Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ 7260cfbe

History | View | Annotate | Download (309.8 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=W0201
25

    
26
# W0201 since most LU attributes are defined in CheckPrereq or similar
27
# functions
28

    
29
import os
30
import os.path
31
import time
32
import re
33
import platform
34
import logging
35
import copy
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 serializer
45
from ganeti import ssconf
46

    
47

    
48
class LogicalUnit(object):
49
  """Logical Unit base class.
50

51
  Subclasses must follow these rules:
52
    - implement ExpandNames
53
    - implement CheckPrereq (except when tasklets are used)
54
    - implement Exec (except when tasklets are used)
55
    - implement BuildHooksEnv
56
    - redefine HPATH and HTYPE
57
    - optionally redefine their run requirements:
58
        REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
59

60
  Note that all commands require root permissions.
61

62
  @ivar dry_run_result: the value (if any) that will be returned to the caller
63
      in dry-run mode (signalled by opcode dry_run parameter)
64

65
  """
66
  HPATH = None
67
  HTYPE = None
68
  _OP_REQP = []
69
  REQ_BGL = True
70

    
71
  def __init__(self, processor, op, context, rpc):
72
    """Constructor for LogicalUnit.
73

74
    This needs to be overridden in derived classes in order to check op
75
    validity.
76

77
    """
78
    self.proc = processor
79
    self.op = op
80
    self.cfg = context.cfg
81
    self.context = context
82
    self.rpc = rpc
83
    # Dicts used to declare locking needs to mcpu
84
    self.needed_locks = None
85
    self.acquired_locks = {}
86
    self.share_locks = dict.fromkeys(locking.LEVELS, 0)
87
    self.add_locks = {}
88
    self.remove_locks = {}
89
    # Used to force good behavior when calling helper functions
90
    self.recalculate_locks = {}
91
    self.__ssh = None
92
    # logging
93
    self.LogWarning = processor.LogWarning # pylint: disable-msg=C0103
94
    self.LogInfo = processor.LogInfo # pylint: disable-msg=C0103
95
    self.LogStep = processor.LogStep # pylint: disable-msg=C0103
96
    # support for dry-run
97
    self.dry_run_result = None
98

    
99
    # Tasklets
100
    self.tasklets = None
101

    
102
    for attr_name in self._OP_REQP:
103
      attr_val = getattr(op, attr_name, None)
104
      if attr_val is None:
105
        raise errors.OpPrereqError("Required parameter '%s' missing" %
106
                                   attr_name, errors.ECODE_INVAL)
107

    
108
    self.CheckArguments()
109

    
110
  def __GetSSH(self):
111
    """Returns the SshRunner object
112

113
    """
114
    if not self.__ssh:
115
      self.__ssh = ssh.SshRunner(self.cfg.GetClusterName())
116
    return self.__ssh
117

    
118
  ssh = property(fget=__GetSSH)
119

    
120
  def CheckArguments(self):
121
    """Check syntactic validity for the opcode arguments.
122

123
    This method is for doing a simple syntactic check and ensure
124
    validity of opcode parameters, without any cluster-related
125
    checks. While the same can be accomplished in ExpandNames and/or
126
    CheckPrereq, doing these separate is better because:
127

128
      - ExpandNames is left as as purely a lock-related function
129
      - CheckPrereq is run after we have acquired locks (and possible
130
        waited for them)
131

132
    The function is allowed to change the self.op attribute so that
133
    later methods can no longer worry about missing parameters.
134

135
    """
136
    pass
137

    
138
  def ExpandNames(self):
139
    """Expand names for this LU.
140

141
    This method is called before starting to execute the opcode, and it should
142
    update all the parameters of the opcode to their canonical form (e.g. a
143
    short node name must be fully expanded after this method has successfully
144
    completed). This way locking, hooks, logging, ecc. can work correctly.
145

146
    LUs which implement this method must also populate the self.needed_locks
147
    member, as a dict with lock levels as keys, and a list of needed lock names
148
    as values. Rules:
149

150
      - use an empty dict if you don't need any lock
151
      - if you don't need any lock at a particular level omit that level
152
      - don't put anything for the BGL level
153
      - if you want all locks at a level use locking.ALL_SET as a value
154

155
    If you need to share locks (rather than acquire them exclusively) at one
156
    level you can modify self.share_locks, setting a true value (usually 1) for
157
    that level. By default locks are not shared.
158

159
    This function can also define a list of tasklets, which then will be
160
    executed in order instead of the usual LU-level CheckPrereq and Exec
161
    functions, if those are not defined by the LU.
162

163
    Examples::
164

165
      # Acquire all nodes and one instance
166
      self.needed_locks = {
167
        locking.LEVEL_NODE: locking.ALL_SET,
168
        locking.LEVEL_INSTANCE: ['instance1.example.tld'],
169
      }
170
      # Acquire just two nodes
171
      self.needed_locks = {
172
        locking.LEVEL_NODE: ['node1.example.tld', 'node2.example.tld'],
173
      }
174
      # Acquire no locks
175
      self.needed_locks = {} # No, you can't leave it to the default value None
176

177
    """
178
    # The implementation of this method is mandatory only if the new LU is
179
    # concurrent, so that old LUs don't need to be changed all at the same
180
    # time.
181
    if self.REQ_BGL:
182
      self.needed_locks = {} # Exclusive LUs don't need locks.
183
    else:
184
      raise NotImplementedError
185

    
186
  def DeclareLocks(self, level):
187
    """Declare LU locking needs for a level
188

189
    While most LUs can just declare their locking needs at ExpandNames time,
190
    sometimes there's the need to calculate some locks after having acquired
191
    the ones before. This function is called just before acquiring locks at a
192
    particular level, but after acquiring the ones at lower levels, and permits
193
    such calculations. It can be used to modify self.needed_locks, and by
194
    default it does nothing.
195

196
    This function is only called if you have something already set in
197
    self.needed_locks for the level.
198

199
    @param level: Locking level which is going to be locked
200
    @type level: member of ganeti.locking.LEVELS
201

202
    """
203

    
204
  def CheckPrereq(self):
205
    """Check prerequisites for this LU.
206

207
    This method should check that the prerequisites for the execution
208
    of this LU are fulfilled. It can do internode communication, but
209
    it should be idempotent - no cluster or system changes are
210
    allowed.
211

212
    The method should raise errors.OpPrereqError in case something is
213
    not fulfilled. Its return value is ignored.
214

215
    This method should also update all the parameters of the opcode to
216
    their canonical form if it hasn't been done by ExpandNames before.
217

218
    """
219
    if self.tasklets is not None:
220
      for (idx, tl) in enumerate(self.tasklets):
221
        logging.debug("Checking prerequisites for tasklet %s/%s",
222
                      idx + 1, len(self.tasklets))
223
        tl.CheckPrereq()
224
    else:
225
      raise NotImplementedError
226

    
227
  def Exec(self, feedback_fn):
228
    """Execute the LU.
229

230
    This method should implement the actual work. It should raise
231
    errors.OpExecError for failures that are somewhat dealt with in
232
    code, or expected.
233

234
    """
235
    if self.tasklets is not None:
236
      for (idx, tl) in enumerate(self.tasklets):
237
        logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets))
238
        tl.Exec(feedback_fn)
239
    else:
240
      raise NotImplementedError
241

    
242
  def BuildHooksEnv(self):
243
    """Build hooks environment for this LU.
244

245
    This method should return a three-node tuple consisting of: a dict
246
    containing the environment that will be used for running the
247
    specific hook for this LU, a list of node names on which the hook
248
    should run before the execution, and a list of node names on which
249
    the hook should run after the execution.
250

251
    The keys of the dict must not have 'GANETI_' prefixed as this will
252
    be handled in the hooks runner. Also note additional keys will be
253
    added by the hooks runner. If the LU doesn't define any
254
    environment, an empty dict (and not None) should be returned.
255

256
    No nodes should be returned as an empty list (and not None).
257

258
    Note that if the HPATH for a LU class is None, this function will
259
    not be called.
260

261
    """
262
    raise NotImplementedError
263

    
264
  # this is valid in this entire class even if added here
265
  # pylint: disable-msg=R0201
266
  def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
267
    """Notify the LU about the results of its hooks.
268

269
    This method is called every time a hooks phase is executed, and notifies
270
    the Logical Unit about the hooks' result. The LU can then use it to alter
271
    its result based on the hooks.  By default the method does nothing and the
272
    previous result is passed back unchanged but any LU can define it if it
273
    wants to use the local cluster hook-scripts somehow.
274

275
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
276
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
277
    @param hook_results: the results of the multi-node hooks rpc call
278
    @param feedback_fn: function used send feedback back to the caller
279
    @param lu_result: the previous Exec result this LU had, or None
280
        in the PRE phase
281
    @return: the new Exec result, based on the previous result
282
        and hook results
283

284
    """
285
    return lu_result
286

    
287
  def _ExpandAndLockInstance(self):
288
    """Helper function to expand and lock an instance.
289

290
    Many LUs that work on an instance take its name in self.op.instance_name
291
    and need to expand it and then declare the expanded name for locking. This
292
    function does it, and then updates self.op.instance_name to the expanded
293
    name. It also initializes needed_locks as a dict, if this hasn't been done
294
    before.
295

296
    """
297
    if self.needed_locks is None:
298
      self.needed_locks = {}
299
    else:
300
      assert locking.LEVEL_INSTANCE not in self.needed_locks, \
301
        "_ExpandAndLockInstance called with instance-level locks set"
302
    expanded_name = self.cfg.ExpandInstanceName(self.op.instance_name)
303
    if expanded_name is None:
304
      raise errors.OpPrereqError("Instance '%s' not known" %
305
                                 self.op.instance_name, errors.ECODE_NOENT)
306
    self.needed_locks[locking.LEVEL_INSTANCE] = expanded_name
307
    self.op.instance_name = expanded_name
308

    
309
  def _LockInstancesNodes(self, primary_only=False):
310
    """Helper function to declare instances' nodes for locking.
311

312
    This function should be called after locking one or more instances to lock
313
    their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
314
    with all primary or secondary nodes for instances already locked and
315
    present in self.needed_locks[locking.LEVEL_INSTANCE].
316

317
    It should be called from DeclareLocks, and for safety only works if
318
    self.recalculate_locks[locking.LEVEL_NODE] is set.
319

320
    In the future it may grow parameters to just lock some instance's nodes, or
321
    to just lock primaries or secondary nodes, if needed.
322

323
    If should be called in DeclareLocks in a way similar to::
324

325
      if level == locking.LEVEL_NODE:
326
        self._LockInstancesNodes()
327

328
    @type primary_only: boolean
329
    @param primary_only: only lock primary nodes of locked instances
330

331
    """
332
    assert locking.LEVEL_NODE in self.recalculate_locks, \
333
      "_LockInstancesNodes helper function called with no nodes to recalculate"
334

    
335
    # TODO: check if we're really been called with the instance locks held
336

    
337
    # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
338
    # future we might want to have different behaviors depending on the value
339
    # of self.recalculate_locks[locking.LEVEL_NODE]
340
    wanted_nodes = []
341
    for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
342
      instance = self.context.cfg.GetInstanceInfo(instance_name)
343
      wanted_nodes.append(instance.primary_node)
344
      if not primary_only:
345
        wanted_nodes.extend(instance.secondary_nodes)
346

    
347
    if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE:
348
      self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
349
    elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND:
350
      self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
351

    
352
    del self.recalculate_locks[locking.LEVEL_NODE]
353

    
354

    
355
class NoHooksLU(LogicalUnit): # pylint: disable-msg=W0223
356
  """Simple LU which runs no hooks.
357

358
  This LU is intended as a parent for other LogicalUnits which will
359
  run no hooks, in order to reduce duplicate code.
360

361
  """
362
  HPATH = None
363
  HTYPE = None
364

    
365
  def BuildHooksEnv(self):
366
    """Empty BuildHooksEnv for NoHooksLu.
367

368
    This just raises an error.
369

370
    """
371
    assert False, "BuildHooksEnv called for NoHooksLUs"
372

    
373

    
374
class Tasklet:
375
  """Tasklet base class.
376

377
  Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
378
  they can mix legacy code with tasklets. Locking needs to be done in the LU,
379
  tasklets know nothing about locks.
380

381
  Subclasses must follow these rules:
382
    - Implement CheckPrereq
383
    - Implement Exec
384

385
  """
386
  def __init__(self, lu):
387
    self.lu = lu
388

    
389
    # Shortcuts
390
    self.cfg = lu.cfg
391
    self.rpc = lu.rpc
392

    
393
  def CheckPrereq(self):
394
    """Check prerequisites for this tasklets.
395

396
    This method should check whether the prerequisites for the execution of
397
    this tasklet are fulfilled. It can do internode communication, but it
398
    should be idempotent - no cluster or system changes are allowed.
399

400
    The method should raise errors.OpPrereqError in case something is not
401
    fulfilled. Its return value is ignored.
402

403
    This method should also update all parameters to their canonical form if it
404
    hasn't been done before.
405

406
    """
407
    raise NotImplementedError
408

    
409
  def Exec(self, feedback_fn):
410
    """Execute the tasklet.
411

412
    This method should implement the actual work. It should raise
413
    errors.OpExecError for failures that are somewhat dealt with in code, or
414
    expected.
415

416
    """
417
    raise NotImplementedError
418

    
419

    
420
def _GetWantedNodes(lu, nodes):
421
  """Returns list of checked and expanded node names.
422

423
  @type lu: L{LogicalUnit}
424
  @param lu: the logical unit on whose behalf we execute
425
  @type nodes: list
426
  @param nodes: list of node names or None for all nodes
427
  @rtype: list
428
  @return: the list of nodes, sorted
429
  @raise errors.OpProgrammerError: if the nodes parameter is wrong type
430

431
  """
432
  if not isinstance(nodes, list):
433
    raise errors.OpPrereqError("Invalid argument type 'nodes'",
434
                               errors.ECODE_INVAL)
435

    
436
  if not nodes:
437
    raise errors.ProgrammerError("_GetWantedNodes should only be called with a"
438
      " non-empty list of nodes whose name is to be expanded.")
439

    
440
  wanted = []
441
  for name in nodes:
442
    node = lu.cfg.ExpandNodeName(name)
443
    if node is None:
444
      raise errors.OpPrereqError("No such node name '%s'" % name,
445
                                 errors.ECODE_NOENT)
446
    wanted.append(node)
447

    
448
  return utils.NiceSort(wanted)
449

    
450

    
451
def _GetWantedInstances(lu, instances):
452
  """Returns list of checked and expanded instance names.
453

454
  @type lu: L{LogicalUnit}
455
  @param lu: the logical unit on whose behalf we execute
456
  @type instances: list
457
  @param instances: list of instance names or None for all instances
458
  @rtype: list
459
  @return: the list of instances, sorted
460
  @raise errors.OpPrereqError: if the instances parameter is wrong type
461
  @raise errors.OpPrereqError: if any of the passed instances is not found
462

463
  """
464
  if not isinstance(instances, list):
465
    raise errors.OpPrereqError("Invalid argument type 'instances'",
466
                               errors.ECODE_INVAL)
467

    
468
  if instances:
469
    wanted = []
470

    
471
    for name in instances:
472
      instance = lu.cfg.ExpandInstanceName(name)
473
      if instance is None:
474
        raise errors.OpPrereqError("No such instance name '%s'" % name,
475
                                   errors.ECODE_NOENT)
476
      wanted.append(instance)
477

    
478
  else:
479
    wanted = utils.NiceSort(lu.cfg.GetInstanceList())
480
  return wanted
481

    
482

    
483
def _CheckOutputFields(static, dynamic, selected):
484
  """Checks whether all selected fields are valid.
485

486
  @type static: L{utils.FieldSet}
487
  @param static: static fields set
488
  @type dynamic: L{utils.FieldSet}
489
  @param dynamic: dynamic fields set
490

491
  """
492
  f = utils.FieldSet()
493
  f.Extend(static)
494
  f.Extend(dynamic)
495

    
496
  delta = f.NonMatching(selected)
497
  if delta:
498
    raise errors.OpPrereqError("Unknown output fields selected: %s"
499
                               % ",".join(delta), errors.ECODE_INVAL)
500

    
501

    
502
def _CheckBooleanOpField(op, name):
503
  """Validates boolean opcode parameters.
504

505
  This will ensure that an opcode parameter is either a boolean value,
506
  or None (but that it always exists).
507

508
  """
509
  val = getattr(op, name, None)
510
  if not (val is None or isinstance(val, bool)):
511
    raise errors.OpPrereqError("Invalid boolean parameter '%s' (%s)" %
512
                               (name, str(val)), errors.ECODE_INVAL)
513
  setattr(op, name, val)
514

    
515

    
516
def _CheckGlobalHvParams(params):
517
  """Validates that given hypervisor params are not global ones.
518

519
  This will ensure that instances don't get customised versions of
520
  global params.
521

522
  """
523
  used_globals = constants.HVC_GLOBALS.intersection(params)
524
  if used_globals:
525
    msg = ("The following hypervisor parameters are global and cannot"
526
           " be customized at instance level, please modify them at"
527
           " cluster level: %s" % utils.CommaJoin(used_globals))
528
    raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
529

    
530

    
531
def _CheckNodeOnline(lu, node):
532
  """Ensure that a given node is online.
533

534
  @param lu: the LU on behalf of which we make the check
535
  @param node: the node to check
536
  @raise errors.OpPrereqError: if the node is offline
537

538
  """
539
  if lu.cfg.GetNodeInfo(node).offline:
540
    raise errors.OpPrereqError("Can't use offline node %s" % node,
541
                               errors.ECODE_INVAL)
542

    
543

    
544
def _CheckNodeNotDrained(lu, node):
545
  """Ensure that a given node is not drained.
546

547
  @param lu: the LU on behalf of which we make the check
548
  @param node: the node to check
549
  @raise errors.OpPrereqError: if the node is drained
550

551
  """
552
  if lu.cfg.GetNodeInfo(node).drained:
553
    raise errors.OpPrereqError("Can't use drained node %s" % node,
554
                               errors.ECODE_INVAL)
555

    
556

    
557
def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
558
                          memory, vcpus, nics, disk_template, disks,
559
                          bep, hvp, hypervisor_name):
560
  """Builds instance related env variables for hooks
561

562
  This builds the hook environment from individual variables.
563

564
  @type name: string
565
  @param name: the name of the instance
566
  @type primary_node: string
567
  @param primary_node: the name of the instance's primary node
568
  @type secondary_nodes: list
569
  @param secondary_nodes: list of secondary nodes as strings
570
  @type os_type: string
571
  @param os_type: the name of the instance's OS
572
  @type status: boolean
573
  @param status: the should_run status of the instance
574
  @type memory: string
575
  @param memory: the memory size of the instance
576
  @type vcpus: string
577
  @param vcpus: the count of VCPUs the instance has
578
  @type nics: list
579
  @param nics: list of tuples (ip, mac, mode, link) representing
580
      the NICs the instance has
581
  @type disk_template: string
582
  @param disk_template: the disk template of the instance
583
  @type disks: list
584
  @param disks: the list of (size, mode) pairs
585
  @type bep: dict
586
  @param bep: the backend parameters for the instance
587
  @type hvp: dict
588
  @param hvp: the hypervisor parameters for the instance
589
  @type hypervisor_name: string
590
  @param hypervisor_name: the hypervisor for the instance
591
  @rtype: dict
592
  @return: the hook environment for this instance
593

594
  """
595
  if status:
596
    str_status = "up"
597
  else:
598
    str_status = "down"
599
  env = {
600
    "OP_TARGET": name,
601
    "INSTANCE_NAME": name,
602
    "INSTANCE_PRIMARY": primary_node,
603
    "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
604
    "INSTANCE_OS_TYPE": os_type,
605
    "INSTANCE_STATUS": str_status,
606
    "INSTANCE_MEMORY": memory,
607
    "INSTANCE_VCPUS": vcpus,
608
    "INSTANCE_DISK_TEMPLATE": disk_template,
609
    "INSTANCE_HYPERVISOR": hypervisor_name,
610
  }
611

    
612
  if nics:
613
    nic_count = len(nics)
614
    for idx, (ip, mac, mode, link) in enumerate(nics):
615
      if ip is None:
616
        ip = ""
617
      env["INSTANCE_NIC%d_IP" % idx] = ip
618
      env["INSTANCE_NIC%d_MAC" % idx] = mac
619
      env["INSTANCE_NIC%d_MODE" % idx] = mode
620
      env["INSTANCE_NIC%d_LINK" % idx] = link
621
      if mode == constants.NIC_MODE_BRIDGED:
622
        env["INSTANCE_NIC%d_BRIDGE" % idx] = link
623
  else:
624
    nic_count = 0
625

    
626
  env["INSTANCE_NIC_COUNT"] = nic_count
627

    
628
  if disks:
629
    disk_count = len(disks)
630
    for idx, (size, mode) in enumerate(disks):
631
      env["INSTANCE_DISK%d_SIZE" % idx] = size
632
      env["INSTANCE_DISK%d_MODE" % idx] = mode
633
  else:
634
    disk_count = 0
635

    
636
  env["INSTANCE_DISK_COUNT"] = disk_count
637

    
638
  for source, kind in [(bep, "BE"), (hvp, "HV")]:
639
    for key, value in source.items():
640
      env["INSTANCE_%s_%s" % (kind, key)] = value
641

    
642
  return env
643

    
644

    
645
def _NICListToTuple(lu, nics):
646
  """Build a list of nic information tuples.
647

648
  This list is suitable to be passed to _BuildInstanceHookEnv or as a return
649
  value in LUQueryInstanceData.
650

651
  @type lu:  L{LogicalUnit}
652
  @param lu: the logical unit on whose behalf we execute
653
  @type nics: list of L{objects.NIC}
654
  @param nics: list of nics to convert to hooks tuples
655

656
  """
657
  hooks_nics = []
658
  c_nicparams = lu.cfg.GetClusterInfo().nicparams[constants.PP_DEFAULT]
659
  for nic in nics:
660
    ip = nic.ip
661
    mac = nic.mac
662
    filled_params = objects.FillDict(c_nicparams, nic.nicparams)
663
    mode = filled_params[constants.NIC_MODE]
664
    link = filled_params[constants.NIC_LINK]
665
    hooks_nics.append((ip, mac, mode, link))
666
  return hooks_nics
667

    
668

    
669
def _BuildInstanceHookEnvByObject(lu, instance, override=None):
670
  """Builds instance related env variables for hooks from an object.
671

672
  @type lu: L{LogicalUnit}
673
  @param lu: the logical unit on whose behalf we execute
674
  @type instance: L{objects.Instance}
675
  @param instance: the instance for which we should build the
676
      environment
677
  @type override: dict
678
  @param override: dictionary with key/values that will override
679
      our values
680
  @rtype: dict
681
  @return: the hook environment dictionary
682

683
  """
684
  cluster = lu.cfg.GetClusterInfo()
685
  bep = cluster.FillBE(instance)
686
  hvp = cluster.FillHV(instance)
687
  args = {
688
    'name': instance.name,
689
    'primary_node': instance.primary_node,
690
    'secondary_nodes': instance.secondary_nodes,
691
    'os_type': instance.os,
692
    'status': instance.admin_up,
693
    'memory': bep[constants.BE_MEMORY],
694
    'vcpus': bep[constants.BE_VCPUS],
695
    'nics': _NICListToTuple(lu, instance.nics),
696
    'disk_template': instance.disk_template,
697
    'disks': [(disk.size, disk.mode) for disk in instance.disks],
698
    'bep': bep,
699
    'hvp': hvp,
700
    'hypervisor_name': instance.hypervisor,
701
  }
702
  if override:
703
    args.update(override)
704
  return _BuildInstanceHookEnv(**args) # pylint: disable-msg=W0142
705

    
706

    
707
def _AdjustCandidatePool(lu, exceptions):
708
  """Adjust the candidate pool after node operations.
709

710
  """
711
  mod_list = lu.cfg.MaintainCandidatePool(exceptions)
712
  if mod_list:
713
    lu.LogInfo("Promoted nodes to master candidate role: %s",
714
               utils.CommaJoin(node.name for node in mod_list))
715
    for name in mod_list:
716
      lu.context.ReaddNode(name)
717
  mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
718
  if mc_now > mc_max:
719
    lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
720
               (mc_now, mc_max))
721

    
722

    
723
def _DecideSelfPromotion(lu, exceptions=None):
724
  """Decide whether I should promote myself as a master candidate.
725

726
  """
727
  cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
728
  mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
729
  # the new node will increase mc_max with one, so:
730
  mc_should = min(mc_should + 1, cp_size)
731
  return mc_now < mc_should
732

    
733

    
734
def _CheckNicsBridgesExist(lu, target_nics, target_node,
735
                               profile=constants.PP_DEFAULT):
736
  """Check that the brigdes needed by a list of nics exist.
737

738
  """
739
  c_nicparams = lu.cfg.GetClusterInfo().nicparams[profile]
740
  paramslist = [objects.FillDict(c_nicparams, nic.nicparams)
741
                for nic in target_nics]
742
  brlist = [params[constants.NIC_LINK] for params in paramslist
743
            if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
744
  if brlist:
745
    result = lu.rpc.call_bridges_exist(target_node, brlist)
746
    result.Raise("Error checking bridges on destination node '%s'" %
747
                 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
748

    
749

    
750
def _CheckInstanceBridgesExist(lu, instance, node=None):
751
  """Check that the brigdes needed by an instance exist.
752

753
  """
754
  if node is None:
755
    node = instance.primary_node
756
  _CheckNicsBridgesExist(lu, instance.nics, node)
757

    
758

    
759
def _CheckOSVariant(os_obj, name):
760
  """Check whether an OS name conforms to the os variants specification.
761

762
  @type os_obj: L{objects.OS}
763
  @param os_obj: OS object to check
764
  @type name: string
765
  @param name: OS name passed by the user, to check for validity
766

767
  """
768
  if not os_obj.supported_variants:
769
    return
770
  try:
771
    variant = name.split("+", 1)[1]
772
  except IndexError:
773
    raise errors.OpPrereqError("OS name must include a variant",
774
                               errors.ECODE_INVAL)
775

    
776
  if variant not in os_obj.supported_variants:
777
    raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
778

    
779

    
780
def _GetNodeInstancesInner(cfg, fn):
781
  return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
782

    
783

    
784
def _GetNodeInstances(cfg, node_name):
785
  """Returns a list of all primary and secondary instances on a node.
786

787
  """
788

    
789
  return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
790

    
791

    
792
def _GetNodePrimaryInstances(cfg, node_name):
793
  """Returns primary instances on a node.
794

795
  """
796
  return _GetNodeInstancesInner(cfg,
797
                                lambda inst: node_name == inst.primary_node)
798

    
799

    
800
def _GetNodeSecondaryInstances(cfg, node_name):
801
  """Returns secondary instances on a node.
802

803
  """
804
  return _GetNodeInstancesInner(cfg,
805
                                lambda inst: node_name in inst.secondary_nodes)
806

    
807

    
808
def _GetStorageTypeArgs(cfg, storage_type):
809
  """Returns the arguments for a storage type.
810

811
  """
812
  # Special case for file storage
813
  if storage_type == constants.ST_FILE:
814
    # storage.FileStorage wants a list of storage directories
815
    return [[cfg.GetFileStorageDir()]]
816

    
817
  return []
818

    
819

    
820
def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq):
821
  faulty = []
822

    
823
  for dev in instance.disks:
824
    cfg.SetDiskID(dev, node_name)
825

    
826
  result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks)
827
  result.Raise("Failed to get disk status from node %s" % node_name,
828
               prereq=prereq, ecode=errors.ECODE_ENVIRON)
829

    
830
  for idx, bdev_status in enumerate(result.payload):
831
    if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
832
      faulty.append(idx)
833

    
834
  return faulty
835

    
836

    
837
class LUPostInitCluster(LogicalUnit):
838
  """Logical unit for running hooks after cluster initialization.
839

840
  """
841
  HPATH = "cluster-init"
842
  HTYPE = constants.HTYPE_CLUSTER
843
  _OP_REQP = []
844

    
845
  def BuildHooksEnv(self):
846
    """Build hooks env.
847

848
    """
849
    env = {"OP_TARGET": self.cfg.GetClusterName()}
850
    mn = self.cfg.GetMasterNode()
851
    return env, [], [mn]
852

    
853
  def CheckPrereq(self):
854
    """No prerequisites to check.
855

856
    """
857
    return True
858

    
859
  def Exec(self, feedback_fn):
860
    """Nothing to do.
861

862
    """
863
    return True
864

    
865

    
866
class LUDestroyCluster(LogicalUnit):
867
  """Logical unit for destroying the cluster.
868

869
  """
870
  HPATH = "cluster-destroy"
871
  HTYPE = constants.HTYPE_CLUSTER
872
  _OP_REQP = []
873

    
874
  def BuildHooksEnv(self):
875
    """Build hooks env.
876

877
    """
878
    env = {"OP_TARGET": self.cfg.GetClusterName()}
879
    return env, [], []
880

    
881
  def CheckPrereq(self):
882
    """Check prerequisites.
883

884
    This checks whether the cluster is empty.
885

886
    Any errors are signaled by raising errors.OpPrereqError.
887

888
    """
889
    master = self.cfg.GetMasterNode()
890

    
891
    nodelist = self.cfg.GetNodeList()
892
    if len(nodelist) != 1 or nodelist[0] != master:
893
      raise errors.OpPrereqError("There are still %d node(s) in"
894
                                 " this cluster." % (len(nodelist) - 1),
895
                                 errors.ECODE_INVAL)
896
    instancelist = self.cfg.GetInstanceList()
897
    if instancelist:
898
      raise errors.OpPrereqError("There are still %d instance(s) in"
899
                                 " this cluster." % len(instancelist),
900
                                 errors.ECODE_INVAL)
901

    
902
  def Exec(self, feedback_fn):
903
    """Destroys the cluster.
904

905
    """
906
    master = self.cfg.GetMasterNode()
907
    modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
908

    
909
    # Run post hooks on master node before it's removed
910
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
911
    try:
912
      hm.RunPhase(constants.HOOKS_PHASE_POST, [master])
913
    except:
914
      # pylint: disable-msg=W0702
915
      self.LogWarning("Errors occurred running hooks on %s" % master)
916

    
917
    result = self.rpc.call_node_stop_master(master, False)
918
    result.Raise("Could not disable the master role")
919

    
920
    if modify_ssh_setup:
921
      priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
922
      utils.CreateBackup(priv_key)
923
      utils.CreateBackup(pub_key)
924

    
925
    return master
926

    
927

    
928
class LUVerifyCluster(LogicalUnit):
929
  """Verifies the cluster status.
930

931
  """
932
  HPATH = "cluster-verify"
933
  HTYPE = constants.HTYPE_CLUSTER
934
  _OP_REQP = ["skip_checks", "verbose", "error_codes", "debug_simulate_errors"]
935
  REQ_BGL = False
936

    
937
  TCLUSTER = "cluster"
938
  TNODE = "node"
939
  TINSTANCE = "instance"
940

    
941
  ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
942
  EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
943
  EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
944
  EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
945
  EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
946
  EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
947
  EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE")
948
  ENODEDRBD = (TNODE, "ENODEDRBD")
949
  ENODEFILECHECK = (TNODE, "ENODEFILECHECK")
950
  ENODEHOOKS = (TNODE, "ENODEHOOKS")
951
  ENODEHV = (TNODE, "ENODEHV")
952
  ENODELVM = (TNODE, "ENODELVM")
953
  ENODEN1 = (TNODE, "ENODEN1")
954
  ENODENET = (TNODE, "ENODENET")
955
  ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE")
956
  ENODEORPHANLV = (TNODE, "ENODEORPHANLV")
957
  ENODERPC = (TNODE, "ENODERPC")
958
  ENODESSH = (TNODE, "ENODESSH")
959
  ENODEVERSION = (TNODE, "ENODEVERSION")
960
  ENODESETUP = (TNODE, "ENODESETUP")
961
  ENODETIME = (TNODE, "ENODETIME")
962

    
963
  ETYPE_FIELD = "code"
964
  ETYPE_ERROR = "ERROR"
965
  ETYPE_WARNING = "WARNING"
966

    
967
  def ExpandNames(self):
968
    self.needed_locks = {
969
      locking.LEVEL_NODE: locking.ALL_SET,
970
      locking.LEVEL_INSTANCE: locking.ALL_SET,
971
    }
972
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
973

    
974
  def _Error(self, ecode, item, msg, *args, **kwargs):
975
    """Format an error message.
976

977
    Based on the opcode's error_codes parameter, either format a
978
    parseable error code, or a simpler error string.
979

980
    This must be called only from Exec and functions called from Exec.
981

982
    """
983
    ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
984
    itype, etxt = ecode
985
    # first complete the msg
986
    if args:
987
      msg = msg % args
988
    # then format the whole message
989
    if self.op.error_codes:
990
      msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
991
    else:
992
      if item:
993
        item = " " + item
994
      else:
995
        item = ""
996
      msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
997
    # and finally report it via the feedback_fn
998
    self._feedback_fn("  - %s" % msg)
999

    
1000
  def _ErrorIf(self, cond, *args, **kwargs):
1001
    """Log an error message if the passed condition is True.
1002

1003
    """
1004
    cond = bool(cond) or self.op.debug_simulate_errors
1005
    if cond:
1006
      self._Error(*args, **kwargs)
1007
    # do not mark the operation as failed for WARN cases only
1008
    if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
1009
      self.bad = self.bad or cond
1010

    
1011
  def _VerifyNode(self, nodeinfo, file_list, local_cksum,
1012
                  node_result, master_files, drbd_map, vg_name):
1013
    """Run multiple tests against a node.
1014

1015
    Test list:
1016

1017
      - compares ganeti version
1018
      - checks vg existence and size > 20G
1019
      - checks config file checksum
1020
      - checks ssh to other nodes
1021

1022
    @type nodeinfo: L{objects.Node}
1023
    @param nodeinfo: the node to check
1024
    @param file_list: required list of files
1025
    @param local_cksum: dictionary of local files and their checksums
1026
    @param node_result: the results from the node
1027
    @param master_files: list of files that only masters should have
1028
    @param drbd_map: the useddrbd minors for this node, in
1029
        form of minor: (instance, must_exist) which correspond to instances
1030
        and their running status
1031
    @param vg_name: Ganeti Volume Group (result of self.cfg.GetVGName())
1032

1033
    """
1034
    node = nodeinfo.name
1035
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1036

    
1037
    # main result, node_result should be a non-empty dict
1038
    test = not node_result or not isinstance(node_result, dict)
1039
    _ErrorIf(test, self.ENODERPC, node,
1040
                  "unable to verify node: no data returned")
1041
    if test:
1042
      return
1043

    
1044
    # compares ganeti version
1045
    local_version = constants.PROTOCOL_VERSION
1046
    remote_version = node_result.get('version', None)
1047
    test = not (remote_version and
1048
                isinstance(remote_version, (list, tuple)) and
1049
                len(remote_version) == 2)
1050
    _ErrorIf(test, self.ENODERPC, node,
1051
             "connection to node returned invalid data")
1052
    if test:
1053
      return
1054

    
1055
    test = local_version != remote_version[0]
1056
    _ErrorIf(test, self.ENODEVERSION, node,
1057
             "incompatible protocol versions: master %s,"
1058
             " node %s", local_version, remote_version[0])
1059
    if test:
1060
      return
1061

    
1062
    # node seems compatible, we can actually try to look into its results
1063

    
1064
    # full package version
1065
    self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
1066
                  self.ENODEVERSION, node,
1067
                  "software version mismatch: master %s, node %s",
1068
                  constants.RELEASE_VERSION, remote_version[1],
1069
                  code=self.ETYPE_WARNING)
1070

    
1071
    # checks vg existence and size > 20G
1072
    if vg_name is not None:
1073
      vglist = node_result.get(constants.NV_VGLIST, None)
1074
      test = not vglist
1075
      _ErrorIf(test, self.ENODELVM, node, "unable to check volume groups")
1076
      if not test:
1077
        vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
1078
                                              constants.MIN_VG_SIZE)
1079
        _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
1080

    
1081
    # checks config file checksum
1082

    
1083
    remote_cksum = node_result.get(constants.NV_FILELIST, None)
1084
    test = not isinstance(remote_cksum, dict)
1085
    _ErrorIf(test, self.ENODEFILECHECK, node,
1086
             "node hasn't returned file checksum data")
1087
    if not test:
1088
      for file_name in file_list:
1089
        node_is_mc = nodeinfo.master_candidate
1090
        must_have = (file_name not in master_files) or node_is_mc
1091
        # missing
1092
        test1 = file_name not in remote_cksum
1093
        # invalid checksum
1094
        test2 = not test1 and remote_cksum[file_name] != local_cksum[file_name]
1095
        # existing and good
1096
        test3 = not test1 and remote_cksum[file_name] == local_cksum[file_name]
1097
        _ErrorIf(test1 and must_have, self.ENODEFILECHECK, node,
1098
                 "file '%s' missing", file_name)
1099
        _ErrorIf(test2 and must_have, self.ENODEFILECHECK, node,
1100
                 "file '%s' has wrong checksum", file_name)
1101
        # not candidate and this is not a must-have file
1102
        _ErrorIf(test2 and not must_have, self.ENODEFILECHECK, node,
1103
                 "file '%s' should not exist on non master"
1104
                 " candidates (and the file is outdated)", file_name)
1105
        # all good, except non-master/non-must have combination
1106
        _ErrorIf(test3 and not must_have, self.ENODEFILECHECK, node,
1107
                 "file '%s' should not exist"
1108
                 " on non master candidates", file_name)
1109

    
1110
    # checks ssh to any
1111

    
1112
    test = constants.NV_NODELIST not in node_result
1113
    _ErrorIf(test, self.ENODESSH, node,
1114
             "node hasn't returned node ssh connectivity data")
1115
    if not test:
1116
      if node_result[constants.NV_NODELIST]:
1117
        for a_node, a_msg in node_result[constants.NV_NODELIST].items():
1118
          _ErrorIf(True, self.ENODESSH, node,
1119
                   "ssh communication with node '%s': %s", a_node, a_msg)
1120

    
1121
    test = constants.NV_NODENETTEST not in node_result
1122
    _ErrorIf(test, self.ENODENET, node,
1123
             "node hasn't returned node tcp connectivity data")
1124
    if not test:
1125
      if node_result[constants.NV_NODENETTEST]:
1126
        nlist = utils.NiceSort(node_result[constants.NV_NODENETTEST].keys())
1127
        for anode in nlist:
1128
          _ErrorIf(True, self.ENODENET, node,
1129
                   "tcp communication with node '%s': %s",
1130
                   anode, node_result[constants.NV_NODENETTEST][anode])
1131

    
1132
    hyp_result = node_result.get(constants.NV_HYPERVISOR, None)
1133
    if isinstance(hyp_result, dict):
1134
      for hv_name, hv_result in hyp_result.iteritems():
1135
        test = hv_result is not None
1136
        _ErrorIf(test, self.ENODEHV, node,
1137
                 "hypervisor %s verify failure: '%s'", hv_name, hv_result)
1138

    
1139
    # check used drbd list
1140
    if vg_name is not None:
1141
      used_minors = node_result.get(constants.NV_DRBDLIST, [])
1142
      test = not isinstance(used_minors, (tuple, list))
1143
      _ErrorIf(test, self.ENODEDRBD, node,
1144
               "cannot parse drbd status file: %s", str(used_minors))
1145
      if not test:
1146
        for minor, (iname, must_exist) in drbd_map.items():
1147
          test = minor not in used_minors and must_exist
1148
          _ErrorIf(test, self.ENODEDRBD, node,
1149
                   "drbd minor %d of instance %s is not active",
1150
                   minor, iname)
1151
        for minor in used_minors:
1152
          test = minor not in drbd_map
1153
          _ErrorIf(test, self.ENODEDRBD, node,
1154
                   "unallocated drbd minor %d is in use", minor)
1155
    test = node_result.get(constants.NV_NODESETUP,
1156
                           ["Missing NODESETUP results"])
1157
    _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s",
1158
             "; ".join(test))
1159

    
1160
    # check pv names
1161
    if vg_name is not None:
1162
      pvlist = node_result.get(constants.NV_PVLIST, None)
1163
      test = pvlist is None
1164
      _ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node")
1165
      if not test:
1166
        # check that ':' is not present in PV names, since it's a
1167
        # special character for lvcreate (denotes the range of PEs to
1168
        # use on the PV)
1169
        for _, pvname, owner_vg in pvlist:
1170
          test = ":" in pvname
1171
          _ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV"
1172
                   " '%s' of VG '%s'", pvname, owner_vg)
1173

    
1174
  def _VerifyInstance(self, instance, instanceconfig, node_vol_is,
1175
                      node_instance, n_offline):
1176
    """Verify an instance.
1177

1178
    This function checks to see if the required block devices are
1179
    available on the instance's node.
1180

1181
    """
1182
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1183
    node_current = instanceconfig.primary_node
1184

    
1185
    node_vol_should = {}
1186
    instanceconfig.MapLVsByNode(node_vol_should)
1187

    
1188
    for node in node_vol_should:
1189
      if node in n_offline:
1190
        # ignore missing volumes on offline nodes
1191
        continue
1192
      for volume in node_vol_should[node]:
1193
        test = node not in node_vol_is or volume not in node_vol_is[node]
1194
        _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
1195
                 "volume %s missing on node %s", volume, node)
1196

    
1197
    if instanceconfig.admin_up:
1198
      test = ((node_current not in node_instance or
1199
               not instance in node_instance[node_current]) and
1200
              node_current not in n_offline)
1201
      _ErrorIf(test, self.EINSTANCEDOWN, instance,
1202
               "instance not running on its primary node %s",
1203
               node_current)
1204

    
1205
    for node in node_instance:
1206
      if (not node == node_current):
1207
        test = instance in node_instance[node]
1208
        _ErrorIf(test, self.EINSTANCEWRONGNODE, instance,
1209
                 "instance should not run on node %s", node)
1210

    
1211
  def _VerifyOrphanVolumes(self, node_vol_should, node_vol_is):
1212
    """Verify if there are any unknown volumes in the cluster.
1213

1214
    The .os, .swap and backup volumes are ignored. All other volumes are
1215
    reported as unknown.
1216

1217
    """
1218
    for node in node_vol_is:
1219
      for volume in node_vol_is[node]:
1220
        test = (node not in node_vol_should or
1221
                volume not in node_vol_should[node])
1222
        self._ErrorIf(test, self.ENODEORPHANLV, node,
1223
                      "volume %s is unknown", volume)
1224

    
1225
  def _VerifyOrphanInstances(self, instancelist, node_instance):
1226
    """Verify the list of running instances.
1227

1228
    This checks what instances are running but unknown to the cluster.
1229

1230
    """
1231
    for node in node_instance:
1232
      for o_inst in node_instance[node]:
1233
        test = o_inst not in instancelist
1234
        self._ErrorIf(test, self.ENODEORPHANINSTANCE, node,
1235
                      "instance %s on node %s should not exist", o_inst, node)
1236

    
1237
  def _VerifyNPlusOneMemory(self, node_info, instance_cfg):
1238
    """Verify N+1 Memory Resilience.
1239

1240
    Check that if one single node dies we can still start all the instances it
1241
    was primary for.
1242

1243
    """
1244
    for node, nodeinfo in node_info.iteritems():
1245
      # This code checks that every node which is now listed as secondary has
1246
      # enough memory to host all instances it is supposed to should a single
1247
      # other node in the cluster fail.
1248
      # FIXME: not ready for failover to an arbitrary node
1249
      # FIXME: does not support file-backed instances
1250
      # WARNING: we currently take into account down instances as well as up
1251
      # ones, considering that even if they're down someone might want to start
1252
      # them even in the event of a node failure.
1253
      for prinode, instances in nodeinfo['sinst-by-pnode'].iteritems():
1254
        needed_mem = 0
1255
        for instance in instances:
1256
          bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
1257
          if bep[constants.BE_AUTO_BALANCE]:
1258
            needed_mem += bep[constants.BE_MEMORY]
1259
        test = nodeinfo['mfree'] < needed_mem
1260
        self._ErrorIf(test, self.ENODEN1, node,
1261
                      "not enough memory on to accommodate"
1262
                      " failovers should peer node %s fail", prinode)
1263

    
1264
  def CheckPrereq(self):
1265
    """Check prerequisites.
1266

1267
    Transform the list of checks we're going to skip into a set and check that
1268
    all its members are valid.
1269

1270
    """
1271
    self.skip_set = frozenset(self.op.skip_checks)
1272
    if not constants.VERIFY_OPTIONAL_CHECKS.issuperset(self.skip_set):
1273
      raise errors.OpPrereqError("Invalid checks to be skipped specified",
1274
                                 errors.ECODE_INVAL)
1275

    
1276
  def BuildHooksEnv(self):
1277
    """Build hooks env.
1278

1279
    Cluster-Verify hooks just ran in the post phase and their failure makes
1280
    the output be logged in the verify output and the verification to fail.
1281

1282
    """
1283
    all_nodes = self.cfg.GetNodeList()
1284
    env = {
1285
      "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
1286
      }
1287
    for node in self.cfg.GetAllNodesInfo().values():
1288
      env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
1289

    
1290
    return env, [], all_nodes
1291

    
1292
  def Exec(self, feedback_fn):
1293
    """Verify integrity of cluster, performing various test on nodes.
1294

1295
    """
1296
    self.bad = False
1297
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1298
    verbose = self.op.verbose
1299
    self._feedback_fn = feedback_fn
1300
    feedback_fn("* Verifying global settings")
1301
    for msg in self.cfg.VerifyConfig():
1302
      _ErrorIf(True, self.ECLUSTERCFG, None, msg)
1303

    
1304
    vg_name = self.cfg.GetVGName()
1305
    hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
1306
    nodelist = utils.NiceSort(self.cfg.GetNodeList())
1307
    nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
1308
    instancelist = utils.NiceSort(self.cfg.GetInstanceList())
1309
    instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
1310
                        for iname in instancelist)
1311
    i_non_redundant = [] # Non redundant instances
1312
    i_non_a_balanced = [] # Non auto-balanced instances
1313
    n_offline = [] # List of offline nodes
1314
    n_drained = [] # List of nodes being drained
1315
    node_volume = {}
1316
    node_instance = {}
1317
    node_info = {}
1318
    instance_cfg = {}
1319

    
1320
    # FIXME: verify OS list
1321
    # do local checksums
1322
    master_files = [constants.CLUSTER_CONF_FILE]
1323

    
1324
    file_names = ssconf.SimpleStore().GetFileList()
1325
    file_names.append(constants.SSL_CERT_FILE)
1326
    file_names.append(constants.RAPI_CERT_FILE)
1327
    file_names.extend(master_files)
1328

    
1329
    local_checksums = utils.FingerprintFiles(file_names)
1330

    
1331
    feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
1332
    node_verify_param = {
1333
      constants.NV_FILELIST: file_names,
1334
      constants.NV_NODELIST: [node.name for node in nodeinfo
1335
                              if not node.offline],
1336
      constants.NV_HYPERVISOR: hypervisors,
1337
      constants.NV_NODENETTEST: [(node.name, node.primary_ip,
1338
                                  node.secondary_ip) for node in nodeinfo
1339
                                 if not node.offline],
1340
      constants.NV_INSTANCELIST: hypervisors,
1341
      constants.NV_VERSION: None,
1342
      constants.NV_HVINFO: self.cfg.GetHypervisorType(),
1343
      constants.NV_NODESETUP: None,
1344
      constants.NV_TIME: None,
1345
      }
1346

    
1347
    if vg_name is not None:
1348
      node_verify_param[constants.NV_VGLIST] = None
1349
      node_verify_param[constants.NV_LVLIST] = vg_name
1350
      node_verify_param[constants.NV_PVLIST] = [vg_name]
1351
      node_verify_param[constants.NV_DRBDLIST] = None
1352

    
1353
    # Due to the way our RPC system works, exact response times cannot be
1354
    # guaranteed (e.g. a broken node could run into a timeout). By keeping the
1355
    # time before and after executing the request, we can at least have a time
1356
    # window.
1357
    nvinfo_starttime = time.time()
1358
    all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
1359
                                           self.cfg.GetClusterName())
1360
    nvinfo_endtime = time.time()
1361

    
1362
    cluster = self.cfg.GetClusterInfo()
1363
    master_node = self.cfg.GetMasterNode()
1364
    all_drbd_map = self.cfg.ComputeDRBDMap()
1365

    
1366
    feedback_fn("* Verifying node status")
1367
    for node_i in nodeinfo:
1368
      node = node_i.name
1369

    
1370
      if node_i.offline:
1371
        if verbose:
1372
          feedback_fn("* Skipping offline node %s" % (node,))
1373
        n_offline.append(node)
1374
        continue
1375

    
1376
      if node == master_node:
1377
        ntype = "master"
1378
      elif node_i.master_candidate:
1379
        ntype = "master candidate"
1380
      elif node_i.drained:
1381
        ntype = "drained"
1382
        n_drained.append(node)
1383
      else:
1384
        ntype = "regular"
1385
      if verbose:
1386
        feedback_fn("* Verifying node %s (%s)" % (node, ntype))
1387

    
1388
      msg = all_nvinfo[node].fail_msg
1389
      _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
1390
      if msg:
1391
        continue
1392

    
1393
      nresult = all_nvinfo[node].payload
1394
      node_drbd = {}
1395
      for minor, instance in all_drbd_map[node].items():
1396
        test = instance not in instanceinfo
1397
        _ErrorIf(test, self.ECLUSTERCFG, None,
1398
                 "ghost instance '%s' in temporary DRBD map", instance)
1399
          # ghost instance should not be running, but otherwise we
1400
          # don't give double warnings (both ghost instance and
1401
          # unallocated minor in use)
1402
        if test:
1403
          node_drbd[minor] = (instance, False)
1404
        else:
1405
          instance = instanceinfo[instance]
1406
          node_drbd[minor] = (instance.name, instance.admin_up)
1407

    
1408
      self._VerifyNode(node_i, file_names, local_checksums,
1409
                       nresult, master_files, node_drbd, vg_name)
1410

    
1411
      lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
1412
      if vg_name is None:
1413
        node_volume[node] = {}
1414
      elif isinstance(lvdata, basestring):
1415
        _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
1416
                 utils.SafeEncode(lvdata))
1417
        node_volume[node] = {}
1418
      elif not isinstance(lvdata, dict):
1419
        _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
1420
        continue
1421
      else:
1422
        node_volume[node] = lvdata
1423

    
1424
      # node_instance
1425
      idata = nresult.get(constants.NV_INSTANCELIST, None)
1426
      test = not isinstance(idata, list)
1427
      _ErrorIf(test, self.ENODEHV, node,
1428
               "rpc call to node failed (instancelist)")
1429
      if test:
1430
        continue
1431

    
1432
      node_instance[node] = idata
1433

    
1434
      # node_info
1435
      nodeinfo = nresult.get(constants.NV_HVINFO, None)
1436
      test = not isinstance(nodeinfo, dict)
1437
      _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
1438
      if test:
1439
        continue
1440

    
1441
      # Node time
1442
      ntime = nresult.get(constants.NV_TIME, None)
1443
      try:
1444
        ntime_merged = utils.MergeTime(ntime)
1445
      except (ValueError, TypeError):
1446
        _ErrorIf(test, self.ENODETIME, node, "Node returned invalid time")
1447

    
1448
      if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
1449
        ntime_diff = abs(nvinfo_starttime - ntime_merged)
1450
      elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
1451
        ntime_diff = abs(ntime_merged - nvinfo_endtime)
1452
      else:
1453
        ntime_diff = None
1454

    
1455
      _ErrorIf(ntime_diff is not None, self.ENODETIME, node,
1456
               "Node time diverges by at least %0.1fs from master node time",
1457
               ntime_diff)
1458

    
1459
      if ntime_diff is not None:
1460
        continue
1461

    
1462
      try:
1463
        node_info[node] = {
1464
          "mfree": int(nodeinfo['memory_free']),
1465
          "pinst": [],
1466
          "sinst": [],
1467
          # dictionary holding all instances this node is secondary for,
1468
          # grouped by their primary node. Each key is a cluster node, and each
1469
          # value is a list of instances which have the key as primary and the
1470
          # current node as secondary.  this is handy to calculate N+1 memory
1471
          # availability if you can only failover from a primary to its
1472
          # secondary.
1473
          "sinst-by-pnode": {},
1474
        }
1475
        # FIXME: devise a free space model for file based instances as well
1476
        if vg_name is not None:
1477
          test = (constants.NV_VGLIST not in nresult or
1478
                  vg_name not in nresult[constants.NV_VGLIST])
1479
          _ErrorIf(test, self.ENODELVM, node,
1480
                   "node didn't return data for the volume group '%s'"
1481
                   " - it is either missing or broken", vg_name)
1482
          if test:
1483
            continue
1484
          node_info[node]["dfree"] = int(nresult[constants.NV_VGLIST][vg_name])
1485
      except (ValueError, KeyError):
1486
        _ErrorIf(True, self.ENODERPC, node,
1487
                 "node returned invalid nodeinfo, check lvm/hypervisor")
1488
        continue
1489

    
1490
    node_vol_should = {}
1491

    
1492
    feedback_fn("* Verifying instance status")
1493
    for instance in instancelist:
1494
      if verbose:
1495
        feedback_fn("* Verifying instance %s" % instance)
1496
      inst_config = instanceinfo[instance]
1497
      self._VerifyInstance(instance, inst_config, node_volume,
1498
                           node_instance, n_offline)
1499
      inst_nodes_offline = []
1500

    
1501
      inst_config.MapLVsByNode(node_vol_should)
1502

    
1503
      instance_cfg[instance] = inst_config
1504

    
1505
      pnode = inst_config.primary_node
1506
      _ErrorIf(pnode not in node_info and pnode not in n_offline,
1507
               self.ENODERPC, pnode, "instance %s, connection to"
1508
               " primary node failed", instance)
1509
      if pnode in node_info:
1510
        node_info[pnode]['pinst'].append(instance)
1511

    
1512
      if pnode in n_offline:
1513
        inst_nodes_offline.append(pnode)
1514

    
1515
      # If the instance is non-redundant we cannot survive losing its primary
1516
      # node, so we are not N+1 compliant. On the other hand we have no disk
1517
      # templates with more than one secondary so that situation is not well
1518
      # supported either.
1519
      # FIXME: does not support file-backed instances
1520
      if len(inst_config.secondary_nodes) == 0:
1521
        i_non_redundant.append(instance)
1522
      _ErrorIf(len(inst_config.secondary_nodes) > 1,
1523
               self.EINSTANCELAYOUT, instance,
1524
               "instance has multiple secondary nodes", code="WARNING")
1525

    
1526
      if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
1527
        i_non_a_balanced.append(instance)
1528

    
1529
      for snode in inst_config.secondary_nodes:
1530
        _ErrorIf(snode not in node_info and snode not in n_offline,
1531
                 self.ENODERPC, snode,
1532
                 "instance %s, connection to secondary node"
1533
                 "failed", instance)
1534

    
1535
        if snode in node_info:
1536
          node_info[snode]['sinst'].append(instance)
1537
          if pnode not in node_info[snode]['sinst-by-pnode']:
1538
            node_info[snode]['sinst-by-pnode'][pnode] = []
1539
          node_info[snode]['sinst-by-pnode'][pnode].append(instance)
1540

    
1541
        if snode in n_offline:
1542
          inst_nodes_offline.append(snode)
1543

    
1544
      # warn that the instance lives on offline nodes
1545
      _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
1546
               "instance lives on offline node(s) %s",
1547
               utils.CommaJoin(inst_nodes_offline))
1548

    
1549
    feedback_fn("* Verifying orphan volumes")
1550
    self._VerifyOrphanVolumes(node_vol_should, node_volume)
1551

    
1552
    feedback_fn("* Verifying remaining instances")
1553
    self._VerifyOrphanInstances(instancelist, node_instance)
1554

    
1555
    if constants.VERIFY_NPLUSONE_MEM not in self.skip_set:
1556
      feedback_fn("* Verifying N+1 Memory redundancy")
1557
      self._VerifyNPlusOneMemory(node_info, instance_cfg)
1558

    
1559
    feedback_fn("* Other Notes")
1560
    if i_non_redundant:
1561
      feedback_fn("  - NOTICE: %d non-redundant instance(s) found."
1562
                  % len(i_non_redundant))
1563

    
1564
    if i_non_a_balanced:
1565
      feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
1566
                  % len(i_non_a_balanced))
1567

    
1568
    if n_offline:
1569
      feedback_fn("  - NOTICE: %d offline node(s) found." % len(n_offline))
1570

    
1571
    if n_drained:
1572
      feedback_fn("  - NOTICE: %d drained node(s) found." % len(n_drained))
1573

    
1574
    return not self.bad
1575

    
1576
  def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
1577
    """Analyze the post-hooks' result
1578

1579
    This method analyses the hook result, handles it, and sends some
1580
    nicely-formatted feedback back to the user.
1581

1582
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
1583
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
1584
    @param hooks_results: the results of the multi-node hooks rpc call
1585
    @param feedback_fn: function used send feedback back to the caller
1586
    @param lu_result: previous Exec result
1587
    @return: the new Exec result, based on the previous result
1588
        and hook results
1589

1590
    """
1591
    # We only really run POST phase hooks, and are only interested in
1592
    # their results
1593
    if phase == constants.HOOKS_PHASE_POST:
1594
      # Used to change hooks' output to proper indentation
1595
      indent_re = re.compile('^', re.M)
1596
      feedback_fn("* Hooks Results")
1597
      assert hooks_results, "invalid result from hooks"
1598

    
1599
      for node_name in hooks_results:
1600
        res = hooks_results[node_name]
1601
        msg = res.fail_msg
1602
        test = msg and not res.offline
1603
        self._ErrorIf(test, self.ENODEHOOKS, node_name,
1604
                      "Communication failure in hooks execution: %s", msg)
1605
        if test:
1606
          # override manually lu_result here as _ErrorIf only
1607
          # overrides self.bad
1608
          lu_result = 1
1609
          continue
1610
        for script, hkr, output in res.payload:
1611
          test = hkr == constants.HKR_FAIL
1612
          self._ErrorIf(test, self.ENODEHOOKS, node_name,
1613
                        "Script %s failed, output:", script)
1614
          if test:
1615
            output = indent_re.sub('      ', output)
1616
            feedback_fn("%s" % output)
1617
            lu_result = 1
1618

    
1619
      return lu_result
1620

    
1621

    
1622
class LUVerifyDisks(NoHooksLU):
1623
  """Verifies the cluster disks status.
1624

1625
  """
1626
  _OP_REQP = []
1627
  REQ_BGL = False
1628

    
1629
  def ExpandNames(self):
1630
    self.needed_locks = {
1631
      locking.LEVEL_NODE: locking.ALL_SET,
1632
      locking.LEVEL_INSTANCE: locking.ALL_SET,
1633
    }
1634
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
1635

    
1636
  def CheckPrereq(self):
1637
    """Check prerequisites.
1638

1639
    This has no prerequisites.
1640

1641
    """
1642
    pass
1643

    
1644
  def Exec(self, feedback_fn):
1645
    """Verify integrity of cluster disks.
1646

1647
    @rtype: tuple of three items
1648
    @return: a tuple of (dict of node-to-node_error, list of instances
1649
        which need activate-disks, dict of instance: (node, volume) for
1650
        missing volumes
1651

1652
    """
1653
    result = res_nodes, res_instances, res_missing = {}, [], {}
1654

    
1655
    vg_name = self.cfg.GetVGName()
1656
    nodes = utils.NiceSort(self.cfg.GetNodeList())
1657
    instances = [self.cfg.GetInstanceInfo(name)
1658
                 for name in self.cfg.GetInstanceList()]
1659

    
1660
    nv_dict = {}
1661
    for inst in instances:
1662
      inst_lvs = {}
1663
      if (not inst.admin_up or
1664
          inst.disk_template not in constants.DTS_NET_MIRROR):
1665
        continue
1666
      inst.MapLVsByNode(inst_lvs)
1667
      # transform { iname: {node: [vol,],},} to {(node, vol): iname}
1668
      for node, vol_list in inst_lvs.iteritems():
1669
        for vol in vol_list:
1670
          nv_dict[(node, vol)] = inst
1671

    
1672
    if not nv_dict:
1673
      return result
1674

    
1675
    node_lvs = self.rpc.call_lv_list(nodes, vg_name)
1676

    
1677
    for node in nodes:
1678
      # node_volume
1679
      node_res = node_lvs[node]
1680
      if node_res.offline:
1681
        continue
1682
      msg = node_res.fail_msg
1683
      if msg:
1684
        logging.warning("Error enumerating LVs on node %s: %s", node, msg)
1685
        res_nodes[node] = msg
1686
        continue
1687

    
1688
      lvs = node_res.payload
1689
      for lv_name, (_, _, lv_online) in lvs.items():
1690
        inst = nv_dict.pop((node, lv_name), None)
1691
        if (not lv_online and inst is not None
1692
            and inst.name not in res_instances):
1693
          res_instances.append(inst.name)
1694

    
1695
    # any leftover items in nv_dict are missing LVs, let's arrange the
1696
    # data better
1697
    for key, inst in nv_dict.iteritems():
1698
      if inst.name not in res_missing:
1699
        res_missing[inst.name] = []
1700
      res_missing[inst.name].append(key)
1701

    
1702
    return result
1703

    
1704

    
1705
class LURepairDiskSizes(NoHooksLU):
1706
  """Verifies the cluster disks sizes.
1707

1708
  """
1709
  _OP_REQP = ["instances"]
1710
  REQ_BGL = False
1711

    
1712
  def ExpandNames(self):
1713
    if not isinstance(self.op.instances, list):
1714
      raise errors.OpPrereqError("Invalid argument type 'instances'",
1715
                                 errors.ECODE_INVAL)
1716

    
1717
    if self.op.instances:
1718
      self.wanted_names = []
1719
      for name in self.op.instances:
1720
        full_name = self.cfg.ExpandInstanceName(name)
1721
        if full_name is None:
1722
          raise errors.OpPrereqError("Instance '%s' not known" % name,
1723
                                     errors.ECODE_NOENT)
1724
        self.wanted_names.append(full_name)
1725
      self.needed_locks = {
1726
        locking.LEVEL_NODE: [],
1727
        locking.LEVEL_INSTANCE: self.wanted_names,
1728
        }
1729
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
1730
    else:
1731
      self.wanted_names = None
1732
      self.needed_locks = {
1733
        locking.LEVEL_NODE: locking.ALL_SET,
1734
        locking.LEVEL_INSTANCE: locking.ALL_SET,
1735
        }
1736
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
1737

    
1738
  def DeclareLocks(self, level):
1739
    if level == locking.LEVEL_NODE and self.wanted_names is not None:
1740
      self._LockInstancesNodes(primary_only=True)
1741

    
1742
  def CheckPrereq(self):
1743
    """Check prerequisites.
1744

1745
    This only checks the optional instance list against the existing names.
1746

1747
    """
1748
    if self.wanted_names is None:
1749
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
1750

    
1751
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
1752
                             in self.wanted_names]
1753

    
1754
  def _EnsureChildSizes(self, disk):
1755
    """Ensure children of the disk have the needed disk size.
1756

1757
    This is valid mainly for DRBD8 and fixes an issue where the
1758
    children have smaller disk size.
1759

1760
    @param disk: an L{ganeti.objects.Disk} object
1761

1762
    """
1763
    if disk.dev_type == constants.LD_DRBD8:
1764
      assert disk.children, "Empty children for DRBD8?"
1765
      fchild = disk.children[0]
1766
      mismatch = fchild.size < disk.size
1767
      if mismatch:
1768
        self.LogInfo("Child disk has size %d, parent %d, fixing",
1769
                     fchild.size, disk.size)
1770
        fchild.size = disk.size
1771

    
1772
      # and we recurse on this child only, not on the metadev
1773
      return self._EnsureChildSizes(fchild) or mismatch
1774
    else:
1775
      return False
1776

    
1777
  def Exec(self, feedback_fn):
1778
    """Verify the size of cluster disks.
1779

1780
    """
1781
    # TODO: check child disks too
1782
    # TODO: check differences in size between primary/secondary nodes
1783
    per_node_disks = {}
1784
    for instance in self.wanted_instances:
1785
      pnode = instance.primary_node
1786
      if pnode not in per_node_disks:
1787
        per_node_disks[pnode] = []
1788
      for idx, disk in enumerate(instance.disks):
1789
        per_node_disks[pnode].append((instance, idx, disk))
1790

    
1791
    changed = []
1792
    for node, dskl in per_node_disks.items():
1793
      newl = [v[2].Copy() for v in dskl]
1794
      for dsk in newl:
1795
        self.cfg.SetDiskID(dsk, node)
1796
      result = self.rpc.call_blockdev_getsizes(node, newl)
1797
      if result.fail_msg:
1798
        self.LogWarning("Failure in blockdev_getsizes call to node"
1799
                        " %s, ignoring", node)
1800
        continue
1801
      if len(result.data) != len(dskl):
1802
        self.LogWarning("Invalid result from node %s, ignoring node results",
1803
                        node)
1804
        continue
1805
      for ((instance, idx, disk), size) in zip(dskl, result.data):
1806
        if size is None:
1807
          self.LogWarning("Disk %d of instance %s did not return size"
1808
                          " information, ignoring", idx, instance.name)
1809
          continue
1810
        if not isinstance(size, (int, long)):
1811
          self.LogWarning("Disk %d of instance %s did not return valid"
1812
                          " size information, ignoring", idx, instance.name)
1813
          continue
1814
        size = size >> 20
1815
        if size != disk.size:
1816
          self.LogInfo("Disk %d of instance %s has mismatched size,"
1817
                       " correcting: recorded %d, actual %d", idx,
1818
                       instance.name, disk.size, size)
1819
          disk.size = size
1820
          self.cfg.Update(instance, feedback_fn)
1821
          changed.append((instance.name, idx, size))
1822
        if self._EnsureChildSizes(disk):
1823
          self.cfg.Update(instance, feedback_fn)
1824
          changed.append((instance.name, idx, disk.size))
1825
    return changed
1826

    
1827

    
1828
class LURenameCluster(LogicalUnit):
1829
  """Rename the cluster.
1830

1831
  """
1832
  HPATH = "cluster-rename"
1833
  HTYPE = constants.HTYPE_CLUSTER
1834
  _OP_REQP = ["name"]
1835

    
1836
  def BuildHooksEnv(self):
1837
    """Build hooks env.
1838

1839
    """
1840
    env = {
1841
      "OP_TARGET": self.cfg.GetClusterName(),
1842
      "NEW_NAME": self.op.name,
1843
      }
1844
    mn = self.cfg.GetMasterNode()
1845
    return env, [mn], [mn]
1846

    
1847
  def CheckPrereq(self):
1848
    """Verify that the passed name is a valid one.
1849

1850
    """
1851
    hostname = utils.GetHostInfo(self.op.name)
1852

    
1853
    new_name = hostname.name
1854
    self.ip = new_ip = hostname.ip
1855
    old_name = self.cfg.GetClusterName()
1856
    old_ip = self.cfg.GetMasterIP()
1857
    if new_name == old_name and new_ip == old_ip:
1858
      raise errors.OpPrereqError("Neither the name nor the IP address of the"
1859
                                 " cluster has changed",
1860
                                 errors.ECODE_INVAL)
1861
    if new_ip != old_ip:
1862
      if utils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
1863
        raise errors.OpPrereqError("The given cluster IP address (%s) is"
1864
                                   " reachable on the network. Aborting." %
1865
                                   new_ip, errors.ECODE_NOTUNIQUE)
1866

    
1867
    self.op.name = new_name
1868

    
1869
  def Exec(self, feedback_fn):
1870
    """Rename the cluster.
1871

1872
    """
1873
    clustername = self.op.name
1874
    ip = self.ip
1875

    
1876
    # shutdown the master IP
1877
    master = self.cfg.GetMasterNode()
1878
    result = self.rpc.call_node_stop_master(master, False)
1879
    result.Raise("Could not disable the master role")
1880

    
1881
    try:
1882
      cluster = self.cfg.GetClusterInfo()
1883
      cluster.cluster_name = clustername
1884
      cluster.master_ip = ip
1885
      self.cfg.Update(cluster, feedback_fn)
1886

    
1887
      # update the known hosts file
1888
      ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
1889
      node_list = self.cfg.GetNodeList()
1890
      try:
1891
        node_list.remove(master)
1892
      except ValueError:
1893
        pass
1894
      result = self.rpc.call_upload_file(node_list,
1895
                                         constants.SSH_KNOWN_HOSTS_FILE)
1896
      for to_node, to_result in result.iteritems():
1897
        msg = to_result.fail_msg
1898
        if msg:
1899
          msg = ("Copy of file %s to node %s failed: %s" %
1900
                 (constants.SSH_KNOWN_HOSTS_FILE, to_node, msg))
1901
          self.proc.LogWarning(msg)
1902

    
1903
    finally:
1904
      result = self.rpc.call_node_start_master(master, False, False)
1905
      msg = result.fail_msg
1906
      if msg:
1907
        self.LogWarning("Could not re-enable the master role on"
1908
                        " the master, please restart manually: %s", msg)
1909

    
1910

    
1911
def _RecursiveCheckIfLVMBased(disk):
1912
  """Check if the given disk or its children are lvm-based.
1913

1914
  @type disk: L{objects.Disk}
1915
  @param disk: the disk to check
1916
  @rtype: boolean
1917
  @return: boolean indicating whether a LD_LV dev_type was found or not
1918

1919
  """
1920
  if disk.children:
1921
    for chdisk in disk.children:
1922
      if _RecursiveCheckIfLVMBased(chdisk):
1923
        return True
1924
  return disk.dev_type == constants.LD_LV
1925

    
1926

    
1927
class LUSetClusterParams(LogicalUnit):
1928
  """Change the parameters of the cluster.
1929

1930
  """
1931
  HPATH = "cluster-modify"
1932
  HTYPE = constants.HTYPE_CLUSTER
1933
  _OP_REQP = []
1934
  REQ_BGL = False
1935

    
1936
  def CheckArguments(self):
1937
    """Check parameters
1938

1939
    """
1940
    if not hasattr(self.op, "candidate_pool_size"):
1941
      self.op.candidate_pool_size = None
1942
    if self.op.candidate_pool_size is not None:
1943
      try:
1944
        self.op.candidate_pool_size = int(self.op.candidate_pool_size)
1945
      except (ValueError, TypeError), err:
1946
        raise errors.OpPrereqError("Invalid candidate_pool_size value: %s" %
1947
                                   str(err), errors.ECODE_INVAL)
1948
      if self.op.candidate_pool_size < 1:
1949
        raise errors.OpPrereqError("At least one master candidate needed",
1950
                                   errors.ECODE_INVAL)
1951

    
1952
  def ExpandNames(self):
1953
    # FIXME: in the future maybe other cluster params won't require checking on
1954
    # all nodes to be modified.
1955
    self.needed_locks = {
1956
      locking.LEVEL_NODE: locking.ALL_SET,
1957
    }
1958
    self.share_locks[locking.LEVEL_NODE] = 1
1959

    
1960
  def BuildHooksEnv(self):
1961
    """Build hooks env.
1962

1963
    """
1964
    env = {
1965
      "OP_TARGET": self.cfg.GetClusterName(),
1966
      "NEW_VG_NAME": self.op.vg_name,
1967
      }
1968
    mn = self.cfg.GetMasterNode()
1969
    return env, [mn], [mn]
1970

    
1971
  def CheckPrereq(self):
1972
    """Check prerequisites.
1973

1974
    This checks whether the given params don't conflict and
1975
    if the given volume group is valid.
1976

1977
    """
1978
    if self.op.vg_name is not None and not self.op.vg_name:
1979
      instances = self.cfg.GetAllInstancesInfo().values()
1980
      for inst in instances:
1981
        for disk in inst.disks:
1982
          if _RecursiveCheckIfLVMBased(disk):
1983
            raise errors.OpPrereqError("Cannot disable lvm storage while"
1984
                                       " lvm-based instances exist",
1985
                                       errors.ECODE_INVAL)
1986

    
1987
    node_list = self.acquired_locks[locking.LEVEL_NODE]
1988

    
1989
    # if vg_name not None, checks given volume group on all nodes
1990
    if self.op.vg_name:
1991
      vglist = self.rpc.call_vg_list(node_list)
1992
      for node in node_list:
1993
        msg = vglist[node].fail_msg
1994
        if msg:
1995
          # ignoring down node
1996
          self.LogWarning("Error while gathering data on node %s"
1997
                          " (ignoring node): %s", node, msg)
1998
          continue
1999
        vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
2000
                                              self.op.vg_name,
2001
                                              constants.MIN_VG_SIZE)
2002
        if vgstatus:
2003
          raise errors.OpPrereqError("Error on node '%s': %s" %
2004
                                     (node, vgstatus), errors.ECODE_ENVIRON)
2005

    
2006
    self.cluster = cluster = self.cfg.GetClusterInfo()
2007
    # validate params changes
2008
    if self.op.beparams:
2009
      utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
2010
      self.new_beparams = objects.FillDict(
2011
        cluster.beparams[constants.PP_DEFAULT], self.op.beparams)
2012

    
2013
    if self.op.nicparams:
2014
      utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
2015
      self.new_nicparams = objects.FillDict(
2016
        cluster.nicparams[constants.PP_DEFAULT], self.op.nicparams)
2017
      objects.NIC.CheckParameterSyntax(self.new_nicparams)
2018
      nic_errors = []
2019

    
2020
      # check all instances for consistency
2021
      for instance in self.cfg.GetAllInstancesInfo().values():
2022
        for nic_idx, nic in enumerate(instance.nics):
2023
          params_copy = copy.deepcopy(nic.nicparams)
2024
          params_filled = objects.FillDict(self.new_nicparams, params_copy)
2025

    
2026
          # check parameter syntax
2027
          try:
2028
            objects.NIC.CheckParameterSyntax(params_filled)
2029
          except errors.ConfigurationError, err:
2030
            nic_errors.append("Instance %s, nic/%d: %s" %
2031
                              (instance.name, nic_idx, err))
2032

    
2033
          # if we're moving instances to routed, check that they have an ip
2034
          target_mode = params_filled[constants.NIC_MODE]
2035
          if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
2036
            nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
2037
                              (instance.name, nic_idx))
2038
      if nic_errors:
2039
        raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
2040
                                   "\n".join(nic_errors))
2041

    
2042
    # hypervisor list/parameters
2043
    self.new_hvparams = objects.FillDict(cluster.hvparams, {})
2044
    if self.op.hvparams:
2045
      if not isinstance(self.op.hvparams, dict):
2046
        raise errors.OpPrereqError("Invalid 'hvparams' parameter on input",
2047
                                   errors.ECODE_INVAL)
2048
      for hv_name, hv_dict in self.op.hvparams.items():
2049
        if hv_name not in self.new_hvparams:
2050
          self.new_hvparams[hv_name] = hv_dict
2051
        else:
2052
          self.new_hvparams[hv_name].update(hv_dict)
2053

    
2054
    if self.op.enabled_hypervisors is not None:
2055
      self.hv_list = self.op.enabled_hypervisors
2056
      if not self.hv_list:
2057
        raise errors.OpPrereqError("Enabled hypervisors list must contain at"
2058
                                   " least one member",
2059
                                   errors.ECODE_INVAL)
2060
      invalid_hvs = set(self.hv_list) - constants.HYPER_TYPES
2061
      if invalid_hvs:
2062
        raise errors.OpPrereqError("Enabled hypervisors contains invalid"
2063
                                   " entries: %s" %
2064
                                   utils.CommaJoin(invalid_hvs),
2065
                                   errors.ECODE_INVAL)
2066
    else:
2067
      self.hv_list = cluster.enabled_hypervisors
2068

    
2069
    if self.op.hvparams or self.op.enabled_hypervisors is not None:
2070
      # either the enabled list has changed, or the parameters have, validate
2071
      for hv_name, hv_params in self.new_hvparams.items():
2072
        if ((self.op.hvparams and hv_name in self.op.hvparams) or
2073
            (self.op.enabled_hypervisors and
2074
             hv_name in self.op.enabled_hypervisors)):
2075
          # either this is a new hypervisor, or its parameters have changed
2076
          hv_class = hypervisor.GetHypervisor(hv_name)
2077
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2078
          hv_class.CheckParameterSyntax(hv_params)
2079
          _CheckHVParams(self, node_list, hv_name, hv_params)
2080

    
2081
  def Exec(self, feedback_fn):
2082
    """Change the parameters of the cluster.
2083

2084
    """
2085
    if self.op.vg_name is not None:
2086
      new_volume = self.op.vg_name
2087
      if not new_volume:
2088
        new_volume = None
2089
      if new_volume != self.cfg.GetVGName():
2090
        self.cfg.SetVGName(new_volume)
2091
      else:
2092
        feedback_fn("Cluster LVM configuration already in desired"
2093
                    " state, not changing")
2094
    if self.op.hvparams:
2095
      self.cluster.hvparams = self.new_hvparams
2096
    if self.op.enabled_hypervisors is not None:
2097
      self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
2098
    if self.op.beparams:
2099
      self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
2100
    if self.op.nicparams:
2101
      self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
2102

    
2103
    if self.op.candidate_pool_size is not None:
2104
      self.cluster.candidate_pool_size = self.op.candidate_pool_size
2105
      # we need to update the pool size here, otherwise the save will fail
2106
      _AdjustCandidatePool(self, [])
2107

    
2108
    self.cfg.Update(self.cluster, feedback_fn)
2109

    
2110

    
2111
def _RedistributeAncillaryFiles(lu, additional_nodes=None):
2112
  """Distribute additional files which are part of the cluster configuration.
2113

2114
  ConfigWriter takes care of distributing the config and ssconf files, but
2115
  there are more files which should be distributed to all nodes. This function
2116
  makes sure those are copied.
2117

2118
  @param lu: calling logical unit
2119
  @param additional_nodes: list of nodes not in the config to distribute to
2120

2121
  """
2122
  # 1. Gather target nodes
2123
  myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
2124
  dist_nodes = lu.cfg.GetNodeList()
2125
  if additional_nodes is not None:
2126
    dist_nodes.extend(additional_nodes)
2127
  if myself.name in dist_nodes:
2128
    dist_nodes.remove(myself.name)
2129

    
2130
  # 2. Gather files to distribute
2131
  dist_files = set([constants.ETC_HOSTS,
2132
                    constants.SSH_KNOWN_HOSTS_FILE,
2133
                    constants.RAPI_CERT_FILE,
2134
                    constants.RAPI_USERS_FILE,
2135
                    constants.HMAC_CLUSTER_KEY,
2136
                   ])
2137

    
2138
  enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
2139
  for hv_name in enabled_hypervisors:
2140
    hv_class = hypervisor.GetHypervisor(hv_name)
2141
    dist_files.update(hv_class.GetAncillaryFiles())
2142

    
2143
  # 3. Perform the files upload
2144
  for fname in dist_files:
2145
    if os.path.exists(fname):
2146
      result = lu.rpc.call_upload_file(dist_nodes, fname)
2147
      for to_node, to_result in result.items():
2148
        msg = to_result.fail_msg
2149
        if msg:
2150
          msg = ("Copy of file %s to node %s failed: %s" %
2151
                 (fname, to_node, msg))
2152
          lu.proc.LogWarning(msg)
2153

    
2154

    
2155
class LURedistributeConfig(NoHooksLU):
2156
  """Force the redistribution of cluster configuration.
2157

2158
  This is a very simple LU.
2159

2160
  """
2161
  _OP_REQP = []
2162
  REQ_BGL = False
2163

    
2164
  def ExpandNames(self):
2165
    self.needed_locks = {
2166
      locking.LEVEL_NODE: locking.ALL_SET,
2167
    }
2168
    self.share_locks[locking.LEVEL_NODE] = 1
2169

    
2170
  def CheckPrereq(self):
2171
    """Check prerequisites.
2172

2173
    """
2174

    
2175
  def Exec(self, feedback_fn):
2176
    """Redistribute the configuration.
2177

2178
    """
2179
    self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
2180
    _RedistributeAncillaryFiles(self)
2181

    
2182

    
2183
def _WaitForSync(lu, instance, oneshot=False):
2184
  """Sleep and poll for an instance's disk to sync.
2185

2186
  """
2187
  if not instance.disks:
2188
    return True
2189

    
2190
  if not oneshot:
2191
    lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
2192

    
2193
  node = instance.primary_node
2194

    
2195
  for dev in instance.disks:
2196
    lu.cfg.SetDiskID(dev, node)
2197

    
2198
  # TODO: Convert to utils.Retry
2199

    
2200
  retries = 0
2201
  degr_retries = 10 # in seconds, as we sleep 1 second each time
2202
  while True:
2203
    max_time = 0
2204
    done = True
2205
    cumul_degraded = False
2206
    rstats = lu.rpc.call_blockdev_getmirrorstatus(node, instance.disks)
2207
    msg = rstats.fail_msg
2208
    if msg:
2209
      lu.LogWarning("Can't get any data from node %s: %s", node, msg)
2210
      retries += 1
2211
      if retries >= 10:
2212
        raise errors.RemoteError("Can't contact node %s for mirror data,"
2213
                                 " aborting." % node)
2214
      time.sleep(6)
2215
      continue
2216
    rstats = rstats.payload
2217
    retries = 0
2218
    for i, mstat in enumerate(rstats):
2219
      if mstat is None:
2220
        lu.LogWarning("Can't compute data for node %s/%s",
2221
                           node, instance.disks[i].iv_name)
2222
        continue
2223

    
2224
      cumul_degraded = (cumul_degraded or
2225
                        (mstat.is_degraded and mstat.sync_percent is None))
2226
      if mstat.sync_percent is not None:
2227
        done = False
2228
        if mstat.estimated_time is not None:
2229
          rem_time = "%d estimated seconds remaining" % mstat.estimated_time
2230
          max_time = mstat.estimated_time
2231
        else:
2232
          rem_time = "no time estimate"
2233
        lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
2234
                        (instance.disks[i].iv_name, mstat.sync_percent,
2235
                         rem_time))
2236

    
2237
    # if we're done but degraded, let's do a few small retries, to
2238
    # make sure we see a stable and not transient situation; therefore
2239
    # we force restart of the loop
2240
    if (done or oneshot) and cumul_degraded and degr_retries > 0:
2241
      logging.info("Degraded disks found, %d retries left", degr_retries)
2242
      degr_retries -= 1
2243
      time.sleep(1)
2244
      continue
2245

    
2246
    if done or oneshot:
2247
      break
2248

    
2249
    time.sleep(min(60, max_time))
2250

    
2251
  if done:
2252
    lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
2253
  return not cumul_degraded
2254

    
2255

    
2256
def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
2257
  """Check that mirrors are not degraded.
2258

2259
  The ldisk parameter, if True, will change the test from the
2260
  is_degraded attribute (which represents overall non-ok status for
2261
  the device(s)) to the ldisk (representing the local storage status).
2262

2263
  """
2264
  lu.cfg.SetDiskID(dev, node)
2265

    
2266
  result = True
2267

    
2268
  if on_primary or dev.AssembleOnSecondary():
2269
    rstats = lu.rpc.call_blockdev_find(node, dev)
2270
    msg = rstats.fail_msg
2271
    if msg:
2272
      lu.LogWarning("Can't find disk on node %s: %s", node, msg)
2273
      result = False
2274
    elif not rstats.payload:
2275
      lu.LogWarning("Can't find disk on node %s", node)
2276
      result = False
2277
    else:
2278
      if ldisk:
2279
        result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
2280
      else:
2281
        result = result and not rstats.payload.is_degraded
2282

    
2283
  if dev.children:
2284
    for child in dev.children:
2285
      result = result and _CheckDiskConsistency(lu, child, node, on_primary)
2286

    
2287
  return result
2288

    
2289

    
2290
class LUDiagnoseOS(NoHooksLU):
2291
  """Logical unit for OS diagnose/query.
2292

2293
  """
2294
  _OP_REQP = ["output_fields", "names"]
2295
  REQ_BGL = False
2296
  _FIELDS_STATIC = utils.FieldSet()
2297
  _FIELDS_DYNAMIC = utils.FieldSet("name", "valid", "node_status", "variants")
2298
  # Fields that need calculation of global os validity
2299
  _FIELDS_NEEDVALID = frozenset(["valid", "variants"])
2300

    
2301
  def ExpandNames(self):
2302
    if self.op.names:
2303
      raise errors.OpPrereqError("Selective OS query not supported",
2304
                                 errors.ECODE_INVAL)
2305

    
2306
    _CheckOutputFields(static=self._FIELDS_STATIC,
2307
                       dynamic=self._FIELDS_DYNAMIC,
2308
                       selected=self.op.output_fields)
2309

    
2310
    # Lock all nodes, in shared mode
2311
    # Temporary removal of locks, should be reverted later
2312
    # TODO: reintroduce locks when they are lighter-weight
2313
    self.needed_locks = {}
2314
    #self.share_locks[locking.LEVEL_NODE] = 1
2315
    #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
2316

    
2317
  def CheckPrereq(self):
2318
    """Check prerequisites.
2319

2320
    """
2321

    
2322
  @staticmethod
2323
  def _DiagnoseByOS(node_list, rlist):
2324
    """Remaps a per-node return list into an a per-os per-node dictionary
2325

2326
    @param node_list: a list with the names of all nodes
2327
    @param rlist: a map with node names as keys and OS objects as values
2328

2329
    @rtype: dict
2330
    @return: a dictionary with osnames as keys and as value another map, with
2331
        nodes as keys and tuples of (path, status, diagnose) as values, eg::
2332

2333
          {"debian-etch": {"node1": [(/usr/lib/..., True, ""),
2334
                                     (/srv/..., False, "invalid api")],
2335
                           "node2": [(/srv/..., True, "")]}
2336
          }
2337

2338
    """
2339
    all_os = {}
2340
    # we build here the list of nodes that didn't fail the RPC (at RPC
2341
    # level), so that nodes with a non-responding node daemon don't
2342
    # make all OSes invalid
2343
    good_nodes = [node_name for node_name in rlist
2344
                  if not rlist[node_name].fail_msg]
2345
    for node_name, nr in rlist.items():
2346
      if nr.fail_msg or not nr.payload:
2347
        continue
2348
      for name, path, status, diagnose, variants in nr.payload:
2349
        if name not in all_os:
2350
          # build a list of nodes for this os containing empty lists
2351
          # for each node in node_list
2352
          all_os[name] = {}
2353
          for nname in good_nodes:
2354
            all_os[name][nname] = []
2355
        all_os[name][node_name].append((path, status, diagnose, variants))
2356
    return all_os
2357

    
2358
  def Exec(self, feedback_fn):
2359
    """Compute the list of OSes.
2360

2361
    """
2362
    valid_nodes = [node for node in self.cfg.GetOnlineNodeList()]
2363
    node_data = self.rpc.call_os_diagnose(valid_nodes)
2364
    pol = self._DiagnoseByOS(valid_nodes, node_data)
2365
    output = []
2366
    calc_valid = self._FIELDS_NEEDVALID.intersection(self.op.output_fields)
2367
    calc_variants = "variants" in self.op.output_fields
2368

    
2369
    for os_name, os_data in pol.items():
2370
      row = []
2371
      if calc_valid:
2372
        valid = True
2373
        variants = None
2374
        for osl in os_data.values():
2375
          valid = valid and osl and osl[0][1]
2376
          if not valid:
2377
            variants = None
2378
            break
2379
          if calc_variants:
2380
            node_variants = osl[0][3]
2381
            if variants is None:
2382
              variants = node_variants
2383
            else:
2384
              variants = [v for v in variants if v in node_variants]
2385

    
2386
      for field in self.op.output_fields:
2387
        if field == "name":
2388
          val = os_name
2389
        elif field == "valid":
2390
          val = valid
2391
        elif field == "node_status":
2392
          # this is just a copy of the dict
2393
          val = {}
2394
          for node_name, nos_list in os_data.items():
2395
            val[node_name] = nos_list
2396
        elif field == "variants":
2397
          val =  variants
2398
        else:
2399
          raise errors.ParameterError(field)
2400
        row.append(val)
2401
      output.append(row)
2402

    
2403
    return output
2404

    
2405

    
2406
class LURemoveNode(LogicalUnit):
2407
  """Logical unit for removing a node.
2408

2409
  """
2410
  HPATH = "node-remove"
2411
  HTYPE = constants.HTYPE_NODE
2412
  _OP_REQP = ["node_name"]
2413

    
2414
  def BuildHooksEnv(self):
2415
    """Build hooks env.
2416

2417
    This doesn't run on the target node in the pre phase as a failed
2418
    node would then be impossible to remove.
2419

2420
    """
2421
    env = {
2422
      "OP_TARGET": self.op.node_name,
2423
      "NODE_NAME": self.op.node_name,
2424
      }
2425
    all_nodes = self.cfg.GetNodeList()
2426
    if self.op.node_name in all_nodes:
2427
      all_nodes.remove(self.op.node_name)
2428
    return env, all_nodes, all_nodes
2429

    
2430
  def CheckPrereq(self):
2431
    """Check prerequisites.
2432

2433
    This checks:
2434
     - the node exists in the configuration
2435
     - it does not have primary or secondary instances
2436
     - it's not the master
2437

2438
    Any errors are signaled by raising errors.OpPrereqError.
2439

2440
    """
2441
    node = self.cfg.GetNodeInfo(self.cfg.ExpandNodeName(self.op.node_name))
2442
    if node is None:
2443
      raise errors.OpPrereqError("Node '%s' is unknown." % self.op.node_name,
2444
                                 errors.ECODE_NOENT)
2445

    
2446
    instance_list = self.cfg.GetInstanceList()
2447

    
2448
    masternode = self.cfg.GetMasterNode()
2449
    if node.name == masternode:
2450
      raise errors.OpPrereqError("Node is the master node,"
2451
                                 " you need to failover first.",
2452
                                 errors.ECODE_INVAL)
2453

    
2454
    for instance_name in instance_list:
2455
      instance = self.cfg.GetInstanceInfo(instance_name)
2456
      if node.name in instance.all_nodes:
2457
        raise errors.OpPrereqError("Instance %s is still running on the node,"
2458
                                   " please remove first." % instance_name,
2459
                                   errors.ECODE_INVAL)
2460
    self.op.node_name = node.name
2461
    self.node = node
2462

    
2463
  def Exec(self, feedback_fn):
2464
    """Removes the node from the cluster.
2465

2466
    """
2467
    node = self.node
2468
    logging.info("Stopping the node daemon and removing configs from node %s",
2469
                 node.name)
2470

    
2471
    modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
2472

    
2473
    # Promote nodes to master candidate as needed
2474
    _AdjustCandidatePool(self, exceptions=[node.name])
2475
    self.context.RemoveNode(node.name)
2476

    
2477
    # Run post hooks on the node before it's removed
2478
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
2479
    try:
2480
      hm.RunPhase(constants.HOOKS_PHASE_POST, [node.name])
2481
    except:
2482
      # pylint: disable-msg=W0702
2483
      self.LogWarning("Errors occurred running hooks on %s" % node.name)
2484

    
2485
    result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
2486
    msg = result.fail_msg
2487
    if msg:
2488
      self.LogWarning("Errors encountered on the remote node while leaving"
2489
                      " the cluster: %s", msg)
2490

    
2491

    
2492
class LUQueryNodes(NoHooksLU):
2493
  """Logical unit for querying nodes.
2494

2495
  """
2496
  # pylint: disable-msg=W0142
2497
  _OP_REQP = ["output_fields", "names", "use_locking"]
2498
  REQ_BGL = False
2499

    
2500
  _SIMPLE_FIELDS = ["name", "serial_no", "ctime", "mtime", "uuid",
2501
                    "master_candidate", "offline", "drained"]
2502

    
2503
  _FIELDS_DYNAMIC = utils.FieldSet(
2504
    "dtotal", "dfree",
2505
    "mtotal", "mnode", "mfree",
2506
    "bootid",
2507
    "ctotal", "cnodes", "csockets",
2508
    )
2509

    
2510
  _FIELDS_STATIC = utils.FieldSet(*[
2511
    "pinst_cnt", "sinst_cnt",
2512
    "pinst_list", "sinst_list",
2513
    "pip", "sip", "tags",
2514
    "master",
2515
    "role"] + _SIMPLE_FIELDS
2516
    )
2517

    
2518
  def ExpandNames(self):
2519
    _CheckOutputFields(static=self._FIELDS_STATIC,
2520
                       dynamic=self._FIELDS_DYNAMIC,
2521
                       selected=self.op.output_fields)
2522

    
2523
    self.needed_locks = {}
2524
    self.share_locks[locking.LEVEL_NODE] = 1
2525

    
2526
    if self.op.names:
2527
      self.wanted = _GetWantedNodes(self, self.op.names)
2528
    else:
2529
      self.wanted = locking.ALL_SET
2530

    
2531
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
2532
    self.do_locking = self.do_node_query and self.op.use_locking
2533
    if self.do_locking:
2534
      # if we don't request only static fields, we need to lock the nodes
2535
      self.needed_locks[locking.LEVEL_NODE] = self.wanted
2536

    
2537
  def CheckPrereq(self):
2538
    """Check prerequisites.
2539

2540
    """
2541
    # The validation of the node list is done in the _GetWantedNodes,
2542
    # if non empty, and if empty, there's no validation to do
2543
    pass
2544

    
2545
  def Exec(self, feedback_fn):
2546
    """Computes the list of nodes and their attributes.
2547

2548
    """
2549
    all_info = self.cfg.GetAllNodesInfo()
2550
    if self.do_locking:
2551
      nodenames = self.acquired_locks[locking.LEVEL_NODE]
2552
    elif self.wanted != locking.ALL_SET:
2553
      nodenames = self.wanted
2554
      missing = set(nodenames).difference(all_info.keys())
2555
      if missing:
2556
        raise errors.OpExecError(
2557
          "Some nodes were removed before retrieving their data: %s" % missing)
2558
    else:
2559
      nodenames = all_info.keys()
2560

    
2561
    nodenames = utils.NiceSort(nodenames)
2562
    nodelist = [all_info[name] for name in nodenames]
2563

    
2564
    # begin data gathering
2565

    
2566
    if self.do_node_query:
2567
      live_data = {}
2568
      node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
2569
                                          self.cfg.GetHypervisorType())
2570
      for name in nodenames:
2571
        nodeinfo = node_data[name]
2572
        if not nodeinfo.fail_msg and nodeinfo.payload:
2573
          nodeinfo = nodeinfo.payload
2574
          fn = utils.TryConvert
2575
          live_data[name] = {
2576
            "mtotal": fn(int, nodeinfo.get('memory_total', None)),
2577
            "mnode": fn(int, nodeinfo.get('memory_dom0', None)),
2578
            "mfree": fn(int, nodeinfo.get('memory_free', None)),
2579
            "dtotal": fn(int, nodeinfo.get('vg_size', None)),
2580
            "dfree": fn(int, nodeinfo.get('vg_free', None)),
2581
            "ctotal": fn(int, nodeinfo.get('cpu_total', None)),
2582
            "bootid": nodeinfo.get('bootid', None),
2583
            "cnodes": fn(int, nodeinfo.get('cpu_nodes', None)),
2584
            "csockets": fn(int, nodeinfo.get('cpu_sockets', None)),
2585
            }
2586
        else:
2587
          live_data[name] = {}
2588
    else:
2589
      live_data = dict.fromkeys(nodenames, {})
2590

    
2591
    node_to_primary = dict([(name, set()) for name in nodenames])
2592
    node_to_secondary = dict([(name, set()) for name in nodenames])
2593

    
2594
    inst_fields = frozenset(("pinst_cnt", "pinst_list",
2595
                             "sinst_cnt", "sinst_list"))
2596
    if inst_fields & frozenset(self.op.output_fields):
2597
      inst_data = self.cfg.GetAllInstancesInfo()
2598

    
2599
      for inst in inst_data.values():
2600
        if inst.primary_node in node_to_primary:
2601
          node_to_primary[inst.primary_node].add(inst.name)
2602
        for secnode in inst.secondary_nodes:
2603
          if secnode in node_to_secondary:
2604
            node_to_secondary[secnode].add(inst.name)
2605

    
2606
    master_node = self.cfg.GetMasterNode()
2607

    
2608
    # end data gathering
2609

    
2610
    output = []
2611
    for node in nodelist:
2612
      node_output = []
2613
      for field in self.op.output_fields:
2614
        if field in self._SIMPLE_FIELDS:
2615
          val = getattr(node, field)
2616
        elif field == "pinst_list":
2617
          val = list(node_to_primary[node.name])
2618
        elif field == "sinst_list":
2619
          val = list(node_to_secondary[node.name])
2620
        elif field == "pinst_cnt":
2621
          val = len(node_to_primary[node.name])
2622
        elif field == "sinst_cnt":
2623
          val = len(node_to_secondary[node.name])
2624
        elif field == "pip":
2625
          val = node.primary_ip
2626
        elif field == "sip":
2627
          val = node.secondary_ip
2628
        elif field == "tags":
2629
          val = list(node.GetTags())
2630
        elif field == "master":
2631
          val = node.name == master_node
2632
        elif self._FIELDS_DYNAMIC.Matches(field):
2633
          val = live_data[node.name].get(field, None)
2634
        elif field == "role":
2635
          if node.name == master_node:
2636
            val = "M"
2637
          elif node.master_candidate:
2638
            val = "C"
2639
          elif node.drained:
2640
            val = "D"
2641
          elif node.offline:
2642
            val = "O"
2643
          else:
2644
            val = "R"
2645
        else:
2646
          raise errors.ParameterError(field)
2647
        node_output.append(val)
2648
      output.append(node_output)
2649

    
2650
    return output
2651

    
2652

    
2653
class LUQueryNodeVolumes(NoHooksLU):
2654
  """Logical unit for getting volumes on node(s).
2655

2656
  """
2657
  _OP_REQP = ["nodes", "output_fields"]
2658
  REQ_BGL = False
2659
  _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
2660
  _FIELDS_STATIC = utils.FieldSet("node")
2661

    
2662
  def ExpandNames(self):
2663
    _CheckOutputFields(static=self._FIELDS_STATIC,
2664
                       dynamic=self._FIELDS_DYNAMIC,
2665
                       selected=self.op.output_fields)
2666

    
2667
    self.needed_locks = {}
2668
    self.share_locks[locking.LEVEL_NODE] = 1
2669
    if not self.op.nodes:
2670
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
2671
    else:
2672
      self.needed_locks[locking.LEVEL_NODE] = \
2673
        _GetWantedNodes(self, self.op.nodes)
2674

    
2675
  def CheckPrereq(self):
2676
    """Check prerequisites.
2677

2678
    This checks that the fields required are valid output fields.
2679

2680
    """
2681
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
2682

    
2683
  def Exec(self, feedback_fn):
2684
    """Computes the list of nodes and their attributes.
2685

2686
    """
2687
    nodenames = self.nodes
2688
    volumes = self.rpc.call_node_volumes(nodenames)
2689

    
2690
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
2691
             in self.cfg.GetInstanceList()]
2692

    
2693
    lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
2694

    
2695
    output = []
2696
    for node in nodenames:
2697
      nresult = volumes[node]
2698
      if nresult.offline:
2699
        continue
2700
      msg = nresult.fail_msg
2701
      if msg:
2702
        self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
2703
        continue
2704

    
2705
      node_vols = nresult.payload[:]
2706
      node_vols.sort(key=lambda vol: vol['dev'])
2707

    
2708
      for vol in node_vols:
2709
        node_output = []
2710
        for field in self.op.output_fields:
2711
          if field == "node":
2712
            val = node
2713
          elif field == "phys":
2714
            val = vol['dev']
2715
          elif field == "vg":
2716
            val = vol['vg']
2717
          elif field == "name":
2718
            val = vol['name']
2719
          elif field == "size":
2720
            val = int(float(vol['size']))
2721
          elif field == "instance":
2722
            for inst in ilist:
2723
              if node not in lv_by_node[inst]:
2724
                continue
2725
              if vol['name'] in lv_by_node[inst][node]:
2726
                val = inst.name
2727
                break
2728
            else:
2729
              val = '-'
2730
          else:
2731
            raise errors.ParameterError(field)
2732
          node_output.append(str(val))
2733

    
2734
        output.append(node_output)
2735

    
2736
    return output
2737

    
2738

    
2739
class LUQueryNodeStorage(NoHooksLU):
2740
  """Logical unit for getting information on storage units on node(s).
2741

2742
  """
2743
  _OP_REQP = ["nodes", "storage_type", "output_fields"]
2744
  REQ_BGL = False
2745
  _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
2746

    
2747
  def ExpandNames(self):
2748
    storage_type = self.op.storage_type
2749

    
2750
    if storage_type not in constants.VALID_STORAGE_TYPES:
2751
      raise errors.OpPrereqError("Unknown storage type: %s" % storage_type,
2752
                                 errors.ECODE_INVAL)
2753

    
2754
    _CheckOutputFields(static=self._FIELDS_STATIC,
2755
                       dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
2756
                       selected=self.op.output_fields)
2757

    
2758
    self.needed_locks = {}
2759
    self.share_locks[locking.LEVEL_NODE] = 1
2760

    
2761
    if self.op.nodes:
2762
      self.needed_locks[locking.LEVEL_NODE] = \
2763
        _GetWantedNodes(self, self.op.nodes)
2764
    else:
2765
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
2766

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

2770
    This checks that the fields required are valid output fields.
2771

2772
    """
2773
    self.op.name = getattr(self.op, "name", None)
2774

    
2775
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
2776

    
2777
  def Exec(self, feedback_fn):
2778
    """Computes the list of nodes and their attributes.
2779

2780
    """
2781
    # Always get name to sort by
2782
    if constants.SF_NAME in self.op.output_fields:
2783
      fields = self.op.output_fields[:]
2784
    else:
2785
      fields = [constants.SF_NAME] + self.op.output_fields
2786

    
2787
    # Never ask for node or type as it's only known to the LU
2788
    for extra in [constants.SF_NODE, constants.SF_TYPE]:
2789
      while extra in fields:
2790
        fields.remove(extra)
2791

    
2792
    field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
2793
    name_idx = field_idx[constants.SF_NAME]
2794

    
2795
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
2796
    data = self.rpc.call_storage_list(self.nodes,
2797
                                      self.op.storage_type, st_args,
2798
                                      self.op.name, fields)
2799

    
2800
    result = []
2801

    
2802
    for node in utils.NiceSort(self.nodes):
2803
      nresult = data[node]
2804
      if nresult.offline:
2805
        continue
2806

    
2807
      msg = nresult.fail_msg
2808
      if msg:
2809
        self.LogWarning("Can't get storage data from node %s: %s", node, msg)
2810
        continue
2811

    
2812
      rows = dict([(row[name_idx], row) for row in nresult.payload])
2813

    
2814
      for name in utils.NiceSort(rows.keys()):
2815
        row = rows[name]
2816

    
2817
        out = []
2818

    
2819
        for field in self.op.output_fields:
2820
          if field == constants.SF_NODE:
2821
            val = node
2822
          elif field == constants.SF_TYPE:
2823
            val = self.op.storage_type
2824
          elif field in field_idx:
2825
            val = row[field_idx[field]]
2826
          else:
2827
            raise errors.ParameterError(field)
2828

    
2829
          out.append(val)
2830

    
2831
        result.append(out)
2832

    
2833
    return result
2834

    
2835

    
2836
class LUModifyNodeStorage(NoHooksLU):
2837
  """Logical unit for modifying a storage volume on a node.
2838

2839
  """
2840
  _OP_REQP = ["node_name", "storage_type", "name", "changes"]
2841
  REQ_BGL = False
2842

    
2843
  def CheckArguments(self):
2844
    node_name = self.cfg.ExpandNodeName(self.op.node_name)
2845
    if node_name is None:
2846
      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name,
2847
                                 errors.ECODE_NOENT)
2848

    
2849
    self.op.node_name = node_name
2850

    
2851
    storage_type = self.op.storage_type
2852
    if storage_type not in constants.VALID_STORAGE_TYPES:
2853
      raise errors.OpPrereqError("Unknown storage type: %s" % storage_type,
2854
                                 errors.ECODE_INVAL)
2855

    
2856
  def ExpandNames(self):
2857
    self.needed_locks = {
2858
      locking.LEVEL_NODE: self.op.node_name,
2859
      }
2860

    
2861
  def CheckPrereq(self):
2862
    """Check prerequisites.
2863

2864
    """
2865
    storage_type = self.op.storage_type
2866

    
2867
    try:
2868
      modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
2869
    except KeyError:
2870
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
2871
                                 " modified" % storage_type,
2872
                                 errors.ECODE_INVAL)
2873

    
2874
    diff = set(self.op.changes.keys()) - modifiable
2875
    if diff:
2876
      raise errors.OpPrereqError("The following fields can not be modified for"
2877
                                 " storage units of type '%s': %r" %
2878
                                 (storage_type, list(diff)),
2879
                                 errors.ECODE_INVAL)
2880

    
2881
  def Exec(self, feedback_fn):
2882
    """Computes the list of nodes and their attributes.
2883

2884
    """
2885
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
2886
    result = self.rpc.call_storage_modify(self.op.node_name,
2887
                                          self.op.storage_type, st_args,
2888
                                          self.op.name, self.op.changes)
2889
    result.Raise("Failed to modify storage unit '%s' on %s" %
2890
                 (self.op.name, self.op.node_name))
2891

    
2892

    
2893
class LUAddNode(LogicalUnit):
2894
  """Logical unit for adding node to the cluster.
2895

2896
  """
2897
  HPATH = "node-add"
2898
  HTYPE = constants.HTYPE_NODE
2899
  _OP_REQP = ["node_name"]
2900

    
2901
  def BuildHooksEnv(self):
2902
    """Build hooks env.
2903

2904
    This will run on all nodes before, and on all nodes + the new node after.
2905

2906
    """
2907
    env = {
2908
      "OP_TARGET": self.op.node_name,
2909
      "NODE_NAME": self.op.node_name,
2910
      "NODE_PIP": self.op.primary_ip,
2911
      "NODE_SIP": self.op.secondary_ip,
2912
      }
2913
    nodes_0 = self.cfg.GetNodeList()
2914
    nodes_1 = nodes_0 + [self.op.node_name, ]
2915
    return env, nodes_0, nodes_1
2916

    
2917
  def CheckPrereq(self):
2918
    """Check prerequisites.
2919

2920
    This checks:
2921
     - the new node is not already in the config
2922
     - it is resolvable
2923
     - its parameters (single/dual homed) matches the cluster
2924

2925
    Any errors are signaled by raising errors.OpPrereqError.
2926

2927
    """
2928
    node_name = self.op.node_name
2929
    cfg = self.cfg
2930

    
2931
    dns_data = utils.GetHostInfo(node_name)
2932

    
2933
    node = dns_data.name
2934
    primary_ip = self.op.primary_ip = dns_data.ip
2935
    secondary_ip = getattr(self.op, "secondary_ip", None)
2936
    if secondary_ip is None:
2937
      secondary_ip = primary_ip
2938
    if not utils.IsValidIP(secondary_ip):
2939
      raise errors.OpPrereqError("Invalid secondary IP given",
2940
                                 errors.ECODE_INVAL)
2941
    self.op.secondary_ip = secondary_ip
2942

    
2943
    node_list = cfg.GetNodeList()
2944
    if not self.op.readd and node in node_list:
2945
      raise errors.OpPrereqError("Node %s is already in the configuration" %
2946
                                 node, errors.ECODE_EXISTS)
2947
    elif self.op.readd and node not in node_list:
2948
      raise errors.OpPrereqError("Node %s is not in the configuration" % node,
2949
                                 errors.ECODE_NOENT)
2950

    
2951
    for existing_node_name in node_list:
2952
      existing_node = cfg.GetNodeInfo(existing_node_name)
2953

    
2954
      if self.op.readd and node == existing_node_name:
2955
        if (existing_node.primary_ip != primary_ip or
2956
            existing_node.secondary_ip != secondary_ip):
2957
          raise errors.OpPrereqError("Readded node doesn't have the same IP"
2958
                                     " address configuration as before",
2959
                                     errors.ECODE_INVAL)
2960
        continue
2961

    
2962
      if (existing_node.primary_ip == primary_ip or
2963
          existing_node.secondary_ip == primary_ip or
2964
          existing_node.primary_ip == secondary_ip or
2965
          existing_node.secondary_ip == secondary_ip):
2966
        raise errors.OpPrereqError("New node ip address(es) conflict with"
2967
                                   " existing node %s" % existing_node.name,
2968
                                   errors.ECODE_NOTUNIQUE)
2969

    
2970
    # check that the type of the node (single versus dual homed) is the
2971
    # same as for the master
2972
    myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
2973
    master_singlehomed = myself.secondary_ip == myself.primary_ip
2974
    newbie_singlehomed = secondary_ip == primary_ip
2975
    if master_singlehomed != newbie_singlehomed:
2976
      if master_singlehomed:
2977
        raise errors.OpPrereqError("The master has no private ip but the"
2978
                                   " new node has one",
2979
                                   errors.ECODE_INVAL)
2980
      else:
2981
        raise errors.OpPrereqError("The master has a private ip but the"
2982
                                   " new node doesn't have one",
2983
                                   errors.ECODE_INVAL)
2984

    
2985
    # checks reachability
2986
    if not utils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
2987
      raise errors.OpPrereqError("Node not reachable by ping",
2988
                                 errors.ECODE_ENVIRON)
2989

    
2990
    if not newbie_singlehomed:
2991
      # check reachability from my secondary ip to newbie's secondary ip
2992
      if not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
2993
                           source=myself.secondary_ip):
2994
        raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
2995
                                   " based ping to noded port",
2996
                                   errors.ECODE_ENVIRON)
2997

    
2998
    if self.op.readd:
2999
      exceptions = [node]
3000
    else:
3001
      exceptions = []
3002

    
3003
    self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
3004

    
3005
    if self.op.readd:
3006
      self.new_node = self.cfg.GetNodeInfo(node)
3007
      assert self.new_node is not None, "Can't retrieve locked node %s" % node
3008
    else:
3009
      self.new_node = objects.Node(name=node,
3010
                                   primary_ip=primary_ip,
3011
                                   secondary_ip=secondary_ip,
3012
                                   master_candidate=self.master_candidate,
3013
                                   offline=False, drained=False)
3014

    
3015
  def Exec(self, feedback_fn):
3016
    """Adds the new node to the cluster.
3017

3018
    """
3019
    new_node = self.new_node
3020
    node = new_node.name
3021

    
3022
    # for re-adds, reset the offline/drained/master-candidate flags;
3023
    # we need to reset here, otherwise offline would prevent RPC calls
3024
    # later in the procedure; this also means that if the re-add
3025
    # fails, we are left with a non-offlined, broken node
3026
    if self.op.readd:
3027
      new_node.drained = new_node.offline = False # pylint: disable-msg=W0201
3028
      self.LogInfo("Readding a node, the offline/drained flags were reset")
3029
      # if we demote the node, we do cleanup later in the procedure
3030
      new_node.master_candidate = self.master_candidate
3031

    
3032
    # notify the user about any possible mc promotion
3033
    if new_node.master_candidate:
3034
      self.LogInfo("Node will be a master candidate")
3035

    
3036
    # check connectivity
3037
    result = self.rpc.call_version([node])[node]
3038
    result.Raise("Can't get version information from node %s" % node)
3039
    if constants.PROTOCOL_VERSION == result.payload:
3040
      logging.info("Communication to node %s fine, sw version %s match",
3041
                   node, result.payload)
3042
    else:
3043
      raise errors.OpExecError("Version mismatch master version %s,"
3044
                               " node version %s" %
3045
                               (constants.PROTOCOL_VERSION, result.payload))
3046

    
3047
    # setup ssh on node
3048
    if self.cfg.GetClusterInfo().modify_ssh_setup:
3049
      logging.info("Copy ssh key to node %s", node)
3050
      priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
3051
      keyarray = []
3052
      keyfiles = [constants.SSH_HOST_DSA_PRIV, constants.SSH_HOST_DSA_PUB,
3053
                  constants.SSH_HOST_RSA_PRIV, constants.SSH_HOST_RSA_PUB,
3054
                  priv_key, pub_key]
3055

    
3056
      for i in keyfiles:
3057
        keyarray.append(utils.ReadFile(i))
3058

    
3059
      result = self.rpc.call_node_add(node, keyarray[0], keyarray[1],
3060
                                      keyarray[2], keyarray[3], keyarray[4],
3061
                                      keyarray[5])
3062
      result.Raise("Cannot transfer ssh keys to the new node")
3063

    
3064
    # Add node to our /etc/hosts, and add key to known_hosts
3065
    if self.cfg.GetClusterInfo().modify_etc_hosts:
3066
      utils.AddHostToEtcHosts(new_node.name)
3067

    
3068
    if new_node.secondary_ip != new_node.primary_ip:
3069
      result = self.rpc.call_node_has_ip_address(new_node.name,
3070
                                                 new_node.secondary_ip)
3071
      result.Raise("Failure checking secondary ip on node %s" % new_node.name,
3072
                   prereq=True, ecode=errors.ECODE_ENVIRON)
3073
      if not result.payload:
3074
        raise errors.OpExecError("Node claims it doesn't have the secondary ip"
3075
                                 " you gave (%s). Please fix and re-run this"
3076
                                 " command." % new_node.secondary_ip)
3077

    
3078
    node_verify_list = [self.cfg.GetMasterNode()]
3079
    node_verify_param = {
3080
      constants.NV_NODELIST: [node],
3081
      # TODO: do a node-net-test as well?
3082
    }
3083

    
3084
    result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
3085
                                       self.cfg.GetClusterName())
3086
    for verifier in node_verify_list:
3087
      result[verifier].Raise("Cannot communicate with node %s" % verifier)
3088
      nl_payload = result[verifier].payload[constants.NV_NODELIST]
3089
      if nl_payload:
3090
        for failed in nl_payload:
3091
          feedback_fn("ssh/hostname verification failed"
3092
                      " (checking from %s): %s" %
3093
                      (verifier, nl_payload[failed]))
3094
        raise errors.OpExecError("ssh/hostname verification failed.")
3095

    
3096
    if self.op.readd:
3097
      _RedistributeAncillaryFiles(self)
3098
      self.context.ReaddNode(new_node)
3099
      # make sure we redistribute the config
3100
      self.cfg.Update(new_node, feedback_fn)
3101
      # and make sure the new node will not have old files around
3102
      if not new_node.master_candidate:
3103
        result = self.rpc.call_node_demote_from_mc(new_node.name)
3104
        msg = result.fail_msg
3105
        if msg:
3106
          self.LogWarning("Node failed to demote itself from master"
3107
                          " candidate status: %s" % msg)
3108
    else:
3109
      _RedistributeAncillaryFiles(self, additional_nodes=[node])
3110
      self.context.AddNode(new_node, self.proc.GetECId())
3111

    
3112

    
3113
class LUSetNodeParams(LogicalUnit):
3114
  """Modifies the parameters of a node.
3115

3116
  """
3117
  HPATH = "node-modify"
3118
  HTYPE = constants.HTYPE_NODE
3119
  _OP_REQP = ["node_name"]
3120
  REQ_BGL = False
3121

    
3122
  def CheckArguments(self):
3123
    node_name = self.cfg.ExpandNodeName(self.op.node_name)
3124
    if node_name is None:
3125
      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name,
3126
                                 errors.ECODE_INVAL)
3127
    self.op.node_name = node_name
3128
    _CheckBooleanOpField(self.op, 'master_candidate')
3129
    _CheckBooleanOpField(self.op, 'offline')
3130
    _CheckBooleanOpField(self.op, 'drained')
3131
    all_mods = [self.op.offline, self.op.master_candidate, self.op.drained]
3132
    if all_mods.count(None) == 3:
3133
      raise errors.OpPrereqError("Please pass at least one modification",
3134
                                 errors.ECODE_INVAL)
3135
    if all_mods.count(True) > 1:
3136
      raise errors.OpPrereqError("Can't set the node into more than one"
3137
                                 " state at the same time",
3138
                                 errors.ECODE_INVAL)
3139

    
3140
  def ExpandNames(self):
3141
    self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
3142

    
3143
  def BuildHooksEnv(self):
3144
    """Build hooks env.
3145

3146
    This runs on the master node.
3147

3148
    """
3149
    env = {
3150
      "OP_TARGET": self.op.node_name,
3151
      "MASTER_CANDIDATE": str(self.op.master_candidate),
3152
      "OFFLINE": str(self.op.offline),
3153
      "DRAINED": str(self.op.drained),
3154
      }
3155
    nl = [self.cfg.GetMasterNode(),
3156
          self.op.node_name]
3157
    return env, nl, nl
3158

    
3159
  def CheckPrereq(self):
3160
    """Check prerequisites.
3161

3162
    This only checks the instance list against the existing names.
3163

3164
    """
3165
    node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
3166

    
3167
    if (self.op.master_candidate is not None or
3168
        self.op.drained is not None or
3169
        self.op.offline is not None):
3170
      # we can't change the master's node flags
3171
      if self.op.node_name == self.cfg.GetMasterNode():
3172
        raise errors.OpPrereqError("The master role can be changed"
3173
                                   " only via masterfailover",
3174
                                   errors.ECODE_INVAL)
3175

    
3176
    # Boolean value that tells us whether we're offlining or draining the node
3177
    offline_or_drain = self.op.offline == True or self.op.drained == True
3178
    deoffline_or_drain = self.op.offline == False or self.op.drained == False
3179

    
3180
    if (node.master_candidate and
3181
        (self.op.master_candidate == False or offline_or_drain)):
3182
      cp_size = self.cfg.GetClusterInfo().candidate_pool_size
3183
      mc_now, mc_should, mc_max = self.cfg.GetMasterCandidateStats()
3184
      if mc_now <= cp_size:
3185
        msg = ("Not enough master candidates (desired"
3186
               " %d, new value will be %d)" % (cp_size, mc_now-1))
3187
        # Only allow forcing the operation if it's an offline/drain operation,
3188
        # and we could not possibly promote more nodes.
3189
        # FIXME: this can still lead to issues if in any way another node which
3190
        # could be promoted appears in the meantime.
3191
        if self.op.force and offline_or_drain and mc_should == mc_max:
3192
          self.LogWarning(msg)
3193
        else:
3194
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
3195

    
3196
    if (self.op.master_candidate == True and
3197
        ((node.offline and not self.op.offline == False) or
3198
         (node.drained and not self.op.drained == False))):
3199
      raise errors.OpPrereqError("Node '%s' is offline or drained, can't set"
3200
                                 " to master_candidate" % node.name,
3201
                                 errors.ECODE_INVAL)
3202

    
3203
    # If we're being deofflined/drained, we'll MC ourself if needed
3204
    if (deoffline_or_drain and not offline_or_drain and not
3205
        self.op.master_candidate == True):
3206
      self.op.master_candidate = _DecideSelfPromotion(self)
3207
      if self.op.master_candidate:
3208
        self.LogInfo("Autopromoting node to master candidate")
3209

    
3210
    return
3211

    
3212
  def Exec(self, feedback_fn):
3213
    """Modifies a node.
3214

3215
    """
3216
    node = self.node
3217

    
3218
    result = []
3219
    changed_mc = False
3220

    
3221
    if self.op.offline is not None:
3222
      node.offline = self.op.offline
3223
      result.append(("offline", str(self.op.offline)))
3224
      if self.op.offline == True:
3225
        if node.master_candidate:
3226
          node.master_candidate = False
3227
          changed_mc = True
3228
          result.append(("master_candidate", "auto-demotion due to offline"))
3229
        if node.drained:
3230
          node.drained = False
3231
          result.append(("drained", "clear drained status due to offline"))
3232

    
3233
    if self.op.master_candidate is not None:
3234
      node.master_candidate = self.op.master_candidate
3235
      changed_mc = True
3236
      result.append(("master_candidate", str(self.op.master_candidate)))
3237
      if self.op.master_candidate == False:
3238
        rrc = self.rpc.call_node_demote_from_mc(node.name)
3239
        msg = rrc.fail_msg
3240
        if msg:
3241
          self.LogWarning("Node failed to demote itself: %s" % msg)
3242

    
3243
    if self.op.drained is not None:
3244
      node.drained = self.op.drained
3245
      result.append(("drained", str(self.op.drained)))
3246
      if self.op.drained == True:
3247
        if node.master_candidate:
3248
          node.master_candidate = False
3249
          changed_mc = True
3250
          result.append(("master_candidate", "auto-demotion due to drain"))
3251
          rrc = self.rpc.call_node_demote_from_mc(node.name)
3252
          msg = rrc.fail_msg
3253
          if msg:
3254
            self.LogWarning("Node failed to demote itself: %s" % msg)
3255
        if node.offline:
3256
          node.offline = False
3257
          result.append(("offline", "clear offline status due to drain"))
3258

    
3259
    # this will trigger configuration file update, if needed
3260
    self.cfg.Update(node, feedback_fn)
3261
    # this will trigger job queue propagation or cleanup
3262
    if changed_mc:
3263
      self.context.ReaddNode(node)
3264

    
3265
    return result
3266

    
3267

    
3268
class LUPowercycleNode(NoHooksLU):
3269
  """Powercycles a node.
3270

3271
  """
3272
  _OP_REQP = ["node_name", "force"]
3273
  REQ_BGL = False
3274

    
3275
  def CheckArguments(self):
3276
    node_name = self.cfg.ExpandNodeName(self.op.node_name)
3277
    if node_name is None:
3278
      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name,
3279
                                 errors.ECODE_NOENT)
3280
    self.op.node_name = node_name
3281
    if node_name == self.cfg.GetMasterNode() and not self.op.force:
3282
      raise errors.OpPrereqError("The node is the master and the force"
3283
                                 " parameter was not set",
3284
                                 errors.ECODE_INVAL)
3285

    
3286
  def ExpandNames(self):
3287
    """Locking for PowercycleNode.
3288

3289
    This is a last-resort option and shouldn't block on other
3290
    jobs. Therefore, we grab no locks.
3291

3292
    """
3293
    self.needed_locks = {}
3294

    
3295
  def CheckPrereq(self):
3296
    """Check prerequisites.
3297

3298
    This LU has no prereqs.
3299

3300
    """
3301
    pass
3302

    
3303
  def Exec(self, feedback_fn):
3304
    """Reboots a node.
3305

3306
    """
3307
    result = self.rpc.call_node_powercycle(self.op.node_name,
3308
                                           self.cfg.GetHypervisorType())
3309
    result.Raise("Failed to schedule the reboot")
3310
    return result.payload
3311

    
3312

    
3313
class LUQueryClusterInfo(NoHooksLU):
3314
  """Query cluster configuration.
3315

3316
  """
3317
  _OP_REQP = []
3318
  REQ_BGL = False
3319

    
3320
  def ExpandNames(self):
3321
    self.needed_locks = {}
3322

    
3323
  def CheckPrereq(self):
3324
    """No prerequsites needed for this LU.
3325

3326
    """
3327
    pass
3328

    
3329
  def Exec(self, feedback_fn):
3330
    """Return cluster config.
3331

3332
    """
3333
    cluster = self.cfg.GetClusterInfo()
3334
    result = {
3335
      "software_version": constants.RELEASE_VERSION,
3336
      "protocol_version": constants.PROTOCOL_VERSION,
3337
      "config_version": constants.CONFIG_VERSION,
3338
      "os_api_version": max(constants.OS_API_VERSIONS),
3339
      "export_version": constants.EXPORT_VERSION,
3340
      "architecture": (platform.architecture()[0], platform.machine()),
3341
      "name": cluster.cluster_name,
3342
      "master": cluster.master_node,
3343
      "default_hypervisor": cluster.enabled_hypervisors[0],
3344
      "enabled_hypervisors": cluster.enabled_hypervisors,
3345
      "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
3346
                        for hypervisor_name in cluster.enabled_hypervisors]),
3347
      "beparams": cluster.beparams,
3348
      "nicparams": cluster.nicparams,
3349
      "candidate_pool_size": cluster.candidate_pool_size,
3350
      "master_netdev": cluster.master_netdev,
3351
      "volume_group_name": cluster.volume_group_name,
3352
      "file_storage_dir": cluster.file_storage_dir,
3353
      "ctime": cluster.ctime,
3354
      "mtime": cluster.mtime,
3355
      "uuid": cluster.uuid,
3356
      "tags": list(cluster.GetTags()),
3357
      }
3358

    
3359
    return result
3360

    
3361

    
3362
class LUQueryConfigValues(NoHooksLU):
3363
  """Return configuration values.
3364

3365
  """
3366
  _OP_REQP = []
3367
  REQ_BGL = False
3368
  _FIELDS_DYNAMIC = utils.FieldSet()
3369
  _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag",
3370
                                  "watcher_pause")
3371

    
3372
  def ExpandNames(self):
3373
    self.needed_locks = {}
3374

    
3375
    _CheckOutputFields(static=self._FIELDS_STATIC,
3376
                       dynamic=self._FIELDS_DYNAMIC,
3377
                       selected=self.op.output_fields)
3378

    
3379
  def CheckPrereq(self):
3380
    """No prerequisites.
3381

3382
    """
3383
    pass
3384

    
3385
  def Exec(self, feedback_fn):
3386
    """Dump a representation of the cluster config to the standard output.
3387

3388
    """
3389
    values = []
3390
    for field in self.op.output_fields:
3391
      if field == "cluster_name":
3392
        entry = self.cfg.GetClusterName()
3393
      elif field == "master_node":
3394
        entry = self.cfg.GetMasterNode()
3395
      elif field == "drain_flag":
3396
        entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
3397
      elif field == "watcher_pause":
3398
        return utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
3399
      else:
3400
        raise errors.ParameterError(field)
3401
      values.append(entry)
3402
    return values
3403

    
3404

    
3405
class LUActivateInstanceDisks(NoHooksLU):
3406
  """Bring up an instance's disks.
3407

3408
  """
3409
  _OP_REQP = ["instance_name"]
3410
  REQ_BGL = False
3411

    
3412
  def ExpandNames(self):
3413
    self._ExpandAndLockInstance()
3414
    self.needed_locks[locking.LEVEL_NODE] = []
3415
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3416

    
3417
  def DeclareLocks(self, level):
3418
    if level == locking.LEVEL_NODE:
3419
      self._LockInstancesNodes()
3420

    
3421
  def CheckPrereq(self):
3422
    """Check prerequisites.
3423

3424
    This checks that the instance is in the cluster.
3425

3426
    """
3427
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3428
    assert self.instance is not None, \
3429
      "Cannot retrieve locked instance %s" % self.op.instance_name
3430
    _CheckNodeOnline(self, self.instance.primary_node)
3431
    if not hasattr(self.op, "ignore_size"):
3432
      self.op.ignore_size = False
3433

    
3434
  def Exec(self, feedback_fn):
3435
    """Activate the disks.
3436

3437
    """
3438
    disks_ok, disks_info = \
3439
              _AssembleInstanceDisks(self, self.instance,
3440
                                     ignore_size=self.op.ignore_size)
3441
    if not disks_ok:
3442
      raise errors.OpExecError("Cannot activate block devices")
3443

    
3444
    return disks_info
3445

    
3446

    
3447
def _AssembleInstanceDisks(lu, instance, ignore_secondaries=False,
3448
                           ignore_size=False):
3449
  """Prepare the block devices for an instance.
3450

3451
  This sets up the block devices on all nodes.
3452

3453
  @type lu: L{LogicalUnit}
3454
  @param lu: the logical unit on whose behalf we execute
3455
  @type instance: L{objects.Instance}
3456
  @param instance: the instance for whose disks we assemble
3457
  @type ignore_secondaries: boolean
3458
  @param ignore_secondaries: if true, errors on secondary nodes
3459
      won't result in an error return from the function
3460
  @type ignore_size: boolean
3461
  @param ignore_size: if true, the current known size of the disk
3462
      will not be used during the disk activation, useful for cases
3463
      when the size is wrong
3464
  @return: False if the operation failed, otherwise a list of
3465
      (host, instance_visible_name, node_visible_name)
3466
      with the mapping from node devices to instance devices
3467

3468
  """
3469
  device_info = []
3470
  disks_ok = True
3471
  iname = instance.name
3472
  # With the two passes mechanism we try to reduce the window of
3473
  # opportunity for the race condition of switching DRBD to primary
3474
  # before handshaking occured, but we do not eliminate it
3475

    
3476
  # The proper fix would be to wait (with some limits) until the
3477
  # connection has been made and drbd transitions from WFConnection
3478
  # into any other network-connected state (Connected, SyncTarget,
3479
  # SyncSource, etc.)
3480

    
3481
  # 1st pass, assemble on all nodes in secondary mode
3482
  for inst_disk in instance.disks:
3483
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
3484
      if ignore_size:
3485
        node_disk = node_disk.Copy()
3486
        node_disk.UnsetSize()
3487
      lu.cfg.SetDiskID(node_disk, node)
3488
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
3489
      msg = result.fail_msg
3490
      if msg:
3491
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
3492
                           " (is_primary=False, pass=1): %s",
3493
                           inst_disk.iv_name, node, msg)
3494
        if not ignore_secondaries:
3495
          disks_ok = False
3496

    
3497
  # FIXME: race condition on drbd migration to primary
3498

    
3499
  # 2nd pass, do only the primary node
3500
  for inst_disk in instance.disks:
3501
    dev_path = None
3502

    
3503
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
3504
      if node != instance.primary_node:
3505
        continue
3506
      if ignore_size:
3507
        node_disk = node_disk.Copy()
3508
        node_disk.UnsetSize()
3509
      lu.cfg.SetDiskID(node_disk, node)
3510
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
3511
      msg = result.fail_msg
3512
      if msg:
3513
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
3514
                           " (is_primary=True, pass=2): %s",
3515
                           inst_disk.iv_name, node, msg)
3516
        disks_ok = False
3517
      else:
3518
        dev_path = result.payload
3519

    
3520
    device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
3521

    
3522
  # leave the disks configured for the primary node
3523
  # this is a workaround that would be fixed better by
3524
  # improving the logical/physical id handling
3525
  for disk in instance.disks:
3526
    lu.cfg.SetDiskID(disk, instance.primary_node)
3527

    
3528
  return disks_ok, device_info
3529

    
3530

    
3531
def _StartInstanceDisks(lu, instance, force):
3532
  """Start the disks of an instance.
3533

3534
  """
3535
  disks_ok, _ = _AssembleInstanceDisks(lu, instance,
3536
                                           ignore_secondaries=force)
3537
  if not disks_ok:
3538
    _ShutdownInstanceDisks(lu, instance)
3539
    if force is not None and not force:
3540
      lu.proc.LogWarning("", hint="If the message above refers to a"
3541
                         " secondary node,"
3542
                         " you can retry the operation using '--force'.")
3543
    raise errors.OpExecError("Disk consistency error")
3544

    
3545

    
3546
class LUDeactivateInstanceDisks(NoHooksLU):
3547
  """Shutdown an instance's disks.
3548

3549
  """
3550
  _OP_REQP = ["instance_name"]
3551
  REQ_BGL = False
3552

    
3553
  def ExpandNames(self):
3554
    self._ExpandAndLockInstance()
3555
    self.needed_locks[locking.LEVEL_NODE] = []
3556
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3557

    
3558
  def DeclareLocks(self, level):
3559
    if level == locking.LEVEL_NODE:
3560
      self._LockInstancesNodes()
3561

    
3562
  def CheckPrereq(self):
3563
    """Check prerequisites.
3564

3565
    This checks that the instance is in the cluster.
3566

3567
    """
3568
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3569
    assert self.instance is not None, \
3570
      "Cannot retrieve locked instance %s" % self.op.instance_name
3571

    
3572
  def Exec(self, feedback_fn):
3573
    """Deactivate the disks
3574

3575
    """
3576
    instance = self.instance
3577
    _SafeShutdownInstanceDisks(self, instance)
3578

    
3579

    
3580
def _SafeShutdownInstanceDisks(lu, instance):
3581
  """Shutdown block devices of an instance.
3582

3583
  This function checks if an instance is running, before calling
3584
  _ShutdownInstanceDisks.
3585

3586
  """
3587
  pnode = instance.primary_node
3588
  ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
3589
  ins_l.Raise("Can't contact node %s" % pnode)
3590

    
3591
  if instance.name in ins_l.payload:
3592
    raise errors.OpExecError("Instance is running, can't shutdown"
3593
                             " block devices.")
3594

    
3595
  _ShutdownInstanceDisks(lu, instance)
3596

    
3597

    
3598
def _ShutdownInstanceDisks(lu, instance, ignore_primary=False):
3599
  """Shutdown block devices of an instance.
3600

3601
  This does the shutdown on all nodes of the instance.
3602

3603
  If the ignore_primary is false, errors on the primary node are
3604
  ignored.
3605

3606
  """
3607
  all_result = True
3608
  for disk in instance.disks:
3609
    for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
3610
      lu.cfg.SetDiskID(top_disk, node)
3611
      result = lu.rpc.call_blockdev_shutdown(node, top_disk)
3612
      msg = result.fail_msg
3613
      if msg:
3614
        lu.LogWarning("Could not shutdown block device %s on node %s: %s",
3615
                      disk.iv_name, node, msg)
3616
        if not ignore_primary or node != instance.primary_node:
3617
          all_result = False
3618
  return all_result
3619

    
3620

    
3621
def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
3622
  """Checks if a node has enough free memory.
3623

3624
  This function check if a given node has the needed amount of free
3625
  memory. In case the node has less memory or we cannot get the
3626
  information from the node, this function raise an OpPrereqError
3627
  exception.
3628

3629
  @type lu: C{LogicalUnit}
3630
  @param lu: a logical unit from which we get configuration data
3631
  @type node: C{str}
3632
  @param node: the node to check
3633
  @type reason: C{str}
3634
  @param reason: string to use in the error message
3635
  @type requested: C{int}
3636
  @param requested: the amount of memory in MiB to check for
3637
  @type hypervisor_name: C{str}
3638
  @param hypervisor_name: the hypervisor to ask for memory stats
3639
  @raise errors.OpPrereqError: if the node doesn't have enough memory, or
3640
      we cannot check the node
3641

3642
  """
3643
  nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
3644
  nodeinfo[node].Raise("Can't get data from node %s" % node,
3645
                       prereq=True, ecode=errors.ECODE_ENVIRON)
3646
  free_mem = nodeinfo[node].payload.get('memory_free', None)
3647
  if not isinstance(free_mem, int):
3648
    raise errors.OpPrereqError("Can't compute free memory on node %s, result"
3649
                               " was '%s'" % (node, free_mem),
3650
                               errors.ECODE_ENVIRON)
3651
  if requested > free_mem:
3652
    raise errors.OpPrereqError("Not enough memory on node %s for %s:"
3653
                               " needed %s MiB, available %s MiB" %
3654
                               (node, reason, requested, free_mem),
3655
                               errors.ECODE_NORES)
3656

    
3657

    
3658
class LUStartupInstance(LogicalUnit):
3659
  """Starts an instance.
3660

3661
  """
3662
  HPATH = "instance-start"
3663
  HTYPE = constants.HTYPE_INSTANCE
3664
  _OP_REQP = ["instance_name", "force"]
3665
  REQ_BGL = False
3666

    
3667
  def ExpandNames(self):
3668
    self._ExpandAndLockInstance()
3669

    
3670
  def BuildHooksEnv(self):
3671
    """Build hooks env.
3672

3673
    This runs on master, primary and secondary nodes of the instance.
3674

3675
    """
3676
    env = {
3677
      "FORCE": self.op.force,
3678
      }
3679
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
3680
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3681
    return env, nl, nl
3682

    
3683
  def CheckPrereq(self):
3684
    """Check prerequisites.
3685

3686
    This checks that the instance is in the cluster.
3687

3688
    """
3689
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3690
    assert self.instance is not None, \
3691
      "Cannot retrieve locked instance %s" % self.op.instance_name
3692

    
3693
    # extra beparams
3694
    self.beparams = getattr(self.op, "beparams", {})
3695
    if self.beparams:
3696
      if not isinstance(self.beparams, dict):
3697
        raise errors.OpPrereqError("Invalid beparams passed: %s, expected"
3698
                                   " dict" % (type(self.beparams), ),
3699
                                   errors.ECODE_INVAL)
3700
      # fill the beparams dict
3701
      utils.ForceDictType(self.beparams, constants.BES_PARAMETER_TYPES)
3702
      self.op.beparams = self.beparams
3703

    
3704
    # extra hvparams
3705
    self.hvparams = getattr(self.op, "hvparams", {})
3706
    if self.hvparams:
3707
      if not isinstance(self.hvparams, dict):
3708
        raise errors.OpPrereqError("Invalid hvparams passed: %s, expected"
3709
                                   " dict" % (type(self.hvparams), ),
3710
                                   errors.ECODE_INVAL)
3711

    
3712
      # check hypervisor parameter syntax (locally)
3713
      cluster = self.cfg.GetClusterInfo()
3714
      utils.ForceDictType(self.hvparams, constants.HVS_PARAMETER_TYPES)
3715
      filled_hvp = objects.FillDict(cluster.hvparams[instance.hypervisor],
3716
                                    instance.hvparams)
3717
      filled_hvp.update(self.hvparams)
3718
      hv_type = hypervisor.GetHypervisor(instance.hypervisor)
3719
      hv_type.CheckParameterSyntax(filled_hvp)
3720
      _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
3721
      self.op.hvparams = self.hvparams
3722

    
3723
    _CheckNodeOnline(self, instance.primary_node)
3724

    
3725
    bep = self.cfg.GetClusterInfo().FillBE(instance)
3726
    # check bridges existence
3727
    _CheckInstanceBridgesExist(self, instance)
3728

    
3729
    remote_info = self.rpc.call_instance_info(instance.primary_node,
3730
                                              instance.name,
3731
                                              instance.hypervisor)
3732
    remote_info.Raise("Error checking node %s" % instance.primary_node,
3733
                      prereq=True, ecode=errors.ECODE_ENVIRON)
3734
    if not remote_info.payload: # not running already
3735
      _CheckNodeFreeMemory(self, instance.primary_node,
3736
                           "starting instance %s" % instance.name,
3737
                           bep[constants.BE_MEMORY], instance.hypervisor)
3738

    
3739
  def Exec(self, feedback_fn):
3740
    """Start the instance.
3741

3742
    """
3743
    instance = self.instance
3744
    force = self.op.force
3745

    
3746
    self.cfg.MarkInstanceUp(instance.name)
3747

    
3748
    node_current = instance.primary_node
3749

    
3750
    _StartInstanceDisks(self, instance, force)
3751

    
3752
    result = self.rpc.call_instance_start(node_current, instance,
3753
                                          self.hvparams, self.beparams)
3754
    msg = result.fail_msg
3755
    if msg:
3756
      _ShutdownInstanceDisks(self, instance)
3757
      raise errors.OpExecError("Could not start instance: %s" % msg)
3758

    
3759

    
3760
class LURebootInstance(LogicalUnit):
3761
  """Reboot an instance.
3762

3763
  """
3764
  HPATH = "instance-reboot"
3765
  HTYPE = constants.HTYPE_INSTANCE
3766
  _OP_REQP = ["instance_name", "ignore_secondaries", "reboot_type"]
3767
  REQ_BGL = False
3768

    
3769
  def CheckArguments(self):
3770
    """Check the arguments.
3771

3772
    """
3773
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
3774
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
3775

    
3776
  def ExpandNames(self):
3777
    if self.op.reboot_type not in [constants.INSTANCE_REBOOT_SOFT,
3778
                                   constants.INSTANCE_REBOOT_HARD,
3779
                                   constants.INSTANCE_REBOOT_FULL]:
3780
      raise errors.ParameterError("reboot type not in [%s, %s, %s]" %
3781
                                  (constants.INSTANCE_REBOOT_SOFT,
3782
                                   constants.INSTANCE_REBOOT_HARD,
3783
                                   constants.INSTANCE_REBOOT_FULL))
3784
    self._ExpandAndLockInstance()
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
      "IGNORE_SECONDARIES": self.op.ignore_secondaries,
3794
      "REBOOT_TYPE": self.op.reboot_type,
3795
      "SHUTDOWN_TIMEOUT": self.shutdown_timeout,
3796
      }
3797
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
3798
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3799
    return env, nl, nl
3800

    
3801
  def CheckPrereq(self):
3802
    """Check prerequisites.
3803

3804
    This checks that the instance is in the cluster.
3805

3806
    """
3807
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3808
    assert self.instance is not None, \
3809
      "Cannot retrieve locked instance %s" % self.op.instance_name
3810

    
3811
    _CheckNodeOnline(self, instance.primary_node)
3812

    
3813
    # check bridges existence
3814
    _CheckInstanceBridgesExist(self, instance)
3815

    
3816
  def Exec(self, feedback_fn):
3817
    """Reboot the instance.
3818

3819
    """
3820
    instance = self.instance
3821
    ignore_secondaries = self.op.ignore_secondaries
3822
    reboot_type = self.op.reboot_type
3823

    
3824
    node_current = instance.primary_node
3825

    
3826
    if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
3827
                       constants.INSTANCE_REBOOT_HARD]:
3828
      for disk in instance.disks:
3829
        self.cfg.SetDiskID(disk, node_current)
3830
      result = self.rpc.call_instance_reboot(node_current, instance,
3831
                                             reboot_type,
3832
                                             self.shutdown_timeout)
3833
      result.Raise("Could not reboot instance")
3834
    else:
3835
      result = self.rpc.call_instance_shutdown(node_current, instance,
3836
                                               self.shutdown_timeout)
3837
      result.Raise("Could not shutdown instance for full reboot")
3838
      _ShutdownInstanceDisks(self, instance)
3839
      _StartInstanceDisks(self, instance, ignore_secondaries)
3840
      result = self.rpc.call_instance_start(node_current, instance, None, None)
3841
      msg = result.fail_msg
3842
      if msg:
3843
        _ShutdownInstanceDisks(self, instance)
3844
        raise errors.OpExecError("Could not start instance for"
3845
                                 " full reboot: %s" % msg)
3846

    
3847
    self.cfg.MarkInstanceUp(instance.name)
3848

    
3849

    
3850
class LUShutdownInstance(LogicalUnit):
3851
  """Shutdown an instance.
3852

3853
  """
3854
  HPATH = "instance-stop"
3855
  HTYPE = constants.HTYPE_INSTANCE
3856
  _OP_REQP = ["instance_name"]
3857
  REQ_BGL = False
3858

    
3859
  def CheckArguments(self):
3860
    """Check the arguments.
3861

3862
    """
3863
    self.timeout = getattr(self.op, "timeout",
3864
                           constants.DEFAULT_SHUTDOWN_TIMEOUT)
3865

    
3866
  def ExpandNames(self):
3867
    self._ExpandAndLockInstance()
3868

    
3869
  def BuildHooksEnv(self):
3870
    """Build hooks env.
3871

3872
    This runs on master, primary and secondary nodes of the instance.
3873

3874
    """
3875
    env = _BuildInstanceHookEnvByObject(self, self.instance)
3876
    env["TIMEOUT"] = self.timeout
3877
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3878
    return env, nl, nl
3879

    
3880
  def CheckPrereq(self):
3881
    """Check prerequisites.
3882

3883
    This checks that the instance is in the cluster.
3884

3885
    """
3886
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3887
    assert self.instance is not None, \
3888
      "Cannot retrieve locked instance %s" % self.op.instance_name
3889
    _CheckNodeOnline(self, self.instance.primary_node)
3890

    
3891
  def Exec(self, feedback_fn):
3892
    """Shutdown the instance.
3893

3894
    """
3895
    instance = self.instance
3896
    node_current = instance.primary_node
3897
    timeout = self.timeout
3898
    self.cfg.MarkInstanceDown(instance.name)
3899
    result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
3900
    msg = result.fail_msg
3901
    if msg:
3902
      self.proc.LogWarning("Could not shutdown instance: %s" % msg)
3903

    
3904
    _ShutdownInstanceDisks(self, instance)
3905

    
3906

    
3907
class LUReinstallInstance(LogicalUnit):
3908
  """Reinstall an instance.
3909

3910
  """
3911
  HPATH = "instance-reinstall"
3912
  HTYPE = constants.HTYPE_INSTANCE
3913
  _OP_REQP = ["instance_name"]
3914
  REQ_BGL = False
3915

    
3916
  def ExpandNames(self):
3917
    self._ExpandAndLockInstance()
3918

    
3919
  def BuildHooksEnv(self):
3920
    """Build hooks env.
3921

3922
    This runs on master, primary and secondary nodes of the instance.
3923

3924
    """
3925
    env = _BuildInstanceHookEnvByObject(self, self.instance)
3926
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3927
    return env, nl, nl
3928

    
3929
  def CheckPrereq(self):
3930
    """Check prerequisites.
3931

3932
    This checks that the instance is in the cluster and is not running.
3933

3934
    """
3935
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3936
    assert instance is not None, \
3937
      "Cannot retrieve locked instance %s" % self.op.instance_name
3938
    _CheckNodeOnline(self, instance.primary_node)
3939

    
3940
    if instance.disk_template == constants.DT_DISKLESS:
3941
      raise errors.OpPrereqError("Instance '%s' has no disks" %
3942
                                 self.op.instance_name,
3943
                                 errors.ECODE_INVAL)
3944
    if instance.admin_up:
3945
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
3946
                                 self.op.instance_name,
3947
                                 errors.ECODE_STATE)
3948
    remote_info = self.rpc.call_instance_info(instance.primary_node,
3949
                                              instance.name,
3950
                                              instance.hypervisor)
3951
    remote_info.Raise("Error checking node %s" % instance.primary_node,
3952
                      prereq=True, ecode=errors.ECODE_ENVIRON)
3953
    if remote_info.payload:
3954
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
3955
                                 (self.op.instance_name,
3956
                                  instance.primary_node),
3957
                                 errors.ECODE_STATE)
3958

    
3959
    self.op.os_type = getattr(self.op, "os_type", None)
3960
    self.op.force_variant = getattr(self.op, "force_variant", False)
3961
    if self.op.os_type is not None:
3962
      # OS verification
3963
      pnode = self.cfg.GetNodeInfo(
3964
        self.cfg.ExpandNodeName(instance.primary_node))
3965
      if pnode is None:
3966
        raise errors.OpPrereqError("Primary node '%s' is unknown" %
3967
                                   self.op.pnode, errors.ECODE_NOENT)
3968
      result = self.rpc.call_os_get(pnode.name, self.op.os_type)
3969
      result.Raise("OS '%s' not in supported OS list for primary node %s" %
3970
                   (self.op.os_type, pnode.name),
3971
                   prereq=True, ecode=errors.ECODE_INVAL)
3972
      if not self.op.force_variant:
3973
        _CheckOSVariant(result.payload, self.op.os_type)
3974

    
3975
    self.instance = instance
3976

    
3977
  def Exec(self, feedback_fn):
3978
    """Reinstall the instance.
3979

3980
    """
3981
    inst = self.instance
3982

    
3983
    if self.op.os_type is not None:
3984
      feedback_fn("Changing OS to '%s'..." % self.op.os_type)
3985
      inst.os = self.op.os_type
3986
      self.cfg.Update(inst, feedback_fn)
3987

    
3988
    _StartInstanceDisks(self, inst, None)
3989
    try:
3990
      feedback_fn("Running the instance OS create scripts...")
3991
      result = self.rpc.call_instance_os_add(inst.primary_node, inst, True)
3992
      result.Raise("Could not install OS for instance %s on node %s" %
3993
                   (inst.name, inst.primary_node))
3994
    finally:
3995
      _ShutdownInstanceDisks(self, inst)
3996

    
3997

    
3998
class LURecreateInstanceDisks(LogicalUnit):
3999
  """Recreate an instance's missing disks.
4000

4001
  """
4002
  HPATH = "instance-recreate-disks"
4003
  HTYPE = constants.HTYPE_INSTANCE
4004
  _OP_REQP = ["instance_name", "disks"]
4005
  REQ_BGL = False
4006

    
4007
  def CheckArguments(self):
4008
    """Check the arguments.
4009

4010
    """
4011
    if not isinstance(self.op.disks, list):
4012
      raise errors.OpPrereqError("Invalid disks parameter", errors.ECODE_INVAL)
4013
    for item in self.op.disks:
4014
      if (not isinstance(item, int) or
4015
          item < 0):
4016
        raise errors.OpPrereqError("Invalid disk specification '%s'" %
4017
                                   str(item), errors.ECODE_INVAL)
4018

    
4019
  def ExpandNames(self):
4020
    self._ExpandAndLockInstance()
4021

    
4022
  def BuildHooksEnv(self):
4023
    """Build hooks env.
4024

4025
    This runs on master, primary and secondary nodes of the instance.
4026

4027
    """
4028
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4029
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4030
    return env, nl, nl
4031

    
4032
  def CheckPrereq(self):
4033
    """Check prerequisites.
4034

4035
    This checks that the instance is in the cluster and is not running.
4036

4037
    """
4038
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4039
    assert instance is not None, \
4040
      "Cannot retrieve locked instance %s" % self.op.instance_name
4041
    _CheckNodeOnline(self, instance.primary_node)
4042

    
4043
    if instance.disk_template == constants.DT_DISKLESS:
4044
      raise errors.OpPrereqError("Instance '%s' has no disks" %
4045
                                 self.op.instance_name, errors.ECODE_INVAL)
4046
    if instance.admin_up:
4047
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
4048
                                 self.op.instance_name, errors.ECODE_STATE)
4049
    remote_info = self.rpc.call_instance_info(instance.primary_node,
4050
                                              instance.name,
4051
                                              instance.hypervisor)
4052
    remote_info.Raise("Error checking node %s" % instance.primary_node,
4053
                      prereq=True, ecode=errors.ECODE_ENVIRON)
4054
    if remote_info.payload:
4055
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
4056
                                 (self.op.instance_name,
4057
                                  instance.primary_node), errors.ECODE_STATE)
4058

    
4059
    if not self.op.disks:
4060
      self.op.disks = range(len(instance.disks))
4061
    else:
4062
      for idx in self.op.disks:
4063
        if idx >= len(instance.disks):
4064
          raise errors.OpPrereqError("Invalid disk index passed '%s'" % idx,
4065
                                     errors.ECODE_INVAL)
4066

    
4067
    self.instance = instance
4068

    
4069
  def Exec(self, feedback_fn):
4070
    """Recreate the disks.
4071

4072
    """
4073
    to_skip = []
4074
    for idx, _ in enumerate(self.instance.disks):
4075
      if idx not in self.op.disks: # disk idx has not been passed in
4076
        to_skip.append(idx)
4077
        continue
4078

    
4079
    _CreateDisks(self, self.instance, to_skip=to_skip)
4080

    
4081

    
4082
class LURenameInstance(LogicalUnit):
4083
  """Rename an instance.
4084

4085
  """
4086
  HPATH = "instance-rename"
4087
  HTYPE = constants.HTYPE_INSTANCE
4088
  _OP_REQP = ["instance_name", "new_name"]
4089

    
4090
  def BuildHooksEnv(self):
4091
    """Build hooks env.
4092

4093
    This runs on master, primary and secondary nodes of the instance.
4094

4095
    """
4096
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4097
    env["INSTANCE_NEW_NAME"] = self.op.new_name
4098
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4099
    return env, nl, nl
4100

    
4101
  def CheckPrereq(self):
4102
    """Check prerequisites.
4103

4104
    This checks that the instance is in the cluster and is not running.
4105

4106
    """
4107
    instance = self.cfg.GetInstanceInfo(
4108
      self.cfg.ExpandInstanceName(self.op.instance_name))
4109
    if instance is None:
4110
      raise errors.OpPrereqError("Instance '%s' not known" %
4111
                                 self.op.instance_name, errors.ECODE_NOENT)
4112
    _CheckNodeOnline(self, instance.primary_node)
4113

    
4114
    if instance.admin_up:
4115
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
4116
                                 self.op.instance_name, errors.ECODE_STATE)
4117
    remote_info = self.rpc.call_instance_info(instance.primary_node,
4118
                                              instance.name,
4119
                                              instance.hypervisor)
4120
    remote_info.Raise("Error checking node %s" % instance.primary_node,
4121
                      prereq=True, ecode=errors.ECODE_ENVIRON)
4122
    if remote_info.payload:
4123
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
4124
                                 (self.op.instance_name,
4125
                                  instance.primary_node), errors.ECODE_STATE)
4126
    self.instance = instance
4127

    
4128
    # new name verification
4129
    name_info = utils.GetHostInfo(self.op.new_name)
4130

    
4131
    self.op.new_name = new_name = name_info.name
4132
    instance_list = self.cfg.GetInstanceList()
4133
    if new_name in instance_list:
4134
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
4135
                                 new_name, errors.ECODE_EXISTS)
4136

    
4137
    if not getattr(self.op, "ignore_ip", False):
4138
      if utils.TcpPing(name_info.ip, constants.DEFAULT_NODED_PORT):
4139
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
4140
                                   (name_info.ip, new_name),
4141
                                   errors.ECODE_NOTUNIQUE)
4142

    
4143

    
4144
  def Exec(self, feedback_fn):
4145
    """Reinstall the instance.
4146

4147
    """
4148
    inst = self.instance
4149
    old_name = inst.name
4150

    
4151
    if inst.disk_template == constants.DT_FILE:
4152
      old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
4153

    
4154
    self.cfg.RenameInstance(inst.name, self.op.new_name)
4155
    # Change the instance lock. This is definitely safe while we hold the BGL
4156
    self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
4157
    self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
4158

    
4159
    # re-read the instance from the configuration after rename
4160
    inst = self.cfg.GetInstanceInfo(self.op.new_name)
4161

    
4162
    if inst.disk_template == constants.DT_FILE:
4163
      new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
4164
      result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
4165
                                                     old_file_storage_dir,
4166
                                                     new_file_storage_dir)
4167
      result.Raise("Could not rename on node %s directory '%s' to '%s'"
4168
                   " (but the instance has been renamed in Ganeti)" %
4169
                   (inst.primary_node, old_file_storage_dir,
4170
                    new_file_storage_dir))
4171

    
4172
    _StartInstanceDisks(self, inst, None)
4173
    try:
4174
      result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
4175
                                                 old_name)
4176
      msg = result.fail_msg
4177
      if msg:
4178
        msg = ("Could not run OS rename script for instance %s on node %s"
4179
               " (but the instance has been renamed in Ganeti): %s" %
4180
               (inst.name, inst.primary_node, msg))
4181
        self.proc.LogWarning(msg)
4182
    finally:
4183
      _ShutdownInstanceDisks(self, inst)
4184

    
4185

    
4186
class LURemoveInstance(LogicalUnit):
4187
  """Remove an instance.
4188

4189
  """
4190
  HPATH = "instance-remove"
4191
  HTYPE = constants.HTYPE_INSTANCE
4192
  _OP_REQP = ["instance_name", "ignore_failures"]
4193
  REQ_BGL = False
4194

    
4195
  def CheckArguments(self):
4196
    """Check the arguments.
4197

4198
    """
4199
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
4200
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
4201

    
4202
  def ExpandNames(self):
4203
    self._ExpandAndLockInstance()
4204
    self.needed_locks[locking.LEVEL_NODE] = []
4205
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4206

    
4207
  def DeclareLocks(self, level):
4208
    if level == locking.LEVEL_NODE:
4209
      self._LockInstancesNodes()
4210

    
4211
  def BuildHooksEnv(self):
4212
    """Build hooks env.
4213

4214
    This runs on master, primary and secondary nodes of the instance.
4215

4216
    """
4217
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4218
    env["SHUTDOWN_TIMEOUT"] = self.shutdown_timeout
4219
    nl = [self.cfg.GetMasterNode()]
4220
    return env, nl, nl
4221

    
4222
  def CheckPrereq(self):
4223
    """Check prerequisites.
4224

4225
    This checks that the instance is in the cluster.
4226

4227
    """
4228
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4229
    assert self.instance is not None, \
4230
      "Cannot retrieve locked instance %s" % self.op.instance_name
4231

    
4232
  def Exec(self, feedback_fn):
4233
    """Remove the instance.
4234

4235
    """
4236
    instance = self.instance
4237
    logging.info("Shutting down instance %s on node %s",
4238
                 instance.name, instance.primary_node)
4239

    
4240
    result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
4241
                                             self.shutdown_timeout)
4242
    msg = result.fail_msg
4243
    if msg:
4244
      if self.op.ignore_failures:
4245
        feedback_fn("Warning: can't shutdown instance: %s" % msg)
4246
      else:
4247
        raise errors.OpExecError("Could not shutdown instance %s on"
4248
                                 " node %s: %s" %
4249
                                 (instance.name, instance.primary_node, msg))
4250

    
4251
    logging.info("Removing block devices for instance %s", instance.name)
4252

    
4253
    if not _RemoveDisks(self, instance):
4254
      if self.op.ignore_failures:
4255
        feedback_fn("Warning: can't remove instance's disks")
4256
      else:
4257
        raise errors.OpExecError("Can't remove instance's disks")
4258

    
4259
    logging.info("Removing instance %s out of cluster config", instance.name)
4260

    
4261
    self.cfg.RemoveInstance(instance.name)
4262
    self.remove_locks[locking.LEVEL_INSTANCE] = instance.name
4263

    
4264

    
4265
class LUQueryInstances(NoHooksLU):
4266
  """Logical unit for querying instances.
4267

4268
  """
4269
  # pylint: disable-msg=W0142
4270
  _OP_REQP = ["output_fields", "names", "use_locking"]
4271
  REQ_BGL = False
4272
  _SIMPLE_FIELDS = ["name", "os", "network_port", "hypervisor",
4273
                    "serial_no", "ctime", "mtime", "uuid"]
4274
  _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
4275
                                    "admin_state",
4276
                                    "disk_template", "ip", "mac", "bridge",
4277
                                    "nic_mode", "nic_link",
4278
                                    "sda_size", "sdb_size", "vcpus", "tags",
4279
                                    "network_port", "beparams",
4280
                                    r"(disk)\.(size)/([0-9]+)",
4281
                                    r"(disk)\.(sizes)", "disk_usage",
4282
                                    r"(nic)\.(mac|ip|mode|link)/([0-9]+)",
4283
                                    r"(nic)\.(bridge)/([0-9]+)",
4284
                                    r"(nic)\.(macs|ips|modes|links|bridges)",
4285
                                    r"(disk|nic)\.(count)",
4286
                                    "hvparams",
4287
                                    ] + _SIMPLE_FIELDS +
4288
                                  ["hv/%s" % name
4289
                                   for name in constants.HVS_PARAMETERS
4290
                                   if name not in constants.HVC_GLOBALS] +
4291
                                  ["be/%s" % name
4292
                                   for name in constants.BES_PARAMETERS])
4293
  _FIELDS_DYNAMIC = utils.FieldSet("oper_state", "oper_ram", "status")
4294

    
4295

    
4296
  def ExpandNames(self):
4297
    _CheckOutputFields(static=self._FIELDS_STATIC,
4298
                       dynamic=self._FIELDS_DYNAMIC,
4299
                       selected=self.op.output_fields)
4300

    
4301
    self.needed_locks = {}
4302
    self.share_locks[locking.LEVEL_INSTANCE] = 1
4303
    self.share_locks[locking.LEVEL_NODE] = 1
4304

    
4305
    if self.op.names:
4306
      self.wanted = _GetWantedInstances(self, self.op.names)
4307
    else:
4308
      self.wanted = locking.ALL_SET
4309

    
4310
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
4311
    self.do_locking = self.do_node_query and self.op.use_locking
4312
    if self.do_locking:
4313
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
4314
      self.needed_locks[locking.LEVEL_NODE] = []
4315
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4316

    
4317
  def DeclareLocks(self, level):
4318
    if level == locking.LEVEL_NODE and self.do_locking:
4319
      self._LockInstancesNodes()
4320

    
4321
  def CheckPrereq(self):
4322
    """Check prerequisites.
4323

4324
    """
4325
    pass
4326

    
4327
  def Exec(self, feedback_fn):
4328
    """Computes the list of nodes and their attributes.
4329

4330
    """
4331
    # pylint: disable-msg=R0912
4332
    # way too many branches here
4333
    all_info = self.cfg.GetAllInstancesInfo()
4334
    if self.wanted == locking.ALL_SET:
4335
      # caller didn't specify instance names, so ordering is not important
4336
      if self.do_locking:
4337
        instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
4338
      else:
4339
        instance_names = all_info.keys()
4340
      instance_names = utils.NiceSort(instance_names)
4341
    else:
4342
      # caller did specify names, so we must keep the ordering
4343
      if self.do_locking:
4344
        tgt_set = self.acquired_locks[locking.LEVEL_INSTANCE]
4345
      else:
4346
        tgt_set = all_info.keys()
4347
      missing = set(self.wanted).difference(tgt_set)
4348
      if missing:
4349
        raise errors.OpExecError("Some instances were removed before"
4350
                                 " retrieving their data: %s" % missing)
4351
      instance_names = self.wanted
4352

    
4353
    instance_list = [all_info[iname] for iname in instance_names]
4354

    
4355
    # begin data gathering
4356

    
4357
    nodes = frozenset([inst.primary_node for inst in instance_list])
4358
    hv_list = list(set([inst.hypervisor for inst in instance_list]))
4359

    
4360
    bad_nodes = []
4361
    off_nodes = []
4362
    if self.do_node_query:
4363
      live_data = {}
4364
      node_data = self.rpc.call_all_instances_info(nodes, hv_list)
4365
      for name in nodes:
4366
        result = node_data[name]
4367
        if result.offline:
4368
          # offline nodes will be in both lists
4369
          off_nodes.append(name)
4370
        if result.fail_msg:
4371
          bad_nodes.append(name)
4372
        else:
4373
          if result.payload:
4374
            live_data.update(result.payload)
4375
          # else no instance is alive
4376
    else:
4377
      live_data = dict([(name, {}) for name in instance_names])
4378

    
4379
    # end data gathering
4380

    
4381
    HVPREFIX = "hv/"
4382
    BEPREFIX = "be/"
4383
    output = []
4384
    cluster = self.cfg.GetClusterInfo()
4385
    for instance in instance_list:
4386
      iout = []
4387
      i_hv = cluster.FillHV(instance, skip_globals=True)
4388
      i_be = cluster.FillBE(instance)
4389
      i_nicp = [objects.FillDict(cluster.nicparams[constants.PP_DEFAULT],
4390
                                 nic.nicparams) for nic in instance.nics]
4391
      for field in self.op.output_fields:
4392
        st_match = self._FIELDS_STATIC.Matches(field)
4393
        if field in self._SIMPLE_FIELDS:
4394
          val = getattr(instance, field)
4395
        elif field == "pnode":
4396
          val = instance.primary_node
4397
        elif field == "snodes":
4398
          val = list(instance.secondary_nodes)
4399
        elif field == "admin_state":
4400
          val = instance.admin_up
4401
        elif field == "oper_state":
4402
          if instance.primary_node in bad_nodes:
4403
            val = None
4404
          else:
4405
            val = bool(live_data.get(instance.name))
4406
        elif field == "status":
4407
          if instance.primary_node in off_nodes:
4408
            val = "ERROR_nodeoffline"
4409
          elif instance.primary_node in bad_nodes:
4410
            val = "ERROR_nodedown"
4411
          else:
4412
            running = bool(live_data.get(instance.name))
4413
            if running:
4414
              if instance.admin_up:
4415
                val = "running"
4416
              else:
4417
                val = "ERROR_up"
4418
            else:
4419
              if instance.admin_up:
4420
                val = "ERROR_down"
4421
              else:
4422
                val = "ADMIN_down"
4423
        elif field == "oper_ram":
4424
          if instance.primary_node in bad_nodes:
4425
            val = None
4426
          elif instance.name in live_data:
4427
            val = live_data[instance.name].get("memory", "?")
4428
          else:
4429
            val = "-"
4430
        elif field == "vcpus":
4431
          val = i_be[constants.BE_VCPUS]
4432
        elif field == "disk_template":
4433
          val = instance.disk_template
4434
        elif field == "ip":
4435
          if instance.nics:
4436
            val = instance.nics[0].ip
4437
          else:
4438
            val = None
4439
        elif field == "nic_mode":
4440
          if instance.nics:
4441
            val = i_nicp[0][constants.NIC_MODE]
4442
          else:
4443
            val = None
4444
        elif field == "nic_link":
4445
          if instance.nics:
4446
            val = i_nicp[0][constants.NIC_LINK]
4447
          else:
4448
            val = None
4449
        elif field == "bridge":
4450
          if (instance.nics and
4451
              i_nicp[0][constants.NIC_MODE] == constants.NIC_MODE_BRIDGED):
4452
            val = i_nicp[0][constants.NIC_LINK]
4453
          else:
4454
            val = None
4455
        elif field == "mac":
4456
          if instance.nics:
4457
            val = instance.nics[0].mac
4458
          else:
4459
            val = None
4460
        elif field == "sda_size" or field == "sdb_size":
4461
          idx = ord(field[2]) - ord('a')
4462
          try:
4463
            val = instance.FindDisk(idx).size
4464
          except errors.OpPrereqError:
4465
            val = None
4466
        elif field == "disk_usage": # total disk usage per node
4467
          disk_sizes = [{'size': disk.size} for disk in instance.disks]
4468
          val = _ComputeDiskSize(instance.disk_template, disk_sizes)
4469
        elif field == "tags":
4470
          val = list(instance.GetTags())
4471
        elif field == "hvparams":
4472
          val = i_hv
4473
        elif (field.startswith(HVPREFIX) and
4474
              field[len(HVPREFIX):] in constants.HVS_PARAMETERS and
4475
              field[len(HVPREFIX):] not in constants.HVC_GLOBALS):
4476
          val = i_hv.get(field[len(HVPREFIX):], None)
4477
        elif field == "beparams":
4478
          val = i_be
4479
        elif (field.startswith(BEPREFIX) and
4480
              field[len(BEPREFIX):] in constants.BES_PARAMETERS):
4481
          val = i_be.get(field[len(BEPREFIX):], None)
4482
        elif st_match and st_match.groups():
4483
          # matches a variable list
4484
          st_groups = st_match.groups()
4485
          if st_groups and st_groups[0] == "disk":
4486
            if st_groups[1] == "count":
4487
              val = len(instance.disks)
4488
            elif st_groups[1] == "sizes":
4489
              val = [disk.size for disk in instance.disks]
4490
            elif st_groups[1] == "size":
4491
              try:
4492
                val = instance.FindDisk(st_groups[2]).size
4493
              except errors.OpPrereqError:
4494
                val = None
4495
            else:
4496
              assert False, "Unhandled disk parameter"
4497
          elif st_groups[0] == "nic":
4498
            if st_groups[1] == "count":
4499
              val = len(instance.nics)
4500
            elif st_groups[1] == "macs":
4501
              val = [nic.mac for nic in instance.nics]
4502
            elif st_groups[1] == "ips":
4503
              val = [nic.ip for nic in instance.nics]
4504
            elif st_groups[1] == "modes":
4505
              val = [nicp[constants.NIC_MODE] for nicp in i_nicp]
4506
            elif st_groups[1] == "links":
4507
              val = [nicp[constants.NIC_LINK] for nicp in i_nicp]
4508
            elif st_groups[1] == "bridges":
4509
              val = []
4510
              for nicp in i_nicp:
4511
                if nicp[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
4512
                  val.append(nicp[constants.NIC_LINK])
4513
                else:
4514
                  val.append(None)
4515
            else:
4516
              # index-based item
4517
              nic_idx = int(st_groups[2])
4518
              if nic_idx >= len(instance.nics):
4519
                val = None
4520
              else:
4521
                if st_groups[1] == "mac":
4522
                  val = instance.nics[nic_idx].mac
4523
                elif st_groups[1] == "ip":
4524
                  val = instance.nics[nic_idx].ip
4525
                elif st_groups[1] == "mode":
4526
                  val = i_nicp[nic_idx][constants.NIC_MODE]
4527
                elif st_groups[1] == "link":
4528
                  val = i_nicp[nic_idx][constants.NIC_LINK]
4529
                elif st_groups[1] == "bridge":
4530
                  nic_mode = i_nicp[nic_idx][constants.NIC_MODE]
4531
                  if nic_mode == constants.NIC_MODE_BRIDGED:
4532
                    val = i_nicp[nic_idx][constants.NIC_LINK]
4533
                  else:
4534
                    val = None
4535
                else:
4536
                  assert False, "Unhandled NIC parameter"
4537
          else:
4538
            assert False, ("Declared but unhandled variable parameter '%s'" %
4539
                           field)
4540
        else:
4541
          assert False, "Declared but unhandled parameter '%s'" % field
4542
        iout.append(val)
4543
      output.append(iout)
4544

    
4545
    return output
4546

    
4547

    
4548
class LUFailoverInstance(LogicalUnit):
4549
  """Failover an instance.
4550

4551
  """
4552
  HPATH = "instance-failover"
4553
  HTYPE = constants.HTYPE_INSTANCE
4554
  _OP_REQP = ["instance_name", "ignore_consistency"]
4555
  REQ_BGL = False
4556

    
4557
  def CheckArguments(self):
4558
    """Check the arguments.
4559

4560
    """
4561
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
4562
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
4563

    
4564
  def ExpandNames(self):
4565
    self._ExpandAndLockInstance()
4566
    self.needed_locks[locking.LEVEL_NODE] = []
4567
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4568

    
4569
  def DeclareLocks(self, level):
4570
    if level == locking.LEVEL_NODE:
4571
      self._LockInstancesNodes()
4572

    
4573
  def BuildHooksEnv(self):
4574
    """Build hooks env.
4575

4576
    This runs on master, primary and secondary nodes of the instance.
4577

4578
    """
4579
    env = {
4580
      "IGNORE_CONSISTENCY": self.op.ignore_consistency,
4581
      "SHUTDOWN_TIMEOUT": self.shutdown_timeout,
4582
      }
4583
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4584
    nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
4585
    return env, nl, nl
4586

    
4587
  def CheckPrereq(self):
4588
    """Check prerequisites.
4589

4590
    This checks that the instance is in the cluster.
4591

4592
    """
4593
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4594
    assert self.instance is not None, \
4595
      "Cannot retrieve locked instance %s" % self.op.instance_name
4596

    
4597
    bep = self.cfg.GetClusterInfo().FillBE(instance)
4598
    if instance.disk_template not in constants.DTS_NET_MIRROR:
4599
      raise errors.OpPrereqError("Instance's disk layout is not"
4600
                                 " network mirrored, cannot failover.",
4601
                                 errors.ECODE_STATE)
4602

    
4603
    secondary_nodes = instance.secondary_nodes
4604
    if not secondary_nodes:
4605
      raise errors.ProgrammerError("no secondary node but using "
4606
                                   "a mirrored disk template")
4607

    
4608
    target_node = secondary_nodes[0]
4609
    _CheckNodeOnline(self, target_node)
4610
    _CheckNodeNotDrained(self, target_node)
4611
    if instance.admin_up:
4612
      # check memory requirements on the secondary node
4613
      _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
4614
                           instance.name, bep[constants.BE_MEMORY],
4615
                           instance.hypervisor)
4616
    else:
4617
      self.LogInfo("Not checking memory on the secondary node as"
4618
                   " instance will not be started")
4619

    
4620
    # check bridge existance
4621
    _CheckInstanceBridgesExist(self, instance, node=target_node)
4622

    
4623
  def Exec(self, feedback_fn):
4624
    """Failover an instance.
4625

4626
    The failover is done by shutting it down on its present node and
4627
    starting it on the secondary.
4628

4629
    """
4630
    instance = self.instance
4631

    
4632
    source_node = instance.primary_node
4633
    target_node = instance.secondary_nodes[0]
4634

    
4635
    if instance.admin_up:
4636
      feedback_fn("* checking disk consistency between source and target")
4637
      for dev in instance.disks:
4638
        # for drbd, these are drbd over lvm
4639
        if not _CheckDiskConsistency(self, dev, target_node, False):
4640
          if not self.op.ignore_consistency:
4641
            raise errors.OpExecError("Disk %s is degraded on target node,"
4642
                                     " aborting failover." % dev.iv_name)
4643
    else:
4644
      feedback_fn("* not checking disk consistency as instance is not running")
4645

    
4646
    feedback_fn("* shutting down instance on source node")
4647
    logging.info("Shutting down instance %s on node %s",
4648
                 instance.name, source_node)
4649

    
4650
    result = self.rpc.call_instance_shutdown(source_node, instance,
4651
                                             self.shutdown_timeout)
4652
    msg = result.fail_msg
4653
    if msg:
4654
      if self.op.ignore_consistency:
4655
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
4656
                             " Proceeding anyway. Please make sure node"
4657
                             " %s is down. Error details: %s",
4658
                             instance.name, source_node, source_node, msg)
4659
      else:
4660
        raise errors.OpExecError("Could not shutdown instance %s on"
4661
                                 " node %s: %s" %
4662
                                 (instance.name, source_node, msg))
4663

    
4664
    feedback_fn("* deactivating the instance's disks on source node")
4665
    if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
4666
      raise errors.OpExecError("Can't shut down the instance's disks.")
4667

    
4668
    instance.primary_node = target_node
4669
    # distribute new instance config to the other nodes
4670
    self.cfg.Update(instance, feedback_fn)
4671

    
4672
    # Only start the instance if it's marked as up
4673
    if instance.admin_up:
4674
      feedback_fn("* activating the instance's disks on target node")
4675
      logging.info("Starting instance %s on node %s",
4676
                   instance.name, target_node)
4677

    
4678
      disks_ok, _ = _AssembleInstanceDisks(self, instance,
4679
                                               ignore_secondaries=True)
4680
      if not disks_ok:
4681
        _ShutdownInstanceDisks(self, instance)
4682
        raise errors.OpExecError("Can't activate the instance's disks")
4683

    
4684
      feedback_fn("* starting the instance on the target node")
4685
      result = self.rpc.call_instance_start(target_node, instance, None, None)
4686
      msg = result.fail_msg
4687
      if msg:
4688
        _ShutdownInstanceDisks(self, instance)
4689
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
4690
                                 (instance.name, target_node, msg))
4691

    
4692

    
4693
class LUMigrateInstance(LogicalUnit):
4694
  """Migrate an instance.
4695

4696
  This is migration without shutting down, compared to the failover,
4697
  which is done with shutdown.
4698

4699
  """
4700
  HPATH = "instance-migrate"
4701
  HTYPE = constants.HTYPE_INSTANCE
4702
  _OP_REQP = ["instance_name", "live", "cleanup"]
4703

    
4704
  REQ_BGL = False
4705

    
4706
  def ExpandNames(self):
4707
    self._ExpandAndLockInstance()
4708

    
4709
    self.needed_locks[locking.LEVEL_NODE] = []
4710
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4711

    
4712
    self._migrater = TLMigrateInstance(self, self.op.instance_name,
4713
                                       self.op.live, self.op.cleanup)
4714
    self.tasklets = [self._migrater]
4715

    
4716
  def DeclareLocks(self, level):
4717
    if level == locking.LEVEL_NODE:
4718
      self._LockInstancesNodes()
4719

    
4720
  def BuildHooksEnv(self):
4721
    """Build hooks env.
4722

4723
    This runs on master, primary and secondary nodes of the instance.
4724

4725
    """
4726
    instance = self._migrater.instance
4727
    env = _BuildInstanceHookEnvByObject(self, instance)
4728
    env["MIGRATE_LIVE"] = self.op.live
4729
    env["MIGRATE_CLEANUP"] = self.op.cleanup
4730
    nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
4731
    return env, nl, nl
4732

    
4733

    
4734
class LUMoveInstance(LogicalUnit):
4735
  """Move an instance by data-copying.
4736

4737
  """
4738
  HPATH = "instance-move"
4739
  HTYPE = constants.HTYPE_INSTANCE
4740
  _OP_REQP = ["instance_name", "target_node"]
4741
  REQ_BGL = False
4742

    
4743
  def CheckArguments(self):
4744
    """Check the arguments.
4745

4746
    """
4747
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
4748
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
4749

    
4750
  def ExpandNames(self):
4751
    self._ExpandAndLockInstance()
4752
    target_node = self.cfg.ExpandNodeName(self.op.target_node)
4753
    if target_node is None:
4754
      raise errors.OpPrereqError("Node '%s' not known" %
4755
                                  self.op.target_node, errors.ECODE_NOENT)
4756
    self.op.target_node = target_node
4757
    self.needed_locks[locking.LEVEL_NODE] = [target_node]
4758
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
4759

    
4760
  def DeclareLocks(self, level):
4761
    if level == locking.LEVEL_NODE:
4762
      self._LockInstancesNodes(primary_only=True)
4763

    
4764
  def BuildHooksEnv(self):
4765
    """Build hooks env.
4766

4767
    This runs on master, primary and secondary nodes of the instance.
4768

4769
    """
4770
    env = {
4771
      "TARGET_NODE": self.op.target_node,
4772
      "SHUTDOWN_TIMEOUT": self.shutdown_timeout,
4773
      }
4774
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4775
    nl = [self.cfg.GetMasterNode()] + [self.instance.primary_node,
4776
                                       self.op.target_node]
4777
    return env, nl, nl
4778

    
4779
  def CheckPrereq(self):
4780
    """Check prerequisites.
4781

4782
    This checks that the instance is in the cluster.
4783

4784
    """
4785
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4786
    assert self.instance is not None, \
4787
      "Cannot retrieve locked instance %s" % self.op.instance_name
4788

    
4789
    node = self.cfg.GetNodeInfo(self.op.target_node)
4790
    assert node is not None, \
4791
      "Cannot retrieve locked node %s" % self.op.target_node
4792

    
4793
    self.target_node = target_node = node.name
4794

    
4795
    if target_node == instance.primary_node:
4796
      raise errors.OpPrereqError("Instance %s is already on the node %s" %
4797
                                 (instance.name, target_node),
4798
                                 errors.ECODE_STATE)
4799

    
4800
    bep = self.cfg.GetClusterInfo().FillBE(instance)
4801

    
4802
    for idx, dsk in enumerate(instance.disks):
4803
      if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
4804
        raise errors.OpPrereqError("Instance disk %d has a complex layout,"
4805
                                   " cannot copy" % idx, errors.ECODE_STATE)
4806

    
4807
    _CheckNodeOnline(self, target_node)
4808
    _CheckNodeNotDrained(self, target_node)
4809

    
4810
    if instance.admin_up:
4811
      # check memory requirements on the secondary node
4812
      _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
4813
                           instance.name, bep[constants.BE_MEMORY],
4814
                           instance.hypervisor)
4815
    else:
4816
      self.LogInfo("Not checking memory on the secondary node as"
4817
                   " instance will not be started")
4818

    
4819
    # check bridge existance
4820
    _CheckInstanceBridgesExist(self, instance, node=target_node)
4821

    
4822
  def Exec(self, feedback_fn):
4823
    """Move an instance.
4824

4825
    The move is done by shutting it down on its present node, copying
4826
    the data over (slow) and starting it on the new node.
4827

4828
    """
4829
    instance = self.instance
4830

    
4831
    source_node = instance.primary_node
4832
    target_node = self.target_node
4833

    
4834
    self.LogInfo("Shutting down instance %s on source node %s",
4835
                 instance.name, source_node)
4836

    
4837
    result = self.rpc.call_instance_shutdown(source_node, instance,
4838
                                             self.shutdown_timeout)
4839
    msg = result.fail_msg
4840
    if msg:
4841
      if self.op.ignore_consistency:
4842
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
4843
                             " Proceeding anyway. Please make sure node"
4844
                             " %s is down. Error details: %s",
4845
                             instance.name, source_node, source_node, msg)
4846
      else:
4847
        raise errors.OpExecError("Could not shutdown instance %s on"
4848
                                 " node %s: %s" %
4849
                                 (instance.name, source_node, msg))
4850

    
4851
    # create the target disks
4852
    try:
4853
      _CreateDisks(self, instance, target_node=target_node)
4854
    except errors.OpExecError:
4855
      self.LogWarning("Device creation failed, reverting...")
4856
      try:
4857
        _RemoveDisks(self, instance, target_node=target_node)
4858
      finally:
4859
        self.cfg.ReleaseDRBDMinors(instance.name)
4860
        raise
4861

    
4862
    cluster_name = self.cfg.GetClusterInfo().cluster_name
4863

    
4864
    errs = []
4865
    # activate, get path, copy the data over
4866
    for idx, disk in enumerate(instance.disks):
4867
      self.LogInfo("Copying data for disk %d", idx)
4868
      result = self.rpc.call_blockdev_assemble(target_node, disk,
4869
                                               instance.name, True)
4870
      if result.fail_msg:
4871
        self.LogWarning("Can't assemble newly created disk %d: %s",
4872
                        idx, result.fail_msg)
4873
        errs.append(result.fail_msg)
4874
        break
4875
      dev_path = result.payload
4876
      result = self.rpc.call_blockdev_export(source_node, disk,
4877
                                             target_node, dev_path,
4878
                                             cluster_name)
4879
      if result.fail_msg:
4880
        self.LogWarning("Can't copy data over for disk %d: %s",
4881
                        idx, result.fail_msg)
4882
        errs.append(result.fail_msg)
4883
        break
4884

    
4885
    if errs:
4886
      self.LogWarning("Some disks failed to copy, aborting")
4887
      try:
4888
        _RemoveDisks(self, instance, target_node=target_node)
4889
      finally:
4890
        self.cfg.ReleaseDRBDMinors(instance.name)
4891
        raise errors.OpExecError("Errors during disk copy: %s" %
4892
                                 (",".join(errs),))
4893

    
4894
    instance.primary_node = target_node
4895
    self.cfg.Update(instance, feedback_fn)
4896

    
4897
    self.LogInfo("Removing the disks on the original node")
4898
    _RemoveDisks(self, instance, target_node=source_node)
4899

    
4900
    # Only start the instance if it's marked as up
4901
    if instance.admin_up:
4902
      self.LogInfo("Starting instance %s on node %s",
4903
                   instance.name, target_node)
4904

    
4905
      disks_ok, _ = _AssembleInstanceDisks(self, instance,
4906
                                           ignore_secondaries=True)
4907
      if not disks_ok:
4908
        _ShutdownInstanceDisks(self, instance)
4909
        raise errors.OpExecError("Can't activate the instance's disks")
4910

    
4911
      result = self.rpc.call_instance_start(target_node, instance, None, None)
4912
      msg = result.fail_msg
4913
      if msg:
4914
        _ShutdownInstanceDisks(self, instance)
4915
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
4916
                                 (instance.name, target_node, msg))
4917

    
4918

    
4919
class LUMigrateNode(LogicalUnit):
4920
  """Migrate all instances from a node.
4921

4922
  """
4923
  HPATH = "node-migrate"
4924
  HTYPE = constants.HTYPE_NODE
4925
  _OP_REQP = ["node_name", "live"]
4926
  REQ_BGL = False
4927

    
4928
  def ExpandNames(self):
4929
    self.op.node_name = self.cfg.ExpandNodeName(self.op.node_name)
4930
    if self.op.node_name is None:
4931
      raise errors.OpPrereqError("Node '%s' not known" % self.op.node_name,
4932
                                 errors.ECODE_NOENT)
4933

    
4934
    self.needed_locks = {
4935
      locking.LEVEL_NODE: [self.op.node_name],
4936
      }
4937

    
4938
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
4939

    
4940
    # Create tasklets for migrating instances for all instances on this node
4941
    names = []
4942
    tasklets = []
4943

    
4944
    for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name):
4945
      logging.debug("Migrating instance %s", inst.name)
4946
      names.append(inst.name)
4947

    
4948
      tasklets.append(TLMigrateInstance(self, inst.name, self.op.live, False))
4949

    
4950
    self.tasklets = tasklets
4951

    
4952
    # Declare instance locks
4953
    self.needed_locks[locking.LEVEL_INSTANCE] = names
4954

    
4955
  def DeclareLocks(self, level):
4956
    if level == locking.LEVEL_NODE:
4957
      self._LockInstancesNodes()
4958

    
4959
  def BuildHooksEnv(self):
4960
    """Build hooks env.
4961

4962
    This runs on the master, the primary and all the secondaries.
4963

4964
    """
4965
    env = {
4966
      "NODE_NAME": self.op.node_name,
4967
      }
4968

    
4969
    nl = [self.cfg.GetMasterNode()]
4970

    
4971
    return (env, nl, nl)
4972

    
4973

    
4974
class TLMigrateInstance(Tasklet):
4975
  def __init__(self, lu, instance_name, live, cleanup):
4976
    """Initializes this class.
4977

4978
    """
4979
    Tasklet.__init__(self, lu)
4980

    
4981
    # Parameters
4982
    self.instance_name = instance_name
4983
    self.live = live
4984
    self.cleanup = cleanup
4985

    
4986
  def CheckPrereq(self):
4987
    """Check prerequisites.
4988

4989
    This checks that the instance is in the cluster.
4990

4991
    """
4992
    instance = self.cfg.GetInstanceInfo(
4993
      self.cfg.ExpandInstanceName(self.instance_name))
4994
    if instance is None:
4995
      raise errors.OpPrereqError("Instance '%s' not known" %
4996
                                 self.instance_name, errors.ECODE_NOENT)
4997

    
4998
    if instance.disk_template != constants.DT_DRBD8:
4999
      raise errors.OpPrereqError("Instance's disk layout is not"
5000
                                 " drbd8, cannot migrate.", errors.ECODE_STATE)
5001

    
5002
    secondary_nodes = instance.secondary_nodes
5003
    if not secondary_nodes:
5004
      raise errors.ConfigurationError("No secondary node but using"
5005
                                      " drbd8 disk template")
5006

    
5007
    i_be = self.cfg.GetClusterInfo().FillBE(instance)
5008

    
5009
    target_node = secondary_nodes[0]
5010
    # check memory requirements on the secondary node
5011
    _CheckNodeFreeMemory(self, target_node, "migrating instance %s" %
5012
                         instance.name, i_be[constants.BE_MEMORY],
5013
                         instance.hypervisor)
5014

    
5015
    # check bridge existance
5016
    _CheckInstanceBridgesExist(self, instance, node=target_node)
5017

    
5018
    if not self.cleanup:
5019
      _CheckNodeNotDrained(self, target_node)
5020
      result = self.rpc.call_instance_migratable(instance.primary_node,
5021
                                                 instance)
5022
      result.Raise("Can't migrate, please use failover",
5023
                   prereq=True, ecode=errors.ECODE_STATE)
5024

    
5025
    self.instance = instance
5026

    
5027
  def _WaitUntilSync(self):
5028
    """Poll with custom rpc for disk sync.
5029

5030
    This uses our own step-based rpc call.
5031

5032
    """
5033
    self.feedback_fn("* wait until resync is done")
5034
    all_done = False
5035
    while not all_done:
5036
      all_done = True
5037
      result = self.rpc.call_drbd_wait_sync(self.all_nodes,
5038
                                            self.nodes_ip,
5039
                                            self.instance.disks)
5040
      min_percent = 100
5041
      for node, nres in result.items():
5042
        nres.Raise("Cannot resync disks on node %s" % node)
5043
        node_done, node_percent = nres.payload
5044
        all_done = all_done and node_done
5045
        if node_percent is not None:
5046
          min_percent = min(min_percent, node_percent)
5047
      if not all_done:
5048
        if min_percent < 100:
5049
          self.feedback_fn("   - progress: %.1f%%" % min_percent)
5050
        time.sleep(2)
5051

    
5052
  def _EnsureSecondary(self, node):
5053
    """Demote a node to secondary.
5054

5055
    """
5056
    self.feedback_fn("* switching node %s to secondary mode" % node)
5057

    
5058
    for dev in self.instance.disks:
5059
      self.cfg.SetDiskID(dev, node)
5060

    
5061
    result = self.rpc.call_blockdev_close(node, self.instance.name,
5062
                                          self.instance.disks)
5063
    result.Raise("Cannot change disk to secondary on node %s" % node)
5064

    
5065
  def _GoStandalone(self):
5066
    """Disconnect from the network.
5067

5068
    """
5069
    self.feedback_fn("* changing into standalone mode")
5070
    result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
5071
                                               self.instance.disks)
5072
    for node, nres in result.items():
5073
      nres.Raise("Cannot disconnect disks node %s" % node)
5074

    
5075
  def _GoReconnect(self, multimaster):
5076
    """Reconnect to the network.
5077

5078
    """
5079
    if multimaster:
5080
      msg = "dual-master"
5081
    else:
5082
      msg = "single-master"
5083
    self.feedback_fn("* changing disks into %s mode" % msg)
5084
    result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
5085
                                           self.instance.disks,
5086
                                           self.instance.name, multimaster)
5087
    for node, nres in result.items():
5088
      nres.Raise("Cannot change disks config on node %s" % node)
5089

    
5090
  def _ExecCleanup(self):
5091
    """Try to cleanup after a failed migration.
5092

5093
    The cleanup is done by:
5094
      - check that the instance is running only on one node
5095
        (and update the config if needed)
5096
      - change disks on its secondary node to secondary
5097
      - wait until disks are fully synchronized
5098
      - disconnect from the network
5099
      - change disks into single-master mode
5100
      - wait again until disks are fully synchronized
5101

5102
    """
5103
    instance = self.instance
5104
    target_node = self.target_node
5105
    source_node = self.source_node
5106

    
5107
    # check running on only one node
5108
    self.feedback_fn("* checking where the instance actually runs"
5109
                     " (if this hangs, the hypervisor might be in"
5110
                     " a bad state)")
5111
    ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
5112
    for node, result in ins_l.items():
5113
      result.Raise("Can't contact node %s" % node)
5114

    
5115
    runningon_source = instance.name in ins_l[source_node].payload
5116
    runningon_target = instance.name in ins_l[target_node].payload
5117

    
5118
    if runningon_source and runningon_target:
5119
      raise errors.OpExecError("Instance seems to be running on two nodes,"
5120
                               " or the hypervisor is confused. You will have"
5121
                               " to ensure manually that it runs only on one"
5122
                               " and restart this operation.")
5123

    
5124
    if not (runningon_source or runningon_target):
5125
      raise errors.OpExecError("Instance does not seem to be running at all."
5126
                               " In this case, it's safer to repair by"
5127
                               " running 'gnt-instance stop' to ensure disk"
5128
                               " shutdown, and then restarting it.")
5129

    
5130
    if runningon_target:
5131
      # the migration has actually succeeded, we need to update the config
5132
      self.feedback_fn("* instance running on secondary node (%s),"
5133
                       " updating config" % target_node)
5134
      instance.primary_node = target_node
5135
      self.cfg.Update(instance, self.feedback_fn)
5136
      demoted_node = source_node
5137
    else:
5138
      self.feedback_fn("* instance confirmed to be running on its"
5139
                       " primary node (%s)" % source_node)
5140
      demoted_node = target_node
5141

    
5142
    self._EnsureSecondary(demoted_node)
5143
    try:
5144
      self._WaitUntilSync()
5145
    except errors.OpExecError:
5146
      # we ignore here errors, since if the device is standalone, it
5147
      # won't be able to sync
5148
      pass
5149
    self._GoStandalone()
5150
    self._GoReconnect(False)
5151
    self._WaitUntilSync()
5152

    
5153
    self.feedback_fn("* done")
5154

    
5155
  def _RevertDiskStatus(self):
5156
    """Try to revert the disk status after a failed migration.
5157

5158
    """
5159
    target_node = self.target_node
5160
    try:
5161
      self._EnsureSecondary(target_node)
5162
      self._GoStandalone()
5163
      self._GoReconnect(False)
5164
      self._WaitUntilSync()
5165
    except errors.OpExecError, err:
5166
      self.lu.LogWarning("Migration failed and I can't reconnect the"
5167
                         " drives: error '%s'\n"
5168
                         "Please look and recover the instance status" %
5169
                         str(err))
5170

    
5171
  def _AbortMigration(self):
5172
    """Call the hypervisor code to abort a started migration.
5173

5174
    """
5175
    instance = self.instance
5176
    target_node = self.target_node
5177
    migration_info = self.migration_info
5178

    
5179
    abort_result = self.rpc.call_finalize_migration(target_node,
5180
                                                    instance,
5181
                                                    migration_info,
5182
                                                    False)
5183
    abort_msg = abort_result.fail_msg
5184
    if abort_msg:
5185
      logging.error("Aborting migration failed on target node %s: %s",
5186
                    target_node, abort_msg)
5187
      # Don't raise an exception here, as we stil have to try to revert the
5188
      # disk status, even if this step failed.
5189

    
5190
  def _ExecMigration(self):
5191
    """Migrate an instance.
5192

5193
    The migrate is done by:
5194
      - change the disks into dual-master mode
5195
      - wait until disks are fully synchronized again
5196
      - migrate the instance
5197
      - change disks on the new secondary node (the old primary) to secondary
5198
      - wait until disks are fully synchronized
5199
      - change disks into single-master mode
5200

5201
    """
5202
    instance = self.instance
5203
    target_node = self.target_node
5204
    source_node = self.source_node
5205

    
5206
    self.feedback_fn("* checking disk consistency between source and target")
5207
    for dev in instance.disks:
5208
      if not _CheckDiskConsistency(self, dev, target_node, False):
5209
        raise errors.OpExecError("Disk %s is degraded or not fully"
5210
                                 " synchronized on target node,"
5211
                                 " aborting migrate." % dev.iv_name)
5212

    
5213
    # First get the migration information from the remote node
5214
    result = self.rpc.call_migration_info(source_node, instance)
5215
    msg = result.fail_msg
5216
    if msg:
5217
      log_err = ("Failed fetching source migration information from %s: %s" %
5218
                 (source_node, msg))
5219
      logging.error(log_err)
5220
      raise errors.OpExecError(log_err)
5221

    
5222
    self.migration_info = migration_info = result.payload
5223

    
5224
    # Then switch the disks to master/master mode
5225
    self._EnsureSecondary(target_node)
5226
    self._GoStandalone()
5227
    self._GoReconnect(True)
5228
    self._WaitUntilSync()
5229

    
5230
    self.feedback_fn("* preparing %s to accept the instance" % target_node)
5231
    result = self.rpc.call_accept_instance(target_node,
5232
                                           instance,
5233
                                           migration_info,
5234
                                           self.nodes_ip[target_node])
5235

    
5236
    msg = result.fail_msg
5237
    if msg:
5238
      logging.error("Instance pre-migration failed, trying to revert"
5239
                    " disk status: %s", msg)
5240
      self.feedback_fn("Pre-migration failed, aborting")
5241
      self._AbortMigration()
5242
      self._RevertDiskStatus()
5243
      raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
5244
                               (instance.name, msg))
5245

    
5246
    self.feedback_fn("* migrating instance to %s" % target_node)
5247
    time.sleep(10)
5248
    result = self.rpc.call_instance_migrate(source_node, instance,
5249
                                            self.nodes_ip[target_node],
5250
                                            self.live)
5251
    msg = result.fail_msg
5252
    if msg:
5253
      logging.error("Instance migration failed, trying to revert"
5254
                    " disk status: %s", msg)
5255
      self.feedback_fn("Migration failed, aborting")
5256
      self._AbortMigration()
5257
      self._RevertDiskStatus()
5258
      raise errors.OpExecError("Could not migrate instance %s: %s" %
5259
                               (instance.name, msg))
5260
    time.sleep(10)
5261

    
5262
    instance.primary_node = target_node
5263
    # distribute new instance config to the other nodes
5264
    self.cfg.Update(instance, self.feedback_fn)
5265

    
5266
    result = self.rpc.call_finalize_migration(target_node,
5267
                                              instance,
5268
                                              migration_info,
5269
                                              True)
5270
    msg = result.fail_msg
5271
    if msg:
5272
      logging.error("Instance migration succeeded, but finalization failed:"
5273
                    " %s", msg)
5274
      raise errors.OpExecError("Could not finalize instance migration: %s" %
5275
                               msg)
5276

    
5277
    self._EnsureSecondary(source_node)
5278
    self._WaitUntilSync()
5279
    self._GoStandalone()
5280
    self._GoReconnect(False)
5281
    self._WaitUntilSync()
5282

    
5283
    self.feedback_fn("* done")
5284

    
5285
  def Exec(self, feedback_fn):
5286
    """Perform the migration.
5287

5288
    """
5289
    feedback_fn("Migrating instance %s" % self.instance.name)
5290

    
5291
    self.feedback_fn = feedback_fn
5292

    
5293
    self.source_node = self.instance.primary_node
5294
    self.target_node = self.instance.secondary_nodes[0]
5295
    self.all_nodes = [self.source_node, self.target_node]
5296
    self.nodes_ip = {
5297
      self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
5298
      self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
5299
      }
5300

    
5301
    if self.cleanup:
5302
      return self._ExecCleanup()
5303
    else:
5304
      return self._ExecMigration()
5305

    
5306

    
5307
def _CreateBlockDev(lu, node, instance, device, force_create,
5308
                    info, force_open):
5309
  """Create a tree of block devices on a given node.
5310

5311
  If this device type has to be created on secondaries, create it and
5312
  all its children.
5313

5314
  If not, just recurse to children keeping the same 'force' value.
5315

5316
  @param lu: the lu on whose behalf we execute
5317
  @param node: the node on which to create the device
5318
  @type instance: L{objects.Instance}
5319
  @param instance: the instance which owns the device
5320
  @type device: L{objects.Disk}
5321
  @param device: the device to create
5322
  @type force_create: boolean
5323
  @param force_create: whether to force creation of this device; this
5324
      will be change to True whenever we find a device which has
5325
      CreateOnSecondary() attribute
5326
  @param info: the extra 'metadata' we should attach to the device
5327
      (this will be represented as a LVM tag)
5328
  @type force_open: boolean
5329
  @param force_open: this parameter will be passes to the
5330
      L{backend.BlockdevCreate} function where it specifies
5331
      whether we run on primary or not, and it affects both
5332
      the child assembly and the device own Open() execution
5333

5334
  """
5335
  if device.CreateOnSecondary():
5336
    force_create = True
5337

    
5338
  if device.children:
5339
    for child in device.children:
5340
      _CreateBlockDev(lu, node, instance, child, force_create,
5341
                      info, force_open)
5342

    
5343
  if not force_create:
5344
    return
5345

    
5346
  _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
5347

    
5348

    
5349
def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
5350
  """Create a single block device on a given node.
5351

5352
  This will not recurse over children of the device, so they must be
5353
  created in advance.
5354

5355
  @param lu: the lu on whose behalf we execute
5356
  @param node: the node on which to create the device
5357
  @type instance: L{objects.Instance}
5358
  @param instance: the instance which owns the device
5359
  @type device: L{objects.Disk}
5360
  @param device: the device to create
5361
  @param info: the extra 'metadata' we should attach to the device
5362
      (this will be represented as a LVM tag)
5363
  @type force_open: boolean
5364
  @param force_open: this parameter will be passes to the
5365
      L{backend.BlockdevCreate} function where it specifies
5366
      whether we run on primary or not, and it affects both
5367
      the child assembly and the device own Open() execution
5368

5369
  """
5370
  lu.cfg.SetDiskID(device, node)
5371
  result = lu.rpc.call_blockdev_create(node, device, device.size,
5372
                                       instance.name, force_open, info)
5373
  result.Raise("Can't create block device %s on"
5374
               " node %s for instance %s" % (device, node, instance.name))
5375
  if device.physical_id is None:
5376
    device.physical_id = result.payload
5377

    
5378

    
5379
def _GenerateUniqueNames(lu, exts):
5380
  """Generate a suitable LV name.
5381

5382
  This will generate a logical volume name for the given instance.
5383

5384
  """
5385
  results = []
5386
  for val in exts:
5387
    new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
5388
    results.append("%s%s" % (new_id, val))
5389
  return results
5390

    
5391

    
5392
def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
5393
                         p_minor, s_minor):
5394
  """Generate a drbd8 device complete with its children.
5395

5396
  """
5397
  port = lu.cfg.AllocatePort()
5398
  vgname = lu.cfg.GetVGName()
5399
  shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
5400
  dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
5401
                          logical_id=(vgname, names[0]))
5402
  dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
5403
                          logical_id=(vgname, names[1]))
5404
  drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
5405
                          logical_id=(primary, secondary, port,
5406
                                      p_minor, s_minor,
5407
                                      shared_secret),
5408
                          children=[dev_data, dev_meta],
5409
                          iv_name=iv_name)
5410
  return drbd_dev
5411

    
5412

    
5413
def _GenerateDiskTemplate(lu, template_name,
5414
                          instance_name, primary_node,
5415
                          secondary_nodes, disk_info,
5416
                          file_storage_dir, file_driver,
5417
                          base_index):
5418
  """Generate the entire disk layout for a given template type.
5419

5420
  """
5421
  #TODO: compute space requirements
5422

    
5423
  vgname = lu.cfg.GetVGName()
5424
  disk_count = len(disk_info)
5425
  disks = []
5426
  if template_name == constants.DT_DISKLESS:
5427
    pass
5428
  elif template_name == constants.DT_PLAIN:
5429
    if len(secondary_nodes) != 0:
5430
      raise errors.ProgrammerError("Wrong template configuration")
5431

    
5432
    names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
5433
                                      for i in range(disk_count)])
5434
    for idx, disk in enumerate(disk_info):
5435
      disk_index = idx + base_index
5436
      disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
5437
                              logical_id=(vgname, names[idx]),
5438
                              iv_name="disk/%d" % disk_index,
5439
                              mode=disk["mode"])
5440
      disks.append(disk_dev)
5441
  elif template_name == constants.DT_DRBD8:
5442
    if len(secondary_nodes) != 1:
5443
      raise errors.ProgrammerError("Wrong template configuration")
5444
    remote_node = secondary_nodes[0]
5445
    minors = lu.cfg.AllocateDRBDMinor(
5446
      [primary_node, remote_node] * len(disk_info), instance_name)
5447

    
5448
    names = []
5449
    for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
5450
                                               for i in range(disk_count)]):
5451
      names.append(lv_prefix + "_data")
5452
      names.append(lv_prefix + "_meta")
5453
    for idx, disk in enumerate(disk_info):
5454
      disk_index = idx + base_index
5455
      disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
5456
                                      disk["size"], names[idx*2:idx*2+2],
5457
                                      "disk/%d" % disk_index,
5458
                                      minors[idx*2], minors[idx*2+1])
5459
      disk_dev.mode = disk["mode"]
5460
      disks.append(disk_dev)
5461
  elif template_name == constants.DT_FILE:
5462
    if len(secondary_nodes) != 0:
5463
      raise errors.ProgrammerError("Wrong template configuration")
5464

    
5465
    for idx, disk in enumerate(disk_info):
5466
      disk_index = idx + base_index
5467
      disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
5468
                              iv_name="disk/%d" % disk_index,
5469
                              logical_id=(file_driver,
5470
                                          "%s/disk%d" % (file_storage_dir,
5471
                                                         disk_index)),
5472
                              mode=disk["mode"])
5473
      disks.append(disk_dev)
5474
  else:
5475
    raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
5476
  return disks
5477

    
5478

    
5479
def _GetInstanceInfoText(instance):
5480
  """Compute that text that should be added to the disk's metadata.
5481

5482
  """
5483
  return "originstname+%s" % instance.name
5484

    
5485

    
5486
def _CreateDisks(lu, instance, to_skip=None, target_node=None):
5487
  """Create all disks for an instance.
5488

5489
  This abstracts away some work from AddInstance.
5490

5491
  @type lu: L{LogicalUnit}
5492
  @param lu: the logical unit on whose behalf we execute
5493
  @type instance: L{objects.Instance}
5494
  @param instance: the instance whose disks we should create
5495
  @type to_skip: list
5496
  @param to_skip: list of indices to skip
5497
  @type target_node: string
5498
  @param target_node: if passed, overrides the target node for creation
5499
  @rtype: boolean
5500
  @return: the success of the creation
5501

5502
  """
5503
  info = _GetInstanceInfoText(instance)
5504
  if target_node is None:
5505
    pnode = instance.primary_node
5506
    all_nodes = instance.all_nodes
5507
  else:
5508
    pnode = target_node
5509
    all_nodes = [pnode]
5510

    
5511
  if instance.disk_template == constants.DT_FILE:
5512
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
5513
    result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
5514

    
5515
    result.Raise("Failed to create directory '%s' on"
5516
                 " node %s" % (file_storage_dir, pnode))
5517

    
5518
  # Note: this needs to be kept in sync with adding of disks in
5519
  # LUSetInstanceParams
5520
  for idx, device in enumerate(instance.disks):
5521
    if to_skip and idx in to_skip:
5522
      continue
5523
    logging.info("Creating volume %s for instance %s",
5524
                 device.iv_name, instance.name)
5525
    #HARDCODE
5526
    for node in all_nodes:
5527
      f_create = node == pnode
5528
      _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
5529

    
5530

    
5531
def _RemoveDisks(lu, instance, target_node=None):
5532
  """Remove all disks for an instance.
5533

5534
  This abstracts away some work from `AddInstance()` and
5535
  `RemoveInstance()`. Note that in case some of the devices couldn't
5536
  be removed, the removal will continue with the other ones (compare
5537
  with `_CreateDisks()`).
5538

5539
  @type lu: L{LogicalUnit}
5540
  @param lu: the logical unit on whose behalf we execute
5541
  @type instance: L{objects.Instance}
5542
  @param instance: the instance whose disks we should remove
5543
  @type target_node: string
5544
  @param target_node: used to override the node on which to remove the disks
5545
  @rtype: boolean
5546
  @return: the success of the removal
5547

5548
  """
5549
  logging.info("Removing block devices for instance %s", instance.name)
5550

    
5551
  all_result = True
5552
  for device in instance.disks:
5553
    if target_node:
5554
      edata = [(target_node, device)]
5555
    else:
5556
      edata = device.ComputeNodeTree(instance.primary_node)
5557
    for node, disk in edata:
5558
      lu.cfg.SetDiskID(disk, node)
5559
      msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
5560
      if msg:
5561
        lu.LogWarning("Could not remove block device %s on node %s,"
5562
                      " continuing anyway: %s", device.iv_name, node, msg)
5563
        all_result = False
5564

    
5565
  if instance.disk_template == constants.DT_FILE:
5566
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
5567
    if target_node:
5568
      tgt = target_node
5569
    else:
5570
      tgt = instance.primary_node
5571
    result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
5572
    if result.fail_msg:
5573
      lu.LogWarning("Could not remove directory '%s' on node %s: %s",
5574
                    file_storage_dir, instance.primary_node, result.fail_msg)
5575
      all_result = False
5576

    
5577
  return all_result
5578

    
5579

    
5580
def _ComputeDiskSize(disk_template, disks):
5581
  """Compute disk size requirements in the volume group
5582

5583
  """
5584
  # Required free disk space as a function of disk and swap space
5585
  req_size_dict = {
5586
    constants.DT_DISKLESS: None,
5587
    constants.DT_PLAIN: sum(d["size"] for d in disks),
5588
    # 128 MB are added for drbd metadata for each disk
5589
    constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
5590
    constants.DT_FILE: None,
5591
  }
5592

    
5593
  if disk_template not in req_size_dict:
5594
    raise errors.ProgrammerError("Disk template '%s' size requirement"
5595
                                 " is unknown" %  disk_template)
5596

    
5597
  return req_size_dict[disk_template]
5598

    
5599

    
5600
def _CheckHVParams(lu, nodenames, hvname, hvparams):
5601
  """Hypervisor parameter validation.
5602

5603
  This function abstract the hypervisor parameter validation to be
5604
  used in both instance create and instance modify.
5605

5606
  @type lu: L{LogicalUnit}
5607
  @param lu: the logical unit for which we check
5608
  @type nodenames: list
5609
  @param nodenames: the list of nodes on which we should check
5610
  @type hvname: string
5611
  @param hvname: the name of the hypervisor we should use
5612
  @type hvparams: dict
5613
  @param hvparams: the parameters which we need to check
5614
  @raise errors.OpPrereqError: if the parameters are not valid
5615

5616
  """
5617
  hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
5618
                                                  hvname,
5619
                                                  hvparams)
5620
  for node in nodenames:
5621
    info = hvinfo[node]
5622
    if info.offline:
5623
      continue
5624
    info.Raise("Hypervisor parameter validation failed on node %s" % node)
5625

    
5626

    
5627
class LUCreateInstance(LogicalUnit):
5628
  """Create an instance.
5629

5630
  """
5631
  HPATH = "instance-add"
5632
  HTYPE = constants.HTYPE_INSTANCE
5633
  _OP_REQP = ["instance_name", "disks", "disk_template",
5634
              "mode", "start",
5635
              "wait_for_sync", "ip_check", "nics",
5636
              "hvparams", "beparams"]
5637
  REQ_BGL = False
5638

    
5639
  def CheckArguments(self):
5640
    """Check arguments.
5641

5642
    """
5643
    # do not require name_check to ease forward/backward compatibility
5644
    # for tools
5645
    if not hasattr(self.op, "name_check"):
5646
      self.op.name_check = True
5647
    if self.op.ip_check and not self.op.name_check:
5648
      # TODO: make the ip check more flexible and not depend on the name check
5649
      raise errors.OpPrereqError("Cannot do ip checks without a name check",
5650
                                 errors.ECODE_INVAL)
5651

    
5652
  def _ExpandNode(self, node):
5653
    """Expands and checks one node name.
5654

5655
    """
5656
    node_full = self.cfg.ExpandNodeName(node)
5657
    if node_full is None:
5658
      raise errors.OpPrereqError("Unknown node %s" % node, errors.ECODE_NOENT)
5659
    return node_full
5660

    
5661
  def ExpandNames(self):
5662
    """ExpandNames for CreateInstance.
5663

5664
    Figure out the right locks for instance creation.
5665

5666
    """
5667
    self.needed_locks = {}
5668

    
5669
    # set optional parameters to none if they don't exist
5670
    for attr in ["pnode", "snode", "iallocator", "hypervisor"]:
5671
      if not hasattr(self.op, attr):
5672
        setattr(self.op, attr, None)
5673

    
5674
    # cheap checks, mostly valid constants given
5675

    
5676
    # verify creation mode
5677
    if self.op.mode not in (constants.INSTANCE_CREATE,
5678
                            constants.INSTANCE_IMPORT):
5679
      raise errors.OpPrereqError("Invalid instance creation mode '%s'" %
5680
                                 self.op.mode, errors.ECODE_INVAL)
5681

    
5682
    # disk template and mirror node verification
5683
    if self.op.disk_template not in constants.DISK_TEMPLATES:
5684
      raise errors.OpPrereqError("Invalid disk template name",
5685
                                 errors.ECODE_INVAL)
5686

    
5687
    if self.op.hypervisor is None:
5688
      self.op.hypervisor = self.cfg.GetHypervisorType()
5689

    
5690
    cluster = self.cfg.GetClusterInfo()
5691
    enabled_hvs = cluster.enabled_hypervisors
5692
    if self.op.hypervisor not in enabled_hvs:
5693
      raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
5694
                                 " cluster (%s)" % (self.op.hypervisor,
5695
                                  ",".join(enabled_hvs)),
5696
                                 errors.ECODE_STATE)
5697

    
5698
    # check hypervisor parameter syntax (locally)
5699
    utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
5700
    filled_hvp = objects.FillDict(cluster.hvparams[self.op.hypervisor],
5701
                                  self.op.hvparams)
5702
    hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
5703
    hv_type.CheckParameterSyntax(filled_hvp)
5704
    self.hv_full = filled_hvp
5705
    # check that we don't specify global parameters on an instance
5706
    _CheckGlobalHvParams(self.op.hvparams)
5707

    
5708
    # fill and remember the beparams dict
5709
    utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
5710
    self.be_full = objects.FillDict(cluster.beparams[constants.PP_DEFAULT],
5711
                                    self.op.beparams)
5712

    
5713
    #### instance parameters check
5714

    
5715
    # instance name verification
5716
    if self.op.name_check:
5717
      hostname1 = utils.GetHostInfo(self.op.instance_name)
5718
      self.op.instance_name = instance_name = hostname1.name
5719
      # used in CheckPrereq for ip ping check
5720
      self.check_ip = hostname1.ip
5721
    else:
5722
      instance_name = self.op.instance_name
5723
      self.check_ip = None
5724

    
5725
    # this is just a preventive check, but someone might still add this
5726
    # instance in the meantime, and creation will fail at lock-add time
5727
    if instance_name in self.cfg.GetInstanceList():
5728
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
5729
                                 instance_name, errors.ECODE_EXISTS)
5730

    
5731
    self.add_locks[locking.LEVEL_INSTANCE] = instance_name
5732

    
5733
    # NIC buildup
5734
    self.nics = []
5735
    for idx, nic in enumerate(self.op.nics):
5736
      nic_mode_req = nic.get("mode", None)
5737
      nic_mode = nic_mode_req
5738
      if nic_mode is None:
5739
        nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
5740

    
5741
      # in routed mode, for the first nic, the default ip is 'auto'
5742
      if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
5743
        default_ip_mode = constants.VALUE_AUTO
5744
      else:
5745
        default_ip_mode = constants.VALUE_NONE
5746

    
5747
      # ip validity checks
5748
      ip = nic.get("ip", default_ip_mode)
5749
      if ip is None or ip.lower() == constants.VALUE_NONE:
5750
        nic_ip = None
5751
      elif ip.lower() == constants.VALUE_AUTO:
5752
        if not self.op.name_check:
5753
          raise errors.OpPrereqError("IP address set to auto but name checks"
5754
                                     " have been skipped. Aborting.",
5755
                                     errors.ECODE_INVAL)
5756
        nic_ip = hostname1.ip
5757
      else:
5758
        if not utils.IsValidIP(ip):
5759
          raise errors.OpPrereqError("Given IP address '%s' doesn't look"
5760
                                     " like a valid IP" % ip,
5761
                                     errors.ECODE_INVAL)
5762
        nic_ip = ip
5763

    
5764
      # TODO: check the ip address for uniqueness
5765
      if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
5766
        raise errors.OpPrereqError("Routed nic mode requires an ip address",
5767
                                   errors.ECODE_INVAL)
5768

    
5769
      # MAC address verification
5770
      mac = nic.get("mac", constants.VALUE_AUTO)
5771
      if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
5772
        if not utils.IsValidMac(mac.lower()):
5773
          raise errors.OpPrereqError("Invalid MAC address specified: %s" %
5774
                                     mac, errors.ECODE_INVAL)
5775
        else:
5776
          try:
5777
            self.cfg.ReserveMAC(mac, self.proc.GetECId())
5778
          except errors.ReservationError:
5779
            raise errors.OpPrereqError("MAC address %s already in use"
5780
                                       " in cluster" % mac,
5781
                                       errors.ECODE_NOTUNIQUE)
5782

    
5783
      # bridge verification
5784
      bridge = nic.get("bridge", None)
5785
      link = nic.get("link", None)
5786
      if bridge and link:
5787
        raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
5788
                                   " at the same time", errors.ECODE_INVAL)
5789
      elif bridge and nic_mode == constants.NIC_MODE_ROUTED:
5790
        raise errors.OpPrereqError("Cannot pass 'bridge' on a routed nic",
5791
                                   errors.ECODE_INVAL)
5792
      elif bridge:
5793
        link = bridge
5794

    
5795
      nicparams = {}
5796
      if nic_mode_req:
5797
        nicparams[constants.NIC_MODE] = nic_mode_req
5798
      if link:
5799
        nicparams[constants.NIC_LINK] = link
5800

    
5801
      check_params = objects.FillDict(cluster.nicparams[constants.PP_DEFAULT],
5802
                                      nicparams)
5803
      objects.NIC.CheckParameterSyntax(check_params)
5804
      self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
5805

    
5806
    # disk checks/pre-build
5807
    self.disks = []
5808
    for disk in self.op.disks:
5809
      mode = disk.get("mode", constants.DISK_RDWR)
5810
      if mode not in constants.DISK_ACCESS_SET:
5811
        raise errors.OpPrereqError("Invalid disk access mode '%s'" %
5812
                                   mode, errors.ECODE_INVAL)
5813
      size = disk.get("size", None)
5814
      if size is None:
5815
        raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
5816
      try:
5817
        size = int(size)
5818
      except ValueError:
5819
        raise errors.OpPrereqError("Invalid disk size '%s'" % size,
5820
                                   errors.ECODE_INVAL)
5821
      self.disks.append({"size": size, "mode": mode})
5822

    
5823
    # file storage checks
5824
    if (self.op.file_driver and
5825
        not self.op.file_driver in constants.FILE_DRIVER):
5826
      raise errors.OpPrereqError("Invalid file driver name '%s'" %
5827
                                 self.op.file_driver, errors.ECODE_INVAL)
5828

    
5829
    if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
5830
      raise errors.OpPrereqError("File storage directory path not absolute",
5831
                                 errors.ECODE_INVAL)
5832

    
5833
    ### Node/iallocator related checks
5834
    if [self.op.iallocator, self.op.pnode].count(None) != 1:
5835
      raise errors.OpPrereqError("One and only one of iallocator and primary"
5836
                                 " node must be given",
5837
                                 errors.ECODE_INVAL)
5838

    
5839
    if self.op.iallocator:
5840
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5841
    else:
5842
      self.op.pnode = self._ExpandNode(self.op.pnode)
5843
      nodelist = [self.op.pnode]
5844
      if self.op.snode is not None:
5845
        self.op.snode = self._ExpandNode(self.op.snode)
5846
        nodelist.append(self.op.snode)
5847
      self.needed_locks[locking.LEVEL_NODE] = nodelist
5848

    
5849
    # in case of import lock the source node too
5850
    if self.op.mode == constants.INSTANCE_IMPORT:
5851
      src_node = getattr(self.op, "src_node", None)
5852
      src_path = getattr(self.op, "src_path", None)
5853

    
5854
      if src_path is None:
5855
        self.op.src_path = src_path = self.op.instance_name
5856

    
5857
      if src_node is None:
5858
        self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5859
        self.op.src_node = None
5860
        if os.path.isabs(src_path):
5861
          raise errors.OpPrereqError("Importing an instance from an absolute"
5862
                                     " path requires a source node option.",
5863
                                     errors.ECODE_INVAL)
5864
      else:
5865
        self.op.src_node = src_node = self._ExpandNode(src_node)
5866
        if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
5867
          self.needed_locks[locking.LEVEL_NODE].append(src_node)
5868
        if not os.path.isabs(src_path):
5869
          self.op.src_path = src_path = \
5870
            os.path.join(constants.EXPORT_DIR, src_path)
5871

    
5872
      # On import force_variant must be True, because if we forced it at
5873
      # initial install, our only chance when importing it back is that it
5874
      # works again!
5875
      self.op.force_variant = True
5876

    
5877
    else: # INSTANCE_CREATE
5878
      if getattr(self.op, "os_type", None) is None:
5879
        raise errors.OpPrereqError("No guest OS specified",
5880
                                   errors.ECODE_INVAL)
5881
      self.op.force_variant = getattr(self.op, "force_variant", False)
5882

    
5883
  def _RunAllocator(self):
5884
    """Run the allocator based on input opcode.
5885

5886
    """
5887
    nics = [n.ToDict() for n in self.nics]
5888
    ial = IAllocator(self.cfg, self.rpc,
5889
                     mode=constants.IALLOCATOR_MODE_ALLOC,
5890
                     name=self.op.instance_name,
5891
                     disk_template=self.op.disk_template,
5892
                     tags=[],
5893
                     os=self.op.os_type,
5894
                     vcpus=self.be_full[constants.BE_VCPUS],
5895
                     mem_size=self.be_full[constants.BE_MEMORY],
5896
                     disks=self.disks,
5897
                     nics=nics,
5898
                     hypervisor=self.op.hypervisor,
5899
                     )
5900

    
5901
    ial.Run(self.op.iallocator)
5902

    
5903
    if not ial.success:
5904
      raise errors.OpPrereqError("Can't compute nodes using"
5905
                                 " iallocator '%s': %s" %
5906
                                 (self.op.iallocator, ial.info),
5907
                                 errors.ECODE_NORES)
5908
    if len(ial.nodes) != ial.required_nodes:
5909
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
5910
                                 " of nodes (%s), required %s" %
5911
                                 (self.op.iallocator, len(ial.nodes),
5912
                                  ial.required_nodes), errors.ECODE_FAULT)
5913
    self.op.pnode = ial.nodes[0]
5914
    self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
5915
                 self.op.instance_name, self.op.iallocator,
5916
                 utils.CommaJoin(ial.nodes))
5917
    if ial.required_nodes == 2:
5918
      self.op.snode = ial.nodes[1]
5919

    
5920
  def BuildHooksEnv(self):
5921
    """Build hooks env.
5922

5923
    This runs on master, primary and secondary nodes of the instance.
5924

5925
    """
5926
    env = {
5927
      "ADD_MODE": self.op.mode,
5928
      }
5929
    if self.op.mode == constants.INSTANCE_IMPORT:
5930
      env["SRC_NODE"] = self.op.src_node
5931
      env["SRC_PATH"] = self.op.src_path
5932
      env["SRC_IMAGES"] = self.src_images
5933

    
5934
    env.update(_BuildInstanceHookEnv(
5935
      name=self.op.instance_name,
5936
      primary_node=self.op.pnode,
5937
      secondary_nodes=self.secondaries,
5938
      status=self.op.start,
5939
      os_type=self.op.os_type,
5940
      memory=self.be_full[constants.BE_MEMORY],
5941
      vcpus=self.be_full[constants.BE_VCPUS],
5942
      nics=_NICListToTuple(self, self.nics),
5943
      disk_template=self.op.disk_template,
5944
      disks=[(d["size"], d["mode"]) for d in self.disks],
5945
      bep=self.be_full,
5946
      hvp=self.hv_full,
5947
      hypervisor_name=self.op.hypervisor,
5948
    ))
5949

    
5950
    nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
5951
          self.secondaries)
5952
    return env, nl, nl
5953

    
5954

    
5955
  def CheckPrereq(self):
5956
    """Check prerequisites.
5957

5958
    """
5959
    if (not self.cfg.GetVGName() and
5960
        self.op.disk_template not in constants.DTS_NOT_LVM):
5961
      raise errors.OpPrereqError("Cluster does not support lvm-based"
5962
                                 " instances", errors.ECODE_STATE)
5963

    
5964
    if self.op.mode == constants.INSTANCE_IMPORT:
5965
      src_node = self.op.src_node
5966
      src_path = self.op.src_path
5967

    
5968
      if src_node is None:
5969
        locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
5970
        exp_list = self.rpc.call_export_list(locked_nodes)
5971
        found = False
5972
        for node in exp_list:
5973
          if exp_list[node].fail_msg:
5974
            continue
5975
          if src_path in exp_list[node].payload:
5976
            found = True
5977
            self.op.src_node = src_node = node
5978
            self.op.src_path = src_path = os.path.join(constants.EXPORT_DIR,
5979
                                                       src_path)
5980
            break
5981
        if not found:
5982
          raise errors.OpPrereqError("No export found for relative path %s" %
5983
                                      src_path, errors.ECODE_INVAL)
5984

    
5985
      _CheckNodeOnline(self, src_node)
5986
      result = self.rpc.call_export_info(src_node, src_path)
5987
      result.Raise("No export or invalid export found in dir %s" % src_path)
5988

    
5989
      export_info = objects.SerializableConfigParser.Loads(str(result.payload))
5990
      if not export_info.has_section(constants.INISECT_EXP):
5991
        raise errors.ProgrammerError("Corrupted export config",
5992
                                     errors.ECODE_ENVIRON)
5993

    
5994
      ei_version = export_info.get(constants.INISECT_EXP, 'version')
5995
      if (int(ei_version) != constants.EXPORT_VERSION):
5996
        raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
5997
                                   (ei_version, constants.EXPORT_VERSION),
5998
                                   errors.ECODE_ENVIRON)
5999

    
6000
      # Check that the new instance doesn't have less disks than the export
6001
      instance_disks = len(self.disks)
6002
      export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
6003
      if instance_disks < export_disks:
6004
        raise errors.OpPrereqError("Not enough disks to import."
6005
                                   " (instance: %d, export: %d)" %
6006
                                   (instance_disks, export_disks),
6007
                                   errors.ECODE_INVAL)
6008

    
6009
      self.op.os_type = export_info.get(constants.INISECT_EXP, 'os')
6010
      disk_images = []
6011
      for idx in range(export_disks):
6012
        option = 'disk%d_dump' % idx
6013
        if export_info.has_option(constants.INISECT_INS, option):
6014
          # FIXME: are the old os-es, disk sizes, etc. useful?
6015
          export_name = export_info.get(constants.INISECT_INS, option)
6016
          image = os.path.join(src_path, export_name)
6017
          disk_images.append(image)
6018
        else:
6019
          disk_images.append(False)
6020

    
6021
      self.src_images = disk_images
6022

    
6023
      old_name = export_info.get(constants.INISECT_INS, 'name')
6024
      # FIXME: int() here could throw a ValueError on broken exports
6025
      exp_nic_count = int(export_info.get(constants.INISECT_INS, 'nic_count'))
6026
      if self.op.instance_name == old_name:
6027
        for idx, nic in enumerate(self.nics):
6028
          if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
6029
            nic_mac_ini = 'nic%d_mac' % idx
6030
            nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
6031

    
6032
    # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
6033

    
6034
    # ip ping checks (we use the same ip that was resolved in ExpandNames)
6035
    if self.op.ip_check:
6036
      if utils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
6037
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
6038
                                   (self.check_ip, self.op.instance_name),
6039
                                   errors.ECODE_NOTUNIQUE)
6040

    
6041
    #### mac address generation
6042
    # By generating here the mac address both the allocator and the hooks get
6043
    # the real final mac address rather than the 'auto' or 'generate' value.
6044
    # There is a race condition between the generation and the instance object
6045
    # creation, which means that we know the mac is valid now, but we're not
6046
    # sure it will be when we actually add the instance. If things go bad
6047
    # adding the instance will abort because of a duplicate mac, and the
6048
    # creation job will fail.
6049
    for nic in self.nics:
6050
      if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
6051
        nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
6052

    
6053
    #### allocator run
6054

    
6055
    if self.op.iallocator is not None:
6056
      self._RunAllocator()
6057

    
6058
    #### node related checks
6059

    
6060
    # check primary node
6061
    self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
6062
    assert self.pnode is not None, \
6063
      "Cannot retrieve locked node %s" % self.op.pnode
6064
    if pnode.offline:
6065
      raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
6066
                                 pnode.name, errors.ECODE_STATE)
6067
    if pnode.drained:
6068
      raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
6069
                                 pnode.name, errors.ECODE_STATE)
6070

    
6071
    self.secondaries = []
6072

    
6073
    # mirror node verification
6074
    if self.op.disk_template in constants.DTS_NET_MIRROR:
6075
      if self.op.snode is None:
6076
        raise errors.OpPrereqError("The networked disk templates need"
6077
                                   " a mirror node", errors.ECODE_INVAL)
6078
      if self.op.snode == pnode.name:
6079
        raise errors.OpPrereqError("The secondary node cannot be the"
6080
                                   " primary node.", errors.ECODE_INVAL)
6081
      _CheckNodeOnline(self, self.op.snode)
6082
      _CheckNodeNotDrained(self, self.op.snode)
6083
      self.secondaries.append(self.op.snode)
6084

    
6085
    nodenames = [pnode.name] + self.secondaries
6086

    
6087
    req_size = _ComputeDiskSize(self.op.disk_template,
6088
                                self.disks)
6089

    
6090
    # Check lv size requirements
6091
    if req_size is not None:
6092
      nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
6093
                                         self.op.hypervisor)
6094
      for node in nodenames:
6095
        info = nodeinfo[node]
6096
        info.Raise("Cannot get current information from node %s" % node)
6097
        info = info.payload
6098
        vg_free = info.get('vg_free', None)
6099
        if not isinstance(vg_free, int):
6100
          raise errors.OpPrereqError("Can't compute free disk space on"
6101
                                     " node %s" % node, errors.ECODE_ENVIRON)
6102
        if req_size > vg_free:
6103
          raise errors.OpPrereqError("Not enough disk space on target node %s."
6104
                                     " %d MB available, %d MB required" %
6105
                                     (node, vg_free, req_size),
6106
                                     errors.ECODE_NORES)
6107

    
6108
    _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
6109

    
6110
    # os verification
6111
    result = self.rpc.call_os_get(pnode.name, self.op.os_type)
6112
    result.Raise("OS '%s' not in supported os list for primary node %s" %
6113
                 (self.op.os_type, pnode.name),
6114
                 prereq=True, ecode=errors.ECODE_INVAL)
6115
    if not self.op.force_variant:
6116
      _CheckOSVariant(result.payload, self.op.os_type)
6117

    
6118
    _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
6119

    
6120
    # memory check on primary node
6121
    if self.op.start:
6122
      _CheckNodeFreeMemory(self, self.pnode.name,
6123
                           "creating instance %s" % self.op.instance_name,
6124
                           self.be_full[constants.BE_MEMORY],
6125
                           self.op.hypervisor)
6126

    
6127
    self.dry_run_result = list(nodenames)
6128

    
6129
  def Exec(self, feedback_fn):
6130
    """Create and add the instance to the cluster.
6131

6132
    """
6133
    instance = self.op.instance_name
6134
    pnode_name = self.pnode.name
6135

    
6136
    ht_kind = self.op.hypervisor
6137
    if ht_kind in constants.HTS_REQ_PORT:
6138
      network_port = self.cfg.AllocatePort()
6139
    else:
6140
      network_port = None
6141

    
6142
    ##if self.op.vnc_bind_address is None:
6143
    ##  self.op.vnc_bind_address = constants.VNC_DEFAULT_BIND_ADDRESS
6144

    
6145
    # this is needed because os.path.join does not accept None arguments
6146
    if self.op.file_storage_dir is None:
6147
      string_file_storage_dir = ""
6148
    else:
6149
      string_file_storage_dir = self.op.file_storage_dir
6150

    
6151
    # build the full file storage dir path
6152
    file_storage_dir = os.path.normpath(os.path.join(
6153
                                        self.cfg.GetFileStorageDir(),
6154
                                        string_file_storage_dir, instance))
6155

    
6156

    
6157
    disks = _GenerateDiskTemplate(self,
6158
                                  self.op.disk_template,
6159
                                  instance, pnode_name,
6160
                                  self.secondaries,
6161
                                  self.disks,
6162
                                  file_storage_dir,
6163
                                  self.op.file_driver,
6164
                                  0)
6165

    
6166
    iobj = objects.Instance(name=instance, os=self.op.os_type,
6167
                            primary_node=pnode_name,
6168
                            nics=self.nics, disks=disks,
6169
                            disk_template=self.op.disk_template,
6170
                            admin_up=False,
6171
                            network_port=network_port,
6172
                            beparams=self.op.beparams,
6173
                            hvparams=self.op.hvparams,
6174
                            hypervisor=self.op.hypervisor,
6175
                            )
6176

    
6177
    feedback_fn("* creating instance disks...")
6178
    try:
6179
      _CreateDisks(self, iobj)
6180
    except errors.OpExecError:
6181
      self.LogWarning("Device creation failed, reverting...")
6182
      try:
6183
        _RemoveDisks(self, iobj)
6184
      finally:
6185
        self.cfg.ReleaseDRBDMinors(instance)
6186
        raise
6187

    
6188
    feedback_fn("adding instance %s to cluster config" % instance)
6189

    
6190
    self.cfg.AddInstance(iobj, self.proc.GetECId())
6191

    
6192
    # Declare that we don't want to remove the instance lock anymore, as we've
6193
    # added the instance to the config
6194
    del self.remove_locks[locking.LEVEL_INSTANCE]
6195
    # Unlock all the nodes
6196
    if self.op.mode == constants.INSTANCE_IMPORT:
6197
      nodes_keep = [self.op.src_node]
6198
      nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
6199
                       if node != self.op.src_node]
6200
      self.context.glm.release(locking.LEVEL_NODE, nodes_release)
6201
      self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
6202
    else:
6203
      self.context.glm.release(locking.LEVEL_NODE)
6204
      del self.acquired_locks[locking.LEVEL_NODE]
6205

    
6206
    if self.op.wait_for_sync:
6207
      disk_abort = not _WaitForSync(self, iobj)
6208
    elif iobj.disk_template in constants.DTS_NET_MIRROR:
6209
      # make sure the disks are not degraded (still sync-ing is ok)
6210
      time.sleep(15)
6211
      feedback_fn("* checking mirrors status")
6212
      disk_abort = not _WaitForSync(self, iobj, oneshot=True)
6213
    else:
6214
      disk_abort = False
6215

    
6216
    if disk_abort:
6217
      _RemoveDisks(self, iobj)
6218
      self.cfg.RemoveInstance(iobj.name)
6219
      # Make sure the instance lock gets removed
6220
      self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
6221
      raise errors.OpExecError("There are some degraded disks for"
6222
                               " this instance")
6223

    
6224
    feedback_fn("creating os for instance %s on node %s" %
6225
                (instance, pnode_name))
6226

    
6227
    if iobj.disk_template != constants.DT_DISKLESS:
6228
      if self.op.mode == constants.INSTANCE_CREATE:
6229
        feedback_fn("* running the instance OS create scripts...")
6230
        result = self.rpc.call_instance_os_add(pnode_name, iobj, False)
6231
        result.Raise("Could not add os for instance %s"
6232
                     " on node %s" % (instance, pnode_name))
6233

    
6234
      elif self.op.mode == constants.INSTANCE_IMPORT:
6235
        feedback_fn("* running the instance OS import scripts...")
6236
        src_node = self.op.src_node
6237
        src_images = self.src_images
6238
        cluster_name = self.cfg.GetClusterName()
6239
        import_result = self.rpc.call_instance_os_import(pnode_name, iobj,
6240
                                                         src_node, src_images,
6241
                                                         cluster_name)
6242
        msg = import_result.fail_msg
6243
        if msg:
6244
          self.LogWarning("Error while importing the disk images for instance"
6245
                          " %s on node %s: %s" % (instance, pnode_name, msg))
6246
      else:
6247
        # also checked in the prereq part
6248
        raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
6249
                                     % self.op.mode)
6250

    
6251
    if self.op.start:
6252
      iobj.admin_up = True
6253
      self.cfg.Update(iobj, feedback_fn)
6254
      logging.info("Starting instance %s on node %s", instance, pnode_name)
6255
      feedback_fn("* starting instance...")
6256
      result = self.rpc.call_instance_start(pnode_name, iobj, None, None)
6257
      result.Raise("Could not start instance")
6258

    
6259
    return list(iobj.all_nodes)
6260

    
6261

    
6262
class LUConnectConsole(NoHooksLU):
6263
  """Connect to an instance's console.
6264

6265
  This is somewhat special in that it returns the command line that
6266
  you need to run on the master node in order to connect to the
6267
  console.
6268

6269
  """
6270
  _OP_REQP = ["instance_name"]
6271
  REQ_BGL = False
6272

    
6273
  def ExpandNames(self):
6274
    self._ExpandAndLockInstance()
6275

    
6276
  def CheckPrereq(self):
6277
    """Check prerequisites.
6278

6279
    This checks that the instance is in the cluster.
6280

6281
    """
6282
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6283
    assert self.instance is not None, \
6284
      "Cannot retrieve locked instance %s" % self.op.instance_name
6285
    _CheckNodeOnline(self, self.instance.primary_node)
6286

    
6287
  def Exec(self, feedback_fn):
6288
    """Connect to the console of an instance
6289

6290
    """
6291
    instance = self.instance
6292
    node = instance.primary_node
6293

    
6294
    node_insts = self.rpc.call_instance_list([node],
6295
                                             [instance.hypervisor])[node]
6296
    node_insts.Raise("Can't get node information from %s" % node)
6297

    
6298
    if instance.name not in node_insts.payload:
6299
      raise errors.OpExecError("Instance %s is not running." % instance.name)
6300

    
6301
    logging.debug("Connecting to console of %s on %s", instance.name, node)
6302

    
6303
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
6304
    cluster = self.cfg.GetClusterInfo()
6305
    # beparams and hvparams are passed separately, to avoid editing the
6306
    # instance and then saving the defaults in the instance itself.
6307
    hvparams = cluster.FillHV(instance)
6308
    beparams = cluster.FillBE(instance)
6309
    console_cmd = hyper.GetShellCommandForConsole(instance, hvparams, beparams)
6310

    
6311
    # build ssh cmdline
6312
    return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
6313

    
6314

    
6315
class LUReplaceDisks(LogicalUnit):
6316
  """Replace the disks of an instance.
6317

6318
  """
6319
  HPATH = "mirrors-replace"
6320
  HTYPE = constants.HTYPE_INSTANCE
6321
  _OP_REQP = ["instance_name", "mode", "disks"]
6322
  REQ_BGL = False
6323

    
6324
  def CheckArguments(self):
6325
    if not hasattr(self.op, "remote_node"):
6326
      self.op.remote_node = None
6327
    if not hasattr(self.op, "iallocator"):
6328
      self.op.iallocator = None
6329

    
6330
    TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
6331
                                  self.op.iallocator)
6332

    
6333
  def ExpandNames(self):
6334
    self._ExpandAndLockInstance()
6335

    
6336
    if self.op.iallocator is not None:
6337
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6338

    
6339
    elif self.op.remote_node is not None:
6340
      remote_node = self.cfg.ExpandNodeName(self.op.remote_node)
6341
      if remote_node is None:
6342
        raise errors.OpPrereqError("Node '%s' not known" %
6343
                                   self.op.remote_node, errors.ECODE_NOENT)
6344

    
6345
      self.op.remote_node = remote_node
6346

    
6347
      # Warning: do not remove the locking of the new secondary here
6348
      # unless DRBD8.AddChildren is changed to work in parallel;
6349
      # currently it doesn't since parallel invocations of
6350
      # FindUnusedMinor will conflict
6351
      self.needed_locks[locking.LEVEL_NODE] = [remote_node]
6352
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
6353

    
6354
    else:
6355
      self.needed_locks[locking.LEVEL_NODE] = []
6356
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
6357

    
6358
    self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
6359
                                   self.op.iallocator, self.op.remote_node,
6360
                                   self.op.disks)
6361

    
6362
    self.tasklets = [self.replacer]
6363

    
6364
  def DeclareLocks(self, level):
6365
    # If we're not already locking all nodes in the set we have to declare the
6366
    # instance's primary/secondary nodes.
6367
    if (level == locking.LEVEL_NODE and
6368
        self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
6369
      self._LockInstancesNodes()
6370

    
6371
  def BuildHooksEnv(self):
6372
    """Build hooks env.
6373

6374
    This runs on the master, the primary and all the secondaries.
6375

6376
    """
6377
    instance = self.replacer.instance
6378
    env = {
6379
      "MODE": self.op.mode,
6380
      "NEW_SECONDARY": self.op.remote_node,
6381
      "OLD_SECONDARY": instance.secondary_nodes[0],
6382
      }
6383
    env.update(_BuildInstanceHookEnvByObject(self, instance))
6384
    nl = [
6385
      self.cfg.GetMasterNode(),
6386
      instance.primary_node,
6387
      ]
6388
    if self.op.remote_node is not None:
6389
      nl.append(self.op.remote_node)
6390
    return env, nl, nl
6391

    
6392

    
6393
class LUEvacuateNode(LogicalUnit):
6394
  """Relocate the secondary instances from a node.
6395

6396
  """
6397
  HPATH = "node-evacuate"
6398
  HTYPE = constants.HTYPE_NODE
6399
  _OP_REQP = ["node_name"]
6400
  REQ_BGL = False
6401

    
6402
  def CheckArguments(self):
6403
    if not hasattr(self.op, "remote_node"):
6404
      self.op.remote_node = None
6405
    if not hasattr(self.op, "iallocator"):
6406
      self.op.iallocator = None
6407

    
6408
    TLReplaceDisks.CheckArguments(constants.REPLACE_DISK_CHG,
6409
                                  self.op.remote_node,
6410
                                  self.op.iallocator)
6411

    
6412
  def ExpandNames(self):
6413
    self.op.node_name = self.cfg.ExpandNodeName(self.op.node_name)
6414
    if self.op.node_name is None:
6415
      raise errors.OpPrereqError("Node '%s' not known" % self.op.node_name,
6416
                                 errors.ECODE_NOENT)
6417

    
6418
    self.needed_locks = {}
6419

    
6420
    # Declare node locks
6421
    if self.op.iallocator is not None:
6422
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6423

    
6424
    elif self.op.remote_node is not None:
6425
      remote_node = self.cfg.ExpandNodeName(self.op.remote_node)
6426
      if remote_node is None:
6427
        raise errors.OpPrereqError("Node '%s' not known" %
6428
                                   self.op.remote_node, errors.ECODE_NOENT)
6429

    
6430
      self.op.remote_node = remote_node
6431

    
6432
      # Warning: do not remove the locking of the new secondary here
6433
      # unless DRBD8.AddChildren is changed to work in parallel;
6434
      # currently it doesn't since parallel invocations of
6435
      # FindUnusedMinor will conflict
6436
      self.needed_locks[locking.LEVEL_NODE] = [remote_node]
6437
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
6438

    
6439
    else:
6440
      raise errors.OpPrereqError("Invalid parameters", errors.ECODE_INVAL)
6441

    
6442
    # Create tasklets for replacing disks for all secondary instances on this
6443
    # node
6444
    names = []
6445
    tasklets = []
6446

    
6447
    for inst in _GetNodeSecondaryInstances(self.cfg, self.op.node_name):
6448
      logging.debug("Replacing disks for instance %s", inst.name)
6449
      names.append(inst.name)
6450

    
6451
      replacer = TLReplaceDisks(self, inst.name, constants.REPLACE_DISK_CHG,
6452
                                self.op.iallocator, self.op.remote_node, [])
6453
      tasklets.append(replacer)
6454

    
6455
    self.tasklets = tasklets
6456
    self.instance_names = names
6457

    
6458
    # Declare instance locks
6459
    self.needed_locks[locking.LEVEL_INSTANCE] = self.instance_names
6460

    
6461
  def DeclareLocks(self, level):
6462
    # If we're not already locking all nodes in the set we have to declare the
6463
    # instance's primary/secondary nodes.
6464
    if (level == locking.LEVEL_NODE and
6465
        self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
6466
      self._LockInstancesNodes()
6467

    
6468
  def BuildHooksEnv(self):
6469
    """Build hooks env.
6470

6471
    This runs on the master, the primary and all the secondaries.
6472

6473
    """
6474
    env = {
6475
      "NODE_NAME": self.op.node_name,
6476
      }
6477

    
6478
    nl = [self.cfg.GetMasterNode()]
6479

    
6480
    if self.op.remote_node is not None:
6481
      env["NEW_SECONDARY"] = self.op.remote_node
6482
      nl.append(self.op.remote_node)
6483

    
6484
    return (env, nl, nl)
6485

    
6486

    
6487
class TLReplaceDisks(Tasklet):
6488
  """Replaces disks for an instance.
6489

6490
  Note: Locking is not within the scope of this class.
6491

6492
  """
6493
  def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
6494
               disks):
6495
    """Initializes this class.
6496

6497
    """
6498
    Tasklet.__init__(self, lu)
6499

    
6500
    # Parameters
6501
    self.instance_name = instance_name
6502
    self.mode = mode
6503
    self.iallocator_name = iallocator_name
6504
    self.remote_node = remote_node
6505
    self.disks = disks
6506

    
6507
    # Runtime data
6508
    self.instance = None
6509
    self.new_node = None
6510
    self.target_node = None
6511
    self.other_node = None
6512
    self.remote_node_info = None
6513
    self.node_secondary_ip = None
6514

    
6515
  @staticmethod
6516
  def CheckArguments(mode, remote_node, iallocator):
6517
    """Helper function for users of this class.
6518

6519
    """
6520
    # check for valid parameter combination
6521
    if mode == constants.REPLACE_DISK_CHG:
6522
      if remote_node is None and iallocator is None:
6523
        raise errors.OpPrereqError("When changing the secondary either an"
6524
                                   " iallocator script must be used or the"
6525
                                   " new node given", errors.ECODE_INVAL)
6526

    
6527
      if remote_node is not None and iallocator is not None:
6528
        raise errors.OpPrereqError("Give either the iallocator or the new"
6529
                                   " secondary, not both", errors.ECODE_INVAL)
6530

    
6531
    elif remote_node is not None or iallocator is not None:
6532
      # Not replacing the secondary
6533
      raise errors.OpPrereqError("The iallocator and new node options can"
6534
                                 " only be used when changing the"
6535
                                 " secondary node", errors.ECODE_INVAL)
6536

    
6537
  @staticmethod
6538
  def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
6539
    """Compute a new secondary node using an IAllocator.
6540

6541
    """
6542
    ial = IAllocator(lu.cfg, lu.rpc,
6543
                     mode=constants.IALLOCATOR_MODE_RELOC,
6544
                     name=instance_name,
6545
                     relocate_from=relocate_from)
6546

    
6547
    ial.Run(iallocator_name)
6548

    
6549
    if not ial.success:
6550
      raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
6551
                                 " %s" % (iallocator_name, ial.info),
6552
                                 errors.ECODE_NORES)
6553

    
6554
    if len(ial.nodes) != ial.required_nodes:
6555
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
6556
                                 " of nodes (%s), required %s" %
6557
                                 (iallocator_name,
6558
                                  len(ial.nodes), ial.required_nodes),
6559
                                 errors.ECODE_FAULT)
6560

    
6561
    remote_node_name = ial.nodes[0]
6562

    
6563
    lu.LogInfo("Selected new secondary for instance '%s': %s",
6564
               instance_name, remote_node_name)
6565

    
6566
    return remote_node_name
6567

    
6568
  def _FindFaultyDisks(self, node_name):
6569
    return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
6570
                                    node_name, True)
6571

    
6572
  def CheckPrereq(self):
6573
    """Check prerequisites.
6574

6575
    This checks that the instance is in the cluster.
6576

6577
    """
6578
    self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
6579
    assert instance is not None, \
6580
      "Cannot retrieve locked instance %s" % self.instance_name
6581

    
6582
    if instance.disk_template != constants.DT_DRBD8:
6583
      raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
6584
                                 " instances", errors.ECODE_INVAL)
6585

    
6586
    if len(instance.secondary_nodes) != 1:
6587
      raise errors.OpPrereqError("The instance has a strange layout,"
6588
                                 " expected one secondary but found %d" %
6589
                                 len(instance.secondary_nodes),
6590
                                 errors.ECODE_FAULT)
6591

    
6592
    secondary_node = instance.secondary_nodes[0]
6593

    
6594
    if self.iallocator_name is None:
6595
      remote_node = self.remote_node
6596
    else:
6597
      remote_node = self._RunAllocator(self.lu, self.iallocator_name,
6598
                                       instance.name, instance.secondary_nodes)
6599

    
6600
    if remote_node is not None:
6601
      self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
6602
      assert self.remote_node_info is not None, \
6603
        "Cannot retrieve locked node %s" % remote_node
6604
    else:
6605
      self.remote_node_info = None
6606

    
6607
    if remote_node == self.instance.primary_node:
6608
      raise errors.OpPrereqError("The specified node is the primary node of"
6609
                                 " the instance.", errors.ECODE_INVAL)
6610

    
6611
    if remote_node == secondary_node:
6612
      raise errors.OpPrereqError("The specified node is already the"
6613
                                 " secondary node of the instance.",
6614
                                 errors.ECODE_INVAL)
6615

    
6616
    if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
6617
                                    constants.REPLACE_DISK_CHG):
6618
      raise errors.OpPrereqError("Cannot specify disks to be replaced",
6619
                                 errors.ECODE_INVAL)
6620

    
6621
    if self.mode == constants.REPLACE_DISK_AUTO:
6622
      faulty_primary = self._FindFaultyDisks(instance.primary_node)
6623
      faulty_secondary = self._FindFaultyDisks(secondary_node)
6624

    
6625
      if faulty_primary and faulty_secondary:
6626
        raise errors.OpPrereqError("Instance %s has faulty disks on more than"
6627
                                   " one node and can not be repaired"
6628
                                   " automatically" % self.instance_name,
6629
                                   errors.ECODE_STATE)
6630

    
6631
      if faulty_primary:
6632
        self.disks = faulty_primary
6633
        self.target_node = instance.primary_node
6634
        self.other_node = secondary_node
6635
        check_nodes = [self.target_node, self.other_node]
6636
      elif faulty_secondary:
6637
        self.disks = faulty_secondary
6638
        self.target_node = secondary_node
6639
        self.other_node = instance.primary_node
6640
        check_nodes = [self.target_node, self.other_node]
6641
      else:
6642
        self.disks = []
6643
        check_nodes = []
6644

    
6645
    else:
6646
      # Non-automatic modes
6647
      if self.mode == constants.REPLACE_DISK_PRI:
6648
        self.target_node = instance.primary_node
6649
        self.other_node = secondary_node
6650
        check_nodes = [self.target_node, self.other_node]
6651

    
6652
      elif self.mode == constants.REPLACE_DISK_SEC:
6653
        self.target_node = secondary_node
6654
        self.other_node = instance.primary_node
6655
        check_nodes = [self.target_node, self.other_node]
6656

    
6657
      elif self.mode == constants.REPLACE_DISK_CHG:
6658
        self.new_node = remote_node
6659
        self.other_node = instance.primary_node
6660
        self.target_node = secondary_node
6661
        check_nodes = [self.new_node, self.other_node]
6662

    
6663
        _CheckNodeNotDrained(self.lu, remote_node)
6664

    
6665
      else:
6666
        raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
6667
                                     self.mode)
6668

    
6669
      # If not specified all disks should be replaced
6670
      if not self.disks:
6671
        self.disks = range(len(self.instance.disks))
6672

    
6673
    for node in check_nodes:
6674
      _CheckNodeOnline(self.lu, node)
6675

    
6676
    # Check whether disks are valid
6677
    for disk_idx in self.disks:
6678
      instance.FindDisk(disk_idx)
6679

    
6680
    # Get secondary node IP addresses
6681
    node_2nd_ip = {}
6682

    
6683
    for node_name in [self.target_node, self.other_node, self.new_node]:
6684
      if node_name is not None:
6685
        node_2nd_ip[node_name] = self.cfg.GetNodeInfo(node_name).secondary_ip
6686

    
6687
    self.node_secondary_ip = node_2nd_ip
6688

    
6689
  def Exec(self, feedback_fn):
6690
    """Execute disk replacement.
6691

6692
    This dispatches the disk replacement to the appropriate handler.
6693

6694
    """
6695
    if not self.disks:
6696
      feedback_fn("No disks need replacement")
6697
      return
6698

    
6699
    feedback_fn("Replacing disk(s) %s for %s" %
6700
                (utils.CommaJoin(self.disks), self.instance.name))
6701

    
6702
    activate_disks = (not self.instance.admin_up)
6703

    
6704
    # Activate the instance disks if we're replacing them on a down instance
6705
    if activate_disks:
6706
      _StartInstanceDisks(self.lu, self.instance, True)
6707

    
6708
    try:
6709
      # Should we replace the secondary node?
6710
      if self.new_node is not None:
6711
        fn = self._ExecDrbd8Secondary
6712
      else:
6713
        fn = self._ExecDrbd8DiskOnly
6714

    
6715
      return fn(feedback_fn)
6716

    
6717
    finally:
6718
      # Deactivate the instance disks if we're replacing them on a
6719
      # down instance
6720
      if activate_disks:
6721
        _SafeShutdownInstanceDisks(self.lu, self.instance)
6722

    
6723
  def _CheckVolumeGroup(self, nodes):
6724
    self.lu.LogInfo("Checking volume groups")
6725

    
6726
    vgname = self.cfg.GetVGName()
6727

    
6728
    # Make sure volume group exists on all involved nodes
6729
    results = self.rpc.call_vg_list(nodes)
6730
    if not results:
6731
      raise errors.OpExecError("Can't list volume groups on the nodes")
6732

    
6733
    for node in nodes:
6734
      res = results[node]
6735
      res.Raise("Error checking node %s" % node)
6736
      if vgname not in res.payload:
6737
        raise errors.OpExecError("Volume group '%s' not found on node %s" %
6738
                                 (vgname, node))
6739

    
6740
  def _CheckDisksExistence(self, nodes):
6741
    # Check disk existence
6742
    for idx, dev in enumerate(self.instance.disks):
6743
      if idx not in self.disks:
6744
        continue
6745

    
6746
      for node in nodes:
6747
        self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
6748
        self.cfg.SetDiskID(dev, node)
6749

    
6750
        result = self.rpc.call_blockdev_find(node, dev)
6751

    
6752
        msg = result.fail_msg
6753
        if msg or not result.payload:
6754
          if not msg:
6755
            msg = "disk not found"
6756
          raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
6757
                                   (idx, node, msg))
6758

    
6759
  def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
6760
    for idx, dev in enumerate(self.instance.disks):
6761
      if idx not in self.disks:
6762
        continue
6763

    
6764
      self.lu.LogInfo("Checking disk/%d consistency on node %s" %
6765
                      (idx, node_name))
6766

    
6767
      if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
6768
                                   ldisk=ldisk):
6769
        raise errors.OpExecError("Node %s has degraded storage, unsafe to"
6770
                                 " replace disks for instance %s" %
6771
                                 (node_name, self.instance.name))
6772

    
6773
  def _CreateNewStorage(self, node_name):
6774
    vgname = self.cfg.GetVGName()
6775
    iv_names = {}
6776

    
6777
    for idx, dev in enumerate(self.instance.disks):
6778
      if idx not in self.disks:
6779
        continue
6780

    
6781
      self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
6782

    
6783
      self.cfg.SetDiskID(dev, node_name)
6784

    
6785
      lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
6786
      names = _GenerateUniqueNames(self.lu, lv_names)
6787

    
6788
      lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
6789
                             logical_id=(vgname, names[0]))
6790
      lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
6791
                             logical_id=(vgname, names[1]))
6792

    
6793
      new_lvs = [lv_data, lv_meta]
6794
      old_lvs = dev.children
6795
      iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
6796

    
6797
      # we pass force_create=True to force the LVM creation
6798
      for new_lv in new_lvs:
6799
        _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
6800
                        _GetInstanceInfoText(self.instance), False)
6801

    
6802
    return iv_names
6803

    
6804
  def _CheckDevices(self, node_name, iv_names):
6805
    for name, (dev, _, _) in iv_names.iteritems():
6806
      self.cfg.SetDiskID(dev, node_name)
6807

    
6808
      result = self.rpc.call_blockdev_find(node_name, dev)
6809

    
6810
      msg = result.fail_msg
6811
      if msg or not result.payload:
6812
        if not msg:
6813
          msg = "disk not found"
6814
        raise errors.OpExecError("Can't find DRBD device %s: %s" %
6815
                                 (name, msg))
6816

    
6817
      if result.payload.is_degraded:
6818
        raise errors.OpExecError("DRBD device %s is degraded!" % name)
6819

    
6820
  def _RemoveOldStorage(self, node_name, iv_names):
6821
    for name, (_, old_lvs, _) in iv_names.iteritems():
6822
      self.lu.LogInfo("Remove logical volumes for %s" % name)
6823

    
6824
      for lv in old_lvs:
6825
        self.cfg.SetDiskID(lv, node_name)
6826

    
6827
        msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
6828
        if msg:
6829
          self.lu.LogWarning("Can't remove old LV: %s" % msg,
6830
                             hint="remove unused LVs manually")
6831

    
6832
  def _ExecDrbd8DiskOnly(self, feedback_fn):
6833
    """Replace a disk on the primary or secondary for DRBD 8.
6834

6835
    The algorithm for replace is quite complicated:
6836

6837
      1. for each disk to be replaced:
6838

6839
        1. create new LVs on the target node with unique names
6840
        1. detach old LVs from the drbd device
6841
        1. rename old LVs to name_replaced.<time_t>
6842
        1. rename new LVs to old LVs
6843
        1. attach the new LVs (with the old names now) to the drbd device
6844

6845
      1. wait for sync across all devices
6846

6847
      1. for each modified disk:
6848

6849
        1. remove old LVs (which have the name name_replaces.<time_t>)
6850

6851
    Failures are not very well handled.
6852

6853
    """
6854
    steps_total = 6
6855

    
6856
    # Step: check device activation
6857
    self.lu.LogStep(1, steps_total, "Check device existence")
6858
    self._CheckDisksExistence([self.other_node, self.target_node])
6859
    self._CheckVolumeGroup([self.target_node, self.other_node])
6860

    
6861
    # Step: check other node consistency
6862
    self.lu.LogStep(2, steps_total, "Check peer consistency")
6863
    self._CheckDisksConsistency(self.other_node,
6864
                                self.other_node == self.instance.primary_node,
6865
                                False)
6866

    
6867
    # Step: create new storage
6868
    self.lu.LogStep(3, steps_total, "Allocate new storage")
6869
    iv_names = self._CreateNewStorage(self.target_node)
6870

    
6871
    # Step: for each lv, detach+rename*2+attach
6872
    self.lu.LogStep(4, steps_total, "Changing drbd configuration")
6873
    for dev, old_lvs, new_lvs in iv_names.itervalues():
6874
      self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
6875

    
6876
      result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
6877
                                                     old_lvs)
6878
      result.Raise("Can't detach drbd from local storage on node"
6879
                   " %s for device %s" % (self.target_node, dev.iv_name))
6880
      #dev.children = []
6881
      #cfg.Update(instance)
6882

    
6883
      # ok, we created the new LVs, so now we know we have the needed
6884
      # storage; as such, we proceed on the target node to rename
6885
      # old_lv to _old, and new_lv to old_lv; note that we rename LVs
6886
      # using the assumption that logical_id == physical_id (which in
6887
      # turn is the unique_id on that node)
6888

    
6889
      # FIXME(iustin): use a better name for the replaced LVs
6890
      temp_suffix = int(time.time())
6891
      ren_fn = lambda d, suff: (d.physical_id[0],
6892
                                d.physical_id[1] + "_replaced-%s" % suff)
6893

    
6894
      # Build the rename list based on what LVs exist on the node
6895
      rename_old_to_new = []
6896
      for to_ren in old_lvs:
6897
        result = self.rpc.call_blockdev_find(self.target_node, to_ren)
6898
        if not result.fail_msg and result.payload:
6899
          # device exists
6900
          rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
6901

    
6902
      self.lu.LogInfo("Renaming the old LVs on the target node")
6903
      result = self.rpc.call_blockdev_rename(self.target_node,
6904
                                             rename_old_to_new)
6905
      result.Raise("Can't rename old LVs on node %s" % self.target_node)
6906

    
6907
      # Now we rename the new LVs to the old LVs
6908
      self.lu.LogInfo("Renaming the new LVs on the target node")
6909
      rename_new_to_old = [(new, old.physical_id)
6910
                           for old, new in zip(old_lvs, new_lvs)]
6911
      result = self.rpc.call_blockdev_rename(self.target_node,
6912
                                             rename_new_to_old)
6913
      result.Raise("Can't rename new LVs on node %s" % self.target_node)
6914

    
6915
      for old, new in zip(old_lvs, new_lvs):
6916
        new.logical_id = old.logical_id
6917
        self.cfg.SetDiskID(new, self.target_node)
6918

    
6919
      for disk in old_lvs:
6920
        disk.logical_id = ren_fn(disk, temp_suffix)
6921
        self.cfg.SetDiskID(disk, self.target_node)
6922

    
6923
      # Now that the new lvs have the old name, we can add them to the device
6924
      self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
6925
      result = self.rpc.call_blockdev_addchildren(self.target_node, dev,
6926
                                                  new_lvs)
6927
      msg = result.fail_msg
6928
      if msg:
6929
        for new_lv in new_lvs:
6930
          msg2 = self.rpc.call_blockdev_remove(self.target_node,
6931
                                               new_lv).fail_msg
6932
          if msg2:
6933
            self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
6934
                               hint=("cleanup manually the unused logical"
6935
                                     "volumes"))
6936
        raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
6937

    
6938
      dev.children = new_lvs
6939

    
6940
      self.cfg.Update(self.instance, feedback_fn)
6941

    
6942
    # Wait for sync
6943
    # This can fail as the old devices are degraded and _WaitForSync
6944
    # does a combined result over all disks, so we don't check its return value
6945
    self.lu.LogStep(5, steps_total, "Sync devices")
6946
    _WaitForSync(self.lu, self.instance)
6947

    
6948
    # Check all devices manually
6949
    self._CheckDevices(self.instance.primary_node, iv_names)
6950

    
6951
    # Step: remove old storage
6952
    self.lu.LogStep(6, steps_total, "Removing old storage")
6953
    self._RemoveOldStorage(self.target_node, iv_names)
6954

    
6955
  def _ExecDrbd8Secondary(self, feedback_fn):
6956
    """Replace the secondary node for DRBD 8.
6957

6958
    The algorithm for replace is quite complicated:
6959
      - for all disks of the instance:
6960
        - create new LVs on the new node with same names
6961
        - shutdown the drbd device on the old secondary
6962
        - disconnect the drbd network on the primary
6963
        - create the drbd device on the new secondary
6964
        - network attach the drbd on the primary, using an artifice:
6965
          the drbd code for Attach() will connect to the network if it
6966
          finds a device which is connected to the good local disks but
6967
          not network enabled
6968
      - wait for sync across all devices
6969
      - remove all disks from the old secondary
6970

6971
    Failures are not very well handled.
6972

6973
    """
6974
    steps_total = 6
6975

    
6976
    # Step: check device activation
6977
    self.lu.LogStep(1, steps_total, "Check device existence")
6978
    self._CheckDisksExistence([self.instance.primary_node])
6979
    self._CheckVolumeGroup([self.instance.primary_node])
6980

    
6981
    # Step: check other node consistency
6982
    self.lu.LogStep(2, steps_total, "Check peer consistency")
6983
    self._CheckDisksConsistency(self.instance.primary_node, True, True)
6984

    
6985
    # Step: create new storage
6986
    self.lu.LogStep(3, steps_total, "Allocate new storage")
6987
    for idx, dev in enumerate(self.instance.disks):
6988
      self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
6989
                      (self.new_node, idx))
6990
      # we pass force_create=True to force LVM creation
6991
      for new_lv in dev.children:
6992
        _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
6993
                        _GetInstanceInfoText(self.instance), False)
6994

    
6995
    # Step 4: dbrd minors and drbd setups changes
6996
    # after this, we must manually remove the drbd minors on both the
6997
    # error and the success paths
6998
    self.lu.LogStep(4, steps_total, "Changing drbd configuration")
6999
    minors = self.cfg.AllocateDRBDMinor([self.new_node
7000
                                         for dev in self.instance.disks],
7001
                                        self.instance.name)
7002
    logging.debug("Allocated minors %r", minors)
7003

    
7004
    iv_names = {}
7005
    for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
7006
      self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
7007
                      (self.new_node, idx))
7008
      # create new devices on new_node; note that we create two IDs:
7009
      # one without port, so the drbd will be activated without
7010
      # networking information on the new node at this stage, and one
7011
      # with network, for the latter activation in step 4
7012
      (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
7013
      if self.instance.primary_node == o_node1:
7014
        p_minor = o_minor1
7015
      else:
7016
        assert self.instance.primary_node == o_node2, "Three-node instance?"
7017
        p_minor = o_minor2
7018

    
7019
      new_alone_id = (self.instance.primary_node, self.new_node, None,
7020
                      p_minor, new_minor, o_secret)
7021
      new_net_id = (self.instance.primary_node, self.new_node, o_port,
7022
                    p_minor, new_minor, o_secret)
7023

    
7024
      iv_names[idx] = (dev, dev.children, new_net_id)
7025
      logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
7026
                    new_net_id)
7027
      new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
7028
                              logical_id=new_alone_id,
7029
                              children=dev.children,
7030
                              size=dev.size)
7031
      try:
7032
        _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
7033
                              _GetInstanceInfoText(self.instance), False)
7034
      except errors.GenericError:
7035
        self.cfg.ReleaseDRBDMinors(self.instance.name)
7036
        raise
7037

    
7038
    # We have new devices, shutdown the drbd on the old secondary
7039
    for idx, dev in enumerate(self.instance.disks):
7040
      self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
7041
      self.cfg.SetDiskID(dev, self.target_node)
7042
      msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
7043
      if msg:
7044
        self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
7045
                           "node: %s" % (idx, msg),
7046
                           hint=("Please cleanup this device manually as"
7047
                                 " soon as possible"))
7048

    
7049
    self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
7050
    result = self.rpc.call_drbd_disconnect_net([self.instance.primary_node],
7051
                                               self.node_secondary_ip,
7052
                                               self.instance.disks)\
7053
                                              [self.instance.primary_node]
7054

    
7055
    msg = result.fail_msg
7056
    if msg:
7057
      # detaches didn't succeed (unlikely)
7058
      self.cfg.ReleaseDRBDMinors(self.instance.name)
7059
      raise errors.OpExecError("Can't detach the disks from the network on"
7060
                               " old node: %s" % (msg,))
7061

    
7062
    # if we managed to detach at least one, we update all the disks of
7063
    # the instance to point to the new secondary
7064
    self.lu.LogInfo("Updating instance configuration")
7065
    for dev, _, new_logical_id in iv_names.itervalues():
7066
      dev.logical_id = new_logical_id
7067
      self.cfg.SetDiskID(dev, self.instance.primary_node)
7068

    
7069
    self.cfg.Update(self.instance, feedback_fn)
7070

    
7071
    # and now perform the drbd attach
7072
    self.lu.LogInfo("Attaching primary drbds to new secondary"
7073
                    " (standalone => connected)")
7074
    result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
7075
                                            self.new_node],
7076
                                           self.node_secondary_ip,
7077
                                           self.instance.disks,
7078
                                           self.instance.name,
7079
                                           False)
7080
    for to_node, to_result in result.items():
7081
      msg = to_result.fail_msg
7082
      if msg:
7083
        self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
7084
                           to_node, msg,
7085
                           hint=("please do a gnt-instance info to see the"
7086
                                 " status of disks"))
7087

    
7088
    # Wait for sync
7089
    # This can fail as the old devices are degraded and _WaitForSync
7090
    # does a combined result over all disks, so we don't check its return value
7091
    self.lu.LogStep(5, steps_total, "Sync devices")
7092
    _WaitForSync(self.lu, self.instance)
7093

    
7094
    # Check all devices manually
7095
    self._CheckDevices(self.instance.primary_node, iv_names)
7096

    
7097
    # Step: remove old storage
7098
    self.lu.LogStep(6, steps_total, "Removing old storage")
7099
    self._RemoveOldStorage(self.target_node, iv_names)
7100

    
7101

    
7102
class LURepairNodeStorage(NoHooksLU):
7103
  """Repairs the volume group on a node.
7104

7105
  """
7106
  _OP_REQP = ["node_name"]
7107
  REQ_BGL = False
7108

    
7109
  def CheckArguments(self):
7110
    node_name = self.cfg.ExpandNodeName(self.op.node_name)
7111
    if node_name is None:
7112
      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name,
7113
                                 errors.ECODE_NOENT)
7114

    
7115
    self.op.node_name = node_name
7116

    
7117
  def ExpandNames(self):
7118
    self.needed_locks = {
7119
      locking.LEVEL_NODE: [self.op.node_name],
7120
      }
7121

    
7122
  def _CheckFaultyDisks(self, instance, node_name):
7123
    """Ensure faulty disks abort the opcode or at least warn."""
7124
    try:
7125
      if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
7126
                                  node_name, True):
7127
        raise errors.OpPrereqError("Instance '%s' has faulty disks on"
7128
                                   " node '%s'" % (instance.name, node_name),
7129
                                   errors.ECODE_STATE)
7130
    except errors.OpPrereqError, err:
7131
      if self.op.ignore_consistency:
7132
        self.proc.LogWarning(str(err.args[0]))
7133
      else:
7134
        raise
7135

    
7136
  def CheckPrereq(self):
7137
    """Check prerequisites.
7138

7139
    """
7140
    storage_type = self.op.storage_type
7141

    
7142
    if (constants.SO_FIX_CONSISTENCY not in
7143
        constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
7144
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
7145
                                 " repaired" % storage_type,
7146
                                 errors.ECODE_INVAL)
7147

    
7148
    # Check whether any instance on this node has faulty disks
7149
    for inst in _GetNodeInstances(self.cfg, self.op.node_name):
7150
      if not inst.admin_up:
7151
        continue
7152
      check_nodes = set(inst.all_nodes)
7153
      check_nodes.discard(self.op.node_name)
7154
      for inst_node_name in check_nodes:
7155
        self._CheckFaultyDisks(inst, inst_node_name)
7156

    
7157
  def Exec(self, feedback_fn):
7158
    feedback_fn("Repairing storage unit '%s' on %s ..." %
7159
                (self.op.name, self.op.node_name))
7160

    
7161
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
7162
    result = self.rpc.call_storage_execute(self.op.node_name,
7163
                                           self.op.storage_type, st_args,
7164
                                           self.op.name,
7165
                                           constants.SO_FIX_CONSISTENCY)
7166
    result.Raise("Failed to repair storage unit '%s' on %s" %
7167
                 (self.op.name, self.op.node_name))
7168

    
7169

    
7170
class LUGrowDisk(LogicalUnit):
7171
  """Grow a disk of an instance.
7172

7173
  """
7174
  HPATH = "disk-grow"
7175
  HTYPE = constants.HTYPE_INSTANCE
7176
  _OP_REQP = ["instance_name", "disk", "amount", "wait_for_sync"]
7177
  REQ_BGL = False
7178

    
7179
  def ExpandNames(self):
7180
    self._ExpandAndLockInstance()
7181
    self.needed_locks[locking.LEVEL_NODE] = []
7182
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7183

    
7184
  def DeclareLocks(self, level):
7185
    if level == locking.LEVEL_NODE:
7186
      self._LockInstancesNodes()
7187

    
7188
  def BuildHooksEnv(self):
7189
    """Build hooks env.
7190

7191
    This runs on the master, the primary and all the secondaries.
7192

7193
    """
7194
    env = {
7195
      "DISK": self.op.disk,
7196
      "AMOUNT": self.op.amount,
7197
      }
7198
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
7199
    nl = [
7200
      self.cfg.GetMasterNode(),
7201
      self.instance.primary_node,
7202
      ]
7203
    return env, nl, nl
7204

    
7205
  def CheckPrereq(self):
7206
    """Check prerequisites.
7207

7208
    This checks that the instance is in the cluster.
7209

7210
    """
7211
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7212
    assert instance is not None, \
7213
      "Cannot retrieve locked instance %s" % self.op.instance_name
7214
    nodenames = list(instance.all_nodes)
7215
    for node in nodenames:
7216
      _CheckNodeOnline(self, node)
7217

    
7218

    
7219
    self.instance = instance
7220

    
7221
    if instance.disk_template not in (constants.DT_PLAIN, constants.DT_DRBD8):
7222
      raise errors.OpPrereqError("Instance's disk layout does not support"
7223
                                 " growing.", errors.ECODE_INVAL)
7224

    
7225
    self.disk = instance.FindDisk(self.op.disk)
7226

    
7227
    nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
7228
                                       instance.hypervisor)
7229
    for node in nodenames:
7230
      info = nodeinfo[node]
7231
      info.Raise("Cannot get current information from node %s" % node)
7232
      vg_free = info.payload.get('vg_free', None)
7233
      if not isinstance(vg_free, int):
7234
        raise errors.OpPrereqError("Can't compute free disk space on"
7235
                                   " node %s" % node, errors.ECODE_ENVIRON)
7236
      if self.op.amount > vg_free:
7237
        raise errors.OpPrereqError("Not enough disk space on target node %s:"
7238
                                   " %d MiB available, %d MiB required" %
7239
                                   (node, vg_free, self.op.amount),
7240
                                   errors.ECODE_NORES)
7241

    
7242
  def Exec(self, feedback_fn):
7243
    """Execute disk grow.
7244

7245
    """
7246
    instance = self.instance
7247
    disk = self.disk
7248
    for node in instance.all_nodes:
7249
      self.cfg.SetDiskID(disk, node)
7250
      result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
7251
      result.Raise("Grow request failed to node %s" % node)
7252

    
7253
      # TODO: Rewrite code to work properly
7254
      # DRBD goes into sync mode for a short amount of time after executing the
7255
      # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
7256
      # calling "resize" in sync mode fails. Sleeping for a short amount of
7257
      # time is a work-around.
7258
      time.sleep(5)
7259

    
7260
    disk.RecordGrow(self.op.amount)
7261
    self.cfg.Update(instance, feedback_fn)
7262
    if self.op.wait_for_sync:
7263
      disk_abort = not _WaitForSync(self, instance)
7264
      if disk_abort:
7265
        self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
7266
                             " status.\nPlease check the instance.")
7267

    
7268

    
7269
class LUQueryInstanceData(NoHooksLU):
7270
  """Query runtime instance data.
7271

7272
  """
7273
  _OP_REQP = ["instances", "static"]
7274
  REQ_BGL = False
7275

    
7276
  def ExpandNames(self):
7277
    self.needed_locks = {}
7278
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
7279

    
7280
    if not isinstance(self.op.instances, list):
7281
      raise errors.OpPrereqError("Invalid argument type 'instances'",
7282
                                 errors.ECODE_INVAL)
7283

    
7284
    if self.op.instances:
7285
      self.wanted_names = []
7286
      for name in self.op.instances:
7287
        full_name = self.cfg.ExpandInstanceName(name)
7288
        if full_name is None:
7289
          raise errors.OpPrereqError("Instance '%s' not known" % name,
7290
                                     errors.ECODE_NOENT)
7291
        self.wanted_names.append(full_name)
7292
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
7293
    else:
7294
      self.wanted_names = None
7295
      self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
7296

    
7297
    self.needed_locks[locking.LEVEL_NODE] = []
7298
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7299

    
7300
  def DeclareLocks(self, level):
7301
    if level == locking.LEVEL_NODE:
7302
      self._LockInstancesNodes()
7303

    
7304
  def CheckPrereq(self):
7305
    """Check prerequisites.
7306

7307
    This only checks the optional instance list against the existing names.
7308

7309
    """
7310
    if self.wanted_names is None:
7311
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
7312

    
7313
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
7314
                             in self.wanted_names]
7315
    return
7316

    
7317
  def _ComputeBlockdevStatus(self, node, instance_name, dev):
7318
    """Returns the status of a block device
7319

7320
    """
7321
    if self.op.static or not node:
7322
      return None
7323

    
7324
    self.cfg.SetDiskID(dev, node)
7325

    
7326
    result = self.rpc.call_blockdev_find(node, dev)
7327
    if result.offline:
7328
      return None
7329

    
7330
    result.Raise("Can't compute disk status for %s" % instance_name)
7331

    
7332
    status = result.payload
7333
    if status is None:
7334
      return None
7335

    
7336
    return (status.dev_path, status.major, status.minor,
7337
            status.sync_percent, status.estimated_time,
7338
            status.is_degraded, status.ldisk_status)
7339

    
7340
  def _ComputeDiskStatus(self, instance, snode, dev):
7341
    """Compute block device status.
7342

7343
    """
7344
    if dev.dev_type in constants.LDS_DRBD:
7345
      # we change the snode then (otherwise we use the one passed in)
7346
      if dev.logical_id[0] == instance.primary_node:
7347
        snode = dev.logical_id[1]
7348
      else:
7349
        snode = dev.logical_id[0]
7350

    
7351
    dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
7352
                                              instance.name, dev)
7353
    dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
7354

    
7355
    if dev.children:
7356
      dev_children = [self._ComputeDiskStatus(instance, snode, child)
7357
                      for child in dev.children]
7358
    else:
7359
      dev_children = []
7360

    
7361
    data = {
7362
      "iv_name": dev.iv_name,
7363
      "dev_type": dev.dev_type,
7364
      "logical_id": dev.logical_id,
7365
      "physical_id": dev.physical_id,
7366
      "pstatus": dev_pstatus,
7367
      "sstatus": dev_sstatus,
7368
      "children": dev_children,
7369
      "mode": dev.mode,
7370
      "size": dev.size,
7371
      }
7372

    
7373
    return data
7374

    
7375
  def Exec(self, feedback_fn):
7376
    """Gather and return data"""
7377
    result = {}
7378

    
7379
    cluster = self.cfg.GetClusterInfo()
7380

    
7381
    for instance in self.wanted_instances:
7382
      if not self.op.static:
7383
        remote_info = self.rpc.call_instance_info(instance.primary_node,
7384
                                                  instance.name,
7385
                                                  instance.hypervisor)
7386
        remote_info.Raise("Error checking node %s" % instance.primary_node)
7387
        remote_info = remote_info.payload
7388
        if remote_info and "state" in remote_info:
7389
          remote_state = "up"
7390
        else:
7391
          remote_state = "down"
7392
      else:
7393
        remote_state = None
7394
      if instance.admin_up:
7395
        config_state = "up"
7396
      else:
7397
        config_state = "down"
7398

    
7399
      disks = [self._ComputeDiskStatus(instance, None, device)
7400
               for device in instance.disks]
7401

    
7402
      idict = {
7403
        "name": instance.name,
7404
        "config_state": config_state,
7405
        "run_state": remote_state,
7406
        "pnode": instance.primary_node,
7407
        "snodes": instance.secondary_nodes,
7408
        "os": instance.os,
7409
        # this happens to be the same format used for hooks
7410
        "nics": _NICListToTuple(self, instance.nics),
7411
        "disks": disks,
7412
        "hypervisor": instance.hypervisor,
7413
        "network_port": instance.network_port,
7414
        "hv_instance": instance.hvparams,
7415
        "hv_actual": cluster.FillHV(instance, skip_globals=True),
7416
        "be_instance": instance.beparams,
7417
        "be_actual": cluster.FillBE(instance),
7418
        "serial_no": instance.serial_no,
7419
        "mtime": instance.mtime,
7420
        "ctime": instance.ctime,
7421
        "uuid": instance.uuid,
7422
        }
7423

    
7424
      result[instance.name] = idict
7425

    
7426
    return result
7427

    
7428

    
7429
class LUSetInstanceParams(LogicalUnit):
7430
  """Modifies an instances's parameters.
7431

7432
  """
7433
  HPATH = "instance-modify"
7434
  HTYPE = constants.HTYPE_INSTANCE
7435
  _OP_REQP = ["instance_name"]
7436
  REQ_BGL = False
7437

    
7438
  def CheckArguments(self):
7439
    if not hasattr(self.op, 'nics'):
7440
      self.op.nics = []
7441
    if not hasattr(self.op, 'disks'):
7442
      self.op.disks = []
7443
    if not hasattr(self.op, 'beparams'):
7444
      self.op.beparams = {}
7445
    if not hasattr(self.op, 'hvparams'):
7446
      self.op.hvparams = {}
7447
    self.op.force = getattr(self.op, "force", False)
7448
    if not (self.op.nics or self.op.disks or
7449
            self.op.hvparams or self.op.beparams):
7450
      raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
7451

    
7452
    if self.op.hvparams:
7453
      _CheckGlobalHvParams(self.op.hvparams)
7454

    
7455
    # Disk validation
7456
    disk_addremove = 0
7457
    for disk_op, disk_dict in self.op.disks:
7458
      if disk_op == constants.DDM_REMOVE:
7459
        disk_addremove += 1
7460
        continue
7461
      elif disk_op == constants.DDM_ADD:
7462
        disk_addremove += 1
7463
      else:
7464
        if not isinstance(disk_op, int):
7465
          raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL)
7466
        if not isinstance(disk_dict, dict):
7467
          msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
7468
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
7469

    
7470
      if disk_op == constants.DDM_ADD:
7471
        mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
7472
        if mode not in constants.DISK_ACCESS_SET:
7473
          raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
7474
                                     errors.ECODE_INVAL)
7475
        size = disk_dict.get('size', None)
7476
        if size is None:
7477
          raise errors.OpPrereqError("Required disk parameter size missing",
7478
                                     errors.ECODE_INVAL)
7479
        try:
7480
          size = int(size)
7481
        except ValueError, err:
7482
          raise errors.OpPrereqError("Invalid disk size parameter: %s" %
7483
                                     str(err), errors.ECODE_INVAL)
7484
        disk_dict['size'] = size
7485
      else:
7486
        # modification of disk
7487
        if 'size' in disk_dict:
7488
          raise errors.OpPrereqError("Disk size change not possible, use"
7489
                                     " grow-disk", errors.ECODE_INVAL)
7490

    
7491
    if disk_addremove > 1:
7492
      raise errors.OpPrereqError("Only one disk add or remove operation"
7493
                                 " supported at a time", errors.ECODE_INVAL)
7494

    
7495
    # NIC validation
7496
    nic_addremove = 0
7497
    for nic_op, nic_dict in self.op.nics:
7498
      if nic_op == constants.DDM_REMOVE:
7499
        nic_addremove += 1
7500
        continue
7501
      elif nic_op == constants.DDM_ADD:
7502
        nic_addremove += 1
7503
      else:
7504
        if not isinstance(nic_op, int):
7505
          raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL)
7506
        if not isinstance(nic_dict, dict):
7507
          msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
7508
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
7509

    
7510
      # nic_dict should be a dict
7511
      nic_ip = nic_dict.get('ip', None)
7512
      if nic_ip is not None:
7513
        if nic_ip.lower() == constants.VALUE_NONE:
7514
          nic_dict['ip'] = None
7515
        else:
7516
          if not utils.IsValidIP(nic_ip):
7517
            raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip,
7518
                                       errors.ECODE_INVAL)
7519

    
7520
      nic_bridge = nic_dict.get('bridge', None)
7521
      nic_link = nic_dict.get('link', None)
7522
      if nic_bridge and nic_link:
7523
        raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
7524
                                   " at the same time", errors.ECODE_INVAL)
7525
      elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
7526
        nic_dict['bridge'] = None
7527
      elif nic_link and nic_link.lower() == constants.VALUE_NONE:
7528
        nic_dict['link'] = None
7529

    
7530
      if nic_op == constants.DDM_ADD:
7531
        nic_mac = nic_dict.get('mac', None)
7532
        if nic_mac is None:
7533
          nic_dict['mac'] = constants.VALUE_AUTO
7534

    
7535
      if 'mac' in nic_dict:
7536
        nic_mac = nic_dict['mac']
7537
        if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7538
          if not utils.IsValidMac(nic_mac):
7539
            raise errors.OpPrereqError("Invalid MAC address %s" % nic_mac,
7540
                                       errors.ECODE_INVAL)
7541
        if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
7542
          raise errors.OpPrereqError("'auto' is not a valid MAC address when"
7543
                                     " modifying an existing nic",
7544
                                     errors.ECODE_INVAL)
7545

    
7546
    if nic_addremove > 1:
7547
      raise errors.OpPrereqError("Only one NIC add or remove operation"
7548
                                 " supported at a time", errors.ECODE_INVAL)
7549

    
7550
  def ExpandNames(self):
7551
    self._ExpandAndLockInstance()
7552
    self.needed_locks[locking.LEVEL_NODE] = []
7553
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7554

    
7555
  def DeclareLocks(self, level):
7556
    if level == locking.LEVEL_NODE:
7557
      self._LockInstancesNodes()
7558

    
7559
  def BuildHooksEnv(self):
7560
    """Build hooks env.
7561

7562
    This runs on the master, primary and secondaries.
7563

7564
    """
7565
    args = dict()
7566
    if constants.BE_MEMORY in self.be_new:
7567
      args['memory'] = self.be_new[constants.BE_MEMORY]
7568
    if constants.BE_VCPUS in self.be_new:
7569
      args['vcpus'] = self.be_new[constants.BE_VCPUS]
7570
    # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
7571
    # information at all.
7572
    if self.op.nics:
7573
      args['nics'] = []
7574
      nic_override = dict(self.op.nics)
7575
      c_nicparams = self.cluster.nicparams[constants.PP_DEFAULT]
7576
      for idx, nic in enumerate(self.instance.nics):
7577
        if idx in nic_override:
7578
          this_nic_override = nic_override[idx]
7579
        else:
7580
          this_nic_override = {}
7581
        if 'ip' in this_nic_override:
7582
          ip = this_nic_override['ip']
7583
        else:
7584
          ip = nic.ip
7585
        if 'mac' in this_nic_override:
7586
          mac = this_nic_override['mac']
7587
        else:
7588
          mac = nic.mac
7589
        if idx in self.nic_pnew:
7590
          nicparams = self.nic_pnew[idx]
7591
        else:
7592
          nicparams = objects.FillDict(c_nicparams, nic.nicparams)
7593
        mode = nicparams[constants.NIC_MODE]
7594
        link = nicparams[constants.NIC_LINK]
7595
        args['nics'].append((ip, mac, mode, link))
7596
      if constants.DDM_ADD in nic_override:
7597
        ip = nic_override[constants.DDM_ADD].get('ip', None)
7598
        mac = nic_override[constants.DDM_ADD]['mac']
7599
        nicparams = self.nic_pnew[constants.DDM_ADD]
7600
        mode = nicparams[constants.NIC_MODE]
7601
        link = nicparams[constants.NIC_LINK]
7602
        args['nics'].append((ip, mac, mode, link))
7603
      elif constants.DDM_REMOVE in nic_override:
7604
        del args['nics'][-1]
7605

    
7606
    env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
7607
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
7608
    return env, nl, nl
7609

    
7610
  def _GetUpdatedParams(self, old_params, update_dict,
7611
                        default_values, parameter_types):
7612
    """Return the new params dict for the given params.
7613

7614
    @type old_params: dict
7615
    @param old_params: old parameters
7616
    @type update_dict: dict
7617
    @param update_dict: dict containing new parameter values,
7618
                        or constants.VALUE_DEFAULT to reset the
7619
                        parameter to its default value
7620
    @type default_values: dict
7621
    @param default_values: default values for the filled parameters
7622
    @type parameter_types: dict
7623
    @param parameter_types: dict mapping target dict keys to types
7624
                            in constants.ENFORCEABLE_TYPES
7625
    @rtype: (dict, dict)
7626
    @return: (new_parameters, filled_parameters)
7627

7628
    """
7629
    params_copy = copy.deepcopy(old_params)
7630
    for key, val in update_dict.iteritems():
7631
      if val == constants.VALUE_DEFAULT:
7632
        try:
7633
          del params_copy[key]
7634
        except KeyError:
7635
          pass
7636
      else:
7637
        params_copy[key] = val
7638
    utils.ForceDictType(params_copy, parameter_types)
7639
    params_filled = objects.FillDict(default_values, params_copy)
7640
    return (params_copy, params_filled)
7641

    
7642
  def CheckPrereq(self):
7643
    """Check prerequisites.
7644

7645
    This only checks the instance list against the existing names.
7646

7647
    """
7648
    self.force = self.op.force
7649

    
7650
    # checking the new params on the primary/secondary nodes
7651

    
7652
    instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7653
    cluster = self.cluster = self.cfg.GetClusterInfo()
7654
    assert self.instance is not None, \
7655
      "Cannot retrieve locked instance %s" % self.op.instance_name
7656
    pnode = instance.primary_node
7657
    nodelist = list(instance.all_nodes)
7658

    
7659
    # hvparams processing
7660
    if self.op.hvparams:
7661
      i_hvdict, hv_new = self._GetUpdatedParams(
7662
                             instance.hvparams, self.op.hvparams,
7663
                             cluster.hvparams[instance.hypervisor],
7664
                             constants.HVS_PARAMETER_TYPES)
7665
      # local check
7666
      hypervisor.GetHypervisor(
7667
        instance.hypervisor).CheckParameterSyntax(hv_new)
7668
      _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
7669
      self.hv_new = hv_new # the new actual values
7670
      self.hv_inst = i_hvdict # the new dict (without defaults)
7671
    else:
7672
      self.hv_new = self.hv_inst = {}
7673

    
7674
    # beparams processing
7675
    if self.op.beparams:
7676
      i_bedict, be_new = self._GetUpdatedParams(
7677
                             instance.beparams, self.op.beparams,
7678
                             cluster.beparams[constants.PP_DEFAULT],
7679
                             constants.BES_PARAMETER_TYPES)
7680
      self.be_new = be_new # the new actual values
7681
      self.be_inst = i_bedict # the new dict (without defaults)
7682
    else:
7683
      self.be_new = self.be_inst = {}
7684

    
7685
    self.warn = []
7686

    
7687
    if constants.BE_MEMORY in self.op.beparams and not self.force:
7688
      mem_check_list = [pnode]
7689
      if be_new[constants.BE_AUTO_BALANCE]:
7690
        # either we changed auto_balance to yes or it was from before
7691
        mem_check_list.extend(instance.secondary_nodes)
7692
      instance_info = self.rpc.call_instance_info(pnode, instance.name,
7693
                                                  instance.hypervisor)
7694
      nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
7695
                                         instance.hypervisor)
7696
      pninfo = nodeinfo[pnode]
7697
      msg = pninfo.fail_msg
7698
      if msg:
7699
        # Assume the primary node is unreachable and go ahead
7700
        self.warn.append("Can't get info from primary node %s: %s" %
7701
                         (pnode,  msg))
7702
      elif not isinstance(pninfo.payload.get('memory_free', None), int):
7703
        self.warn.append("Node data from primary node %s doesn't contain"
7704
                         " free memory information" % pnode)
7705
      elif instance_info.fail_msg:
7706
        self.warn.append("Can't get instance runtime information: %s" %
7707
                        instance_info.fail_msg)
7708
      else:
7709
        if instance_info.payload:
7710
          current_mem = int(instance_info.payload['memory'])
7711
        else:
7712
          # Assume instance not running
7713
          # (there is a slight race condition here, but it's not very probable,
7714
          # and we have no other way to check)
7715
          current_mem = 0
7716
        miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
7717
                    pninfo.payload['memory_free'])
7718
        if miss_mem > 0:
7719
          raise errors.OpPrereqError("This change will prevent the instance"
7720
                                     " from starting, due to %d MB of memory"
7721
                                     " missing on its primary node" % miss_mem,
7722
                                     errors.ECODE_NORES)
7723

    
7724
      if be_new[constants.BE_AUTO_BALANCE]:
7725
        for node, nres in nodeinfo.items():
7726
          if node not in instance.secondary_nodes:
7727
            continue
7728
          msg = nres.fail_msg
7729
          if msg:
7730
            self.warn.append("Can't get info from secondary node %s: %s" %
7731
                             (node, msg))
7732
          elif not isinstance(nres.payload.get('memory_free', None), int):
7733
            self.warn.append("Secondary node %s didn't return free"
7734
                             " memory information" % node)
7735
          elif be_new[constants.BE_MEMORY] > nres.payload['memory_free']:
7736
            self.warn.append("Not enough memory to failover instance to"
7737
                             " secondary node %s" % node)
7738

    
7739
    # NIC processing
7740
    self.nic_pnew = {}
7741
    self.nic_pinst = {}
7742
    for nic_op, nic_dict in self.op.nics:
7743
      if nic_op == constants.DDM_REMOVE:
7744
        if not instance.nics:
7745
          raise errors.OpPrereqError("Instance has no NICs, cannot remove",
7746
                                     errors.ECODE_INVAL)
7747
        continue
7748
      if nic_op != constants.DDM_ADD:
7749
        # an existing nic
7750
        if not instance.nics:
7751
          raise errors.OpPrereqError("Invalid NIC index %s, instance has"
7752
                                     " no NICs" % nic_op,
7753
                                     errors.ECODE_INVAL)
7754
        if nic_op < 0 or nic_op >= len(instance.nics):
7755
          raise errors.OpPrereqError("Invalid NIC index %s, valid values"
7756
                                     " are 0 to %d" %
7757
                                     (nic_op, len(instance.nics) - 1),
7758
                                     errors.ECODE_INVAL)
7759
        old_nic_params = instance.nics[nic_op].nicparams
7760
        old_nic_ip = instance.nics[nic_op].ip
7761
      else:
7762
        old_nic_params = {}
7763
        old_nic_ip = None
7764

    
7765
      update_params_dict = dict([(key, nic_dict[key])
7766
                                 for key in constants.NICS_PARAMETERS
7767
                                 if key in nic_dict])
7768

    
7769
      if 'bridge' in nic_dict:
7770
        update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
7771

    
7772
      new_nic_params, new_filled_nic_params = \
7773
          self._GetUpdatedParams(old_nic_params, update_params_dict,
7774
                                 cluster.nicparams[constants.PP_DEFAULT],
7775
                                 constants.NICS_PARAMETER_TYPES)
7776
      objects.NIC.CheckParameterSyntax(new_filled_nic_params)
7777
      self.nic_pinst[nic_op] = new_nic_params
7778
      self.nic_pnew[nic_op] = new_filled_nic_params
7779
      new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
7780

    
7781
      if new_nic_mode == constants.NIC_MODE_BRIDGED:
7782
        nic_bridge = new_filled_nic_params[constants.NIC_LINK]
7783
        msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
7784
        if msg:
7785
          msg = "Error checking bridges on node %s: %s" % (pnode, msg)
7786
          if self.force:
7787
            self.warn.append(msg)
7788
          else:
7789
            raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
7790
      if new_nic_mode == constants.NIC_MODE_ROUTED:
7791
        if 'ip' in nic_dict:
7792
          nic_ip = nic_dict['ip']
7793
        else:
7794
          nic_ip = old_nic_ip
7795
        if nic_ip is None:
7796
          raise errors.OpPrereqError('Cannot set the nic ip to None'
7797
                                     ' on a routed nic', errors.ECODE_INVAL)
7798
      if 'mac' in nic_dict:
7799
        nic_mac = nic_dict['mac']
7800
        if nic_mac is None:
7801
          raise errors.OpPrereqError('Cannot set the nic mac to None',
7802
                                     errors.ECODE_INVAL)
7803
        elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7804
          # otherwise generate the mac
7805
          nic_dict['mac'] = self.cfg.GenerateMAC(self.proc.GetECId())
7806
        else:
7807
          # or validate/reserve the current one
7808
          try:
7809
            self.cfg.ReserveMAC(nic_mac, self.proc.GetECId())
7810
          except errors.ReservationError:
7811
            raise errors.OpPrereqError("MAC address %s already in use"
7812
                                       " in cluster" % nic_mac,
7813
                                       errors.ECODE_NOTUNIQUE)
7814

    
7815
    # DISK processing
7816
    if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
7817
      raise errors.OpPrereqError("Disk operations not supported for"
7818
                                 " diskless instances",
7819
                                 errors.ECODE_INVAL)
7820
    for disk_op, _ in self.op.disks:
7821
      if disk_op == constants.DDM_REMOVE:
7822
        if len(instance.disks) == 1:
7823
          raise errors.OpPrereqError("Cannot remove the last disk of"
7824
                                     " an instance",
7825
                                     errors.ECODE_INVAL)
7826
        ins_l = self.rpc.call_instance_list([pnode], [instance.hypervisor])
7827
        ins_l = ins_l[pnode]
7828
        msg = ins_l.fail_msg
7829
        if msg:
7830
          raise errors.OpPrereqError("Can't contact node %s: %s" %
7831
                                     (pnode, msg), errors.ECODE_ENVIRON)
7832
        if instance.name in ins_l.payload:
7833
          raise errors.OpPrereqError("Instance is running, can't remove"
7834
                                     " disks.", errors.ECODE_STATE)
7835

    
7836
      if (disk_op == constants.DDM_ADD and
7837
          len(instance.nics) >= constants.MAX_DISKS):
7838
        raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
7839
                                   " add more" % constants.MAX_DISKS,
7840
                                   errors.ECODE_STATE)
7841
      if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
7842
        # an existing disk
7843
        if disk_op < 0 or disk_op >= len(instance.disks):
7844
          raise errors.OpPrereqError("Invalid disk index %s, valid values"
7845
                                     " are 0 to %d" %
7846
                                     (disk_op, len(instance.disks)),
7847
                                     errors.ECODE_INVAL)
7848

    
7849
    return
7850

    
7851
  def Exec(self, feedback_fn):
7852
    """Modifies an instance.
7853

7854
    All parameters take effect only at the next restart of the instance.
7855

7856
    """
7857
    # Process here the warnings from CheckPrereq, as we don't have a
7858
    # feedback_fn there.
7859
    for warn in self.warn:
7860
      feedback_fn("WARNING: %s" % warn)
7861

    
7862
    result = []
7863
    instance = self.instance
7864
    # disk changes
7865
    for disk_op, disk_dict in self.op.disks:
7866
      if disk_op == constants.DDM_REMOVE:
7867
        # remove the last disk
7868
        device = instance.disks.pop()
7869
        device_idx = len(instance.disks)
7870
        for node, disk in device.ComputeNodeTree(instance.primary_node):
7871
          self.cfg.SetDiskID(disk, node)
7872
          msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
7873
          if msg:
7874
            self.LogWarning("Could not remove disk/%d on node %s: %s,"
7875
                            " continuing anyway", device_idx, node, msg)
7876
        result.append(("disk/%d" % device_idx, "remove"))
7877
      elif disk_op == constants.DDM_ADD:
7878
        # add a new disk
7879
        if instance.disk_template == constants.DT_FILE:
7880
          file_driver, file_path = instance.disks[0].logical_id
7881
          file_path = os.path.dirname(file_path)
7882
        else:
7883
          file_driver = file_path = None
7884
        disk_idx_base = len(instance.disks)
7885
        new_disk = _GenerateDiskTemplate(self,
7886
                                         instance.disk_template,
7887
                                         instance.name, instance.primary_node,
7888
                                         instance.secondary_nodes,
7889
                                         [disk_dict],
7890
                                         file_path,
7891
                                         file_driver,
7892
                                         disk_idx_base)[0]
7893
        instance.disks.append(new_disk)
7894
        info = _GetInstanceInfoText(instance)
7895

    
7896
        logging.info("Creating volume %s for instance %s",
7897
                     new_disk.iv_name, instance.name)
7898
        # Note: this needs to be kept in sync with _CreateDisks
7899
        #HARDCODE
7900
        for node in instance.all_nodes:
7901
          f_create = node == instance.primary_node
7902
          try:
7903
            _CreateBlockDev(self, node, instance, new_disk,
7904
                            f_create, info, f_create)
7905
          except errors.OpExecError, err:
7906
            self.LogWarning("Failed to create volume %s (%s) on"
7907
                            " node %s: %s",
7908
                            new_disk.iv_name, new_disk, node, err)
7909
        result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
7910
                       (new_disk.size, new_disk.mode)))
7911
      else:
7912
        # change a given disk
7913
        instance.disks[disk_op].mode = disk_dict['mode']
7914
        result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
7915
    # NIC changes
7916
    for nic_op, nic_dict in self.op.nics:
7917
      if nic_op == constants.DDM_REMOVE:
7918
        # remove the last nic
7919
        del instance.nics[-1]
7920
        result.append(("nic.%d" % len(instance.nics), "remove"))
7921
      elif nic_op == constants.DDM_ADD:
7922
        # mac and bridge should be set, by now
7923
        mac = nic_dict['mac']
7924
        ip = nic_dict.get('ip', None)
7925
        nicparams = self.nic_pinst[constants.DDM_ADD]
7926
        new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
7927
        instance.nics.append(new_nic)
7928
        result.append(("nic.%d" % (len(instance.nics) - 1),
7929
                       "add:mac=%s,ip=%s,mode=%s,link=%s" %
7930
                       (new_nic.mac, new_nic.ip,
7931
                        self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
7932
                        self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
7933
                       )))
7934
      else:
7935
        for key in 'mac', 'ip':
7936
          if key in nic_dict:
7937
            setattr(instance.nics[nic_op], key, nic_dict[key])
7938
        if nic_op in self.nic_pinst:
7939
          instance.nics[nic_op].nicparams = self.nic_pinst[nic_op]
7940
        for key, val in nic_dict.iteritems():
7941
          result.append(("nic.%s/%d" % (key, nic_op), val))
7942

    
7943
    # hvparams changes
7944
    if self.op.hvparams:
7945
      instance.hvparams = self.hv_inst
7946
      for key, val in self.op.hvparams.iteritems():
7947
        result.append(("hv/%s" % key, val))
7948

    
7949
    # beparams changes
7950
    if self.op.beparams:
7951
      instance.beparams = self.be_inst
7952
      for key, val in self.op.beparams.iteritems():
7953
        result.append(("be/%s" % key, val))
7954

    
7955
    self.cfg.Update(instance, feedback_fn)
7956

    
7957
    return result
7958

    
7959

    
7960
class LUQueryExports(NoHooksLU):
7961
  """Query the exports list
7962

7963
  """
7964
  _OP_REQP = ['nodes']
7965
  REQ_BGL = False
7966

    
7967
  def ExpandNames(self):
7968
    self.needed_locks = {}
7969
    self.share_locks[locking.LEVEL_NODE] = 1
7970
    if not self.op.nodes:
7971
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7972
    else:
7973
      self.needed_locks[locking.LEVEL_NODE] = \
7974
        _GetWantedNodes(self, self.op.nodes)
7975

    
7976
  def CheckPrereq(self):
7977
    """Check prerequisites.
7978

7979
    """
7980
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
7981

    
7982
  def Exec(self, feedback_fn):
7983
    """Compute the list of all the exported system images.
7984

7985
    @rtype: dict
7986
    @return: a dictionary with the structure node->(export-list)
7987
        where export-list is a list of the instances exported on
7988
        that node.
7989

7990
    """
7991
    rpcresult = self.rpc.call_export_list(self.nodes)
7992
    result = {}
7993
    for node in rpcresult:
7994
      if rpcresult[node].fail_msg:
7995
        result[node] = False
7996
      else:
7997
        result[node] = rpcresult[node].payload
7998

    
7999
    return result
8000

    
8001

    
8002
class LUExportInstance(LogicalUnit):
8003
  """Export an instance to an image in the cluster.
8004

8005
  """
8006
  HPATH = "instance-export"
8007
  HTYPE = constants.HTYPE_INSTANCE
8008
  _OP_REQP = ["instance_name", "target_node", "shutdown"]
8009
  REQ_BGL = False
8010

    
8011
  def CheckArguments(self):
8012
    """Check the arguments.
8013

8014
    """
8015
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
8016
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
8017

    
8018
  def ExpandNames(self):
8019
    self._ExpandAndLockInstance()
8020
    # FIXME: lock only instance primary and destination node
8021
    #
8022
    # Sad but true, for now we have do lock all nodes, as we don't know where
8023
    # the previous export might be, and and in this LU we search for it and
8024
    # remove it from its current node. In the future we could fix this by:
8025
    #  - making a tasklet to search (share-lock all), then create the new one,
8026
    #    then one to remove, after
8027
    #  - removing the removal operation altogether
8028
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
8029

    
8030
  def DeclareLocks(self, level):
8031
    """Last minute lock declaration."""
8032
    # All nodes are locked anyway, so nothing to do here.
8033

    
8034
  def BuildHooksEnv(self):
8035
    """Build hooks env.
8036

8037
    This will run on the master, primary node and target node.
8038

8039
    """
8040
    env = {
8041
      "EXPORT_NODE": self.op.target_node,
8042
      "EXPORT_DO_SHUTDOWN": self.op.shutdown,
8043
      "SHUTDOWN_TIMEOUT": self.shutdown_timeout,
8044
      }
8045
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
8046
    nl = [self.cfg.GetMasterNode(), self.instance.primary_node,
8047
          self.op.target_node]
8048
    return env, nl, nl
8049

    
8050
  def CheckPrereq(self):
8051
    """Check prerequisites.
8052

8053
    This checks that the instance and node names are valid.
8054

8055
    """
8056
    instance_name = self.op.instance_name
8057
    self.instance = self.cfg.GetInstanceInfo(instance_name)
8058
    assert self.instance is not None, \
8059
          "Cannot retrieve locked instance %s" % self.op.instance_name
8060
    _CheckNodeOnline(self, self.instance.primary_node)
8061

    
8062
    self.dst_node = self.cfg.GetNodeInfo(
8063
      self.cfg.ExpandNodeName(self.op.target_node))
8064

    
8065
    if self.dst_node is None:
8066
      # This is wrong node name, not a non-locked node
8067
      raise errors.OpPrereqError("Wrong node name %s" % self.op.target_node,
8068
                                 errors.ECODE_NOENT)
8069
    _CheckNodeOnline(self, self.dst_node.name)
8070
    _CheckNodeNotDrained(self, self.dst_node.name)
8071

    
8072
    # instance disk type verification
8073
    for disk in self.instance.disks:
8074
      if disk.dev_type == constants.LD_FILE:
8075
        raise errors.OpPrereqError("Export not supported for instances with"
8076
                                   " file-based disks", errors.ECODE_INVAL)
8077

    
8078
  def Exec(self, feedback_fn):
8079
    """Export an instance to an image in the cluster.
8080

8081
    """
8082
    instance = self.instance
8083
    dst_node = self.dst_node
8084
    src_node = instance.primary_node
8085

    
8086
    if self.op.shutdown:
8087
      # shutdown the instance, but not the disks
8088
      feedback_fn("Shutting down instance %s" % instance.name)
8089
      result = self.rpc.call_instance_shutdown(src_node, instance,
8090
                                               self.shutdown_timeout)
8091
      result.Raise("Could not shutdown instance %s on"
8092
                   " node %s" % (instance.name, src_node))
8093

    
8094
    vgname = self.cfg.GetVGName()
8095

    
8096
    snap_disks = []
8097

    
8098
    # set the disks ID correctly since call_instance_start needs the
8099
    # correct drbd minor to create the symlinks
8100
    for disk in instance.disks:
8101
      self.cfg.SetDiskID(disk, src_node)
8102

    
8103
    activate_disks = (not instance.admin_up)
8104

    
8105
    if activate_disks:
8106
      # Activate the instance disks if we'exporting a stopped instance
8107
      feedback_fn("Activating disks for %s" % instance.name)
8108
      _StartInstanceDisks(self, instance, None)
8109

    
8110
    try:
8111
      # per-disk results
8112
      dresults = []
8113
      try:
8114
        for idx, disk in enumerate(instance.disks):
8115
          feedback_fn("Creating a snapshot of disk/%s on node %s" %
8116
                      (idx, src_node))
8117

    
8118
          # result.payload will be a snapshot of an lvm leaf of the one we
8119
          # passed
8120
          result = self.rpc.call_blockdev_snapshot(src_node, disk)
8121
          msg = result.fail_msg
8122
          if msg:
8123
            self.LogWarning("Could not snapshot disk/%s on node %s: %s",
8124
                            idx, src_node, msg)
8125
            snap_disks.append(False)
8126
          else:
8127
            disk_id = (vgname, result.payload)
8128
            new_dev = objects.Disk(dev_type=constants.LD_LV, size=disk.size,
8129
                                   logical_id=disk_id, physical_id=disk_id,
8130
                                   iv_name=disk.iv_name)
8131
            snap_disks.append(new_dev)
8132

    
8133
      finally:
8134
        if self.op.shutdown and instance.admin_up:
8135
          feedback_fn("Starting instance %s" % instance.name)
8136
          result = self.rpc.call_instance_start(src_node, instance, None, None)
8137
          msg = result.fail_msg
8138
          if msg:
8139
            _ShutdownInstanceDisks(self, instance)
8140
            raise errors.OpExecError("Could not start instance: %s" % msg)
8141

    
8142
      # TODO: check for size
8143

    
8144
      cluster_name = self.cfg.GetClusterName()
8145
      for idx, dev in enumerate(snap_disks):
8146
        feedback_fn("Exporting snapshot %s from %s to %s" %
8147
                    (idx, src_node, dst_node.name))
8148
        if dev:
8149
          result = self.rpc.call_snapshot_export(src_node, dev, dst_node.name,
8150
                                                 instance, cluster_name, idx)
8151
          msg = result.fail_msg
8152
          if msg:
8153
            self.LogWarning("Could not export disk/%s from node %s to"
8154
                            " node %s: %s", idx, src_node, dst_node.name, msg)
8155
            dresults.append(False)
8156
          else:
8157
            dresults.append(True)
8158
          msg = self.rpc.call_blockdev_remove(src_node, dev).fail_msg
8159
          if msg:
8160
            self.LogWarning("Could not remove snapshot for disk/%d from node"
8161
                            " %s: %s", idx, src_node, msg)
8162
        else:
8163
          dresults.append(False)
8164

    
8165
      feedback_fn("Finalizing export on %s" % dst_node.name)
8166
      result = self.rpc.call_finalize_export(dst_node.name, instance,
8167
                                             snap_disks)
8168
      fin_resu = True
8169
      msg = result.fail_msg
8170
      if msg:
8171
        self.LogWarning("Could not finalize export for instance %s"
8172
                        " on node %s: %s", instance.name, dst_node.name, msg)
8173
        fin_resu = False
8174

    
8175
    finally:
8176
      if activate_disks:
8177
        feedback_fn("Deactivating disks for %s" % instance.name)
8178
        _ShutdownInstanceDisks(self, instance)
8179

    
8180
    nodelist = self.cfg.GetNodeList()
8181
    nodelist.remove(dst_node.name)
8182

    
8183
    # on one-node clusters nodelist will be empty after the removal
8184
    # if we proceed the backup would be removed because OpQueryExports
8185
    # substitutes an empty list with the full cluster node list.
8186
    iname = instance.name
8187
    if nodelist:
8188
      feedback_fn("Removing old exports for instance %s" % iname)
8189
      exportlist = self.rpc.call_export_list(nodelist)
8190
      for node in exportlist:
8191
        if exportlist[node].fail_msg:
8192
          continue
8193
        if iname in exportlist[node].payload:
8194
          msg = self.rpc.call_export_remove(node, iname).fail_msg
8195
          if msg:
8196
            self.LogWarning("Could not remove older export for instance %s"
8197
                            " on node %s: %s", iname, node, msg)
8198
    return fin_resu, dresults
8199

    
8200

    
8201
class LURemoveExport(NoHooksLU):
8202
  """Remove exports related to the named instance.
8203

8204
  """
8205
  _OP_REQP = ["instance_name"]
8206
  REQ_BGL = False
8207

    
8208
  def ExpandNames(self):
8209
    self.needed_locks = {}
8210
    # We need all nodes to be locked in order for RemoveExport to work, but we
8211
    # don't need to lock the instance itself, as nothing will happen to it (and
8212
    # we can remove exports also for a removed instance)
8213
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
8214

    
8215
  def CheckPrereq(self):
8216
    """Check prerequisites.
8217
    """
8218
    pass
8219

    
8220
  def Exec(self, feedback_fn):
8221
    """Remove any export.
8222

8223
    """
8224
    instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
8225
    # If the instance was not found we'll try with the name that was passed in.
8226
    # This will only work if it was an FQDN, though.
8227
    fqdn_warn = False
8228
    if not instance_name:
8229
      fqdn_warn = True
8230
      instance_name = self.op.instance_name
8231

    
8232
    locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
8233
    exportlist = self.rpc.call_export_list(locked_nodes)
8234
    found = False
8235
    for node in exportlist:
8236
      msg = exportlist[node].fail_msg
8237
      if msg:
8238
        self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
8239
        continue
8240
      if instance_name in exportlist[node].payload:
8241
        found = True
8242
        result = self.rpc.call_export_remove(node, instance_name)
8243
        msg = result.fail_msg
8244
        if msg:
8245
          logging.error("Could not remove export for instance %s"
8246
                        " on node %s: %s", instance_name, node, msg)
8247

    
8248
    if fqdn_warn and not found:
8249
      feedback_fn("Export not found. If trying to remove an export belonging"
8250
                  " to a deleted instance please use its Fully Qualified"
8251
                  " Domain Name.")
8252

    
8253

    
8254
class TagsLU(NoHooksLU): # pylint: disable-msg=W0223
8255
  """Generic tags LU.
8256

8257
  This is an abstract class which is the parent of all the other tags LUs.
8258

8259
  """
8260

    
8261
  def ExpandNames(self):
8262
    self.needed_locks = {}
8263
    if self.op.kind == constants.TAG_NODE:
8264
      name = self.cfg.ExpandNodeName(self.op.name)
8265
      if name is None:
8266
        raise errors.OpPrereqError("Invalid node name (%s)" %
8267
                                   (self.op.name,), errors.ECODE_NOENT)
8268
      self.op.name = name
8269
      self.needed_locks[locking.LEVEL_NODE] = name
8270
    elif self.op.kind == constants.TAG_INSTANCE:
8271
      name = self.cfg.ExpandInstanceName(self.op.name)
8272
      if name is None:
8273
        raise errors.OpPrereqError("Invalid instance name (%s)" %
8274
                                   (self.op.name,), errors.ECODE_NOENT)
8275
      self.op.name = name
8276
      self.needed_locks[locking.LEVEL_INSTANCE] = name
8277

    
8278
  def CheckPrereq(self):
8279
    """Check prerequisites.
8280

8281
    """
8282
    if self.op.kind == constants.TAG_CLUSTER:
8283
      self.target = self.cfg.GetClusterInfo()
8284
    elif self.op.kind == constants.TAG_NODE:
8285
      self.target = self.cfg.GetNodeInfo(self.op.name)
8286
    elif self.op.kind == constants.TAG_INSTANCE:
8287
      self.target = self.cfg.GetInstanceInfo(self.op.name)
8288
    else:
8289
      raise errors.OpPrereqError("Wrong tag type requested (%s)" %
8290
                                 str(self.op.kind), errors.ECODE_INVAL)
8291

    
8292

    
8293
class LUGetTags(TagsLU):
8294
  """Returns the tags of a given object.
8295

8296
  """
8297
  _OP_REQP = ["kind", "name"]
8298
  REQ_BGL = False
8299

    
8300
  def Exec(self, feedback_fn):
8301
    """Returns the tag list.
8302

8303
    """
8304
    return list(self.target.GetTags())
8305

    
8306

    
8307
class LUSearchTags(NoHooksLU):
8308
  """Searches the tags for a given pattern.
8309

8310
  """
8311
  _OP_REQP = ["pattern"]
8312
  REQ_BGL = False
8313

    
8314
  def ExpandNames(self):
8315
    self.needed_locks = {}
8316

    
8317
  def CheckPrereq(self):
8318
    """Check prerequisites.
8319

8320
    This checks the pattern passed for validity by compiling it.
8321

8322
    """
8323
    try:
8324
      self.re = re.compile(self.op.pattern)
8325
    except re.error, err:
8326
      raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
8327
                                 (self.op.pattern, err), errors.ECODE_INVAL)
8328

    
8329
  def Exec(self, feedback_fn):
8330
    """Returns the tag list.
8331

8332
    """
8333
    cfg = self.cfg
8334
    tgts = [("/cluster", cfg.GetClusterInfo())]
8335
    ilist = cfg.GetAllInstancesInfo().values()
8336
    tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
8337
    nlist = cfg.GetAllNodesInfo().values()
8338
    tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
8339
    results = []
8340
    for path, target in tgts:
8341
      for tag in target.GetTags():
8342
        if self.re.search(tag):
8343
          results.append((path, tag))
8344
    return results
8345

    
8346

    
8347
class LUAddTags(TagsLU):
8348
  """Sets a tag on a given object.
8349

8350
  """
8351
  _OP_REQP = ["kind", "name", "tags"]
8352
  REQ_BGL = False
8353

    
8354
  def CheckPrereq(self):
8355
    """Check prerequisites.
8356

8357
    This checks the type and length of the tag name and value.
8358

8359
    """
8360
    TagsLU.CheckPrereq(self)
8361
    for tag in self.op.tags:
8362
      objects.TaggableObject.ValidateTag(tag)
8363

    
8364
  def Exec(self, feedback_fn):
8365
    """Sets the tag.
8366

8367
    """
8368
    try:
8369
      for tag in self.op.tags:
8370
        self.target.AddTag(tag)
8371
    except errors.TagError, err:
8372
      raise errors.OpExecError("Error while setting tag: %s" % str(err))
8373
    self.cfg.Update(self.target, feedback_fn)
8374

    
8375

    
8376
class LUDelTags(TagsLU):
8377
  """Delete a list of tags from a given object.
8378

8379
  """
8380
  _OP_REQP = ["kind", "name", "tags"]
8381
  REQ_BGL = False
8382

    
8383
  def CheckPrereq(self):
8384
    """Check prerequisites.
8385

8386
    This checks that we have the given tag.
8387

8388
    """
8389
    TagsLU.CheckPrereq(self)
8390
    for tag in self.op.tags:
8391
      objects.TaggableObject.ValidateTag(tag)
8392
    del_tags = frozenset(self.op.tags)
8393
    cur_tags = self.target.GetTags()
8394
    if not del_tags <= cur_tags:
8395
      diff_tags = del_tags - cur_tags
8396
      diff_names = ["'%s'" % tag for tag in diff_tags]
8397
      diff_names.sort()
8398
      raise errors.OpPrereqError("Tag(s) %s not found" %
8399
                                 (",".join(diff_names)), errors.ECODE_NOENT)
8400

    
8401
  def Exec(self, feedback_fn):
8402
    """Remove the tag from the object.
8403

8404
    """
8405
    for tag in self.op.tags:
8406
      self.target.RemoveTag(tag)
8407
    self.cfg.Update(self.target, feedback_fn)
8408

    
8409

    
8410
class LUTestDelay(NoHooksLU):
8411
  """Sleep for a specified amount of time.
8412

8413
  This LU sleeps on the master and/or nodes for a specified amount of
8414
  time.
8415

8416
  """
8417
  _OP_REQP = ["duration", "on_master", "on_nodes"]
8418
  REQ_BGL = False
8419

    
8420
  def ExpandNames(self):
8421
    """Expand names and set required locks.
8422

8423
    This expands the node list, if any.
8424

8425
    """
8426
    self.needed_locks = {}
8427
    if self.op.on_nodes:
8428
      # _GetWantedNodes can be used here, but is not always appropriate to use
8429
      # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
8430
      # more information.
8431
      self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
8432
      self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
8433

    
8434
  def CheckPrereq(self):
8435
    """Check prerequisites.
8436

8437
    """
8438

    
8439
  def Exec(self, feedback_fn):
8440
    """Do the actual sleep.
8441

8442
    """
8443
    if self.op.on_master:
8444
      if not utils.TestDelay(self.op.duration):
8445
        raise errors.OpExecError("Error during master delay test")
8446
    if self.op.on_nodes:
8447
      result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
8448
      for node, node_result in result.items():
8449
        node_result.Raise("Failure during rpc call to node %s" % node)
8450

    
8451

    
8452
class IAllocator(object):
8453
  """IAllocator framework.
8454

8455
  An IAllocator instance has three sets of attributes:
8456
    - cfg that is needed to query the cluster
8457
    - input data (all members of the _KEYS class attribute are required)
8458
    - four buffer attributes (in|out_data|text), that represent the
8459
      input (to the external script) in text and data structure format,
8460
      and the output from it, again in two formats
8461
    - the result variables from the script (success, info, nodes) for
8462
      easy usage
8463

8464
  """
8465
  # pylint: disable-msg=R0902
8466
  # lots of instance attributes
8467
  _ALLO_KEYS = [
8468
    "mem_size", "disks", "disk_template",
8469
    "os", "tags", "nics", "vcpus", "hypervisor",
8470
    ]
8471
  _RELO_KEYS = [
8472
    "relocate_from",
8473
    ]
8474

    
8475
  def __init__(self, cfg, rpc, mode, name, **kwargs):
8476
    self.cfg = cfg
8477
    self.rpc = rpc
8478
    # init buffer variables
8479
    self.in_text = self.out_text = self.in_data = self.out_data = None
8480
    # init all input fields so that pylint is happy
8481
    self.mode = mode
8482
    self.name = name
8483
    self.mem_size = self.disks = self.disk_template = None
8484
    self.os = self.tags = self.nics = self.vcpus = None
8485
    self.hypervisor = None
8486
    self.relocate_from = None
8487
    # computed fields
8488
    self.required_nodes = None
8489
    # init result fields
8490
    self.success = self.info = self.nodes = None
8491
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
8492
      keyset = self._ALLO_KEYS
8493
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
8494
      keyset = self._RELO_KEYS
8495
    else:
8496
      raise errors.ProgrammerError("Unknown mode '%s' passed to the"
8497
                                   " IAllocator" % self.mode)
8498
    for key in kwargs:
8499
      if key not in keyset:
8500
        raise errors.ProgrammerError("Invalid input parameter '%s' to"
8501
                                     " IAllocator" % key)
8502
      setattr(self, key, kwargs[key])
8503
    for key in keyset:
8504
      if key not in kwargs:
8505
        raise errors.ProgrammerError("Missing input parameter '%s' to"
8506
                                     " IAllocator" % key)
8507
    self._BuildInputData()
8508

    
8509
  def _ComputeClusterData(self):
8510
    """Compute the generic allocator input data.
8511

8512
    This is the data that is independent of the actual operation.
8513

8514
    """
8515
    cfg = self.cfg
8516
    cluster_info = cfg.GetClusterInfo()
8517
    # cluster data
8518
    data = {
8519
      "version": constants.IALLOCATOR_VERSION,
8520
      "cluster_name": cfg.GetClusterName(),
8521
      "cluster_tags": list(cluster_info.GetTags()),
8522
      "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
8523
      # we don't have job IDs
8524
      }
8525
    iinfo = cfg.GetAllInstancesInfo().values()
8526
    i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
8527

    
8528
    # node data
8529
    node_results = {}
8530
    node_list = cfg.GetNodeList()
8531

    
8532
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
8533
      hypervisor_name = self.hypervisor
8534
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
8535
      hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
8536

    
8537
    node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(),
8538
                                        hypervisor_name)
8539
    node_iinfo = \
8540
      self.rpc.call_all_instances_info(node_list,
8541
                                       cluster_info.enabled_hypervisors)
8542
    for nname, nresult in node_data.items():
8543
      # first fill in static (config-based) values
8544
      ninfo = cfg.GetNodeInfo(nname)
8545
      pnr = {
8546
        "tags": list(ninfo.GetTags()),
8547
        "primary_ip": ninfo.primary_ip,
8548
        "secondary_ip": ninfo.secondary_ip,
8549
        "offline": ninfo.offline,
8550
        "drained": ninfo.drained,
8551
        "master_candidate": ninfo.master_candidate,
8552
        }
8553

    
8554
      if not (ninfo.offline or ninfo.drained):
8555
        nresult.Raise("Can't get data for node %s" % nname)
8556
        node_iinfo[nname].Raise("Can't get node instance info from node %s" %
8557
                                nname)
8558
        remote_info = nresult.payload
8559

    
8560
        for attr in ['memory_total', 'memory_free', 'memory_dom0',
8561
                     'vg_size', 'vg_free', 'cpu_total']:
8562
          if attr not in remote_info:
8563
            raise errors.OpExecError("Node '%s' didn't return attribute"
8564
                                     " '%s'" % (nname, attr))
8565
          if not isinstance(remote_info[attr], int):
8566
            raise errors.OpExecError("Node '%s' returned invalid value"
8567
                                     " for '%s': %s" %
8568
                                     (nname, attr, remote_info[attr]))
8569
        # compute memory used by primary instances
8570
        i_p_mem = i_p_up_mem = 0
8571
        for iinfo, beinfo in i_list:
8572
          if iinfo.primary_node == nname:
8573
            i_p_mem += beinfo[constants.BE_MEMORY]
8574
            if iinfo.name not in node_iinfo[nname].payload:
8575
              i_used_mem = 0
8576
            else:
8577
              i_used_mem = int(node_iinfo[nname].payload[iinfo.name]['memory'])
8578
            i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
8579
            remote_info['memory_free'] -= max(0, i_mem_diff)
8580

    
8581
            if iinfo.admin_up:
8582
              i_p_up_mem += beinfo[constants.BE_MEMORY]
8583

    
8584
        # compute memory used by instances
8585
        pnr_dyn = {
8586
          "total_memory": remote_info['memory_total'],
8587
          "reserved_memory": remote_info['memory_dom0'],
8588
          "free_memory": remote_info['memory_free'],
8589
          "total_disk": remote_info['vg_size'],
8590
          "free_disk": remote_info['vg_free'],
8591
          "total_cpus": remote_info['cpu_total'],
8592
          "i_pri_memory": i_p_mem,
8593
          "i_pri_up_memory": i_p_up_mem,
8594
          }
8595
        pnr.update(pnr_dyn)
8596

    
8597
      node_results[nname] = pnr
8598
    data["nodes"] = node_results
8599

    
8600
    # instance data
8601
    instance_data = {}
8602
    for iinfo, beinfo in i_list:
8603
      nic_data = []
8604
      for nic in iinfo.nics:
8605
        filled_params = objects.FillDict(
8606
            cluster_info.nicparams[constants.PP_DEFAULT],
8607
            nic.nicparams)
8608
        nic_dict = {"mac": nic.mac,
8609
                    "ip": nic.ip,
8610
                    "mode": filled_params[constants.NIC_MODE],
8611
                    "link": filled_params[constants.NIC_LINK],
8612
                   }
8613
        if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
8614
          nic_dict["bridge"] = filled_params[constants.NIC_LINK]
8615
        nic_data.append(nic_dict)
8616
      pir = {
8617
        "tags": list(iinfo.GetTags()),
8618
        "admin_up": iinfo.admin_up,
8619
        "vcpus": beinfo[constants.BE_VCPUS],
8620
        "memory": beinfo[constants.BE_MEMORY],
8621
        "os": iinfo.os,
8622
        "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
8623
        "nics": nic_data,
8624
        "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
8625
        "disk_template": iinfo.disk_template,
8626
        "hypervisor": iinfo.hypervisor,
8627
        }
8628
      pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
8629
                                                 pir["disks"])
8630
      instance_data[iinfo.name] = pir
8631

    
8632
    data["instances"] = instance_data
8633

    
8634
    self.in_data = data
8635

    
8636
  def _AddNewInstance(self):
8637
    """Add new instance data to allocator structure.
8638

8639
    This in combination with _AllocatorGetClusterData will create the
8640
    correct structure needed as input for the allocator.
8641

8642
    The checks for the completeness of the opcode must have already been
8643
    done.
8644

8645
    """
8646
    data = self.in_data
8647

    
8648
    disk_space = _ComputeDiskSize(self.disk_template, self.disks)
8649

    
8650
    if self.disk_template in constants.DTS_NET_MIRROR:
8651
      self.required_nodes = 2
8652
    else:
8653
      self.required_nodes = 1
8654
    request = {
8655
      "type": "allocate",
8656
      "name": self.name,
8657
      "disk_template": self.disk_template,
8658
      "tags": self.tags,
8659
      "os": self.os,
8660
      "vcpus": self.vcpus,
8661
      "memory": self.mem_size,
8662
      "disks": self.disks,
8663
      "disk_space_total": disk_space,
8664
      "nics": self.nics,
8665
      "required_nodes": self.required_nodes,
8666
      }
8667
    data["request"] = request
8668

    
8669
  def _AddRelocateInstance(self):
8670
    """Add relocate instance data to allocator structure.
8671

8672
    This in combination with _IAllocatorGetClusterData will create the
8673
    correct structure needed as input for the allocator.
8674

8675
    The checks for the completeness of the opcode must have already been
8676
    done.
8677

8678
    """
8679
    instance = self.cfg.GetInstanceInfo(self.name)
8680
    if instance is None:
8681
      raise errors.ProgrammerError("Unknown instance '%s' passed to"
8682
                                   " IAllocator" % self.name)
8683

    
8684
    if instance.disk_template not in constants.DTS_NET_MIRROR:
8685
      raise errors.OpPrereqError("Can't relocate non-mirrored instances",
8686
                                 errors.ECODE_INVAL)
8687

    
8688
    if len(instance.secondary_nodes) != 1:
8689
      raise errors.OpPrereqError("Instance has not exactly one secondary node",
8690
                                 errors.ECODE_STATE)
8691

    
8692
    self.required_nodes = 1
8693
    disk_sizes = [{'size': disk.size} for disk in instance.disks]
8694
    disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
8695

    
8696
    request = {
8697
      "type": "relocate",
8698
      "name": self.name,
8699
      "disk_space_total": disk_space,
8700
      "required_nodes": self.required_nodes,
8701
      "relocate_from": self.relocate_from,
8702
      }
8703
    self.in_data["request"] = request
8704

    
8705
  def _BuildInputData(self):
8706
    """Build input data structures.
8707

8708
    """
8709
    self._ComputeClusterData()
8710

    
8711
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
8712
      self._AddNewInstance()
8713
    else:
8714
      self._AddRelocateInstance()
8715

    
8716
    self.in_text = serializer.Dump(self.in_data)
8717

    
8718
  def Run(self, name, validate=True, call_fn=None):
8719
    """Run an instance allocator and return the results.
8720

8721
    """
8722
    if call_fn is None:
8723
      call_fn = self.rpc.call_iallocator_runner
8724

    
8725
    result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
8726
    result.Raise("Failure while running the iallocator script")
8727

    
8728
    self.out_text = result.payload
8729
    if validate:
8730
      self._ValidateResult()
8731

    
8732
  def _ValidateResult(self):
8733
    """Process the allocator results.
8734

8735
    This will process and if successful save the result in
8736
    self.out_data and the other parameters.
8737

8738
    """
8739
    try:
8740
      rdict = serializer.Load(self.out_text)
8741
    except Exception, err:
8742
      raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
8743

    
8744
    if not isinstance(rdict, dict):
8745
      raise errors.OpExecError("Can't parse iallocator results: not a dict")
8746

    
8747
    for key in "success", "info", "nodes":
8748
      if key not in rdict:
8749
        raise errors.OpExecError("Can't parse iallocator results:"
8750
                                 " missing key '%s'" % key)
8751
      setattr(self, key, rdict[key])
8752

    
8753
    if not isinstance(rdict["nodes"], list):
8754
      raise errors.OpExecError("Can't parse iallocator results: 'nodes' key"
8755
                               " is not a list")
8756
    self.out_data = rdict
8757

    
8758

    
8759
class LUTestAllocator(NoHooksLU):
8760
  """Run allocator tests.
8761

8762
  This LU runs the allocator tests
8763

8764
  """
8765
  _OP_REQP = ["direction", "mode", "name"]
8766

    
8767
  def CheckPrereq(self):
8768
    """Check prerequisites.
8769

8770
    This checks the opcode parameters depending on the director and mode test.
8771

8772
    """
8773
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
8774
      for attr in ["name", "mem_size", "disks", "disk_template",
8775
                   "os", "tags", "nics", "vcpus"]:
8776
        if not hasattr(self.op, attr):
8777
          raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
8778
                                     attr, errors.ECODE_INVAL)
8779
      iname = self.cfg.ExpandInstanceName(self.op.name)
8780
      if iname is not None:
8781
        raise errors.OpPrereqError("Instance '%s' already in the cluster" %
8782
                                   iname, errors.ECODE_EXISTS)
8783
      if not isinstance(self.op.nics, list):
8784
        raise errors.OpPrereqError("Invalid parameter 'nics'",
8785
                                   errors.ECODE_INVAL)
8786
      for row in self.op.nics:
8787
        if (not isinstance(row, dict) or
8788
            "mac" not in row or
8789
            "ip" not in row or
8790
            "bridge" not in row):
8791
          raise errors.OpPrereqError("Invalid contents of the 'nics'"
8792
                                     " parameter", errors.ECODE_INVAL)
8793
      if not isinstance(self.op.disks, list):
8794
        raise errors.OpPrereqError("Invalid parameter 'disks'",
8795
                                   errors.ECODE_INVAL)
8796
      for row in self.op.disks:
8797
        if (not isinstance(row, dict) or
8798
            "size" not in row or
8799
            not isinstance(row["size"], int) or
8800
            "mode" not in row or
8801
            row["mode"] not in ['r', 'w']):
8802
          raise errors.OpPrereqError("Invalid contents of the 'disks'"
8803
                                     " parameter", errors.ECODE_INVAL)
8804
      if not hasattr(self.op, "hypervisor") or self.op.hypervisor is None:
8805
        self.op.hypervisor = self.cfg.GetHypervisorType()
8806
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
8807
      if not hasattr(self.op, "name"):
8808
        raise errors.OpPrereqError("Missing attribute 'name' on opcode input",
8809
                                   errors.ECODE_INVAL)
8810
      fname = self.cfg.ExpandInstanceName(self.op.name)
8811
      if fname is None:
8812
        raise errors.OpPrereqError("Instance '%s' not found for relocation" %
8813
                                   self.op.name, errors.ECODE_NOENT)
8814
      self.op.name = fname
8815
      self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
8816
    else:
8817
      raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
8818
                                 self.op.mode, errors.ECODE_INVAL)
8819

    
8820
    if self.op.direction == constants.IALLOCATOR_DIR_OUT:
8821
      if not hasattr(self.op, "allocator") or self.op.allocator is None:
8822
        raise errors.OpPrereqError("Missing allocator name",
8823
                                   errors.ECODE_INVAL)
8824
    elif self.op.direction != constants.IALLOCATOR_DIR_IN:
8825
      raise errors.OpPrereqError("Wrong allocator test '%s'" %
8826
                                 self.op.direction, errors.ECODE_INVAL)
8827

    
8828
  def Exec(self, feedback_fn):
8829
    """Run the allocator test.
8830

8831
    """
8832
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
8833
      ial = IAllocator(self.cfg, self.rpc,
8834
                       mode=self.op.mode,
8835
                       name=self.op.name,
8836
                       mem_size=self.op.mem_size,
8837
                       disks=self.op.disks,
8838
                       disk_template=self.op.disk_template,
8839
                       os=self.op.os,
8840
                       tags=self.op.tags,
8841
                       nics=self.op.nics,
8842
                       vcpus=self.op.vcpus,
8843
                       hypervisor=self.op.hypervisor,
8844
                       )
8845
    else:
8846
      ial = IAllocator(self.cfg, self.rpc,
8847
                       mode=self.op.mode,
8848
                       name=self.op.name,
8849
                       relocate_from=list(self.relocate_from),
8850
                       )
8851

    
8852
    if self.op.direction == constants.IALLOCATOR_DIR_IN:
8853
      result = ial.in_text
8854
    else:
8855
      ial.Run(self.op.allocator, validate=False)
8856
      result = ial.out_text
8857
    return result