = instance.primary_node
3041
        elif field == "snodes":
3042
          val = list(instance.secondary_nodes)
3043
        elif field == "admin_state":
3044
          val = (instance.status != "down")
3045
        elif field == "oper_state":
3046
          if instance.primary_node in bad_nodes:
3047
            val = None
3048
          else:
3049
            val = bool(live_data.get(instance.name))
3050
        elif field == "status":
3051
          if instance.primary_node in bad_nodes:
3052
            val = "ERROR_nodedown"
3053
          else:
3054
            running = bool(live_data.get(instance.name))
3055
            if running:
3056
              if instance.status != "down":
3057
                val = "running"
3058
              else:
3059
                val = "ERROR_up"
3060
            else:
3061
              if instance.status != "down":
3062
                val = "ERROR_down"
3063
              else:
3064
                val = "ADMIN_down"
3065
        elif field == "oper_ram":
3066
          if instance.primary_node in bad_nodes:
3067
            val = None
3068
          elif instance.name in live_data:
3069
            val = live_data[instance.name].get("memory", "?")
3070
          else:
3071
            val = "-"
3072
        elif field == "disk_template":
3073
          val = instance.disk_template
3074
        elif field == "ip":
3075
          val = instance.nics[0].ip
3076
        elif field == "bridge":
3077
          val = instance.nics[0].bridge
3078
        elif field == "mac":
3079
          val = instance.nics[0].mac
3080
        elif field == "sda_size" or field == "sdb_size":
3081
          idx = ord(field[2]) - ord('a')
3082
          try:
3083
            val = instance.FindDisk(idx).size
3084
          except errors.OpPrereqError:
3085
            val = None
3086
        elif field == "tags":
3087
          val = list(instance.GetTags())
3088
        elif field == "serial_no":
3089
          val = instance.serial_no
3090
        elif field == "network_port":
3091
          val = instance.network_port
3092
        elif field == "hypervisor":
3093
          val = instance.hypervisor
3094
        elif field == "hvparams":
3095
          val = i_hv
3096
        elif (field.startswith(HVPREFIX) and
3097
              field[len(HVPREFIX):] in constants.HVS_PARAMETERS):
3098
          val = i_hv.get(field[len(HVPREFIX):], None)
3099
        elif field == "beparams":
3100
          val = i_be
3101
        elif (field.startswith(BEPREFIX) and
3102
              field[len(BEPREFIX):] in constants.BES_PARAMETERS):
3103
          val = i_be.get(field[len(BEPREFIX):], None)
3104
        elif st_match and st_match.groups():
3105
          # matches a variable list
3106
          st_groups = st_match.groups()
3107
          if st_groups and st_groups[0] == "disk":
3108
            if st_groups[1] == "count":
3109
              val = len(instance.disks)
3110
            elif st_groups[1] == "sizes":
3111
              val = [disk.size for disk in instance.disks]
3112
            elif st_groups[1] == "size":
3113
              try:
3114
                val = instance.FindDisk(st_groups[2]).size
3115
              except errors.OpPrereqError:
3116
                val = None
3117
            else:
3118
              assert False, "Unhandled disk parameter"
3119
          elif st_groups[0] == "nic":
3120
            if st_groups[1] == "count":
3121
              val = len(instance.nics)
3122
            elif st_groups[1] == "macs":
3123
              val = [nic.mac for nic in instance.nics]
3124
            elif st_groups[1] == "ips":
3125
              val = [nic.ip for nic in instance.nics]
3126
            elif st_groups[1] == "bridges":
3127
              val = [nic.bridge for nic in instance.nics]
3128
            else:
3129
              # index-based item
3130
              nic_idx = int(st_groups[2])
3131
              if nic_idx >= len(instance.nics):
3132
                val = None
3133
              else:
3134
                if st_groups[1] == "mac":
3135
                  val = instance.nics[nic_idx].mac
3136
                elif st_groups[1] == "ip":
3137
                  val = instance.nics[nic_idx].ip
3138
                elif st_groups[1] == "bridge":
3139
                  val = instance.nics[nic_idx].bridge
3140
                else:
3141
                  assert False, "Unhandled NIC parameter"
3142
          else:
3143
            assert False, "Unhandled variable parameter"
3144
        else:
3145
          raise errors.ParameterError(field)
3146
        iout.append(val)
3147
      output.append(iout)
3148

    
3149
    return output
3150

    
3151

    
3152
class LUFailoverInstance(LogicalUnit):
3153
  """Failover an instance.
3154

3155
  """
3156
  HPATH = "instance-failover"
3157
  HTYPE = constants.HTYPE_INSTANCE
3158
  _OP_REQP = ["instance_name", "ignore_consistency"]
3159
  REQ_BGL = False
3160

    
3161
  def ExpandNames(self):
3162
    self._ExpandAndLockInstance()
3163
    self.needed_locks[locking.LEVEL_NODE] = []
3164
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3165

    
3166
  def DeclareLocks(self, level):
3167
    if level == locking.LEVEL_NODE:
3168
      self._LockInstancesNodes()
3169

    
3170
  def BuildHooksEnv(self):
3171
    """Build hooks env.
3172

3173
    This runs on master, primary and secondary nodes of the instance.
3174

3175
    """
3176
    env = {
3177
      "IGNORE_CONSISTENCY": self.op.ignore_consistency,
3178
      }
3179
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
3180
    nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
3181
    return env, nl, nl
3182

    
3183
  def CheckPrereq(self):
3184
    """Check prerequisites.
3185

3186
    This checks that the instance is in the cluster.
3187

3188
    """
3189
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3190
    assert self.instance is not None, \
3191
      "Cannot retrieve locked instance %s" % self.op.instance_name
3192

    
3193
    bep = self.cfg.GetClusterInfo().FillBE(instance)
3194
    if instance.disk_template not in constants.DTS_NET_MIRROR:
3195
      raise errors.OpPrereqError("Instance's disk layout is not"
3196
                                 " network mirrored, cannot failover.")
3197

    
3198
    secondary_nodes = instance.secondary_nodes
3199
    if not secondary_nodes:
3200
      raise errors.ProgrammerError("no secondary node but using "
3201
                                   "a mirrored disk template")
3202

    
3203
    target_node = secondary_nodes[0]
3204
    # check memory requirements on the secondary node
3205
    _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
3206
                         instance.name, bep[constants.BE_MEMORY],
3207
                         instance.hypervisor)
3208

    
3209
    # check bridge existance
3210
    brlist = [nic.bridge for nic in instance.nics]
3211
    result = self.rpc.call_bridges_exist(target_node, brlist)
3212
    result.Raise()
3213
    if not result.data:
3214
      raise errors.OpPrereqError("One or more target bridges %s does not"
3215
                                 " exist on destination node '%s'" %
3216
                                 (brlist, target_node))
3217

    
3218
  def Exec(self, feedback_fn):
3219
    """Failover an instance.
3220

3221
    The failover is done by shutting it down on its present node and
3222
    starting it on the secondary.
3223

3224
    """
3225
    instance = self.instance
3226

    
3227
    source_node = instance.primary_node
3228
    target_node = instance.secondary_nodes[0]
3229

    
3230
    feedback_fn("* checking disk consistency between source and target")
3231
    for dev in instance.disks:
3232
      # for drbd, these are drbd over lvm
3233
      if not _CheckDiskConsistency(self, dev, target_node, False):
3234
        if instance.status == "up" and not self.op.ignore_consistency:
3235
          raise errors.OpExecError("Disk %s is degraded on target node,"
3236
                                   " aborting failover." % dev.iv_name)
3237

    
3238
    feedback_fn("* shutting down instance on source node")
3239
    logging.info("Shutting down instance %s on node %s",
3240
                 instance.name, source_node)
3241

    
3242
    result = self.rpc.call_instance_shutdown(source_node, instance)
3243
    if result.failed or not result.data:
3244
      if self.op.ignore_consistency:
3245
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
3246
                             " Proceeding"
3247
                             " anyway. Please make sure node %s is down",
3248
                             instance.name, source_node, source_node)
3249
      else:
3250
        raise errors.OpExecError("Could not shutdown instance %s on node %s" %
3251
                                 (instance.name, source_node))
3252

    
3253
    feedback_fn("* deactivating the instance's disks on source node")
3254
    if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
3255
      raise errors.OpExecError("Can't shut down the instance's disks.")
3256

    
3257
    instance.primary_node = target_node
3258
    # distribute new instance config to the other nodes
3259
    self.cfg.Update(instance)
3260

    
3261
    # Only start the instance if it's marked as up
3262
    if instance.status == "up":
3263
      feedback_fn("* activating the instance's disks on target node")
3264
      logging.info("Starting instance %s on node %s",
3265
                   instance.name, target_node)
3266

    
3267
      disks_ok, dummy = _AssembleInstanceDisks(self, instance,
3268
                                               ignore_secondaries=True)
3269
      if not disks_ok:
3270
        _ShutdownInstanceDisks(self, instance)
3271
        raise errors.OpExecError("Can't activate the instance's disks")
3272

    
3273
      feedback_fn("* starting the instance on the target node")
3274
      result = self.rpc.call_instance_start(target_node, instance, None)
3275
      if result.failed or not result.data:
3276
        _ShutdownInstanceDisks(self, instance)
3277
        raise errors.OpExecError("Could not start instance %s on node %s." %
3278
                                 (instance.name, target_node))
3279

    
3280

    
3281
def _CreateBlockDevOnPrimary(lu, node, instance, device, info):
3282
  """Create a tree of block devices on the primary node.
3283

3284
  This always creates all devices.
3285

3286
  """
3287
  if device.children:
3288
    for child in device.children:
3289
      if not _CreateBlockDevOnPrimary(lu, node, instance, child, info):
3290
        return False
3291

    
3292
  lu.cfg.SetDiskID(device, node)
3293
  new_id = lu.rpc.call_blockdev_create(node, device, device.size,
3294
                                       instance.name, True, info)
3295
  if new_id.failed or not new_id.data:
3296
    return False
3297
  if device.physical_id is None:
3298
    device.physical_id = new_id
3299
  return True
3300

    
3301

    
3302
def _CreateBlockDevOnSecondary(lu, node, instance, device, force, info):
3303
  """Create a tree of block devices on a secondary node.
3304

3305
  If this device type has to be created on secondaries, create it and
3306
  all its children.
3307

3308
  If not, just recurse to children keeping the same 'force' value.
3309

3310
  """
3311
  if device.CreateOnSecondary():
3312
    force = True
3313
  if device.children:
3314
    for child in device.children:
3315
      if not _CreateBlockDevOnSecondary(lu, node, instance,
3316
                                        child, force, info):
3317
        return False
3318

    
3319
  if not force:
3320
    return True
3321
  lu.cfg.SetDiskID(device, node)
3322
  new_id = lu.rpc.call_blockdev_create(node, device, device.size,
3323
                                       instance.name, False, info)
3324
  if new_id.failed or not new_id.data:
3325
    return False
3326
  if device.physical_id is None:
3327
    device.physical_id = new_id
3328
  return True
3329

    
3330

    
3331
def _GenerateUniqueNames(lu, exts):
3332
  """Generate a suitable LV name.
3333

3334
  This will generate a logical volume name for the given instance.
3335

3336
  """
3337
  results = []
3338
  for val in exts:
3339
    new_id = lu.cfg.GenerateUniqueID()
3340
    results.append("%s%s" % (new_id, val))
3341
  return results
3342

    
3343

    
3344
def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
3345
                         p_minor, s_minor):
3346
  """Generate a drbd8 device complete with its children.
3347

3348
  """
3349
  port = lu.cfg.AllocatePort()
3350
  vgname = lu.cfg.GetVGName()
3351
  shared_secret = lu.cfg.GenerateDRBDSecret()
3352
  dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
3353
                          logical_id=(vgname, names[0]))
3354
  dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
3355
                          logical_id=(vgname, names[1]))
3356
  drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
3357
                          logical_id=(primary, secondary, port,
3358
                                      p_minor, s_minor,
3359
                                      shared_secret),
3360
                          children=[dev_data, dev_meta],
3361
                          iv_name=iv_name)
3362
  return drbd_dev
3363

    
3364

    
3365
def _GenerateDiskTemplate(lu, template_name,
3366
                          instance_name, primary_node,
3367
                          secondary_nodes, disk_info,
3368
                          file_storage_dir, file_driver,
3369
                          base_index):
3370
  """Generate the entire disk layout for a given template type.
3371

3372
  """
3373
  #TODO: compute space requirements
3374

    
3375
  vgname = lu.cfg.GetVGName()
3376
  disk_count = len(disk_info)
3377
  disks = []
3378
  if template_name == constants.DT_DISKLESS:
3379
    pass
3380
  elif template_name == constants.DT_PLAIN:
3381
    if len(secondary_nodes) != 0:
3382
      raise errors.ProgrammerError("Wrong template configuration")
3383

    
3384
    names = _GenerateUniqueNames(lu, [".disk%d" % i
3385
                                      for i in range(disk_count)])
3386
    for idx, disk in enumerate(disk_info):
3387
      disk_index = idx + base_index
3388
      disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
3389
                              logical_id=(vgname, names[idx]),
3390
                              iv_name="disk/%d" % disk_index)
3391
      disks.append(disk_dev)
3392
  elif template_name == constants.DT_DRBD8:
3393
    if len(secondary_nodes) != 1:
3394
      raise errors.ProgrammerError("Wrong template configuration")
3395
    remote_node = secondary_nodes[0]
3396
    minors = lu.cfg.AllocateDRBDMinor(
3397
      [primary_node, remote_node] * len(disk_info), instance_name)
3398

    
3399
    names = _GenerateUniqueNames(lu,
3400
                                 [".disk%d_%s" % (i, s)
3401
                                  for i in range(disk_count)
3402
                                  for s in ("data", "meta")
3403
                                  ])
3404
    for idx, disk in enumerate(disk_info):
3405
      disk_index = idx + base_index
3406
      disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
3407
                                      disk["size"], names[idx*2:idx*2+2],
3408
                                      "disk/%d" % disk_index,
3409
                                      minors[idx*2], minors[idx*2+1])
3410
      disks.append(disk_dev)
3411
  elif template_name == constants.DT_FILE:
3412
    if len(secondary_nodes) != 0:
3413
      raise errors.ProgrammerError("Wrong template configuration")
3414

    
3415
    for idx, disk in enumerate(disk_info):
3416
      disk_index = idx + base_index
3417
      disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
3418
                              iv_name="disk/%d" % disk_index,
3419
                              logical_id=(file_driver,
3420
                                          "%s/disk%d" % (file_storage_dir,
3421
                                                         idx)))
3422
      disks.append(disk_dev)
3423
  else:
3424
    raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
3425
  return disks
3426

    
3427

    
3428
def _GetInstanceInfoText(instance):
3429
  """Compute that text that should be added to the disk's metadata.
3430

3431
  """
3432
  return "originstname+%s" % instance.name
3433

    
3434

    
3435
def _CreateDisks(lu, instance):
3436
  """Create all disks for an instance.
3437

3438
  This abstracts away some work from AddInstance.
3439

3440
  @type lu: L{LogicalUnit}
3441
  @param lu: the logical unit on whose behalf we execute
3442
  @type instance: L{objects.Instance}
3443
  @param instance: the instance whose disks we should create
3444
  @rtype: boolean
3445
  @return: the success of the creation
3446

3447
  """
3448
  info = _GetInstanceInfoText(instance)
3449

    
3450
  if instance.disk_template == constants.DT_FILE:
3451
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
3452
    result = lu.rpc.call_file_storage_dir_create(instance.primary_node,
3453
                                                 file_storage_dir)
3454

    
3455
    if result.failed or not result.data:
3456
      logging.error("Could not connect to node '%s'", instance.primary_node)
3457
      return False
3458

    
3459
    if not result.data[0]:
3460
      logging.error("Failed to create directory '%s'", file_storage_dir)
3461
      return False
3462

    
3463
  # Note: this needs to be kept in sync with adding of disks in
3464
  # LUSetInstanceParams
3465
  for device in instance.disks:
3466
    logging.info("Creating volume %s for instance %s",
3467
                 device.iv_name, instance.name)
3468
    #HARDCODE
3469
    for secondary_node in instance.secondary_nodes:
3470
      if not _CreateBlockDevOnSecondary(lu, secondary_node, instance,
3471
                                        device, False, info):
3472
        logging.error("Failed to create volume %s (%s) on secondary node %s!",
3473
                      device.iv_name, device, secondary_node)
3474
        return False
3475
    #HARDCODE
3476
    if not _CreateBlockDevOnPrimary(lu, instance.primary_node,
3477
                                    instance, device, info):
3478
      logging.error("Failed to create volume %s on primary!", device.iv_name)
3479
      return False
3480

    
3481
  return True
3482

    
3483

    
3484
def _RemoveDisks(lu, instance):
3485
  """Remove all disks for an instance.
3486

3487
  This abstracts away some work from `AddInstance()` and
3488
  `RemoveInstance()`. Note that in case some of the devices couldn't
3489
  be removed, the removal will continue with the other ones (compare
3490
  with `_CreateDisks()`).
3491

3492
  @type lu: L{LogicalUnit}
3493
  @param lu: the logical unit on whose behalf we execute
3494
  @type instance: L{objects.Instance}
3495
  @param instance: the instance whose disks we should remove
3496
  @rtype: boolean
3497
  @return: the success of the removal
3498

3499
  """
3500
  logging.info("Removing block devices for instance %s", instance.name)
3501

    
3502
  result = True
3503
  for device in instance.disks:
3504
    for node, disk in device.ComputeNodeTree(instance.primary_node):
3505
      lu.cfg.SetDiskID(disk, node)
3506
      result = lu.rpc.call_blockdev_remove(node, disk)
3507
      if result.failed or not result.data:
3508
        lu.proc.LogWarning("Could not remove block device %s on node %s,"
3509
                           " continuing anyway", device.iv_name, node)
3510
        result = False
3511

    
3512
  if instance.disk_template == constants.DT_FILE:
3513
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
3514
    result = lu.rpc.call_file_storage_dir_remove(instance.primary_node,
3515
                                                 file_storage_dir)
3516
    if result.failed or not result.data:
3517
      logging.error("Could not remove directory '%s'", file_storage_dir)
3518
      result = False
3519

    
3520
  return result
3521

    
3522

    
3523
def _ComputeDiskSize(disk_template, disks):
3524
  """Compute disk size requirements in the volume group
3525

3526
  """
3527
  # Required free disk space as a function of disk and swap space
3528
  req_size_dict = {
3529
    constants.DT_DISKLESS: None,
3530
    constants.DT_PLAIN: sum(d["size"] for d in disks),
3531
    # 128 MB are added for drbd metadata for each disk
3532
    constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
3533
    constants.DT_FILE: None,
3534
  }
3535

    
3536
  if disk_template not in req_size_dict:
3537
    raise errors.ProgrammerError("Disk template '%s' size requirement"
3538
                                 " is unknown" %  disk_template)
3539

    
3540
  return req_size_dict[disk_template]
3541

    
3542

    
3543
def _CheckHVParams(lu, nodenames, hvname, hvparams):
3544
  """Hypervisor parameter validation.
3545

3546
  This function abstract the hypervisor parameter validation to be
3547
  used in both instance create and instance modify.
3548

3549
  @type lu: L{LogicalUnit}
3550
  @param lu: the logical unit for which we check
3551
  @type nodenames: list
3552
  @param nodenames: the list of nodes on which we should check
3553
  @type hvname: string
3554
  @param hvname: the name of the hypervisor we should use
3555
  @type hvparams: dict
3556
  @param hvparams: the parameters which we need to check
3557
  @raise errors.OpPrereqError: if the parameters are not valid
3558

3559
  """
3560
  hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
3561
                                                  hvname,
3562
                                                  hvparams)
3563
  for node in nodenames:
3564
    info = hvinfo[node]
3565
    info.Raise()
3566
    if not info.data or not isinstance(info.data, (tuple, list)):
3567
      raise errors.OpPrereqError("Cannot get current information"
3568
                                 " from node '%s' (%s)" % (node, info.data))
3569
    if not info.data[0]:
3570
      raise errors.OpPrereqError("Hypervisor parameter validation failed:"
3571
                                 " %s" % info.data[1])
3572

    
3573

    
3574
class LUCreateInstance(LogicalUnit):
3575
  """Create an instance.
3576

3577
  """
3578
  HPATH = "instance-add"
3579
  HTYPE = constants.HTYPE_INSTANCE
3580
  _OP_REQP = ["instance_name", "disks", "disk_template",
3581
              "mode", "start",
3582
              "wait_for_sync", "ip_check", "nics",
3583
              "hvparams", "beparams"]
3584
  REQ_BGL = False
3585

    
3586
  def _ExpandNode(self, node):
3587
    """Expands and checks one node name.
3588

3589
    """
3590
    node_full = self.cfg.ExpandNodeName(node)
3591
    if node_full is None:
3592
      raise errors.OpPrereqError("Unknown node %s" % node)
3593
    return node_full
3594

    
3595
  def ExpandNames(self):
3596
    """ExpandNames for CreateInstance.
3597

3598
    Figure out the right locks for instance creation.
3599

3600
    """
3601
    self.needed_locks = {}
3602

    
3603
    # set optional parameters to none if they don't exist
3604
    for attr in ["pnode", "snode", "iallocator", "hypervisor"]:
3605
      if not hasattr(self.op, attr):
3606
        setattr(self.op, attr, None)
3607

    
3608
    # cheap checks, mostly valid constants given
3609

    
3610
    # verify creation mode
3611
    if self.op.mode not in (constants.INSTANCE_CREATE,
3612
                            constants.INSTANCE_IMPORT):
3613
      raise errors.OpPrereqError("Invalid instance creation mode '%s'" %
3614
                                 self.op.mode)
3615

    
3616
    # disk template and mirror node verification
3617
    if self.op.disk_template not in constants.DISK_TEMPLATES:
3618
      raise errors.OpPrereqError("Invalid disk template name")
3619

    
3620
    if self.op.hypervisor is None:
3621
      self.op.hypervisor = self.cfg.GetHypervisorType()
3622

    
3623
    cluster = self.cfg.GetClusterInfo()
3624
    enabled_hvs = cluster.enabled_hypervisors
3625
    if self.op.hypervisor not in enabled_hvs:
3626
      raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
3627
                                 " cluster (%s)" % (self.op.hypervisor,
3628
                                  ",".join(enabled_hvs)))
3629

    
3630
    # check hypervisor parameter syntax (locally)
3631

    
3632
    filled_hvp = cluster.FillDict(cluster.hvparams[self.op.hypervisor],
3633
                                  self.op.hvparams)
3634
    hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
3635
    hv_type.CheckParameterSyntax(filled_hvp)
3636

    
3637
    # fill and remember the beparams dict
3638
    utils.CheckBEParams(self.op.beparams)
3639
    self.be_full = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
3640
                                    self.op.beparams)
3641

    
3642
    #### instance parameters check
3643

    
3644
    # instance name verification
3645
    hostname1 = utils.HostInfo(self.op.instance_name)
3646
    self.op.instance_name = instance_name = hostname1.name
3647

    
3648
    # this is just a preventive check, but someone might still add this
3649
    # instance in the meantime, and creation will fail at lock-add time
3650
    if instance_name in self.cfg.GetInstanceList():
3651
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
3652
                                 instance_name)
3653

    
3654
    self.add_locks[locking.LEVEL_INSTANCE] = instance_name
3655

    
3656
    # NIC buildup
3657
    self.nics = []
3658
    for nic in self.op.nics:
3659
      # ip validity checks
3660
      ip = nic.get("ip", None)
3661
      if ip is None or ip.lower() == "none":
3662
        nic_ip = None
3663
      elif ip.lower() == constants.VALUE_AUTO:
3664
        nic_ip = hostname1.ip
3665
      else:
3666
        if not utils.IsValidIP(ip):
3667
          raise errors.OpPrereqError("Given IP address '%s' doesn't look"
3668
                                     " like a valid IP" % ip)
3669
        nic_ip = ip
3670

    
3671
      # MAC address verification
3672
      mac = nic.get("mac", constants.VALUE_AUTO)
3673
      if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
3674
        if not utils.IsValidMac(mac.lower()):
3675
          raise errors.OpPrereqError("Invalid MAC address specified: %s" %
3676
                                     mac)
3677
      # bridge verification
3678
      bridge = nic.get("bridge", self.cfg.GetDefBridge())
3679
      self.nics.append(objects.NIC(mac=mac, ip=nic_ip, bridge=bridge))
3680

    
3681
    # disk checks/pre-build
3682
    self.disks = []
3683
    for disk in self.op.disks:
3684
      mode = disk.get("mode", constants.DISK_RDWR)
3685
      if mode not in constants.DISK_ACCESS_SET:
3686
        raise errors.OpPrereqError("Invalid disk access mode '%s'" %
3687
                                   mode)
3688
      size = disk.get("size", None)
3689
      if size is None:
3690
        raise errors.OpPrereqError("Missing disk size")
3691
      try:
3692
        size = int(size)
3693
      except ValueError:
3694
        raise errors.OpPrereqError("Invalid disk size '%s'" % size)
3695
      self.disks.append({"size": size, "mode": mode})
3696

    
3697
    # used in CheckPrereq for ip ping check
3698
    self.check_ip = hostname1.ip
3699

    
3700
    # file storage checks
3701
    if (self.op.file_driver and
3702
        not self.op.file_driver in constants.FILE_DRIVER):
3703
      raise errors.OpPrereqError("Invalid file driver name '%s'" %
3704
                                 self.op.file_driver)
3705

    
3706
    if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
3707
      raise errors.OpPrereqError("File storage directory path not absolute")
3708

    
3709
    ### Node/iallocator related checks
3710
    if [self.op.iallocator, self.op.pnode].count(None) != 1:
3711
      raise errors.OpPrereqError("One and only one of iallocator and primary"
3712
                                 " node must be given")
3713

    
3714
    if self.op.iallocator:
3715
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3716
    else:
3717
      self.op.pnode = self._ExpandNode(self.op.pnode)
3718
      nodelist = [self.op.pnode]
3719
      if self.op.snode is not None:
3720
        self.op.snode = self._ExpandNode(self.op.snode)
3721
        nodelist.append(self.op.snode)
3722
      self.needed_locks[locking.LEVEL_NODE] = nodelist
3723

    
3724
    # in case of import lock the source node too
3725
    if self.op.mode == constants.INSTANCE_IMPORT:
3726
      src_node = getattr(self.op, "src_node", None)
3727
      src_path = getattr(self.op, "src_path", None)
3728

    
3729
      if src_path is None:
3730
        self.op.src_path = src_path = self.op.instance_name
3731

    
3732
      if src_node is None:
3733
        self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3734
        self.op.src_node = None
3735
        if os.path.isabs(src_path):
3736
          raise errors.OpPrereqError("Importing an instance from an absolute"
3737
                                     " path requires a source node option.")
3738
      else:
3739
        self.op.src_node = src_node = self._ExpandNode(src_node)
3740
        if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
3741
          self.needed_locks[locking.LEVEL_NODE].append(src_node)
3742
        if not os.path.isabs(src_path):
3743
          self.op.src_path = src_path = \
3744
            os.path.join(constants.EXPORT_DIR, src_path)
3745

    
3746
    else: # INSTANCE_CREATE
3747
      if getattr(self.op, "os_type", None) is None:
3748
        raise errors.OpPrereqError("No guest OS specified")
3749

    
3750
  def _RunAllocator(self):
3751
    """Run the allocator based on input opcode.
3752

3753
    """
3754
    nics = [n.ToDict() for n in self.nics]
3755
    ial = IAllocator(self,
3756
                     mode=constants.IALLOCATOR_MODE_ALLOC,
3757
                     name=self.op.instance_name,
3758
                     disk_template=self.op.disk_template,
3759
                     tags=[],
3760
                     os=self.op.os_type,
3761
                     vcpus=self.be_full[constants.BE_VCPUS],
3762
                     mem_size=self.be_full[constants.BE_MEMORY],
3763
                     disks=self.disks,
3764
                     nics=nics,
3765
                     hypervisor=self.op.hypervisor,
3766
                     )
3767

    
3768
    ial.Run(self.op.iallocator)
3769

    
3770
    if not ial.success:
3771
      raise errors.OpPrereqError("Can't compute nodes using"
3772
                                 " iallocator '%s': %s" % (self.op.iallocator,
3773
                                                           ial.info))
3774
    if len(ial.nodes) != ial.required_nodes:
3775
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
3776
                                 " of nodes (%s), required %s" %
3777
                                 (self.op.iallocator, len(ial.nodes),
3778
                                  ial.required_nodes))
3779
    self.op.pnode = ial.nodes[0]
3780
    self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
3781
                 self.op.instance_name, self.op.iallocator,
3782
                 ", ".join(ial.nodes))
3783
    if ial.required_nodes == 2:
3784
      self.op.snode = ial.nodes[1]
3785

    
3786
  def BuildHooksEnv(self):
3787
    """Build hooks env.
3788

3789
    This runs on master, primary and secondary nodes of the instance.
3790

3791
    """
3792
    env = {
3793
      "INSTANCE_DISK_TEMPLATE": self.op.disk_template,
3794
      "INSTANCE_DISK_SIZE": ",".join(str(d["size"]) for d in self.disks),
3795
      "INSTANCE_ADD_MODE": self.op.mode,
3796
      }
3797
    if self.op.mode == constants.INSTANCE_IMPORT:
3798
      env["INSTANCE_SRC_NODE"] = self.op.src_node
3799
      env["INSTANCE_SRC_PATH"] = self.op.src_path
3800
      env["INSTANCE_SRC_IMAGES"] = self.src_images
3801

    
3802
    env.update(_BuildInstanceHookEnv(name=self.op.instance_name,
3803
      primary_node=self.op.pnode,
3804
      secondary_nodes=self.secondaries,
3805
      status=self.instance_status,
3806
      os_type=self.op.os_type,
3807
      memory=self.be_full[constants.BE_MEMORY],
3808
      vcpus=self.be_full[constants.BE_VCPUS],
3809
      nics=[(n.ip, n.bridge, n.mac) for n in self.nics],
3810
    ))
3811

    
3812
    nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
3813
          self.secondaries)
3814
    return env, nl, nl
3815

    
3816

    
3817
  def CheckPrereq(self):
3818
    """Check prerequisites.
3819

3820
    """
3821
    if (not self.cfg.GetVGName() and
3822
        self.op.disk_template not in constants.DTS_NOT_LVM):
3823
      raise errors.OpPrereqError("Cluster does not support lvm-based"
3824
                                 " instances")
3825

    
3826

    
3827
    if self.op.mode == constants.INSTANCE_IMPORT:
3828
      src_node = self.op.src_node
3829
      src_path = self.op.src_path
3830

    
3831
      if src_node is None:
3832
        exp_list = self.rpc.call_export_list(
3833
          self.acquired_locks[locking.LEVEL_NODE])
3834
        found = False
3835
        for node in exp_list:
3836
          if not exp_list[node].failed and src_path in exp_list[node].data:
3837
            found = True
3838
            self.op.src_node = src_node = node
3839
            self.op.src_path = src_path = os.path.join(constants.EXPORT_DIR,
3840
                                                       src_path)
3841
            break
3842
        if not found:
3843
          raise errors.OpPrereqError("No export found for relative path %s" %
3844
                                      src_path)
3845

    
3846
      result = self.rpc.call_export_info(src_node, src_path)
3847
      result.Raise()
3848
      if not result.data:
3849
        raise errors.OpPrereqError("No export found in dir %s" % src_path)
3850

    
3851
      export_info = result.data
3852
      if not export_info.has_section(constants.INISECT_EXP):
3853
        raise errors.ProgrammerError("Corrupted export config")
3854

    
3855
      ei_version = export_info.get(constants.INISECT_EXP, 'version')
3856
      if (int(ei_version) != constants.EXPORT_VERSION):
3857
        raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
3858
                                   (ei_version, constants.EXPORT_VERSION))
3859

    
3860
      # Check that the new instance doesn't have less disks than the export
3861
      instance_disks = len(self.disks)
3862
      export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
3863
      if instance_disks < export_disks:
3864
        raise errors.OpPrereqError("Not enough disks to import."
3865
                                   " (instance: %d, export: %d)" %
3866
                                   (instance_disks, export_disks))
3867

    
3868
      self.op.os_type = export_info.get(constants.INISECT_EXP, 'os')
3869
      disk_images = []
3870
      for idx in range(export_disks):
3871
        option = 'disk%d_dump' % idx
3872
        if export_info.has_option(constants.INISECT_INS, option):
3873
          # FIXME: are the old os-es, disk sizes, etc. useful?
3874
          export_name = export_info.get(constants.INISECT_INS, option)
3875
          image = os.path.join(src_path, export_name)
3876
          disk_images.append(image)
3877
        else:
3878
          disk_images.append(False)
3879

    
3880
      self.src_images = disk_images
3881

    
3882
      old_name = export_info.get(constants.INISECT_INS, 'name')
3883
      # FIXME: int() here could throw a ValueError on broken exports
3884
      exp_nic_count = int(export_info.get(constants.INISECT_INS, 'nic_count'))
3885
      if self.op.instance_name == old_name:
3886
        for idx, nic in enumerate(self.nics):
3887
          if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
3888
            nic_mac_ini = 'nic%d_mac' % idx
3889
            nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
3890

    
3891
    # ip ping checks (we use the same ip that was resolved in ExpandNames)
3892
    if self.op.start and not self.op.ip_check:
3893
      raise errors.OpPrereqError("Cannot ignore IP address conflicts when"
3894
                                 " adding an instance in start mode")
3895

    
3896
    if self.op.ip_check:
3897
      if utils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
3898
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
3899
                                   (self.check_ip, self.op.instance_name))
3900

    
3901
    #### allocator run
3902

    
3903
    if self.op.iallocator is not None:
3904
      self._RunAllocator()
3905

    
3906
    #### node related checks
3907

    
3908
    # check primary node
3909
    self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
3910
    assert self.pnode is not None, \
3911
      "Cannot retrieve locked node %s" % self.op.pnode
3912
    self.secondaries = []
3913

    
3914
    # mirror node verification
3915
    if self.op.disk_template in constants.DTS_NET_MIRROR:
3916
      if self.op.snode is None:
3917
        raise errors.OpPrereqError("The networked disk templates need"
3918
                                   " a mirror node")
3919
      if self.op.snode == pnode.name:
3920
        raise errors.OpPrereqError("The secondary node cannot be"
3921
                                   " the primary node.")
3922
      self.secondaries.append(self.op.snode)
3923

    
3924
    nodenames = [pnode.name] + self.secondaries
3925

    
3926
    req_size = _ComputeDiskSize(self.op.disk_template,
3927
                                self.disks)
3928

    
3929
    # Check lv size requirements
3930
    if req_size is not None:
3931
      nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
3932
                                         self.op.hypervisor)
3933
      for node in nodenames:
3934
        info = nodeinfo[node]
3935
        info.Raise()
3936
        info = info.data
3937
        if not info:
3938
          raise errors.OpPrereqError("Cannot get current information"
3939
                                     " from node '%s'" % node)
3940
        vg_free = info.get('vg_free', None)
3941
        if not isinstance(vg_free, int):
3942
          raise errors.OpPrereqError("Can't compute free disk space on"
3943
                                     " node %s" % node)
3944
        if req_size > info['vg_free']:
3945
          raise errors.OpPrereqError("Not enough disk space on target node %s."
3946
                                     " %d MB available, %d MB required" %
3947
                                     (node, info['vg_free'], req_size))
3948

    
3949
    _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
3950

    
3951
    # os verification
3952
    result = self.rpc.call_os_get(pnode.name, self.op.os_type)
3953
    result.Raise()
3954
    if not isinstance(result.data, objects.OS):
3955
      raise errors.OpPrereqError("OS '%s' not in supported os list for"
3956
                                 " primary node"  % self.op.os_type)
3957

    
3958
    # bridge check on primary node
3959
    bridges = [n.bridge for n in self.nics]
3960
    result = self.rpc.call_bridges_exist(self.pnode.name, bridges)
3961
    result.Raise()
3962
    if not result.data:
3963
      raise errors.OpPrereqError("One of the target bridges '%s' does not"
3964
                                 " exist on destination node '%s'" %
3965
                                 (",".join(bridges), pnode.name))
3966

    
3967
    # memory check on primary node
3968
    if self.op.start:
3969
      _CheckNodeFreeMemory(self, self.pnode.name,
3970
                           "creating instance %s" % self.op.instance_name,
3971
                           self.be_full[constants.BE_MEMORY],
3972
                           self.op.hypervisor)
3973

    
3974
    if self.op.start:
3975
      self.instance_status = 'up'
3976
    else:
3977
      self.instance_status = 'down'
3978

    
3979
  def Exec(self, feedback_fn):
3980
    """Create and add the instance to the cluster.
3981

3982
    """
3983
    instance = self.op.instance_name
3984
    pnode_name = self.pnode.name
3985

    
3986
    for nic in self.nics:
3987
      if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
3988
        nic.mac = self.cfg.GenerateMAC()
3989

    
3990
    ht_kind = self.op.hypervisor
3991
    if ht_kind in constants.HTS_REQ_PORT:
3992
      network_port = self.cfg.AllocatePort()
3993
    else:
3994
      network_port = None
3995

    
3996
    ##if self.op.vnc_bind_address is None:
3997
    ##  self.op.vnc_bind_address = constants.VNC_DEFAULT_BIND_ADDRESS
3998

    
3999
    # this is needed because os.path.join does not accept None arguments
4000
    if self.op.file_storage_dir is None:
4001
      string_file_storage_dir = ""
4002
    else:
4003
      string_file_storage_dir = self.op.file_storage_dir
4004

    
4005
    # build the full file storage dir path
4006
    file_storage_dir = os.path.normpath(os.path.join(
4007
                                        self.cfg.GetFileStorageDir(),
4008
                                        string_file_storage_dir, instance))
4009

    
4010

    
4011
    disks = _GenerateDiskTemplate(self,
4012
                                  self.op.disk_template,
4013
                                  instance, pnode_name,
4014
                                  self.secondaries,
4015
                                  self.disks,
4016
                                  file_storage_dir,
4017
                                  self.op.file_driver,
4018
                                  0)
4019

    
4020
    iobj = objects.Instance(name=instance, os=self.op.os_type,
4021
                            primary_node=pnode_name,
4022
                            nics=self.nics, disks=disks,
4023
                            disk_template=self.op.disk_template,
4024
                            status=self.instance_status,
4025
                            network_port=network_port,
4026
                            beparams=self.op.beparams,
4027
                            hvparams=self.op.hvparams,
4028
                            hypervisor=self.op.hypervisor,
4029
                            )
4030

    
4031
    feedback_fn("* creating instance disks...")
4032
    if not _CreateDisks(self, iobj):
4033
      _RemoveDisks(self, iobj)
4034
      self.cfg.ReleaseDRBDMinors(instance)
4035
      raise errors.OpExecError("Device creation failed, reverting...")
4036

    
4037
    feedback_fn("adding instance %s to cluster config" % instance)
4038

    
4039
    self.cfg.AddInstance(iobj)
4040
    # Declare that we don't want to remove the instance lock anymore, as we've
4041
    # added the instance to the config
4042
    del self.remove_locks[locking.LEVEL_INSTANCE]
4043
    # Remove the temp. assignements for the instance's drbds
4044
    self.cfg.ReleaseDRBDMinors(instance)
4045
    # Unlock all the nodes
4046
    if self.op.mode == constants.INSTANCE_IMPORT:
4047
      nodes_keep = [self.op.src_node]
4048
      nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
4049
                       if node != self.op.src_node]
4050
      self.context.glm.release(locking.LEVEL_NODE, nodes_release)
4051
      self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
4052
    else:
4053
      self.context.glm.release(locking.LEVEL_NODE)
4054
      del self.acquired_locks[locking.LEVEL_NODE]
4055

    
4056
    if self.op.wait_for_sync:
4057
      disk_abort = not _WaitForSync(self, iobj)
4058
    elif iobj.disk_template in constants.DTS_NET_MIRROR:
4059
      # make sure the disks are not degraded (still sync-ing is ok)
4060
      time.sleep(15)
4061
      feedback_fn("* checking mirrors status")
4062
      disk_abort = not _WaitForSync(self, iobj, oneshot=True)
4063
    else:
4064
      disk_abort = False
4065

    
4066
    if disk_abort:
4067
      _RemoveDisks(self, iobj)
4068
      self.cfg.RemoveInstance(iobj.name)
4069
      # Make sure the instance lock gets removed
4070
      self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
4071
      raise errors.OpExecError("There are some degraded disks for"
4072
                               " this instance")
4073

    
4074
    feedback_fn("creating os for instance %s on node %s" %
4075
                (instance, pnode_name))
4076

    
4077
    if iobj.disk_template != constants.DT_DISKLESS:
4078
      if self.op.mode == constants.INSTANCE_CREATE:
4079
        feedback_fn("* running the instance OS create scripts...")
4080
        result = self.rpc.call_instance_os_add(pnode_name, iobj)
4081
        result.Raise()
4082
        if not result.data:
4083
          raise errors.OpExecError("Could not add os for instance %s"
4084
                                   " on node %s" %
4085
                                   (instance, pnode_name))
4086

    
4087
      elif self.op.mode == constants.INSTANCE_IMPORT:
4088
        feedback_fn("* running the instance OS import scripts...")
4089
        src_node = self.op.src_node
4090
        src_images = self.src_images
4091
        cluster_name = self.cfg.GetClusterName()
4092
        import_result = self.rpc.call_instance_os_import(pnode_name, iobj,
4093
                                                         src_node, src_images,
4094
                                                         cluster_name)
4095
        import_result.Raise()
4096
        for idx, result in enumerate(import_result.data):
4097
          if not result:
4098
            self.LogWarning("Could not import the image %s for instance"
4099
                            " %s, disk %d, on node %s" %
4100
                            (src_images[idx], instance, idx, pnode_name))
4101
      else:
4102
        # also checked in the prereq part
4103
        raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
4104
                                     % self.op.mode)
4105

    
4106
    if self.op.start:
4107
      logging.info("Starting instance %s on node %s", instance, pnode_name)
4108
      feedback_fn("* starting instance...")
4109
      result = self.rpc.call_instance_start(pnode_name, iobj, None)
4110
      result.Raise()
4111
      if not result.data:
4112
        raise errors.OpExecError("Could not start instance")
4113

    
4114

    
4115
class LUConnectConsole(NoHooksLU):
4116
  """Connect to an instance's console.
4117

4118
  This is somewhat special in that it returns the command line that
4119
  you need to run on the master node in order to connect to the
4120
  console.
4121

4122
  """
4123
  _OP_REQP = ["instance_name"]
4124
  REQ_BGL = False
4125

    
4126
  def ExpandNames(self):
4127
    self._ExpandAndLockInstance()
4128

    
4129
  def CheckPrereq(self):
4130
    """Check prerequisites.
4131

4132
    This checks that the instance is in the cluster.
4133

4134
    """
4135
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4136
    assert self.instance is not None, \
4137
      "Cannot retrieve locked instance %s" % self.op.instance_name
4138

    
4139
  def Exec(self, feedback_fn):
4140
    """Connect to the console of an instance
4141

4142
    """
4143
    instance = self.instance
4144
    node = instance.primary_node
4145

    
4146
    node_insts = self.rpc.call_instance_list([node],
4147
                                             [instance.hypervisor])[node]
4148
    node_insts.Raise()
4149

    
4150
    if instance.name not in node_insts.data:
4151
      raise errors.OpExecError("Instance %s is not running." % instance.name)
4152

    
4153
    logging.debug("Connecting to console of %s on %s", instance.name, node)
4154

    
4155
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
4156
    console_cmd = hyper.GetShellCommandForConsole(instance)
4157

    
4158
    # build ssh cmdline
4159
    return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
4160

    
4161

    
4162
class LUReplaceDisks(LogicalUnit):
4163
  """Replace the disks of an instance.
4164

4165
  """
4166
  HPATH = "mirrors-replace"
4167
  HTYPE = constants.HTYPE_INSTANCE
4168
  _OP_REQP = ["instance_name", "mode", "disks"]
4169
  REQ_BGL = False
4170

    
4171
  def ExpandNames(self):
4172
    self._ExpandAndLockInstance()
4173

    
4174
    if not hasattr(self.op, "remote_node"):
4175
      self.op.remote_node = None
4176

    
4177
    ia_name = getattr(self.op, "iallocator", None)
4178
    if ia_name is not None:
4179
      if self.op.remote_node is not None:
4180
        raise errors.OpPrereqError("Give either the iallocator or the new"
4181
                                   " secondary, not both")
4182
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4183
    elif self.op.remote_node is not None:
4184
      remote_node = self.cfg.ExpandNodeName(self.op.remote_node)
4185
      if remote_node is None:
4186
        raise errors.OpPrereqError("Node '%s' not known" %
4187
                                   self.op.remote_node)
4188
      self.op.remote_node = remote_node
4189
      self.needed_locks[locking.LEVEL_NODE] = [remote_node]
4190
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
4191
    else:
4192
      self.needed_locks[locking.LEVEL_NODE] = []
4193
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4194

    
4195
  def DeclareLocks(self, level):
4196
    # If we're not already locking all nodes in the set we have to declare the
4197
    # instance's primary/secondary nodes.
4198
    if (level == locking.LEVEL_NODE and
4199
        self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
4200
      self._LockInstancesNodes()
4201

    
4202
  def _RunAllocator(self):
4203
    """Compute a new secondary node using an IAllocator.
4204

4205
    """
4206
    ial = IAllocator(self,
4207
                     mode=constants.IALLOCATOR_MODE_RELOC,
4208
                     name=self.op.instance_name,
4209
                     relocate_from=[self.sec_node])
4210

    
4211
    ial.Run(self.op.iallocator)
4212

    
4213
    if not ial.success:
4214
      raise errors.OpPrereqError("Can't compute nodes using"
4215
                                 " iallocator '%s': %s" % (self.op.iallocator,
4216
                                                           ial.info))
4217
    if len(ial.nodes) != ial.required_nodes:
4218
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
4219
                                 " of nodes (%s), required %s" %
4220
                                 (len(ial.nodes), ial.required_nodes))
4221
    self.op.remote_node = ial.nodes[0]
4222
    self.LogInfo("Selected new secondary for the instance: %s",
4223
                 self.op.remote_node)
4224

    
4225
  def BuildHooksEnv(self):
4226
    """Build hooks env.
4227

4228
    This runs on the master, the primary and all the secondaries.
4229

4230
    """
4231
    env = {
4232
      "MODE": self.op.mode,
4233
      "NEW_SECONDARY": self.op.remote_node,
4234
      "OLD_SECONDARY": self.instance.secondary_nodes[0],
4235
      }
4236
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4237
    nl = [
4238
      self.cfg.GetMasterNode(),
4239
      self.instance.primary_node,
4240
      ]
4241
    if self.op.remote_node is not None:
4242
      nl.append(self.op.remote_node)
4243
    return env, nl, nl
4244

    
4245
  def CheckPrereq(self):
4246
    """Check prerequisites.
4247

4248
    This checks that the instance is in the cluster.
4249

4250
    """
4251
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4252
    assert instance is not None, \
4253
      "Cannot retrieve locked instance %s" % self.op.instance_name
4254
    self.instance = instance
4255

    
4256
    if instance.disk_template not in constants.DTS_NET_MIRROR:
4257
      raise errors.OpPrereqError("Instance's disk layout is not"
4258
                                 " network mirrored.")
4259

    
4260
    if len(instance.secondary_nodes) != 1:
4261
      raise errors.OpPrereqError("The instance has a strange layout,"
4262
                                 " expected one secondary but found %d" %
4263
                                 len(instance.secondary_nodes))
4264

    
4265
    self.sec_node = instance.secondary_nodes[0]
4266

    
4267
    ia_name = getattr(self.op, "iallocator", None)
4268
    if ia_name is not None:
4269
      self._RunAllocator()
4270

    
4271
    remote_node = self.op.remote_node
4272
    if remote_node is not None:
4273
      self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
4274
      assert self.remote_node_info is not None, \
4275
        "Cannot retrieve locked node %s" % remote_node
4276
    else:
4277
      self.remote_node_info = None
4278
    if remote_node == instance.primary_node:
4279
      raise errors.OpPrereqError("The specified node is the primary node of"
4280
                                 " the instance.")
4281
    elif remote_node == self.sec_node:
4282
      if self.op.mode == constants.REPLACE_DISK_SEC:
4283
        # this is for DRBD8, where we can't execute the same mode of
4284
        # replacement as for drbd7 (no different port allocated)
4285
        raise errors.OpPrereqError("Same secondary given, cannot execute"
4286
                                   " replacement")
4287
    if instance.disk_template == constants.DT_DRBD8:
4288
      if (self.op.mode == constants.REPLACE_DISK_ALL and
4289
          remote_node is not None):
4290
        # switch to replace secondary mode
4291
        self.op.mode = constants.REPLACE_DISK_SEC
4292

    
4293
      if self.op.mode == constants.REPLACE_DISK_ALL:
4294
        raise errors.OpPrereqError("Template 'drbd' only allows primary or"
4295
                                   " secondary disk replacement, not"
4296
                                   " both at once")
4297
      elif self.op.mode == constants.REPLACE_DISK_PRI:
4298
        if remote_node is not None:
4299
          raise errors.OpPrereqError("Template 'drbd' does not allow changing"
4300
                                     " the secondary while doing a primary"
4301
                                     " node disk replacement")
4302
        self.tgt_node = instance.primary_node
4303
        self.oth_node = instance.secondary_nodes[0]
4304
      elif self.op.mode == constants.REPLACE_DISK_SEC:
4305
        self.new_node = remote_node # this can be None, in which case
4306
                                    # we don't change the secondary
4307
        self.tgt_node = instance.secondary_nodes[0]
4308
        self.oth_node = instance.primary_node
4309
      else:
4310
        raise errors.ProgrammerError("Unhandled disk replace mode")
4311

    
4312
    if not self.op.disks:
4313
      self.op.disks = range(len(instance.disks))
4314

    
4315
    for disk_idx in self.op.disks:
4316
      instance.FindDisk(disk_idx)
4317

    
4318
  def _ExecD8DiskOnly(self, feedback_fn):
4319
    """Replace a disk on the primary or secondary for dbrd8.
4320

4321
    The algorithm for replace is quite complicated:
4322

4323
      1. for each disk to be replaced:
4324

4325
        1. create new LVs on the target node with unique names
4326
        1. detach old LVs from the drbd device
4327
        1. rename old LVs to name_replaced.<time_t>
4328
        1. rename new LVs to old LVs
4329
        1. attach the new LVs (with the old names now) to the drbd device
4330

4331
      1. wait for sync across all devices
4332

4333
      1. for each modified disk:
4334

4335
        1. remove old LVs (which have the name name_replaces.<time_t>)
4336

4337
    Failures are not very well handled.
4338

4339
    """
4340
    steps_total = 6
4341
    warning, info = (self.proc.LogWarning, self.proc.LogInfo)
4342
    instance = self.instance
4343
    iv_names = {}
4344
    vgname = self.cfg.GetVGName()
4345
    # start of work
4346
    cfg = self.cfg
4347
    tgt_node = self.tgt_node
4348
    oth_node = self.oth_node
4349

    
4350
    # Step: check device activation
4351
    self.proc.LogStep(1, steps_total, "check device existence")
4352
    info("checking volume groups")
4353
    my_vg = cfg.GetVGName()
4354
    results = self.rpc.call_vg_list([oth_node, tgt_node])
4355
    if not results:
4356
      raise errors.OpExecError("Can't list volume groups on the nodes")
4357
    for node in oth_node, tgt_node:
4358
      res = results[node]
4359
      if res.failed or not res.data or my_vg not in res.data:
4360
        raise errors.OpExecError("Volume group '%s' not found on %s" %
4361
                                 (my_vg, node))
4362
    for idx, dev in enumerate(instance.disks):
4363
      if idx not in self.op.disks:
4364
        continue
4365
      for node in tgt_node, oth_node:
4366
        info("checking disk/%d on %s" % (idx, node))
4367
        cfg.SetDiskID(dev, node)
4368
        if not self.rpc.call_blockdev_find(node, dev):
4369
          raise errors.OpExecError("Can't find disk/%d on node %s" %
4370
                                   (idx, node))
4371

    
4372
    # Step: check other node consistency
4373
    self.proc.LogStep(2, steps_total, "check peer consistency")
4374
    for idx, dev in enumerate(instance.disks):
4375
      if idx not in self.op.disks:
4376
        continue
4377
      info("checking disk/%d consistency on %s" % (idx, oth_node))
4378
      if not _CheckDiskConsistency(self, dev, oth_node,
4379
                                   oth_node==instance.primary_node):
4380
        raise errors.OpExecError("Peer node (%s) has degraded storage, unsafe"
4381
                                 " to replace disks on this node (%s)" %
4382
                                 (oth_node, tgt_node))
4383

    
4384
    # Step: create new storage
4385
    self.proc.LogStep(3, steps_total, "allocate new storage")
4386
    for idx, dev in enumerate(instance.disks):
4387
      if idx not in self.op.disks:
4388
        continue
4389
      size = dev.size
4390
      cfg.SetDiskID(dev, tgt_node)
4391
      lv_names = [".disk%d_%s" % (idx, suf)
4392
                  for suf in ["data", "meta"]]
4393
      names = _GenerateUniqueNames(self, lv_names)
4394
      lv_data = objects.Disk(dev_type=constants.LD_LV, size=size,
4395
                             logical_id=(vgname, names[0]))
4396
      lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
4397
                             logical_id=(vgname, names[1]))
4398
      new_lvs = [lv_data, lv_meta]
4399
      old_lvs = dev.children
4400
      iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
4401
      info("creating new local storage on %s for %s" %
4402
           (tgt_node, dev.iv_name))
4403
      # since we *always* want to create this LV, we use the
4404
      # _Create...OnPrimary (which forces the creation), even if we
4405
      # are talking about the secondary node
4406
      for new_lv in new_lvs:
4407
        if not _CreateBlockDevOnPrimary(self, tgt_node, instance, new_lv,
4408
                                        _GetInstanceInfoText(instance)):
4409
          raise errors.OpExecError("Failed to create new LV named '%s' on"
4410
                                   " node '%s'" %
4411
                                   (new_lv.logical_id[1], tgt_node))
4412

    
4413
    # Step: for each lv, detach+rename*2+attach
4414
    self.proc.LogStep(4, steps_total, "change drbd configuration")
4415
    for dev, old_lvs, new_lvs in iv_names.itervalues():
4416
      info("detaching %s drbd from local storage" % dev.iv_name)
4417
      result = self.rpc.call_blockdev_removechildren(tgt_node, dev, old_lvs)
4418
      result.Raise()
4419
      if not result.data:
4420
        raise errors.OpExecError("Can't detach drbd from local storage on node"
4421
                                 " %s for device %s" % (tgt_node, dev.iv_name))
4422
      #dev.children = []
4423
      #cfg.Update(instance)
4424

    
4425
      # ok, we created the new LVs, so now we know we have the needed
4426
      # storage; as such, we proceed on the target node to rename
4427
      # old_lv to _old, and new_lv to old_lv; note that we rename LVs
4428
      # using the assumption that logical_id == physical_id (which in
4429
      # turn is the unique_id on that node)
4430

    
4431
      # FIXME(iustin): use a better name for the replaced LVs
4432
      temp_suffix = int(time.time())
4433
      ren_fn = lambda d, suff: (d.physical_id[0],
4434
                                d.physical_id[1] + "_replaced-%s" % suff)
4435
      # build the rename list based on what LVs exist on the node
4436
      rlist = []
4437
      for to_ren in old_lvs:
4438
        find_res = self.rpc.call_blockdev_find(tgt_node, to_ren)
4439
        if not find_res.failed and find_res.data is not None: # device exists
4440
          rlist.append((to_ren, ren_fn(to_ren, temp_suffix)))
4441

    
4442
      info("renaming the old LVs on the target node")
4443
      result = self.rpc.call_blockdev_rename(tgt_node, rlist)
4444
      result.Raise()
4445
      if not result.data:
4446
        raise errors.OpExecError("Can't rename old LVs on node %s" % tgt_node)
4447
      # now we rename the new LVs to the old LVs
4448
      info("renaming the new LVs on the target node")
4449
      rlist = [(new, old.physical_id) for old, new in zip(old_lvs, new_lvs)]
4450
      result = self.rpc.call_blockdev_rename(tgt_node, rlist)
4451
      result.Raise()
4452
      if not result.data:
4453
        raise errors.OpExecError("Can't rename new LVs on node %s" % tgt_node)
4454

    
4455
      for old, new in zip(old_lvs, new_lvs):
4456
        new.logical_id = old.logical_id
4457
        cfg.SetDiskID(new, tgt_node)
4458

    
4459
      for disk in old_lvs:
4460
        disk.logical_id = ren_fn(disk, temp_suffix)
4461
        cfg.SetDiskID(disk, tgt_node)
4462

    
4463
      # now that the new lvs have the old name, we can add them to the device
4464
      info("adding new mirror component on %s" % tgt_node)
4465
      result =self.rpc.call_blockdev_addchildren(tgt_node, dev, new_lvs)
4466
      if result.failed or not result.data:
4467
        for new_lv in new_lvs:
4468
          result = self.rpc.call_blockdev_remove(tgt_node, new_lv)
4469
          if result.failed or not result.data:
4470
            warning("Can't rollback device %s", hint="manually cleanup unused"
4471
                    " logical volumes")
4472
        raise errors.OpExecError("Can't add local storage to drbd")
4473

    
4474
      dev.children = new_lvs
4475
      cfg.Update(instance)
4476

    
4477
    # Step: wait for sync
4478

    
4479
    # this can fail as the old devices are degraded and _WaitForSync
4480
    # does a combined result over all disks, so we don't check its
4481
    # return value
4482
    self.proc.LogStep(5, steps_total, "sync devices")
4483
    _WaitForSync(self, instance, unlock=True)
4484

    
4485
    # so check manually all the devices
4486
    for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
4487
      cfg.SetDiskID(dev, instance.primary_node)
4488
      result = self.rpc.call_blockdev_find(instance.primary_node, dev)
4489
      if result.failed or result.data[5]:
4490
        raise errors.OpExecError("DRBD device %s is degraded!" % name)
4491

    
4492
    # Step: remove old storage
4493
    self.proc.LogStep(6, steps_total, "removing old storage")
4494
    for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
4495
      info("remove logical volumes for %s" % name)
4496
      for lv in old_lvs:
4497
        cfg.SetDiskID(lv, tgt_node)
4498
        result = self.rpc.call_blockdev_remove(tgt_node, lv)
4499
        if result.failed or not result.data:
4500
          warning("Can't remove old LV", hint="manually remove unused LVs")
4501
          continue
4502

    
4503
  def _ExecD8Secondary(self, feedback_fn):
4504
    """Replace the secondary node for drbd8.
4505

4506
    The algorithm for replace is quite complicated:
4507
      - for all disks of the instance:
4508
        - create new LVs on the new node with same names
4509
        - shutdown the drbd device on the old secondary
4510
        - disconnect the drbd network on the primary
4511
        - create the drbd device on the new secondary
4512
        - network attach the drbd on the primary, using an artifice:
4513
          the drbd code for Attach() will connect to the network if it
4514
          finds a device which is connected to the good local disks but
4515
          not network enabled
4516
      - wait for sync across all devices
4517
      - remove all disks from the old secondary
4518

4519
    Failures are not very well handled.
4520

4521
    """
4522
    steps_total = 6
4523
    warning, info = (self.proc.LogWarning, self.proc.LogInfo)
4524
    instance = self.instance
4525
    iv_names = {}
4526
    vgname = self.cfg.GetVGName()
4527
    # start of work
4528
    cfg = self.cfg
4529
    old_node = self.tgt_node
4530
    new_node = self.new_node
4531
    pri_node = instance.primary_node
4532

    
4533
    # Step: check device activation
4534
    self.proc.LogStep(1, steps_total, "check device existence")
4535
    info("checking volume groups")
4536
    my_vg = cfg.GetVGName()
4537
    results = self.rpc.call_vg_list([pri_node, new_node])
4538
    for node in pri_node, new_node:
4539
      res = results[node]
4540
      if res.failed or not res.data or my_vg not in res.data:
4541
        raise errors.OpExecError("Volume group '%s' not found on %s" %
4542
                                 (my_vg, node))
4543
    for idx, dev in enumerate(instance.disks):
4544
      if idx not in self.op.disks:
4545
        continue
4546
      info("checking disk/%d on %s" % (idx, pri_node))
4547
      cfg.SetDiskID(dev, pri_node)
4548
      result = self.rpc.call_blockdev_find(pri_node, dev)
4549
      result.Raise()
4550
      if not result.data:
4551
        raise errors.OpExecError("Can't find disk/%d on node %s" %
4552
                                 (idx, pri_node))
4553

    
4554
    # Step: check other node consistency
4555
    self.proc.LogStep(2, steps_total, "check peer consistency")
4556
    for idx, dev in enumerate(instance.disks):
4557
      if idx not in self.op.disks:
4558
        continue
4559
      info("checking disk/%d consistency on %s" % (idx, pri_node))
4560
      if not _CheckDiskConsistency(self, dev, pri_node, True, ldisk=True):
4561
        raise errors.OpExecError("Primary node (%s) has degraded storage,"
4562
                                 " unsafe to replace the secondary" %
4563
                                 pri_node)
4564

    
4565
    # Step: create new storage
4566
    self.proc.LogStep(3, steps_total, "allocate new storage")
4567
    for idx, dev in enumerate(instance.disks):
4568
      size = dev.size
4569
      info("adding new local storage on %s for disk/%d" %
4570
           (new_node, idx))
4571
      # since we *always* want to create this LV, we use the
4572
      # _Create...OnPrimary (which forces the creation), even if we
4573
      # are talking about the secondary node
4574
      for new_lv in dev.children:
4575
        if not _CreateBlockDevOnPrimary(self, new_node, instance, new_lv,
4576
                                        _GetInstanceInfoText(instance)):
4577
          raise errors.OpExecError("Failed to create new LV named '%s' on"
4578
                                   " node '%s'" %
4579
                                   (new_lv.logical_id[1], new_node))
4580

    
4581
    # Step 4: dbrd minors and drbd setups changes
4582
    # after this, we must manually remove the drbd minors on both the
4583
    # error and the success paths
4584
    minors = cfg.AllocateDRBDMinor([new_node for dev in instance.disks],
4585
                                   instance.name)
4586
    logging.debug("Allocated minors %s" % (minors,))
4587
    self.proc.LogStep(4, steps_total, "changing drbd configuration")
4588
    for idx, (dev, new_minor) in enumerate(zip(instance.disks, minors)):
4589
      size = dev.size
4590
      info("activating a new drbd on %s for disk/%d" % (new_node, idx))
4591
      # create new devices on new_node
4592
      if pri_node == dev.logical_id[0]:
4593
        new_logical_id = (pri_node, new_node,
4594
                          dev.logical_id[2], dev.logical_id[3], new_minor,
4595
                          dev.logical_id[5])
4596
      else:
4597
        new_logical_id = (new_node, pri_node,
4598
                          dev.logical_id[2], new_minor, dev.logical_id[4],
4599
                          dev.logical_id[5])
4600
      iv_names[idx] = (dev, dev.children, new_logical_id)
4601
      logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
4602
                    new_logical_id)
4603
      new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
4604
                              logical_id=new_logical_id,
4605
                              children=dev.children)
4606
      if not _CreateBlockDevOnSecondary(self, new_node, instance,
4607
                                        new_drbd, False,
4608
                                        _GetInstanceInfoText(instance)):
4609
        self.cfg.ReleaseDRBDMinors(instance.name)
4610
        raise errors.OpExecError("Failed to create new DRBD on"
4611
                                 " node '%s'" % new_node)
4612

    
4613
    for idx, dev in enumerate(instance.disks):
4614
      # we have new devices, shutdown the drbd on the old secondary
4615
      info("shutting down drbd for disk/%d on old node" % idx)
4616
      cfg.SetDiskID(dev, old_node)
4617
      result = self.rpc.call_blockdev_shutdown(old_node, dev)
4618
      if result.failed or not result.data:
4619
        warning("Failed to shutdown drbd for disk/%d on old node" % idx,
4620
                hint="Please cleanup this device manually as soon as possible")
4621

    
4622
    info("detaching primary drbds from the network (=> standalone)")
4623
    done = 0
4624
    for idx, dev in enumerate(instance.disks):
4625
      cfg.SetDiskID(dev, pri_node)
4626
      # set the network part of the physical (unique in bdev terms) id
4627
      # to None, meaning detach from network
4628
      dev.physical_id = (None, None, None, None) + dev.physical_id[4:]
4629
      # and 'find' the device, which will 'fix' it to match the
4630
      # standalone state
4631
      result = self.rpc.call_blockdev_find(pri_node, dev)
4632
      if not result.failed and result.data:
4633
        done += 1
4634
      else:
4635
        warning("Failed to detach drbd disk/%d from network, unusual case" %
4636
                idx)
4637

    
4638
    if not done:
4639
      # no detaches succeeded (very unlikely)
4640
      self.cfg.ReleaseDRBDMinors(instance.name)
4641
      raise errors.OpExecError("Can't detach at least one DRBD from old node")
4642

    
4643
    # if we managed to detach at least one, we update all the disks of
4644
    # the instance to point to the new secondary
4645
    info("updating instance configuration")
4646
    for dev, _, new_logical_id in iv_names.itervalues():
4647
      dev.logical_id = new_logical_id
4648
      cfg.SetDiskID(dev, pri_node)
4649
    cfg.Update(instance)
4650
    # we can remove now the temp minors as now the new values are
4651
    # written to the config file (and therefore stable)
4652
    self.cfg.ReleaseDRBDMinors(instance.name)
4653

    
4654
    # and now perform the drbd attach
4655
    info("attaching primary drbds to new secondary (standalone => connected)")
4656
    failures = []
4657
    for idx, dev in enumerate(instance.disks):
4658
      info("attaching primary drbd for disk/%d to new secondary node" % idx)
4659
      # since the attach is smart, it's enough to 'find' the device,
4660
      # it will automatically activate the network, if the physical_id
4661
      # is correct
4662
      cfg.SetDiskID(dev, pri_node)
4663
      logging.debug("Disk to attach: %s", dev)
4664
      result = self.rpc.call_blockdev_find(pri_node, dev)
4665
      if result.failed or not result.data:
4666
        warning("can't attach drbd disk/%d to new secondary!" % idx,
4667
                "please do a gnt-instance info to see the status of disks")
4668

    
4669
    # this can fail as the old devices are degraded and _WaitForSync
4670
    # does a combined result over all disks, so we don't check its
4671
    # return value
4672
    self.proc.LogStep(5, steps_total, "sync devices")
4673
    _WaitForSync(self, instance, unlock=True)
4674

    
4675
    # so check manually all the devices
4676
    for idx, (dev, old_lvs, _) in iv_names.iteritems():
4677
      cfg.SetDiskID(dev, pri_node)
4678
      result = self.rpc.call_blockdev_find(pri_node, dev)
4679
      result.Raise()
4680
      if result.data[5]:
4681
        raise errors.OpExecError("DRBD device disk/%d is degraded!" % idx)
4682

    
4683
    self.proc.LogStep(6, steps_total, "removing old storage")
4684
    for idx, (dev, old_lvs, _) in iv_names.iteritems():
4685
      info("remove logical volumes for disk/%d" % idx)
4686
      for lv in old_lvs:
4687
        cfg.SetDiskID(lv, old_node)
4688
        result = self.rpc.call_blockdev_remove(old_node, lv)
4689
        if result.failed or not result.data:
4690
          warning("Can't remove LV on old secondary",
4691
                  hint="Cleanup stale volumes by hand")
4692

    
4693
  def Exec(self, feedback_fn):
4694
    """Execute disk replacement.
4695

4696
    This dispatches the disk replacement to the appropriate handler.
4697

4698
    """
4699
    instance = self.instance
4700

    
4701
    # Activate the instance disks if we're replacing them on a down instance
4702
    if instance.status == "down":
4703
      _StartInstanceDisks(self, instance, True)
4704

    
4705
    if instance.disk_template == constants.DT_DRBD8:
4706
      if self.op.remote_node is None:
4707
        fn = self._ExecD8DiskOnly
4708
      else:
4709
        fn = self._ExecD8Secondary
4710
    else:
4711
      raise errors.ProgrammerError("Unhandled disk replacement case")
4712

    
4713
    ret = fn(feedback_fn)
4714

    
4715
    # Deactivate the instance disks if we're replacing them on a down instance
4716
    if instance.status == "down":
4717
      _SafeShutdownInstanceDisks(self, instance)
4718

    
4719
    return ret
4720

    
4721

    
4722
class LUGrowDisk(LogicalUnit):
4723
  """Grow a disk of an instance.
4724

4725
  """
4726
  HPATH = "disk-grow"
4727
  HTYPE = constants.HTYPE_INSTANCE
4728
  _OP_REQP = ["instance_name", "disk", "amount", "wait_for_sync"]
4729
  REQ_BGL = False
4730

    
4731
  def ExpandNames(self):
4732
    self._ExpandAndLockInstance()
4733
    self.needed_locks[locking.LEVEL_NODE] = []
4734
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4735

    
4736
  def DeclareLocks(self, level):
4737
    if level == locking.LEVEL_NODE:
4738
      self._LockInstancesNodes()
4739

    
4740
  def BuildHooksEnv(self):
4741
    """Build hooks env.
4742

4743
    This runs on the master, the primary and all the secondaries.
4744

4745
    """
4746
    env = {
4747
      "DISK": self.op.disk,
4748
      "AMOUNT": self.op.amount,
4749
      }
4750
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4751
    nl = [
4752
      self.cfg.GetMasterNode(),
4753
      self.instance.primary_node,
4754
      ]
4755
    return env, nl, nl
4756

    
4757
  def CheckPrereq(self):
4758
    """Check prerequisites.
4759

4760
    This checks that the instance is in the cluster.
4761

4762
    """
4763
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4764
    assert instance is not None, \
4765
      "Cannot retrieve locked instance %s" % self.op.instance_name
4766

    
4767
    self.instance = instance
4768

    
4769
    if instance.disk_template not in (constants.DT_PLAIN, constants.DT_DRBD8):
4770
      raise errors.OpPrereqError("Instance's disk layout does not support"
4771
                                 " growing.")
4772

    
4773
    self.disk = instance.FindDisk(self.op.disk)
4774

    
4775
    nodenames = [instance.primary_node] + list(instance.secondary_nodes)
4776
    nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
4777
                                       instance.hypervisor)
4778
    for node in nodenames:
4779
      info = nodeinfo[node]
4780
      if info.failed or not info.data:
4781
        raise errors.OpPrereqError("Cannot get current information"
4782
                                   " from node '%s'" % node)
4783
      vg_free = info.data.get('vg_free', None)
4784
      if not isinstance(vg_free, int):
4785
        raise errors.OpPrereqError("Can't compute free disk space on"
4786
                                   " node %s" % node)
4787
      if self.op.amount > vg_free:
4788
        raise errors.OpPrereqError("Not enough disk space on target node %s:"
4789
                                   " %d MiB available, %d MiB required" %
4790
                                   (node, vg_free, self.op.amount))
4791

    
4792
  def Exec(self, feedback_fn):
4793
    """Execute disk grow.
4794

4795
    """
4796
    instance = self.instance
4797
    disk = self.disk
4798
    for node in (instance.secondary_nodes + (instance.primary_node,)):
4799
      self.cfg.SetDiskID(disk, node)
4800
      result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
4801
      result.Raise()
4802
      if (not result.data or not isinstance(result.data, (list, tuple)) or
4803
          len(result.data) != 2):
4804
        raise errors.OpExecError("Grow request failed to node %s" % node)
4805
      elif not result.data[0]:
4806
        raise errors.OpExecError("Grow request failed to node %s: %s" %
4807
                                 (node, result.data[1]))
4808
    disk.RecordGrow(self.op.amount)
4809
    self.cfg.Update(instance)
4810
    if self.op.wait_for_sync:
4811
      disk_abort = not _WaitForSync(self, instance)
4812
      if disk_abort:
4813
        self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
4814
                             " status.\nPlease check the instance.")
4815

    
4816

    
4817
class LUQueryInstanceData(NoHooksLU):
4818
  """Query runtime instance data.
4819

4820
  """
4821
  _OP_REQP = ["instances", "static"]
4822
  REQ_BGL = False
4823

    
4824
  def ExpandNames(self):
4825
    self.needed_locks = {}
4826
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
4827

    
4828
    if not isinstance(self.op.instances, list):
4829
      raise errors.OpPrereqError("Invalid argument type 'instances'")
4830

    
4831
    if self.op.instances:
4832
      self.wanted_names = []
4833
      for name in self.op.instances:
4834
        full_name = self.cfg.ExpandInstanceName(name)
4835
        if full_name is None:
4836
          raise errors.OpPrereqError("Instance '%s' not known" %
4837
                                     self.op.instance_name)
4838
        self.wanted_names.append(full_name)
4839
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
4840
    else:
4841
      self.wanted_names = None
4842
      self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
4843

    
4844
    self.needed_locks[locking.LEVEL_NODE] = []
4845
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4846

    
4847
  def DeclareLocks(self, level):
4848
    if level == locking.LEVEL_NODE:
4849
      self._LockInstancesNodes()
4850

    
4851
  def CheckPrereq(self):
4852
    """Check prerequisites.
4853

4854
    This only checks the optional instance list against the existing names.
4855

4856
    """
4857
    if self.wanted_names is None:
4858
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
4859

    
4860
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
4861
                             in self.wanted_names]
4862
    return
4863

    
4864
  def _ComputeDiskStatus(self, instance, snode, dev):
4865
    """Compute block device status.
4866

4867
    """
4868
    static = self.op.static
4869
    if not static:
4870
      self.cfg.SetDiskID(dev, instance.primary_node)
4871
      dev_pstatus = self.rpc.call_blockdev_find(instance.primary_node, dev)
4872
      dev_pstatus.Raise()
4873
      dev_pstatus = dev_pstatus.data
4874
    else:
4875
      dev_pstatus = None
4876

    
4877
    if dev.dev_type in constants.LDS_DRBD:
4878
      # we change the snode then (otherwise we use the one passed in)
4879
      if dev.logical_id[0] == instance.primary_node:
4880
        snode = dev.logical_id[1]
4881
      else:
4882
        snode = dev.logical_id[0]
4883

    
4884
    if snode and not static:
4885
      self.cfg.SetDiskID(dev, snode)
4886
      dev_sstatus = self.rpc.call_blockdev_find(snode, dev)
4887
      dev_sstatus.Raise()
4888
      dev_sstatus = dev_sstatus.data
4889
    else:
4890
      dev_sstatus = None
4891

    
4892
    if dev.children:
4893
      dev_children = [self._ComputeDiskStatus(instance, snode, child)
4894
                      for child in dev.children]
4895
    else:
4896
      dev_children = []
4897

    
4898
    data = {
4899
      "iv_name": dev.iv_name,
4900
      "dev_type": dev.dev_type,
4901
      "logical_id": dev.logical_id,
4902
      "physical_id": dev.physical_id,
4903
      "pstatus": dev_pstatus,
4904
      "sstatus": dev_sstatus,
4905
      "children": dev_children,
4906
      "mode": dev.mode,
4907
      }
4908

    
4909
    return data
4910

    
4911
  def Exec(self, feedback_fn):
4912
    """Gather and return data"""
4913
    result = {}
4914

    
4915
    cluster = self.cfg.GetClusterInfo()
4916

    
4917
    for instance in self.wanted_instances:
4918
      if not self.op.static:
4919
        remote_info = self.rpc.call_instance_info(instance.primary_node,
4920
                                                  instance.name,
4921
                                                  instance.hypervisor)
4922
        remote_info.Raise()
4923
        remote_info = remote_info.data
4924
        if remote_info and "state" in remote_info:
4925
          remote_state = "up"
4926
        else:
4927
          remote_state = "down"
4928
      else:
4929
        remote_state = None
4930
      if instance.status == "down":
4931
        config_state = "down"
4932
      else:
4933
        config_state = "up"
4934

    
4935
      disks = [self._ComputeDiskStatus(instance, None, device)
4936
               for device in instance.disks]
4937

    
4938
      idict = {
4939
        "name": instance.name,
4940
        "config_state": config_state,
4941
        "run_state": remote_state,
4942
        "pnode": instance.primary_node,
4943
        "snodes": instance.secondary_nodes,
4944
        "os": instance.os,
4945
        "nics": [(nic.mac, nic.ip, nic.bridge) for nic in instance.nics],
4946
        "disks": disks,
4947
        "hypervisor": instance.hypervisor,
4948
        "network_port": instance.network_port,
4949
        "hv_instance": instance.hvparams,
4950
        "hv_actual": cluster.FillHV(instance),
4951
        "be_instance": instance.beparams,
4952
        "be_actual": cluster.FillBE(instance),
4953
        }
4954

    
4955
      result[instance.name] = idict
4956

    
4957
    return result
4958

    
4959

    
4960
class LUSetInstanceParams(LogicalUnit):
4961
  """Modifies an instances's parameters.
4962

4963
  """
4964
  HPATH = "instance-modify"
4965
  HTYPE = constants.HTYPE_INSTANCE
4966
  _OP_REQP = ["instance_name"]
4967
  REQ_BGL = False
4968

    
4969
  def CheckArguments(self):
4970
    if not hasattr(self.op, 'nics'):
4971
      self.op.nics = []
4972
    if not hasattr(self.op, 'disks'):
4973
      self.op.disks = []
4974
    if not hasattr(self.op, 'beparams'):
4975
      self.op.beparams = {}
4976
    if not hasattr(self.op, 'hvparams'):
4977
      self.op.hvparams = {}
4978
    self.op.force = getattr(self.op, "force", False)
4979
    if not (self.op.nics or self.op.disks or
4980
            self.op.hvparams or self.op.beparams):
4981
      raise errors.OpPrereqError("No changes submitted")
4982

    
4983
    utils.CheckBEParams(self.op.beparams)
4984

    
4985
    # Disk validation
4986
    disk_addremove = 0
4987
    for disk_op, disk_dict in self.op.disks:
4988
      if disk_op == constants.DDM_REMOVE:
4989
        disk_addremove += 1
4990
        continue
4991
      elif disk_op == constants.DDM_ADD:
4992
        disk_addremove += 1
4993
      else:
4994
        if not isinstance(disk_op, int):
4995
          raise errors.OpPrereqError("Invalid disk index")
4996
      if disk_op == constants.DDM_ADD:
4997
        mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
4998
        if mode not in (constants.DISK_RDONLY, constants.DISK_RDWR):
4999
          raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode)
5000
        size = disk_dict.get('size', None)
5001
        if size is None:
5002
          raise errors.OpPrereqError("Required disk parameter size missing")
5003
        try:
5004
          size = int(size)
5005
        except ValueError, err:
5006
          raise errors.OpPrereqError("Invalid disk size parameter: %s" %
5007
                                     str(err))
5008
        disk_dict['size'] = size
5009
      else:
5010
        # modification of disk
5011
        if 'size' in disk_dict:
5012
          raise errors.OpPrereqError("Disk size change not possible, use"
5013
                                     " grow-disk")
5014

    
5015
    if disk_addremove > 1:
5016
      raise errors.OpPrereqError("Only one disk add or remove operation"
5017
                                 " supported at a time")
5018

    
5019
    # NIC validation
5020
    nic_addremove = 0
5021
    for nic_op, nic_dict in self.op.nics:
5022
      if nic_op == constants.DDM_REMOVE:
5023
        nic_addremove += 1
5024
        continue
5025
      elif nic_op == constants.DDM_ADD:
5026
        nic_addremove += 1
5027
      else:
5028
        if not isinstance(nic_op, int):
5029
          raise errors.OpPrereqError("Invalid nic index")
5030

    
5031
      # nic_dict should be a dict
5032
      nic_ip = nic_dict.get('ip', None)
5033
      if nic_ip is not None:
5034
        if nic_ip.lower() == "none":
5035
          nic_dict['ip'] = None
5036
        else:
5037
          if not utils.IsValidIP(nic_ip):
5038
            raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip)
5039
      # we can only check None bridges and assign the default one
5040
      nic_bridge = nic_dict.get('bridge', None)
5041
      if nic_bridge is None:
5042
        nic_dict['bridge'] = self.cfg.GetDefBridge()
5043
      # but we can validate MACs
5044
      nic_mac = nic_dict.get('mac', None)
5045
      if nic_mac is not None:
5046
        if self.cfg.IsMacInUse(nic_mac):
5047
          raise errors.OpPrereqError("MAC address %s already in use"
5048
                                     " in cluster" % nic_mac)
5049
        if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
5050
          if not utils.IsValidMac(nic_mac):
5051
            raise errors.OpPrereqError("Invalid MAC address %s" % nic_mac)
5052
    if nic_addremove > 1:
5053
      raise errors.OpPrereqError("Only one NIC add or remove operation"
5054
                                 " supported at a time")
5055

    
5056
  def ExpandNames(self):
5057
    self._ExpandAndLockInstance()
5058
    self.needed_locks[locking.LEVEL_NODE] = []
5059
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
5060

    
5061
  def DeclareLocks(self, level):
5062
    if level == locking.LEVEL_NODE:
5063
      self._LockInstancesNodes()
5064

    
5065
  def BuildHooksEnv(self):
5066
    """Build hooks env.
5067

5068
    This runs on the master, primary and secondaries.
5069

5070
    """
5071
    args = dict()
5072
    if constants.BE_MEMORY in self.be_new:
5073
      args['memory'] = self.be_new[constants.BE_MEMORY]
5074
    if constants.BE_VCPUS in self.be_new:
5075
      args['vcpus'] = self.be_new[constants.BE_VCPUS]
5076
    # FIXME: readd disk/nic changes
5077
    env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
5078
    nl = [self.cfg.GetMasterNode(),
5079
          self.instance.primary_node] + list(self.instance.secondary_nodes)
5080
    return env, nl, nl
5081

    
5082
  def CheckPrereq(self):
5083
    """Check prerequisites.
5084

5085
    This only checks the instance list against the existing names.
5086

5087
    """
5088
    force = self.force = self.op.force
5089

    
5090
    # checking the new params on the primary/secondary nodes
5091

    
5092
    instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
5093
    assert self.instance is not None, \
5094
      "Cannot retrieve locked instance %s" % self.op.instance_name
5095
    pnode = self.instance.primary_node
5096
    nodelist = [pnode]
5097
    nodelist.extend(instance.secondary_nodes)
5098

    
5099
    # hvparams processing
5100
    if self.op.hvparams:
5101
      i_hvdict = copy.deepcopy(instance.hvparams)
5102
      for key, val in self.op.hvparams.iteritems():
5103
        if val == constants.VALUE_DEFAULT:
5104
          try:
5105
            del i_hvdict[key]
5106
          except KeyError:
5107
            pass
5108
        elif val == constants.VALUE_NONE:
5109
          i_hvdict[key] = None
5110
        else:
5111
          i_hvdict[key] = val
5112
      cluster = self.cfg.GetClusterInfo()
5113
      hv_new = cluster.FillDict(cluster.hvparams[instance.hypervisor],
5114
                                i_hvdict)
5115
      # local check
5116
      hypervisor.GetHypervisor(
5117
        instance.hypervisor).CheckParameterSyntax(hv_new)
5118
      _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
5119
      self.hv_new = hv_new # the new actual values
5120
      self.hv_inst = i_hvdict # the new dict (without defaults)
5121
    else:
5122
      self.hv_new = self.hv_inst = {}
5123

    
5124
    # beparams processing
5125
    if self.op.beparams:
5126
      i_bedict = copy.deepcopy(instance.beparams)
5127
      for key, val in self.op.beparams.iteritems():
5128
        if val == constants.VALUE_DEFAULT:
5129
          try:
5130
            del i_bedict[key]
5131
          except KeyError:
5132
            pass
5133
        else:
5134
          i_bedict[key] = val
5135
      cluster = self.cfg.GetClusterInfo()
5136
      be_new = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
5137
                                i_bedict)
5138
      self.be_new = be_new # the new actual values
5139
      self.be_inst = i_bedict # the new dict (without defaults)
5140
    else:
5141
      self.be_new = self.be_inst = {}
5142

    
5143
    self.warn = []
5144

    
5145
    if constants.BE_MEMORY in self.op.beparams and not self.force:
5146
      mem_check_list = [pnode]
5147
      if be_new[constants.BE_AUTO_BALANCE]:
5148
        # either we changed auto_balance to yes or it was from before
5149
        mem_check_list.extend(instance.secondary_nodes)
5150
      instance_info = self.rpc.call_instance_info(pnode, instance.name,
5151
                                                  instance.hypervisor)
5152
      nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
5153
                                         instance.hypervisor)
5154
      if nodeinfo[pnode].failed or not isinstance(nodeinfo[pnode].data, dict):
5155
        # Assume the primary node is unreachable and go ahead
5156
        self.warn.append("Can't get info from primary node %s" % pnode)
5157
      else:
5158
        if not instance_info.failed and instance_info.data:
5159
          current_mem = instance_info.data['memory']
5160
        else:
5161
          # Assume instance not running
5162
          # (there is a slight race condition here, but it's not very probable,
5163
          # and we have no other way to check)
5164
          current_mem = 0
5165
        miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
5166
                    nodeinfo[pnode].data['memory_free'])
5167
        if miss_mem > 0:
5168
          raise errors.OpPrereqError("This change will prevent the instance"
5169
                                     " from starting, due to %d MB of memory"
5170
                                     " missing on its primary node" % miss_mem)
5171

    
5172
      if be_new[constants.BE_AUTO_BALANCE]:
5173
        for node, nres in instance.secondary_nodes.iteritems():
5174
          if nres.failed or not isinstance(nres.data, dict):
5175
            self.warn.append("Can't get info from secondary node %s" % node)
5176
          elif be_new[constants.BE_MEMORY] > nres.data['memory_free']:
5177
            self.warn.append("Not enough memory to failover instance to"
5178
                             " secondary node %s" % node)
5179

    
5180
    # NIC processing
5181
    for nic_op, nic_dict in self.op.nics:
5182
      if nic_op == constants.DDM_REMOVE:
5183
        if not instance.nics:
5184
          raise errors.OpPrereqError("Instance has no NICs, cannot remove")
5185
        continue
5186
      if nic_op != constants.DDM_ADD:
5187
        # an existing nic
5188
        if nic_op < 0 or nic_op >= len(instance.nics):
5189
          raise errors.OpPrereqError("Invalid NIC index %s, valid values"
5190
                                     " are 0 to %d" %
5191
                                     (nic_op, len(instance.nics)))
5192
      nic_bridge = nic_dict.get('bridge', None)
5193
      if nic_bridge is not None:
5194
        if not self.rpc.call_bridges_exist(pnode, [nic_bridge]):
5195
          msg = ("Bridge '%s' doesn't exist on one of"
5196
                 " the instance nodes" % nic_bridge)
5197
          if self.force:
5198
            self.warn.append(msg)
5199
          else:
5200
            raise errors.OpPrereqError(msg)
5201

    
5202
    # DISK processing
5203
    if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
5204
      raise errors.OpPrereqError("Disk operations not supported for"
5205
                                 " diskless instances")
5206
    for disk_op, disk_dict in self.op.disks:
5207
      if disk_op == constants.DDM_REMOVE:
5208
        if len(instance.disks) == 1:
5209
          raise errors.OpPrereqError("Cannot remove the last disk of"
5210
                                     " an instance")
5211
        ins_l = self.rpc.call_instance_list([pnode], [instance.hypervisor])
5212
        ins_l = ins_l[pnode]
5213
        if not type(ins_l) is list:
5214
          raise errors.OpPrereqError("Can't contact node '%s'" % pnode)
5215
        if instance.name in ins_l:
5216
          raise errors.OpPrereqError("Instance is running, can't remove"
5217
                                     " disks.")
5218

    
5219
      if (disk_op == constants.DDM_ADD and
5220
          len(instance.nics) >= constants.MAX_DISKS):
5221
        raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
5222
                                   " add more" % constants.MAX_DISKS)
5223
      if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
5224
        # an existing disk
5225
        if disk_op < 0 or disk_op >= len(instance.disks):
5226
          raise errors.OpPrereqError("Invalid disk index %s, valid values"
5227
                                     " are 0 to %d" %
5228
                                     (disk_op, len(instance.disks)))
5229

    
5230
    return
5231

    
5232
  def Exec(self, feedback_fn):
5233
    """Modifies an instance.
5234

5235
    All parameters take effect only at the next restart of the instance.
5236

5237
    """
5238
    # Process here the warnings from CheckPrereq, as we don't have a
5239
    # feedback_fn there.
5240
    for warn in self.warn:
5241
      feedback_fn("WARNING: %s" % warn)
5242

    
5243
    result = []
5244
    instance = self.instance
5245
    # disk changes
5246
    for disk_op, disk_dict in self.op.disks:
5247
      if disk_op == constants.DDM_REMOVE:
5248
        # remove the last disk
5249
        device = instance.disks.pop()
5250
        device_idx = len(instance.disks)
5251
        for node, disk in device.ComputeNodeTree(instance.primary_node):
5252
          self.cfg.SetDiskID(disk, node)
5253
          result = self.rpc.call_blockdev_remove(node, disk)
5254
          if result.failed or not result.data:
5255
            self.proc.LogWarning("Could not remove disk/%d on node %s,"
5256
                                 " continuing anyway", device_idx, node)
5257
        result.append(("disk/%d" % device_idx, "remove"))
5258
      elif disk_op == constants.DDM_ADD:
5259
        # add a new disk
5260
        if instance.disk_template == constants.DT_FILE:
5261
          file_driver, file_path = instance.disks[0].logical_id
5262
          file_path = os.path.dirname(file_path)
5263
        else:
5264
          file_driver = file_path = None
5265
        disk_idx_base = len(instance.disks)
5266
        new_disk = _GenerateDiskTemplate(self,
5267
                                         instance.disk_template,
5268
                                         instance, instance.primary_node,
5269
                                         instance.secondary_nodes,
5270
                                         [disk_dict],
5271
                                         file_path,
5272
                                         file_driver,
5273
                                         disk_idx_base)[0]
5274
        new_disk.mode = disk_dict['mode']
5275
        instance.disks.append(new_disk)
5276
        info = _GetInstanceInfoText(instance)
5277

    
5278
        logging.info("Creating volume %s for instance %s",
5279
                     new_disk.iv_name, instance.name)
5280
        # Note: this needs to be kept in sync with _CreateDisks
5281
        #HARDCODE
5282
        for secondary_node in instance.secondary_nodes:
5283
          if not _CreateBlockDevOnSecondary(self, secondary_node, instance,
5284
                                            new_disk, False, info):
5285
            self.LogWarning("Failed to create volume %s (%s) on"
5286
                            " secondary node %s!",
5287
                            new_disk.iv_name, new_disk, secondary_node)
5288
        #HARDCODE
5289
        if not _CreateBlockDevOnPrimary(self, instance.primary_node,
5290
                                        instance, new_disk, info):
5291
          self.LogWarning("Failed to create volume %s on primary!",
5292
                          new_disk.iv_name)
5293
        result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
5294
                       (new_disk.size, new_disk.mode)))
5295
      else:
5296
        # change a given disk
5297
        instance.disks[disk_op].mode = disk_dict['mode']
5298
        result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
5299
    # NIC changes
5300
    for nic_op, nic_dict in self.op.nics:
5301
      if nic_op == constants.DDM_REMOVE:
5302
        # remove the last nic
5303
        del instance.nics[-1]
5304
        result.append(("nic.%d" % len(instance.nics), "remove"))
5305
      elif nic_op == constants.DDM_ADD:
5306
        # add a new nic
5307
        if 'mac' not in nic_dict:
5308
          mac = constants.VALUE_GENERATE
5309
        else:
5310
          mac = nic_dict['mac']
5311
        if mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
5312
          mac = self.cfg.GenerateMAC()
5313
        new_nic = objects.NIC(mac=mac, ip=nic_dict.get('ip', None),
5314
                              bridge=nic_dict.get('bridge', None))
5315
        instance.nics.append(new_nic)
5316
        result.append(("nic.%d" % (len(instance.nics) - 1),
5317
                       "add:mac=%s,ip=%s,bridge=%s" %
5318
                       (new_nic.mac, new_nic.ip, new_nic.bridge)))
5319
      else:
5320
        # change a given nic
5321
        for key in 'mac', 'ip', 'bridge':
5322
          if key in nic_dict:
5323
            setattr(instance.nics[nic_op], key, nic_dict[key])
5324
            result.append(("nic.%s/%d" % (key, nic_op), nic_dict[key]))
5325

    
5326
    # hvparams changes
5327
    if self.op.hvparams:
5328
      instance.hvparams = self.hv_new
5329
      for key, val in self.op.hvparams.iteritems():
5330
        result.append(("hv/%s" % key, val))
5331

    
5332
    # beparams changes
5333
    if self.op.beparams:
5334
      instance.beparams = self.be_inst
5335
      for key, val in self.op.beparams.iteritems():
5336
        result.append(("be/%s" % key, val))
5337

    
5338
    self.cfg.Update(instance)
5339

    
5340
    return result
5341

    
5342

    
5343
class LUQueryExports(NoHooksLU):
5344
  """Query the exports list
5345

5346
  """
5347
  _OP_REQP = ['nodes']
5348
  REQ_BGL = False
5349

    
5350
  def ExpandNames(self):
5351
    self.needed_locks = {}
5352
    self.share_locks[locking.LEVEL_NODE] = 1
5353
    if not self.op.nodes:
5354
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5355
    else:
5356
      self.needed_locks[locking.LEVEL_NODE] = \
5357
        _GetWantedNodes(self, self.op.nodes)
5358

    
5359
  def CheckPrereq(self):
5360
    """Check prerequisites.
5361

5362
    """
5363
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
5364

    
5365
  def Exec(self, feedback_fn):
5366
    """Compute the list of all the exported system images.
5367

5368
    @rtype: dict
5369
    @return: a dictionary with the structure node->(export-list)
5370
        where export-list is a list of the instances exported on
5371
        that node.
5372

5373
    """
5374
    rpcresult = self.rpc.call_export_list(self.nodes)
5375
    result = {}
5376
    for node in rpcresult:
5377
      if rpcresult[node].failed:
5378
        result[node] = False
5379
      else:
5380
        result[node] = rpcresult[node].data
5381

    
5382
    return result
5383

    
5384

    
5385
class LUExportInstance(LogicalUnit):
5386
  """Export an instance to an image in the cluster.
5387

5388
  """
5389
  HPATH = "instance-export"
5390
  HTYPE = constants.HTYPE_INSTANCE
5391
  _OP_REQP = ["instance_name", "target_node", "shutdown"]
5392
  REQ_BGL = False
5393

    
5394
  def ExpandNames(self):
5395
    self._ExpandAndLockInstance()
5396
    # FIXME: lock only instance primary and destination node
5397
    #
5398
    # Sad but true, for now we have do lock all nodes, as we don't know where
5399
    # the previous export might be, and and in this LU we search for it and
5400
    # remove it from its current node. In the future we could fix this by:
5401
    #  - making a tasklet to search (share-lock all), then create the new one,
5402
    #    then one to remove, after
5403
    #  - removing the removal operation altoghether
5404
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5405

    
5406
  def DeclareLocks(self, level):
5407
    """Last minute lock declaration."""
5408
    # All nodes are locked anyway, so nothing to do here.
5409

    
5410
  def BuildHooksEnv(self):
5411
    """Build hooks env.
5412

5413
    This will run on the master, primary node and target node.
5414

5415
    """
5416
    env = {
5417
      "EXPORT_NODE": self.op.target_node,
5418
      "EXPORT_DO_SHUTDOWN": self.op.shutdown,
5419
      }
5420
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
5421
    nl = [self.cfg.GetMasterNode(), self.instance.primary_node,
5422
          self.op.target_node]
5423
    return env, nl, nl
5424

    
5425
  def CheckPrereq(self):
5426
    """Check prerequisites.
5427

5428
    This checks that the instance and node names are valid.
5429

5430
    """
5431
    instance_name = self.op.instance_name
5432
    self.instance = self.cfg.GetInstanceInfo(instance_name)
5433
    assert self.instance is not None, \
5434
          "Cannot retrieve locked instance %s" % self.op.instance_name
5435

    
5436
    self.dst_node = self.cfg.GetNodeInfo(
5437
      self.cfg.ExpandNodeName(self.op.target_node))
5438

    
5439
    if self.dst_node is None:
5440
      # This is wrong node name, not a non-locked node
5441
      raise errors.OpPrereqError("Wrong node name %s" % self.op.target_node)
5442

    
5443
    # instance disk type verification
5444
    for disk in self.instance.disks:
5445
      if disk.dev_type == constants.LD_FILE:
5446
        raise errors.OpPrereqError("Export not supported for instances with"
5447
                                   " file-based disks")
5448

    
5449
  def Exec(self, feedback_fn):
5450
    """Export an instance to an image in the cluster.
5451

5452
    """
5453
    instance = self.instance
5454
    dst_node = self.dst_node
5455
    src_node = instance.primary_node
5456
    if self.op.shutdown:
5457
      # shutdown the instance, but not the disks
5458
      result = self.rpc.call_instance_shutdown(src_node, instance)
5459
      result.Raise()
5460
      if not result.data:
5461
        raise errors.OpExecError("Could not shutdown instance %s on node %s" %
5462
                                 (instance.name, src_node))
5463

    
5464
    vgname = self.cfg.GetVGName()
5465

    
5466
    snap_disks = []
5467

    
5468
    try:
5469
      for disk in instance.disks:
5470
        # new_dev_name will be a snapshot of an lvm leaf of the one we passed
5471
        new_dev_name = self.rpc.call_blockdev_snapshot(src_node, disk)
5472
        if new_dev_name.failed or not new_dev_name.data:
5473
          self.LogWarning("Could not snapshot block device %s on node %s",
5474
                          disk.logical_id[1], src_node)
5475
          snap_disks.append(False)
5476
        else:
5477
          new_dev = objects.Disk(dev_type=constants.LD_LV, size=disk.size,
5478
                                 logical_id=(vgname, new_dev_name.data),
5479
                                 physical_id=(vgname, new_dev_name.data),
5480
                                 iv_name=disk.iv_name)
5481
          snap_disks.append(new_dev)
5482

    
5483
    finally:
5484
      if self.op.shutdown and instance.status == "up":
5485
        result = self.rpc.call_instance_start(src_node, instance, None)
5486
        if result.failed or not result.data:
5487
          _ShutdownInstanceDisks(self, instance)
5488
          raise errors.OpExecError("Could not start instance")
5489

    
5490
    # TODO: check for size
5491

    
5492
    cluster_name = self.cfg.GetClusterName()
5493
    for idx, dev in enumerate(snap_disks):
5494
      if dev:
5495
        result = self.rpc.call_snapshot_export(src_node, dev, dst_node.name,
5496
                                               instance, cluster_name, idx)
5497
        if result.failed or not result.data:
5498
          self.LogWarning("Could not export block device %s from node %s to"
5499
                          " node %s", dev.logical_id[1], src_node,
5500
                          dst_node.name)
5501
        result = self.rpc.call_blockdev_remove(src_node, dev)
5502
        if result.failed or not result.data:
5503
          self.LogWarning("Could not remove snapshot block device %s from node"
5504
                          " %s", dev.logical_id[1], src_node)
5505

    
5506
    result = self.rpc.call_finalize_export(dst_node.name, instance, snap_disks)
5507
    if result.failed or not result.data:
5508
      self.LogWarning("Could not finalize export for instance %s on node %s",
5509
                      instance.name, dst_node.name)
5510

    
5511
    nodelist = self.cfg.GetNodeList()
5512
    nodelist.remove(dst_node.name)
5513

    
5514
    # on one-node clusters nodelist will be empty after the removal
5515
    # if we proceed the backup would be removed because OpQueryExports
5516
    # substitutes an empty list with the full cluster node list.
5517
    if nodelist:
5518
      exportlist = self.rpc.call_export_list(nodelist)
5519
      for node in exportlist:
5520
        if exportlist[node].failed:
5521
          continue
5522
        if instance.name in exportlist[node].data:
5523
          if not self.rpc.call_export_remove(node, instance.name):
5524
            self.LogWarning("Could not remove older export for instance %s"
5525
                            " on node %s", instance.name, node)
5526

    
5527

    
5528
class LURemoveExport(NoHooksLU):
5529
  """Remove exports related to the named instance.
5530

5531
  """
5532
  _OP_REQP = ["instance_name"]
5533
  REQ_BGL = False
5534

    
5535
  def ExpandNames(self):
5536
    self.needed_locks = {}
5537
    # We need all nodes to be locked in order for RemoveExport to work, but we
5538
    # don't need to lock the instance itself, as nothing will happen to it (and
5539
    # we can remove exports also for a removed instance)
5540
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5541

    
5542
  def CheckPrereq(self):
5543
    """Check prerequisites.
5544
    """
5545
    pass
5546

    
5547
  def Exec(self, feedback_fn):
5548
    """Remove any export.
5549

5550
    """
5551
    instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
5552
    # If the instance was not found we'll try with the name that was passed in.
5553
    # This will only work if it was an FQDN, though.
5554
    fqdn_warn = False
5555
    if not instance_name:
5556
      fqdn_warn = True
5557
      instance_name = self.op.instance_name
5558

    
5559
    exportlist = self.rpc.call_export_list(self.acquired_locks[
5560
      locking.LEVEL_NODE])
5561
    found = False
5562
    for node in exportlist:
5563
      if exportlist[node].failed:
5564
        self.LogWarning("Failed to query node %s, continuing" % node)
5565
        continue
5566
      if instance_name in exportlist[node].data:
5567
        found = True
5568
        result = self.rpc.call_export_remove(node, instance_name)
5569
        if result.failed or not result.data:
5570
          logging.error("Could not remove export for instance %s"
5571
                        " on node %s", instance_name, node)
5572

    
5573
    if fqdn_warn and not found:
5574
      feedback_fn("Export not found. If trying to remove an export belonging"
5575
                  " to a deleted instance please use its Fully Qualified"
5576
                  " Domain Name.")
5577

    
5578

    
5579
class TagsLU(NoHooksLU):
5580
  """Generic tags LU.
5581

5582
  This is an abstract class which is the parent of all the other tags LUs.
5583

5584
  """
5585

    
5586
  def ExpandNames(self):
5587
    self.needed_locks = {}
5588
    if self.op.kind == constants.TAG_NODE:
5589
      name = self.cfg.ExpandNodeName(self.op.name)
5590
      if name is None:
5591
        raise errors.OpPrereqError("Invalid node name (%s)" %
5592
                                   (self.op.name,))
5593
      self.op.name = name
5594
      self.needed_locks[locking.LEVEL_NODE] = name
5595
    elif self.op.kind == constants.TAG_INSTANCE:
5596
      name = self.cfg.ExpandInstanceName(self.op.name)
5597
      if name is None:
5598
        raise errors.OpPrereqError("Invalid instance name (%s)" %
5599
                                   (self.op.name,))
5600
      self.op.name = name
5601
      self.needed_locks[locking.LEVEL_INSTANCE] = name
5602

    
5603
  def CheckPrereq(self):
5604
    """Check prerequisites.
5605

5606
    """
5607
    if self.op.kind == constants.TAG_CLUSTER:
5608
      self.target = self.cfg.GetClusterInfo()
5609
    elif self.op.kind == constants.TAG_NODE:
5610
      self.target = self.cfg.GetNodeInfo(self.op.name)
5611
    elif self.op.kind == constants.TAG_INSTANCE:
5612
      self.target = self.cfg.GetInstanceInfo(self.op.name)
5613
    else:
5614
      raise errors.OpPrereqError("Wrong tag type requested (%s)" %
5615
                                 str(self.op.kind))
5616

    
5617

    
5618
class LUGetTags(TagsLU):
5619
  """Returns the tags of a given object.
5620

5621
  """
5622
  _OP_REQP = ["kind", "name"]
5623
  REQ_BGL = False
5624

    
5625
  def Exec(self, feedback_fn):
5626
    """Returns the tag list.
5627

5628
    """
5629
    return list(self.target.GetTags())
5630

    
5631

    
5632
class LUSearchTags(NoHooksLU):
5633
  """Searches the tags for a given pattern.
5634

5635
  """
5636
  _OP_REQP = ["pattern"]
5637
  REQ_BGL = False
5638

    
5639
  def ExpandNames(self):
5640
    self.needed_locks = {}
5641

    
5642
  def CheckPrereq(self):
5643
    """Check prerequisites.
5644

5645
    This checks the pattern passed for validity by compiling it.
5646

5647
    """
5648
    try:
5649
      self.re = re.compile(self.op.pattern)
5650
    except re.error, err:
5651
      raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
5652
                                 (self.op.pattern, err))
5653

    
5654
  def Exec(self, feedback_fn):
5655
    """Returns the tag list.
5656

5657
    """
5658
    cfg = self.cfg
5659
    tgts = [("/cluster", cfg.GetClusterInfo())]
5660
    ilist = cfg.GetAllInstancesInfo().values()
5661
    tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
5662
    nlist = cfg.GetAllNodesInfo().values()
5663
    tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
5664
    results = []
5665
    for path, target in tgts:
5666
      for tag in target.GetTags():
5667
        if self.re.search(tag):
5668
          results.append((path, tag))
5669
    return results
5670

    
5671

    
5672
class LUAddTags(TagsLU):
5673
  """Sets a tag on a given object.
5674

5675
  """
5676
  _OP_REQP = ["kind", "name", "tags"]
5677
  REQ_BGL = False
5678

    
5679
  def CheckPrereq(self):
5680
    """Check prerequisites.
5681

5682
    This checks the type and length of the tag name and value.
5683

5684
    """
5685
    TagsLU.CheckPrereq(self)
5686
    for tag in self.op.tags:
5687
      objects.TaggableObject.ValidateTag(tag)
5688

    
5689
  def Exec(self, feedback_fn):
5690
    """Sets the tag.
5691

5692
    """
5693
    try:
5694
      for tag in self.op.tags:
5695
        self.target.AddTag(tag)
5696
    except errors.TagError, err:
5697
      raise errors.OpExecError("Error while setting tag: %s" % str(err))
5698
    try:
5699
      self.cfg.Update(self.target)
5700
    except errors.ConfigurationError:
5701
      raise errors.OpRetryError("There has been a modification to the"
5702
                                " config file and the operation has been"
5703
                                " aborted. Please retry.")
5704

    
5705

    
5706
class LUDelTags(TagsLU):
5707
  """Delete a list of tags from a given object.
5708

5709
  """
5710
  _OP_REQP = ["kind", "name", "tags"]
5711
  REQ_BGL = False
5712

    
5713
  def CheckPrereq(self):
5714
    """Check prerequisites.
5715

5716
    This checks that we have the given tag.
5717

5718
    """
5719
    TagsLU.CheckPrereq(self)
5720
    for tag in self.op.tags:
5721
      objects.TaggableObject.ValidateTag(tag)
5722
    del_tags = frozenset(self.op.tags)
5723
    cur_tags = self.target.GetTags()
5724
    if not del_tags <= cur_tags:
5725
      diff_tags = del_tags - cur_tags
5726
      diff_names = ["'%s'" % tag for tag in diff_tags]
5727
      diff_names.sort()
5728
      raise errors.OpPrereqError("Tag(s) %s not found" %
5729
                                 (",".join(diff_names)))
5730

    
5731
  def Exec(self, feedback_fn):
5732
    """Remove the tag from the object.
5733

5734
    """
5735
    for tag in self.op.tags:
5736
      self.target.RemoveTag(tag)
5737
    try:
5738
      self.cfg.Update(self.target)
5739
    except errors.ConfigurationError:
5740
      raise errors.OpRetryError("There has been a modification to the"
5741
                                " config file and the operation has been"
5742
                                " aborted. Please retry.")
5743

    
5744

    
5745
class LUTestDelay(NoHooksLU):
5746
  """Sleep for a specified amount of time.
5747

5748
  This LU sleeps on the master and/or nodes for a specified amount of
5749
  time.
5750

5751
  """
5752
  _OP_REQP = ["duration", "on_master", "on_nodes"]
5753
  REQ_BGL = False
5754

    
5755
  def ExpandNames(self):
5756
    """Expand names and set required locks.
5757

5758
    This expands the node list, if any.
5759

5760
    """
5761
    self.needed_locks = {}
5762
    if self.op.on_nodes:
5763
      # _GetWantedNodes can be used here, but is not always appropriate to use
5764
      # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
5765
      # more information.
5766
      self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
5767
      self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
5768

    
5769
  def CheckPrereq(self):
5770
    """Check prerequisites.
5771

5772
    """
5773

    
5774
  def Exec(self, feedback_fn):
5775
    """Do the actual sleep.
5776

5777
    """
5778
    if self.op.on_master:
5779
      if not utils.TestDelay(self.op.duration):
5780
        raise errors.OpExecError("Error during master delay test")
5781
    if self.op.on_nodes:
5782
      result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
5783
      if not result:
5784
        raise errors.OpExecError("Complete failure from rpc call")
5785
      for node, node_result in result.items():
5786
        node_result.Raise()
5787
        if not node_result.data:
5788
          raise errors.OpExecError("Failure during rpc call to node %s,"
5789
                                   " result: %s" % (node, node_result.data))
5790

    
5791

    
5792
class IAllocator(object):
5793
  """IAllocator framework.
5794

5795
  An IAllocator instance has three sets of attributes:
5796
    - cfg that is needed to query the cluster
5797
    - input data (all members of the _KEYS class attribute are required)
5798
    - four buffer attributes (in|out_data|text), that represent the
5799
      input (to the external script) in text and data structure format,
5800
      and the output from it, again in two formats
5801
    - the result variables from the script (success, info, nodes) for
5802
      easy usage
5803

5804
  """
5805
  _ALLO_KEYS = [
5806
    "mem_size", "disks", "disk_template",
5807
    "os", "tags", "nics", "vcpus", "hypervisor",
5808
    ]
5809
  _RELO_KEYS = [
5810
    "relocate_from",
5811
    ]
5812

    
5813
  def __init__(self, lu, mode, name, **kwargs):
5814
    self.lu = lu
5815
    # init buffer variables
5816
    self.in_text = self.out_text = self.in_data = self.out_data = None
5817
    # init all input fields so that pylint is happy
5818
    self.mode = mode
5819
    self.name = name
5820
    self.mem_size = self.disks = self.disk_template = None
5821
    self.os = self.tags = self.nics = self.vcpus = None
5822
    self.relocate_from = None
5823
    # computed fields
5824
    self.required_nodes = None
5825
    # init result fields
5826
    self.success = self.info = self.nodes = None
5827
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
5828
      keyset = self._ALLO_KEYS
5829
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
5830
      keyset = self._RELO_KEYS
5831
    else:
5832
      raise errors.ProgrammerError("Unknown mode '%s' passed to the"
5833
                                   " IAllocator" % self.mode)
5834
    for key in kwargs:
5835
      if key not in keyset:
5836
        raise errors.ProgrammerError("Invalid input parameter '%s' to"
5837
                                     " IAllocator" % key)
5838
      setattr(self, key, kwargs[key])
5839
    for key in keyset:
5840
      if key not in kwargs:
5841
        raise errors.ProgrammerError("Missing input parameter '%s' to"
5842
                                     " IAllocator" % key)
5843
    self._BuildInputData()
5844

    
5845
  def _ComputeClusterData(self):
5846
    """Compute the generic allocator input data.
5847

5848
    This is the data that is independent of the actual operation.
5849

5850
    """
5851
    cfg = self.lu.cfg
5852
    cluster_info = cfg.GetClusterInfo()
5853
    # cluster data
5854
    data = {
5855
      "version": 1,
5856
      "cluster_name": cfg.GetClusterName(),
5857
      "cluster_tags": list(cluster_info.GetTags()),
5858
      "enable_hypervisors": list(cluster_info.enabled_hypervisors),
5859
      # we don't have job IDs
5860
      }
5861
    iinfo = cfg.GetAllInstancesInfo().values()
5862
    i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
5863

    
5864
    # node data
5865
    node_results = {}
5866
    node_list = cfg.GetNodeList()
5867

    
5868
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
5869
      hypervisor = self.hypervisor
5870
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
5871
      hypervisor = cfg.GetInstanceInfo(self.name).hypervisor
5872

    
5873
    node_data = self.lu.rpc.call_node_info(node_list, cfg.GetVGName(),
5874
                                           hypervisor)
5875
    node_iinfo = self.lu.rpc.call_all_instances_info(node_list,
5876
                       cluster_info.enabled_hypervisors)
5877
    for nname in node_list:
5878
      ninfo = cfg.GetNodeInfo(nname)
5879
      node_data[nname].Raise()
5880
      if not isinstance(node_data[nname].data, dict):
5881
        raise errors.OpExecError("Can't get data for node %s" % nname)
5882
      remote_info = node_data[nname].data
5883
      for attr in ['memory_total', 'memory_free', 'memory_dom0',
5884
                   'vg_size', 'vg_free', 'cpu_total']:
5885
        if attr not in remote_info:
5886
          raise errors.OpExecError("Node '%s' didn't return attribute '%s'" %
5887
                                   (nname, attr))
5888
        try:
5889
          remote_info[attr] = int(remote_info[attr])
5890
        except ValueError, err:
5891
          raise errors.OpExecError("Node '%s' returned invalid value for '%s':"
5892
                                   " %s" % (nname, attr, str(err)))
5893
      # compute memory used by primary instances
5894
      i_p_mem = i_p_up_mem = 0
5895
      for iinfo, beinfo in i_list:
5896
        if iinfo.primary_node == nname:
5897
          i_p_mem += beinfo[constants.BE_MEMORY]
5898
          if iinfo.name not in node_iinfo[nname]:
5899
            i_used_mem = 0
5900
          else:
5901
            i_used_mem = int(node_iinfo[nname][iinfo.name]['memory'])
5902
          i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
5903
          remote_info['memory_free'] -= max(0, i_mem_diff)
5904

    
5905
          if iinfo.status == "up":
5906
            i_p_up_mem += beinfo[constants.BE_MEMORY]
5907

    
5908
      # compute memory used by instances
5909
      pnr = {
5910
        "tags": list(ninfo.GetTags()),
5911
        "total_memory": remote_info['memory_total'],
5912
        "reserved_memory": remote_info['memory_dom0'],
5913
        "free_memory": remote_info['memory_free'],
5914
        "i_pri_memory": i_p_mem,
5915
        "i_pri_up_memory": i_p_up_mem,
5916
        "total_disk": remote_info['vg_size'],
5917
        "free_disk": remote_info['vg_free'],
5918
        "primary_ip": ninfo.primary_ip,
5919
        "secondary_ip": ninfo.secondary_ip,
5920
        "total_cpus": remote_info['cpu_total'],
5921
        "offline": ninfo.offline,
5922
        }
5923
      node_results[nname] = pnr
5924
    data["nodes"] = node_results
5925

    
5926
    # instance data
5927
    instance_data = {}
5928
    for iinfo, beinfo in i_list:
5929
      nic_data = [{"mac": n.mac, "ip": n.ip, "bridge": n.bridge}
5930
                  for n in iinfo.nics]
5931
      pir = {
5932
        "tags": list(iinfo.GetTags()),
5933
        "should_run": iinfo.status == "up",
5934
        "vcpus": beinfo[constants.BE_VCPUS],
5935
        "memory": beinfo[constants.BE_MEMORY],
5936
        "os": iinfo.os,
5937
        "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
5938
        "nics": nic_data,
5939
        "disks": [{"size": dsk.size, "mode": "w"} for dsk in iinfo.disks],
5940
        "disk_template": iinfo.disk_template,
5941
        "hypervisor": iinfo.hypervisor,
5942
        }
5943
      instance_data[iinfo.name] = pir
5944

    
5945
    data["instances"] = instance_data
5946

    
5947
    self.in_data = data
5948

    
5949
  def _AddNewInstance(self):
5950
    """Add new instance data to allocator structure.
5951

5952
    This in combination with _AllocatorGetClusterData will create the
5953
    correct structure needed as input for the allocator.
5954

5955
    The checks for the completeness of the opcode must have already been
5956
    done.
5957

5958
    """
5959
    data = self.in_data
5960
    if len(self.disks) != 2:
5961
      raise errors.OpExecError("Only two-disk configurations supported")
5962

    
5963
    disk_space = _ComputeDiskSize(self.disk_template, self.disks)
5964

    
5965
    if self.disk_template in constants.DTS_NET_MIRROR:
5966
      self.required_nodes = 2
5967
    else:
5968
      self.required_nodes = 1
5969
    request = {
5970
      "type": "allocate",
5971
      "name": self.name,
5972
      "disk_template": self.disk_template,
5973
      "tags": self.tags,
5974
      "os": self.os,
5975
      "vcpus": self.vcpus,
5976
      "memory": self.mem_size,
5977
      "disks": self.disks,
5978
      "disk_space_total": disk_space,
5979
      "nics": self.nics,
5980
      "required_nodes": self.required_nodes,
5981
      }
5982
    data["request"] = request
5983

    
5984
  def _AddRelocateInstance(self):
5985
    """Add relocate instance data to allocator structure.
5986

5987
    This in combination with _IAllocatorGetClusterData will create the
5988
    correct structure needed as input for the allocator.
5989

5990
    The checks for the completeness of the opcode must have already been
5991
    done.
5992

5993
    """
5994
    instance = self.lu.cfg.GetInstanceInfo(self.name)
5995
    if instance is None:
5996
      raise errors.ProgrammerError("Unknown instance '%s' passed to"
5997
                                   " IAllocator" % self.name)
5998

    
5999
    if instance.disk_template not in constants.DTS_NET_MIRROR:
6000
      raise errors.OpPrereqError("Can't relocate non-mirrored instances")
6001

    
6002
    if len(instance.secondary_nodes) != 1:
6003
      raise errors.OpPrereqError("Instance has not exactly one secondary node")
6004

    
6005
    self.required_nodes = 1
6006
    disk_sizes = [{'size': disk.size} for disk in instance.disks]
6007
    disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
6008

    
6009
    request = {
6010
      "type": "relocate",
6011
      "name": self.name,
6012
      "disk_space_total": disk_space,
6013
      "required_nodes": self.required_nodes,
6014
      "relocate_from": self.relocate_from,
6015
      }
6016
    self.in_data["request"] = request
6017

    
6018
  def _BuildInputData(self):
6019
    """Build input data structures.
6020

6021
    """
6022
    self._ComputeClusterData()
6023

    
6024
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
6025
      self._AddNewInstance()
6026
    else:
6027
      self._AddRelocateInstance()
6028

    
6029
    self.in_text = serializer.Dump(self.in_data)
6030

    
6031
  def Run(self, name, validate=True, call_fn=None):
6032
    """Run an instance allocator and return the results.
6033

6034
    """
6035
    if call_fn is None:
6036
      call_fn = self.lu.rpc.call_iallocator_runner
6037
    data = self.in_text
6038

    
6039
    result = call_fn(self.lu.cfg.GetMasterNode(), name, self.in_text)
6040
    result.Raise()
6041

    
6042
    if not isinstance(result.data, (list, tuple)) or len(result.data) != 4:
6043
      raise errors.OpExecError("Invalid result from master iallocator runner")
6044

    
6045
    rcode, stdout, stderr, fail = result.data
6046

    
6047
    if rcode == constants.IARUN_NOTFOUND:
6048
      raise errors.OpExecError("Can't find allocator '%s'" % name)
6049
    elif rcode == constants.IARUN_FAILURE:
6050
      raise errors.OpExecError("Instance allocator call failed: %s,"
6051
                               " output: %s" % (fail, stdout+stderr))
6052
    self.out_text = stdout
6053
    if validate:
6054
      self._ValidateResult()
6055

    
6056
  def _ValidateResult(self):
6057
    """Process the allocator results.
6058

6059
    This will process and if successful save the result in
6060
    self.out_data and the other parameters.
6061

6062
    """
6063
    try:
6064
      rdict = serializer.Load(self.out_text)
6065
    except Exception, err:
6066
      raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
6067

    
6068
    if not isinstance(rdict, dict):
6069
      raise errors.OpExecError("Can't parse iallocator results: not a dict")
6070

    
6071
    for key in "success", "info", "nodes":
6072
      if key not in rdict:
6073
        raise errors.OpExecError("Can't parse iallocator results:"
6074
                                 " missing key '%s'" % key)
6075
      setattr(self, key, rdict[key])
6076

    
6077
    if not isinstance(rdict["nodes"], list):
6078
      raise errors.OpExecError("Can't parse iallocator results: 'nodes' key"
6079
                               " is not a list")
6080
    self.out_data = rdict
6081

    
6082

    
6083
class LUTestAllocator(NoHooksLU):
6084
  """Run allocator tests.
6085

6086
  This LU runs the allocator tests
6087

6088
  """
6089
  _OP_REQP = ["direction", "mode", "name"]
6090

    
6091
  def CheckPrereq(self):
6092
    """Check prerequisites.
6093

6094
    This checks the opcode parameters depending on the director and mode test.
6095

6096
    """
6097
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
6098
      for attr in ["name", "mem_size", "disks", "disk_template",
6099
                   "os", "tags", "nics", "vcpus"]:
6100
        if not hasattr(self.op, attr):
6101
          raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
6102
                                     attr)
6103
      iname = self.cfg.ExpandInstanceName(self.op.name)
6104
      if iname is not None:
6105
        raise errors.OpPrereqError("Instance '%s' already in the cluster" %
6106
                                   iname)
6107
      if not isinstance(self.op.nics, list):
6108
        raise errors.OpPrereqError("Invalid parameter 'nics'")
6109
      for row in self.op.nics:
6110
        if (not isinstance(row, dict) or
6111
            "mac" not in row or
6112
            "ip" not in row or
6113
            "bridge" not in row):
6114
          raise errors.OpPrereqError("Invalid contents of the"
6115
                                     " 'nics' parameter")
6116
      if not isinstance(self.op.disks, list):
6117
        raise errors.OpPrereqError("Invalid parameter 'disks'")
6118
      if len(self.op.disks) != 2:
6119
        raise errors.OpPrereqError("Only two-disk configurations supported")
6120
      for row in self.op.disks:
6121
        if (not isinstance(row, dict) or
6122
            "size" not in row or
6123
            not isinstance(row["size"], int) or
6124
            "mode" not in row or
6125
            row["mode"] not in ['r', 'w']):
6126
          raise errors.OpPrereqError("Invalid contents of the"
6127
                                     " 'disks' parameter")
6128
      if self.op.hypervisor is None:
6129
        self.op.hypervisor = self.cfg.GetHypervisorType()
6130
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
6131
      if not hasattr(self.op, "name"):
6132
        raise errors.OpPrereqError("Missing attribute 'name' on opcode input")
6133
      fname = self.cfg.ExpandInstanceName(self.op.name)
6134
      if fname is None:
6135
        raise errors.OpPrereqError("Instance '%s' not found for relocation" %
6136
                                   self.op.name)
6137
      self.op.name = fname
6138
      self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
6139
    else:
6140
      raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
6141
                                 self.op.mode)
6142

    
6143
    if self.op.direction == constants.IALLOCATOR_DIR_OUT:
6144
      if not hasattr(self.op, "allocator") or self.op.allocator is None:
6145
        raise errors.OpPrereqError("Missing allocator name")
6146
    elif self.op.direction != constants.IALLOCATOR_DIR_IN:
6147
      raise errors.OpPrereqError("Wrong allocator test '%s'" %
6148
                                 self.op.direction)
6149

    
6150
  def Exec(self, feedback_fn):
6151
    """Run the allocator test.
6152

6153
    """
6154
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
6155
      ial = IAllocator(self,
6156
                       mode=self.op.mode,
6157
                       name=self.op.name,
6158
                       mem_size=self.op.mem_size,
6159
                       disks=self.op.disks,
6160
                       disk_template=self.op.disk_template,
6161
                       os=self.op.os,
6162
                       tags=self.op.tags,
6163
                       nics=self.op.nics,
6164
                       vcpus=self.op.vcpus,
6165
                       hypervisor=self.op.hypervisor,
6166
                       )
6167
    else:
6168
      ial = IAllocator(self,
6169
                       mode=self.op.mode,
6170
                       name=self.op.name,
6171
                       relocate_from=list(self.relocate_from),
6172
                       )
6173

    
6174
    if self.op.direction == constants.IALLOCATOR_DIR_IN:
6175
      result = ial.in_text
6176
    else:
6177
      ial.Run(self.op.allocator, validate=False)
6178
      result = ial.out_text
6179
    return result