Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ 08eec276

History | View | Annotate | Download (312.7 kB)

1
#
2
#
3

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

    
21

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

    
24
# pylint: disable-msg=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
    # support for generic debug attribute
99
    if (not hasattr(self.op, "debug_level") or
100
        not isinstance(self.op.debug_level, int)):
101
      self.op.debug_level = 0
102

    
103
    # Tasklets
104
    self.tasklets = None
105

    
106
    for attr_name in self._OP_REQP:
107
      attr_val = getattr(op, attr_name, None)
108
      if attr_val is None:
109
        raise errors.OpPrereqError("Required parameter '%s' missing" %
110
                                   attr_name, errors.ECODE_INVAL)
111

    
112
    self.CheckArguments()
113

    
114
  def __GetSSH(self):
115
    """Returns the SshRunner object
116

117
    """
118
    if not self.__ssh:
119
      self.__ssh = ssh.SshRunner(self.cfg.GetClusterName())
120
    return self.__ssh
121

    
122
  ssh = property(fget=__GetSSH)
123

    
124
  def CheckArguments(self):
125
    """Check syntactic validity for the opcode arguments.
126

127
    This method is for doing a simple syntactic check and ensure
128
    validity of opcode parameters, without any cluster-related
129
    checks. While the same can be accomplished in ExpandNames and/or
130
    CheckPrereq, doing these separate is better because:
131

132
      - ExpandNames is left as as purely a lock-related function
133
      - CheckPrereq is run after we have acquired locks (and possible
134
        waited for them)
135

136
    The function is allowed to change the self.op attribute so that
137
    later methods can no longer worry about missing parameters.
138

139
    """
140
    pass
141

    
142
  def ExpandNames(self):
143
    """Expand names for this LU.
144

145
    This method is called before starting to execute the opcode, and it should
146
    update all the parameters of the opcode to their canonical form (e.g. a
147
    short node name must be fully expanded after this method has successfully
148
    completed). This way locking, hooks, logging, ecc. can work correctly.
149

150
    LUs which implement this method must also populate the self.needed_locks
151
    member, as a dict with lock levels as keys, and a list of needed lock names
152
    as values. Rules:
153

154
      - use an empty dict if you don't need any lock
155
      - if you don't need any lock at a particular level omit that level
156
      - don't put anything for the BGL level
157
      - if you want all locks at a level use locking.ALL_SET as a value
158

159
    If you need to share locks (rather than acquire them exclusively) at one
160
    level you can modify self.share_locks, setting a true value (usually 1) for
161
    that level. By default locks are not shared.
162

163
    This function can also define a list of tasklets, which then will be
164
    executed in order instead of the usual LU-level CheckPrereq and Exec
165
    functions, if those are not defined by the LU.
166

167
    Examples::
168

169
      # Acquire all nodes and one instance
170
      self.needed_locks = {
171
        locking.LEVEL_NODE: locking.ALL_SET,
172
        locking.LEVEL_INSTANCE: ['instance1.example.tld'],
173
      }
174
      # Acquire just two nodes
175
      self.needed_locks = {
176
        locking.LEVEL_NODE: ['node1.example.tld', 'node2.example.tld'],
177
      }
178
      # Acquire no locks
179
      self.needed_locks = {} # No, you can't leave it to the default value None
180

181
    """
182
    # The implementation of this method is mandatory only if the new LU is
183
    # concurrent, so that old LUs don't need to be changed all at the same
184
    # time.
185
    if self.REQ_BGL:
186
      self.needed_locks = {} # Exclusive LUs don't need locks.
187
    else:
188
      raise NotImplementedError
189

    
190
  def DeclareLocks(self, level):
191
    """Declare LU locking needs for a level
192

193
    While most LUs can just declare their locking needs at ExpandNames time,
194
    sometimes there's the need to calculate some locks after having acquired
195
    the ones before. This function is called just before acquiring locks at a
196
    particular level, but after acquiring the ones at lower levels, and permits
197
    such calculations. It can be used to modify self.needed_locks, and by
198
    default it does nothing.
199

200
    This function is only called if you have something already set in
201
    self.needed_locks for the level.
202

203
    @param level: Locking level which is going to be locked
204
    @type level: member of ganeti.locking.LEVELS
205

206
    """
207

    
208
  def CheckPrereq(self):
209
    """Check prerequisites for this LU.
210

211
    This method should check that the prerequisites for the execution
212
    of this LU are fulfilled. It can do internode communication, but
213
    it should be idempotent - no cluster or system changes are
214
    allowed.
215

216
    The method should raise errors.OpPrereqError in case something is
217
    not fulfilled. Its return value is ignored.
218

219
    This method should also update all the parameters of the opcode to
220
    their canonical form if it hasn't been done by ExpandNames before.
221

222
    """
223
    if self.tasklets is not None:
224
      for (idx, tl) in enumerate(self.tasklets):
225
        logging.debug("Checking prerequisites for tasklet %s/%s",
226
                      idx + 1, len(self.tasklets))
227
        tl.CheckPrereq()
228
    else:
229
      raise NotImplementedError
230

    
231
  def Exec(self, feedback_fn):
232
    """Execute the LU.
233

234
    This method should implement the actual work. It should raise
235
    errors.OpExecError for failures that are somewhat dealt with in
236
    code, or expected.
237

238
    """
239
    if self.tasklets is not None:
240
      for (idx, tl) in enumerate(self.tasklets):
241
        logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets))
242
        tl.Exec(feedback_fn)
243
    else:
244
      raise NotImplementedError
245

    
246
  def BuildHooksEnv(self):
247
    """Build hooks environment for this LU.
248

249
    This method should return a three-node tuple consisting of: a dict
250
    containing the environment that will be used for running the
251
    specific hook for this LU, a list of node names on which the hook
252
    should run before the execution, and a list of node names on which
253
    the hook should run after the execution.
254

255
    The keys of the dict must not have 'GANETI_' prefixed as this will
256
    be handled in the hooks runner. Also note additional keys will be
257
    added by the hooks runner. If the LU doesn't define any
258
    environment, an empty dict (and not None) should be returned.
259

260
    No nodes should be returned as an empty list (and not None).
261

262
    Note that if the HPATH for a LU class is None, this function will
263
    not be called.
264

265
    """
266
    raise NotImplementedError
267

    
268
  def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
269
    """Notify the LU about the results of its hooks.
270

271
    This method is called every time a hooks phase is executed, and notifies
272
    the Logical Unit about the hooks' result. The LU can then use it to alter
273
    its result based on the hooks.  By default the method does nothing and the
274
    previous result is passed back unchanged but any LU can define it if it
275
    wants to use the local cluster hook-scripts somehow.
276

277
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
278
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
279
    @param hook_results: the results of the multi-node hooks rpc call
280
    @param feedback_fn: function used send feedback back to the caller
281
    @param lu_result: the previous Exec result this LU had, or None
282
        in the PRE phase
283
    @return: the new Exec result, based on the previous result
284
        and hook results
285

286
    """
287
    # API must be kept, thus we ignore the unused argument and could
288
    # be a function warnings
289
    # pylint: disable-msg=W0613,R0201
290
    return lu_result
291

    
292
  def _ExpandAndLockInstance(self):
293
    """Helper function to expand and lock an instance.
294

295
    Many LUs that work on an instance take its name in self.op.instance_name
296
    and need to expand it and then declare the expanded name for locking. This
297
    function does it, and then updates self.op.instance_name to the expanded
298
    name. It also initializes needed_locks as a dict, if this hasn't been done
299
    before.
300

301
    """
302
    if self.needed_locks is None:
303
      self.needed_locks = {}
304
    else:
305
      assert locking.LEVEL_INSTANCE not in self.needed_locks, \
306
        "_ExpandAndLockInstance called with instance-level locks set"
307
    expanded_name = self.cfg.ExpandInstanceName(self.op.instance_name)
308
    if expanded_name is None:
309
      raise errors.OpPrereqError("Instance '%s' not known" %
310
                                 self.op.instance_name, errors.ECODE_NOENT)
311
    self.needed_locks[locking.LEVEL_INSTANCE] = expanded_name
312
    self.op.instance_name = expanded_name
313

    
314
  def _LockInstancesNodes(self, primary_only=False):
315
    """Helper function to declare instances' nodes for locking.
316

317
    This function should be called after locking one or more instances to lock
318
    their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
319
    with all primary or secondary nodes for instances already locked and
320
    present in self.needed_locks[locking.LEVEL_INSTANCE].
321

322
    It should be called from DeclareLocks, and for safety only works if
323
    self.recalculate_locks[locking.LEVEL_NODE] is set.
324

325
    In the future it may grow parameters to just lock some instance's nodes, or
326
    to just lock primaries or secondary nodes, if needed.
327

328
    If should be called in DeclareLocks in a way similar to::
329

330
      if level == locking.LEVEL_NODE:
331
        self._LockInstancesNodes()
332

333
    @type primary_only: boolean
334
    @param primary_only: only lock primary nodes of locked instances
335

336
    """
337
    assert locking.LEVEL_NODE in self.recalculate_locks, \
338
      "_LockInstancesNodes helper function called with no nodes to recalculate"
339

    
340
    # TODO: check if we're really been called with the instance locks held
341

    
342
    # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
343
    # future we might want to have different behaviors depending on the value
344
    # of self.recalculate_locks[locking.LEVEL_NODE]
345
    wanted_nodes = []
346
    for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
347
      instance = self.context.cfg.GetInstanceInfo(instance_name)
348
      wanted_nodes.append(instance.primary_node)
349
      if not primary_only:
350
        wanted_nodes.extend(instance.secondary_nodes)
351

    
352
    if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE:
353
      self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
354
    elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND:
355
      self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
356

    
357
    del self.recalculate_locks[locking.LEVEL_NODE]
358

    
359

    
360
class NoHooksLU(LogicalUnit): # pylint: disable-msg=W0223
361
  """Simple LU which runs no hooks.
362

363
  This LU is intended as a parent for other LogicalUnits which will
364
  run no hooks, in order to reduce duplicate code.
365

366
  """
367
  HPATH = None
368
  HTYPE = None
369

    
370
  def BuildHooksEnv(self):
371
    """Empty BuildHooksEnv for NoHooksLu.
372

373
    This just raises an error.
374

375
    """
376
    assert False, "BuildHooksEnv called for NoHooksLUs"
377

    
378

    
379
class Tasklet:
380
  """Tasklet base class.
381

382
  Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
383
  they can mix legacy code with tasklets. Locking needs to be done in the LU,
384
  tasklets know nothing about locks.
385

386
  Subclasses must follow these rules:
387
    - Implement CheckPrereq
388
    - Implement Exec
389

390
  """
391
  def __init__(self, lu):
392
    self.lu = lu
393

    
394
    # Shortcuts
395
    self.cfg = lu.cfg
396
    self.rpc = lu.rpc
397

    
398
  def CheckPrereq(self):
399
    """Check prerequisites for this tasklets.
400

401
    This method should check whether the prerequisites for the execution of
402
    this tasklet are fulfilled. It can do internode communication, but it
403
    should be idempotent - no cluster or system changes are allowed.
404

405
    The method should raise errors.OpPrereqError in case something is not
406
    fulfilled. Its return value is ignored.
407

408
    This method should also update all parameters to their canonical form if it
409
    hasn't been done before.
410

411
    """
412
    raise NotImplementedError
413

    
414
  def Exec(self, feedback_fn):
415
    """Execute the tasklet.
416

417
    This method should implement the actual work. It should raise
418
    errors.OpExecError for failures that are somewhat dealt with in code, or
419
    expected.
420

421
    """
422
    raise NotImplementedError
423

    
424

    
425
def _GetWantedNodes(lu, nodes):
426
  """Returns list of checked and expanded node names.
427

428
  @type lu: L{LogicalUnit}
429
  @param lu: the logical unit on whose behalf we execute
430
  @type nodes: list
431
  @param nodes: list of node names or None for all nodes
432
  @rtype: list
433
  @return: the list of nodes, sorted
434
  @raise errors.OpProgrammerError: if the nodes parameter is wrong type
435

436
  """
437
  if not isinstance(nodes, list):
438
    raise errors.OpPrereqError("Invalid argument type 'nodes'",
439
                               errors.ECODE_INVAL)
440

    
441
  if not nodes:
442
    raise errors.ProgrammerError("_GetWantedNodes should only be called with a"
443
      " non-empty list of nodes whose name is to be expanded.")
444

    
445
  wanted = []
446
  for name in nodes:
447
    node = lu.cfg.ExpandNodeName(name)
448
    if node is None:
449
      raise errors.OpPrereqError("No such node name '%s'" % name,
450
                                 errors.ECODE_NOENT)
451
    wanted.append(node)
452

    
453
  return utils.NiceSort(wanted)
454

    
455

    
456
def _GetWantedInstances(lu, instances):
457
  """Returns list of checked and expanded instance names.
458

459
  @type lu: L{LogicalUnit}
460
  @param lu: the logical unit on whose behalf we execute
461
  @type instances: list
462
  @param instances: list of instance names or None for all instances
463
  @rtype: list
464
  @return: the list of instances, sorted
465
  @raise errors.OpPrereqError: if the instances parameter is wrong type
466
  @raise errors.OpPrereqError: if any of the passed instances is not found
467

468
  """
469
  if not isinstance(instances, list):
470
    raise errors.OpPrereqError("Invalid argument type 'instances'",
471
                               errors.ECODE_INVAL)
472

    
473
  if instances:
474
    wanted = []
475

    
476
    for name in instances:
477
      instance = lu.cfg.ExpandInstanceName(name)
478
      if instance is None:
479
        raise errors.OpPrereqError("No such instance name '%s'" % name,
480
                                   errors.ECODE_NOENT)
481
      wanted.append(instance)
482

    
483
  else:
484
    wanted = utils.NiceSort(lu.cfg.GetInstanceList())
485
  return wanted
486

    
487

    
488
def _CheckOutputFields(static, dynamic, selected):
489
  """Checks whether all selected fields are valid.
490

491
  @type static: L{utils.FieldSet}
492
  @param static: static fields set
493
  @type dynamic: L{utils.FieldSet}
494
  @param dynamic: dynamic fields set
495

496
  """
497
  f = utils.FieldSet()
498
  f.Extend(static)
499
  f.Extend(dynamic)
500

    
501
  delta = f.NonMatching(selected)
502
  if delta:
503
    raise errors.OpPrereqError("Unknown output fields selected: %s"
504
                               % ",".join(delta), errors.ECODE_INVAL)
505

    
506

    
507
def _CheckBooleanOpField(op, name):
508
  """Validates boolean opcode parameters.
509

510
  This will ensure that an opcode parameter is either a boolean value,
511
  or None (but that it always exists).
512

513
  """
514
  val = getattr(op, name, None)
515
  if not (val is None or isinstance(val, bool)):
516
    raise errors.OpPrereqError("Invalid boolean parameter '%s' (%s)" %
517
                               (name, str(val)), errors.ECODE_INVAL)
518
  setattr(op, name, val)
519

    
520

    
521
def _CheckGlobalHvParams(params):
522
  """Validates that given hypervisor params are not global ones.
523

524
  This will ensure that instances don't get customised versions of
525
  global params.
526

527
  """
528
  used_globals = constants.HVC_GLOBALS.intersection(params)
529
  if used_globals:
530
    msg = ("The following hypervisor parameters are global and cannot"
531
           " be customized at instance level, please modify them at"
532
           " cluster level: %s" % utils.CommaJoin(used_globals))
533
    raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
534

    
535

    
536
def _CheckNodeOnline(lu, node):
537
  """Ensure that a given node is online.
538

539
  @param lu: the LU on behalf of which we make the check
540
  @param node: the node to check
541
  @raise errors.OpPrereqError: if the node is offline
542

543
  """
544
  if lu.cfg.GetNodeInfo(node).offline:
545
    raise errors.OpPrereqError("Can't use offline node %s" % node,
546
                               errors.ECODE_INVAL)
547

    
548

    
549
def _CheckNodeNotDrained(lu, node):
550
  """Ensure that a given node is not drained.
551

552
  @param lu: the LU on behalf of which we make the check
553
  @param node: the node to check
554
  @raise errors.OpPrereqError: if the node is drained
555

556
  """
557
  if lu.cfg.GetNodeInfo(node).drained:
558
    raise errors.OpPrereqError("Can't use drained node %s" % node,
559
                               errors.ECODE_INVAL)
560

    
561

    
562
def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
563
                          memory, vcpus, nics, disk_template, disks,
564
                          bep, hvp, hypervisor_name):
565
  """Builds instance related env variables for hooks
566

567
  This builds the hook environment from individual variables.
568

569
  @type name: string
570
  @param name: the name of the instance
571
  @type primary_node: string
572
  @param primary_node: the name of the instance's primary node
573
  @type secondary_nodes: list
574
  @param secondary_nodes: list of secondary nodes as strings
575
  @type os_type: string
576
  @param os_type: the name of the instance's OS
577
  @type status: boolean
578
  @param status: the should_run status of the instance
579
  @type memory: string
580
  @param memory: the memory size of the instance
581
  @type vcpus: string
582
  @param vcpus: the count of VCPUs the instance has
583
  @type nics: list
584
  @param nics: list of tuples (ip, mac, mode, link) representing
585
      the NICs the instance has
586
  @type disk_template: string
587
  @param disk_template: the disk template of the instance
588
  @type disks: list
589
  @param disks: the list of (size, mode) pairs
590
  @type bep: dict
591
  @param bep: the backend parameters for the instance
592
  @type hvp: dict
593
  @param hvp: the hypervisor parameters for the instance
594
  @type hypervisor_name: string
595
  @param hypervisor_name: the hypervisor for the instance
596
  @rtype: dict
597
  @return: the hook environment for this instance
598

599
  """
600
  if status:
601
    str_status = "up"
602
  else:
603
    str_status = "down"
604
  env = {
605
    "OP_TARGET": name,
606
    "INSTANCE_NAME": name,
607
    "INSTANCE_PRIMARY": primary_node,
608
    "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
609
    "INSTANCE_OS_TYPE": os_type,
610
    "INSTANCE_STATUS": str_status,
611
    "INSTANCE_MEMORY": memory,
612
    "INSTANCE_VCPUS": vcpus,
613
    "INSTANCE_DISK_TEMPLATE": disk_template,
614
    "INSTANCE_HYPERVISOR": hypervisor_name,
615
  }
616

    
617
  if nics:
618
    nic_count = len(nics)
619
    for idx, (ip, mac, mode, link) in enumerate(nics):
620
      if ip is None:
621
        ip = ""
622
      env["INSTANCE_NIC%d_IP" % idx] = ip
623
      env["INSTANCE_NIC%d_MAC" % idx] = mac
624
      env["INSTANCE_NIC%d_MODE" % idx] = mode
625
      env["INSTANCE_NIC%d_LINK" % idx] = link
626
      if mode == constants.NIC_MODE_BRIDGED:
627
        env["INSTANCE_NIC%d_BRIDGE" % idx] = link
628
  else:
629
    nic_count = 0
630

    
631
  env["INSTANCE_NIC_COUNT"] = nic_count
632

    
633
  if disks:
634
    disk_count = len(disks)
635
    for idx, (size, mode) in enumerate(disks):
636
      env["INSTANCE_DISK%d_SIZE" % idx] = size
637
      env["INSTANCE_DISK%d_MODE" % idx] = mode
638
  else:
639
    disk_count = 0
640

    
641
  env["INSTANCE_DISK_COUNT"] = disk_count
642

    
643
  for source, kind in [(bep, "BE"), (hvp, "HV")]:
644
    for key, value in source.items():
645
      env["INSTANCE_%s_%s" % (kind, key)] = value
646

    
647
  return env
648

    
649

    
650
def _NICListToTuple(lu, nics):
651
  """Build a list of nic information tuples.
652

653
  This list is suitable to be passed to _BuildInstanceHookEnv or as a return
654
  value in LUQueryInstanceData.
655

656
  @type lu:  L{LogicalUnit}
657
  @param lu: the logical unit on whose behalf we execute
658
  @type nics: list of L{objects.NIC}
659
  @param nics: list of nics to convert to hooks tuples
660

661
  """
662
  hooks_nics = []
663
  c_nicparams = lu.cfg.GetClusterInfo().nicparams[constants.PP_DEFAULT]
664
  for nic in nics:
665
    ip = nic.ip
666
    mac = nic.mac
667
    filled_params = objects.FillDict(c_nicparams, nic.nicparams)
668
    mode = filled_params[constants.NIC_MODE]
669
    link = filled_params[constants.NIC_LINK]
670
    hooks_nics.append((ip, mac, mode, link))
671
  return hooks_nics
672

    
673

    
674
def _BuildInstanceHookEnvByObject(lu, instance, override=None):
675
  """Builds instance related env variables for hooks from an object.
676

677
  @type lu: L{LogicalUnit}
678
  @param lu: the logical unit on whose behalf we execute
679
  @type instance: L{objects.Instance}
680
  @param instance: the instance for which we should build the
681
      environment
682
  @type override: dict
683
  @param override: dictionary with key/values that will override
684
      our values
685
  @rtype: dict
686
  @return: the hook environment dictionary
687

688
  """
689
  cluster = lu.cfg.GetClusterInfo()
690
  bep = cluster.FillBE(instance)
691
  hvp = cluster.FillHV(instance)
692
  args = {
693
    'name': instance.name,
694
    'primary_node': instance.primary_node,
695
    'secondary_nodes': instance.secondary_nodes,
696
    'os_type': instance.os,
697
    'status': instance.admin_up,
698
    'memory': bep[constants.BE_MEMORY],
699
    'vcpus': bep[constants.BE_VCPUS],
700
    'nics': _NICListToTuple(lu, instance.nics),
701
    'disk_template': instance.disk_template,
702
    'disks': [(disk.size, disk.mode) for disk in instance.disks],
703
    'bep': bep,
704
    'hvp': hvp,
705
    'hypervisor_name': instance.hypervisor,
706
  }
707
  if override:
708
    args.update(override)
709
  return _BuildInstanceHookEnv(**args) # pylint: disable-msg=W0142
710

    
711

    
712
def _AdjustCandidatePool(lu, exceptions):
713
  """Adjust the candidate pool after node operations.
714

715
  """
716
  mod_list = lu.cfg.MaintainCandidatePool(exceptions)
717
  if mod_list:
718
    lu.LogInfo("Promoted nodes to master candidate role: %s",
719
               utils.CommaJoin(node.name for node in mod_list))
720
    for name in mod_list:
721
      lu.context.ReaddNode(name)
722
  mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
723
  if mc_now > mc_max:
724
    lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
725
               (mc_now, mc_max))
726

    
727

    
728
def _DecideSelfPromotion(lu, exceptions=None):
729
  """Decide whether I should promote myself as a master candidate.
730

731
  """
732
  cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
733
  mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
734
  # the new node will increase mc_max with one, so:
735
  mc_should = min(mc_should + 1, cp_size)
736
  return mc_now < mc_should
737

    
738

    
739
def _CheckNicsBridgesExist(lu, target_nics, target_node,
740
                               profile=constants.PP_DEFAULT):
741
  """Check that the brigdes needed by a list of nics exist.
742

743
  """
744
  c_nicparams = lu.cfg.GetClusterInfo().nicparams[profile]
745
  paramslist = [objects.FillDict(c_nicparams, nic.nicparams)
746
                for nic in target_nics]
747
  brlist = [params[constants.NIC_LINK] for params in paramslist
748
            if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
749
  if brlist:
750
    result = lu.rpc.call_bridges_exist(target_node, brlist)
751
    result.Raise("Error checking bridges on destination node '%s'" %
752
                 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
753

    
754

    
755
def _CheckInstanceBridgesExist(lu, instance, node=None):
756
  """Check that the brigdes needed by an instance exist.
757

758
  """
759
  if node is None:
760
    node = instance.primary_node
761
  _CheckNicsBridgesExist(lu, instance.nics, node)
762

    
763

    
764
def _CheckOSVariant(os_obj, name):
765
  """Check whether an OS name conforms to the os variants specification.
766

767
  @type os_obj: L{objects.OS}
768
  @param os_obj: OS object to check
769
  @type name: string
770
  @param name: OS name passed by the user, to check for validity
771

772
  """
773
  if not os_obj.supported_variants:
774
    return
775
  try:
776
    variant = name.split("+", 1)[1]
777
  except IndexError:
778
    raise errors.OpPrereqError("OS name must include a variant",
779
                               errors.ECODE_INVAL)
780

    
781
  if variant not in os_obj.supported_variants:
782
    raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
783

    
784

    
785
def _GetNodeInstancesInner(cfg, fn):
786
  return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
787

    
788

    
789
def _GetNodeInstances(cfg, node_name):
790
  """Returns a list of all primary and secondary instances on a node.
791

792
  """
793

    
794
  return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
795

    
796

    
797
def _GetNodePrimaryInstances(cfg, node_name):
798
  """Returns primary instances on a node.
799

800
  """
801
  return _GetNodeInstancesInner(cfg,
802
                                lambda inst: node_name == inst.primary_node)
803

    
804

    
805
def _GetNodeSecondaryInstances(cfg, node_name):
806
  """Returns secondary instances on a node.
807

808
  """
809
  return _GetNodeInstancesInner(cfg,
810
                                lambda inst: node_name in inst.secondary_nodes)
811

    
812

    
813
def _GetStorageTypeArgs(cfg, storage_type):
814
  """Returns the arguments for a storage type.
815

816
  """
817
  # Special case for file storage
818
  if storage_type == constants.ST_FILE:
819
    # storage.FileStorage wants a list of storage directories
820
    return [[cfg.GetFileStorageDir()]]
821

    
822
  return []
823

    
824

    
825
def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq):
826
  faulty = []
827

    
828
  for dev in instance.disks:
829
    cfg.SetDiskID(dev, node_name)
830

    
831
  result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks)
832
  result.Raise("Failed to get disk status from node %s" % node_name,
833
               prereq=prereq, ecode=errors.ECODE_ENVIRON)
834

    
835
  for idx, bdev_status in enumerate(result.payload):
836
    if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
837
      faulty.append(idx)
838

    
839
  return faulty
840

    
841

    
842
class LUPostInitCluster(LogicalUnit):
843
  """Logical unit for running hooks after cluster initialization.
844

845
  """
846
  HPATH = "cluster-init"
847
  HTYPE = constants.HTYPE_CLUSTER
848
  _OP_REQP = []
849

    
850
  def BuildHooksEnv(self):
851
    """Build hooks env.
852

853
    """
854
    env = {"OP_TARGET": self.cfg.GetClusterName()}
855
    mn = self.cfg.GetMasterNode()
856
    return env, [], [mn]
857

    
858
  def CheckPrereq(self):
859
    """No prerequisites to check.
860

861
    """
862
    return True
863

    
864
  def Exec(self, feedback_fn):
865
    """Nothing to do.
866

867
    """
868
    return True
869

    
870

    
871
class LUDestroyCluster(LogicalUnit):
872
  """Logical unit for destroying the cluster.
873

874
  """
875
  HPATH = "cluster-destroy"
876
  HTYPE = constants.HTYPE_CLUSTER
877
  _OP_REQP = []
878

    
879
  def BuildHooksEnv(self):
880
    """Build hooks env.
881

882
    """
883
    env = {"OP_TARGET": self.cfg.GetClusterName()}
884
    return env, [], []
885

    
886
  def CheckPrereq(self):
887
    """Check prerequisites.
888

889
    This checks whether the cluster is empty.
890

891
    Any errors are signaled by raising errors.OpPrereqError.
892

893
    """
894
    master = self.cfg.GetMasterNode()
895

    
896
    nodelist = self.cfg.GetNodeList()
897
    if len(nodelist) != 1 or nodelist[0] != master:
898
      raise errors.OpPrereqError("There are still %d node(s) in"
899
                                 " this cluster." % (len(nodelist) - 1),
900
                                 errors.ECODE_INVAL)
901
    instancelist = self.cfg.GetInstanceList()
902
    if instancelist:
903
      raise errors.OpPrereqError("There are still %d instance(s) in"
904
                                 " this cluster." % len(instancelist),
905
                                 errors.ECODE_INVAL)
906

    
907
  def Exec(self, feedback_fn):
908
    """Destroys the cluster.
909

910
    """
911
    master = self.cfg.GetMasterNode()
912
    modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
913

    
914
    # Run post hooks on master node before it's removed
915
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
916
    try:
917
      hm.RunPhase(constants.HOOKS_PHASE_POST, [master])
918
    except:
919
      # pylint: disable-msg=W0702
920
      self.LogWarning("Errors occurred running hooks on %s" % master)
921

    
922
    result = self.rpc.call_node_stop_master(master, False)
923
    result.Raise("Could not disable the master role")
924

    
925
    if modify_ssh_setup:
926
      priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
927
      utils.CreateBackup(priv_key)
928
      utils.CreateBackup(pub_key)
929

    
930
    return master
931

    
932

    
933
class LUVerifyCluster(LogicalUnit):
934
  """Verifies the cluster status.
935

936
  """
937
  HPATH = "cluster-verify"
938
  HTYPE = constants.HTYPE_CLUSTER
939
  _OP_REQP = ["skip_checks", "verbose", "error_codes", "debug_simulate_errors"]
940
  REQ_BGL = False
941

    
942
  TCLUSTER = "cluster"
943
  TNODE = "node"
944
  TINSTANCE = "instance"
945

    
946
  ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
947
  EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
948
  EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
949
  EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
950
  EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
951
  EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
952
  EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE")
953
  ENODEDRBD = (TNODE, "ENODEDRBD")
954
  ENODEFILECHECK = (TNODE, "ENODEFILECHECK")
955
  ENODEHOOKS = (TNODE, "ENODEHOOKS")
956
  ENODEHV = (TNODE, "ENODEHV")
957
  ENODELVM = (TNODE, "ENODELVM")
958
  ENODEN1 = (TNODE, "ENODEN1")
959
  ENODENET = (TNODE, "ENODENET")
960
  ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE")
961
  ENODEORPHANLV = (TNODE, "ENODEORPHANLV")
962
  ENODERPC = (TNODE, "ENODERPC")
963
  ENODESSH = (TNODE, "ENODESSH")
964
  ENODEVERSION = (TNODE, "ENODEVERSION")
965
  ENODESETUP = (TNODE, "ENODESETUP")
966
  ENODETIME = (TNODE, "ENODETIME")
967

    
968
  ETYPE_FIELD = "code"
969
  ETYPE_ERROR = "ERROR"
970
  ETYPE_WARNING = "WARNING"
971

    
972
  def ExpandNames(self):
973
    self.needed_locks = {
974
      locking.LEVEL_NODE: locking.ALL_SET,
975
      locking.LEVEL_INSTANCE: locking.ALL_SET,
976
    }
977
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
978

    
979
  def _Error(self, ecode, item, msg, *args, **kwargs):
980
    """Format an error message.
981

982
    Based on the opcode's error_codes parameter, either format a
983
    parseable error code, or a simpler error string.
984

985
    This must be called only from Exec and functions called from Exec.
986

987
    """
988
    ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
989
    itype, etxt = ecode
990
    # first complete the msg
991
    if args:
992
      msg = msg % args
993
    # then format the whole message
994
    if self.op.error_codes:
995
      msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
996
    else:
997
      if item:
998
        item = " " + item
999
      else:
1000
        item = ""
1001
      msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
1002
    # and finally report it via the feedback_fn
1003
    self._feedback_fn("  - %s" % msg)
1004

    
1005
  def _ErrorIf(self, cond, *args, **kwargs):
1006
    """Log an error message if the passed condition is True.
1007

1008
    """
1009
    cond = bool(cond) or self.op.debug_simulate_errors
1010
    if cond:
1011
      self._Error(*args, **kwargs)
1012
    # do not mark the operation as failed for WARN cases only
1013
    if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
1014
      self.bad = self.bad or cond
1015

    
1016
  def _VerifyNode(self, nodeinfo, file_list, local_cksum,
1017
                  node_result, master_files, drbd_map, vg_name):
1018
    """Run multiple tests against a node.
1019

1020
    Test list:
1021

1022
      - compares ganeti version
1023
      - checks vg existence and size > 20G
1024
      - checks config file checksum
1025
      - checks ssh to other nodes
1026

1027
    @type nodeinfo: L{objects.Node}
1028
    @param nodeinfo: the node to check
1029
    @param file_list: required list of files
1030
    @param local_cksum: dictionary of local files and their checksums
1031
    @param node_result: the results from the node
1032
    @param master_files: list of files that only masters should have
1033
    @param drbd_map: the useddrbd minors for this node, in
1034
        form of minor: (instance, must_exist) which correspond to instances
1035
        and their running status
1036
    @param vg_name: Ganeti Volume Group (result of self.cfg.GetVGName())
1037

1038
    """
1039
    node = nodeinfo.name
1040
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1041

    
1042
    # main result, node_result should be a non-empty dict
1043
    test = not node_result or not isinstance(node_result, dict)
1044
    _ErrorIf(test, self.ENODERPC, node,
1045
                  "unable to verify node: no data returned")
1046
    if test:
1047
      return
1048

    
1049
    # compares ganeti version
1050
    local_version = constants.PROTOCOL_VERSION
1051
    remote_version = node_result.get('version', None)
1052
    test = not (remote_version and
1053
                isinstance(remote_version, (list, tuple)) and
1054
                len(remote_version) == 2)
1055
    _ErrorIf(test, self.ENODERPC, node,
1056
             "connection to node returned invalid data")
1057
    if test:
1058
      return
1059

    
1060
    test = local_version != remote_version[0]
1061
    _ErrorIf(test, self.ENODEVERSION, node,
1062
             "incompatible protocol versions: master %s,"
1063
             " node %s", local_version, remote_version[0])
1064
    if test:
1065
      return
1066

    
1067
    # node seems compatible, we can actually try to look into its results
1068

    
1069
    # full package version
1070
    self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
1071
                  self.ENODEVERSION, node,
1072
                  "software version mismatch: master %s, node %s",
1073
                  constants.RELEASE_VERSION, remote_version[1],
1074
                  code=self.ETYPE_WARNING)
1075

    
1076
    # checks vg existence and size > 20G
1077
    if vg_name is not None:
1078
      vglist = node_result.get(constants.NV_VGLIST, None)
1079
      test = not vglist
1080
      _ErrorIf(test, self.ENODELVM, node, "unable to check volume groups")
1081
      if not test:
1082
        vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
1083
                                              constants.MIN_VG_SIZE)
1084
        _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
1085

    
1086
    # checks config file checksum
1087

    
1088
    remote_cksum = node_result.get(constants.NV_FILELIST, None)
1089
    test = not isinstance(remote_cksum, dict)
1090
    _ErrorIf(test, self.ENODEFILECHECK, node,
1091
             "node hasn't returned file checksum data")
1092
    if not test:
1093
      for file_name in file_list:
1094
        node_is_mc = nodeinfo.master_candidate
1095
        must_have = (file_name not in master_files) or node_is_mc
1096
        # missing
1097
        test1 = file_name not in remote_cksum
1098
        # invalid checksum
1099
        test2 = not test1 and remote_cksum[file_name] != local_cksum[file_name]
1100
        # existing and good
1101
        test3 = not test1 and remote_cksum[file_name] == local_cksum[file_name]
1102
        _ErrorIf(test1 and must_have, self.ENODEFILECHECK, node,
1103
                 "file '%s' missing", file_name)
1104
        _ErrorIf(test2 and must_have, self.ENODEFILECHECK, node,
1105
                 "file '%s' has wrong checksum", file_name)
1106
        # not candidate and this is not a must-have file
1107
        _ErrorIf(test2 and not must_have, self.ENODEFILECHECK, node,
1108
                 "file '%s' should not exist on non master"
1109
                 " candidates (and the file is outdated)", file_name)
1110
        # all good, except non-master/non-must have combination
1111
        _ErrorIf(test3 and not must_have, self.ENODEFILECHECK, node,
1112
                 "file '%s' should not exist"
1113
                 " on non master candidates", file_name)
1114

    
1115
    # checks ssh to any
1116

    
1117
    test = constants.NV_NODELIST not in node_result
1118
    _ErrorIf(test, self.ENODESSH, node,
1119
             "node hasn't returned node ssh connectivity data")
1120
    if not test:
1121
      if node_result[constants.NV_NODELIST]:
1122
        for a_node, a_msg in node_result[constants.NV_NODELIST].items():
1123
          _ErrorIf(True, self.ENODESSH, node,
1124
                   "ssh communication with node '%s': %s", a_node, a_msg)
1125

    
1126
    test = constants.NV_NODENETTEST not in node_result
1127
    _ErrorIf(test, self.ENODENET, node,
1128
             "node hasn't returned node tcp connectivity data")
1129
    if not test:
1130
      if node_result[constants.NV_NODENETTEST]:
1131
        nlist = utils.NiceSort(node_result[constants.NV_NODENETTEST].keys())
1132
        for anode in nlist:
1133
          _ErrorIf(True, self.ENODENET, node,
1134
                   "tcp communication with node '%s': %s",
1135
                   anode, node_result[constants.NV_NODENETTEST][anode])
1136

    
1137
    hyp_result = node_result.get(constants.NV_HYPERVISOR, None)
1138
    if isinstance(hyp_result, dict):
1139
      for hv_name, hv_result in hyp_result.iteritems():
1140
        test = hv_result is not None
1141
        _ErrorIf(test, self.ENODEHV, node,
1142
                 "hypervisor %s verify failure: '%s'", hv_name, hv_result)
1143

    
1144
    # check used drbd list
1145
    if vg_name is not None:
1146
      used_minors = node_result.get(constants.NV_DRBDLIST, [])
1147
      test = not isinstance(used_minors, (tuple, list))
1148
      _ErrorIf(test, self.ENODEDRBD, node,
1149
               "cannot parse drbd status file: %s", str(used_minors))
1150
      if not test:
1151
        for minor, (iname, must_exist) in drbd_map.items():
1152
          test = minor not in used_minors and must_exist
1153
          _ErrorIf(test, self.ENODEDRBD, node,
1154
                   "drbd minor %d of instance %s is not active",
1155
                   minor, iname)
1156
        for minor in used_minors:
1157
          test = minor not in drbd_map
1158
          _ErrorIf(test, self.ENODEDRBD, node,
1159
                   "unallocated drbd minor %d is in use", minor)
1160
    test = node_result.get(constants.NV_NODESETUP,
1161
                           ["Missing NODESETUP results"])
1162
    _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s",
1163
             "; ".join(test))
1164

    
1165
    # check pv names
1166
    if vg_name is not None:
1167
      pvlist = node_result.get(constants.NV_PVLIST, None)
1168
      test = pvlist is None
1169
      _ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node")
1170
      if not test:
1171
        # check that ':' is not present in PV names, since it's a
1172
        # special character for lvcreate (denotes the range of PEs to
1173
        # use on the PV)
1174
        for _, pvname, owner_vg in pvlist:
1175
          test = ":" in pvname
1176
          _ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV"
1177
                   " '%s' of VG '%s'", pvname, owner_vg)
1178

    
1179
  def _VerifyInstance(self, instance, instanceconfig, node_vol_is,
1180
                      node_instance, n_offline):
1181
    """Verify an instance.
1182

1183
    This function checks to see if the required block devices are
1184
    available on the instance's node.
1185

1186
    """
1187
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1188
    node_current = instanceconfig.primary_node
1189

    
1190
    node_vol_should = {}
1191
    instanceconfig.MapLVsByNode(node_vol_should)
1192

    
1193
    for node in node_vol_should:
1194
      if node in n_offline:
1195
        # ignore missing volumes on offline nodes
1196
        continue
1197
      for volume in node_vol_should[node]:
1198
        test = node not in node_vol_is or volume not in node_vol_is[node]
1199
        _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
1200
                 "volume %s missing on node %s", volume, node)
1201

    
1202
    if instanceconfig.admin_up:
1203
      test = ((node_current not in node_instance or
1204
               not instance in node_instance[node_current]) and
1205
              node_current not in n_offline)
1206
      _ErrorIf(test, self.EINSTANCEDOWN, instance,
1207
               "instance not running on its primary node %s",
1208
               node_current)
1209

    
1210
    for node in node_instance:
1211
      if (not node == node_current):
1212
        test = instance in node_instance[node]
1213
        _ErrorIf(test, self.EINSTANCEWRONGNODE, instance,
1214
                 "instance should not run on node %s", node)
1215

    
1216
  def _VerifyOrphanVolumes(self, node_vol_should, node_vol_is):
1217
    """Verify if there are any unknown volumes in the cluster.
1218

1219
    The .os, .swap and backup volumes are ignored. All other volumes are
1220
    reported as unknown.
1221

1222
    """
1223
    for node in node_vol_is:
1224
      for volume in node_vol_is[node]:
1225
        test = (node not in node_vol_should or
1226
                volume not in node_vol_should[node])
1227
        self._ErrorIf(test, self.ENODEORPHANLV, node,
1228
                      "volume %s is unknown", volume)
1229

    
1230
  def _VerifyOrphanInstances(self, instancelist, node_instance):
1231
    """Verify the list of running instances.
1232

1233
    This checks what instances are running but unknown to the cluster.
1234

1235
    """
1236
    for node in node_instance:
1237
      for o_inst in node_instance[node]:
1238
        test = o_inst not in instancelist
1239
        self._ErrorIf(test, self.ENODEORPHANINSTANCE, node,
1240
                      "instance %s on node %s should not exist", o_inst, node)
1241

    
1242
  def _VerifyNPlusOneMemory(self, node_info, instance_cfg):
1243
    """Verify N+1 Memory Resilience.
1244

1245
    Check that if one single node dies we can still start all the instances it
1246
    was primary for.
1247

1248
    """
1249
    for node, nodeinfo in node_info.iteritems():
1250
      # This code checks that every node which is now listed as secondary has
1251
      # enough memory to host all instances it is supposed to should a single
1252
      # other node in the cluster fail.
1253
      # FIXME: not ready for failover to an arbitrary node
1254
      # FIXME: does not support file-backed instances
1255
      # WARNING: we currently take into account down instances as well as up
1256
      # ones, considering that even if they're down someone might want to start
1257
      # them even in the event of a node failure.
1258
      for prinode, instances in nodeinfo['sinst-by-pnode'].iteritems():
1259
        needed_mem = 0
1260
        for instance in instances:
1261
          bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
1262
          if bep[constants.BE_AUTO_BALANCE]:
1263
            needed_mem += bep[constants.BE_MEMORY]
1264
        test = nodeinfo['mfree'] < needed_mem
1265
        self._ErrorIf(test, self.ENODEN1, node,
1266
                      "not enough memory on to accommodate"
1267
                      " failovers should peer node %s fail", prinode)
1268

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

1272
    Transform the list of checks we're going to skip into a set and check that
1273
    all its members are valid.
1274

1275
    """
1276
    self.skip_set = frozenset(self.op.skip_checks)
1277
    if not constants.VERIFY_OPTIONAL_CHECKS.issuperset(self.skip_set):
1278
      raise errors.OpPrereqError("Invalid checks to be skipped specified",
1279
                                 errors.ECODE_INVAL)
1280

    
1281
  def BuildHooksEnv(self):
1282
    """Build hooks env.
1283

1284
    Cluster-Verify hooks just ran in the post phase and their failure makes
1285
    the output be logged in the verify output and the verification to fail.
1286

1287
    """
1288
    all_nodes = self.cfg.GetNodeList()
1289
    env = {
1290
      "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
1291
      }
1292
    for node in self.cfg.GetAllNodesInfo().values():
1293
      env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
1294

    
1295
    return env, [], all_nodes
1296

    
1297
  def Exec(self, feedback_fn):
1298
    """Verify integrity of cluster, performing various test on nodes.
1299

1300
    """
1301
    self.bad = False
1302
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1303
    verbose = self.op.verbose
1304
    self._feedback_fn = feedback_fn
1305
    feedback_fn("* Verifying global settings")
1306
    for msg in self.cfg.VerifyConfig():
1307
      _ErrorIf(True, self.ECLUSTERCFG, None, msg)
1308

    
1309
    vg_name = self.cfg.GetVGName()
1310
    hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
1311
    nodelist = utils.NiceSort(self.cfg.GetNodeList())
1312
    nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
1313
    instancelist = utils.NiceSort(self.cfg.GetInstanceList())
1314
    instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
1315
                        for iname in instancelist)
1316
    i_non_redundant = [] # Non redundant instances
1317
    i_non_a_balanced = [] # Non auto-balanced instances
1318
    n_offline = [] # List of offline nodes
1319
    n_drained = [] # List of nodes being drained
1320
    node_volume = {}
1321
    node_instance = {}
1322
    node_info = {}
1323
    instance_cfg = {}
1324

    
1325
    # FIXME: verify OS list
1326
    # do local checksums
1327
    master_files = [constants.CLUSTER_CONF_FILE]
1328

    
1329
    file_names = ssconf.SimpleStore().GetFileList()
1330
    file_names.append(constants.SSL_CERT_FILE)
1331
    file_names.append(constants.RAPI_CERT_FILE)
1332
    file_names.extend(master_files)
1333

    
1334
    local_checksums = utils.FingerprintFiles(file_names)
1335

    
1336
    feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
1337
    node_verify_param = {
1338
      constants.NV_FILELIST: file_names,
1339
      constants.NV_NODELIST: [node.name for node in nodeinfo
1340
                              if not node.offline],
1341
      constants.NV_HYPERVISOR: hypervisors,
1342
      constants.NV_NODENETTEST: [(node.name, node.primary_ip,
1343
                                  node.secondary_ip) for node in nodeinfo
1344
                                 if not node.offline],
1345
      constants.NV_INSTANCELIST: hypervisors,
1346
      constants.NV_VERSION: None,
1347
      constants.NV_HVINFO: self.cfg.GetHypervisorType(),
1348
      constants.NV_NODESETUP: None,
1349
      constants.NV_TIME: None,
1350
      }
1351

    
1352
    if vg_name is not None:
1353
      node_verify_param[constants.NV_VGLIST] = None
1354
      node_verify_param[constants.NV_LVLIST] = vg_name
1355
      node_verify_param[constants.NV_PVLIST] = [vg_name]
1356
      node_verify_param[constants.NV_DRBDLIST] = None
1357

    
1358
    # Due to the way our RPC system works, exact response times cannot be
1359
    # guaranteed (e.g. a broken node could run into a timeout). By keeping the
1360
    # time before and after executing the request, we can at least have a time
1361
    # window.
1362
    nvinfo_starttime = time.time()
1363
    all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
1364
                                           self.cfg.GetClusterName())
1365
    nvinfo_endtime = time.time()
1366

    
1367
    cluster = self.cfg.GetClusterInfo()
1368
    master_node = self.cfg.GetMasterNode()
1369
    all_drbd_map = self.cfg.ComputeDRBDMap()
1370

    
1371
    feedback_fn("* Verifying node status")
1372
    for node_i in nodeinfo:
1373
      node = node_i.name
1374

    
1375
      if node_i.offline:
1376
        if verbose:
1377
          feedback_fn("* Skipping offline node %s" % (node,))
1378
        n_offline.append(node)
1379
        continue
1380

    
1381
      if node == master_node:
1382
        ntype = "master"
1383
      elif node_i.master_candidate:
1384
        ntype = "master candidate"
1385
      elif node_i.drained:
1386
        ntype = "drained"
1387
        n_drained.append(node)
1388
      else:
1389
        ntype = "regular"
1390
      if verbose:
1391
        feedback_fn("* Verifying node %s (%s)" % (node, ntype))
1392

    
1393
      msg = all_nvinfo[node].fail_msg
1394
      _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
1395
      if msg:
1396
        continue
1397

    
1398
      nresult = all_nvinfo[node].payload
1399
      node_drbd = {}
1400
      for minor, instance in all_drbd_map[node].items():
1401
        test = instance not in instanceinfo
1402
        _ErrorIf(test, self.ECLUSTERCFG, None,
1403
                 "ghost instance '%s' in temporary DRBD map", instance)
1404
          # ghost instance should not be running, but otherwise we
1405
          # don't give double warnings (both ghost instance and
1406
          # unallocated minor in use)
1407
        if test:
1408
          node_drbd[minor] = (instance, False)
1409
        else:
1410
          instance = instanceinfo[instance]
1411
          node_drbd[minor] = (instance.name, instance.admin_up)
1412

    
1413
      self._VerifyNode(node_i, file_names, local_checksums,
1414
                       nresult, master_files, node_drbd, vg_name)
1415

    
1416
      lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
1417
      if vg_name is None:
1418
        node_volume[node] = {}
1419
      elif isinstance(lvdata, basestring):
1420
        _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
1421
                 utils.SafeEncode(lvdata))
1422
        node_volume[node] = {}
1423
      elif not isinstance(lvdata, dict):
1424
        _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
1425
        continue
1426
      else:
1427
        node_volume[node] = lvdata
1428

    
1429
      # node_instance
1430
      idata = nresult.get(constants.NV_INSTANCELIST, None)
1431
      test = not isinstance(idata, list)
1432
      _ErrorIf(test, self.ENODEHV, node,
1433
               "rpc call to node failed (instancelist)")
1434
      if test:
1435
        continue
1436

    
1437
      node_instance[node] = idata
1438

    
1439
      # node_info
1440
      nodeinfo = nresult.get(constants.NV_HVINFO, None)
1441
      test = not isinstance(nodeinfo, dict)
1442
      _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
1443
      if test:
1444
        continue
1445

    
1446
      # Node time
1447
      ntime = nresult.get(constants.NV_TIME, None)
1448
      try:
1449
        ntime_merged = utils.MergeTime(ntime)
1450
      except (ValueError, TypeError):
1451
        _ErrorIf(test, self.ENODETIME, node, "Node returned invalid time")
1452

    
1453
      if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
1454
        ntime_diff = abs(nvinfo_starttime - ntime_merged)
1455
      elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
1456
        ntime_diff = abs(ntime_merged - nvinfo_endtime)
1457
      else:
1458
        ntime_diff = None
1459

    
1460
      _ErrorIf(ntime_diff is not None, self.ENODETIME, node,
1461
               "Node time diverges by at least %0.1fs from master node time",
1462
               ntime_diff)
1463

    
1464
      if ntime_diff is not None:
1465
        continue
1466

    
1467
      try:
1468
        node_info[node] = {
1469
          "mfree": int(nodeinfo['memory_free']),
1470
          "pinst": [],
1471
          "sinst": [],
1472
          # dictionary holding all instances this node is secondary for,
1473
          # grouped by their primary node. Each key is a cluster node, and each
1474
          # value is a list of instances which have the key as primary and the
1475
          # current node as secondary.  this is handy to calculate N+1 memory
1476
          # availability if you can only failover from a primary to its
1477
          # secondary.
1478
          "sinst-by-pnode": {},
1479
        }
1480
        # FIXME: devise a free space model for file based instances as well
1481
        if vg_name is not None:
1482
          test = (constants.NV_VGLIST not in nresult or
1483
                  vg_name not in nresult[constants.NV_VGLIST])
1484
          _ErrorIf(test, self.ENODELVM, node,
1485
                   "node didn't return data for the volume group '%s'"
1486
                   " - it is either missing or broken", vg_name)
1487
          if test:
1488
            continue
1489
          node_info[node]["dfree"] = int(nresult[constants.NV_VGLIST][vg_name])
1490
      except (ValueError, KeyError):
1491
        _ErrorIf(True, self.ENODERPC, node,
1492
                 "node returned invalid nodeinfo, check lvm/hypervisor")
1493
        continue
1494

    
1495
    node_vol_should = {}
1496

    
1497
    feedback_fn("* Verifying instance status")
1498
    for instance in instancelist:
1499
      if verbose:
1500
        feedback_fn("* Verifying instance %s" % instance)
1501
      inst_config = instanceinfo[instance]
1502
      self._VerifyInstance(instance, inst_config, node_volume,
1503
                           node_instance, n_offline)
1504
      inst_nodes_offline = []
1505

    
1506
      inst_config.MapLVsByNode(node_vol_should)
1507

    
1508
      instance_cfg[instance] = inst_config
1509

    
1510
      pnode = inst_config.primary_node
1511
      _ErrorIf(pnode not in node_info and pnode not in n_offline,
1512
               self.ENODERPC, pnode, "instance %s, connection to"
1513
               " primary node failed", instance)
1514
      if pnode in node_info:
1515
        node_info[pnode]['pinst'].append(instance)
1516

    
1517
      if pnode in n_offline:
1518
        inst_nodes_offline.append(pnode)
1519

    
1520
      # If the instance is non-redundant we cannot survive losing its primary
1521
      # node, so we are not N+1 compliant. On the other hand we have no disk
1522
      # templates with more than one secondary so that situation is not well
1523
      # supported either.
1524
      # FIXME: does not support file-backed instances
1525
      if len(inst_config.secondary_nodes) == 0:
1526
        i_non_redundant.append(instance)
1527
      _ErrorIf(len(inst_config.secondary_nodes) > 1,
1528
               self.EINSTANCELAYOUT, instance,
1529
               "instance has multiple secondary nodes", code="WARNING")
1530

    
1531
      if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
1532
        i_non_a_balanced.append(instance)
1533

    
1534
      for snode in inst_config.secondary_nodes:
1535
        _ErrorIf(snode not in node_info and snode not in n_offline,
1536
                 self.ENODERPC, snode,
1537
                 "instance %s, connection to secondary node"
1538
                 "failed", instance)
1539

    
1540
        if snode in node_info:
1541
          node_info[snode]['sinst'].append(instance)
1542
          if pnode not in node_info[snode]['sinst-by-pnode']:
1543
            node_info[snode]['sinst-by-pnode'][pnode] = []
1544
          node_info[snode]['sinst-by-pnode'][pnode].append(instance)
1545

    
1546
        if snode in n_offline:
1547
          inst_nodes_offline.append(snode)
1548

    
1549
      # warn that the instance lives on offline nodes
1550
      _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
1551
               "instance lives on offline node(s) %s",
1552
               utils.CommaJoin(inst_nodes_offline))
1553

    
1554
    feedback_fn("* Verifying orphan volumes")
1555
    self._VerifyOrphanVolumes(node_vol_should, node_volume)
1556

    
1557
    feedback_fn("* Verifying remaining instances")
1558
    self._VerifyOrphanInstances(instancelist, node_instance)
1559

    
1560
    if constants.VERIFY_NPLUSONE_MEM not in self.skip_set:
1561
      feedback_fn("* Verifying N+1 Memory redundancy")
1562
      self._VerifyNPlusOneMemory(node_info, instance_cfg)
1563

    
1564
    feedback_fn("* Other Notes")
1565
    if i_non_redundant:
1566
      feedback_fn("  - NOTICE: %d non-redundant instance(s) found."
1567
                  % len(i_non_redundant))
1568

    
1569
    if i_non_a_balanced:
1570
      feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
1571
                  % len(i_non_a_balanced))
1572

    
1573
    if n_offline:
1574
      feedback_fn("  - NOTICE: %d offline node(s) found." % len(n_offline))
1575

    
1576
    if n_drained:
1577
      feedback_fn("  - NOTICE: %d drained node(s) found." % len(n_drained))
1578

    
1579
    return not self.bad
1580

    
1581
  def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
1582
    """Analyze the post-hooks' result
1583

1584
    This method analyses the hook result, handles it, and sends some
1585
    nicely-formatted feedback back to the user.
1586

1587
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
1588
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
1589
    @param hooks_results: the results of the multi-node hooks rpc call
1590
    @param feedback_fn: function used send feedback back to the caller
1591
    @param lu_result: previous Exec result
1592
    @return: the new Exec result, based on the previous result
1593
        and hook results
1594

1595
    """
1596
    # We only really run POST phase hooks, and are only interested in
1597
    # their results
1598
    if phase == constants.HOOKS_PHASE_POST:
1599
      # Used to change hooks' output to proper indentation
1600
      indent_re = re.compile('^', re.M)
1601
      feedback_fn("* Hooks Results")
1602
      assert hooks_results, "invalid result from hooks"
1603

    
1604
      for node_name in hooks_results:
1605
        res = hooks_results[node_name]
1606
        msg = res.fail_msg
1607
        test = msg and not res.offline
1608
        self._ErrorIf(test, self.ENODEHOOKS, node_name,
1609
                      "Communication failure in hooks execution: %s", msg)
1610
        if res.offline or msg:
1611
          # No need to investigate payload if node is offline or gave an error.
1612
          # override manually lu_result here as _ErrorIf only
1613
          # overrides self.bad
1614
          lu_result = 1
1615
          continue
1616
        for script, hkr, output in res.payload:
1617
          test = hkr == constants.HKR_FAIL
1618
          self._ErrorIf(test, self.ENODEHOOKS, node_name,
1619
                        "Script %s failed, output:", script)
1620
          if test:
1621
            output = indent_re.sub('      ', output)
1622
            feedback_fn("%s" % output)
1623
            lu_result = 1
1624

    
1625
      return lu_result
1626

    
1627

    
1628
class LUVerifyDisks(NoHooksLU):
1629
  """Verifies the cluster disks status.
1630

1631
  """
1632
  _OP_REQP = []
1633
  REQ_BGL = False
1634

    
1635
  def ExpandNames(self):
1636
    self.needed_locks = {
1637
      locking.LEVEL_NODE: locking.ALL_SET,
1638
      locking.LEVEL_INSTANCE: locking.ALL_SET,
1639
    }
1640
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
1641

    
1642
  def CheckPrereq(self):
1643
    """Check prerequisites.
1644

1645
    This has no prerequisites.
1646

1647
    """
1648
    pass
1649

    
1650
  def Exec(self, feedback_fn):
1651
    """Verify integrity of cluster disks.
1652

1653
    @rtype: tuple of three items
1654
    @return: a tuple of (dict of node-to-node_error, list of instances
1655
        which need activate-disks, dict of instance: (node, volume) for
1656
        missing volumes
1657

1658
    """
1659
    result = res_nodes, res_instances, res_missing = {}, [], {}
1660

    
1661
    vg_name = self.cfg.GetVGName()
1662
    nodes = utils.NiceSort(self.cfg.GetNodeList())
1663
    instances = [self.cfg.GetInstanceInfo(name)
1664
                 for name in self.cfg.GetInstanceList()]
1665

    
1666
    nv_dict = {}
1667
    for inst in instances:
1668
      inst_lvs = {}
1669
      if (not inst.admin_up or
1670
          inst.disk_template not in constants.DTS_NET_MIRROR):
1671
        continue
1672
      inst.MapLVsByNode(inst_lvs)
1673
      # transform { iname: {node: [vol,],},} to {(node, vol): iname}
1674
      for node, vol_list in inst_lvs.iteritems():
1675
        for vol in vol_list:
1676
          nv_dict[(node, vol)] = inst
1677

    
1678
    if not nv_dict:
1679
      return result
1680

    
1681
    node_lvs = self.rpc.call_lv_list(nodes, vg_name)
1682

    
1683
    for node in nodes:
1684
      # node_volume
1685
      node_res = node_lvs[node]
1686
      if node_res.offline:
1687
        continue
1688
      msg = node_res.fail_msg
1689
      if msg:
1690
        logging.warning("Error enumerating LVs on node %s: %s", node, msg)
1691
        res_nodes[node] = msg
1692
        continue
1693

    
1694
      lvs = node_res.payload
1695
      for lv_name, (_, _, lv_online) in lvs.items():
1696
        inst = nv_dict.pop((node, lv_name), None)
1697
        if (not lv_online and inst is not None
1698
            and inst.name not in res_instances):
1699
          res_instances.append(inst.name)
1700

    
1701
    # any leftover items in nv_dict are missing LVs, let's arrange the
1702
    # data better
1703
    for key, inst in nv_dict.iteritems():
1704
      if inst.name not in res_missing:
1705
        res_missing[inst.name] = []
1706
      res_missing[inst.name].append(key)
1707

    
1708
    return result
1709

    
1710

    
1711
class LURepairDiskSizes(NoHooksLU):
1712
  """Verifies the cluster disks sizes.
1713

1714
  """
1715
  _OP_REQP = ["instances"]
1716
  REQ_BGL = False
1717

    
1718
  def ExpandNames(self):
1719
    if not isinstance(self.op.instances, list):
1720
      raise errors.OpPrereqError("Invalid argument type 'instances'",
1721
                                 errors.ECODE_INVAL)
1722

    
1723
    if self.op.instances:
1724
      self.wanted_names = []
1725
      for name in self.op.instances:
1726
        full_name = self.cfg.ExpandInstanceName(name)
1727
        if full_name is None:
1728
          raise errors.OpPrereqError("Instance '%s' not known" % name,
1729
                                     errors.ECODE_NOENT)
1730
        self.wanted_names.append(full_name)
1731
      self.needed_locks = {
1732
        locking.LEVEL_NODE: [],
1733
        locking.LEVEL_INSTANCE: self.wanted_names,
1734
        }
1735
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
1736
    else:
1737
      self.wanted_names = None
1738
      self.needed_locks = {
1739
        locking.LEVEL_NODE: locking.ALL_SET,
1740
        locking.LEVEL_INSTANCE: locking.ALL_SET,
1741
        }
1742
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
1743

    
1744
  def DeclareLocks(self, level):
1745
    if level == locking.LEVEL_NODE and self.wanted_names is not None:
1746
      self._LockInstancesNodes(primary_only=True)
1747

    
1748
  def CheckPrereq(self):
1749
    """Check prerequisites.
1750

1751
    This only checks the optional instance list against the existing names.
1752

1753
    """
1754
    if self.wanted_names is None:
1755
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
1756

    
1757
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
1758
                             in self.wanted_names]
1759

    
1760
  def _EnsureChildSizes(self, disk):
1761
    """Ensure children of the disk have the needed disk size.
1762

1763
    This is valid mainly for DRBD8 and fixes an issue where the
1764
    children have smaller disk size.
1765

1766
    @param disk: an L{ganeti.objects.Disk} object
1767

1768
    """
1769
    if disk.dev_type == constants.LD_DRBD8:
1770
      assert disk.children, "Empty children for DRBD8?"
1771
      fchild = disk.children[0]
1772
      mismatch = fchild.size < disk.size
1773
      if mismatch:
1774
        self.LogInfo("Child disk has size %d, parent %d, fixing",
1775
                     fchild.size, disk.size)
1776
        fchild.size = disk.size
1777

    
1778
      # and we recurse on this child only, not on the metadev
1779
      return self._EnsureChildSizes(fchild) or mismatch
1780
    else:
1781
      return False
1782

    
1783
  def Exec(self, feedback_fn):
1784
    """Verify the size of cluster disks.
1785

1786
    """
1787
    # TODO: check child disks too
1788
    # TODO: check differences in size between primary/secondary nodes
1789
    per_node_disks = {}
1790
    for instance in self.wanted_instances:
1791
      pnode = instance.primary_node
1792
      if pnode not in per_node_disks:
1793
        per_node_disks[pnode] = []
1794
      for idx, disk in enumerate(instance.disks):
1795
        per_node_disks[pnode].append((instance, idx, disk))
1796

    
1797
    changed = []
1798
    for node, dskl in per_node_disks.items():
1799
      newl = [v[2].Copy() for v in dskl]
1800
      for dsk in newl:
1801
        self.cfg.SetDiskID(dsk, node)
1802
      result = self.rpc.call_blockdev_getsizes(node, newl)
1803
      if result.fail_msg:
1804
        self.LogWarning("Failure in blockdev_getsizes call to node"
1805
                        " %s, ignoring", node)
1806
        continue
1807
      if len(result.data) != len(dskl):
1808
        self.LogWarning("Invalid result from node %s, ignoring node results",
1809
                        node)
1810
        continue
1811
      for ((instance, idx, disk), size) in zip(dskl, result.data):
1812
        if size is None:
1813
          self.LogWarning("Disk %d of instance %s did not return size"
1814
                          " information, ignoring", idx, instance.name)
1815
          continue
1816
        if not isinstance(size, (int, long)):
1817
          self.LogWarning("Disk %d of instance %s did not return valid"
1818
                          " size information, ignoring", idx, instance.name)
1819
          continue
1820
        size = size >> 20
1821
        if size != disk.size:
1822
          self.LogInfo("Disk %d of instance %s has mismatched size,"
1823
                       " correcting: recorded %d, actual %d", idx,
1824
                       instance.name, disk.size, size)
1825
          disk.size = size
1826
          self.cfg.Update(instance, feedback_fn)
1827
          changed.append((instance.name, idx, size))
1828
        if self._EnsureChildSizes(disk):
1829
          self.cfg.Update(instance, feedback_fn)
1830
          changed.append((instance.name, idx, disk.size))
1831
    return changed
1832

    
1833

    
1834
class LURenameCluster(LogicalUnit):
1835
  """Rename the cluster.
1836

1837
  """
1838
  HPATH = "cluster-rename"
1839
  HTYPE = constants.HTYPE_CLUSTER
1840
  _OP_REQP = ["name"]
1841

    
1842
  def BuildHooksEnv(self):
1843
    """Build hooks env.
1844

1845
    """
1846
    env = {
1847
      "OP_TARGET": self.cfg.GetClusterName(),
1848
      "NEW_NAME": self.op.name,
1849
      }
1850
    mn = self.cfg.GetMasterNode()
1851
    all_nodes = self.cfg.GetNodeList()
1852
    return env, [mn], all_nodes
1853

    
1854
  def CheckPrereq(self):
1855
    """Verify that the passed name is a valid one.
1856

1857
    """
1858
    hostname = utils.GetHostInfo(self.op.name)
1859

    
1860
    new_name = hostname.name
1861
    self.ip = new_ip = hostname.ip
1862
    old_name = self.cfg.GetClusterName()
1863
    old_ip = self.cfg.GetMasterIP()
1864
    if new_name == old_name and new_ip == old_ip:
1865
      raise errors.OpPrereqError("Neither the name nor the IP address of the"
1866
                                 " cluster has changed",
1867
                                 errors.ECODE_INVAL)
1868
    if new_ip != old_ip:
1869
      if utils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
1870
        raise errors.OpPrereqError("The given cluster IP address (%s) is"
1871
                                   " reachable on the network. Aborting." %
1872
                                   new_ip, errors.ECODE_NOTUNIQUE)
1873

    
1874
    self.op.name = new_name
1875

    
1876
  def Exec(self, feedback_fn):
1877
    """Rename the cluster.
1878

1879
    """
1880
    clustername = self.op.name
1881
    ip = self.ip
1882

    
1883
    # shutdown the master IP
1884
    master = self.cfg.GetMasterNode()
1885
    result = self.rpc.call_node_stop_master(master, False)
1886
    result.Raise("Could not disable the master role")
1887

    
1888
    try:
1889
      cluster = self.cfg.GetClusterInfo()
1890
      cluster.cluster_name = clustername
1891
      cluster.master_ip = ip
1892
      self.cfg.Update(cluster, feedback_fn)
1893

    
1894
      # update the known hosts file
1895
      ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
1896
      node_list = self.cfg.GetNodeList()
1897
      try:
1898
        node_list.remove(master)
1899
      except ValueError:
1900
        pass
1901
      result = self.rpc.call_upload_file(node_list,
1902
                                         constants.SSH_KNOWN_HOSTS_FILE)
1903
      for to_node, to_result in result.iteritems():
1904
        msg = to_result.fail_msg
1905
        if msg:
1906
          msg = ("Copy of file %s to node %s failed: %s" %
1907
                 (constants.SSH_KNOWN_HOSTS_FILE, to_node, msg))
1908
          self.proc.LogWarning(msg)
1909

    
1910
    finally:
1911
      result = self.rpc.call_node_start_master(master, False, False)
1912
      msg = result.fail_msg
1913
      if msg:
1914
        self.LogWarning("Could not re-enable the master role on"
1915
                        " the master, please restart manually: %s", msg)
1916

    
1917

    
1918
def _RecursiveCheckIfLVMBased(disk):
1919
  """Check if the given disk or its children are lvm-based.
1920

1921
  @type disk: L{objects.Disk}
1922
  @param disk: the disk to check
1923
  @rtype: boolean
1924
  @return: boolean indicating whether a LD_LV dev_type was found or not
1925

1926
  """
1927
  if disk.children:
1928
    for chdisk in disk.children:
1929
      if _RecursiveCheckIfLVMBased(chdisk):
1930
        return True
1931
  return disk.dev_type == constants.LD_LV
1932

    
1933

    
1934
class LUSetClusterParams(LogicalUnit):
1935
  """Change the parameters of the cluster.
1936

1937
  """
1938
  HPATH = "cluster-modify"
1939
  HTYPE = constants.HTYPE_CLUSTER
1940
  _OP_REQP = []
1941
  REQ_BGL = False
1942

    
1943
  def CheckArguments(self):
1944
    """Check parameters
1945

1946
    """
1947
    if not hasattr(self.op, "candidate_pool_size"):
1948
      self.op.candidate_pool_size = None
1949
    if self.op.candidate_pool_size is not None:
1950
      try:
1951
        self.op.candidate_pool_size = int(self.op.candidate_pool_size)
1952
      except (ValueError, TypeError), err:
1953
        raise errors.OpPrereqError("Invalid candidate_pool_size value: %s" %
1954
                                   str(err), errors.ECODE_INVAL)
1955
      if self.op.candidate_pool_size < 1:
1956
        raise errors.OpPrereqError("At least one master candidate needed",
1957
                                   errors.ECODE_INVAL)
1958

    
1959
  def ExpandNames(self):
1960
    # FIXME: in the future maybe other cluster params won't require checking on
1961
    # all nodes to be modified.
1962
    self.needed_locks = {
1963
      locking.LEVEL_NODE: locking.ALL_SET,
1964
    }
1965
    self.share_locks[locking.LEVEL_NODE] = 1
1966

    
1967
  def BuildHooksEnv(self):
1968
    """Build hooks env.
1969

1970
    """
1971
    env = {
1972
      "OP_TARGET": self.cfg.GetClusterName(),
1973
      "NEW_VG_NAME": self.op.vg_name,
1974
      }
1975
    mn = self.cfg.GetMasterNode()
1976
    return env, [mn], [mn]
1977

    
1978
  def CheckPrereq(self):
1979
    """Check prerequisites.
1980

1981
    This checks whether the given params don't conflict and
1982
    if the given volume group is valid.
1983

1984
    """
1985
    if self.op.vg_name is not None and not self.op.vg_name:
1986
      instances = self.cfg.GetAllInstancesInfo().values()
1987
      for inst in instances:
1988
        for disk in inst.disks:
1989
          if _RecursiveCheckIfLVMBased(disk):
1990
            raise errors.OpPrereqError("Cannot disable lvm storage while"
1991
                                       " lvm-based instances exist",
1992
                                       errors.ECODE_INVAL)
1993

    
1994
    node_list = self.acquired_locks[locking.LEVEL_NODE]
1995

    
1996
    # if vg_name not None, checks given volume group on all nodes
1997
    if self.op.vg_name:
1998
      vglist = self.rpc.call_vg_list(node_list)
1999
      for node in node_list:
2000
        msg = vglist[node].fail_msg
2001
        if msg:
2002
          # ignoring down node
2003
          self.LogWarning("Error while gathering data on node %s"
2004
                          " (ignoring node): %s", node, msg)
2005
          continue
2006
        vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
2007
                                              self.op.vg_name,
2008
                                              constants.MIN_VG_SIZE)
2009
        if vgstatus:
2010
          raise errors.OpPrereqError("Error on node '%s': %s" %
2011
                                     (node, vgstatus), errors.ECODE_ENVIRON)
2012

    
2013
    self.cluster = cluster = self.cfg.GetClusterInfo()
2014
    # validate params changes
2015
    if self.op.beparams:
2016
      utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
2017
      self.new_beparams = objects.FillDict(
2018
        cluster.beparams[constants.PP_DEFAULT], self.op.beparams)
2019

    
2020
    if self.op.nicparams:
2021
      utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
2022
      self.new_nicparams = objects.FillDict(
2023
        cluster.nicparams[constants.PP_DEFAULT], self.op.nicparams)
2024
      objects.NIC.CheckParameterSyntax(self.new_nicparams)
2025
      nic_errors = []
2026

    
2027
      # check all instances for consistency
2028
      for instance in self.cfg.GetAllInstancesInfo().values():
2029
        for nic_idx, nic in enumerate(instance.nics):
2030
          params_copy = copy.deepcopy(nic.nicparams)
2031
          params_filled = objects.FillDict(self.new_nicparams, params_copy)
2032

    
2033
          # check parameter syntax
2034
          try:
2035
            objects.NIC.CheckParameterSyntax(params_filled)
2036
          except errors.ConfigurationError, err:
2037
            nic_errors.append("Instance %s, nic/%d: %s" %
2038
                              (instance.name, nic_idx, err))
2039

    
2040
          # if we're moving instances to routed, check that they have an ip
2041
          target_mode = params_filled[constants.NIC_MODE]
2042
          if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
2043
            nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
2044
                              (instance.name, nic_idx))
2045
      if nic_errors:
2046
        raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
2047
                                   "\n".join(nic_errors))
2048

    
2049
    # hypervisor list/parameters
2050
    self.new_hvparams = objects.FillDict(cluster.hvparams, {})
2051
    if self.op.hvparams:
2052
      if not isinstance(self.op.hvparams, dict):
2053
        raise errors.OpPrereqError("Invalid 'hvparams' parameter on input",
2054
                                   errors.ECODE_INVAL)
2055
      for hv_name, hv_dict in self.op.hvparams.items():
2056
        if hv_name not in self.new_hvparams:
2057
          self.new_hvparams[hv_name] = hv_dict
2058
        else:
2059
          self.new_hvparams[hv_name].update(hv_dict)
2060

    
2061
    if self.op.enabled_hypervisors is not None:
2062
      self.hv_list = self.op.enabled_hypervisors
2063
      if not self.hv_list:
2064
        raise errors.OpPrereqError("Enabled hypervisors list must contain at"
2065
                                   " least one member",
2066
                                   errors.ECODE_INVAL)
2067
      invalid_hvs = set(self.hv_list) - constants.HYPER_TYPES
2068
      if invalid_hvs:
2069
        raise errors.OpPrereqError("Enabled hypervisors contains invalid"
2070
                                   " entries: %s" %
2071
                                   utils.CommaJoin(invalid_hvs),
2072
                                   errors.ECODE_INVAL)
2073
    else:
2074
      self.hv_list = cluster.enabled_hypervisors
2075

    
2076
    if self.op.hvparams or self.op.enabled_hypervisors is not None:
2077
      # either the enabled list has changed, or the parameters have, validate
2078
      for hv_name, hv_params in self.new_hvparams.items():
2079
        if ((self.op.hvparams and hv_name in self.op.hvparams) or
2080
            (self.op.enabled_hypervisors and
2081
             hv_name in self.op.enabled_hypervisors)):
2082
          # either this is a new hypervisor, or its parameters have changed
2083
          hv_class = hypervisor.GetHypervisor(hv_name)
2084
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2085
          hv_class.CheckParameterSyntax(hv_params)
2086
          _CheckHVParams(self, node_list, hv_name, hv_params)
2087

    
2088
  def Exec(self, feedback_fn):
2089
    """Change the parameters of the cluster.
2090

2091
    """
2092
    if self.op.vg_name is not None:
2093
      new_volume = self.op.vg_name
2094
      if not new_volume:
2095
        new_volume = None
2096
      if new_volume != self.cfg.GetVGName():
2097
        self.cfg.SetVGName(new_volume)
2098
      else:
2099
        feedback_fn("Cluster LVM configuration already in desired"
2100
                    " state, not changing")
2101
    if self.op.hvparams:
2102
      self.cluster.hvparams = self.new_hvparams
2103
    if self.op.enabled_hypervisors is not None:
2104
      self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
2105
    if self.op.beparams:
2106
      self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
2107
    if self.op.nicparams:
2108
      self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
2109

    
2110
    if self.op.candidate_pool_size is not None:
2111
      self.cluster.candidate_pool_size = self.op.candidate_pool_size
2112
      # we need to update the pool size here, otherwise the save will fail
2113
      _AdjustCandidatePool(self, [])
2114

    
2115
    self.cfg.Update(self.cluster, feedback_fn)
2116

    
2117

    
2118
def _RedistributeAncillaryFiles(lu, additional_nodes=None):
2119
  """Distribute additional files which are part of the cluster configuration.
2120

2121
  ConfigWriter takes care of distributing the config and ssconf files, but
2122
  there are more files which should be distributed to all nodes. This function
2123
  makes sure those are copied.
2124

2125
  @param lu: calling logical unit
2126
  @param additional_nodes: list of nodes not in the config to distribute to
2127

2128
  """
2129
  # 1. Gather target nodes
2130
  myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
2131
  dist_nodes = lu.cfg.GetNodeList()
2132
  if additional_nodes is not None:
2133
    dist_nodes.extend(additional_nodes)
2134
  if myself.name in dist_nodes:
2135
    dist_nodes.remove(myself.name)
2136

    
2137
  # 2. Gather files to distribute
2138
  dist_files = set([constants.ETC_HOSTS,
2139
                    constants.SSH_KNOWN_HOSTS_FILE,
2140
                    constants.RAPI_CERT_FILE,
2141
                    constants.RAPI_USERS_FILE,
2142
                    constants.HMAC_CLUSTER_KEY,
2143
                   ])
2144

    
2145
  enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
2146
  for hv_name in enabled_hypervisors:
2147
    hv_class = hypervisor.GetHypervisor(hv_name)
2148
    dist_files.update(hv_class.GetAncillaryFiles())
2149

    
2150
  # 3. Perform the files upload
2151
  for fname in dist_files:
2152
    if os.path.exists(fname):
2153
      result = lu.rpc.call_upload_file(dist_nodes, fname)
2154
      for to_node, to_result in result.items():
2155
        msg = to_result.fail_msg
2156
        if msg:
2157
          msg = ("Copy of file %s to node %s failed: %s" %
2158
                 (fname, to_node, msg))
2159
          lu.proc.LogWarning(msg)
2160

    
2161

    
2162
class LURedistributeConfig(NoHooksLU):
2163
  """Force the redistribution of cluster configuration.
2164

2165
  This is a very simple LU.
2166

2167
  """
2168
  _OP_REQP = []
2169
  REQ_BGL = False
2170

    
2171
  def ExpandNames(self):
2172
    self.needed_locks = {
2173
      locking.LEVEL_NODE: locking.ALL_SET,
2174
    }
2175
    self.share_locks[locking.LEVEL_NODE] = 1
2176

    
2177
  def CheckPrereq(self):
2178
    """Check prerequisites.
2179

2180
    """
2181

    
2182
  def Exec(self, feedback_fn):
2183
    """Redistribute the configuration.
2184

2185
    """
2186
    self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
2187
    _RedistributeAncillaryFiles(self)
2188

    
2189

    
2190
def _WaitForSync(lu, instance, oneshot=False):
2191
  """Sleep and poll for an instance's disk to sync.
2192

2193
  """
2194
  if not instance.disks:
2195
    return True
2196

    
2197
  if not oneshot:
2198
    lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
2199

    
2200
  node = instance.primary_node
2201

    
2202
  for dev in instance.disks:
2203
    lu.cfg.SetDiskID(dev, node)
2204

    
2205
  # TODO: Convert to utils.Retry
2206

    
2207
  retries = 0
2208
  degr_retries = 10 # in seconds, as we sleep 1 second each time
2209
  while True:
2210
    max_time = 0
2211
    done = True
2212
    cumul_degraded = False
2213
    rstats = lu.rpc.call_blockdev_getmirrorstatus(node, instance.disks)
2214
    msg = rstats.fail_msg
2215
    if msg:
2216
      lu.LogWarning("Can't get any data from node %s: %s", node, msg)
2217
      retries += 1
2218
      if retries >= 10:
2219
        raise errors.RemoteError("Can't contact node %s for mirror data,"
2220
                                 " aborting." % node)
2221
      time.sleep(6)
2222
      continue
2223
    rstats = rstats.payload
2224
    retries = 0
2225
    for i, mstat in enumerate(rstats):
2226
      if mstat is None:
2227
        lu.LogWarning("Can't compute data for node %s/%s",
2228
                           node, instance.disks[i].iv_name)
2229
        continue
2230

    
2231
      cumul_degraded = (cumul_degraded or
2232
                        (mstat.is_degraded and mstat.sync_percent is None))
2233
      if mstat.sync_percent is not None:
2234
        done = False
2235
        if mstat.estimated_time is not None:
2236
          rem_time = "%d estimated seconds remaining" % mstat.estimated_time
2237
          max_time = mstat.estimated_time
2238
        else:
2239
          rem_time = "no time estimate"
2240
        lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
2241
                        (instance.disks[i].iv_name, mstat.sync_percent,
2242
                         rem_time))
2243

    
2244
    # if we're done but degraded, let's do a few small retries, to
2245
    # make sure we see a stable and not transient situation; therefore
2246
    # we force restart of the loop
2247
    if (done or oneshot) and cumul_degraded and degr_retries > 0:
2248
      logging.info("Degraded disks found, %d retries left", degr_retries)
2249
      degr_retries -= 1
2250
      time.sleep(1)
2251
      continue
2252

    
2253
    if done or oneshot:
2254
      break
2255

    
2256
    time.sleep(min(60, max_time))
2257

    
2258
  if done:
2259
    lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
2260
  return not cumul_degraded
2261

    
2262

    
2263
def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
2264
  """Check that mirrors are not degraded.
2265

2266
  The ldisk parameter, if True, will change the test from the
2267
  is_degraded attribute (which represents overall non-ok status for
2268
  the device(s)) to the ldisk (representing the local storage status).
2269

2270
  """
2271
  lu.cfg.SetDiskID(dev, node)
2272

    
2273
  result = True
2274

    
2275
  if on_primary or dev.AssembleOnSecondary():
2276
    rstats = lu.rpc.call_blockdev_find(node, dev)
2277
    msg = rstats.fail_msg
2278
    if msg:
2279
      lu.LogWarning("Can't find disk on node %s: %s", node, msg)
2280
      result = False
2281
    elif not rstats.payload:
2282
      lu.LogWarning("Can't find disk on node %s", node)
2283
      result = False
2284
    else:
2285
      if ldisk:
2286
        result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
2287
      else:
2288
        result = result and not rstats.payload.is_degraded
2289

    
2290
  if dev.children:
2291
    for child in dev.children:
2292
      result = result and _CheckDiskConsistency(lu, child, node, on_primary)
2293

    
2294
  return result
2295

    
2296

    
2297
class LUDiagnoseOS(NoHooksLU):
2298
  """Logical unit for OS diagnose/query.
2299

2300
  """
2301
  _OP_REQP = ["output_fields", "names"]
2302
  REQ_BGL = False
2303
  _FIELDS_STATIC = utils.FieldSet()
2304
  _FIELDS_DYNAMIC = utils.FieldSet("name", "valid", "node_status", "variants")
2305
  # Fields that need calculation of global os validity
2306
  _FIELDS_NEEDVALID = frozenset(["valid", "variants"])
2307

    
2308
  def ExpandNames(self):
2309
    if self.op.names:
2310
      raise errors.OpPrereqError("Selective OS query not supported",
2311
                                 errors.ECODE_INVAL)
2312

    
2313
    _CheckOutputFields(static=self._FIELDS_STATIC,
2314
                       dynamic=self._FIELDS_DYNAMIC,
2315
                       selected=self.op.output_fields)
2316

    
2317
    # Lock all nodes, in shared mode
2318
    # Temporary removal of locks, should be reverted later
2319
    # TODO: reintroduce locks when they are lighter-weight
2320
    self.needed_locks = {}
2321
    #self.share_locks[locking.LEVEL_NODE] = 1
2322
    #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
2323

    
2324
  def CheckPrereq(self):
2325
    """Check prerequisites.
2326

2327
    """
2328

    
2329
  @staticmethod
2330
  def _DiagnoseByOS(rlist):
2331
    """Remaps a per-node return list into an a per-os per-node dictionary
2332

2333
    @param rlist: a map with node names as keys and OS objects as values
2334

2335
    @rtype: dict
2336
    @return: a dictionary with osnames as keys and as value another map, with
2337
        nodes as keys and tuples of (path, status, diagnose) as values, eg::
2338

2339
          {"debian-etch": {"node1": [(/usr/lib/..., True, ""),
2340
                                     (/srv/..., False, "invalid api")],
2341
                           "node2": [(/srv/..., True, "")]}
2342
          }
2343

2344
    """
2345
    all_os = {}
2346
    # we build here the list of nodes that didn't fail the RPC (at RPC
2347
    # level), so that nodes with a non-responding node daemon don't
2348
    # make all OSes invalid
2349
    good_nodes = [node_name for node_name in rlist
2350
                  if not rlist[node_name].fail_msg]
2351
    for node_name, nr in rlist.items():
2352
      if nr.fail_msg or not nr.payload:
2353
        continue
2354
      for name, path, status, diagnose, variants in nr.payload:
2355
        if name not in all_os:
2356
          # build a list of nodes for this os containing empty lists
2357
          # for each node in node_list
2358
          all_os[name] = {}
2359
          for nname in good_nodes:
2360
            all_os[name][nname] = []
2361
        all_os[name][node_name].append((path, status, diagnose, variants))
2362
    return all_os
2363

    
2364
  def Exec(self, feedback_fn):
2365
    """Compute the list of OSes.
2366

2367
    """
2368
    valid_nodes = [node for node in self.cfg.GetOnlineNodeList()]
2369
    node_data = self.rpc.call_os_diagnose(valid_nodes)
2370
    pol = self._DiagnoseByOS(node_data)
2371
    output = []
2372
    calc_valid = self._FIELDS_NEEDVALID.intersection(self.op.output_fields)
2373
    calc_variants = "variants" in self.op.output_fields
2374

    
2375
    for os_name, os_data in pol.items():
2376
      row = []
2377
      if calc_valid:
2378
        valid = True
2379
        variants = None
2380
        for osl in os_data.values():
2381
          valid = valid and osl and osl[0][1]
2382
          if not valid:
2383
            variants = None
2384
            break
2385
          if calc_variants:
2386
            node_variants = osl[0][3]
2387
            if variants is None:
2388
              variants = node_variants
2389
            else:
2390
              variants = [v for v in variants if v in node_variants]
2391

    
2392
      for field in self.op.output_fields:
2393
        if field == "name":
2394
          val = os_name
2395
        elif field == "valid":
2396
          val = valid
2397
        elif field == "node_status":
2398
          # this is just a copy of the dict
2399
          val = {}
2400
          for node_name, nos_list in os_data.items():
2401
            val[node_name] = nos_list
2402
        elif field == "variants":
2403
          val =  variants
2404
        else:
2405
          raise errors.ParameterError(field)
2406
        row.append(val)
2407
      output.append(row)
2408

    
2409
    return output
2410

    
2411

    
2412
class LURemoveNode(LogicalUnit):
2413
  """Logical unit for removing a node.
2414

2415
  """
2416
  HPATH = "node-remove"
2417
  HTYPE = constants.HTYPE_NODE
2418
  _OP_REQP = ["node_name"]
2419

    
2420
  def BuildHooksEnv(self):
2421
    """Build hooks env.
2422

2423
    This doesn't run on the target node in the pre phase as a failed
2424
    node would then be impossible to remove.
2425

2426
    """
2427
    env = {
2428
      "OP_TARGET": self.op.node_name,
2429
      "NODE_NAME": self.op.node_name,
2430
      }
2431
    all_nodes = self.cfg.GetNodeList()
2432
    try:
2433
      all_nodes.remove(self.op.node_name)
2434
    except ValueError:
2435
      logging.warning("Node %s which is about to be removed not found"
2436
                      " in the all nodes list", self.op.node_name)
2437
    return env, all_nodes, all_nodes
2438

    
2439
  def CheckPrereq(self):
2440
    """Check prerequisites.
2441

2442
    This checks:
2443
     - the node exists in the configuration
2444
     - it does not have primary or secondary instances
2445
     - it's not the master
2446

2447
    Any errors are signaled by raising errors.OpPrereqError.
2448

2449
    """
2450
    node = self.cfg.GetNodeInfo(self.cfg.ExpandNodeName(self.op.node_name))
2451
    if node is None:
2452
      raise errors.OpPrereqError("Node '%s' is unknown." % self.op.node_name,
2453
                                 errors.ECODE_NOENT)
2454

    
2455
    instance_list = self.cfg.GetInstanceList()
2456

    
2457
    masternode = self.cfg.GetMasterNode()
2458
    if node.name == masternode:
2459
      raise errors.OpPrereqError("Node is the master node,"
2460
                                 " you need to failover first.",
2461
                                 errors.ECODE_INVAL)
2462

    
2463
    for instance_name in instance_list:
2464
      instance = self.cfg.GetInstanceInfo(instance_name)
2465
      if node.name in instance.all_nodes:
2466
        raise errors.OpPrereqError("Instance %s is still running on the node,"
2467
                                   " please remove first." % instance_name,
2468
                                   errors.ECODE_INVAL)
2469
    self.op.node_name = node.name
2470
    self.node = node
2471

    
2472
  def Exec(self, feedback_fn):
2473
    """Removes the node from the cluster.
2474

2475
    """
2476
    node = self.node
2477
    logging.info("Stopping the node daemon and removing configs from node %s",
2478
                 node.name)
2479

    
2480
    modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
2481

    
2482
    # Promote nodes to master candidate as needed
2483
    _AdjustCandidatePool(self, exceptions=[node.name])
2484
    self.context.RemoveNode(node.name)
2485

    
2486
    # Run post hooks on the node before it's removed
2487
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
2488
    try:
2489
      hm.RunPhase(constants.HOOKS_PHASE_POST, [node.name])
2490
    except:
2491
      # pylint: disable-msg=W0702
2492
      self.LogWarning("Errors occurred running hooks on %s" % node.name)
2493

    
2494
    result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
2495
    msg = result.fail_msg
2496
    if msg:
2497
      self.LogWarning("Errors encountered on the remote node while leaving"
2498
                      " the cluster: %s", msg)
2499

    
2500

    
2501
class LUQueryNodes(NoHooksLU):
2502
  """Logical unit for querying nodes.
2503

2504
  """
2505
  # pylint: disable-msg=W0142
2506
  _OP_REQP = ["output_fields", "names", "use_locking"]
2507
  REQ_BGL = False
2508

    
2509
  _SIMPLE_FIELDS = ["name", "serial_no", "ctime", "mtime", "uuid",
2510
                    "master_candidate", "offline", "drained"]
2511

    
2512
  _FIELDS_DYNAMIC = utils.FieldSet(
2513
    "dtotal", "dfree",
2514
    "mtotal", "mnode", "mfree",
2515
    "bootid",
2516
    "ctotal", "cnodes", "csockets",
2517
    )
2518

    
2519
  _FIELDS_STATIC = utils.FieldSet(*[
2520
    "pinst_cnt", "sinst_cnt",
2521
    "pinst_list", "sinst_list",
2522
    "pip", "sip", "tags",
2523
    "master",
2524
    "role"] + _SIMPLE_FIELDS
2525
    )
2526

    
2527
  def ExpandNames(self):
2528
    _CheckOutputFields(static=self._FIELDS_STATIC,
2529
                       dynamic=self._FIELDS_DYNAMIC,
2530
                       selected=self.op.output_fields)
2531

    
2532
    self.needed_locks = {}
2533
    self.share_locks[locking.LEVEL_NODE] = 1
2534

    
2535
    if self.op.names:
2536
      self.wanted = _GetWantedNodes(self, self.op.names)
2537
    else:
2538
      self.wanted = locking.ALL_SET
2539

    
2540
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
2541
    self.do_locking = self.do_node_query and self.op.use_locking
2542
    if self.do_locking:
2543
      # if we don't request only static fields, we need to lock the nodes
2544
      self.needed_locks[locking.LEVEL_NODE] = self.wanted
2545

    
2546
  def CheckPrereq(self):
2547
    """Check prerequisites.
2548

2549
    """
2550
    # The validation of the node list is done in the _GetWantedNodes,
2551
    # if non empty, and if empty, there's no validation to do
2552
    pass
2553

    
2554
  def Exec(self, feedback_fn):
2555
    """Computes the list of nodes and their attributes.
2556

2557
    """
2558
    all_info = self.cfg.GetAllNodesInfo()
2559
    if self.do_locking:
2560
      nodenames = self.acquired_locks[locking.LEVEL_NODE]
2561
    elif self.wanted != locking.ALL_SET:
2562
      nodenames = self.wanted
2563
      missing = set(nodenames).difference(all_info.keys())
2564
      if missing:
2565
        raise errors.OpExecError(
2566
          "Some nodes were removed before retrieving their data: %s" % missing)
2567
    else:
2568
      nodenames = all_info.keys()
2569

    
2570
    nodenames = utils.NiceSort(nodenames)
2571
    nodelist = [all_info[name] for name in nodenames]
2572

    
2573
    # begin data gathering
2574

    
2575
    if self.do_node_query:
2576
      live_data = {}
2577
      node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
2578
                                          self.cfg.GetHypervisorType())
2579
      for name in nodenames:
2580
        nodeinfo = node_data[name]
2581
        if not nodeinfo.fail_msg and nodeinfo.payload:
2582
          nodeinfo = nodeinfo.payload
2583
          fn = utils.TryConvert
2584
          live_data[name] = {
2585
            "mtotal": fn(int, nodeinfo.get('memory_total', None)),
2586
            "mnode": fn(int, nodeinfo.get('memory_dom0', None)),
2587
            "mfree": fn(int, nodeinfo.get('memory_free', None)),
2588
            "dtotal": fn(int, nodeinfo.get('vg_size', None)),
2589
            "dfree": fn(int, nodeinfo.get('vg_free', None)),
2590
            "ctotal": fn(int, nodeinfo.get('cpu_total', None)),
2591
            "bootid": nodeinfo.get('bootid', None),
2592
            "cnodes": fn(int, nodeinfo.get('cpu_nodes', None)),
2593
            "csockets": fn(int, nodeinfo.get('cpu_sockets', None)),
2594
            }
2595
        else:
2596
          live_data[name] = {}
2597
    else:
2598
      live_data = dict.fromkeys(nodenames, {})
2599

    
2600
    node_to_primary = dict([(name, set()) for name in nodenames])
2601
    node_to_secondary = dict([(name, set()) for name in nodenames])
2602

    
2603
    inst_fields = frozenset(("pinst_cnt", "pinst_list",
2604
                             "sinst_cnt", "sinst_list"))
2605
    if inst_fields & frozenset(self.op.output_fields):
2606
      inst_data = self.cfg.GetAllInstancesInfo()
2607

    
2608
      for inst in inst_data.values():
2609
        if inst.primary_node in node_to_primary:
2610
          node_to_primary[inst.primary_node].add(inst.name)
2611
        for secnode in inst.secondary_nodes:
2612
          if secnode in node_to_secondary:
2613
            node_to_secondary[secnode].add(inst.name)
2614

    
2615
    master_node = self.cfg.GetMasterNode()
2616

    
2617
    # end data gathering
2618

    
2619
    output = []
2620
    for node in nodelist:
2621
      node_output = []
2622
      for field in self.op.output_fields:
2623
        if field in self._SIMPLE_FIELDS:
2624
          val = getattr(node, field)
2625
        elif field == "pinst_list":
2626
          val = list(node_to_primary[node.name])
2627
        elif field == "sinst_list":
2628
          val = list(node_to_secondary[node.name])
2629
        elif field == "pinst_cnt":
2630
          val = len(node_to_primary[node.name])
2631
        elif field == "sinst_cnt":
2632
          val = len(node_to_secondary[node.name])
2633
        elif field == "pip":
2634
          val = node.primary_ip
2635
        elif field == "sip":
2636
          val = node.secondary_ip
2637
        elif field == "tags":
2638
          val = list(node.GetTags())
2639
        elif field == "master":
2640
          val = node.name == master_node
2641
        elif self._FIELDS_DYNAMIC.Matches(field):
2642
          val = live_data[node.name].get(field, None)
2643
        elif field == "role":
2644
          if node.name == master_node:
2645
            val = "M"
2646
          elif node.master_candidate:
2647
            val = "C"
2648
          elif node.drained:
2649
            val = "D"
2650
          elif node.offline:
2651
            val = "O"
2652
          else:
2653
            val = "R"
2654
        else:
2655
          raise errors.ParameterError(field)
2656
        node_output.append(val)
2657
      output.append(node_output)
2658

    
2659
    return output
2660

    
2661

    
2662
class LUQueryNodeVolumes(NoHooksLU):
2663
  """Logical unit for getting volumes on node(s).
2664

2665
  """
2666
  _OP_REQP = ["nodes", "output_fields"]
2667
  REQ_BGL = False
2668
  _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
2669
  _FIELDS_STATIC = utils.FieldSet("node")
2670

    
2671
  def ExpandNames(self):
2672
    _CheckOutputFields(static=self._FIELDS_STATIC,
2673
                       dynamic=self._FIELDS_DYNAMIC,
2674
                       selected=self.op.output_fields)
2675

    
2676
    self.needed_locks = {}
2677
    self.share_locks[locking.LEVEL_NODE] = 1
2678
    if not self.op.nodes:
2679
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
2680
    else:
2681
      self.needed_locks[locking.LEVEL_NODE] = \
2682
        _GetWantedNodes(self, self.op.nodes)
2683

    
2684
  def CheckPrereq(self):
2685
    """Check prerequisites.
2686

2687
    This checks that the fields required are valid output fields.
2688

2689
    """
2690
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
2691

    
2692
  def Exec(self, feedback_fn):
2693
    """Computes the list of nodes and their attributes.
2694

2695
    """
2696
    nodenames = self.nodes
2697
    volumes = self.rpc.call_node_volumes(nodenames)
2698

    
2699
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
2700
             in self.cfg.GetInstanceList()]
2701

    
2702
    lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
2703

    
2704
    output = []
2705
    for node in nodenames:
2706
      nresult = volumes[node]
2707
      if nresult.offline:
2708
        continue
2709
      msg = nresult.fail_msg
2710
      if msg:
2711
        self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
2712
        continue
2713

    
2714
      node_vols = nresult.payload[:]
2715
      node_vols.sort(key=lambda vol: vol['dev'])
2716

    
2717
      for vol in node_vols:
2718
        node_output = []
2719
        for field in self.op.output_fields:
2720
          if field == "node":
2721
            val = node
2722
          elif field == "phys":
2723
            val = vol['dev']
2724
          elif field == "vg":
2725
            val = vol['vg']
2726
          elif field == "name":
2727
            val = vol['name']
2728
          elif field == "size":
2729
            val = int(float(vol['size']))
2730
          elif field == "instance":
2731
            for inst in ilist:
2732
              if node not in lv_by_node[inst]:
2733
                continue
2734
              if vol['name'] in lv_by_node[inst][node]:
2735
                val = inst.name
2736
                break
2737
            else:
2738
              val = '-'
2739
          else:
2740
            raise errors.ParameterError(field)
2741
          node_output.append(str(val))
2742

    
2743
        output.append(node_output)
2744

    
2745
    return output
2746

    
2747

    
2748
class LUQueryNodeStorage(NoHooksLU):
2749
  """Logical unit for getting information on storage units on node(s).
2750

2751
  """
2752
  _OP_REQP = ["nodes", "storage_type", "output_fields"]
2753
  REQ_BGL = False
2754
  _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
2755

    
2756
  def ExpandNames(self):
2757
    storage_type = self.op.storage_type
2758

    
2759
    if storage_type not in constants.VALID_STORAGE_TYPES:
2760
      raise errors.OpPrereqError("Unknown storage type: %s" % storage_type,
2761
                                 errors.ECODE_INVAL)
2762

    
2763
    _CheckOutputFields(static=self._FIELDS_STATIC,
2764
                       dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
2765
                       selected=self.op.output_fields)
2766

    
2767
    self.needed_locks = {}
2768
    self.share_locks[locking.LEVEL_NODE] = 1
2769

    
2770
    if self.op.nodes:
2771
      self.needed_locks[locking.LEVEL_NODE] = \
2772
        _GetWantedNodes(self, self.op.nodes)
2773
    else:
2774
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
2775

    
2776
  def CheckPrereq(self):
2777
    """Check prerequisites.
2778

2779
    This checks that the fields required are valid output fields.
2780

2781
    """
2782
    self.op.name = getattr(self.op, "name", None)
2783

    
2784
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
2785

    
2786
  def Exec(self, feedback_fn):
2787
    """Computes the list of nodes and their attributes.
2788

2789
    """
2790
    # Always get name to sort by
2791
    if constants.SF_NAME in self.op.output_fields:
2792
      fields = self.op.output_fields[:]
2793
    else:
2794
      fields = [constants.SF_NAME] + self.op.output_fields
2795

    
2796
    # Never ask for node or type as it's only known to the LU
2797
    for extra in [constants.SF_NODE, constants.SF_TYPE]:
2798
      while extra in fields:
2799
        fields.remove(extra)
2800

    
2801
    field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
2802
    name_idx = field_idx[constants.SF_NAME]
2803

    
2804
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
2805
    data = self.rpc.call_storage_list(self.nodes,
2806
                                      self.op.storage_type, st_args,
2807
                                      self.op.name, fields)
2808

    
2809
    result = []
2810

    
2811
    for node in utils.NiceSort(self.nodes):
2812
      nresult = data[node]
2813
      if nresult.offline:
2814
        continue
2815

    
2816
      msg = nresult.fail_msg
2817
      if msg:
2818
        self.LogWarning("Can't get storage data from node %s: %s", node, msg)
2819
        continue
2820

    
2821
      rows = dict([(row[name_idx], row) for row in nresult.payload])
2822

    
2823
      for name in utils.NiceSort(rows.keys()):
2824
        row = rows[name]
2825

    
2826
        out = []
2827

    
2828
        for field in self.op.output_fields:
2829
          if field == constants.SF_NODE:
2830
            val = node
2831
          elif field == constants.SF_TYPE:
2832
            val = self.op.storage_type
2833
          elif field in field_idx:
2834
            val = row[field_idx[field]]
2835
          else:
2836
            raise errors.ParameterError(field)
2837

    
2838
          out.append(val)
2839

    
2840
        result.append(out)
2841

    
2842
    return result
2843

    
2844

    
2845
class LUModifyNodeStorage(NoHooksLU):
2846
  """Logical unit for modifying a storage volume on a node.
2847

2848
  """
2849
  _OP_REQP = ["node_name", "storage_type", "name", "changes"]
2850
  REQ_BGL = False
2851

    
2852
  def CheckArguments(self):
2853
    node_name = self.cfg.ExpandNodeName(self.op.node_name)
2854
    if node_name is None:
2855
      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name,
2856
                                 errors.ECODE_NOENT)
2857

    
2858
    self.op.node_name = node_name
2859

    
2860
    storage_type = self.op.storage_type
2861
    if storage_type not in constants.VALID_STORAGE_TYPES:
2862
      raise errors.OpPrereqError("Unknown storage type: %s" % storage_type,
2863
                                 errors.ECODE_INVAL)
2864

    
2865
  def ExpandNames(self):
2866
    self.needed_locks = {
2867
      locking.LEVEL_NODE: self.op.node_name,
2868
      }
2869

    
2870
  def CheckPrereq(self):
2871
    """Check prerequisites.
2872

2873
    """
2874
    storage_type = self.op.storage_type
2875

    
2876
    try:
2877
      modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
2878
    except KeyError:
2879
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
2880
                                 " modified" % storage_type,
2881
                                 errors.ECODE_INVAL)
2882

    
2883
    diff = set(self.op.changes.keys()) - modifiable
2884
    if diff:
2885
      raise errors.OpPrereqError("The following fields can not be modified for"
2886
                                 " storage units of type '%s': %r" %
2887
                                 (storage_type, list(diff)),
2888
                                 errors.ECODE_INVAL)
2889

    
2890
  def Exec(self, feedback_fn):
2891
    """Computes the list of nodes and their attributes.
2892

2893
    """
2894
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
2895
    result = self.rpc.call_storage_modify(self.op.node_name,
2896
                                          self.op.storage_type, st_args,
2897
                                          self.op.name, self.op.changes)
2898
    result.Raise("Failed to modify storage unit '%s' on %s" %
2899
                 (self.op.name, self.op.node_name))
2900

    
2901

    
2902
class LUAddNode(LogicalUnit):
2903
  """Logical unit for adding node to the cluster.
2904

2905
  """
2906
  HPATH = "node-add"
2907
  HTYPE = constants.HTYPE_NODE
2908
  _OP_REQP = ["node_name"]
2909

    
2910
  def BuildHooksEnv(self):
2911
    """Build hooks env.
2912

2913
    This will run on all nodes before, and on all nodes + the new node after.
2914

2915
    """
2916
    env = {
2917
      "OP_TARGET": self.op.node_name,
2918
      "NODE_NAME": self.op.node_name,
2919
      "NODE_PIP": self.op.primary_ip,
2920
      "NODE_SIP": self.op.secondary_ip,
2921
      }
2922
    nodes_0 = self.cfg.GetNodeList()
2923
    nodes_1 = nodes_0 + [self.op.node_name, ]
2924
    return env, nodes_0, nodes_1
2925

    
2926
  def CheckPrereq(self):
2927
    """Check prerequisites.
2928

2929
    This checks:
2930
     - the new node is not already in the config
2931
     - it is resolvable
2932
     - its parameters (single/dual homed) matches the cluster
2933

2934
    Any errors are signaled by raising errors.OpPrereqError.
2935

2936
    """
2937
    node_name = self.op.node_name
2938
    cfg = self.cfg
2939

    
2940
    dns_data = utils.GetHostInfo(node_name)
2941

    
2942
    node = dns_data.name
2943
    primary_ip = self.op.primary_ip = dns_data.ip
2944
    secondary_ip = getattr(self.op, "secondary_ip", None)
2945
    if secondary_ip is None:
2946
      secondary_ip = primary_ip
2947
    if not utils.IsValidIP(secondary_ip):
2948
      raise errors.OpPrereqError("Invalid secondary IP given",
2949
                                 errors.ECODE_INVAL)
2950
    self.op.secondary_ip = secondary_ip
2951

    
2952
    node_list = cfg.GetNodeList()
2953
    if not self.op.readd and node in node_list:
2954
      raise errors.OpPrereqError("Node %s is already in the configuration" %
2955
                                 node, errors.ECODE_EXISTS)
2956
    elif self.op.readd and node not in node_list:
2957
      raise errors.OpPrereqError("Node %s is not in the configuration" % node,
2958
                                 errors.ECODE_NOENT)
2959

    
2960
    for existing_node_name in node_list:
2961
      existing_node = cfg.GetNodeInfo(existing_node_name)
2962

    
2963
      if self.op.readd and node == existing_node_name:
2964
        if (existing_node.primary_ip != primary_ip or
2965
            existing_node.secondary_ip != secondary_ip):
2966
          raise errors.OpPrereqError("Readded node doesn't have the same IP"
2967
                                     " address configuration as before",
2968
                                     errors.ECODE_INVAL)
2969
        continue
2970

    
2971
      if (existing_node.primary_ip == primary_ip or
2972
          existing_node.secondary_ip == primary_ip or
2973
          existing_node.primary_ip == secondary_ip or
2974
          existing_node.secondary_ip == secondary_ip):
2975
        raise errors.OpPrereqError("New node ip address(es) conflict with"
2976
                                   " existing node %s" % existing_node.name,
2977
                                   errors.ECODE_NOTUNIQUE)
2978

    
2979
    # check that the type of the node (single versus dual homed) is the
2980
    # same as for the master
2981
    myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
2982
    master_singlehomed = myself.secondary_ip == myself.primary_ip
2983
    newbie_singlehomed = secondary_ip == primary_ip
2984
    if master_singlehomed != newbie_singlehomed:
2985
      if master_singlehomed:
2986
        raise errors.OpPrereqError("The master has no private ip but the"
2987
                                   " new node has one",
2988
                                   errors.ECODE_INVAL)
2989
      else:
2990
        raise errors.OpPrereqError("The master has a private ip but the"
2991
                                   " new node doesn't have one",
2992
                                   errors.ECODE_INVAL)
2993

    
2994
    # checks reachability
2995
    if not utils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
2996
      raise errors.OpPrereqError("Node not reachable by ping",
2997
                                 errors.ECODE_ENVIRON)
2998

    
2999
    if not newbie_singlehomed:
3000
      # check reachability from my secondary ip to newbie's secondary ip
3001
      if not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
3002
                           source=myself.secondary_ip):
3003
        raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
3004
                                   " based ping to noded port",
3005
                                   errors.ECODE_ENVIRON)
3006

    
3007
    if self.op.readd:
3008
      exceptions = [node]
3009
    else:
3010
      exceptions = []
3011

    
3012
    self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
3013

    
3014
    if self.op.readd:
3015
      self.new_node = self.cfg.GetNodeInfo(node)
3016
      assert self.new_node is not None, "Can't retrieve locked node %s" % node
3017
    else:
3018
      self.new_node = objects.Node(name=node,
3019
                                   primary_ip=primary_ip,
3020
                                   secondary_ip=secondary_ip,
3021
                                   master_candidate=self.master_candidate,
3022
                                   offline=False, drained=False)
3023

    
3024
  def Exec(self, feedback_fn):
3025
    """Adds the new node to the cluster.
3026

3027
    """
3028
    new_node = self.new_node
3029
    node = new_node.name
3030

    
3031
    # for re-adds, reset the offline/drained/master-candidate flags;
3032
    # we need to reset here, otherwise offline would prevent RPC calls
3033
    # later in the procedure; this also means that if the re-add
3034
    # fails, we are left with a non-offlined, broken node
3035
    if self.op.readd:
3036
      new_node.drained = new_node.offline = False # pylint: disable-msg=W0201
3037
      self.LogInfo("Readding a node, the offline/drained flags were reset")
3038
      # if we demote the node, we do cleanup later in the procedure
3039
      new_node.master_candidate = self.master_candidate
3040

    
3041
    # notify the user about any possible mc promotion
3042
    if new_node.master_candidate:
3043
      self.LogInfo("Node will be a master candidate")
3044

    
3045
    # check connectivity
3046
    result = self.rpc.call_version([node])[node]
3047
    result.Raise("Can't get version information from node %s" % node)
3048
    if constants.PROTOCOL_VERSION == result.payload:
3049
      logging.info("Communication to node %s fine, sw version %s match",
3050
                   node, result.payload)
3051
    else:
3052
      raise errors.OpExecError("Version mismatch master version %s,"
3053
                               " node version %s" %
3054
                               (constants.PROTOCOL_VERSION, result.payload))
3055

    
3056
    # setup ssh on node
3057
    if self.cfg.GetClusterInfo().modify_ssh_setup:
3058
      logging.info("Copy ssh key to node %s", node)
3059
      priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
3060
      keyarray = []
3061
      keyfiles = [constants.SSH_HOST_DSA_PRIV, constants.SSH_HOST_DSA_PUB,
3062
                  constants.SSH_HOST_RSA_PRIV, constants.SSH_HOST_RSA_PUB,
3063
                  priv_key, pub_key]
3064

    
3065
      for i in keyfiles:
3066
        keyarray.append(utils.ReadFile(i))
3067

    
3068
      result = self.rpc.call_node_add(node, keyarray[0], keyarray[1],
3069
                                      keyarray[2], keyarray[3], keyarray[4],
3070
                                      keyarray[5])
3071
      result.Raise("Cannot transfer ssh keys to the new node")
3072

    
3073
    # Add node to our /etc/hosts, and add key to known_hosts
3074
    if self.cfg.GetClusterInfo().modify_etc_hosts:
3075
      utils.AddHostToEtcHosts(new_node.name)
3076

    
3077
    if new_node.secondary_ip != new_node.primary_ip:
3078
      result = self.rpc.call_node_has_ip_address(new_node.name,
3079
                                                 new_node.secondary_ip)
3080
      result.Raise("Failure checking secondary ip on node %s" % new_node.name,
3081
                   prereq=True, ecode=errors.ECODE_ENVIRON)
3082
      if not result.payload:
3083
        raise errors.OpExecError("Node claims it doesn't have the secondary ip"
3084
                                 " you gave (%s). Please fix and re-run this"
3085
                                 " command." % new_node.secondary_ip)
3086

    
3087
    node_verify_list = [self.cfg.GetMasterNode()]
3088
    node_verify_param = {
3089
      constants.NV_NODELIST: [node],
3090
      # TODO: do a node-net-test as well?
3091
    }
3092

    
3093
    result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
3094
                                       self.cfg.GetClusterName())
3095
    for verifier in node_verify_list:
3096
      result[verifier].Raise("Cannot communicate with node %s" % verifier)
3097
      nl_payload = result[verifier].payload[constants.NV_NODELIST]
3098
      if nl_payload:
3099
        for failed in nl_payload:
3100
          feedback_fn("ssh/hostname verification failed"
3101
                      " (checking from %s): %s" %
3102
                      (verifier, nl_payload[failed]))
3103
        raise errors.OpExecError("ssh/hostname verification failed.")
3104

    
3105
    if self.op.readd:
3106
      _RedistributeAncillaryFiles(self)
3107
      self.context.ReaddNode(new_node)
3108
      # make sure we redistribute the config
3109
      self.cfg.Update(new_node, feedback_fn)
3110
      # and make sure the new node will not have old files around
3111
      if not new_node.master_candidate:
3112
        result = self.rpc.call_node_demote_from_mc(new_node.name)
3113
        msg = result.fail_msg
3114
        if msg:
3115
          self.LogWarning("Node failed to demote itself from master"
3116
                          " candidate status: %s" % msg)
3117
    else:
3118
      _RedistributeAncillaryFiles(self, additional_nodes=[node])
3119
      self.context.AddNode(new_node, self.proc.GetECId())
3120

    
3121

    
3122
class LUSetNodeParams(LogicalUnit):
3123
  """Modifies the parameters of a node.
3124

3125
  """
3126
  HPATH = "node-modify"
3127
  HTYPE = constants.HTYPE_NODE
3128
  _OP_REQP = ["node_name"]
3129
  REQ_BGL = False
3130

    
3131
  def CheckArguments(self):
3132
    node_name = self.cfg.ExpandNodeName(self.op.node_name)
3133
    if node_name is None:
3134
      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name,
3135
                                 errors.ECODE_INVAL)
3136
    self.op.node_name = node_name
3137
    _CheckBooleanOpField(self.op, 'master_candidate')
3138
    _CheckBooleanOpField(self.op, 'offline')
3139
    _CheckBooleanOpField(self.op, 'drained')
3140
    all_mods = [self.op.offline, self.op.master_candidate, self.op.drained]
3141
    if all_mods.count(None) == 3:
3142
      raise errors.OpPrereqError("Please pass at least one modification",
3143
                                 errors.ECODE_INVAL)
3144
    if all_mods.count(True) > 1:
3145
      raise errors.OpPrereqError("Can't set the node into more than one"
3146
                                 " state at the same time",
3147
                                 errors.ECODE_INVAL)
3148

    
3149
  def ExpandNames(self):
3150
    self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
3151

    
3152
  def BuildHooksEnv(self):
3153
    """Build hooks env.
3154

3155
    This runs on the master node.
3156

3157
    """
3158
    env = {
3159
      "OP_TARGET": self.op.node_name,
3160
      "MASTER_CANDIDATE": str(self.op.master_candidate),
3161
      "OFFLINE": str(self.op.offline),
3162
      "DRAINED": str(self.op.drained),
3163
      }
3164
    nl = [self.cfg.GetMasterNode(),
3165
          self.op.node_name]
3166
    return env, nl, nl
3167

    
3168
  def CheckPrereq(self):
3169
    """Check prerequisites.
3170

3171
    This only checks the instance list against the existing names.
3172

3173
    """
3174
    node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
3175

    
3176
    if (self.op.master_candidate is not None or
3177
        self.op.drained is not None or
3178
        self.op.offline is not None):
3179
      # we can't change the master's node flags
3180
      if self.op.node_name == self.cfg.GetMasterNode():
3181
        raise errors.OpPrereqError("The master role can be changed"
3182
                                   " only via masterfailover",
3183
                                   errors.ECODE_INVAL)
3184

    
3185
    # Boolean value that tells us whether we're offlining or draining the node
3186
    offline_or_drain = self.op.offline == True or self.op.drained == True
3187
    deoffline_or_drain = self.op.offline == False or self.op.drained == False
3188

    
3189
    if (node.master_candidate and
3190
        (self.op.master_candidate == False or offline_or_drain)):
3191
      cp_size = self.cfg.GetClusterInfo().candidate_pool_size
3192
      mc_now, mc_should, mc_max = self.cfg.GetMasterCandidateStats()
3193
      if mc_now <= cp_size:
3194
        msg = ("Not enough master candidates (desired"
3195
               " %d, new value will be %d)" % (cp_size, mc_now-1))
3196
        # Only allow forcing the operation if it's an offline/drain operation,
3197
        # and we could not possibly promote more nodes.
3198
        # FIXME: this can still lead to issues if in any way another node which
3199
        # could be promoted appears in the meantime.
3200
        if self.op.force and offline_or_drain and mc_should == mc_max:
3201
          self.LogWarning(msg)
3202
        else:
3203
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
3204

    
3205
    if (self.op.master_candidate == True and
3206
        ((node.offline and not self.op.offline == False) or
3207
         (node.drained and not self.op.drained == False))):
3208
      raise errors.OpPrereqError("Node '%s' is offline or drained, can't set"
3209
                                 " to master_candidate" % node.name,
3210
                                 errors.ECODE_INVAL)
3211

    
3212
    # If we're being deofflined/drained, we'll MC ourself if needed
3213
    if (deoffline_or_drain and not offline_or_drain and not
3214
        self.op.master_candidate == True and not node.master_candidate):
3215
      self.op.master_candidate = _DecideSelfPromotion(self)
3216
      if self.op.master_candidate:
3217
        self.LogInfo("Autopromoting node to master candidate")
3218

    
3219
    return
3220

    
3221
  def Exec(self, feedback_fn):
3222
    """Modifies a node.
3223

3224
    """
3225
    node = self.node
3226

    
3227
    result = []
3228
    changed_mc = False
3229

    
3230
    if self.op.offline is not None:
3231
      node.offline = self.op.offline
3232
      result.append(("offline", str(self.op.offline)))
3233
      if self.op.offline == True:
3234
        if node.master_candidate:
3235
          node.master_candidate = False
3236
          changed_mc = True
3237
          result.append(("master_candidate", "auto-demotion due to offline"))
3238
        if node.drained:
3239
          node.drained = False
3240
          result.append(("drained", "clear drained status due to offline"))
3241

    
3242
    if self.op.master_candidate is not None:
3243
      node.master_candidate = self.op.master_candidate
3244
      changed_mc = True
3245
      result.append(("master_candidate", str(self.op.master_candidate)))
3246
      if self.op.master_candidate == False:
3247
        rrc = self.rpc.call_node_demote_from_mc(node.name)
3248
        msg = rrc.fail_msg
3249
        if msg:
3250
          self.LogWarning("Node failed to demote itself: %s" % msg)
3251

    
3252
    if self.op.drained is not None:
3253
      node.drained = self.op.drained
3254
      result.append(("drained", str(self.op.drained)))
3255
      if self.op.drained == True:
3256
        if node.master_candidate:
3257
          node.master_candidate = False
3258
          changed_mc = True
3259
          result.append(("master_candidate", "auto-demotion due to drain"))
3260
          rrc = self.rpc.call_node_demote_from_mc(node.name)
3261
          msg = rrc.fail_msg
3262
          if msg:
3263
            self.LogWarning("Node failed to demote itself: %s" % msg)
3264
        if node.offline:
3265
          node.offline = False
3266
          result.append(("offline", "clear offline status due to drain"))
3267

    
3268
    # this will trigger configuration file update, if needed
3269
    self.cfg.Update(node, feedback_fn)
3270
    # this will trigger job queue propagation or cleanup
3271
    if changed_mc:
3272
      self.context.ReaddNode(node)
3273

    
3274
    return result
3275

    
3276

    
3277
class LUPowercycleNode(NoHooksLU):
3278
  """Powercycles a node.
3279

3280
  """
3281
  _OP_REQP = ["node_name", "force"]
3282
  REQ_BGL = False
3283

    
3284
  def CheckArguments(self):
3285
    node_name = self.cfg.ExpandNodeName(self.op.node_name)
3286
    if node_name is None:
3287
      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name,
3288
                                 errors.ECODE_NOENT)
3289
    self.op.node_name = node_name
3290
    if node_name == self.cfg.GetMasterNode() and not self.op.force:
3291
      raise errors.OpPrereqError("The node is the master and the force"
3292
                                 " parameter was not set",
3293
                                 errors.ECODE_INVAL)
3294

    
3295
  def ExpandNames(self):
3296
    """Locking for PowercycleNode.
3297

3298
    This is a last-resort option and shouldn't block on other
3299
    jobs. Therefore, we grab no locks.
3300

3301
    """
3302
    self.needed_locks = {}
3303

    
3304
  def CheckPrereq(self):
3305
    """Check prerequisites.
3306

3307
    This LU has no prereqs.
3308

3309
    """
3310
    pass
3311

    
3312
  def Exec(self, feedback_fn):
3313
    """Reboots a node.
3314

3315
    """
3316
    result = self.rpc.call_node_powercycle(self.op.node_name,
3317
                                           self.cfg.GetHypervisorType())
3318
    result.Raise("Failed to schedule the reboot")
3319
    return result.payload
3320

    
3321

    
3322
class LUQueryClusterInfo(NoHooksLU):
3323
  """Query cluster configuration.
3324

3325
  """
3326
  _OP_REQP = []
3327
  REQ_BGL = False
3328

    
3329
  def ExpandNames(self):
3330
    self.needed_locks = {}
3331

    
3332
  def CheckPrereq(self):
3333
    """No prerequsites needed for this LU.
3334

3335
    """
3336
    pass
3337

    
3338
  def Exec(self, feedback_fn):
3339
    """Return cluster config.
3340

3341
    """
3342
    cluster = self.cfg.GetClusterInfo()
3343
    result = {
3344
      "software_version": constants.RELEASE_VERSION,
3345
      "protocol_version": constants.PROTOCOL_VERSION,
3346
      "config_version": constants.CONFIG_VERSION,
3347
      "os_api_version": max(constants.OS_API_VERSIONS),
3348
      "export_version": constants.EXPORT_VERSION,
3349
      "architecture": (platform.architecture()[0], platform.machine()),
3350
      "name": cluster.cluster_name,
3351
      "master": cluster.master_node,
3352
      "default_hypervisor": cluster.enabled_hypervisors[0],
3353
      "enabled_hypervisors": cluster.enabled_hypervisors,
3354
      "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
3355
                        for hypervisor_name in cluster.enabled_hypervisors]),
3356
      "beparams": cluster.beparams,
3357
      "nicparams": cluster.nicparams,
3358
      "candidate_pool_size": cluster.candidate_pool_size,
3359
      "master_netdev": cluster.master_netdev,
3360
      "volume_group_name": cluster.volume_group_name,
3361
      "file_storage_dir": cluster.file_storage_dir,
3362
      "ctime": cluster.ctime,
3363
      "mtime": cluster.mtime,
3364
      "uuid": cluster.uuid,
3365
      "tags": list(cluster.GetTags()),
3366
      }
3367

    
3368
    return result
3369

    
3370

    
3371
class LUQueryConfigValues(NoHooksLU):
3372
  """Return configuration values.
3373

3374
  """
3375
  _OP_REQP = []
3376
  REQ_BGL = False
3377
  _FIELDS_DYNAMIC = utils.FieldSet()
3378
  _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag",
3379
                                  "watcher_pause")
3380

    
3381
  def ExpandNames(self):
3382
    self.needed_locks = {}
3383

    
3384
    _CheckOutputFields(static=self._FIELDS_STATIC,
3385
                       dynamic=self._FIELDS_DYNAMIC,
3386
                       selected=self.op.output_fields)
3387

    
3388
  def CheckPrereq(self):
3389
    """No prerequisites.
3390

3391
    """
3392
    pass
3393

    
3394
  def Exec(self, feedback_fn):
3395
    """Dump a representation of the cluster config to the standard output.
3396

3397
    """
3398
    values = []
3399
    for field in self.op.output_fields:
3400
      if field == "cluster_name":
3401
        entry = self.cfg.GetClusterName()
3402
      elif field == "master_node":
3403
        entry = self.cfg.GetMasterNode()
3404
      elif field == "drain_flag":
3405
        entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
3406
      elif field == "watcher_pause":
3407
        return utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
3408
      else:
3409
        raise errors.ParameterError(field)
3410
      values.append(entry)
3411
    return values
3412

    
3413

    
3414
class LUActivateInstanceDisks(NoHooksLU):
3415
  """Bring up an instance's disks.
3416

3417
  """
3418
  _OP_REQP = ["instance_name"]
3419
  REQ_BGL = False
3420

    
3421
  def ExpandNames(self):
3422
    self._ExpandAndLockInstance()
3423
    self.needed_locks[locking.LEVEL_NODE] = []
3424
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3425

    
3426
  def DeclareLocks(self, level):
3427
    if level == locking.LEVEL_NODE:
3428
      self._LockInstancesNodes()
3429

    
3430
  def CheckPrereq(self):
3431
    """Check prerequisites.
3432

3433
    This checks that the instance is in the cluster.
3434

3435
    """
3436
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3437
    assert self.instance is not None, \
3438
      "Cannot retrieve locked instance %s" % self.op.instance_name
3439
    _CheckNodeOnline(self, self.instance.primary_node)
3440
    if not hasattr(self.op, "ignore_size"):
3441
      self.op.ignore_size = False
3442

    
3443
  def Exec(self, feedback_fn):
3444
    """Activate the disks.
3445

3446
    """
3447
    disks_ok, disks_info = \
3448
              _AssembleInstanceDisks(self, self.instance,
3449
                                     ignore_size=self.op.ignore_size)
3450
    if not disks_ok:
3451
      raise errors.OpExecError("Cannot activate block devices")
3452

    
3453
    return disks_info
3454

    
3455

    
3456
def _AssembleInstanceDisks(lu, instance, ignore_secondaries=False,
3457
                           ignore_size=False):
3458
  """Prepare the block devices for an instance.
3459

3460
  This sets up the block devices on all nodes.
3461

3462
  @type lu: L{LogicalUnit}
3463
  @param lu: the logical unit on whose behalf we execute
3464
  @type instance: L{objects.Instance}
3465
  @param instance: the instance for whose disks we assemble
3466
  @type ignore_secondaries: boolean
3467
  @param ignore_secondaries: if true, errors on secondary nodes
3468
      won't result in an error return from the function
3469
  @type ignore_size: boolean
3470
  @param ignore_size: if true, the current known size of the disk
3471
      will not be used during the disk activation, useful for cases
3472
      when the size is wrong
3473
  @return: False if the operation failed, otherwise a list of
3474
      (host, instance_visible_name, node_visible_name)
3475
      with the mapping from node devices to instance devices
3476

3477
  """
3478
  device_info = []
3479
  disks_ok = True
3480
  iname = instance.name
3481
  # With the two passes mechanism we try to reduce the window of
3482
  # opportunity for the race condition of switching DRBD to primary
3483
  # before handshaking occured, but we do not eliminate it
3484

    
3485
  # The proper fix would be to wait (with some limits) until the
3486
  # connection has been made and drbd transitions from WFConnection
3487
  # into any other network-connected state (Connected, SyncTarget,
3488
  # SyncSource, etc.)
3489

    
3490
  # 1st pass, assemble on all nodes in secondary mode
3491
  for inst_disk in instance.disks:
3492
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
3493
      if ignore_size:
3494
        node_disk = node_disk.Copy()
3495
        node_disk.UnsetSize()
3496
      lu.cfg.SetDiskID(node_disk, node)
3497
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
3498
      msg = result.fail_msg
3499
      if msg:
3500
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
3501
                           " (is_primary=False, pass=1): %s",
3502
                           inst_disk.iv_name, node, msg)
3503
        if not ignore_secondaries:
3504
          disks_ok = False
3505

    
3506
  # FIXME: race condition on drbd migration to primary
3507

    
3508
  # 2nd pass, do only the primary node
3509
  for inst_disk in instance.disks:
3510
    dev_path = None
3511

    
3512
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
3513
      if node != instance.primary_node:
3514
        continue
3515
      if ignore_size:
3516
        node_disk = node_disk.Copy()
3517
        node_disk.UnsetSize()
3518
      lu.cfg.SetDiskID(node_disk, node)
3519
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
3520
      msg = result.fail_msg
3521
      if msg:
3522
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
3523
                           " (is_primary=True, pass=2): %s",
3524
                           inst_disk.iv_name, node, msg)
3525
        disks_ok = False
3526
      else:
3527
        dev_path = result.payload
3528

    
3529
    device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
3530

    
3531
  # leave the disks configured for the primary node
3532
  # this is a workaround that would be fixed better by
3533
  # improving the logical/physical id handling
3534
  for disk in instance.disks:
3535
    lu.cfg.SetDiskID(disk, instance.primary_node)
3536

    
3537
  return disks_ok, device_info
3538

    
3539

    
3540
def _StartInstanceDisks(lu, instance, force):
3541
  """Start the disks of an instance.
3542

3543
  """
3544
  disks_ok, _ = _AssembleInstanceDisks(lu, instance,
3545
                                           ignore_secondaries=force)
3546
  if not disks_ok:
3547
    _ShutdownInstanceDisks(lu, instance)
3548
    if force is not None and not force:
3549
      lu.proc.LogWarning("", hint="If the message above refers to a"
3550
                         " secondary node,"
3551
                         " you can retry the operation using '--force'.")
3552
    raise errors.OpExecError("Disk consistency error")
3553

    
3554

    
3555
class LUDeactivateInstanceDisks(NoHooksLU):
3556
  """Shutdown an instance's disks.
3557

3558
  """
3559
  _OP_REQP = ["instance_name"]
3560
  REQ_BGL = False
3561

    
3562
  def ExpandNames(self):
3563
    self._ExpandAndLockInstance()
3564
    self.needed_locks[locking.LEVEL_NODE] = []
3565
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3566

    
3567
  def DeclareLocks(self, level):
3568
    if level == locking.LEVEL_NODE:
3569
      self._LockInstancesNodes()
3570

    
3571
  def CheckPrereq(self):
3572
    """Check prerequisites.
3573

3574
    This checks that the instance is in the cluster.
3575

3576
    """
3577
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3578
    assert self.instance is not None, \
3579
      "Cannot retrieve locked instance %s" % self.op.instance_name
3580

    
3581
  def Exec(self, feedback_fn):
3582
    """Deactivate the disks
3583

3584
    """
3585
    instance = self.instance
3586
    _SafeShutdownInstanceDisks(self, instance)
3587

    
3588

    
3589
def _SafeShutdownInstanceDisks(lu, instance):
3590
  """Shutdown block devices of an instance.
3591

3592
  This function checks if an instance is running, before calling
3593
  _ShutdownInstanceDisks.
3594

3595
  """
3596
  pnode = instance.primary_node
3597
  ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
3598
  ins_l.Raise("Can't contact node %s" % pnode)
3599

    
3600
  if instance.name in ins_l.payload:
3601
    raise errors.OpExecError("Instance is running, can't shutdown"
3602
                             " block devices.")
3603

    
3604
  _ShutdownInstanceDisks(lu, instance)
3605

    
3606

    
3607
def _ShutdownInstanceDisks(lu, instance, ignore_primary=False):
3608
  """Shutdown block devices of an instance.
3609

3610
  This does the shutdown on all nodes of the instance.
3611

3612
  If the ignore_primary is false, errors on the primary node are
3613
  ignored.
3614

3615
  """
3616
  all_result = True
3617
  for disk in instance.disks:
3618
    for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
3619
      lu.cfg.SetDiskID(top_disk, node)
3620
      result = lu.rpc.call_blockdev_shutdown(node, top_disk)
3621
      msg = result.fail_msg
3622
      if msg:
3623
        lu.LogWarning("Could not shutdown block device %s on node %s: %s",
3624
                      disk.iv_name, node, msg)
3625
        if not ignore_primary or node != instance.primary_node:
3626
          all_result = False
3627
  return all_result
3628

    
3629

    
3630
def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
3631
  """Checks if a node has enough free memory.
3632

3633
  This function check if a given node has the needed amount of free
3634
  memory. In case the node has less memory or we cannot get the
3635
  information from the node, this function raise an OpPrereqError
3636
  exception.
3637

3638
  @type lu: C{LogicalUnit}
3639
  @param lu: a logical unit from which we get configuration data
3640
  @type node: C{str}
3641
  @param node: the node to check
3642
  @type reason: C{str}
3643
  @param reason: string to use in the error message
3644
  @type requested: C{int}
3645
  @param requested: the amount of memory in MiB to check for
3646
  @type hypervisor_name: C{str}
3647
  @param hypervisor_name: the hypervisor to ask for memory stats
3648
  @raise errors.OpPrereqError: if the node doesn't have enough memory, or
3649
      we cannot check the node
3650

3651
  """
3652
  nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
3653
  nodeinfo[node].Raise("Can't get data from node %s" % node,
3654
                       prereq=True, ecode=errors.ECODE_ENVIRON)
3655
  free_mem = nodeinfo[node].payload.get('memory_free', None)
3656
  if not isinstance(free_mem, int):
3657
    raise errors.OpPrereqError("Can't compute free memory on node %s, result"
3658
                               " was '%s'" % (node, free_mem),
3659
                               errors.ECODE_ENVIRON)
3660
  if requested > free_mem:
3661
    raise errors.OpPrereqError("Not enough memory on node %s for %s:"
3662
                               " needed %s MiB, available %s MiB" %
3663
                               (node, reason, requested, free_mem),
3664
                               errors.ECODE_NORES)
3665

    
3666

    
3667
class LUStartupInstance(LogicalUnit):
3668
  """Starts an instance.
3669

3670
  """
3671
  HPATH = "instance-start"
3672
  HTYPE = constants.HTYPE_INSTANCE
3673
  _OP_REQP = ["instance_name", "force"]
3674
  REQ_BGL = False
3675

    
3676
  def ExpandNames(self):
3677
    self._ExpandAndLockInstance()
3678

    
3679
  def BuildHooksEnv(self):
3680
    """Build hooks env.
3681

3682
    This runs on master, primary and secondary nodes of the instance.
3683

3684
    """
3685
    env = {
3686
      "FORCE": self.op.force,
3687
      }
3688
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
3689
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3690
    return env, nl, nl
3691

    
3692
  def CheckPrereq(self):
3693
    """Check prerequisites.
3694

3695
    This checks that the instance is in the cluster.
3696

3697
    """
3698
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3699
    assert self.instance is not None, \
3700
      "Cannot retrieve locked instance %s" % self.op.instance_name
3701

    
3702
    # extra beparams
3703
    self.beparams = getattr(self.op, "beparams", {})
3704
    if self.beparams:
3705
      if not isinstance(self.beparams, dict):
3706
        raise errors.OpPrereqError("Invalid beparams passed: %s, expected"
3707
                                   " dict" % (type(self.beparams), ),
3708
                                   errors.ECODE_INVAL)
3709
      # fill the beparams dict
3710
      utils.ForceDictType(self.beparams, constants.BES_PARAMETER_TYPES)
3711
      self.op.beparams = self.beparams
3712

    
3713
    # extra hvparams
3714
    self.hvparams = getattr(self.op, "hvparams", {})
3715
    if self.hvparams:
3716
      if not isinstance(self.hvparams, dict):
3717
        raise errors.OpPrereqError("Invalid hvparams passed: %s, expected"
3718
                                   " dict" % (type(self.hvparams), ),
3719
                                   errors.ECODE_INVAL)
3720

    
3721
      # check hypervisor parameter syntax (locally)
3722
      cluster = self.cfg.GetClusterInfo()
3723
      utils.ForceDictType(self.hvparams, constants.HVS_PARAMETER_TYPES)
3724
      filled_hvp = objects.FillDict(cluster.hvparams[instance.hypervisor],
3725
                                    instance.hvparams)
3726
      filled_hvp.update(self.hvparams)
3727
      hv_type = hypervisor.GetHypervisor(instance.hypervisor)
3728
      hv_type.CheckParameterSyntax(filled_hvp)
3729
      _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
3730
      self.op.hvparams = self.hvparams
3731

    
3732
    _CheckNodeOnline(self, instance.primary_node)
3733

    
3734
    bep = self.cfg.GetClusterInfo().FillBE(instance)
3735
    # check bridges existence
3736
    _CheckInstanceBridgesExist(self, instance)
3737

    
3738
    remote_info = self.rpc.call_instance_info(instance.primary_node,
3739
                                              instance.name,
3740
                                              instance.hypervisor)
3741
    remote_info.Raise("Error checking node %s" % instance.primary_node,
3742
                      prereq=True, ecode=errors.ECODE_ENVIRON)
3743
    if not remote_info.payload: # not running already
3744
      _CheckNodeFreeMemory(self, instance.primary_node,
3745
                           "starting instance %s" % instance.name,
3746
                           bep[constants.BE_MEMORY], instance.hypervisor)
3747

    
3748
  def Exec(self, feedback_fn):
3749
    """Start the instance.
3750

3751
    """
3752
    instance = self.instance
3753
    force = self.op.force
3754

    
3755
    self.cfg.MarkInstanceUp(instance.name)
3756

    
3757
    node_current = instance.primary_node
3758

    
3759
    _StartInstanceDisks(self, instance, force)
3760

    
3761
    result = self.rpc.call_instance_start(node_current, instance,
3762
                                          self.hvparams, self.beparams)
3763
    msg = result.fail_msg
3764
    if msg:
3765
      _ShutdownInstanceDisks(self, instance)
3766
      raise errors.OpExecError("Could not start instance: %s" % msg)
3767

    
3768

    
3769
class LURebootInstance(LogicalUnit):
3770
  """Reboot an instance.
3771

3772
  """
3773
  HPATH = "instance-reboot"
3774
  HTYPE = constants.HTYPE_INSTANCE
3775
  _OP_REQP = ["instance_name", "ignore_secondaries", "reboot_type"]
3776
  REQ_BGL = False
3777

    
3778
  def CheckArguments(self):
3779
    """Check the arguments.
3780

3781
    """
3782
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
3783
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
3784

    
3785
  def ExpandNames(self):
3786
    if self.op.reboot_type not in [constants.INSTANCE_REBOOT_SOFT,
3787
                                   constants.INSTANCE_REBOOT_HARD,
3788
                                   constants.INSTANCE_REBOOT_FULL]:
3789
      raise errors.ParameterError("reboot type not in [%s, %s, %s]" %
3790
                                  (constants.INSTANCE_REBOOT_SOFT,
3791
                                   constants.INSTANCE_REBOOT_HARD,
3792
                                   constants.INSTANCE_REBOOT_FULL))
3793
    self._ExpandAndLockInstance()
3794

    
3795
  def BuildHooksEnv(self):
3796
    """Build hooks env.
3797

3798
    This runs on master, primary and secondary nodes of the instance.
3799

3800
    """
3801
    env = {
3802
      "IGNORE_SECONDARIES": self.op.ignore_secondaries,
3803
      "REBOOT_TYPE": self.op.reboot_type,
3804
      "SHUTDOWN_TIMEOUT": self.shutdown_timeout,
3805
      }
3806
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
3807
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3808
    return env, nl, nl
3809

    
3810
  def CheckPrereq(self):
3811
    """Check prerequisites.
3812

3813
    This checks that the instance is in the cluster.
3814

3815
    """
3816
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3817
    assert self.instance is not None, \
3818
      "Cannot retrieve locked instance %s" % self.op.instance_name
3819

    
3820
    _CheckNodeOnline(self, instance.primary_node)
3821

    
3822
    # check bridges existence
3823
    _CheckInstanceBridgesExist(self, instance)
3824

    
3825
  def Exec(self, feedback_fn):
3826
    """Reboot the instance.
3827

3828
    """
3829
    instance = self.instance
3830
    ignore_secondaries = self.op.ignore_secondaries
3831
    reboot_type = self.op.reboot_type
3832

    
3833
    node_current = instance.primary_node
3834

    
3835
    if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
3836
                       constants.INSTANCE_REBOOT_HARD]:
3837
      for disk in instance.disks:
3838
        self.cfg.SetDiskID(disk, node_current)
3839
      result = self.rpc.call_instance_reboot(node_current, instance,
3840
                                             reboot_type,
3841
                                             self.shutdown_timeout)
3842
      result.Raise("Could not reboot instance")
3843
    else:
3844
      result = self.rpc.call_instance_shutdown(node_current, instance,
3845
                                               self.shutdown_timeout)
3846
      result.Raise("Could not shutdown instance for full reboot")
3847
      _ShutdownInstanceDisks(self, instance)
3848
      _StartInstanceDisks(self, instance, ignore_secondaries)
3849
      result = self.rpc.call_instance_start(node_current, instance, None, None)
3850
      msg = result.fail_msg
3851
      if msg:
3852
        _ShutdownInstanceDisks(self, instance)
3853
        raise errors.OpExecError("Could not start instance for"
3854
                                 " full reboot: %s" % msg)
3855

    
3856
    self.cfg.MarkInstanceUp(instance.name)
3857

    
3858

    
3859
class LUShutdownInstance(LogicalUnit):
3860
  """Shutdown an instance.
3861

3862
  """
3863
  HPATH = "instance-stop"
3864
  HTYPE = constants.HTYPE_INSTANCE
3865
  _OP_REQP = ["instance_name"]
3866
  REQ_BGL = False
3867

    
3868
  def CheckArguments(self):
3869
    """Check the arguments.
3870

3871
    """
3872
    self.timeout = getattr(self.op, "timeout",
3873
                           constants.DEFAULT_SHUTDOWN_TIMEOUT)
3874

    
3875
  def ExpandNames(self):
3876
    self._ExpandAndLockInstance()
3877

    
3878
  def BuildHooksEnv(self):
3879
    """Build hooks env.
3880

3881
    This runs on master, primary and secondary nodes of the instance.
3882

3883
    """
3884
    env = _BuildInstanceHookEnvByObject(self, self.instance)
3885
    env["TIMEOUT"] = self.timeout
3886
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3887
    return env, nl, nl
3888

    
3889
  def CheckPrereq(self):
3890
    """Check prerequisites.
3891

3892
    This checks that the instance is in the cluster.
3893

3894
    """
3895
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3896
    assert self.instance is not None, \
3897
      "Cannot retrieve locked instance %s" % self.op.instance_name
3898
    _CheckNodeOnline(self, self.instance.primary_node)
3899

    
3900
  def Exec(self, feedback_fn):
3901
    """Shutdown the instance.
3902

3903
    """
3904
    instance = self.instance
3905
    node_current = instance.primary_node
3906
    timeout = self.timeout
3907
    self.cfg.MarkInstanceDown(instance.name)
3908
    result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
3909
    msg = result.fail_msg
3910
    if msg:
3911
      self.proc.LogWarning("Could not shutdown instance: %s" % msg)
3912

    
3913
    _ShutdownInstanceDisks(self, instance)
3914

    
3915

    
3916
class LUReinstallInstance(LogicalUnit):
3917
  """Reinstall an instance.
3918

3919
  """
3920
  HPATH = "instance-reinstall"
3921
  HTYPE = constants.HTYPE_INSTANCE
3922
  _OP_REQP = ["instance_name"]
3923
  REQ_BGL = False
3924

    
3925
  def ExpandNames(self):
3926
    self._ExpandAndLockInstance()
3927

    
3928
  def BuildHooksEnv(self):
3929
    """Build hooks env.
3930

3931
    This runs on master, primary and secondary nodes of the instance.
3932

3933
    """
3934
    env = _BuildInstanceHookEnvByObject(self, self.instance)
3935
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3936
    return env, nl, nl
3937

    
3938
  def CheckPrereq(self):
3939
    """Check prerequisites.
3940

3941
    This checks that the instance is in the cluster and is not running.
3942

3943
    """
3944
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3945
    assert instance is not None, \
3946
      "Cannot retrieve locked instance %s" % self.op.instance_name
3947
    _CheckNodeOnline(self, instance.primary_node)
3948

    
3949
    if instance.disk_template == constants.DT_DISKLESS:
3950
      raise errors.OpPrereqError("Instance '%s' has no disks" %
3951
                                 self.op.instance_name,
3952
                                 errors.ECODE_INVAL)
3953
    if instance.admin_up:
3954
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
3955
                                 self.op.instance_name,
3956
                                 errors.ECODE_STATE)
3957
    remote_info = self.rpc.call_instance_info(instance.primary_node,
3958
                                              instance.name,
3959
                                              instance.hypervisor)
3960
    remote_info.Raise("Error checking node %s" % instance.primary_node,
3961
                      prereq=True, ecode=errors.ECODE_ENVIRON)
3962
    if remote_info.payload:
3963
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
3964
                                 (self.op.instance_name,
3965
                                  instance.primary_node),
3966
                                 errors.ECODE_STATE)
3967

    
3968
    self.op.os_type = getattr(self.op, "os_type", None)
3969
    self.op.force_variant = getattr(self.op, "force_variant", False)
3970
    if self.op.os_type is not None:
3971
      # OS verification
3972
      pnode = self.cfg.GetNodeInfo(
3973
        self.cfg.ExpandNodeName(instance.primary_node))
3974
      if pnode is None:
3975
        raise errors.OpPrereqError("Primary node '%s' is unknown" %
3976
                                   self.op.pnode, errors.ECODE_NOENT)
3977
      result = self.rpc.call_os_get(pnode.name, self.op.os_type)
3978
      result.Raise("OS '%s' not in supported OS list for primary node %s" %
3979
                   (self.op.os_type, pnode.name),
3980
                   prereq=True, ecode=errors.ECODE_INVAL)
3981
      if not self.op.force_variant:
3982
        _CheckOSVariant(result.payload, self.op.os_type)
3983

    
3984
    self.instance = instance
3985

    
3986
  def Exec(self, feedback_fn):
3987
    """Reinstall the instance.
3988

3989
    """
3990
    inst = self.instance
3991

    
3992
    if self.op.os_type is not None:
3993
      feedback_fn("Changing OS to '%s'..." % self.op.os_type)
3994
      inst.os = self.op.os_type
3995
      self.cfg.Update(inst, feedback_fn)
3996

    
3997
    _StartInstanceDisks(self, inst, None)
3998
    try:
3999
      feedback_fn("Running the instance OS create scripts...")
4000
      # FIXME: pass debug option from opcode to backend
4001
      result = self.rpc.call_instance_os_add(inst.primary_node, inst, True,
4002
                                             self.op.debug_level)
4003
      result.Raise("Could not install OS for instance %s on node %s" %
4004
                   (inst.name, inst.primary_node))
4005
    finally:
4006
      _ShutdownInstanceDisks(self, inst)
4007

    
4008

    
4009
class LURecreateInstanceDisks(LogicalUnit):
4010
  """Recreate an instance's missing disks.
4011

4012
  """
4013
  HPATH = "instance-recreate-disks"
4014
  HTYPE = constants.HTYPE_INSTANCE
4015
  _OP_REQP = ["instance_name", "disks"]
4016
  REQ_BGL = False
4017

    
4018
  def CheckArguments(self):
4019
    """Check the arguments.
4020

4021
    """
4022
    if not isinstance(self.op.disks, list):
4023
      raise errors.OpPrereqError("Invalid disks parameter", errors.ECODE_INVAL)
4024
    for item in self.op.disks:
4025
      if (not isinstance(item, int) or
4026
          item < 0):
4027
        raise errors.OpPrereqError("Invalid disk specification '%s'" %
4028
                                   str(item), errors.ECODE_INVAL)
4029

    
4030
  def ExpandNames(self):
4031
    self._ExpandAndLockInstance()
4032

    
4033
  def BuildHooksEnv(self):
4034
    """Build hooks env.
4035

4036
    This runs on master, primary and secondary nodes of the instance.
4037

4038
    """
4039
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4040
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4041
    return env, nl, nl
4042

    
4043
  def CheckPrereq(self):
4044
    """Check prerequisites.
4045

4046
    This checks that the instance is in the cluster and is not running.
4047

4048
    """
4049
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4050
    assert instance is not None, \
4051
      "Cannot retrieve locked instance %s" % self.op.instance_name
4052
    _CheckNodeOnline(self, instance.primary_node)
4053

    
4054
    if instance.disk_template == constants.DT_DISKLESS:
4055
      raise errors.OpPrereqError("Instance '%s' has no disks" %
4056
                                 self.op.instance_name, errors.ECODE_INVAL)
4057
    if instance.admin_up:
4058
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
4059
                                 self.op.instance_name, errors.ECODE_STATE)
4060
    remote_info = self.rpc.call_instance_info(instance.primary_node,
4061
                                              instance.name,
4062
                                              instance.hypervisor)
4063
    remote_info.Raise("Error checking node %s" % instance.primary_node,
4064
                      prereq=True, ecode=errors.ECODE_ENVIRON)
4065
    if remote_info.payload:
4066
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
4067
                                 (self.op.instance_name,
4068
                                  instance.primary_node), errors.ECODE_STATE)
4069

    
4070
    if not self.op.disks:
4071
      self.op.disks = range(len(instance.disks))
4072
    else:
4073
      for idx in self.op.disks:
4074
        if idx >= len(instance.disks):
4075
          raise errors.OpPrereqError("Invalid disk index passed '%s'" % idx,
4076
                                     errors.ECODE_INVAL)
4077

    
4078
    self.instance = instance
4079

    
4080
  def Exec(self, feedback_fn):
4081
    """Recreate the disks.
4082

4083
    """
4084
    to_skip = []
4085
    for idx, _ in enumerate(self.instance.disks):
4086
      if idx not in self.op.disks: # disk idx has not been passed in
4087
        to_skip.append(idx)
4088
        continue
4089

    
4090
    _CreateDisks(self, self.instance, to_skip=to_skip)
4091

    
4092

    
4093
class LURenameInstance(LogicalUnit):
4094
  """Rename an instance.
4095

4096
  """
4097
  HPATH = "instance-rename"
4098
  HTYPE = constants.HTYPE_INSTANCE
4099
  _OP_REQP = ["instance_name", "new_name"]
4100

    
4101
  def BuildHooksEnv(self):
4102
    """Build hooks env.
4103

4104
    This runs on master, primary and secondary nodes of the instance.
4105

4106
    """
4107
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4108
    env["INSTANCE_NEW_NAME"] = self.op.new_name
4109
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4110
    return env, nl, nl
4111

    
4112
  def CheckPrereq(self):
4113
    """Check prerequisites.
4114

4115
    This checks that the instance is in the cluster and is not running.
4116

4117
    """
4118
    instance = self.cfg.GetInstanceInfo(
4119
      self.cfg.ExpandInstanceName(self.op.instance_name))
4120
    if instance is None:
4121
      raise errors.OpPrereqError("Instance '%s' not known" %
4122
                                 self.op.instance_name, errors.ECODE_NOENT)
4123
    _CheckNodeOnline(self, instance.primary_node)
4124

    
4125
    if instance.admin_up:
4126
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
4127
                                 self.op.instance_name, errors.ECODE_STATE)
4128
    remote_info = self.rpc.call_instance_info(instance.primary_node,
4129
                                              instance.name,
4130
                                              instance.hypervisor)
4131
    remote_info.Raise("Error checking node %s" % instance.primary_node,
4132
                      prereq=True, ecode=errors.ECODE_ENVIRON)
4133
    if remote_info.payload:
4134
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
4135
                                 (self.op.instance_name,
4136
                                  instance.primary_node), errors.ECODE_STATE)
4137
    self.instance = instance
4138

    
4139
    # new name verification
4140
    name_info = utils.GetHostInfo(self.op.new_name)
4141

    
4142
    self.op.new_name = new_name = name_info.name
4143
    instance_list = self.cfg.GetInstanceList()
4144
    if new_name in instance_list:
4145
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
4146
                                 new_name, errors.ECODE_EXISTS)
4147

    
4148
    if not getattr(self.op, "ignore_ip", False):
4149
      if utils.TcpPing(name_info.ip, constants.DEFAULT_NODED_PORT):
4150
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
4151
                                   (name_info.ip, new_name),
4152
                                   errors.ECODE_NOTUNIQUE)
4153

    
4154

    
4155
  def Exec(self, feedback_fn):
4156
    """Reinstall the instance.
4157

4158
    """
4159
    inst = self.instance
4160
    old_name = inst.name
4161

    
4162
    if inst.disk_template == constants.DT_FILE:
4163
      old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
4164

    
4165
    self.cfg.RenameInstance(inst.name, self.op.new_name)
4166
    # Change the instance lock. This is definitely safe while we hold the BGL
4167
    self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
4168
    self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
4169

    
4170
    # re-read the instance from the configuration after rename
4171
    inst = self.cfg.GetInstanceInfo(self.op.new_name)
4172

    
4173
    if inst.disk_template == constants.DT_FILE:
4174
      new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
4175
      result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
4176
                                                     old_file_storage_dir,
4177
                                                     new_file_storage_dir)
4178
      result.Raise("Could not rename on node %s directory '%s' to '%s'"
4179
                   " (but the instance has been renamed in Ganeti)" %
4180
                   (inst.primary_node, old_file_storage_dir,
4181
                    new_file_storage_dir))
4182

    
4183
    _StartInstanceDisks(self, inst, None)
4184
    try:
4185
      result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
4186
                                                 old_name, self.op.debug_level)
4187
      msg = result.fail_msg
4188
      if msg:
4189
        msg = ("Could not run OS rename script for instance %s on node %s"
4190
               " (but the instance has been renamed in Ganeti): %s" %
4191
               (inst.name, inst.primary_node, msg))
4192
        self.proc.LogWarning(msg)
4193
    finally:
4194
      _ShutdownInstanceDisks(self, inst)
4195

    
4196

    
4197
class LURemoveInstance(LogicalUnit):
4198
  """Remove an instance.
4199

4200
  """
4201
  HPATH = "instance-remove"
4202
  HTYPE = constants.HTYPE_INSTANCE
4203
  _OP_REQP = ["instance_name", "ignore_failures"]
4204
  REQ_BGL = False
4205

    
4206
  def CheckArguments(self):
4207
    """Check the arguments.
4208

4209
    """
4210
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
4211
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
4212

    
4213
  def ExpandNames(self):
4214
    self._ExpandAndLockInstance()
4215
    self.needed_locks[locking.LEVEL_NODE] = []
4216
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4217

    
4218
  def DeclareLocks(self, level):
4219
    if level == locking.LEVEL_NODE:
4220
      self._LockInstancesNodes()
4221

    
4222
  def BuildHooksEnv(self):
4223
    """Build hooks env.
4224

4225
    This runs on master, primary and secondary nodes of the instance.
4226

4227
    """
4228
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4229
    env["SHUTDOWN_TIMEOUT"] = self.shutdown_timeout
4230
    nl = [self.cfg.GetMasterNode()]
4231
    return env, nl, nl
4232

    
4233
  def CheckPrereq(self):
4234
    """Check prerequisites.
4235

4236
    This checks that the instance is in the cluster.
4237

4238
    """
4239
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4240
    assert self.instance is not None, \
4241
      "Cannot retrieve locked instance %s" % self.op.instance_name
4242

    
4243
  def Exec(self, feedback_fn):
4244
    """Remove the instance.
4245

4246
    """
4247
    instance = self.instance
4248
    logging.info("Shutting down instance %s on node %s",
4249
                 instance.name, instance.primary_node)
4250

    
4251
    result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
4252
                                             self.shutdown_timeout)
4253
    msg = result.fail_msg
4254
    if msg:
4255
      if self.op.ignore_failures:
4256
        feedback_fn("Warning: can't shutdown instance: %s" % msg)
4257
      else:
4258
        raise errors.OpExecError("Could not shutdown instance %s on"
4259
                                 " node %s: %s" %
4260
                                 (instance.name, instance.primary_node, msg))
4261

    
4262
    logging.info("Removing block devices for instance %s", instance.name)
4263

    
4264
    if not _RemoveDisks(self, instance):
4265
      if self.op.ignore_failures:
4266
        feedback_fn("Warning: can't remove instance's disks")
4267
      else:
4268
        raise errors.OpExecError("Can't remove instance's disks")
4269

    
4270
    logging.info("Removing instance %s out of cluster config", instance.name)
4271

    
4272
    self.cfg.RemoveInstance(instance.name)
4273
    self.remove_locks[locking.LEVEL_INSTANCE] = instance.name
4274

    
4275

    
4276
class LUQueryInstances(NoHooksLU):
4277
  """Logical unit for querying instances.
4278

4279
  """
4280
  # pylint: disable-msg=W0142
4281
  _OP_REQP = ["output_fields", "names", "use_locking"]
4282
  REQ_BGL = False
4283
  _SIMPLE_FIELDS = ["name", "os", "network_port", "hypervisor",
4284
                    "serial_no", "ctime", "mtime", "uuid"]
4285
  _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
4286
                                    "admin_state",
4287
                                    "disk_template", "ip", "mac", "bridge",
4288
                                    "nic_mode", "nic_link",
4289
                                    "sda_size", "sdb_size", "vcpus", "tags",
4290
                                    "network_port", "beparams",
4291
                                    r"(disk)\.(size)/([0-9]+)",
4292
                                    r"(disk)\.(sizes)", "disk_usage",
4293
                                    r"(nic)\.(mac|ip|mode|link)/([0-9]+)",
4294
                                    r"(nic)\.(bridge)/([0-9]+)",
4295
                                    r"(nic)\.(macs|ips|modes|links|bridges)",
4296
                                    r"(disk|nic)\.(count)",
4297
                                    "hvparams",
4298
                                    ] + _SIMPLE_FIELDS +
4299
                                  ["hv/%s" % name
4300
                                   for name in constants.HVS_PARAMETERS
4301
                                   if name not in constants.HVC_GLOBALS] +
4302
                                  ["be/%s" % name
4303
                                   for name in constants.BES_PARAMETERS])
4304
  _FIELDS_DYNAMIC = utils.FieldSet("oper_state", "oper_ram", "status")
4305

    
4306

    
4307
  def ExpandNames(self):
4308
    _CheckOutputFields(static=self._FIELDS_STATIC,
4309
                       dynamic=self._FIELDS_DYNAMIC,
4310
                       selected=self.op.output_fields)
4311

    
4312
    self.needed_locks = {}
4313
    self.share_locks[locking.LEVEL_INSTANCE] = 1
4314
    self.share_locks[locking.LEVEL_NODE] = 1
4315

    
4316
    if self.op.names:
4317
      self.wanted = _GetWantedInstances(self, self.op.names)
4318
    else:
4319
      self.wanted = locking.ALL_SET
4320

    
4321
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
4322
    self.do_locking = self.do_node_query and self.op.use_locking
4323
    if self.do_locking:
4324
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
4325
      self.needed_locks[locking.LEVEL_NODE] = []
4326
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4327

    
4328
  def DeclareLocks(self, level):
4329
    if level == locking.LEVEL_NODE and self.do_locking:
4330
      self._LockInstancesNodes()
4331

    
4332
  def CheckPrereq(self):
4333
    """Check prerequisites.
4334

4335
    """
4336
    pass
4337

    
4338
  def Exec(self, feedback_fn):
4339
    """Computes the list of nodes and their attributes.
4340

4341
    """
4342
    # pylint: disable-msg=R0912
4343
    # way too many branches here
4344
    all_info = self.cfg.GetAllInstancesInfo()
4345
    if self.wanted == locking.ALL_SET:
4346
      # caller didn't specify instance names, so ordering is not important
4347
      if self.do_locking:
4348
        instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
4349
      else:
4350
        instance_names = all_info.keys()
4351
      instance_names = utils.NiceSort(instance_names)
4352
    else:
4353
      # caller did specify names, so we must keep the ordering
4354
      if self.do_locking:
4355
        tgt_set = self.acquired_locks[locking.LEVEL_INSTANCE]
4356
      else:
4357
        tgt_set = all_info.keys()
4358
      missing = set(self.wanted).difference(tgt_set)
4359
      if missing:
4360
        raise errors.OpExecError("Some instances were removed before"
4361
                                 " retrieving their data: %s" % missing)
4362
      instance_names = self.wanted
4363

    
4364
    instance_list = [all_info[iname] for iname in instance_names]
4365

    
4366
    # begin data gathering
4367

    
4368
    nodes = frozenset([inst.primary_node for inst in instance_list])
4369
    hv_list = list(set([inst.hypervisor for inst in instance_list]))
4370

    
4371
    bad_nodes = []
4372
    off_nodes = []
4373
    if self.do_node_query:
4374
      live_data = {}
4375
      node_data = self.rpc.call_all_instances_info(nodes, hv_list)
4376
      for name in nodes:
4377
        result = node_data[name]
4378
        if result.offline:
4379
          # offline nodes will be in both lists
4380
          off_nodes.append(name)
4381
        if result.fail_msg:
4382
          bad_nodes.append(name)
4383
        else:
4384
          if result.payload:
4385
            live_data.update(result.payload)
4386
          # else no instance is alive
4387
    else:
4388
      live_data = dict([(name, {}) for name in instance_names])
4389

    
4390
    # end data gathering
4391

    
4392
    HVPREFIX = "hv/"
4393
    BEPREFIX = "be/"
4394
    output = []
4395
    cluster = self.cfg.GetClusterInfo()
4396
    for instance in instance_list:
4397
      iout = []
4398
      i_hv = cluster.FillHV(instance, skip_globals=True)
4399
      i_be = cluster.FillBE(instance)
4400
      i_nicp = [objects.FillDict(cluster.nicparams[constants.PP_DEFAULT],
4401
                                 nic.nicparams) for nic in instance.nics]
4402
      for field in self.op.output_fields:
4403
        st_match = self._FIELDS_STATIC.Matches(field)
4404
        if field in self._SIMPLE_FIELDS:
4405
          val = getattr(instance, field)
4406
        elif field == "pnode":
4407
          val = instance.primary_node
4408
        elif field == "snodes":
4409
          val = list(instance.secondary_nodes)
4410
        elif field == "admin_state":
4411
          val = instance.admin_up
4412
        elif field == "oper_state":
4413
          if instance.primary_node in bad_nodes:
4414
            val = None
4415
          else:
4416
            val = bool(live_data.get(instance.name))
4417
        elif field == "status":
4418
          if instance.primary_node in off_nodes:
4419
            val = "ERROR_nodeoffline"
4420
          elif instance.primary_node in bad_nodes:
4421
            val = "ERROR_nodedown"
4422
          else:
4423
            running = bool(live_data.get(instance.name))
4424
            if running:
4425
              if instance.admin_up:
4426
                val = "running"
4427
              else:
4428
                val = "ERROR_up"
4429
            else:
4430
              if instance.admin_up:
4431
                val = "ERROR_down"
4432
              else:
4433
                val = "ADMIN_down"
4434
        elif field == "oper_ram":
4435
          if instance.primary_node in bad_nodes:
4436
            val = None
4437
          elif instance.name in live_data:
4438
            val = live_data[instance.name].get("memory", "?")
4439
          else:
4440
            val = "-"
4441
        elif field == "vcpus":
4442
          val = i_be[constants.BE_VCPUS]
4443
        elif field == "disk_template":
4444
          val = instance.disk_template
4445
        elif field == "ip":
4446
          if instance.nics:
4447
            val = instance.nics[0].ip
4448
          else:
4449
            val = None
4450
        elif field == "nic_mode":
4451
          if instance.nics:
4452
            val = i_nicp[0][constants.NIC_MODE]
4453
          else:
4454
            val = None
4455
        elif field == "nic_link":
4456
          if instance.nics:
4457
            val = i_nicp[0][constants.NIC_LINK]
4458
          else:
4459
            val = None
4460
        elif field == "bridge":
4461
          if (instance.nics and
4462
              i_nicp[0][constants.NIC_MODE] == constants.NIC_MODE_BRIDGED):
4463
            val = i_nicp[0][constants.NIC_LINK]
4464
          else:
4465
            val = None
4466
        elif field == "mac":
4467
          if instance.nics:
4468
            val = instance.nics[0].mac
4469
          else:
4470
            val = None
4471
        elif field == "sda_size" or field == "sdb_size":
4472
          idx = ord(field[2]) - ord('a')
4473
          try:
4474
            val = instance.FindDisk(idx).size
4475
          except errors.OpPrereqError:
4476
            val = None
4477
        elif field == "disk_usage": # total disk usage per node
4478
          disk_sizes = [{'size': disk.size} for disk in instance.disks]
4479
          val = _ComputeDiskSize(instance.disk_template, disk_sizes)
4480
        elif field == "tags":
4481
          val = list(instance.GetTags())
4482
        elif field == "hvparams":
4483
          val = i_hv
4484
        elif (field.startswith(HVPREFIX) and
4485
              field[len(HVPREFIX):] in constants.HVS_PARAMETERS and
4486
              field[len(HVPREFIX):] not in constants.HVC_GLOBALS):
4487
          val = i_hv.get(field[len(HVPREFIX):], None)
4488
        elif field == "beparams":
4489
          val = i_be
4490
        elif (field.startswith(BEPREFIX) and
4491
              field[len(BEPREFIX):] in constants.BES_PARAMETERS):
4492
          val = i_be.get(field[len(BEPREFIX):], None)
4493
        elif st_match and st_match.groups():
4494
          # matches a variable list
4495
          st_groups = st_match.groups()
4496
          if st_groups and st_groups[0] == "disk":
4497
            if st_groups[1] == "count":
4498
              val = len(instance.disks)
4499
            elif st_groups[1] == "sizes":
4500
              val = [disk.size for disk in instance.disks]
4501
            elif st_groups[1] == "size":
4502
              try:
4503
                val = instance.FindDisk(st_groups[2]).size
4504
              except errors.OpPrereqError:
4505
                val = None
4506
            else:
4507
              assert False, "Unhandled disk parameter"
4508
          elif st_groups[0] == "nic":
4509
            if st_groups[1] == "count":
4510
              val = len(instance.nics)
4511
            elif st_groups[1] == "macs":
4512
              val = [nic.mac for nic in instance.nics]
4513
            elif st_groups[1] == "ips":
4514
              val = [nic.ip for nic in instance.nics]
4515
            elif st_groups[1] == "modes":
4516
              val = [nicp[constants.NIC_MODE] for nicp in i_nicp]
4517
            elif st_groups[1] == "links":
4518
              val = [nicp[constants.NIC_LINK] for nicp in i_nicp]
4519
            elif st_groups[1] == "bridges":
4520
              val = []
4521
              for nicp in i_nicp:
4522
                if nicp[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
4523
                  val.append(nicp[constants.NIC_LINK])
4524
                else:
4525
                  val.append(None)
4526
            else:
4527
              # index-based item
4528
              nic_idx = int(st_groups[2])
4529
              if nic_idx >= len(instance.nics):
4530
                val = None
4531
              else:
4532
                if st_groups[1] == "mac":
4533
                  val = instance.nics[nic_idx].mac
4534
                elif st_groups[1] == "ip":
4535
                  val = instance.nics[nic_idx].ip
4536
                elif st_groups[1] == "mode":
4537
                  val = i_nicp[nic_idx][constants.NIC_MODE]
4538
                elif st_groups[1] == "link":
4539
                  val = i_nicp[nic_idx][constants.NIC_LINK]
4540
                elif st_groups[1] == "bridge":
4541
                  nic_mode = i_nicp[nic_idx][constants.NIC_MODE]
4542
                  if nic_mode == constants.NIC_MODE_BRIDGED:
4543
                    val = i_nicp[nic_idx][constants.NIC_LINK]
4544
                  else:
4545
                    val = None
4546
                else:
4547
                  assert False, "Unhandled NIC parameter"
4548
          else:
4549
            assert False, ("Declared but unhandled variable parameter '%s'" %
4550
                           field)
4551
        else:
4552
          assert False, "Declared but unhandled parameter '%s'" % field
4553
        iout.append(val)
4554
      output.append(iout)
4555

    
4556
    return output
4557

    
4558

    
4559
class LUFailoverInstance(LogicalUnit):
4560
  """Failover an instance.
4561

4562
  """
4563
  HPATH = "instance-failover"
4564
  HTYPE = constants.HTYPE_INSTANCE
4565
  _OP_REQP = ["instance_name", "ignore_consistency"]
4566
  REQ_BGL = False
4567

    
4568
  def CheckArguments(self):
4569
    """Check the arguments.
4570

4571
    """
4572
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
4573
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
4574

    
4575
  def ExpandNames(self):
4576
    self._ExpandAndLockInstance()
4577
    self.needed_locks[locking.LEVEL_NODE] = []
4578
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4579

    
4580
  def DeclareLocks(self, level):
4581
    if level == locking.LEVEL_NODE:
4582
      self._LockInstancesNodes()
4583

    
4584
  def BuildHooksEnv(self):
4585
    """Build hooks env.
4586

4587
    This runs on master, primary and secondary nodes of the instance.
4588

4589
    """
4590
    instance = self.instance
4591
    source_node = instance.primary_node
4592
    target_node = instance.secondary_nodes[0]
4593
    env = {
4594
      "IGNORE_CONSISTENCY": self.op.ignore_consistency,
4595
      "SHUTDOWN_TIMEOUT": self.shutdown_timeout,
4596
      "OLD_PRIMARY": source_node,
4597
      "OLD_SECONDARY": target_node,
4598
      "NEW_PRIMARY": target_node,
4599
      "NEW_SECONDARY": source_node,
4600
      }
4601
    env.update(_BuildInstanceHookEnvByObject(self, instance))
4602
    nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
4603
    return env, nl, nl
4604

    
4605
  def CheckPrereq(self):
4606
    """Check prerequisites.
4607

4608
    This checks that the instance is in the cluster.
4609

4610
    """
4611
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4612
    assert self.instance is not None, \
4613
      "Cannot retrieve locked instance %s" % self.op.instance_name
4614

    
4615
    bep = self.cfg.GetClusterInfo().FillBE(instance)
4616
    if instance.disk_template not in constants.DTS_NET_MIRROR:
4617
      raise errors.OpPrereqError("Instance's disk layout is not"
4618
                                 " network mirrored, cannot failover.",
4619
                                 errors.ECODE_STATE)
4620

    
4621
    secondary_nodes = instance.secondary_nodes
4622
    if not secondary_nodes:
4623
      raise errors.ProgrammerError("no secondary node but using "
4624
                                   "a mirrored disk template")
4625

    
4626
    target_node = secondary_nodes[0]
4627
    _CheckNodeOnline(self, target_node)
4628
    _CheckNodeNotDrained(self, target_node)
4629
    if instance.admin_up:
4630
      # check memory requirements on the secondary node
4631
      _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
4632
                           instance.name, bep[constants.BE_MEMORY],
4633
                           instance.hypervisor)
4634
    else:
4635
      self.LogInfo("Not checking memory on the secondary node as"
4636
                   " instance will not be started")
4637

    
4638
    # check bridge existance
4639
    _CheckInstanceBridgesExist(self, instance, node=target_node)
4640

    
4641
  def Exec(self, feedback_fn):
4642
    """Failover an instance.
4643

4644
    The failover is done by shutting it down on its present node and
4645
    starting it on the secondary.
4646

4647
    """
4648
    instance = self.instance
4649

    
4650
    source_node = instance.primary_node
4651
    target_node = instance.secondary_nodes[0]
4652

    
4653
    if instance.admin_up:
4654
      feedback_fn("* checking disk consistency between source and target")
4655
      for dev in instance.disks:
4656
        # for drbd, these are drbd over lvm
4657
        if not _CheckDiskConsistency(self, dev, target_node, False):
4658
          if not self.op.ignore_consistency:
4659
            raise errors.OpExecError("Disk %s is degraded on target node,"
4660
                                     " aborting failover." % dev.iv_name)
4661
    else:
4662
      feedback_fn("* not checking disk consistency as instance is not running")
4663

    
4664
    feedback_fn("* shutting down instance on source node")
4665
    logging.info("Shutting down instance %s on node %s",
4666
                 instance.name, source_node)
4667

    
4668
    result = self.rpc.call_instance_shutdown(source_node, instance,
4669
                                             self.shutdown_timeout)
4670
    msg = result.fail_msg
4671
    if msg:
4672
      if self.op.ignore_consistency:
4673
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
4674
                             " Proceeding anyway. Please make sure node"
4675
                             " %s is down. Error details: %s",
4676
                             instance.name, source_node, source_node, msg)
4677
      else:
4678
        raise errors.OpExecError("Could not shutdown instance %s on"
4679
                                 " node %s: %s" %
4680
                                 (instance.name, source_node, msg))
4681

    
4682
    feedback_fn("* deactivating the instance's disks on source node")
4683
    if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
4684
      raise errors.OpExecError("Can't shut down the instance's disks.")
4685

    
4686
    instance.primary_node = target_node
4687
    # distribute new instance config to the other nodes
4688
    self.cfg.Update(instance, feedback_fn)
4689

    
4690
    # Only start the instance if it's marked as up
4691
    if instance.admin_up:
4692
      feedback_fn("* activating the instance's disks on target node")
4693
      logging.info("Starting instance %s on node %s",
4694
                   instance.name, target_node)
4695

    
4696
      disks_ok, _ = _AssembleInstanceDisks(self, instance,
4697
                                               ignore_secondaries=True)
4698
      if not disks_ok:
4699
        _ShutdownInstanceDisks(self, instance)
4700
        raise errors.OpExecError("Can't activate the instance's disks")
4701

    
4702
      feedback_fn("* starting the instance on the target node")
4703
      result = self.rpc.call_instance_start(target_node, instance, None, None)
4704
      msg = result.fail_msg
4705
      if msg:
4706
        _ShutdownInstanceDisks(self, instance)
4707
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
4708
                                 (instance.name, target_node, msg))
4709

    
4710

    
4711
class LUMigrateInstance(LogicalUnit):
4712
  """Migrate an instance.
4713

4714
  This is migration without shutting down, compared to the failover,
4715
  which is done with shutdown.
4716

4717
  """
4718
  HPATH = "instance-migrate"
4719
  HTYPE = constants.HTYPE_INSTANCE
4720
  _OP_REQP = ["instance_name", "live", "cleanup"]
4721

    
4722
  REQ_BGL = False
4723

    
4724
  def ExpandNames(self):
4725
    self._ExpandAndLockInstance()
4726

    
4727
    self.needed_locks[locking.LEVEL_NODE] = []
4728
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4729

    
4730
    self._migrater = TLMigrateInstance(self, self.op.instance_name,
4731
                                       self.op.live, self.op.cleanup)
4732
    self.tasklets = [self._migrater]
4733

    
4734
  def DeclareLocks(self, level):
4735
    if level == locking.LEVEL_NODE:
4736
      self._LockInstancesNodes()
4737

    
4738
  def BuildHooksEnv(self):
4739
    """Build hooks env.
4740

4741
    This runs on master, primary and secondary nodes of the instance.
4742

4743
    """
4744
    instance = self._migrater.instance
4745
    source_node = instance.primary_node
4746
    target_node = instance.secondary_nodes[0]
4747
    env = _BuildInstanceHookEnvByObject(self, instance)
4748
    env["MIGRATE_LIVE"] = self.op.live
4749
    env["MIGRATE_CLEANUP"] = self.op.cleanup
4750
    env.update({
4751
        "OLD_PRIMARY": source_node,
4752
        "OLD_SECONDARY": target_node,
4753
        "NEW_PRIMARY": target_node,
4754
        "NEW_SECONDARY": source_node,
4755
        })
4756
    nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
4757
    return env, nl, nl
4758

    
4759

    
4760
class LUMoveInstance(LogicalUnit):
4761
  """Move an instance by data-copying.
4762

4763
  """
4764
  HPATH = "instance-move"
4765
  HTYPE = constants.HTYPE_INSTANCE
4766
  _OP_REQP = ["instance_name", "target_node"]
4767
  REQ_BGL = False
4768

    
4769
  def CheckArguments(self):
4770
    """Check the arguments.
4771

4772
    """
4773
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
4774
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
4775

    
4776
  def ExpandNames(self):
4777
    self._ExpandAndLockInstance()
4778
    target_node = self.cfg.ExpandNodeName(self.op.target_node)
4779
    if target_node is None:
4780
      raise errors.OpPrereqError("Node '%s' not known" %
4781
                                  self.op.target_node, errors.ECODE_NOENT)
4782
    self.op.target_node = target_node
4783
    self.needed_locks[locking.LEVEL_NODE] = [target_node]
4784
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
4785

    
4786
  def DeclareLocks(self, level):
4787
    if level == locking.LEVEL_NODE:
4788
      self._LockInstancesNodes(primary_only=True)
4789

    
4790
  def BuildHooksEnv(self):
4791
    """Build hooks env.
4792

4793
    This runs on master, primary and secondary nodes of the instance.
4794

4795
    """
4796
    env = {
4797
      "TARGET_NODE": self.op.target_node,
4798
      "SHUTDOWN_TIMEOUT": self.shutdown_timeout,
4799
      }
4800
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4801
    nl = [self.cfg.GetMasterNode()] + [self.instance.primary_node,
4802
                                       self.op.target_node]
4803
    return env, nl, nl
4804

    
4805
  def CheckPrereq(self):
4806
    """Check prerequisites.
4807

4808
    This checks that the instance is in the cluster.
4809

4810
    """
4811
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4812
    assert self.instance is not None, \
4813
      "Cannot retrieve locked instance %s" % self.op.instance_name
4814

    
4815
    node = self.cfg.GetNodeInfo(self.op.target_node)
4816
    assert node is not None, \
4817
      "Cannot retrieve locked node %s" % self.op.target_node
4818

    
4819
    self.target_node = target_node = node.name
4820

    
4821
    if target_node == instance.primary_node:
4822
      raise errors.OpPrereqError("Instance %s is already on the node %s" %
4823
                                 (instance.name, target_node),
4824
                                 errors.ECODE_STATE)
4825

    
4826
    bep = self.cfg.GetClusterInfo().FillBE(instance)
4827

    
4828
    for idx, dsk in enumerate(instance.disks):
4829
      if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
4830
        raise errors.OpPrereqError("Instance disk %d has a complex layout,"
4831
                                   " cannot copy" % idx, errors.ECODE_STATE)
4832

    
4833
    _CheckNodeOnline(self, target_node)
4834
    _CheckNodeNotDrained(self, target_node)
4835

    
4836
    if instance.admin_up:
4837
      # check memory requirements on the secondary node
4838
      _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
4839
                           instance.name, bep[constants.BE_MEMORY],
4840
                           instance.hypervisor)
4841
    else:
4842
      self.LogInfo("Not checking memory on the secondary node as"
4843
                   " instance will not be started")
4844

    
4845
    # check bridge existance
4846
    _CheckInstanceBridgesExist(self, instance, node=target_node)
4847

    
4848
  def Exec(self, feedback_fn):
4849
    """Move an instance.
4850

4851
    The move is done by shutting it down on its present node, copying
4852
    the data over (slow) and starting it on the new node.
4853

4854
    """
4855
    instance = self.instance
4856

    
4857
    source_node = instance.primary_node
4858
    target_node = self.target_node
4859

    
4860
    self.LogInfo("Shutting down instance %s on source node %s",
4861
                 instance.name, source_node)
4862

    
4863
    result = self.rpc.call_instance_shutdown(source_node, instance,
4864
                                             self.shutdown_timeout)
4865
    msg = result.fail_msg
4866
    if msg:
4867
      if self.op.ignore_consistency:
4868
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
4869
                             " Proceeding anyway. Please make sure node"
4870
                             " %s is down. Error details: %s",
4871
                             instance.name, source_node, source_node, msg)
4872
      else:
4873
        raise errors.OpExecError("Could not shutdown instance %s on"
4874
                                 " node %s: %s" %
4875
                                 (instance.name, source_node, msg))
4876

    
4877
    # create the target disks
4878
    try:
4879
      _CreateDisks(self, instance, target_node=target_node)
4880
    except errors.OpExecError:
4881
      self.LogWarning("Device creation failed, reverting...")
4882
      try:
4883
        _RemoveDisks(self, instance, target_node=target_node)
4884
      finally:
4885
        self.cfg.ReleaseDRBDMinors(instance.name)
4886
        raise
4887

    
4888
    cluster_name = self.cfg.GetClusterInfo().cluster_name
4889

    
4890
    errs = []
4891
    # activate, get path, copy the data over
4892
    for idx, disk in enumerate(instance.disks):
4893
      self.LogInfo("Copying data for disk %d", idx)
4894
      result = self.rpc.call_blockdev_assemble(target_node, disk,
4895
                                               instance.name, True)
4896
      if result.fail_msg:
4897
        self.LogWarning("Can't assemble newly created disk %d: %s",
4898
                        idx, result.fail_msg)
4899
        errs.append(result.fail_msg)
4900
        break
4901
      dev_path = result.payload
4902
      result = self.rpc.call_blockdev_export(source_node, disk,
4903
                                             target_node, dev_path,
4904
                                             cluster_name)
4905
      if result.fail_msg:
4906
        self.LogWarning("Can't copy data over for disk %d: %s",
4907
                        idx, result.fail_msg)
4908
        errs.append(result.fail_msg)
4909
        break
4910

    
4911
    if errs:
4912
      self.LogWarning("Some disks failed to copy, aborting")
4913
      try:
4914
        _RemoveDisks(self, instance, target_node=target_node)
4915
      finally:
4916
        self.cfg.ReleaseDRBDMinors(instance.name)
4917
        raise errors.OpExecError("Errors during disk copy: %s" %
4918
                                 (",".join(errs),))
4919

    
4920
    instance.primary_node = target_node
4921
    self.cfg.Update(instance, feedback_fn)
4922

    
4923
    self.LogInfo("Removing the disks on the original node")
4924
    _RemoveDisks(self, instance, target_node=source_node)
4925

    
4926
    # Only start the instance if it's marked as up
4927
    if instance.admin_up:
4928
      self.LogInfo("Starting instance %s on node %s",
4929
                   instance.name, target_node)
4930

    
4931
      disks_ok, _ = _AssembleInstanceDisks(self, instance,
4932
                                           ignore_secondaries=True)
4933
      if not disks_ok:
4934
        _ShutdownInstanceDisks(self, instance)
4935
        raise errors.OpExecError("Can't activate the instance's disks")
4936

    
4937
      result = self.rpc.call_instance_start(target_node, instance, None, None)
4938
      msg = result.fail_msg
4939
      if msg:
4940
        _ShutdownInstanceDisks(self, instance)
4941
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
4942
                                 (instance.name, target_node, msg))
4943

    
4944

    
4945
class LUMigrateNode(LogicalUnit):
4946
  """Migrate all instances from a node.
4947

4948
  """
4949
  HPATH = "node-migrate"
4950
  HTYPE = constants.HTYPE_NODE
4951
  _OP_REQP = ["node_name", "live"]
4952
  REQ_BGL = False
4953

    
4954
  def ExpandNames(self):
4955
    self.op.node_name = self.cfg.ExpandNodeName(self.op.node_name)
4956
    if self.op.node_name is None:
4957
      raise errors.OpPrereqError("Node '%s' not known" % self.op.node_name,
4958
                                 errors.ECODE_NOENT)
4959

    
4960
    self.needed_locks = {
4961
      locking.LEVEL_NODE: [self.op.node_name],
4962
      }
4963

    
4964
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
4965

    
4966
    # Create tasklets for migrating instances for all instances on this node
4967
    names = []
4968
    tasklets = []
4969

    
4970
    for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name):
4971
      logging.debug("Migrating instance %s", inst.name)
4972
      names.append(inst.name)
4973

    
4974
      tasklets.append(TLMigrateInstance(self, inst.name, self.op.live, False))
4975

    
4976
    self.tasklets = tasklets
4977

    
4978
    # Declare instance locks
4979
    self.needed_locks[locking.LEVEL_INSTANCE] = names
4980

    
4981
  def DeclareLocks(self, level):
4982
    if level == locking.LEVEL_NODE:
4983
      self._LockInstancesNodes()
4984

    
4985
  def BuildHooksEnv(self):
4986
    """Build hooks env.
4987

4988
    This runs on the master, the primary and all the secondaries.
4989

4990
    """
4991
    env = {
4992
      "NODE_NAME": self.op.node_name,
4993
      }
4994

    
4995
    nl = [self.cfg.GetMasterNode()]
4996

    
4997
    return (env, nl, nl)
4998

    
4999

    
5000
class TLMigrateInstance(Tasklet):
5001
  def __init__(self, lu, instance_name, live, cleanup):
5002
    """Initializes this class.
5003

5004
    """
5005
    Tasklet.__init__(self, lu)
5006

    
5007
    # Parameters
5008
    self.instance_name = instance_name
5009
    self.live = live
5010
    self.cleanup = cleanup
5011

    
5012
  def CheckPrereq(self):
5013
    """Check prerequisites.
5014

5015
    This checks that the instance is in the cluster.
5016

5017
    """
5018
    instance = self.cfg.GetInstanceInfo(
5019
      self.cfg.ExpandInstanceName(self.instance_name))
5020
    if instance is None:
5021
      raise errors.OpPrereqError("Instance '%s' not known" %
5022
                                 self.instance_name, errors.ECODE_NOENT)
5023

    
5024
    if instance.disk_template != constants.DT_DRBD8:
5025
      raise errors.OpPrereqError("Instance's disk layout is not"
5026
                                 " drbd8, cannot migrate.", errors.ECODE_STATE)
5027

    
5028
    secondary_nodes = instance.secondary_nodes
5029
    if not secondary_nodes:
5030
      raise errors.ConfigurationError("No secondary node but using"
5031
                                      " drbd8 disk template")
5032

    
5033
    i_be = self.cfg.GetClusterInfo().FillBE(instance)
5034

    
5035
    target_node = secondary_nodes[0]
5036
    # check memory requirements on the secondary node
5037
    _CheckNodeFreeMemory(self, target_node, "migrating instance %s" %
5038
                         instance.name, i_be[constants.BE_MEMORY],
5039
                         instance.hypervisor)
5040

    
5041
    # check bridge existance
5042
    _CheckInstanceBridgesExist(self, instance, node=target_node)
5043

    
5044
    if not self.cleanup:
5045
      _CheckNodeNotDrained(self, target_node)
5046
      result = self.rpc.call_instance_migratable(instance.primary_node,
5047
                                                 instance)
5048
      result.Raise("Can't migrate, please use failover",
5049
                   prereq=True, ecode=errors.ECODE_STATE)
5050

    
5051
    self.instance = instance
5052

    
5053
  def _WaitUntilSync(self):
5054
    """Poll with custom rpc for disk sync.
5055

5056
    This uses our own step-based rpc call.
5057

5058
    """
5059
    self.feedback_fn("* wait until resync is done")
5060
    all_done = False
5061
    while not all_done:
5062
      all_done = True
5063
      result = self.rpc.call_drbd_wait_sync(self.all_nodes,
5064
                                            self.nodes_ip,
5065
                                            self.instance.disks)
5066
      min_percent = 100
5067
      for node, nres in result.items():
5068
        nres.Raise("Cannot resync disks on node %s" % node)
5069
        node_done, node_percent = nres.payload
5070
        all_done = all_done and node_done
5071
        if node_percent is not None:
5072
          min_percent = min(min_percent, node_percent)
5073
      if not all_done:
5074
        if min_percent < 100:
5075
          self.feedback_fn("   - progress: %.1f%%" % min_percent)
5076
        time.sleep(2)
5077

    
5078
  def _EnsureSecondary(self, node):
5079
    """Demote a node to secondary.
5080

5081
    """
5082
    self.feedback_fn("* switching node %s to secondary mode" % node)
5083

    
5084
    for dev in self.instance.disks:
5085
      self.cfg.SetDiskID(dev, node)
5086

    
5087
    result = self.rpc.call_blockdev_close(node, self.instance.name,
5088
                                          self.instance.disks)
5089
    result.Raise("Cannot change disk to secondary on node %s" % node)
5090

    
5091
  def _GoStandalone(self):
5092
    """Disconnect from the network.
5093

5094
    """
5095
    self.feedback_fn("* changing into standalone mode")
5096
    result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
5097
                                               self.instance.disks)
5098
    for node, nres in result.items():
5099
      nres.Raise("Cannot disconnect disks node %s" % node)
5100

    
5101
  def _GoReconnect(self, multimaster):
5102
    """Reconnect to the network.
5103

5104
    """
5105
    if multimaster:
5106
      msg = "dual-master"
5107
    else:
5108
      msg = "single-master"
5109
    self.feedback_fn("* changing disks into %s mode" % msg)
5110
    result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
5111
                                           self.instance.disks,
5112
                                           self.instance.name, multimaster)
5113
    for node, nres in result.items():
5114
      nres.Raise("Cannot change disks config on node %s" % node)
5115

    
5116
  def _ExecCleanup(self):
5117
    """Try to cleanup after a failed migration.
5118

5119
    The cleanup is done by:
5120
      - check that the instance is running only on one node
5121
        (and update the config if needed)
5122
      - change disks on its secondary node to secondary
5123
      - wait until disks are fully synchronized
5124
      - disconnect from the network
5125
      - change disks into single-master mode
5126
      - wait again until disks are fully synchronized
5127

5128
    """
5129
    instance = self.instance
5130
    target_node = self.target_node
5131
    source_node = self.source_node
5132

    
5133
    # check running on only one node
5134
    self.feedback_fn("* checking where the instance actually runs"
5135
                     " (if this hangs, the hypervisor might be in"
5136
                     " a bad state)")
5137
    ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
5138
    for node, result in ins_l.items():
5139
      result.Raise("Can't contact node %s" % node)
5140

    
5141
    runningon_source = instance.name in ins_l[source_node].payload
5142
    runningon_target = instance.name in ins_l[target_node].payload
5143

    
5144
    if runningon_source and runningon_target:
5145
      raise errors.OpExecError("Instance seems to be running on two nodes,"
5146
                               " or the hypervisor is confused. You will have"
5147
                               " to ensure manually that it runs only on one"
5148
                               " and restart this operation.")
5149

    
5150
    if not (runningon_source or runningon_target):
5151
      raise errors.OpExecError("Instance does not seem to be running at all."
5152
                               " In this case, it's safer to repair by"
5153
                               " running 'gnt-instance stop' to ensure disk"
5154
                               " shutdown, and then restarting it.")
5155

    
5156
    if runningon_target:
5157
      # the migration has actually succeeded, we need to update the config
5158
      self.feedback_fn("* instance running on secondary node (%s),"
5159
                       " updating config" % target_node)
5160
      instance.primary_node = target_node
5161
      self.cfg.Update(instance, self.feedback_fn)
5162
      demoted_node = source_node
5163
    else:
5164
      self.feedback_fn("* instance confirmed to be running on its"
5165
                       " primary node (%s)" % source_node)
5166
      demoted_node = target_node
5167

    
5168
    self._EnsureSecondary(demoted_node)
5169
    try:
5170
      self._WaitUntilSync()
5171
    except errors.OpExecError:
5172
      # we ignore here errors, since if the device is standalone, it
5173
      # won't be able to sync
5174
      pass
5175
    self._GoStandalone()
5176
    self._GoReconnect(False)
5177
    self._WaitUntilSync()
5178

    
5179
    self.feedback_fn("* done")
5180

    
5181
  def _RevertDiskStatus(self):
5182
    """Try to revert the disk status after a failed migration.
5183

5184
    """
5185
    target_node = self.target_node
5186
    try:
5187
      self._EnsureSecondary(target_node)
5188
      self._GoStandalone()
5189
      self._GoReconnect(False)
5190
      self._WaitUntilSync()
5191
    except errors.OpExecError, err:
5192
      self.lu.LogWarning("Migration failed and I can't reconnect the"
5193
                         " drives: error '%s'\n"
5194
                         "Please look and recover the instance status" %
5195
                         str(err))
5196

    
5197
  def _AbortMigration(self):
5198
    """Call the hypervisor code to abort a started migration.
5199

5200
    """
5201
    instance = self.instance
5202
    target_node = self.target_node
5203
    migration_info = self.migration_info
5204

    
5205
    abort_result = self.rpc.call_finalize_migration(target_node,
5206
                                                    instance,
5207
                                                    migration_info,
5208
                                                    False)
5209
    abort_msg = abort_result.fail_msg
5210
    if abort_msg:
5211
      logging.error("Aborting migration failed on target node %s: %s",
5212
                    target_node, abort_msg)
5213
      # Don't raise an exception here, as we stil have to try to revert the
5214
      # disk status, even if this step failed.
5215

    
5216
  def _ExecMigration(self):
5217
    """Migrate an instance.
5218

5219
    The migrate is done by:
5220
      - change the disks into dual-master mode
5221
      - wait until disks are fully synchronized again
5222
      - migrate the instance
5223
      - change disks on the new secondary node (the old primary) to secondary
5224
      - wait until disks are fully synchronized
5225
      - change disks into single-master mode
5226

5227
    """
5228
    instance = self.instance
5229
    target_node = self.target_node
5230
    source_node = self.source_node
5231

    
5232
    self.feedback_fn("* checking disk consistency between source and target")
5233
    for dev in instance.disks:
5234
      if not _CheckDiskConsistency(self, dev, target_node, False):
5235
        raise errors.OpExecError("Disk %s is degraded or not fully"
5236
                                 " synchronized on target node,"
5237
                                 " aborting migrate." % dev.iv_name)
5238

    
5239
    # First get the migration information from the remote node
5240
    result = self.rpc.call_migration_info(source_node, instance)
5241
    msg = result.fail_msg
5242
    if msg:
5243
      log_err = ("Failed fetching source migration information from %s: %s" %
5244
                 (source_node, msg))
5245
      logging.error(log_err)
5246
      raise errors.OpExecError(log_err)
5247

    
5248
    self.migration_info = migration_info = result.payload
5249

    
5250
    # Then switch the disks to master/master mode
5251
    self._EnsureSecondary(target_node)
5252
    self._GoStandalone()
5253
    self._GoReconnect(True)
5254
    self._WaitUntilSync()
5255

    
5256
    self.feedback_fn("* preparing %s to accept the instance" % target_node)
5257
    result = self.rpc.call_accept_instance(target_node,
5258
                                           instance,
5259
                                           migration_info,
5260
                                           self.nodes_ip[target_node])
5261

    
5262
    msg = result.fail_msg
5263
    if msg:
5264
      logging.error("Instance pre-migration failed, trying to revert"
5265
                    " disk status: %s", msg)
5266
      self.feedback_fn("Pre-migration failed, aborting")
5267
      self._AbortMigration()
5268
      self._RevertDiskStatus()
5269
      raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
5270
                               (instance.name, msg))
5271

    
5272
    self.feedback_fn("* migrating instance to %s" % target_node)
5273
    time.sleep(10)
5274
    result = self.rpc.call_instance_migrate(source_node, instance,
5275
                                            self.nodes_ip[target_node],
5276
                                            self.live)
5277
    msg = result.fail_msg
5278
    if msg:
5279
      logging.error("Instance migration failed, trying to revert"
5280
                    " disk status: %s", msg)
5281
      self.feedback_fn("Migration failed, aborting")
5282
      self._AbortMigration()
5283
      self._RevertDiskStatus()
5284
      raise errors.OpExecError("Could not migrate instance %s: %s" %
5285
                               (instance.name, msg))
5286
    time.sleep(10)
5287

    
5288
    instance.primary_node = target_node
5289
    # distribute new instance config to the other nodes
5290
    self.cfg.Update(instance, self.feedback_fn)
5291

    
5292
    result = self.rpc.call_finalize_migration(target_node,
5293
                                              instance,
5294
                                              migration_info,
5295
                                              True)
5296
    msg = result.fail_msg
5297
    if msg:
5298
      logging.error("Instance migration succeeded, but finalization failed:"
5299
                    " %s", msg)
5300
      raise errors.OpExecError("Could not finalize instance migration: %s" %
5301
                               msg)
5302

    
5303
    self._EnsureSecondary(source_node)
5304
    self._WaitUntilSync()
5305
    self._GoStandalone()
5306
    self._GoReconnect(False)
5307
    self._WaitUntilSync()
5308

    
5309
    self.feedback_fn("* done")
5310

    
5311
  def Exec(self, feedback_fn):
5312
    """Perform the migration.
5313

5314
    """
5315
    feedback_fn("Migrating instance %s" % self.instance.name)
5316

    
5317
    self.feedback_fn = feedback_fn
5318

    
5319
    self.source_node = self.instance.primary_node
5320
    self.target_node = self.instance.secondary_nodes[0]
5321
    self.all_nodes = [self.source_node, self.target_node]
5322
    self.nodes_ip = {
5323
      self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
5324
      self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
5325
      }
5326

    
5327
    if self.cleanup:
5328
      return self._ExecCleanup()
5329
    else:
5330
      return self._ExecMigration()
5331

    
5332

    
5333
def _CreateBlockDev(lu, node, instance, device, force_create,
5334
                    info, force_open):
5335
  """Create a tree of block devices on a given node.
5336

5337
  If this device type has to be created on secondaries, create it and
5338
  all its children.
5339

5340
  If not, just recurse to children keeping the same 'force' value.
5341

5342
  @param lu: the lu on whose behalf we execute
5343
  @param node: the node on which to create the device
5344
  @type instance: L{objects.Instance}
5345
  @param instance: the instance which owns the device
5346
  @type device: L{objects.Disk}
5347
  @param device: the device to create
5348
  @type force_create: boolean
5349
  @param force_create: whether to force creation of this device; this
5350
      will be change to True whenever we find a device which has
5351
      CreateOnSecondary() attribute
5352
  @param info: the extra 'metadata' we should attach to the device
5353
      (this will be represented as a LVM tag)
5354
  @type force_open: boolean
5355
  @param force_open: this parameter will be passes to the
5356
      L{backend.BlockdevCreate} function where it specifies
5357
      whether we run on primary or not, and it affects both
5358
      the child assembly and the device own Open() execution
5359

5360
  """
5361
  if device.CreateOnSecondary():
5362
    force_create = True
5363

    
5364
  if device.children:
5365
    for child in device.children:
5366
      _CreateBlockDev(lu, node, instance, child, force_create,
5367
                      info, force_open)
5368

    
5369
  if not force_create:
5370
    return
5371

    
5372
  _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
5373

    
5374

    
5375
def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
5376
  """Create a single block device on a given node.
5377

5378
  This will not recurse over children of the device, so they must be
5379
  created in advance.
5380

5381
  @param lu: the lu on whose behalf we execute
5382
  @param node: the node on which to create the device
5383
  @type instance: L{objects.Instance}
5384
  @param instance: the instance which owns the device
5385
  @type device: L{objects.Disk}
5386
  @param device: the device to create
5387
  @param info: the extra 'metadata' we should attach to the device
5388
      (this will be represented as a LVM tag)
5389
  @type force_open: boolean
5390
  @param force_open: this parameter will be passes to the
5391
      L{backend.BlockdevCreate} function where it specifies
5392
      whether we run on primary or not, and it affects both
5393
      the child assembly and the device own Open() execution
5394

5395
  """
5396
  lu.cfg.SetDiskID(device, node)
5397
  result = lu.rpc.call_blockdev_create(node, device, device.size,
5398
                                       instance.name, force_open, info)
5399
  result.Raise("Can't create block device %s on"
5400
               " node %s for instance %s" % (device, node, instance.name))
5401
  if device.physical_id is None:
5402
    device.physical_id = result.payload
5403

    
5404

    
5405
def _GenerateUniqueNames(lu, exts):
5406
  """Generate a suitable LV name.
5407

5408
  This will generate a logical volume name for the given instance.
5409

5410
  """
5411
  results = []
5412
  for val in exts:
5413
    new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
5414
    results.append("%s%s" % (new_id, val))
5415
  return results
5416

    
5417

    
5418
def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
5419
                         p_minor, s_minor):
5420
  """Generate a drbd8 device complete with its children.
5421

5422
  """
5423
  port = lu.cfg.AllocatePort()
5424
  vgname = lu.cfg.GetVGName()
5425
  shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
5426
  dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
5427
                          logical_id=(vgname, names[0]))
5428
  dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
5429
                          logical_id=(vgname, names[1]))
5430
  drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
5431
                          logical_id=(primary, secondary, port,
5432
                                      p_minor, s_minor,
5433
                                      shared_secret),
5434
                          children=[dev_data, dev_meta],
5435
                          iv_name=iv_name)
5436
  return drbd_dev
5437

    
5438

    
5439
def _GenerateDiskTemplate(lu, template_name,
5440
                          instance_name, primary_node,
5441
                          secondary_nodes, disk_info,
5442
                          file_storage_dir, file_driver,
5443
                          base_index):
5444
  """Generate the entire disk layout for a given template type.
5445

5446
  """
5447
  #TODO: compute space requirements
5448

    
5449
  vgname = lu.cfg.GetVGName()
5450
  disk_count = len(disk_info)
5451
  disks = []
5452
  if template_name == constants.DT_DISKLESS:
5453
    pass
5454
  elif template_name == constants.DT_PLAIN:
5455
    if len(secondary_nodes) != 0:
5456
      raise errors.ProgrammerError("Wrong template configuration")
5457

    
5458
    names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
5459
                                      for i in range(disk_count)])
5460
    for idx, disk in enumerate(disk_info):
5461
      disk_index = idx + base_index
5462
      disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
5463
                              logical_id=(vgname, names[idx]),
5464
                              iv_name="disk/%d" % disk_index,
5465
                              mode=disk["mode"])
5466
      disks.append(disk_dev)
5467
  elif template_name == constants.DT_DRBD8:
5468
    if len(secondary_nodes) != 1:
5469
      raise errors.ProgrammerError("Wrong template configuration")
5470
    remote_node = secondary_nodes[0]
5471
    minors = lu.cfg.AllocateDRBDMinor(
5472
      [primary_node, remote_node] * len(disk_info), instance_name)
5473

    
5474
    names = []
5475
    for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
5476
                                               for i in range(disk_count)]):
5477
      names.append(lv_prefix + "_data")
5478
      names.append(lv_prefix + "_meta")
5479
    for idx, disk in enumerate(disk_info):
5480
      disk_index = idx + base_index
5481
      disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
5482
                                      disk["size"], names[idx*2:idx*2+2],
5483
                                      "disk/%d" % disk_index,
5484
                                      minors[idx*2], minors[idx*2+1])
5485
      disk_dev.mode = disk["mode"]
5486
      disks.append(disk_dev)
5487
  elif template_name == constants.DT_FILE:
5488
    if len(secondary_nodes) != 0:
5489
      raise errors.ProgrammerError("Wrong template configuration")
5490

    
5491
    for idx, disk in enumerate(disk_info):
5492
      disk_index = idx + base_index
5493
      disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
5494
                              iv_name="disk/%d" % disk_index,
5495
                              logical_id=(file_driver,
5496
                                          "%s/disk%d" % (file_storage_dir,
5497
                                                         disk_index)),
5498
                              mode=disk["mode"])
5499
      disks.append(disk_dev)
5500
  else:
5501
    raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
5502
  return disks
5503

    
5504

    
5505
def _GetInstanceInfoText(instance):
5506
  """Compute that text that should be added to the disk's metadata.
5507

5508
  """
5509
  return "originstname+%s" % instance.name
5510

    
5511

    
5512
def _CreateDisks(lu, instance, to_skip=None, target_node=None):
5513
  """Create all disks for an instance.
5514

5515
  This abstracts away some work from AddInstance.
5516

5517
  @type lu: L{LogicalUnit}
5518
  @param lu: the logical unit on whose behalf we execute
5519
  @type instance: L{objects.Instance}
5520
  @param instance: the instance whose disks we should create
5521
  @type to_skip: list
5522
  @param to_skip: list of indices to skip
5523
  @type target_node: string
5524
  @param target_node: if passed, overrides the target node for creation
5525
  @rtype: boolean
5526
  @return: the success of the creation
5527

5528
  """
5529
  info = _GetInstanceInfoText(instance)
5530
  if target_node is None:
5531
    pnode = instance.primary_node
5532
    all_nodes = instance.all_nodes
5533
  else:
5534
    pnode = target_node
5535
    all_nodes = [pnode]
5536

    
5537
  if instance.disk_template == constants.DT_FILE:
5538
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
5539
    result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
5540

    
5541
    result.Raise("Failed to create directory '%s' on"
5542
                 " node %s" % (file_storage_dir, pnode))
5543

    
5544
  # Note: this needs to be kept in sync with adding of disks in
5545
  # LUSetInstanceParams
5546
  for idx, device in enumerate(instance.disks):
5547
    if to_skip and idx in to_skip:
5548
      continue
5549
    logging.info("Creating volume %s for instance %s",
5550
                 device.iv_name, instance.name)
5551
    #HARDCODE
5552
    for node in all_nodes:
5553
      f_create = node == pnode
5554
      _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
5555

    
5556

    
5557
def _RemoveDisks(lu, instance, target_node=None):
5558
  """Remove all disks for an instance.
5559

5560
  This abstracts away some work from `AddInstance()` and
5561
  `RemoveInstance()`. Note that in case some of the devices couldn't
5562
  be removed, the removal will continue with the other ones (compare
5563
  with `_CreateDisks()`).
5564

5565
  @type lu: L{LogicalUnit}
5566
  @param lu: the logical unit on whose behalf we execute
5567
  @type instance: L{objects.Instance}
5568
  @param instance: the instance whose disks we should remove
5569
  @type target_node: string
5570
  @param target_node: used to override the node on which to remove the disks
5571
  @rtype: boolean
5572
  @return: the success of the removal
5573

5574
  """
5575
  logging.info("Removing block devices for instance %s", instance.name)
5576

    
5577
  all_result = True
5578
  for device in instance.disks:
5579
    if target_node:
5580
      edata = [(target_node, device)]
5581
    else:
5582
      edata = device.ComputeNodeTree(instance.primary_node)
5583
    for node, disk in edata:
5584
      lu.cfg.SetDiskID(disk, node)
5585
      msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
5586
      if msg:
5587
        lu.LogWarning("Could not remove block device %s on node %s,"
5588
                      " continuing anyway: %s", device.iv_name, node, msg)
5589
        all_result = False
5590

    
5591
  if instance.disk_template == constants.DT_FILE:
5592
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
5593
    if target_node:
5594
      tgt = target_node
5595
    else:
5596
      tgt = instance.primary_node
5597
    result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
5598
    if result.fail_msg:
5599
      lu.LogWarning("Could not remove directory '%s' on node %s: %s",
5600
                    file_storage_dir, instance.primary_node, result.fail_msg)
5601
      all_result = False
5602

    
5603
  return all_result
5604

    
5605

    
5606
def _ComputeDiskSize(disk_template, disks):
5607
  """Compute disk size requirements in the volume group
5608

5609
  """
5610
  # Required free disk space as a function of disk and swap space
5611
  req_size_dict = {
5612
    constants.DT_DISKLESS: None,
5613
    constants.DT_PLAIN: sum(d["size"] for d in disks),
5614
    # 128 MB are added for drbd metadata for each disk
5615
    constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
5616
    constants.DT_FILE: None,
5617
  }
5618

    
5619
  if disk_template not in req_size_dict:
5620
    raise errors.ProgrammerError("Disk template '%s' size requirement"
5621
                                 " is unknown" %  disk_template)
5622

    
5623
  return req_size_dict[disk_template]
5624

    
5625

    
5626
def _CheckHVParams(lu, nodenames, hvname, hvparams):
5627
  """Hypervisor parameter validation.
5628

5629
  This function abstract the hypervisor parameter validation to be
5630
  used in both instance create and instance modify.
5631

5632
  @type lu: L{LogicalUnit}
5633
  @param lu: the logical unit for which we check
5634
  @type nodenames: list
5635
  @param nodenames: the list of nodes on which we should check
5636
  @type hvname: string
5637
  @param hvname: the name of the hypervisor we should use
5638
  @type hvparams: dict
5639
  @param hvparams: the parameters which we need to check
5640
  @raise errors.OpPrereqError: if the parameters are not valid
5641

5642
  """
5643
  hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
5644
                                                  hvname,
5645
                                                  hvparams)
5646
  for node in nodenames:
5647
    info = hvinfo[node]
5648
    if info.offline:
5649
      continue
5650
    info.Raise("Hypervisor parameter validation failed on node %s" % node)
5651

    
5652

    
5653
class LUCreateInstance(LogicalUnit):
5654
  """Create an instance.
5655

5656
  """
5657
  HPATH = "instance-add"
5658
  HTYPE = constants.HTYPE_INSTANCE
5659
  _OP_REQP = ["instance_name", "disks", "disk_template",
5660
              "mode", "start",
5661
              "wait_for_sync", "ip_check", "nics",
5662
              "hvparams", "beparams"]
5663
  REQ_BGL = False
5664

    
5665
  def CheckArguments(self):
5666
    """Check arguments.
5667

5668
    """
5669
    # do not require name_check to ease forward/backward compatibility
5670
    # for tools
5671
    if not hasattr(self.op, "name_check"):
5672
      self.op.name_check = True
5673
    if self.op.ip_check and not self.op.name_check:
5674
      # TODO: make the ip check more flexible and not depend on the name check
5675
      raise errors.OpPrereqError("Cannot do ip checks without a name check",
5676
                                 errors.ECODE_INVAL)
5677

    
5678
  def _ExpandNode(self, node):
5679
    """Expands and checks one node name.
5680

5681
    """
5682
    node_full = self.cfg.ExpandNodeName(node)
5683
    if node_full is None:
5684
      raise errors.OpPrereqError("Unknown node %s" % node, errors.ECODE_NOENT)
5685
    return node_full
5686

    
5687
  def ExpandNames(self):
5688
    """ExpandNames for CreateInstance.
5689

5690
    Figure out the right locks for instance creation.
5691

5692
    """
5693
    self.needed_locks = {}
5694

    
5695
    # set optional parameters to none if they don't exist
5696
    for attr in ["pnode", "snode", "iallocator", "hypervisor"]:
5697
      if not hasattr(self.op, attr):
5698
        setattr(self.op, attr, None)
5699

    
5700
    # cheap checks, mostly valid constants given
5701

    
5702
    # verify creation mode
5703
    if self.op.mode not in (constants.INSTANCE_CREATE,
5704
                            constants.INSTANCE_IMPORT):
5705
      raise errors.OpPrereqError("Invalid instance creation mode '%s'" %
5706
                                 self.op.mode, errors.ECODE_INVAL)
5707

    
5708
    # disk template and mirror node verification
5709
    if self.op.disk_template not in constants.DISK_TEMPLATES:
5710
      raise errors.OpPrereqError("Invalid disk template name",
5711
                                 errors.ECODE_INVAL)
5712

    
5713
    if self.op.hypervisor is None:
5714
      self.op.hypervisor = self.cfg.GetHypervisorType()
5715

    
5716
    cluster = self.cfg.GetClusterInfo()
5717
    enabled_hvs = cluster.enabled_hypervisors
5718
    if self.op.hypervisor not in enabled_hvs:
5719
      raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
5720
                                 " cluster (%s)" % (self.op.hypervisor,
5721
                                  ",".join(enabled_hvs)),
5722
                                 errors.ECODE_STATE)
5723

    
5724
    # check hypervisor parameter syntax (locally)
5725
    utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
5726
    filled_hvp = objects.FillDict(cluster.hvparams[self.op.hypervisor],
5727
                                  self.op.hvparams)
5728
    hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
5729
    hv_type.CheckParameterSyntax(filled_hvp)
5730
    self.hv_full = filled_hvp
5731
    # check that we don't specify global parameters on an instance
5732
    _CheckGlobalHvParams(self.op.hvparams)
5733

    
5734
    # fill and remember the beparams dict
5735
    utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
5736
    self.be_full = objects.FillDict(cluster.beparams[constants.PP_DEFAULT],
5737
                                    self.op.beparams)
5738

    
5739
    #### instance parameters check
5740

    
5741
    # instance name verification
5742
    if self.op.name_check:
5743
      hostname1 = utils.GetHostInfo(self.op.instance_name)
5744
      self.op.instance_name = instance_name = hostname1.name
5745
      # used in CheckPrereq for ip ping check
5746
      self.check_ip = hostname1.ip
5747
    else:
5748
      instance_name = self.op.instance_name
5749
      self.check_ip = None
5750

    
5751
    # this is just a preventive check, but someone might still add this
5752
    # instance in the meantime, and creation will fail at lock-add time
5753
    if instance_name in self.cfg.GetInstanceList():
5754
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
5755
                                 instance_name, errors.ECODE_EXISTS)
5756

    
5757
    self.add_locks[locking.LEVEL_INSTANCE] = instance_name
5758

    
5759
    # NIC buildup
5760
    self.nics = []
5761
    for idx, nic in enumerate(self.op.nics):
5762
      nic_mode_req = nic.get("mode", None)
5763
      nic_mode = nic_mode_req
5764
      if nic_mode is None:
5765
        nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
5766

    
5767
      # in routed mode, for the first nic, the default ip is 'auto'
5768
      if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
5769
        default_ip_mode = constants.VALUE_AUTO
5770
      else:
5771
        default_ip_mode = constants.VALUE_NONE
5772

    
5773
      # ip validity checks
5774
      ip = nic.get("ip", default_ip_mode)
5775
      if ip is None or ip.lower() == constants.VALUE_NONE:
5776
        nic_ip = None
5777
      elif ip.lower() == constants.VALUE_AUTO:
5778
        if not self.op.name_check:
5779
          raise errors.OpPrereqError("IP address set to auto but name checks"
5780
                                     " have been skipped. Aborting.",
5781
                                     errors.ECODE_INVAL)
5782
        nic_ip = hostname1.ip
5783
      else:
5784
        if not utils.IsValidIP(ip):
5785
          raise errors.OpPrereqError("Given IP address '%s' doesn't look"
5786
                                     " like a valid IP" % ip,
5787
                                     errors.ECODE_INVAL)
5788
        nic_ip = ip
5789

    
5790
      # TODO: check the ip address for uniqueness
5791
      if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
5792
        raise errors.OpPrereqError("Routed nic mode requires an ip address",
5793
                                   errors.ECODE_INVAL)
5794

    
5795
      # MAC address verification
5796
      mac = nic.get("mac", constants.VALUE_AUTO)
5797
      if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
5798
        mac = utils.NormalizeAndValidateMac(mac)
5799

    
5800
        try:
5801
          self.cfg.ReserveMAC(mac, self.proc.GetECId())
5802
        except errors.ReservationError:
5803
          raise errors.OpPrereqError("MAC address %s already in use"
5804
                                     " in cluster" % mac,
5805
                                     errors.ECODE_NOTUNIQUE)
5806

    
5807
      # bridge verification
5808
      bridge = nic.get("bridge", None)
5809
      link = nic.get("link", None)
5810
      if bridge and link:
5811
        raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
5812
                                   " at the same time", errors.ECODE_INVAL)
5813
      elif bridge and nic_mode == constants.NIC_MODE_ROUTED:
5814
        raise errors.OpPrereqError("Cannot pass 'bridge' on a routed nic",
5815
                                   errors.ECODE_INVAL)
5816
      elif bridge:
5817
        link = bridge
5818

    
5819
      nicparams = {}
5820
      if nic_mode_req:
5821
        nicparams[constants.NIC_MODE] = nic_mode_req
5822
      if link:
5823
        nicparams[constants.NIC_LINK] = link
5824

    
5825
      check_params = objects.FillDict(cluster.nicparams[constants.PP_DEFAULT],
5826
                                      nicparams)
5827
      objects.NIC.CheckParameterSyntax(check_params)
5828
      self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
5829

    
5830
    # disk checks/pre-build
5831
    self.disks = []
5832
    for disk in self.op.disks:
5833
      mode = disk.get("mode", constants.DISK_RDWR)
5834
      if mode not in constants.DISK_ACCESS_SET:
5835
        raise errors.OpPrereqError("Invalid disk access mode '%s'" %
5836
                                   mode, errors.ECODE_INVAL)
5837
      size = disk.get("size", None)
5838
      if size is None:
5839
        raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
5840
      try:
5841
        size = int(size)
5842
      except (TypeError, ValueError):
5843
        raise errors.OpPrereqError("Invalid disk size '%s'" % size,
5844
                                   errors.ECODE_INVAL)
5845
      self.disks.append({"size": size, "mode": mode})
5846

    
5847
    # file storage checks
5848
    if (self.op.file_driver and
5849
        not self.op.file_driver in constants.FILE_DRIVER):
5850
      raise errors.OpPrereqError("Invalid file driver name '%s'" %
5851
                                 self.op.file_driver, errors.ECODE_INVAL)
5852

    
5853
    if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
5854
      raise errors.OpPrereqError("File storage directory path not absolute",
5855
                                 errors.ECODE_INVAL)
5856

    
5857
    ### Node/iallocator related checks
5858
    if [self.op.iallocator, self.op.pnode].count(None) != 1:
5859
      raise errors.OpPrereqError("One and only one of iallocator and primary"
5860
                                 " node must be given",
5861
                                 errors.ECODE_INVAL)
5862

    
5863
    if self.op.iallocator:
5864
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5865
    else:
5866
      self.op.pnode = self._ExpandNode(self.op.pnode)
5867
      nodelist = [self.op.pnode]
5868
      if self.op.snode is not None:
5869
        self.op.snode = self._ExpandNode(self.op.snode)
5870
        nodelist.append(self.op.snode)
5871
      self.needed_locks[locking.LEVEL_NODE] = nodelist
5872

    
5873
    # in case of import lock the source node too
5874
    if self.op.mode == constants.INSTANCE_IMPORT:
5875
      src_node = getattr(self.op, "src_node", None)
5876
      src_path = getattr(self.op, "src_path", None)
5877

    
5878
      if src_path is None:
5879
        self.op.src_path = src_path = self.op.instance_name
5880

    
5881
      if src_node is None:
5882
        self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5883
        self.op.src_node = None
5884
        if os.path.isabs(src_path):
5885
          raise errors.OpPrereqError("Importing an instance from an absolute"
5886
                                     " path requires a source node option.",
5887
                                     errors.ECODE_INVAL)
5888
      else:
5889
        self.op.src_node = src_node = self._ExpandNode(src_node)
5890
        if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
5891
          self.needed_locks[locking.LEVEL_NODE].append(src_node)
5892
        if not os.path.isabs(src_path):
5893
          self.op.src_path = src_path = \
5894
            os.path.join(constants.EXPORT_DIR, src_path)
5895

    
5896
      # On import force_variant must be True, because if we forced it at
5897
      # initial install, our only chance when importing it back is that it
5898
      # works again!
5899
      self.op.force_variant = True
5900

    
5901
    else: # INSTANCE_CREATE
5902
      if getattr(self.op, "os_type", None) is None:
5903
        raise errors.OpPrereqError("No guest OS specified",
5904
                                   errors.ECODE_INVAL)
5905
      self.op.force_variant = getattr(self.op, "force_variant", False)
5906

    
5907
  def _RunAllocator(self):
5908
    """Run the allocator based on input opcode.
5909

5910
    """
5911
    nics = [n.ToDict() for n in self.nics]
5912
    ial = IAllocator(self.cfg, self.rpc,
5913
                     mode=constants.IALLOCATOR_MODE_ALLOC,
5914
                     name=self.op.instance_name,
5915
                     disk_template=self.op.disk_template,
5916
                     tags=[],
5917
                     os=self.op.os_type,
5918
                     vcpus=self.be_full[constants.BE_VCPUS],
5919
                     mem_size=self.be_full[constants.BE_MEMORY],
5920
                     disks=self.disks,
5921
                     nics=nics,
5922
                     hypervisor=self.op.hypervisor,
5923
                     )
5924

    
5925
    ial.Run(self.op.iallocator)
5926

    
5927
    if not ial.success:
5928
      raise errors.OpPrereqError("Can't compute nodes using"
5929
                                 " iallocator '%s': %s" %
5930
                                 (self.op.iallocator, ial.info),
5931
                                 errors.ECODE_NORES)
5932
    if len(ial.nodes) != ial.required_nodes:
5933
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
5934
                                 " of nodes (%s), required %s" %
5935
                                 (self.op.iallocator, len(ial.nodes),
5936
                                  ial.required_nodes), errors.ECODE_FAULT)
5937
    self.op.pnode = ial.nodes[0]
5938
    self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
5939
                 self.op.instance_name, self.op.iallocator,
5940
                 utils.CommaJoin(ial.nodes))
5941
    if ial.required_nodes == 2:
5942
      self.op.snode = ial.nodes[1]
5943

    
5944
  def BuildHooksEnv(self):
5945
    """Build hooks env.
5946

5947
    This runs on master, primary and secondary nodes of the instance.
5948

5949
    """
5950
    env = {
5951
      "ADD_MODE": self.op.mode,
5952
      }
5953
    if self.op.mode == constants.INSTANCE_IMPORT:
5954
      env["SRC_NODE"] = self.op.src_node
5955
      env["SRC_PATH"] = self.op.src_path
5956
      env["SRC_IMAGES"] = self.src_images
5957

    
5958
    env.update(_BuildInstanceHookEnv(
5959
      name=self.op.instance_name,
5960
      primary_node=self.op.pnode,
5961
      secondary_nodes=self.secondaries,
5962
      status=self.op.start,
5963
      os_type=self.op.os_type,
5964
      memory=self.be_full[constants.BE_MEMORY],
5965
      vcpus=self.be_full[constants.BE_VCPUS],
5966
      nics=_NICListToTuple(self, self.nics),
5967
      disk_template=self.op.disk_template,
5968
      disks=[(d["size"], d["mode"]) for d in self.disks],
5969
      bep=self.be_full,
5970
      hvp=self.hv_full,
5971
      hypervisor_name=self.op.hypervisor,
5972
    ))
5973

    
5974
    nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
5975
          self.secondaries)
5976
    return env, nl, nl
5977

    
5978

    
5979
  def CheckPrereq(self):
5980
    """Check prerequisites.
5981

5982
    """
5983
    if (not self.cfg.GetVGName() and
5984
        self.op.disk_template not in constants.DTS_NOT_LVM):
5985
      raise errors.OpPrereqError("Cluster does not support lvm-based"
5986
                                 " instances", errors.ECODE_STATE)
5987

    
5988
    if self.op.mode == constants.INSTANCE_IMPORT:
5989
      src_node = self.op.src_node
5990
      src_path = self.op.src_path
5991

    
5992
      if src_node is None:
5993
        locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
5994
        exp_list = self.rpc.call_export_list(locked_nodes)
5995
        found = False
5996
        for node in exp_list:
5997
          if exp_list[node].fail_msg:
5998
            continue
5999
          if src_path in exp_list[node].payload:
6000
            found = True
6001
            self.op.src_node = src_node = node
6002
            self.op.src_path = src_path = os.path.join(constants.EXPORT_DIR,
6003
                                                       src_path)
6004
            break
6005
        if not found:
6006
          raise errors.OpPrereqError("No export found for relative path %s" %
6007
                                      src_path, errors.ECODE_INVAL)
6008

    
6009
      _CheckNodeOnline(self, src_node)
6010
      result = self.rpc.call_export_info(src_node, src_path)
6011
      result.Raise("No export or invalid export found in dir %s" % src_path)
6012

    
6013
      export_info = objects.SerializableConfigParser.Loads(str(result.payload))
6014
      if not export_info.has_section(constants.INISECT_EXP):
6015
        raise errors.ProgrammerError("Corrupted export config",
6016
                                     errors.ECODE_ENVIRON)
6017

    
6018
      ei_version = export_info.get(constants.INISECT_EXP, 'version')
6019
      if (int(ei_version) != constants.EXPORT_VERSION):
6020
        raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
6021
                                   (ei_version, constants.EXPORT_VERSION),
6022
                                   errors.ECODE_ENVIRON)
6023

    
6024
      # Check that the new instance doesn't have less disks than the export
6025
      instance_disks = len(self.disks)
6026
      export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
6027
      if instance_disks < export_disks:
6028
        raise errors.OpPrereqError("Not enough disks to import."
6029
                                   " (instance: %d, export: %d)" %
6030
                                   (instance_disks, export_disks),
6031
                                   errors.ECODE_INVAL)
6032

    
6033
      self.op.os_type = export_info.get(constants.INISECT_EXP, 'os')
6034
      disk_images = []
6035
      for idx in range(export_disks):
6036
        option = 'disk%d_dump' % idx
6037
        if export_info.has_option(constants.INISECT_INS, option):
6038
          # FIXME: are the old os-es, disk sizes, etc. useful?
6039
          export_name = export_info.get(constants.INISECT_INS, option)
6040
          image = os.path.join(src_path, export_name)
6041
          disk_images.append(image)
6042
        else:
6043
          disk_images.append(False)
6044

    
6045
      self.src_images = disk_images
6046

    
6047
      old_name = export_info.get(constants.INISECT_INS, 'name')
6048
      # FIXME: int() here could throw a ValueError on broken exports
6049
      exp_nic_count = int(export_info.get(constants.INISECT_INS, 'nic_count'))
6050
      if self.op.instance_name == old_name:
6051
        for idx, nic in enumerate(self.nics):
6052
          if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
6053
            nic_mac_ini = 'nic%d_mac' % idx
6054
            nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
6055

    
6056
    # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
6057

    
6058
    # ip ping checks (we use the same ip that was resolved in ExpandNames)
6059
    if self.op.ip_check:
6060
      if utils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
6061
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
6062
                                   (self.check_ip, self.op.instance_name),
6063
                                   errors.ECODE_NOTUNIQUE)
6064

    
6065
    #### mac address generation
6066
    # By generating here the mac address both the allocator and the hooks get
6067
    # the real final mac address rather than the 'auto' or 'generate' value.
6068
    # There is a race condition between the generation and the instance object
6069
    # creation, which means that we know the mac is valid now, but we're not
6070
    # sure it will be when we actually add the instance. If things go bad
6071
    # adding the instance will abort because of a duplicate mac, and the
6072
    # creation job will fail.
6073
    for nic in self.nics:
6074
      if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
6075
        nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
6076

    
6077
    #### allocator run
6078

    
6079
    if self.op.iallocator is not None:
6080
      self._RunAllocator()
6081

    
6082
    #### node related checks
6083

    
6084
    # check primary node
6085
    self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
6086
    assert self.pnode is not None, \
6087
      "Cannot retrieve locked node %s" % self.op.pnode
6088
    if pnode.offline:
6089
      raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
6090
                                 pnode.name, errors.ECODE_STATE)
6091
    if pnode.drained:
6092
      raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
6093
                                 pnode.name, errors.ECODE_STATE)
6094

    
6095
    self.secondaries = []
6096

    
6097
    # mirror node verification
6098
    if self.op.disk_template in constants.DTS_NET_MIRROR:
6099
      if self.op.snode is None:
6100
        raise errors.OpPrereqError("The networked disk templates need"
6101
                                   " a mirror node", errors.ECODE_INVAL)
6102
      if self.op.snode == pnode.name:
6103
        raise errors.OpPrereqError("The secondary node cannot be the"
6104
                                   " primary node.", errors.ECODE_INVAL)
6105
      _CheckNodeOnline(self, self.op.snode)
6106
      _CheckNodeNotDrained(self, self.op.snode)
6107
      self.secondaries.append(self.op.snode)
6108

    
6109
    nodenames = [pnode.name] + self.secondaries
6110

    
6111
    req_size = _ComputeDiskSize(self.op.disk_template,
6112
                                self.disks)
6113

    
6114
    # Check lv size requirements
6115
    if req_size is not None:
6116
      nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
6117
                                         self.op.hypervisor)
6118
      for node in nodenames:
6119
        info = nodeinfo[node]
6120
        info.Raise("Cannot get current information from node %s" % node)
6121
        info = info.payload
6122
        vg_free = info.get('vg_free', None)
6123
        if not isinstance(vg_free, int):
6124
          raise errors.OpPrereqError("Can't compute free disk space on"
6125
                                     " node %s" % node, errors.ECODE_ENVIRON)
6126
        if req_size > vg_free:
6127
          raise errors.OpPrereqError("Not enough disk space on target node %s."
6128
                                     " %d MB available, %d MB required" %
6129
                                     (node, vg_free, req_size),
6130
                                     errors.ECODE_NORES)
6131

    
6132
    _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
6133

    
6134
    # os verification
6135
    result = self.rpc.call_os_get(pnode.name, self.op.os_type)
6136
    result.Raise("OS '%s' not in supported os list for primary node %s" %
6137
                 (self.op.os_type, pnode.name),
6138
                 prereq=True, ecode=errors.ECODE_INVAL)
6139
    if not self.op.force_variant:
6140
      _CheckOSVariant(result.payload, self.op.os_type)
6141

    
6142
    _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
6143

    
6144
    # memory check on primary node
6145
    if self.op.start:
6146
      _CheckNodeFreeMemory(self, self.pnode.name,
6147
                           "creating instance %s" % self.op.instance_name,
6148
                           self.be_full[constants.BE_MEMORY],
6149
                           self.op.hypervisor)
6150

    
6151
    self.dry_run_result = list(nodenames)
6152

    
6153
  def Exec(self, feedback_fn):
6154
    """Create and add the instance to the cluster.
6155

6156
    """
6157
    instance = self.op.instance_name
6158
    pnode_name = self.pnode.name
6159

    
6160
    ht_kind = self.op.hypervisor
6161
    if ht_kind in constants.HTS_REQ_PORT:
6162
      network_port = self.cfg.AllocatePort()
6163
    else:
6164
      network_port = None
6165

    
6166
    ##if self.op.vnc_bind_address is None:
6167
    ##  self.op.vnc_bind_address = constants.VNC_DEFAULT_BIND_ADDRESS
6168

    
6169
    # this is needed because os.path.join does not accept None arguments
6170
    if self.op.file_storage_dir is None:
6171
      string_file_storage_dir = ""
6172
    else:
6173
      string_file_storage_dir = self.op.file_storage_dir
6174

    
6175
    # build the full file storage dir path
6176
    file_storage_dir = os.path.normpath(os.path.join(
6177
                                        self.cfg.GetFileStorageDir(),
6178
                                        string_file_storage_dir, instance))
6179

    
6180

    
6181
    disks = _GenerateDiskTemplate(self,
6182
                                  self.op.disk_template,
6183
                                  instance, pnode_name,
6184
                                  self.secondaries,
6185
                                  self.disks,
6186
                                  file_storage_dir,
6187
                                  self.op.file_driver,
6188
                                  0)
6189

    
6190
    iobj = objects.Instance(name=instance, os=self.op.os_type,
6191
                            primary_node=pnode_name,
6192
                            nics=self.nics, disks=disks,
6193
                            disk_template=self.op.disk_template,
6194
                            admin_up=False,
6195
                            network_port=network_port,
6196
                            beparams=self.op.beparams,
6197
                            hvparams=self.op.hvparams,
6198
                            hypervisor=self.op.hypervisor,
6199
                            )
6200

    
6201
    feedback_fn("* creating instance disks...")
6202
    try:
6203
      _CreateDisks(self, iobj)
6204
    except errors.OpExecError:
6205
      self.LogWarning("Device creation failed, reverting...")
6206
      try:
6207
        _RemoveDisks(self, iobj)
6208
      finally:
6209
        self.cfg.ReleaseDRBDMinors(instance)
6210
        raise
6211

    
6212
    feedback_fn("adding instance %s to cluster config" % instance)
6213

    
6214
    self.cfg.AddInstance(iobj, self.proc.GetECId())
6215

    
6216
    # Declare that we don't want to remove the instance lock anymore, as we've
6217
    # added the instance to the config
6218
    del self.remove_locks[locking.LEVEL_INSTANCE]
6219
    # Unlock all the nodes
6220
    if self.op.mode == constants.INSTANCE_IMPORT:
6221
      nodes_keep = [self.op.src_node]
6222
      nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
6223
                       if node != self.op.src_node]
6224
      self.context.glm.release(locking.LEVEL_NODE, nodes_release)
6225
      self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
6226
    else:
6227
      self.context.glm.release(locking.LEVEL_NODE)
6228
      del self.acquired_locks[locking.LEVEL_NODE]
6229

    
6230
    if self.op.wait_for_sync:
6231
      disk_abort = not _WaitForSync(self, iobj)
6232
    elif iobj.disk_template in constants.DTS_NET_MIRROR:
6233
      # make sure the disks are not degraded (still sync-ing is ok)
6234
      time.sleep(15)
6235
      feedback_fn("* checking mirrors status")
6236
      disk_abort = not _WaitForSync(self, iobj, oneshot=True)
6237
    else:
6238
      disk_abort = False
6239

    
6240
    if disk_abort:
6241
      _RemoveDisks(self, iobj)
6242
      self.cfg.RemoveInstance(iobj.name)
6243
      # Make sure the instance lock gets removed
6244
      self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
6245
      raise errors.OpExecError("There are some degraded disks for"
6246
                               " this instance")
6247

    
6248
    feedback_fn("creating os for instance %s on node %s" %
6249
                (instance, pnode_name))
6250

    
6251
    if iobj.disk_template != constants.DT_DISKLESS:
6252
      if self.op.mode == constants.INSTANCE_CREATE:
6253
        feedback_fn("* running the instance OS create scripts...")
6254
        # FIXME: pass debug option from opcode to backend
6255
        result = self.rpc.call_instance_os_add(pnode_name, iobj, False,
6256
                                               self.op.debug_level)
6257
        result.Raise("Could not add os for instance %s"
6258
                     " on node %s" % (instance, pnode_name))
6259

    
6260
      elif self.op.mode == constants.INSTANCE_IMPORT:
6261
        feedback_fn("* running the instance OS import scripts...")
6262
        src_node = self.op.src_node
6263
        src_images = self.src_images
6264
        cluster_name = self.cfg.GetClusterName()
6265
        # FIXME: pass debug option from opcode to backend
6266
        import_result = self.rpc.call_instance_os_import(pnode_name, iobj,
6267
                                                         src_node, src_images,
6268
                                                         cluster_name,
6269
                                                         self.op.debug_level)
6270
        msg = import_result.fail_msg
6271
        if msg:
6272
          self.LogWarning("Error while importing the disk images for instance"
6273
                          " %s on node %s: %s" % (instance, pnode_name, msg))
6274
      else:
6275
        # also checked in the prereq part
6276
        raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
6277
                                     % self.op.mode)
6278

    
6279
    if self.op.start:
6280
      iobj.admin_up = True
6281
      self.cfg.Update(iobj, feedback_fn)
6282
      logging.info("Starting instance %s on node %s", instance, pnode_name)
6283
      feedback_fn("* starting instance...")
6284
      result = self.rpc.call_instance_start(pnode_name, iobj, None, None)
6285
      result.Raise("Could not start instance")
6286

    
6287
    return list(iobj.all_nodes)
6288

    
6289

    
6290
class LUConnectConsole(NoHooksLU):
6291
  """Connect to an instance's console.
6292

6293
  This is somewhat special in that it returns the command line that
6294
  you need to run on the master node in order to connect to the
6295
  console.
6296

6297
  """
6298
  _OP_REQP = ["instance_name"]
6299
  REQ_BGL = False
6300

    
6301
  def ExpandNames(self):
6302
    self._ExpandAndLockInstance()
6303

    
6304
  def CheckPrereq(self):
6305
    """Check prerequisites.
6306

6307
    This checks that the instance is in the cluster.
6308

6309
    """
6310
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6311
    assert self.instance is not None, \
6312
      "Cannot retrieve locked instance %s" % self.op.instance_name
6313
    _CheckNodeOnline(self, self.instance.primary_node)
6314

    
6315
  def Exec(self, feedback_fn):
6316
    """Connect to the console of an instance
6317

6318
    """
6319
    instance = self.instance
6320
    node = instance.primary_node
6321

    
6322
    node_insts = self.rpc.call_instance_list([node],
6323
                                             [instance.hypervisor])[node]
6324
    node_insts.Raise("Can't get node information from %s" % node)
6325

    
6326
    if instance.name not in node_insts.payload:
6327
      raise errors.OpExecError("Instance %s is not running." % instance.name)
6328

    
6329
    logging.debug("Connecting to console of %s on %s", instance.name, node)
6330

    
6331
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
6332
    cluster = self.cfg.GetClusterInfo()
6333
    # beparams and hvparams are passed separately, to avoid editing the
6334
    # instance and then saving the defaults in the instance itself.
6335
    hvparams = cluster.FillHV(instance)
6336
    beparams = cluster.FillBE(instance)
6337
    console_cmd = hyper.GetShellCommandForConsole(instance, hvparams, beparams)
6338

    
6339
    # build ssh cmdline
6340
    return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
6341

    
6342

    
6343
class LUReplaceDisks(LogicalUnit):
6344
  """Replace the disks of an instance.
6345

6346
  """
6347
  HPATH = "mirrors-replace"
6348
  HTYPE = constants.HTYPE_INSTANCE
6349
  _OP_REQP = ["instance_name", "mode", "disks"]
6350
  REQ_BGL = False
6351

    
6352
  def CheckArguments(self):
6353
    if not hasattr(self.op, "remote_node"):
6354
      self.op.remote_node = None
6355
    if not hasattr(self.op, "iallocator"):
6356
      self.op.iallocator = None
6357
    if not hasattr(self.op, "early_release"):
6358
      self.op.early_release = False
6359

    
6360
    TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
6361
                                  self.op.iallocator)
6362

    
6363
  def ExpandNames(self):
6364
    self._ExpandAndLockInstance()
6365

    
6366
    if self.op.iallocator is not None:
6367
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6368

    
6369
    elif self.op.remote_node is not None:
6370
      remote_node = self.cfg.ExpandNodeName(self.op.remote_node)
6371
      if remote_node is None:
6372
        raise errors.OpPrereqError("Node '%s' not known" %
6373
                                   self.op.remote_node, errors.ECODE_NOENT)
6374

    
6375
      self.op.remote_node = remote_node
6376

    
6377
      # Warning: do not remove the locking of the new secondary here
6378
      # unless DRBD8.AddChildren is changed to work in parallel;
6379
      # currently it doesn't since parallel invocations of
6380
      # FindUnusedMinor will conflict
6381
      self.needed_locks[locking.LEVEL_NODE] = [remote_node]
6382
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
6383

    
6384
    else:
6385
      self.needed_locks[locking.LEVEL_NODE] = []
6386
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
6387

    
6388
    self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
6389
                                   self.op.iallocator, self.op.remote_node,
6390
                                   self.op.disks, False, self.op.early_release)
6391

    
6392
    self.tasklets = [self.replacer]
6393

    
6394
  def DeclareLocks(self, level):
6395
    # If we're not already locking all nodes in the set we have to declare the
6396
    # instance's primary/secondary nodes.
6397
    if (level == locking.LEVEL_NODE and
6398
        self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
6399
      self._LockInstancesNodes()
6400

    
6401
  def BuildHooksEnv(self):
6402
    """Build hooks env.
6403

6404
    This runs on the master, the primary and all the secondaries.
6405

6406
    """
6407
    instance = self.replacer.instance
6408
    env = {
6409
      "MODE": self.op.mode,
6410
      "NEW_SECONDARY": self.op.remote_node,
6411
      "OLD_SECONDARY": instance.secondary_nodes[0],
6412
      }
6413
    env.update(_BuildInstanceHookEnvByObject(self, instance))
6414
    nl = [
6415
      self.cfg.GetMasterNode(),
6416
      instance.primary_node,
6417
      ]
6418
    if self.op.remote_node is not None:
6419
      nl.append(self.op.remote_node)
6420
    return env, nl, nl
6421

    
6422

    
6423
class LUEvacuateNode(LogicalUnit):
6424
  """Relocate the secondary instances from a node.
6425

6426
  """
6427
  HPATH = "node-evacuate"
6428
  HTYPE = constants.HTYPE_NODE
6429
  _OP_REQP = ["node_name"]
6430
  REQ_BGL = False
6431

    
6432
  def CheckArguments(self):
6433
    if not hasattr(self.op, "remote_node"):
6434
      self.op.remote_node = None
6435
    if not hasattr(self.op, "iallocator"):
6436
      self.op.iallocator = None
6437
    if not hasattr(self.op, "early_release"):
6438
      self.op.early_release = False
6439

    
6440
    TLReplaceDisks.CheckArguments(constants.REPLACE_DISK_CHG,
6441
                                  self.op.remote_node,
6442
                                  self.op.iallocator)
6443

    
6444
  def ExpandNames(self):
6445
    self.op.node_name = self.cfg.ExpandNodeName(self.op.node_name)
6446
    if self.op.node_name is None:
6447
      raise errors.OpPrereqError("Node '%s' not known" % self.op.node_name,
6448
                                 errors.ECODE_NOENT)
6449

    
6450
    self.needed_locks = {}
6451

    
6452
    # Declare node locks
6453
    if self.op.iallocator is not None:
6454
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6455

    
6456
    elif self.op.remote_node is not None:
6457
      remote_node = self.cfg.ExpandNodeName(self.op.remote_node)
6458
      if remote_node is None:
6459
        raise errors.OpPrereqError("Node '%s' not known" %
6460
                                   self.op.remote_node, errors.ECODE_NOENT)
6461

    
6462
      self.op.remote_node = remote_node
6463

    
6464
      # Warning: do not remove the locking of the new secondary here
6465
      # unless DRBD8.AddChildren is changed to work in parallel;
6466
      # currently it doesn't since parallel invocations of
6467
      # FindUnusedMinor will conflict
6468
      self.needed_locks[locking.LEVEL_NODE] = [remote_node]
6469
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
6470

    
6471
    else:
6472
      raise errors.OpPrereqError("Invalid parameters", errors.ECODE_INVAL)
6473

    
6474
    # Create tasklets for replacing disks for all secondary instances on this
6475
    # node
6476
    names = []
6477
    tasklets = []
6478

    
6479
    for inst in _GetNodeSecondaryInstances(self.cfg, self.op.node_name):
6480
      logging.debug("Replacing disks for instance %s", inst.name)
6481
      names.append(inst.name)
6482

    
6483
      replacer = TLReplaceDisks(self, inst.name, constants.REPLACE_DISK_CHG,
6484
                                self.op.iallocator, self.op.remote_node, [],
6485
                                True, self.op.early_release)
6486
      tasklets.append(replacer)
6487

    
6488
    self.tasklets = tasklets
6489
    self.instance_names = names
6490

    
6491
    # Declare instance locks
6492
    self.needed_locks[locking.LEVEL_INSTANCE] = self.instance_names
6493

    
6494
  def DeclareLocks(self, level):
6495
    # If we're not already locking all nodes in the set we have to declare the
6496
    # instance's primary/secondary nodes.
6497
    if (level == locking.LEVEL_NODE and
6498
        self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
6499
      self._LockInstancesNodes()
6500

    
6501
  def BuildHooksEnv(self):
6502
    """Build hooks env.
6503

6504
    This runs on the master, the primary and all the secondaries.
6505

6506
    """
6507
    env = {
6508
      "NODE_NAME": self.op.node_name,
6509
      }
6510

    
6511
    nl = [self.cfg.GetMasterNode()]
6512

    
6513
    if self.op.remote_node is not None:
6514
      env["NEW_SECONDARY"] = self.op.remote_node
6515
      nl.append(self.op.remote_node)
6516

    
6517
    return (env, nl, nl)
6518

    
6519

    
6520
class TLReplaceDisks(Tasklet):
6521
  """Replaces disks for an instance.
6522

6523
  Note: Locking is not within the scope of this class.
6524

6525
  """
6526
  def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
6527
               disks, delay_iallocator, early_release):
6528
    """Initializes this class.
6529

6530
    """
6531
    Tasklet.__init__(self, lu)
6532

    
6533
    # Parameters
6534
    self.instance_name = instance_name
6535
    self.mode = mode
6536
    self.iallocator_name = iallocator_name
6537
    self.remote_node = remote_node
6538
    self.disks = disks
6539
    self.delay_iallocator = delay_iallocator
6540
    self.early_release = early_release
6541

    
6542
    # Runtime data
6543
    self.instance = None
6544
    self.new_node = None
6545
    self.target_node = None
6546
    self.other_node = None
6547
    self.remote_node_info = None
6548
    self.node_secondary_ip = None
6549

    
6550
  @staticmethod
6551
  def CheckArguments(mode, remote_node, iallocator):
6552
    """Helper function for users of this class.
6553

6554
    """
6555
    # check for valid parameter combination
6556
    if mode == constants.REPLACE_DISK_CHG:
6557
      if remote_node is None and iallocator is None:
6558
        raise errors.OpPrereqError("When changing the secondary either an"
6559
                                   " iallocator script must be used or the"
6560
                                   " new node given", errors.ECODE_INVAL)
6561

    
6562
      if remote_node is not None and iallocator is not None:
6563
        raise errors.OpPrereqError("Give either the iallocator or the new"
6564
                                   " secondary, not both", errors.ECODE_INVAL)
6565

    
6566
    elif remote_node is not None or iallocator is not None:
6567
      # Not replacing the secondary
6568
      raise errors.OpPrereqError("The iallocator and new node options can"
6569
                                 " only be used when changing the"
6570
                                 " secondary node", errors.ECODE_INVAL)
6571

    
6572
  @staticmethod
6573
  def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
6574
    """Compute a new secondary node using an IAllocator.
6575

6576
    """
6577
    ial = IAllocator(lu.cfg, lu.rpc,
6578
                     mode=constants.IALLOCATOR_MODE_RELOC,
6579
                     name=instance_name,
6580
                     relocate_from=relocate_from)
6581

    
6582
    ial.Run(iallocator_name)
6583

    
6584
    if not ial.success:
6585
      raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
6586
                                 " %s" % (iallocator_name, ial.info),
6587
                                 errors.ECODE_NORES)
6588

    
6589
    if len(ial.nodes) != ial.required_nodes:
6590
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
6591
                                 " of nodes (%s), required %s" %
6592
                                 (iallocator_name,
6593
                                  len(ial.nodes), ial.required_nodes),
6594
                                 errors.ECODE_FAULT)
6595

    
6596
    remote_node_name = ial.nodes[0]
6597

    
6598
    lu.LogInfo("Selected new secondary for instance '%s': %s",
6599
               instance_name, remote_node_name)
6600

    
6601
    return remote_node_name
6602

    
6603
  def _FindFaultyDisks(self, node_name):
6604
    return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
6605
                                    node_name, True)
6606

    
6607
  def CheckPrereq(self):
6608
    """Check prerequisites.
6609

6610
    This checks that the instance is in the cluster.
6611

6612
    """
6613
    self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
6614
    assert instance is not None, \
6615
      "Cannot retrieve locked instance %s" % self.instance_name
6616

    
6617
    if instance.disk_template != constants.DT_DRBD8:
6618
      raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
6619
                                 " instances", errors.ECODE_INVAL)
6620

    
6621
    if len(instance.secondary_nodes) != 1:
6622
      raise errors.OpPrereqError("The instance has a strange layout,"
6623
                                 " expected one secondary but found %d" %
6624
                                 len(instance.secondary_nodes),
6625
                                 errors.ECODE_FAULT)
6626

    
6627
    if not self.delay_iallocator:
6628
      self._CheckPrereq2()
6629

    
6630
  def _CheckPrereq2(self):
6631
    """Check prerequisites, second part.
6632

6633
    This function should always be part of CheckPrereq. It was separated and is
6634
    now called from Exec because during node evacuation iallocator was only
6635
    called with an unmodified cluster model, not taking planned changes into
6636
    account.
6637

6638
    """
6639
    instance = self.instance
6640
    secondary_node = instance.secondary_nodes[0]
6641

    
6642
    if self.iallocator_name is None:
6643
      remote_node = self.remote_node
6644
    else:
6645
      remote_node = self._RunAllocator(self.lu, self.iallocator_name,
6646
                                       instance.name, instance.secondary_nodes)
6647

    
6648
    if remote_node is not None:
6649
      self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
6650
      assert self.remote_node_info is not None, \
6651
        "Cannot retrieve locked node %s" % remote_node
6652
    else:
6653
      self.remote_node_info = None
6654

    
6655
    if remote_node == self.instance.primary_node:
6656
      raise errors.OpPrereqError("The specified node is the primary node of"
6657
                                 " the instance.", errors.ECODE_INVAL)
6658

    
6659
    if remote_node == secondary_node:
6660
      raise errors.OpPrereqError("The specified node is already the"
6661
                                 " secondary node of the instance.",
6662
                                 errors.ECODE_INVAL)
6663

    
6664
    if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
6665
                                    constants.REPLACE_DISK_CHG):
6666
      raise errors.OpPrereqError("Cannot specify disks to be replaced",
6667
                                 errors.ECODE_INVAL)
6668

    
6669
    if self.mode == constants.REPLACE_DISK_AUTO:
6670
      faulty_primary = self._FindFaultyDisks(instance.primary_node)
6671
      faulty_secondary = self._FindFaultyDisks(secondary_node)
6672

    
6673
      if faulty_primary and faulty_secondary:
6674
        raise errors.OpPrereqError("Instance %s has faulty disks on more than"
6675
                                   " one node and can not be repaired"
6676
                                   " automatically" % self.instance_name,
6677
                                   errors.ECODE_STATE)
6678

    
6679
      if faulty_primary:
6680
        self.disks = faulty_primary
6681
        self.target_node = instance.primary_node
6682
        self.other_node = secondary_node
6683
        check_nodes = [self.target_node, self.other_node]
6684
      elif faulty_secondary:
6685
        self.disks = faulty_secondary
6686
        self.target_node = secondary_node
6687
        self.other_node = instance.primary_node
6688
        check_nodes = [self.target_node, self.other_node]
6689
      else:
6690
        self.disks = []
6691
        check_nodes = []
6692

    
6693
    else:
6694
      # Non-automatic modes
6695
      if self.mode == constants.REPLACE_DISK_PRI:
6696
        self.target_node = instance.primary_node
6697
        self.other_node = secondary_node
6698
        check_nodes = [self.target_node, self.other_node]
6699

    
6700
      elif self.mode == constants.REPLACE_DISK_SEC:
6701
        self.target_node = secondary_node
6702
        self.other_node = instance.primary_node
6703
        check_nodes = [self.target_node, self.other_node]
6704

    
6705
      elif self.mode == constants.REPLACE_DISK_CHG:
6706
        self.new_node = remote_node
6707
        self.other_node = instance.primary_node
6708
        self.target_node = secondary_node
6709
        check_nodes = [self.new_node, self.other_node]
6710

    
6711
        _CheckNodeNotDrained(self.lu, remote_node)
6712

    
6713
      else:
6714
        raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
6715
                                     self.mode)
6716

    
6717
      # If not specified all disks should be replaced
6718
      if not self.disks:
6719
        self.disks = range(len(self.instance.disks))
6720

    
6721
    for node in check_nodes:
6722
      _CheckNodeOnline(self.lu, node)
6723

    
6724
    # Check whether disks are valid
6725
    for disk_idx in self.disks:
6726
      instance.FindDisk(disk_idx)
6727

    
6728
    # Get secondary node IP addresses
6729
    node_2nd_ip = {}
6730

    
6731
    for node_name in [self.target_node, self.other_node, self.new_node]:
6732
      if node_name is not None:
6733
        node_2nd_ip[node_name] = self.cfg.GetNodeInfo(node_name).secondary_ip
6734

    
6735
    self.node_secondary_ip = node_2nd_ip
6736

    
6737
  def Exec(self, feedback_fn):
6738
    """Execute disk replacement.
6739

6740
    This dispatches the disk replacement to the appropriate handler.
6741

6742
    """
6743
    if self.delay_iallocator:
6744
      self._CheckPrereq2()
6745

    
6746
    if not self.disks:
6747
      feedback_fn("No disks need replacement")
6748
      return
6749

    
6750
    feedback_fn("Replacing disk(s) %s for %s" %
6751
                (utils.CommaJoin(self.disks), self.instance.name))
6752

    
6753
    activate_disks = (not self.instance.admin_up)
6754

    
6755
    # Activate the instance disks if we're replacing them on a down instance
6756
    if activate_disks:
6757
      _StartInstanceDisks(self.lu, self.instance, True)
6758

    
6759
    try:
6760
      # Should we replace the secondary node?
6761
      if self.new_node is not None:
6762
        fn = self._ExecDrbd8Secondary
6763
      else:
6764
        fn = self._ExecDrbd8DiskOnly
6765

    
6766
      return fn(feedback_fn)
6767

    
6768
    finally:
6769
      # Deactivate the instance disks if we're replacing them on a
6770
      # down instance
6771
      if activate_disks:
6772
        _SafeShutdownInstanceDisks(self.lu, self.instance)
6773

    
6774
  def _CheckVolumeGroup(self, nodes):
6775
    self.lu.LogInfo("Checking volume groups")
6776

    
6777
    vgname = self.cfg.GetVGName()
6778

    
6779
    # Make sure volume group exists on all involved nodes
6780
    results = self.rpc.call_vg_list(nodes)
6781
    if not results:
6782
      raise errors.OpExecError("Can't list volume groups on the nodes")
6783

    
6784
    for node in nodes:
6785
      res = results[node]
6786
      res.Raise("Error checking node %s" % node)
6787
      if vgname not in res.payload:
6788
        raise errors.OpExecError("Volume group '%s' not found on node %s" %
6789
                                 (vgname, node))
6790

    
6791
  def _CheckDisksExistence(self, nodes):
6792
    # Check disk existence
6793
    for idx, dev in enumerate(self.instance.disks):
6794
      if idx not in self.disks:
6795
        continue
6796

    
6797
      for node in nodes:
6798
        self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
6799
        self.cfg.SetDiskID(dev, node)
6800

    
6801
        result = self.rpc.call_blockdev_find(node, dev)
6802

    
6803
        msg = result.fail_msg
6804
        if msg or not result.payload:
6805
          if not msg:
6806
            msg = "disk not found"
6807
          raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
6808
                                   (idx, node, msg))
6809

    
6810
  def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
6811
    for idx, dev in enumerate(self.instance.disks):
6812
      if idx not in self.disks:
6813
        continue
6814

    
6815
      self.lu.LogInfo("Checking disk/%d consistency on node %s" %
6816
                      (idx, node_name))
6817

    
6818
      if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
6819
                                   ldisk=ldisk):
6820
        raise errors.OpExecError("Node %s has degraded storage, unsafe to"
6821
                                 " replace disks for instance %s" %
6822
                                 (node_name, self.instance.name))
6823

    
6824
  def _CreateNewStorage(self, node_name):
6825
    vgname = self.cfg.GetVGName()
6826
    iv_names = {}
6827

    
6828
    for idx, dev in enumerate(self.instance.disks):
6829
      if idx not in self.disks:
6830
        continue
6831

    
6832
      self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
6833

    
6834
      self.cfg.SetDiskID(dev, node_name)
6835

    
6836
      lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
6837
      names = _GenerateUniqueNames(self.lu, lv_names)
6838

    
6839
      lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
6840
                             logical_id=(vgname, names[0]))
6841
      lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
6842
                             logical_id=(vgname, names[1]))
6843

    
6844
      new_lvs = [lv_data, lv_meta]
6845
      old_lvs = dev.children
6846
      iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
6847

    
6848
      # we pass force_create=True to force the LVM creation
6849
      for new_lv in new_lvs:
6850
        _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
6851
                        _GetInstanceInfoText(self.instance), False)
6852

    
6853
    return iv_names
6854

    
6855
  def _CheckDevices(self, node_name, iv_names):
6856
    for name, (dev, _, _) in iv_names.iteritems():
6857
      self.cfg.SetDiskID(dev, node_name)
6858

    
6859
      result = self.rpc.call_blockdev_find(node_name, dev)
6860

    
6861
      msg = result.fail_msg
6862
      if msg or not result.payload:
6863
        if not msg:
6864
          msg = "disk not found"
6865
        raise errors.OpExecError("Can't find DRBD device %s: %s" %
6866
                                 (name, msg))
6867

    
6868
      if result.payload.is_degraded:
6869
        raise errors.OpExecError("DRBD device %s is degraded!" % name)
6870

    
6871
  def _RemoveOldStorage(self, node_name, iv_names):
6872
    for name, (_, old_lvs, _) in iv_names.iteritems():
6873
      self.lu.LogInfo("Remove logical volumes for %s" % name)
6874

    
6875
      for lv in old_lvs:
6876
        self.cfg.SetDiskID(lv, node_name)
6877

    
6878
        msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
6879
        if msg:
6880
          self.lu.LogWarning("Can't remove old LV: %s" % msg,
6881
                             hint="remove unused LVs manually")
6882

    
6883
  def _ReleaseNodeLock(self, node_name):
6884
    """Releases the lock for a given node."""
6885
    self.lu.context.glm.release(locking.LEVEL_NODE, node_name)
6886

    
6887
  def _ExecDrbd8DiskOnly(self, feedback_fn):
6888
    """Replace a disk on the primary or secondary for DRBD 8.
6889

6890
    The algorithm for replace is quite complicated:
6891

6892
      1. for each disk to be replaced:
6893

6894
        1. create new LVs on the target node with unique names
6895
        1. detach old LVs from the drbd device
6896
        1. rename old LVs to name_replaced.<time_t>
6897
        1. rename new LVs to old LVs
6898
        1. attach the new LVs (with the old names now) to the drbd device
6899

6900
      1. wait for sync across all devices
6901

6902
      1. for each modified disk:
6903

6904
        1. remove old LVs (which have the name name_replaces.<time_t>)
6905

6906
    Failures are not very well handled.
6907

6908
    """
6909
    steps_total = 6
6910

    
6911
    # Step: check device activation
6912
    self.lu.LogStep(1, steps_total, "Check device existence")
6913
    self._CheckDisksExistence([self.other_node, self.target_node])
6914
    self._CheckVolumeGroup([self.target_node, self.other_node])
6915

    
6916
    # Step: check other node consistency
6917
    self.lu.LogStep(2, steps_total, "Check peer consistency")
6918
    self._CheckDisksConsistency(self.other_node,
6919
                                self.other_node == self.instance.primary_node,
6920
                                False)
6921

    
6922
    # Step: create new storage
6923
    self.lu.LogStep(3, steps_total, "Allocate new storage")
6924
    iv_names = self._CreateNewStorage(self.target_node)
6925

    
6926
    # Step: for each lv, detach+rename*2+attach
6927
    self.lu.LogStep(4, steps_total, "Changing drbd configuration")
6928
    for dev, old_lvs, new_lvs in iv_names.itervalues():
6929
      self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
6930

    
6931
      result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
6932
                                                     old_lvs)
6933
      result.Raise("Can't detach drbd from local storage on node"
6934
                   " %s for device %s" % (self.target_node, dev.iv_name))
6935
      #dev.children = []
6936
      #cfg.Update(instance)
6937

    
6938
      # ok, we created the new LVs, so now we know we have the needed
6939
      # storage; as such, we proceed on the target node to rename
6940
      # old_lv to _old, and new_lv to old_lv; note that we rename LVs
6941
      # using the assumption that logical_id == physical_id (which in
6942
      # turn is the unique_id on that node)
6943

    
6944
      # FIXME(iustin): use a better name for the replaced LVs
6945
      temp_suffix = int(time.time())
6946
      ren_fn = lambda d, suff: (d.physical_id[0],
6947
                                d.physical_id[1] + "_replaced-%s" % suff)
6948

    
6949
      # Build the rename list based on what LVs exist on the node
6950
      rename_old_to_new = []
6951
      for to_ren in old_lvs:
6952
        result = self.rpc.call_blockdev_find(self.target_node, to_ren)
6953
        if not result.fail_msg and result.payload:
6954
          # device exists
6955
          rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
6956

    
6957
      self.lu.LogInfo("Renaming the old LVs on the target node")
6958
      result = self.rpc.call_blockdev_rename(self.target_node,
6959
                                             rename_old_to_new)
6960
      result.Raise("Can't rename old LVs on node %s" % self.target_node)
6961

    
6962
      # Now we rename the new LVs to the old LVs
6963
      self.lu.LogInfo("Renaming the new LVs on the target node")
6964
      rename_new_to_old = [(new, old.physical_id)
6965
                           for old, new in zip(old_lvs, new_lvs)]
6966
      result = self.rpc.call_blockdev_rename(self.target_node,
6967
                                             rename_new_to_old)
6968
      result.Raise("Can't rename new LVs on node %s" % self.target_node)
6969

    
6970
      for old, new in zip(old_lvs, new_lvs):
6971
        new.logical_id = old.logical_id
6972
        self.cfg.SetDiskID(new, self.target_node)
6973

    
6974
      for disk in old_lvs:
6975
        disk.logical_id = ren_fn(disk, temp_suffix)
6976
        self.cfg.SetDiskID(disk, self.target_node)
6977

    
6978
      # Now that the new lvs have the old name, we can add them to the device
6979
      self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
6980
      result = self.rpc.call_blockdev_addchildren(self.target_node, dev,
6981
                                                  new_lvs)
6982
      msg = result.fail_msg
6983
      if msg:
6984
        for new_lv in new_lvs:
6985
          msg2 = self.rpc.call_blockdev_remove(self.target_node,
6986
                                               new_lv).fail_msg
6987
          if msg2:
6988
            self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
6989
                               hint=("cleanup manually the unused logical"
6990
                                     "volumes"))
6991
        raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
6992

    
6993
      dev.children = new_lvs
6994

    
6995
      self.cfg.Update(self.instance, feedback_fn)
6996

    
6997
    cstep = 5
6998
    if self.early_release:
6999
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
7000
      cstep += 1
7001
      self._RemoveOldStorage(self.target_node, iv_names)
7002
      # only release the lock if we're doing secondary replace, since
7003
      # we use the primary node later
7004
      if self.target_node != self.instance.primary_node:
7005
        self._ReleaseNodeLock(self.target_node)
7006

    
7007
    # Wait for sync
7008
    # This can fail as the old devices are degraded and _WaitForSync
7009
    # does a combined result over all disks, so we don't check its return value
7010
    self.lu.LogStep(cstep, steps_total, "Sync devices")
7011
    cstep += 1
7012
    _WaitForSync(self.lu, self.instance)
7013

    
7014
    # Check all devices manually
7015
    self._CheckDevices(self.instance.primary_node, iv_names)
7016

    
7017
    # Step: remove old storage
7018
    if not self.early_release:
7019
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
7020
      cstep += 1
7021
      self._RemoveOldStorage(self.target_node, iv_names)
7022

    
7023
  def _ExecDrbd8Secondary(self, feedback_fn):
7024
    """Replace the secondary node for DRBD 8.
7025

7026
    The algorithm for replace is quite complicated:
7027
      - for all disks of the instance:
7028
        - create new LVs on the new node with same names
7029
        - shutdown the drbd device on the old secondary
7030
        - disconnect the drbd network on the primary
7031
        - create the drbd device on the new secondary
7032
        - network attach the drbd on the primary, using an artifice:
7033
          the drbd code for Attach() will connect to the network if it
7034
          finds a device which is connected to the good local disks but
7035
          not network enabled
7036
      - wait for sync across all devices
7037
      - remove all disks from the old secondary
7038

7039
    Failures are not very well handled.
7040

7041
    """
7042
    steps_total = 6
7043

    
7044
    # Step: check device activation
7045
    self.lu.LogStep(1, steps_total, "Check device existence")
7046
    self._CheckDisksExistence([self.instance.primary_node])
7047
    self._CheckVolumeGroup([self.instance.primary_node])
7048

    
7049
    # Step: check other node consistency
7050
    self.lu.LogStep(2, steps_total, "Check peer consistency")
7051
    self._CheckDisksConsistency(self.instance.primary_node, True, True)
7052

    
7053
    # Step: create new storage
7054
    self.lu.LogStep(3, steps_total, "Allocate new storage")
7055
    for idx, dev in enumerate(self.instance.disks):
7056
      self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
7057
                      (self.new_node, idx))
7058
      # we pass force_create=True to force LVM creation
7059
      for new_lv in dev.children:
7060
        _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
7061
                        _GetInstanceInfoText(self.instance), False)
7062

    
7063
    # Step 4: dbrd minors and drbd setups changes
7064
    # after this, we must manually remove the drbd minors on both the
7065
    # error and the success paths
7066
    self.lu.LogStep(4, steps_total, "Changing drbd configuration")
7067
    minors = self.cfg.AllocateDRBDMinor([self.new_node
7068
                                         for dev in self.instance.disks],
7069
                                        self.instance.name)
7070
    logging.debug("Allocated minors %r", minors)
7071

    
7072
    iv_names = {}
7073
    for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
7074
      self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
7075
                      (self.new_node, idx))
7076
      # create new devices on new_node; note that we create two IDs:
7077
      # one without port, so the drbd will be activated without
7078
      # networking information on the new node at this stage, and one
7079
      # with network, for the latter activation in step 4
7080
      (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
7081
      if self.instance.primary_node == o_node1:
7082
        p_minor = o_minor1
7083
      else:
7084
        assert self.instance.primary_node == o_node2, "Three-node instance?"
7085
        p_minor = o_minor2
7086

    
7087
      new_alone_id = (self.instance.primary_node, self.new_node, None,
7088
                      p_minor, new_minor, o_secret)
7089
      new_net_id = (self.instance.primary_node, self.new_node, o_port,
7090
                    p_minor, new_minor, o_secret)
7091

    
7092
      iv_names[idx] = (dev, dev.children, new_net_id)
7093
      logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
7094
                    new_net_id)
7095
      new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
7096
                              logical_id=new_alone_id,
7097
                              children=dev.children,
7098
                              size=dev.size)
7099
      try:
7100
        _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
7101
                              _GetInstanceInfoText(self.instance), False)
7102
      except errors.GenericError:
7103
        self.cfg.ReleaseDRBDMinors(self.instance.name)
7104
        raise
7105

    
7106
    # We have new devices, shutdown the drbd on the old secondary
7107
    for idx, dev in enumerate(self.instance.disks):
7108
      self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
7109
      self.cfg.SetDiskID(dev, self.target_node)
7110
      msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
7111
      if msg:
7112
        self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
7113
                           "node: %s" % (idx, msg),
7114
                           hint=("Please cleanup this device manually as"
7115
                                 " soon as possible"))
7116

    
7117
    self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
7118
    result = self.rpc.call_drbd_disconnect_net([self.instance.primary_node],
7119
                                               self.node_secondary_ip,
7120
                                               self.instance.disks)\
7121
                                              [self.instance.primary_node]
7122

    
7123
    msg = result.fail_msg
7124
    if msg:
7125
      # detaches didn't succeed (unlikely)
7126
      self.cfg.ReleaseDRBDMinors(self.instance.name)
7127
      raise errors.OpExecError("Can't detach the disks from the network on"
7128
                               " old node: %s" % (msg,))
7129

    
7130
    # if we managed to detach at least one, we update all the disks of
7131
    # the instance to point to the new secondary
7132
    self.lu.LogInfo("Updating instance configuration")
7133
    for dev, _, new_logical_id in iv_names.itervalues():
7134
      dev.logical_id = new_logical_id
7135
      self.cfg.SetDiskID(dev, self.instance.primary_node)
7136

    
7137
    self.cfg.Update(self.instance, feedback_fn)
7138

    
7139
    # and now perform the drbd attach
7140
    self.lu.LogInfo("Attaching primary drbds to new secondary"
7141
                    " (standalone => connected)")
7142
    result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
7143
                                            self.new_node],
7144
                                           self.node_secondary_ip,
7145
                                           self.instance.disks,
7146
                                           self.instance.name,
7147
                                           False)
7148
    for to_node, to_result in result.items():
7149
      msg = to_result.fail_msg
7150
      if msg:
7151
        self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
7152
                           to_node, msg,
7153
                           hint=("please do a gnt-instance info to see the"
7154
                                 " status of disks"))
7155
    cstep = 5
7156
    if self.early_release:
7157
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
7158
      cstep += 1
7159
      self._RemoveOldStorage(self.target_node, iv_names)
7160
      self._ReleaseNodeLock([self.target_node, self.new_node])
7161

    
7162
    # Wait for sync
7163
    # This can fail as the old devices are degraded and _WaitForSync
7164
    # does a combined result over all disks, so we don't check its return value
7165
    self.lu.LogStep(cstep, steps_total, "Sync devices")
7166
    cstep += 1
7167
    _WaitForSync(self.lu, self.instance)
7168

    
7169
    # Check all devices manually
7170
    self._CheckDevices(self.instance.primary_node, iv_names)
7171

    
7172
    # Step: remove old storage
7173
    if not self.early_release:
7174
      self.lu.LogStep(cstep, steps_total, "Removing old storage")
7175
      self._RemoveOldStorage(self.target_node, iv_names)
7176

    
7177

    
7178
class LURepairNodeStorage(NoHooksLU):
7179
  """Repairs the volume group on a node.
7180

7181
  """
7182
  _OP_REQP = ["node_name"]
7183
  REQ_BGL = False
7184

    
7185
  def CheckArguments(self):
7186
    node_name = self.cfg.ExpandNodeName(self.op.node_name)
7187
    if node_name is None:
7188
      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name,
7189
                                 errors.ECODE_NOENT)
7190

    
7191
    self.op.node_name = node_name
7192

    
7193
  def ExpandNames(self):
7194
    self.needed_locks = {
7195
      locking.LEVEL_NODE: [self.op.node_name],
7196
      }
7197

    
7198
  def _CheckFaultyDisks(self, instance, node_name):
7199
    """Ensure faulty disks abort the opcode or at least warn."""
7200
    try:
7201
      if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
7202
                                  node_name, True):
7203
        raise errors.OpPrereqError("Instance '%s' has faulty disks on"
7204
                                   " node '%s'" % (instance.name, node_name),
7205
                                   errors.ECODE_STATE)
7206
    except errors.OpPrereqError, err:
7207
      if self.op.ignore_consistency:
7208
        self.proc.LogWarning(str(err.args[0]))
7209
      else:
7210
        raise
7211

    
7212
  def CheckPrereq(self):
7213
    """Check prerequisites.
7214

7215
    """
7216
    storage_type = self.op.storage_type
7217

    
7218
    if (constants.SO_FIX_CONSISTENCY not in
7219
        constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
7220
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
7221
                                 " repaired" % storage_type,
7222
                                 errors.ECODE_INVAL)
7223

    
7224
    # Check whether any instance on this node has faulty disks
7225
    for inst in _GetNodeInstances(self.cfg, self.op.node_name):
7226
      if not inst.admin_up:
7227
        continue
7228
      check_nodes = set(inst.all_nodes)
7229
      check_nodes.discard(self.op.node_name)
7230
      for inst_node_name in check_nodes:
7231
        self._CheckFaultyDisks(inst, inst_node_name)
7232

    
7233
  def Exec(self, feedback_fn):
7234
    feedback_fn("Repairing storage unit '%s' on %s ..." %
7235
                (self.op.name, self.op.node_name))
7236

    
7237
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
7238
    result = self.rpc.call_storage_execute(self.op.node_name,
7239
                                           self.op.storage_type, st_args,
7240
                                           self.op.name,
7241
                                           constants.SO_FIX_CONSISTENCY)
7242
    result.Raise("Failed to repair storage unit '%s' on %s" %
7243
                 (self.op.name, self.op.node_name))
7244

    
7245

    
7246
class LUGrowDisk(LogicalUnit):
7247
  """Grow a disk of an instance.
7248

7249
  """
7250
  HPATH = "disk-grow"
7251
  HTYPE = constants.HTYPE_INSTANCE
7252
  _OP_REQP = ["instance_name", "disk", "amount", "wait_for_sync"]
7253
  REQ_BGL = False
7254

    
7255
  def ExpandNames(self):
7256
    self._ExpandAndLockInstance()
7257
    self.needed_locks[locking.LEVEL_NODE] = []
7258
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7259

    
7260
  def DeclareLocks(self, level):
7261
    if level == locking.LEVEL_NODE:
7262
      self._LockInstancesNodes()
7263

    
7264
  def BuildHooksEnv(self):
7265
    """Build hooks env.
7266

7267
    This runs on the master, the primary and all the secondaries.
7268

7269
    """
7270
    env = {
7271
      "DISK": self.op.disk,
7272
      "AMOUNT": self.op.amount,
7273
      }
7274
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
7275
    nl = [
7276
      self.cfg.GetMasterNode(),
7277
      self.instance.primary_node,
7278
      ]
7279
    return env, nl, nl
7280

    
7281
  def CheckPrereq(self):
7282
    """Check prerequisites.
7283

7284
    This checks that the instance is in the cluster.
7285

7286
    """
7287
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7288
    assert instance is not None, \
7289
      "Cannot retrieve locked instance %s" % self.op.instance_name
7290
    nodenames = list(instance.all_nodes)
7291
    for node in nodenames:
7292
      _CheckNodeOnline(self, node)
7293

    
7294

    
7295
    self.instance = instance
7296

    
7297
    if instance.disk_template not in (constants.DT_PLAIN, constants.DT_DRBD8):
7298
      raise errors.OpPrereqError("Instance's disk layout does not support"
7299
                                 " growing.", errors.ECODE_INVAL)
7300

    
7301
    self.disk = instance.FindDisk(self.op.disk)
7302

    
7303
    nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
7304
                                       instance.hypervisor)
7305
    for node in nodenames:
7306
      info = nodeinfo[node]
7307
      info.Raise("Cannot get current information from node %s" % node)
7308
      vg_free = info.payload.get('vg_free', None)
7309
      if not isinstance(vg_free, int):
7310
        raise errors.OpPrereqError("Can't compute free disk space on"
7311
                                   " node %s" % node, errors.ECODE_ENVIRON)
7312
      if self.op.amount > vg_free:
7313
        raise errors.OpPrereqError("Not enough disk space on target node %s:"
7314
                                   " %d MiB available, %d MiB required" %
7315
                                   (node, vg_free, self.op.amount),
7316
                                   errors.ECODE_NORES)
7317

    
7318
  def Exec(self, feedback_fn):
7319
    """Execute disk grow.
7320

7321
    """
7322
    instance = self.instance
7323
    disk = self.disk
7324
    for node in instance.all_nodes:
7325
      self.cfg.SetDiskID(disk, node)
7326
      result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
7327
      result.Raise("Grow request failed to node %s" % node)
7328

    
7329
      # TODO: Rewrite code to work properly
7330
      # DRBD goes into sync mode for a short amount of time after executing the
7331
      # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
7332
      # calling "resize" in sync mode fails. Sleeping for a short amount of
7333
      # time is a work-around.
7334
      time.sleep(5)
7335

    
7336
    disk.RecordGrow(self.op.amount)
7337
    self.cfg.Update(instance, feedback_fn)
7338
    if self.op.wait_for_sync:
7339
      disk_abort = not _WaitForSync(self, instance)
7340
      if disk_abort:
7341
        self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
7342
                             " status.\nPlease check the instance.")
7343

    
7344

    
7345
class LUQueryInstanceData(NoHooksLU):
7346
  """Query runtime instance data.
7347

7348
  """
7349
  _OP_REQP = ["instances", "static"]
7350
  REQ_BGL = False
7351

    
7352
  def ExpandNames(self):
7353
    self.needed_locks = {}
7354
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
7355

    
7356
    if not isinstance(self.op.instances, list):
7357
      raise errors.OpPrereqError("Invalid argument type 'instances'",
7358
                                 errors.ECODE_INVAL)
7359

    
7360
    if self.op.instances:
7361
      self.wanted_names = []
7362
      for name in self.op.instances:
7363
        full_name = self.cfg.ExpandInstanceName(name)
7364
        if full_name is None:
7365
          raise errors.OpPrereqError("Instance '%s' not known" % name,
7366
                                     errors.ECODE_NOENT)
7367
        self.wanted_names.append(full_name)
7368
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
7369
    else:
7370
      self.wanted_names = None
7371
      self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
7372

    
7373
    self.needed_locks[locking.LEVEL_NODE] = []
7374
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7375

    
7376
  def DeclareLocks(self, level):
7377
    if level == locking.LEVEL_NODE:
7378
      self._LockInstancesNodes()
7379

    
7380
  def CheckPrereq(self):
7381
    """Check prerequisites.
7382

7383
    This only checks the optional instance list against the existing names.
7384

7385
    """
7386
    if self.wanted_names is None:
7387
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
7388

    
7389
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
7390
                             in self.wanted_names]
7391
    return
7392

    
7393
  def _ComputeBlockdevStatus(self, node, instance_name, dev):
7394
    """Returns the status of a block device
7395

7396
    """
7397
    if self.op.static or not node:
7398
      return None
7399

    
7400
    self.cfg.SetDiskID(dev, node)
7401

    
7402
    result = self.rpc.call_blockdev_find(node, dev)
7403
    if result.offline:
7404
      return None
7405

    
7406
    result.Raise("Can't compute disk status for %s" % instance_name)
7407

    
7408
    status = result.payload
7409
    if status is None:
7410
      return None
7411

    
7412
    return (status.dev_path, status.major, status.minor,
7413
            status.sync_percent, status.estimated_time,
7414
            status.is_degraded, status.ldisk_status)
7415

    
7416
  def _ComputeDiskStatus(self, instance, snode, dev):
7417
    """Compute block device status.
7418

7419
    """
7420
    if dev.dev_type in constants.LDS_DRBD:
7421
      # we change the snode then (otherwise we use the one passed in)
7422
      if dev.logical_id[0] == instance.primary_node:
7423
        snode = dev.logical_id[1]
7424
      else:
7425
        snode = dev.logical_id[0]
7426

    
7427
    dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
7428
                                              instance.name, dev)
7429
    dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
7430

    
7431
    if dev.children:
7432
      dev_children = [self._ComputeDiskStatus(instance, snode, child)
7433
                      for child in dev.children]
7434
    else:
7435
      dev_children = []
7436

    
7437
    data = {
7438
      "iv_name": dev.iv_name,
7439
      "dev_type": dev.dev_type,
7440
      "logical_id": dev.logical_id,
7441
      "physical_id": dev.physical_id,
7442
      "pstatus": dev_pstatus,
7443
      "sstatus": dev_sstatus,
7444
      "children": dev_children,
7445
      "mode": dev.mode,
7446
      "size": dev.size,
7447
      }
7448

    
7449
    return data
7450

    
7451
  def Exec(self, feedback_fn):
7452
    """Gather and return data"""
7453
    result = {}
7454

    
7455
    cluster = self.cfg.GetClusterInfo()
7456

    
7457
    for instance in self.wanted_instances:
7458
      if not self.op.static:
7459
        remote_info = self.rpc.call_instance_info(instance.primary_node,
7460
                                                  instance.name,
7461
                                                  instance.hypervisor)
7462
        remote_info.Raise("Error checking node %s" % instance.primary_node)
7463
        remote_info = remote_info.payload
7464
        if remote_info and "state" in remote_info:
7465
          remote_state = "up"
7466
        else:
7467
          remote_state = "down"
7468
      else:
7469
        remote_state = None
7470
      if instance.admin_up:
7471
        config_state = "up"
7472
      else:
7473
        config_state = "down"
7474

    
7475
      disks = [self._ComputeDiskStatus(instance, None, device)
7476
               for device in instance.disks]
7477

    
7478
      idict = {
7479
        "name": instance.name,
7480
        "config_state": config_state,
7481
        "run_state": remote_state,
7482
        "pnode": instance.primary_node,
7483
        "snodes": instance.secondary_nodes,
7484
        "os": instance.os,
7485
        # this happens to be the same format used for hooks
7486
        "nics": _NICListToTuple(self, instance.nics),
7487
        "disks": disks,
7488
        "hypervisor": instance.hypervisor,
7489
        "network_port": instance.network_port,
7490
        "hv_instance": instance.hvparams,
7491
        "hv_actual": cluster.FillHV(instance, skip_globals=True),
7492
        "be_instance": instance.beparams,
7493
        "be_actual": cluster.FillBE(instance),
7494
        "serial_no": instance.serial_no,
7495
        "mtime": instance.mtime,
7496
        "ctime": instance.ctime,
7497
        "uuid": instance.uuid,
7498
        }
7499

    
7500
      result[instance.name] = idict
7501

    
7502
    return result
7503

    
7504

    
7505
class LUSetInstanceParams(LogicalUnit):
7506
  """Modifies an instances's parameters.
7507

7508
  """
7509
  HPATH = "instance-modify"
7510
  HTYPE = constants.HTYPE_INSTANCE
7511
  _OP_REQP = ["instance_name"]
7512
  REQ_BGL = False
7513

    
7514
  def CheckArguments(self):
7515
    if not hasattr(self.op, 'nics'):
7516
      self.op.nics = []
7517
    if not hasattr(self.op, 'disks'):
7518
      self.op.disks = []
7519
    if not hasattr(self.op, 'beparams'):
7520
      self.op.beparams = {}
7521
    if not hasattr(self.op, 'hvparams'):
7522
      self.op.hvparams = {}
7523
    self.op.force = getattr(self.op, "force", False)
7524
    if not (self.op.nics or self.op.disks or
7525
            self.op.hvparams or self.op.beparams):
7526
      raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
7527

    
7528
    if self.op.hvparams:
7529
      _CheckGlobalHvParams(self.op.hvparams)
7530

    
7531
    # Disk validation
7532
    disk_addremove = 0
7533
    for disk_op, disk_dict in self.op.disks:
7534
      if disk_op == constants.DDM_REMOVE:
7535
        disk_addremove += 1
7536
        continue
7537
      elif disk_op == constants.DDM_ADD:
7538
        disk_addremove += 1
7539
      else:
7540
        if not isinstance(disk_op, int):
7541
          raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL)
7542
        if not isinstance(disk_dict, dict):
7543
          msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
7544
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
7545

    
7546
      if disk_op == constants.DDM_ADD:
7547
        mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
7548
        if mode not in constants.DISK_ACCESS_SET:
7549
          raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
7550
                                     errors.ECODE_INVAL)
7551
        size = disk_dict.get('size', None)
7552
        if size is None:
7553
          raise errors.OpPrereqError("Required disk parameter size missing",
7554
                                     errors.ECODE_INVAL)
7555
        try:
7556
          size = int(size)
7557
        except (TypeError, ValueError), err:
7558
          raise errors.OpPrereqError("Invalid disk size parameter: %s" %
7559
                                     str(err), errors.ECODE_INVAL)
7560
        disk_dict['size'] = size
7561
      else:
7562
        # modification of disk
7563
        if 'size' in disk_dict:
7564
          raise errors.OpPrereqError("Disk size change not possible, use"
7565
                                     " grow-disk", errors.ECODE_INVAL)
7566

    
7567
    if disk_addremove > 1:
7568
      raise errors.OpPrereqError("Only one disk add or remove operation"
7569
                                 " supported at a time", errors.ECODE_INVAL)
7570

    
7571
    # NIC validation
7572
    nic_addremove = 0
7573
    for nic_op, nic_dict in self.op.nics:
7574
      if nic_op == constants.DDM_REMOVE:
7575
        nic_addremove += 1
7576
        continue
7577
      elif nic_op == constants.DDM_ADD:
7578
        nic_addremove += 1
7579
      else:
7580
        if not isinstance(nic_op, int):
7581
          raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL)
7582
        if not isinstance(nic_dict, dict):
7583
          msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
7584
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
7585

    
7586
      # nic_dict should be a dict
7587
      nic_ip = nic_dict.get('ip', None)
7588
      if nic_ip is not None:
7589
        if nic_ip.lower() == constants.VALUE_NONE:
7590
          nic_dict['ip'] = None
7591
        else:
7592
          if not utils.IsValidIP(nic_ip):
7593
            raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip,
7594
                                       errors.ECODE_INVAL)
7595

    
7596
      nic_bridge = nic_dict.get('bridge', None)
7597
      nic_link = nic_dict.get('link', None)
7598
      if nic_bridge and nic_link:
7599
        raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
7600
                                   " at the same time", errors.ECODE_INVAL)
7601
      elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
7602
        nic_dict['bridge'] = None
7603
      elif nic_link and nic_link.lower() == constants.VALUE_NONE:
7604
        nic_dict['link'] = None
7605

    
7606
      if nic_op == constants.DDM_ADD:
7607
        nic_mac = nic_dict.get('mac', None)
7608
        if nic_mac is None:
7609
          nic_dict['mac'] = constants.VALUE_AUTO
7610

    
7611
      if 'mac' in nic_dict:
7612
        nic_mac = nic_dict['mac']
7613
        if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7614
          nic_mac = utils.NormalizeAndValidateMac(nic_mac)
7615

    
7616
        if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
7617
          raise errors.OpPrereqError("'auto' is not a valid MAC address when"
7618
                                     " modifying an existing nic",
7619
                                     errors.ECODE_INVAL)
7620

    
7621
    if nic_addremove > 1:
7622
      raise errors.OpPrereqError("Only one NIC add or remove operation"
7623
                                 " supported at a time", errors.ECODE_INVAL)
7624

    
7625
  def ExpandNames(self):
7626
    self._ExpandAndLockInstance()
7627
    self.needed_locks[locking.LEVEL_NODE] = []
7628
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7629

    
7630
  def DeclareLocks(self, level):
7631
    if level == locking.LEVEL_NODE:
7632
      self._LockInstancesNodes()
7633

    
7634
  def BuildHooksEnv(self):
7635
    """Build hooks env.
7636

7637
    This runs on the master, primary and secondaries.
7638

7639
    """
7640
    args = dict()
7641
    if constants.BE_MEMORY in self.be_new:
7642
      args['memory'] = self.be_new[constants.BE_MEMORY]
7643
    if constants.BE_VCPUS in self.be_new:
7644
      args['vcpus'] = self.be_new[constants.BE_VCPUS]
7645
    # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
7646
    # information at all.
7647
    if self.op.nics:
7648
      args['nics'] = []
7649
      nic_override = dict(self.op.nics)
7650
      c_nicparams = self.cluster.nicparams[constants.PP_DEFAULT]
7651
      for idx, nic in enumerate(self.instance.nics):
7652
        if idx in nic_override:
7653
          this_nic_override = nic_override[idx]
7654
        else:
7655
          this_nic_override = {}
7656
        if 'ip' in this_nic_override:
7657
          ip = this_nic_override['ip']
7658
        else:
7659
          ip = nic.ip
7660
        if 'mac' in this_nic_override:
7661
          mac = this_nic_override['mac']
7662
        else:
7663
          mac = nic.mac
7664
        if idx in self.nic_pnew:
7665
          nicparams = self.nic_pnew[idx]
7666
        else:
7667
          nicparams = objects.FillDict(c_nicparams, nic.nicparams)
7668
        mode = nicparams[constants.NIC_MODE]
7669
        link = nicparams[constants.NIC_LINK]
7670
        args['nics'].append((ip, mac, mode, link))
7671
      if constants.DDM_ADD in nic_override:
7672
        ip = nic_override[constants.DDM_ADD].get('ip', None)
7673
        mac = nic_override[constants.DDM_ADD]['mac']
7674
        nicparams = self.nic_pnew[constants.DDM_ADD]
7675
        mode = nicparams[constants.NIC_MODE]
7676
        link = nicparams[constants.NIC_LINK]
7677
        args['nics'].append((ip, mac, mode, link))
7678
      elif constants.DDM_REMOVE in nic_override:
7679
        del args['nics'][-1]
7680

    
7681
    env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
7682
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
7683
    return env, nl, nl
7684

    
7685
  @staticmethod
7686
  def _GetUpdatedParams(old_params, update_dict,
7687
                        default_values, parameter_types):
7688
    """Return the new params dict for the given params.
7689

7690
    @type old_params: dict
7691
    @param old_params: old parameters
7692
    @type update_dict: dict
7693
    @param update_dict: dict containing new parameter values,
7694
                        or constants.VALUE_DEFAULT to reset the
7695
                        parameter to its default value
7696
    @type default_values: dict
7697
    @param default_values: default values for the filled parameters
7698
    @type parameter_types: dict
7699
    @param parameter_types: dict mapping target dict keys to types
7700
                            in constants.ENFORCEABLE_TYPES
7701
    @rtype: (dict, dict)
7702
    @return: (new_parameters, filled_parameters)
7703

7704
    """
7705
    params_copy = copy.deepcopy(old_params)
7706
    for key, val in update_dict.iteritems():
7707
      if val == constants.VALUE_DEFAULT:
7708
        try:
7709
          del params_copy[key]
7710
        except KeyError:
7711
          pass
7712
      else:
7713
        params_copy[key] = val
7714
    utils.ForceDictType(params_copy, parameter_types)
7715
    params_filled = objects.FillDict(default_values, params_copy)
7716
    return (params_copy, params_filled)
7717

    
7718
  def CheckPrereq(self):
7719
    """Check prerequisites.
7720

7721
    This only checks the instance list against the existing names.
7722

7723
    """
7724
    self.force = self.op.force
7725

    
7726
    # checking the new params on the primary/secondary nodes
7727

    
7728
    instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7729
    cluster = self.cluster = self.cfg.GetClusterInfo()
7730
    assert self.instance is not None, \
7731
      "Cannot retrieve locked instance %s" % self.op.instance_name
7732
    pnode = instance.primary_node
7733
    nodelist = list(instance.all_nodes)
7734

    
7735
    # hvparams processing
7736
    if self.op.hvparams:
7737
      i_hvdict, hv_new = self._GetUpdatedParams(
7738
                             instance.hvparams, self.op.hvparams,
7739
                             cluster.hvparams[instance.hypervisor],
7740
                             constants.HVS_PARAMETER_TYPES)
7741
      # local check
7742
      hypervisor.GetHypervisor(
7743
        instance.hypervisor).CheckParameterSyntax(hv_new)
7744
      _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
7745
      self.hv_new = hv_new # the new actual values
7746
      self.hv_inst = i_hvdict # the new dict (without defaults)
7747
    else:
7748
      self.hv_new = self.hv_inst = {}
7749

    
7750
    # beparams processing
7751
    if self.op.beparams:
7752
      i_bedict, be_new = self._GetUpdatedParams(
7753
                             instance.beparams, self.op.beparams,
7754
                             cluster.beparams[constants.PP_DEFAULT],
7755
                             constants.BES_PARAMETER_TYPES)
7756
      self.be_new = be_new # the new actual values
7757
      self.be_inst = i_bedict # the new dict (without defaults)
7758
    else:
7759
      self.be_new = self.be_inst = {}
7760

    
7761
    self.warn = []
7762

    
7763
    if constants.BE_MEMORY in self.op.beparams and not self.force:
7764
      mem_check_list = [pnode]
7765
      if be_new[constants.BE_AUTO_BALANCE]:
7766
        # either we changed auto_balance to yes or it was from before
7767
        mem_check_list.extend(instance.secondary_nodes)
7768
      instance_info = self.rpc.call_instance_info(pnode, instance.name,
7769
                                                  instance.hypervisor)
7770
      nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
7771
                                         instance.hypervisor)
7772
      pninfo = nodeinfo[pnode]
7773
      msg = pninfo.fail_msg
7774
      if msg:
7775
        # Assume the primary node is unreachable and go ahead
7776
        self.warn.append("Can't get info from primary node %s: %s" %
7777
                         (pnode,  msg))
7778
      elif not isinstance(pninfo.payload.get('memory_free', None), int):
7779
        self.warn.append("Node data from primary node %s doesn't contain"
7780
                         " free memory information" % pnode)
7781
      elif instance_info.fail_msg:
7782
        self.warn.append("Can't get instance runtime information: %s" %
7783
                        instance_info.fail_msg)
7784
      else:
7785
        if instance_info.payload:
7786
          current_mem = int(instance_info.payload['memory'])
7787
        else:
7788
          # Assume instance not running
7789
          # (there is a slight race condition here, but it's not very probable,
7790
          # and we have no other way to check)
7791
          current_mem = 0
7792
        miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
7793
                    pninfo.payload['memory_free'])
7794
        if miss_mem > 0:
7795
          raise errors.OpPrereqError("This change will prevent the instance"
7796
                                     " from starting, due to %d MB of memory"
7797
                                     " missing on its primary node" % miss_mem,
7798
                                     errors.ECODE_NORES)
7799

    
7800
      if be_new[constants.BE_AUTO_BALANCE]:
7801
        for node, nres in nodeinfo.items():
7802
          if node not in instance.secondary_nodes:
7803
            continue
7804
          msg = nres.fail_msg
7805
          if msg:
7806
            self.warn.append("Can't get info from secondary node %s: %s" %
7807
                             (node, msg))
7808
          elif not isinstance(nres.payload.get('memory_free', None), int):
7809
            self.warn.append("Secondary node %s didn't return free"
7810
                             " memory information" % node)
7811
          elif be_new[constants.BE_MEMORY] > nres.payload['memory_free']:
7812
            self.warn.append("Not enough memory to failover instance to"
7813
                             " secondary node %s" % node)
7814

    
7815
    # NIC processing
7816
    self.nic_pnew = {}
7817
    self.nic_pinst = {}
7818
    for nic_op, nic_dict in self.op.nics:
7819
      if nic_op == constants.DDM_REMOVE:
7820
        if not instance.nics:
7821
          raise errors.OpPrereqError("Instance has no NICs, cannot remove",
7822
                                     errors.ECODE_INVAL)
7823
        continue
7824
      if nic_op != constants.DDM_ADD:
7825
        # an existing nic
7826
        if not instance.nics:
7827
          raise errors.OpPrereqError("Invalid NIC index %s, instance has"
7828
                                     " no NICs" % nic_op,
7829
                                     errors.ECODE_INVAL)
7830
        if nic_op < 0 or nic_op >= len(instance.nics):
7831
          raise errors.OpPrereqError("Invalid NIC index %s, valid values"
7832
                                     " are 0 to %d" %
7833
                                     (nic_op, len(instance.nics) - 1),
7834
                                     errors.ECODE_INVAL)
7835
        old_nic_params = instance.nics[nic_op].nicparams
7836
        old_nic_ip = instance.nics[nic_op].ip
7837
      else:
7838
        old_nic_params = {}
7839
        old_nic_ip = None
7840

    
7841
      update_params_dict = dict([(key, nic_dict[key])
7842
                                 for key in constants.NICS_PARAMETERS
7843
                                 if key in nic_dict])
7844

    
7845
      if 'bridge' in nic_dict:
7846
        update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
7847

    
7848
      new_nic_params, new_filled_nic_params = \
7849
          self._GetUpdatedParams(old_nic_params, update_params_dict,
7850
                                 cluster.nicparams[constants.PP_DEFAULT],
7851
                                 constants.NICS_PARAMETER_TYPES)
7852
      objects.NIC.CheckParameterSyntax(new_filled_nic_params)
7853
      self.nic_pinst[nic_op] = new_nic_params
7854
      self.nic_pnew[nic_op] = new_filled_nic_params
7855
      new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
7856

    
7857
      if new_nic_mode == constants.NIC_MODE_BRIDGED:
7858
        nic_bridge = new_filled_nic_params[constants.NIC_LINK]
7859
        msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
7860
        if msg:
7861
          msg = "Error checking bridges on node %s: %s" % (pnode, msg)
7862
          if self.force:
7863
            self.warn.append(msg)
7864
          else:
7865
            raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
7866
      if new_nic_mode == constants.NIC_MODE_ROUTED:
7867
        if 'ip' in nic_dict:
7868
          nic_ip = nic_dict['ip']
7869
        else:
7870
          nic_ip = old_nic_ip
7871
        if nic_ip is None:
7872
          raise errors.OpPrereqError('Cannot set the nic ip to None'
7873
                                     ' on a routed nic', errors.ECODE_INVAL)
7874
      if 'mac' in nic_dict:
7875
        nic_mac = nic_dict['mac']
7876
        if nic_mac is None:
7877
          raise errors.OpPrereqError('Cannot set the nic mac to None',
7878
                                     errors.ECODE_INVAL)
7879
        elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7880
          # otherwise generate the mac
7881
          nic_dict['mac'] = self.cfg.GenerateMAC(self.proc.GetECId())
7882
        else:
7883
          # or validate/reserve the current one
7884
          try:
7885
            self.cfg.ReserveMAC(nic_mac, self.proc.GetECId())
7886
          except errors.ReservationError:
7887
            raise errors.OpPrereqError("MAC address %s already in use"
7888
                                       " in cluster" % nic_mac,
7889
                                       errors.ECODE_NOTUNIQUE)
7890

    
7891
    # DISK processing
7892
    if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
7893
      raise errors.OpPrereqError("Disk operations not supported for"
7894
                                 " diskless instances",
7895
                                 errors.ECODE_INVAL)
7896
    for disk_op, _ in self.op.disks:
7897
      if disk_op == constants.DDM_REMOVE:
7898
        if len(instance.disks) == 1:
7899
          raise errors.OpPrereqError("Cannot remove the last disk of"
7900
                                     " an instance",
7901
                                     errors.ECODE_INVAL)
7902
        ins_l = self.rpc.call_instance_list([pnode], [instance.hypervisor])
7903
        ins_l = ins_l[pnode]
7904
        msg = ins_l.fail_msg
7905
        if msg:
7906
          raise errors.OpPrereqError("Can't contact node %s: %s" %
7907
                                     (pnode, msg), errors.ECODE_ENVIRON)
7908
        if instance.name in ins_l.payload:
7909
          raise errors.OpPrereqError("Instance is running, can't remove"
7910
                                     " disks.", errors.ECODE_STATE)
7911

    
7912
      if (disk_op == constants.DDM_ADD and
7913
          len(instance.nics) >= constants.MAX_DISKS):
7914
        raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
7915
                                   " add more" % constants.MAX_DISKS,
7916
                                   errors.ECODE_STATE)
7917
      if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
7918
        # an existing disk
7919
        if disk_op < 0 or disk_op >= len(instance.disks):
7920
          raise errors.OpPrereqError("Invalid disk index %s, valid values"
7921
                                     " are 0 to %d" %
7922
                                     (disk_op, len(instance.disks)),
7923
                                     errors.ECODE_INVAL)
7924

    
7925
    return
7926

    
7927
  def Exec(self, feedback_fn):
7928
    """Modifies an instance.
7929

7930
    All parameters take effect only at the next restart of the instance.
7931

7932
    """
7933
    # Process here the warnings from CheckPrereq, as we don't have a
7934
    # feedback_fn there.
7935
    for warn in self.warn:
7936
      feedback_fn("WARNING: %s" % warn)
7937

    
7938
    result = []
7939
    instance = self.instance
7940
    # disk changes
7941
    for disk_op, disk_dict in self.op.disks:
7942
      if disk_op == constants.DDM_REMOVE:
7943
        # remove the last disk
7944
        device = instance.disks.pop()
7945
        device_idx = len(instance.disks)
7946
        for node, disk in device.ComputeNodeTree(instance.primary_node):
7947
          self.cfg.SetDiskID(disk, node)
7948
          msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
7949
          if msg:
7950
            self.LogWarning("Could not remove disk/%d on node %s: %s,"
7951
                            " continuing anyway", device_idx, node, msg)
7952
        result.append(("disk/%d" % device_idx, "remove"))
7953
      elif disk_op == constants.DDM_ADD:
7954
        # add a new disk
7955
        if instance.disk_template == constants.DT_FILE:
7956
          file_driver, file_path = instance.disks[0].logical_id
7957
          file_path = os.path.dirname(file_path)
7958
        else:
7959
          file_driver = file_path = None
7960
        disk_idx_base = len(instance.disks)
7961
        new_disk = _GenerateDiskTemplate(self,
7962
                                         instance.disk_template,
7963
                                         instance.name, instance.primary_node,
7964
                                         instance.secondary_nodes,
7965
                                         [disk_dict],
7966
                                         file_path,
7967
                                         file_driver,
7968
                                         disk_idx_base)[0]
7969
        instance.disks.append(new_disk)
7970
        info = _GetInstanceInfoText(instance)
7971

    
7972
        logging.info("Creating volume %s for instance %s",
7973
                     new_disk.iv_name, instance.name)
7974
        # Note: this needs to be kept in sync with _CreateDisks
7975
        #HARDCODE
7976
        for node in instance.all_nodes:
7977
          f_create = node == instance.primary_node
7978
          try:
7979
            _CreateBlockDev(self, node, instance, new_disk,
7980
                            f_create, info, f_create)
7981
          except errors.OpExecError, err:
7982
            self.LogWarning("Failed to create volume %s (%s) on"
7983
                            " node %s: %s",
7984
                            new_disk.iv_name, new_disk, node, err)
7985
        result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
7986
                       (new_disk.size, new_disk.mode)))
7987
      else:
7988
        # change a given disk
7989
        instance.disks[disk_op].mode = disk_dict['mode']
7990
        result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
7991
    # NIC changes
7992
    for nic_op, nic_dict in self.op.nics:
7993
      if nic_op == constants.DDM_REMOVE:
7994
        # remove the last nic
7995
        del instance.nics[-1]
7996
        result.append(("nic.%d" % len(instance.nics), "remove"))
7997
      elif nic_op == constants.DDM_ADD:
7998
        # mac and bridge should be set, by now
7999
        mac = nic_dict['mac']
8000
        ip = nic_dict.get('ip', None)
8001
        nicparams = self.nic_pinst[constants.DDM_ADD]
8002
        new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
8003
        instance.nics.append(new_nic)
8004
        result.append(("nic.%d" % (len(instance.nics) - 1),
8005
                       "add:mac=%s,ip=%s,mode=%s,link=%s" %
8006
                       (new_nic.mac, new_nic.ip,
8007
                        self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
8008
                        self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
8009
                       )))
8010
      else:
8011
        for key in 'mac', 'ip':
8012
          if key in nic_dict:
8013
            setattr(instance.nics[nic_op], key, nic_dict[key])
8014
        if nic_op in self.nic_pinst:
8015
          instance.nics[nic_op].nicparams = self.nic_pinst[nic_op]
8016
        for key, val in nic_dict.iteritems():
8017
          result.append(("nic.%s/%d" % (key, nic_op), val))
8018

    
8019
    # hvparams changes
8020
    if self.op.hvparams:
8021
      instance.hvparams = self.hv_inst
8022
      for key, val in self.op.hvparams.iteritems():
8023
        result.append(("hv/%s" % key, val))
8024

    
8025
    # beparams changes
8026
    if self.op.beparams:
8027
      instance.beparams = self.be_inst
8028
      for key, val in self.op.beparams.iteritems():
8029
        result.append(("be/%s" % key, val))
8030

    
8031
    self.cfg.Update(instance, feedback_fn)
8032

    
8033
    return result
8034

    
8035

    
8036
class LUQueryExports(NoHooksLU):
8037
  """Query the exports list
8038

8039
  """
8040
  _OP_REQP = ['nodes']
8041
  REQ_BGL = False
8042

    
8043
  def ExpandNames(self):
8044
    self.needed_locks = {}
8045
    self.share_locks[locking.LEVEL_NODE] = 1
8046
    if not self.op.nodes:
8047
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
8048
    else:
8049
      self.needed_locks[locking.LEVEL_NODE] = \
8050
        _GetWantedNodes(self, self.op.nodes)
8051

    
8052
  def CheckPrereq(self):
8053
    """Check prerequisites.
8054

8055
    """
8056
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
8057

    
8058
  def Exec(self, feedback_fn):
8059
    """Compute the list of all the exported system images.
8060

8061
    @rtype: dict
8062
    @return: a dictionary with the structure node->(export-list)
8063
        where export-list is a list of the instances exported on
8064
        that node.
8065

8066
    """
8067
    rpcresult = self.rpc.call_export_list(self.nodes)
8068
    result = {}
8069
    for node in rpcresult:
8070
      if rpcresult[node].fail_msg:
8071
        result[node] = False
8072
      else:
8073
        result[node] = rpcresult[node].payload
8074

    
8075
    return result
8076

    
8077

    
8078
class LUExportInstance(LogicalUnit):
8079
  """Export an instance to an image in the cluster.
8080

8081
  """
8082
  HPATH = "instance-export"
8083
  HTYPE = constants.HTYPE_INSTANCE
8084
  _OP_REQP = ["instance_name", "target_node", "shutdown"]
8085
  REQ_BGL = False
8086

    
8087
  def CheckArguments(self):
8088
    """Check the arguments.
8089

8090
    """
8091
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
8092
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
8093

    
8094
  def ExpandNames(self):
8095
    self._ExpandAndLockInstance()
8096
    # FIXME: lock only instance primary and destination node
8097
    #
8098
    # Sad but true, for now we have do lock all nodes, as we don't know where
8099
    # the previous export might be, and and in this LU we search for it and
8100
    # remove it from its current node. In the future we could fix this by:
8101
    #  - making a tasklet to search (share-lock all), then create the new one,
8102
    #    then one to remove, after
8103
    #  - removing the removal operation altogether
8104
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
8105

    
8106
  def DeclareLocks(self, level):
8107
    """Last minute lock declaration."""
8108
    # All nodes are locked anyway, so nothing to do here.
8109

    
8110
  def BuildHooksEnv(self):
8111
    """Build hooks env.
8112

8113
    This will run on the master, primary node and target node.
8114

8115
    """
8116
    env = {
8117
      "EXPORT_NODE": self.op.target_node,
8118
      "EXPORT_DO_SHUTDOWN": self.op.shutdown,
8119
      "SHUTDOWN_TIMEOUT": self.shutdown_timeout,
8120
      }
8121
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
8122
    nl = [self.cfg.GetMasterNode(), self.instance.primary_node,
8123
          self.op.target_node]
8124
    return env, nl, nl
8125

    
8126
  def CheckPrereq(self):
8127
    """Check prerequisites.
8128

8129
    This checks that the instance and node names are valid.
8130

8131
    """
8132
    instance_name = self.op.instance_name
8133
    self.instance = self.cfg.GetInstanceInfo(instance_name)
8134
    assert self.instance is not None, \
8135
          "Cannot retrieve locked instance %s" % self.op.instance_name
8136
    _CheckNodeOnline(self, self.instance.primary_node)
8137

    
8138
    self.dst_node = self.cfg.GetNodeInfo(
8139
      self.cfg.ExpandNodeName(self.op.target_node))
8140

    
8141
    if self.dst_node is None:
8142
      # This is wrong node name, not a non-locked node
8143
      raise errors.OpPrereqError("Wrong node name %s" % self.op.target_node,
8144
                                 errors.ECODE_NOENT)
8145
    _CheckNodeOnline(self, self.dst_node.name)
8146
    _CheckNodeNotDrained(self, self.dst_node.name)
8147

    
8148
    # instance disk type verification
8149
    for disk in self.instance.disks:
8150
      if disk.dev_type == constants.LD_FILE:
8151
        raise errors.OpPrereqError("Export not supported for instances with"
8152
                                   " file-based disks", errors.ECODE_INVAL)
8153

    
8154
  def Exec(self, feedback_fn):
8155
    """Export an instance to an image in the cluster.
8156

8157
    """
8158
    instance = self.instance
8159
    dst_node = self.dst_node
8160
    src_node = instance.primary_node
8161

    
8162
    if self.op.shutdown:
8163
      # shutdown the instance, but not the disks
8164
      feedback_fn("Shutting down instance %s" % instance.name)
8165
      result = self.rpc.call_instance_shutdown(src_node, instance,
8166
                                               self.shutdown_timeout)
8167
      result.Raise("Could not shutdown instance %s on"
8168
                   " node %s" % (instance.name, src_node))
8169

    
8170
    vgname = self.cfg.GetVGName()
8171

    
8172
    snap_disks = []
8173

    
8174
    # set the disks ID correctly since call_instance_start needs the
8175
    # correct drbd minor to create the symlinks
8176
    for disk in instance.disks:
8177
      self.cfg.SetDiskID(disk, src_node)
8178

    
8179
    activate_disks = (not instance.admin_up)
8180

    
8181
    if activate_disks:
8182
      # Activate the instance disks if we'exporting a stopped instance
8183
      feedback_fn("Activating disks for %s" % instance.name)
8184
      _StartInstanceDisks(self, instance, None)
8185

    
8186
    try:
8187
      # per-disk results
8188
      dresults = []
8189
      try:
8190
        for idx, disk in enumerate(instance.disks):
8191
          feedback_fn("Creating a snapshot of disk/%s on node %s" %
8192
                      (idx, src_node))
8193

    
8194
          # result.payload will be a snapshot of an lvm leaf of the one we
8195
          # passed
8196
          result = self.rpc.call_blockdev_snapshot(src_node, disk)
8197
          msg = result.fail_msg
8198
          if msg:
8199
            self.LogWarning("Could not snapshot disk/%s on node %s: %s",
8200
                            idx, src_node, msg)
8201
            snap_disks.append(False)
8202
          else:
8203
            disk_id = (vgname, result.payload)
8204
            new_dev = objects.Disk(dev_type=constants.LD_LV, size=disk.size,
8205
                                   logical_id=disk_id, physical_id=disk_id,
8206
                                   iv_name=disk.iv_name)
8207
            snap_disks.append(new_dev)
8208

    
8209
      finally:
8210
        if self.op.shutdown and instance.admin_up:
8211
          feedback_fn("Starting instance %s" % instance.name)
8212
          result = self.rpc.call_instance_start(src_node, instance, None, None)
8213
          msg = result.fail_msg
8214
          if msg:
8215
            _ShutdownInstanceDisks(self, instance)
8216
            raise errors.OpExecError("Could not start instance: %s" % msg)
8217

    
8218
      # TODO: check for size
8219

    
8220
      cluster_name = self.cfg.GetClusterName()
8221
      for idx, dev in enumerate(snap_disks):
8222
        feedback_fn("Exporting snapshot %s from %s to %s" %
8223
                    (idx, src_node, dst_node.name))
8224
        if dev:
8225
          # FIXME: pass debug from opcode to backend
8226
          result = self.rpc.call_snapshot_export(src_node, dev, dst_node.name,
8227
                                                 instance, cluster_name,
8228
                                                 idx, self.op.debug_level)
8229
          msg = result.fail_msg
8230
          if msg:
8231
            self.LogWarning("Could not export disk/%s from node %s to"
8232
                            " node %s: %s", idx, src_node, dst_node.name, msg)
8233
            dresults.append(False)
8234
          else:
8235
            dresults.append(True)
8236
          msg = self.rpc.call_blockdev_remove(src_node, dev).fail_msg
8237
          if msg:
8238
            self.LogWarning("Could not remove snapshot for disk/%d from node"
8239
                            " %s: %s", idx, src_node, msg)
8240
        else:
8241
          dresults.append(False)
8242

    
8243
      feedback_fn("Finalizing export on %s" % dst_node.name)
8244
      result = self.rpc.call_finalize_export(dst_node.name, instance,
8245
                                             snap_disks)
8246
      fin_resu = True
8247
      msg = result.fail_msg
8248
      if msg:
8249
        self.LogWarning("Could not finalize export for instance %s"
8250
                        " on node %s: %s", instance.name, dst_node.name, msg)
8251
        fin_resu = False
8252

    
8253
    finally:
8254
      if activate_disks:
8255
        feedback_fn("Deactivating disks for %s" % instance.name)
8256
        _ShutdownInstanceDisks(self, instance)
8257

    
8258
    nodelist = self.cfg.GetNodeList()
8259
    nodelist.remove(dst_node.name)
8260

    
8261
    # on one-node clusters nodelist will be empty after the removal
8262
    # if we proceed the backup would be removed because OpQueryExports
8263
    # substitutes an empty list with the full cluster node list.
8264
    iname = instance.name
8265
    if nodelist:
8266
      feedback_fn("Removing old exports for instance %s" % iname)
8267
      exportlist = self.rpc.call_export_list(nodelist)
8268
      for node in exportlist:
8269
        if exportlist[node].fail_msg:
8270
          continue
8271
        if iname in exportlist[node].payload:
8272
          msg = self.rpc.call_export_remove(node, iname).fail_msg
8273
          if msg:
8274
            self.LogWarning("Could not remove older export for instance %s"
8275
                            " on node %s: %s", iname, node, msg)
8276
    return fin_resu, dresults
8277

    
8278

    
8279
class LURemoveExport(NoHooksLU):
8280
  """Remove exports related to the named instance.
8281

8282
  """
8283
  _OP_REQP = ["instance_name"]
8284
  REQ_BGL = False
8285

    
8286
  def ExpandNames(self):
8287
    self.needed_locks = {}
8288
    # We need all nodes to be locked in order for RemoveExport to work, but we
8289
    # don't need to lock the instance itself, as nothing will happen to it (and
8290
    # we can remove exports also for a removed instance)
8291
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
8292

    
8293
  def CheckPrereq(self):
8294
    """Check prerequisites.
8295
    """
8296
    pass
8297

    
8298
  def Exec(self, feedback_fn):
8299
    """Remove any export.
8300

8301
    """
8302
    instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
8303
    # If the instance was not found we'll try with the name that was passed in.
8304
    # This will only work if it was an FQDN, though.
8305
    fqdn_warn = False
8306
    if not instance_name:
8307
      fqdn_warn = True
8308
      instance_name = self.op.instance_name
8309

    
8310
    locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
8311
    exportlist = self.rpc.call_export_list(locked_nodes)
8312
    found = False
8313
    for node in exportlist:
8314
      msg = exportlist[node].fail_msg
8315
      if msg:
8316
        self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
8317
        continue
8318
      if instance_name in exportlist[node].payload:
8319
        found = True
8320
        result = self.rpc.call_export_remove(node, instance_name)
8321
        msg = result.fail_msg
8322
        if msg:
8323
          logging.error("Could not remove export for instance %s"
8324
                        " on node %s: %s", instance_name, node, msg)
8325

    
8326
    if fqdn_warn and not found:
8327
      feedback_fn("Export not found. If trying to remove an export belonging"
8328
                  " to a deleted instance please use its Fully Qualified"
8329
                  " Domain Name.")
8330

    
8331

    
8332
class TagsLU(NoHooksLU): # pylint: disable-msg=W0223
8333
  """Generic tags LU.
8334

8335
  This is an abstract class which is the parent of all the other tags LUs.
8336

8337
  """
8338

    
8339
  def ExpandNames(self):
8340
    self.needed_locks = {}
8341
    if self.op.kind == constants.TAG_NODE:
8342
      name = self.cfg.ExpandNodeName(self.op.name)
8343
      if name is None:
8344
        raise errors.OpPrereqError("Invalid node name (%s)" %
8345
                                   (self.op.name,), errors.ECODE_NOENT)
8346
      self.op.name = name
8347
      self.needed_locks[locking.LEVEL_NODE] = name
8348
    elif self.op.kind == constants.TAG_INSTANCE:
8349
      name = self.cfg.ExpandInstanceName(self.op.name)
8350
      if name is None:
8351
        raise errors.OpPrereqError("Invalid instance name (%s)" %
8352
                                   (self.op.name,), errors.ECODE_NOENT)
8353
      self.op.name = name
8354
      self.needed_locks[locking.LEVEL_INSTANCE] = name
8355

    
8356
  def CheckPrereq(self):
8357
    """Check prerequisites.
8358

8359
    """
8360
    if self.op.kind == constants.TAG_CLUSTER:
8361
      self.target = self.cfg.GetClusterInfo()
8362
    elif self.op.kind == constants.TAG_NODE:
8363
      self.target = self.cfg.GetNodeInfo(self.op.name)
8364
    elif self.op.kind == constants.TAG_INSTANCE:
8365
      self.target = self.cfg.GetInstanceInfo(self.op.name)
8366
    else:
8367
      raise errors.OpPrereqError("Wrong tag type requested (%s)" %
8368
                                 str(self.op.kind), errors.ECODE_INVAL)
8369

    
8370

    
8371
class LUGetTags(TagsLU):
8372
  """Returns the tags of a given object.
8373

8374
  """
8375
  _OP_REQP = ["kind", "name"]
8376
  REQ_BGL = False
8377

    
8378
  def Exec(self, feedback_fn):
8379
    """Returns the tag list.
8380

8381
    """
8382
    return list(self.target.GetTags())
8383

    
8384

    
8385
class LUSearchTags(NoHooksLU):
8386
  """Searches the tags for a given pattern.
8387

8388
  """
8389
  _OP_REQP = ["pattern"]
8390
  REQ_BGL = False
8391

    
8392
  def ExpandNames(self):
8393
    self.needed_locks = {}
8394

    
8395
  def CheckPrereq(self):
8396
    """Check prerequisites.
8397

8398
    This checks the pattern passed for validity by compiling it.
8399

8400
    """
8401
    try:
8402
      self.re = re.compile(self.op.pattern)
8403
    except re.error, err:
8404
      raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
8405
                                 (self.op.pattern, err), errors.ECODE_INVAL)
8406

    
8407
  def Exec(self, feedback_fn):
8408
    """Returns the tag list.
8409

8410
    """
8411
    cfg = self.cfg
8412
    tgts = [("/cluster", cfg.GetClusterInfo())]
8413
    ilist = cfg.GetAllInstancesInfo().values()
8414
    tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
8415
    nlist = cfg.GetAllNodesInfo().values()
8416
    tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
8417
    results = []
8418
    for path, target in tgts:
8419
      for tag in target.GetTags():
8420
        if self.re.search(tag):
8421
          results.append((path, tag))
8422
    return results
8423

    
8424

    
8425
class LUAddTags(TagsLU):
8426
  """Sets a tag on a given object.
8427

8428
  """
8429
  _OP_REQP = ["kind", "name", "tags"]
8430
  REQ_BGL = False
8431

    
8432
  def CheckPrereq(self):
8433
    """Check prerequisites.
8434

8435
    This checks the type and length of the tag name and value.
8436

8437
    """
8438
    TagsLU.CheckPrereq(self)
8439
    for tag in self.op.tags:
8440
      objects.TaggableObject.ValidateTag(tag)
8441

    
8442
  def Exec(self, feedback_fn):
8443
    """Sets the tag.
8444

8445
    """
8446
    try:
8447
      for tag in self.op.tags:
8448
        self.target.AddTag(tag)
8449
    except errors.TagError, err:
8450
      raise errors.OpExecError("Error while setting tag: %s" % str(err))
8451
    self.cfg.Update(self.target, feedback_fn)
8452

    
8453

    
8454
class LUDelTags(TagsLU):
8455
  """Delete a list of tags from a given object.
8456

8457
  """
8458
  _OP_REQP = ["kind", "name", "tags"]
8459
  REQ_BGL = False
8460

    
8461
  def CheckPrereq(self):
8462
    """Check prerequisites.
8463

8464
    This checks that we have the given tag.
8465

8466
    """
8467
    TagsLU.CheckPrereq(self)
8468
    for tag in self.op.tags:
8469
      objects.TaggableObject.ValidateTag(tag)
8470
    del_tags = frozenset(self.op.tags)
8471
    cur_tags = self.target.GetTags()
8472
    if not del_tags <= cur_tags:
8473
      diff_tags = del_tags - cur_tags
8474
      diff_names = ["'%s'" % tag for tag in diff_tags]
8475
      diff_names.sort()
8476
      raise errors.OpPrereqError("Tag(s) %s not found" %
8477
                                 (",".join(diff_names)), errors.ECODE_NOENT)
8478

    
8479
  def Exec(self, feedback_fn):
8480
    """Remove the tag from the object.
8481

8482
    """
8483
    for tag in self.op.tags:
8484
      self.target.RemoveTag(tag)
8485
    self.cfg.Update(self.target, feedback_fn)
8486

    
8487

    
8488
class LUTestDelay(NoHooksLU):
8489
  """Sleep for a specified amount of time.
8490

8491
  This LU sleeps on the master and/or nodes for a specified amount of
8492
  time.
8493

8494
  """
8495
  _OP_REQP = ["duration", "on_master", "on_nodes"]
8496
  REQ_BGL = False
8497

    
8498
  def ExpandNames(self):
8499
    """Expand names and set required locks.
8500

8501
    This expands the node list, if any.
8502

8503
    """
8504
    self.needed_locks = {}
8505
    if self.op.on_nodes:
8506
      # _GetWantedNodes can be used here, but is not always appropriate to use
8507
      # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
8508
      # more information.
8509
      self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
8510
      self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
8511

    
8512
  def CheckPrereq(self):
8513
    """Check prerequisites.
8514

8515
    """
8516

    
8517
  def Exec(self, feedback_fn):
8518
    """Do the actual sleep.
8519

8520
    """
8521
    if self.op.on_master:
8522
      if not utils.TestDelay(self.op.duration):
8523
        raise errors.OpExecError("Error during master delay test")
8524
    if self.op.on_nodes:
8525
      result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
8526
      for node, node_result in result.items():
8527
        node_result.Raise("Failure during rpc call to node %s" % node)
8528

    
8529

    
8530
class IAllocator(object):
8531
  """IAllocator framework.
8532

8533
  An IAllocator instance has three sets of attributes:
8534
    - cfg that is needed to query the cluster
8535
    - input data (all members of the _KEYS class attribute are required)
8536
    - four buffer attributes (in|out_data|text), that represent the
8537
      input (to the external script) in text and data structure format,
8538
      and the output from it, again in two formats
8539
    - the result variables from the script (success, info, nodes) for
8540
      easy usage
8541

8542
  """
8543
  # pylint: disable-msg=R0902
8544
  # lots of instance attributes
8545
  _ALLO_KEYS = [
8546
    "mem_size", "disks", "disk_template",
8547
    "os", "tags", "nics", "vcpus", "hypervisor",
8548
    ]
8549
  _RELO_KEYS = [
8550
    "relocate_from",
8551
    ]
8552

    
8553
  def __init__(self, cfg, rpc, mode, name, **kwargs):
8554
    self.cfg = cfg
8555
    self.rpc = rpc
8556
    # init buffer variables
8557
    self.in_text = self.out_text = self.in_data = self.out_data = None
8558
    # init all input fields so that pylint is happy
8559
    self.mode = mode
8560
    self.name = name
8561
    self.mem_size = self.disks = self.disk_template = None
8562
    self.os = self.tags = self.nics = self.vcpus = None
8563
    self.hypervisor = None
8564
    self.relocate_from = None
8565
    # computed fields
8566
    self.required_nodes = None
8567
    # init result fields
8568
    self.success = self.info = self.nodes = None
8569
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
8570
      keyset = self._ALLO_KEYS
8571
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
8572
      keyset = self._RELO_KEYS
8573
    else:
8574
      raise errors.ProgrammerError("Unknown mode '%s' passed to the"
8575
                                   " IAllocator" % self.mode)
8576
    for key in kwargs:
8577
      if key not in keyset:
8578
        raise errors.ProgrammerError("Invalid input parameter '%s' to"
8579
                                     " IAllocator" % key)
8580
      setattr(self, key, kwargs[key])
8581
    for key in keyset:
8582
      if key not in kwargs:
8583
        raise errors.ProgrammerError("Missing input parameter '%s' to"
8584
                                     " IAllocator" % key)
8585
    self._BuildInputData()
8586

    
8587
  def _ComputeClusterData(self):
8588
    """Compute the generic allocator input data.
8589

8590
    This is the data that is independent of the actual operation.
8591

8592
    """
8593
    cfg = self.cfg
8594
    cluster_info = cfg.GetClusterInfo()
8595
    # cluster data
8596
    data = {
8597
      "version": constants.IALLOCATOR_VERSION,
8598
      "cluster_name": cfg.GetClusterName(),
8599
      "cluster_tags": list(cluster_info.GetTags()),
8600
      "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
8601
      # we don't have job IDs
8602
      }
8603
    iinfo = cfg.GetAllInstancesInfo().values()
8604
    i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
8605

    
8606
    # node data
8607
    node_results = {}
8608
    node_list = cfg.GetNodeList()
8609

    
8610
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
8611
      hypervisor_name = self.hypervisor
8612
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
8613
      hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
8614

    
8615
    node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(),
8616
                                        hypervisor_name)
8617
    node_iinfo = \
8618
      self.rpc.call_all_instances_info(node_list,
8619
                                       cluster_info.enabled_hypervisors)
8620
    for nname, nresult in node_data.items():
8621
      # first fill in static (config-based) values
8622
      ninfo = cfg.GetNodeInfo(nname)
8623
      pnr = {
8624
        "tags": list(ninfo.GetTags()),
8625
        "primary_ip": ninfo.primary_ip,
8626
        "secondary_ip": ninfo.secondary_ip,
8627
        "offline": ninfo.offline,
8628
        "drained": ninfo.drained,
8629
        "master_candidate": ninfo.master_candidate,
8630
        }
8631

    
8632
      if not (ninfo.offline or ninfo.drained):
8633
        nresult.Raise("Can't get data for node %s" % nname)
8634
        node_iinfo[nname].Raise("Can't get node instance info from node %s" %
8635
                                nname)
8636
        remote_info = nresult.payload
8637

    
8638
        for attr in ['memory_total', 'memory_free', 'memory_dom0',
8639
                     'vg_size', 'vg_free', 'cpu_total']:
8640
          if attr not in remote_info:
8641
            raise errors.OpExecError("Node '%s' didn't return attribute"
8642
                                     " '%s'" % (nname, attr))
8643
          if not isinstance(remote_info[attr], int):
8644
            raise errors.OpExecError("Node '%s' returned invalid value"
8645
                                     " for '%s': %s" %
8646
                                     (nname, attr, remote_info[attr]))
8647
        # compute memory used by primary instances
8648
        i_p_mem = i_p_up_mem = 0
8649
        for iinfo, beinfo in i_list:
8650
          if iinfo.primary_node == nname:
8651
            i_p_mem += beinfo[constants.BE_MEMORY]
8652
            if iinfo.name not in node_iinfo[nname].payload:
8653
              i_used_mem = 0
8654
            else:
8655
              i_used_mem = int(node_iinfo[nname].payload[iinfo.name]['memory'])
8656
            i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
8657
            remote_info['memory_free'] -= max(0, i_mem_diff)
8658

    
8659
            if iinfo.admin_up:
8660
              i_p_up_mem += beinfo[constants.BE_MEMORY]
8661

    
8662
        # compute memory used by instances
8663
        pnr_dyn = {
8664
          "total_memory": remote_info['memory_total'],
8665
          "reserved_memory": remote_info['memory_dom0'],
8666
          "free_memory": remote_info['memory_free'],
8667
          "total_disk": remote_info['vg_size'],
8668
          "free_disk": remote_info['vg_free'],
8669
          "total_cpus": remote_info['cpu_total'],
8670
          "i_pri_memory": i_p_mem,
8671
          "i_pri_up_memory": i_p_up_mem,
8672
          }
8673
        pnr.update(pnr_dyn)
8674

    
8675
      node_results[nname] = pnr
8676
    data["nodes"] = node_results
8677

    
8678
    # instance data
8679
    instance_data = {}
8680
    for iinfo, beinfo in i_list:
8681
      nic_data = []
8682
      for nic in iinfo.nics:
8683
        filled_params = objects.FillDict(
8684
            cluster_info.nicparams[constants.PP_DEFAULT],
8685
            nic.nicparams)
8686
        nic_dict = {"mac": nic.mac,
8687
                    "ip": nic.ip,
8688
                    "mode": filled_params[constants.NIC_MODE],
8689
                    "link": filled_params[constants.NIC_LINK],
8690
                   }
8691
        if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
8692
          nic_dict["bridge"] = filled_params[constants.NIC_LINK]
8693
        nic_data.append(nic_dict)
8694
      pir = {
8695
        "tags": list(iinfo.GetTags()),
8696
        "admin_up": iinfo.admin_up,
8697
        "vcpus": beinfo[constants.BE_VCPUS],
8698
        "memory": beinfo[constants.BE_MEMORY],
8699
        "os": iinfo.os,
8700
        "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
8701
        "nics": nic_data,
8702
        "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
8703
        "disk_template": iinfo.disk_template,
8704
        "hypervisor": iinfo.hypervisor,
8705
        }
8706
      pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
8707
                                                 pir["disks"])
8708
      instance_data[iinfo.name] = pir
8709

    
8710
    data["instances"] = instance_data
8711

    
8712
    self.in_data = data
8713

    
8714
  def _AddNewInstance(self):
8715
    """Add new instance data to allocator structure.
8716

8717
    This in combination with _AllocatorGetClusterData will create the
8718
    correct structure needed as input for the allocator.
8719

8720
    The checks for the completeness of the opcode must have already been
8721
    done.
8722

8723
    """
8724
    data = self.in_data
8725

    
8726
    disk_space = _ComputeDiskSize(self.disk_template, self.disks)
8727

    
8728
    if self.disk_template in constants.DTS_NET_MIRROR:
8729
      self.required_nodes = 2
8730
    else:
8731
      self.required_nodes = 1
8732
    request = {
8733
      "type": "allocate",
8734
      "name": self.name,
8735
      "disk_template": self.disk_template,
8736
      "tags": self.tags,
8737
      "os": self.os,
8738
      "vcpus": self.vcpus,
8739
      "memory": self.mem_size,
8740
      "disks": self.disks,
8741
      "disk_space_total": disk_space,
8742
      "nics": self.nics,
8743
      "required_nodes": self.required_nodes,
8744
      }
8745
    data["request"] = request
8746

    
8747
  def _AddRelocateInstance(self):
8748
    """Add relocate instance data to allocator structure.
8749

8750
    This in combination with _IAllocatorGetClusterData will create the
8751
    correct structure needed as input for the allocator.
8752

8753
    The checks for the completeness of the opcode must have already been
8754
    done.
8755

8756
    """
8757
    instance = self.cfg.GetInstanceInfo(self.name)
8758
    if instance is None:
8759
      raise errors.ProgrammerError("Unknown instance '%s' passed to"
8760
                                   " IAllocator" % self.name)
8761

    
8762
    if instance.disk_template not in constants.DTS_NET_MIRROR:
8763
      raise errors.OpPrereqError("Can't relocate non-mirrored instances",
8764
                                 errors.ECODE_INVAL)
8765

    
8766
    if len(instance.secondary_nodes) != 1:
8767
      raise errors.OpPrereqError("Instance has not exactly one secondary node",
8768
                                 errors.ECODE_STATE)
8769

    
8770
    self.required_nodes = 1
8771
    disk_sizes = [{'size': disk.size} for disk in instance.disks]
8772
    disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
8773

    
8774
    request = {
8775
      "type": "relocate",
8776
      "name": self.name,
8777
      "disk_space_total": disk_space,
8778
      "required_nodes": self.required_nodes,
8779
      "relocate_from": self.relocate_from,
8780
      }
8781
    self.in_data["request"] = request
8782

    
8783
  def _BuildInputData(self):
8784
    """Build input data structures.
8785

8786
    """
8787
    self._ComputeClusterData()
8788

    
8789
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
8790
      self._AddNewInstance()
8791
    else:
8792
      self._AddRelocateInstance()
8793

    
8794
    self.in_text = serializer.Dump(self.in_data)
8795

    
8796
  def Run(self, name, validate=True, call_fn=None):
8797
    """Run an instance allocator and return the results.
8798

8799
    """
8800
    if call_fn is None:
8801
      call_fn = self.rpc.call_iallocator_runner
8802

    
8803
    result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
8804
    result.Raise("Failure while running the iallocator script")
8805

    
8806
    self.out_text = result.payload
8807
    if validate:
8808
      self._ValidateResult()
8809

    
8810
  def _ValidateResult(self):
8811
    """Process the allocator results.
8812

8813
    This will process and if successful save the result in
8814
    self.out_data and the other parameters.
8815

8816
    """
8817
    try:
8818
      rdict = serializer.Load(self.out_text)
8819
    except Exception, err:
8820
      raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
8821

    
8822
    if not isinstance(rdict, dict):
8823
      raise errors.OpExecError("Can't parse iallocator results: not a dict")
8824

    
8825
    for key in "success", "info", "nodes":
8826
      if key not in rdict:
8827
        raise errors.OpExecError("Can't parse iallocator results:"
8828
                                 " missing key '%s'" % key)
8829
      setattr(self, key, rdict[key])
8830

    
8831
    if not isinstance(rdict["nodes"], list):
8832
      raise errors.OpExecError("Can't parse iallocator results: 'nodes' key"
8833
                               " is not a list")
8834
    self.out_data = rdict
8835

    
8836

    
8837
class LUTestAllocator(NoHooksLU):
8838
  """Run allocator tests.
8839

8840
  This LU runs the allocator tests
8841

8842
  """
8843
  _OP_REQP = ["direction", "mode", "name"]
8844

    
8845
  def CheckPrereq(self):
8846
    """Check prerequisites.
8847

8848
    This checks the opcode parameters depending on the director and mode test.
8849

8850
    """
8851
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
8852
      for attr in ["name", "mem_size", "disks", "disk_template",
8853
                   "os", "tags", "nics", "vcpus"]:
8854
        if not hasattr(self.op, attr):
8855
          raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
8856
                                     attr, errors.ECODE_INVAL)
8857
      iname = self.cfg.ExpandInstanceName(self.op.name)
8858
      if iname is not None:
8859
        raise errors.OpPrereqError("Instance '%s' already in the cluster" %
8860
                                   iname, errors.ECODE_EXISTS)
8861
      if not isinstance(self.op.nics, list):
8862
        raise errors.OpPrereqError("Invalid parameter 'nics'",
8863
                                   errors.ECODE_INVAL)
8864
      for row in self.op.nics:
8865
        if (not isinstance(row, dict) or
8866
            "mac" not in row or
8867
            "ip" not in row or
8868
            "bridge" not in row):
8869
          raise errors.OpPrereqError("Invalid contents of the 'nics'"
8870
                                     " parameter", errors.ECODE_INVAL)
8871
      if not isinstance(self.op.disks, list):
8872
        raise errors.OpPrereqError("Invalid parameter 'disks'",
8873
                                   errors.ECODE_INVAL)
8874
      for row in self.op.disks:
8875
        if (not isinstance(row, dict) or
8876
            "size" not in row or
8877
            not isinstance(row["size"], int) or
8878
            "mode" not in row or
8879
            row["mode"] not in ['r', 'w']):
8880
          raise errors.OpPrereqError("Invalid contents of the 'disks'"
8881
                                     " parameter", errors.ECODE_INVAL)
8882
      if not hasattr(self.op, "hypervisor") or self.op.hypervisor is None:
8883
        self.op.hypervisor = self.cfg.GetHypervisorType()
8884
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
8885
      if not hasattr(self.op, "name"):
8886
        raise errors.OpPrereqError("Missing attribute 'name' on opcode input",
8887
                                   errors.ECODE_INVAL)
8888
      fname = self.cfg.ExpandInstanceName(self.op.name)
8889
      if fname is None:
8890
        raise errors.OpPrereqError("Instance '%s' not found for relocation" %
8891
                                   self.op.name, errors.ECODE_NOENT)
8892
      self.op.name = fname
8893
      self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
8894
    else:
8895
      raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
8896
                                 self.op.mode, errors.ECODE_INVAL)
8897

    
8898
    if self.op.direction == constants.IALLOCATOR_DIR_OUT:
8899
      if not hasattr(self.op, "allocator") or self.op.allocator is None:
8900
        raise errors.OpPrereqError("Missing allocator name",
8901
                                   errors.ECODE_INVAL)
8902
    elif self.op.direction != constants.IALLOCATOR_DIR_IN:
8903
      raise errors.OpPrereqError("Wrong allocator test '%s'" %
8904
                                 self.op.direction, errors.ECODE_INVAL)
8905

    
8906
  def Exec(self, feedback_fn):
8907
    """Run the allocator test.
8908

8909
    """
8910
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
8911
      ial = IAllocator(self.cfg, self.rpc,
8912
                       mode=self.op.mode,
8913
                       name=self.op.name,
8914
                       mem_size=self.op.mem_size,
8915
                       disks=self.op.disks,
8916
                       disk_template=self.op.disk_template,
8917
                       os=self.op.os,
8918
                       tags=self.op.tags,
8919
                       nics=self.op.nics,
8920
                       vcpus=self.op.vcpus,
8921
                       hypervisor=self.op.hypervisor,
8922
                       )
8923
    else:
8924
      ial = IAllocator(self.cfg, self.rpc,
8925
                       mode=self.op.mode,
8926
                       name=self.op.name,
8927
                       relocate_from=list(self.relocate_from),
8928
                       )
8929

    
8930
    if self.op.direction == constants.IALLOCATOR_DIR_IN:
8931
      result = ial.in_text
8932
    else:
8933
      ial.Run(self.op.allocator, validate=False)
8934
      result = ial.out_text
8935
    return result