Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ 36b66e6e

History | View | Annotate | Download (305.7 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=W0613,W0201
25

    
26
import os
27
import os.path
28
import time
29
import re
30
import platform
31
import logging
32
import copy
33

    
34
from ganeti import ssh
35
from ganeti import utils
36
from ganeti import errors
37
from ganeti import hypervisor
38
from ganeti import locking
39
from ganeti import constants
40
from ganeti import objects
41
from ganeti import serializer
42
from ganeti import ssconf
43

    
44

    
45
class LogicalUnit(object):
46
  """Logical Unit base class.
47

48
  Subclasses must follow these rules:
49
    - implement ExpandNames
50
    - implement CheckPrereq (except when tasklets are used)
51
    - implement Exec (except when tasklets are used)
52
    - implement BuildHooksEnv
53
    - redefine HPATH and HTYPE
54
    - optionally redefine their run requirements:
55
        REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
56

57
  Note that all commands require root permissions.
58

59
  @ivar dry_run_result: the value (if any) that will be returned to the caller
60
      in dry-run mode (signalled by opcode dry_run parameter)
61

62
  """
63
  HPATH = None
64
  HTYPE = None
65
  _OP_REQP = []
66
  REQ_BGL = True
67

    
68
  def __init__(self, processor, op, context, rpc):
69
    """Constructor for LogicalUnit.
70

71
    This needs to be overridden in derived classes in order to check op
72
    validity.
73

74
    """
75
    self.proc = processor
76
    self.op = op
77
    self.cfg = context.cfg
78
    self.context = context
79
    self.rpc = rpc
80
    # Dicts used to declare locking needs to mcpu
81
    self.needed_locks = None
82
    self.acquired_locks = {}
83
    self.share_locks = dict.fromkeys(locking.LEVELS, 0)
84
    self.add_locks = {}
85
    self.remove_locks = {}
86
    # Used to force good behavior when calling helper functions
87
    self.recalculate_locks = {}
88
    self.__ssh = None
89
    # logging
90
    self.LogWarning = processor.LogWarning
91
    self.LogInfo = processor.LogInfo
92
    self.LogStep = processor.LogStep
93
    # support for dry-run
94
    self.dry_run_result = None
95

    
96
    # Tasklets
97
    self.tasklets = None
98

    
99
    for attr_name in self._OP_REQP:
100
      attr_val = getattr(op, attr_name, None)
101
      if attr_val is None:
102
        raise errors.OpPrereqError("Required parameter '%s' missing" %
103
                                   attr_name, errors.ECODE_INVAL)
104

    
105
    self.CheckArguments()
106

    
107
  def __GetSSH(self):
108
    """Returns the SshRunner object
109

110
    """
111
    if not self.__ssh:
112
      self.__ssh = ssh.SshRunner(self.cfg.GetClusterName())
113
    return self.__ssh
114

    
115
  ssh = property(fget=__GetSSH)
116

    
117
  def CheckArguments(self):
118
    """Check syntactic validity for the opcode arguments.
119

120
    This method is for doing a simple syntactic check and ensure
121
    validity of opcode parameters, without any cluster-related
122
    checks. While the same can be accomplished in ExpandNames and/or
123
    CheckPrereq, doing these separate is better because:
124

125
      - ExpandNames is left as as purely a lock-related function
126
      - CheckPrereq is run after we have acquired locks (and possible
127
        waited for them)
128

129
    The function is allowed to change the self.op attribute so that
130
    later methods can no longer worry about missing parameters.
131

132
    """
133
    pass
134

    
135
  def ExpandNames(self):
136
    """Expand names for this LU.
137

138
    This method is called before starting to execute the opcode, and it should
139
    update all the parameters of the opcode to their canonical form (e.g. a
140
    short node name must be fully expanded after this method has successfully
141
    completed). This way locking, hooks, logging, ecc. can work correctly.
142

143
    LUs which implement this method must also populate the self.needed_locks
144
    member, as a dict with lock levels as keys, and a list of needed lock names
145
    as values. Rules:
146

147
      - use an empty dict if you don't need any lock
148
      - if you don't need any lock at a particular level omit that level
149
      - don't put anything for the BGL level
150
      - if you want all locks at a level use locking.ALL_SET as a value
151

152
    If you need to share locks (rather than acquire them exclusively) at one
153
    level you can modify self.share_locks, setting a true value (usually 1) for
154
    that level. By default locks are not shared.
155

156
    This function can also define a list of tasklets, which then will be
157
    executed in order instead of the usual LU-level CheckPrereq and Exec
158
    functions, if those are not defined by the LU.
159

160
    Examples::
161

162
      # Acquire all nodes and one instance
163
      self.needed_locks = {
164
        locking.LEVEL_NODE: locking.ALL_SET,
165
        locking.LEVEL_INSTANCE: ['instance1.example.tld'],
166
      }
167
      # Acquire just two nodes
168
      self.needed_locks = {
169
        locking.LEVEL_NODE: ['node1.example.tld', 'node2.example.tld'],
170
      }
171
      # Acquire no locks
172
      self.needed_locks = {} # No, you can't leave it to the default value None
173

174
    """
175
    # The implementation of this method is mandatory only if the new LU is
176
    # concurrent, so that old LUs don't need to be changed all at the same
177
    # time.
178
    if self.REQ_BGL:
179
      self.needed_locks = {} # Exclusive LUs don't need locks.
180
    else:
181
      raise NotImplementedError
182

    
183
  def DeclareLocks(self, level):
184
    """Declare LU locking needs for a level
185

186
    While most LUs can just declare their locking needs at ExpandNames time,
187
    sometimes there's the need to calculate some locks after having acquired
188
    the ones before. This function is called just before acquiring locks at a
189
    particular level, but after acquiring the ones at lower levels, and permits
190
    such calculations. It can be used to modify self.needed_locks, and by
191
    default it does nothing.
192

193
    This function is only called if you have something already set in
194
    self.needed_locks for the level.
195

196
    @param level: Locking level which is going to be locked
197
    @type level: member of ganeti.locking.LEVELS
198

199
    """
200

    
201
  def CheckPrereq(self):
202
    """Check prerequisites for this LU.
203

204
    This method should check that the prerequisites for the execution
205
    of this LU are fulfilled. It can do internode communication, but
206
    it should be idempotent - no cluster or system changes are
207
    allowed.
208

209
    The method should raise errors.OpPrereqError in case something is
210
    not fulfilled. Its return value is ignored.
211

212
    This method should also update all the parameters of the opcode to
213
    their canonical form if it hasn't been done by ExpandNames before.
214

215
    """
216
    if self.tasklets is not None:
217
      for (idx, tl) in enumerate(self.tasklets):
218
        logging.debug("Checking prerequisites for tasklet %s/%s",
219
                      idx + 1, len(self.tasklets))
220
        tl.CheckPrereq()
221
    else:
222
      raise NotImplementedError
223

    
224
  def Exec(self, feedback_fn):
225
    """Execute the LU.
226

227
    This method should implement the actual work. It should raise
228
    errors.OpExecError for failures that are somewhat dealt with in
229
    code, or expected.
230

231
    """
232
    if self.tasklets is not None:
233
      for (idx, tl) in enumerate(self.tasklets):
234
        logging.debug("Executing tasklet %s/%s", idx + 1, len(self.tasklets))
235
        tl.Exec(feedback_fn)
236
    else:
237
      raise NotImplementedError
238

    
239
  def BuildHooksEnv(self):
240
    """Build hooks environment for this LU.
241

242
    This method should return a three-node tuple consisting of: a dict
243
    containing the environment that will be used for running the
244
    specific hook for this LU, a list of node names on which the hook
245
    should run before the execution, and a list of node names on which
246
    the hook should run after the execution.
247

248
    The keys of the dict must not have 'GANETI_' prefixed as this will
249
    be handled in the hooks runner. Also note additional keys will be
250
    added by the hooks runner. If the LU doesn't define any
251
    environment, an empty dict (and not None) should be returned.
252

253
    No nodes should be returned as an empty list (and not None).
254

255
    Note that if the HPATH for a LU class is None, this function will
256
    not be called.
257

258
    """
259
    raise NotImplementedError
260

    
261
  def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
262
    """Notify the LU about the results of its hooks.
263

264
    This method is called every time a hooks phase is executed, and notifies
265
    the Logical Unit about the hooks' result. The LU can then use it to alter
266
    its result based on the hooks.  By default the method does nothing and the
267
    previous result is passed back unchanged but any LU can define it if it
268
    wants to use the local cluster hook-scripts somehow.
269

270
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
271
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
272
    @param hook_results: the results of the multi-node hooks rpc call
273
    @param feedback_fn: function used send feedback back to the caller
274
    @param lu_result: the previous Exec result this LU had, or None
275
        in the PRE phase
276
    @return: the new Exec result, based on the previous result
277
        and hook results
278

279
    """
280
    return lu_result
281

    
282
  def _ExpandAndLockInstance(self):
283
    """Helper function to expand and lock an instance.
284

285
    Many LUs that work on an instance take its name in self.op.instance_name
286
    and need to expand it and then declare the expanded name for locking. This
287
    function does it, and then updates self.op.instance_name to the expanded
288
    name. It also initializes needed_locks as a dict, if this hasn't been done
289
    before.
290

291
    """
292
    if self.needed_locks is None:
293
      self.needed_locks = {}
294
    else:
295
      assert locking.LEVEL_INSTANCE not in self.needed_locks, \
296
        "_ExpandAndLockInstance called with instance-level locks set"
297
    expanded_name = self.cfg.ExpandInstanceName(self.op.instance_name)
298
    if expanded_name is None:
299
      raise errors.OpPrereqError("Instance '%s' not known" %
300
                                 self.op.instance_name, errors.ECODE_NOENT)
301
    self.needed_locks[locking.LEVEL_INSTANCE] = expanded_name
302
    self.op.instance_name = expanded_name
303

    
304
  def _LockInstancesNodes(self, primary_only=False):
305
    """Helper function to declare instances' nodes for locking.
306

307
    This function should be called after locking one or more instances to lock
308
    their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
309
    with all primary or secondary nodes for instances already locked and
310
    present in self.needed_locks[locking.LEVEL_INSTANCE].
311

312
    It should be called from DeclareLocks, and for safety only works if
313
    self.recalculate_locks[locking.LEVEL_NODE] is set.
314

315
    In the future it may grow parameters to just lock some instance's nodes, or
316
    to just lock primaries or secondary nodes, if needed.
317

318
    If should be called in DeclareLocks in a way similar to::
319

320
      if level == locking.LEVEL_NODE:
321
        self._LockInstancesNodes()
322

323
    @type primary_only: boolean
324
    @param primary_only: only lock primary nodes of locked instances
325

326
    """
327
    assert locking.LEVEL_NODE in self.recalculate_locks, \
328
      "_LockInstancesNodes helper function called with no nodes to recalculate"
329

    
330
    # TODO: check if we're really been called with the instance locks held
331

    
332
    # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
333
    # future we might want to have different behaviors depending on the value
334
    # of self.recalculate_locks[locking.LEVEL_NODE]
335
    wanted_nodes = []
336
    for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
337
      instance = self.context.cfg.GetInstanceInfo(instance_name)
338
      wanted_nodes.append(instance.primary_node)
339
      if not primary_only:
340
        wanted_nodes.extend(instance.secondary_nodes)
341

    
342
    if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE:
343
      self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
344
    elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND:
345
      self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
346

    
347
    del self.recalculate_locks[locking.LEVEL_NODE]
348

    
349

    
350
class NoHooksLU(LogicalUnit):
351
  """Simple LU which runs no hooks.
352

353
  This LU is intended as a parent for other LogicalUnits which will
354
  run no hooks, in order to reduce duplicate code.
355

356
  """
357
  HPATH = None
358
  HTYPE = None
359

    
360

    
361
class Tasklet:
362
  """Tasklet base class.
363

364
  Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
365
  they can mix legacy code with tasklets. Locking needs to be done in the LU,
366
  tasklets know nothing about locks.
367

368
  Subclasses must follow these rules:
369
    - Implement CheckPrereq
370
    - Implement Exec
371

372
  """
373
  def __init__(self, lu):
374
    self.lu = lu
375

    
376
    # Shortcuts
377
    self.cfg = lu.cfg
378
    self.rpc = lu.rpc
379

    
380
  def CheckPrereq(self):
381
    """Check prerequisites for this tasklets.
382

383
    This method should check whether the prerequisites for the execution of
384
    this tasklet are fulfilled. It can do internode communication, but it
385
    should be idempotent - no cluster or system changes are allowed.
386

387
    The method should raise errors.OpPrereqError in case something is not
388
    fulfilled. Its return value is ignored.
389

390
    This method should also update all parameters to their canonical form if it
391
    hasn't been done before.
392

393
    """
394
    raise NotImplementedError
395

    
396
  def Exec(self, feedback_fn):
397
    """Execute the tasklet.
398

399
    This method should implement the actual work. It should raise
400
    errors.OpExecError for failures that are somewhat dealt with in code, or
401
    expected.
402

403
    """
404
    raise NotImplementedError
405

    
406

    
407
def _GetWantedNodes(lu, nodes):
408
  """Returns list of checked and expanded node names.
409

410
  @type lu: L{LogicalUnit}
411
  @param lu: the logical unit on whose behalf we execute
412
  @type nodes: list
413
  @param nodes: list of node names or None for all nodes
414
  @rtype: list
415
  @return: the list of nodes, sorted
416
  @raise errors.OpProgrammerError: if the nodes parameter is wrong type
417

418
  """
419
  if not isinstance(nodes, list):
420
    raise errors.OpPrereqError("Invalid argument type 'nodes'",
421
                               errors.ECODE_INVAL)
422

    
423
  if not nodes:
424
    raise errors.ProgrammerError("_GetWantedNodes should only be called with a"
425
      " non-empty list of nodes whose name is to be expanded.")
426

    
427
  wanted = []
428
  for name in nodes:
429
    node = lu.cfg.ExpandNodeName(name)
430
    if node is None:
431
      raise errors.OpPrereqError("No such node name '%s'" % name,
432
                                 errors.ECODE_NOENT)
433
    wanted.append(node)
434

    
435
  return utils.NiceSort(wanted)
436

    
437

    
438
def _GetWantedInstances(lu, instances):
439
  """Returns list of checked and expanded instance names.
440

441
  @type lu: L{LogicalUnit}
442
  @param lu: the logical unit on whose behalf we execute
443
  @type instances: list
444
  @param instances: list of instance names or None for all instances
445
  @rtype: list
446
  @return: the list of instances, sorted
447
  @raise errors.OpPrereqError: if the instances parameter is wrong type
448
  @raise errors.OpPrereqError: if any of the passed instances is not found
449

450
  """
451
  if not isinstance(instances, list):
452
    raise errors.OpPrereqError("Invalid argument type 'instances'",
453
                               errors.ECODE_INVAL)
454

    
455
  if instances:
456
    wanted = []
457

    
458
    for name in instances:
459
      instance = lu.cfg.ExpandInstanceName(name)
460
      if instance is None:
461
        raise errors.OpPrereqError("No such instance name '%s'" % name,
462
                                   errors.ECODE_NOENT)
463
      wanted.append(instance)
464

    
465
  else:
466
    wanted = utils.NiceSort(lu.cfg.GetInstanceList())
467
  return wanted
468

    
469

    
470
def _CheckOutputFields(static, dynamic, selected):
471
  """Checks whether all selected fields are valid.
472

473
  @type static: L{utils.FieldSet}
474
  @param static: static fields set
475
  @type dynamic: L{utils.FieldSet}
476
  @param dynamic: dynamic fields set
477

478
  """
479
  f = utils.FieldSet()
480
  f.Extend(static)
481
  f.Extend(dynamic)
482

    
483
  delta = f.NonMatching(selected)
484
  if delta:
485
    raise errors.OpPrereqError("Unknown output fields selected: %s"
486
                               % ",".join(delta), errors.ECODE_INVAL)
487

    
488

    
489
def _CheckBooleanOpField(op, name):
490
  """Validates boolean opcode parameters.
491

492
  This will ensure that an opcode parameter is either a boolean value,
493
  or None (but that it always exists).
494

495
  """
496
  val = getattr(op, name, None)
497
  if not (val is None or isinstance(val, bool)):
498
    raise errors.OpPrereqError("Invalid boolean parameter '%s' (%s)" %
499
                               (name, str(val)), errors.ECODE_INVAL)
500
  setattr(op, name, val)
501

    
502

    
503
def _CheckGlobalHvParams(params):
504
  """Validates that given hypervisor params are not global ones.
505

506
  This will ensure that instances don't get customised versions of
507
  global params.
508

509
  """
510
  used_globals = constants.HVC_GLOBALS.intersection(params)
511
  if used_globals:
512
    msg = ("The following hypervisor parameters are global and cannot"
513
           " be customized at instance level, please modify them at"
514
           " cluster level: %s" % ", ".join(used_globals))
515
    raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
516

    
517

    
518
def _CheckNodeOnline(lu, node):
519
  """Ensure that a given node is online.
520

521
  @param lu: the LU on behalf of which we make the check
522
  @param node: the node to check
523
  @raise errors.OpPrereqError: if the node is offline
524

525
  """
526
  if lu.cfg.GetNodeInfo(node).offline:
527
    raise errors.OpPrereqError("Can't use offline node %s" % node,
528
                               errors.ECODE_INVAL)
529

    
530

    
531
def _CheckNodeNotDrained(lu, node):
532
  """Ensure that a given node is not drained.
533

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

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

    
543

    
544
def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
545
                          memory, vcpus, nics, disk_template, disks,
546
                          bep, hvp, hypervisor_name):
547
  """Builds instance related env variables for hooks
548

549
  This builds the hook environment from individual variables.
550

551
  @type name: string
552
  @param name: the name of the instance
553
  @type primary_node: string
554
  @param primary_node: the name of the instance's primary node
555
  @type secondary_nodes: list
556
  @param secondary_nodes: list of secondary nodes as strings
557
  @type os_type: string
558
  @param os_type: the name of the instance's OS
559
  @type status: boolean
560
  @param status: the should_run status of the instance
561
  @type memory: string
562
  @param memory: the memory size of the instance
563
  @type vcpus: string
564
  @param vcpus: the count of VCPUs the instance has
565
  @type nics: list
566
  @param nics: list of tuples (ip, mac, mode, link) representing
567
      the NICs the instance has
568
  @type disk_template: string
569
  @param disk_template: the disk template of the instance
570
  @type disks: list
571
  @param disks: the list of (size, mode) pairs
572
  @type bep: dict
573
  @param bep: the backend parameters for the instance
574
  @type hvp: dict
575
  @param hvp: the hypervisor parameters for the instance
576
  @type hypervisor_name: string
577
  @param hypervisor_name: the hypervisor for the instance
578
  @rtype: dict
579
  @return: the hook environment for this instance
580

581
  """
582
  if status:
583
    str_status = "up"
584
  else:
585
    str_status = "down"
586
  env = {
587
    "OP_TARGET": name,
588
    "INSTANCE_NAME": name,
589
    "INSTANCE_PRIMARY": primary_node,
590
    "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
591
    "INSTANCE_OS_TYPE": os_type,
592
    "INSTANCE_STATUS": str_status,
593
    "INSTANCE_MEMORY": memory,
594
    "INSTANCE_VCPUS": vcpus,
595
    "INSTANCE_DISK_TEMPLATE": disk_template,
596
    "INSTANCE_HYPERVISOR": hypervisor_name,
597
  }
598

    
599
  if nics:
600
    nic_count = len(nics)
601
    for idx, (ip, mac, mode, link) in enumerate(nics):
602
      if ip is None:
603
        ip = ""
604
      env["INSTANCE_NIC%d_IP" % idx] = ip
605
      env["INSTANCE_NIC%d_MAC" % idx] = mac
606
      env["INSTANCE_NIC%d_MODE" % idx] = mode
607
      env["INSTANCE_NIC%d_LINK" % idx] = link
608
      if mode == constants.NIC_MODE_BRIDGED:
609
        env["INSTANCE_NIC%d_BRIDGE" % idx] = link
610
  else:
611
    nic_count = 0
612

    
613
  env["INSTANCE_NIC_COUNT"] = nic_count
614

    
615
  if disks:
616
    disk_count = len(disks)
617
    for idx, (size, mode) in enumerate(disks):
618
      env["INSTANCE_DISK%d_SIZE" % idx] = size
619
      env["INSTANCE_DISK%d_MODE" % idx] = mode
620
  else:
621
    disk_count = 0
622

    
623
  env["INSTANCE_DISK_COUNT"] = disk_count
624

    
625
  for source, kind in [(bep, "BE"), (hvp, "HV")]:
626
    for key, value in source.items():
627
      env["INSTANCE_%s_%s" % (kind, key)] = value
628

    
629
  return env
630

    
631

    
632
def _NICListToTuple(lu, nics):
633
  """Build a list of nic information tuples.
634

635
  This list is suitable to be passed to _BuildInstanceHookEnv or as a return
636
  value in LUQueryInstanceData.
637

638
  @type lu:  L{LogicalUnit}
639
  @param lu: the logical unit on whose behalf we execute
640
  @type nics: list of L{objects.NIC}
641
  @param nics: list of nics to convert to hooks tuples
642

643
  """
644
  hooks_nics = []
645
  c_nicparams = lu.cfg.GetClusterInfo().nicparams[constants.PP_DEFAULT]
646
  for nic in nics:
647
    ip = nic.ip
648
    mac = nic.mac
649
    filled_params = objects.FillDict(c_nicparams, nic.nicparams)
650
    mode = filled_params[constants.NIC_MODE]
651
    link = filled_params[constants.NIC_LINK]
652
    hooks_nics.append((ip, mac, mode, link))
653
  return hooks_nics
654

    
655

    
656
def _BuildInstanceHookEnvByObject(lu, instance, override=None):
657
  """Builds instance related env variables for hooks from an object.
658

659
  @type lu: L{LogicalUnit}
660
  @param lu: the logical unit on whose behalf we execute
661
  @type instance: L{objects.Instance}
662
  @param instance: the instance for which we should build the
663
      environment
664
  @type override: dict
665
  @param override: dictionary with key/values that will override
666
      our values
667
  @rtype: dict
668
  @return: the hook environment dictionary
669

670
  """
671
  cluster = lu.cfg.GetClusterInfo()
672
  bep = cluster.FillBE(instance)
673
  hvp = cluster.FillHV(instance)
674
  args = {
675
    'name': instance.name,
676
    'primary_node': instance.primary_node,
677
    'secondary_nodes': instance.secondary_nodes,
678
    'os_type': instance.os,
679
    'status': instance.admin_up,
680
    'memory': bep[constants.BE_MEMORY],
681
    'vcpus': bep[constants.BE_VCPUS],
682
    'nics': _NICListToTuple(lu, instance.nics),
683
    'disk_template': instance.disk_template,
684
    'disks': [(disk.size, disk.mode) for disk in instance.disks],
685
    'bep': bep,
686
    'hvp': hvp,
687
    'hypervisor_name': instance.hypervisor,
688
  }
689
  if override:
690
    args.update(override)
691
  return _BuildInstanceHookEnv(**args)
692

    
693

    
694
def _AdjustCandidatePool(lu, exceptions):
695
  """Adjust the candidate pool after node operations.
696

697
  """
698
  mod_list = lu.cfg.MaintainCandidatePool(exceptions)
699
  if mod_list:
700
    lu.LogInfo("Promoted nodes to master candidate role: %s",
701
               ", ".join(node.name for node in mod_list))
702
    for name in mod_list:
703
      lu.context.ReaddNode(name)
704
  mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
705
  if mc_now > mc_max:
706
    lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
707
               (mc_now, mc_max))
708

    
709

    
710
def _DecideSelfPromotion(lu, exceptions=None):
711
  """Decide whether I should promote myself as a master candidate.
712

713
  """
714
  cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
715
  mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
716
  # the new node will increase mc_max with one, so:
717
  mc_should = min(mc_should + 1, cp_size)
718
  return mc_now < mc_should
719

    
720

    
721
def _CheckNicsBridgesExist(lu, target_nics, target_node,
722
                               profile=constants.PP_DEFAULT):
723
  """Check that the brigdes needed by a list of nics exist.
724

725
  """
726
  c_nicparams = lu.cfg.GetClusterInfo().nicparams[profile]
727
  paramslist = [objects.FillDict(c_nicparams, nic.nicparams)
728
                for nic in target_nics]
729
  brlist = [params[constants.NIC_LINK] for params in paramslist
730
            if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
731
  if brlist:
732
    result = lu.rpc.call_bridges_exist(target_node, brlist)
733
    result.Raise("Error checking bridges on destination node '%s'" %
734
                 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
735

    
736

    
737
def _CheckInstanceBridgesExist(lu, instance, node=None):
738
  """Check that the brigdes needed by an instance exist.
739

740
  """
741
  if node is None:
742
    node = instance.primary_node
743
  _CheckNicsBridgesExist(lu, instance.nics, node)
744

    
745

    
746
def _CheckOSVariant(os_obj, name):
747
  """Check whether an OS name conforms to the os variants specification.
748

749
  @type os_obj: L{objects.OS}
750
  @param os_obj: OS object to check
751
  @type name: string
752
  @param name: OS name passed by the user, to check for validity
753

754
  """
755
  if not os_obj.supported_variants:
756
    return
757
  try:
758
    variant = name.split("+", 1)[1]
759
  except IndexError:
760
    raise errors.OpPrereqError("OS name must include a variant",
761
                               errors.ECODE_INVAL)
762

    
763
  if variant not in os_obj.supported_variants:
764
    raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
765

    
766

    
767
def _GetNodeInstancesInner(cfg, fn):
768
  return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
769

    
770

    
771
def _GetNodeInstances(cfg, node_name):
772
  """Returns a list of all primary and secondary instances on a node.
773

774
  """
775

    
776
  return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
777

    
778

    
779
def _GetNodePrimaryInstances(cfg, node_name):
780
  """Returns primary instances on a node.
781

782
  """
783
  return _GetNodeInstancesInner(cfg,
784
                                lambda inst: node_name == inst.primary_node)
785

    
786

    
787
def _GetNodeSecondaryInstances(cfg, node_name):
788
  """Returns secondary instances on a node.
789

790
  """
791
  return _GetNodeInstancesInner(cfg,
792
                                lambda inst: node_name in inst.secondary_nodes)
793

    
794

    
795
def _GetStorageTypeArgs(cfg, storage_type):
796
  """Returns the arguments for a storage type.
797

798
  """
799
  # Special case for file storage
800
  if storage_type == constants.ST_FILE:
801
    # storage.FileStorage wants a list of storage directories
802
    return [[cfg.GetFileStorageDir()]]
803

    
804
  return []
805

    
806

    
807
def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq):
808
  faulty = []
809

    
810
  for dev in instance.disks:
811
    cfg.SetDiskID(dev, node_name)
812

    
813
  result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks)
814
  result.Raise("Failed to get disk status from node %s" % node_name,
815
               prereq=prereq, ecode=errors.ECODE_ENVIRON)
816

    
817
  for idx, bdev_status in enumerate(result.payload):
818
    if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
819
      faulty.append(idx)
820

    
821
  return faulty
822

    
823

    
824
class LUPostInitCluster(LogicalUnit):
825
  """Logical unit for running hooks after cluster initialization.
826

827
  """
828
  HPATH = "cluster-init"
829
  HTYPE = constants.HTYPE_CLUSTER
830
  _OP_REQP = []
831

    
832
  def BuildHooksEnv(self):
833
    """Build hooks env.
834

835
    """
836
    env = {"OP_TARGET": self.cfg.GetClusterName()}
837
    mn = self.cfg.GetMasterNode()
838
    return env, [], [mn]
839

    
840
  def CheckPrereq(self):
841
    """No prerequisites to check.
842

843
    """
844
    return True
845

    
846
  def Exec(self, feedback_fn):
847
    """Nothing to do.
848

849
    """
850
    return True
851

    
852

    
853
class LUDestroyCluster(LogicalUnit):
854
  """Logical unit for destroying the cluster.
855

856
  """
857
  HPATH = "cluster-destroy"
858
  HTYPE = constants.HTYPE_CLUSTER
859
  _OP_REQP = []
860

    
861
  def BuildHooksEnv(self):
862
    """Build hooks env.
863

864
    """
865
    env = {"OP_TARGET": self.cfg.GetClusterName()}
866
    return env, [], []
867

    
868
  def CheckPrereq(self):
869
    """Check prerequisites.
870

871
    This checks whether the cluster is empty.
872

873
    Any errors are signaled by raising errors.OpPrereqError.
874

875
    """
876
    master = self.cfg.GetMasterNode()
877

    
878
    nodelist = self.cfg.GetNodeList()
879
    if len(nodelist) != 1 or nodelist[0] != master:
880
      raise errors.OpPrereqError("There are still %d node(s) in"
881
                                 " this cluster." % (len(nodelist) - 1),
882
                                 errors.ECODE_INVAL)
883
    instancelist = self.cfg.GetInstanceList()
884
    if instancelist:
885
      raise errors.OpPrereqError("There are still %d instance(s) in"
886
                                 " this cluster." % len(instancelist),
887
                                 errors.ECODE_INVAL)
888

    
889
  def Exec(self, feedback_fn):
890
    """Destroys the cluster.
891

892
    """
893
    master = self.cfg.GetMasterNode()
894
    modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
895

    
896
    # Run post hooks on master node before it's removed
897
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
898
    try:
899
      hm.RunPhase(constants.HOOKS_PHASE_POST, [master])
900
    except:
901
      self.LogWarning("Errors occurred running hooks on %s" % master)
902

    
903
    result = self.rpc.call_node_stop_master(master, False)
904
    result.Raise("Could not disable the master role")
905

    
906
    if modify_ssh_setup:
907
      priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
908
      utils.CreateBackup(priv_key)
909
      utils.CreateBackup(pub_key)
910

    
911
    return master
912

    
913

    
914
class LUVerifyCluster(LogicalUnit):
915
  """Verifies the cluster status.
916

917
  """
918
  HPATH = "cluster-verify"
919
  HTYPE = constants.HTYPE_CLUSTER
920
  _OP_REQP = ["skip_checks", "verbose", "error_codes", "debug_simulate_errors"]
921
  REQ_BGL = False
922

    
923
  TCLUSTER = "cluster"
924
  TNODE = "node"
925
  TINSTANCE = "instance"
926

    
927
  ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
928
  EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
929
  EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
930
  EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
931
  EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
932
  EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
933
  EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE")
934
  ENODEDRBD = (TNODE, "ENODEDRBD")
935
  ENODEFILECHECK = (TNODE, "ENODEFILECHECK")
936
  ENODEHOOKS = (TNODE, "ENODEHOOKS")
937
  ENODEHV = (TNODE, "ENODEHV")
938
  ENODELVM = (TNODE, "ENODELVM")
939
  ENODEN1 = (TNODE, "ENODEN1")
940
  ENODENET = (TNODE, "ENODENET")
941
  ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE")
942
  ENODEORPHANLV = (TNODE, "ENODEORPHANLV")
943
  ENODERPC = (TNODE, "ENODERPC")
944
  ENODESSH = (TNODE, "ENODESSH")
945
  ENODEVERSION = (TNODE, "ENODEVERSION")
946
  ENODESETUP = (TNODE, "ENODESETUP")
947

    
948
  ETYPE_FIELD = "code"
949
  ETYPE_ERROR = "ERROR"
950
  ETYPE_WARNING = "WARNING"
951

    
952
  def ExpandNames(self):
953
    self.needed_locks = {
954
      locking.LEVEL_NODE: locking.ALL_SET,
955
      locking.LEVEL_INSTANCE: locking.ALL_SET,
956
    }
957
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
958

    
959
  def _Error(self, ecode, item, msg, *args, **kwargs):
960
    """Format an error message.
961

962
    Based on the opcode's error_codes parameter, either format a
963
    parseable error code, or a simpler error string.
964

965
    This must be called only from Exec and functions called from Exec.
966

967
    """
968
    ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
969
    itype, etxt = ecode
970
    # first complete the msg
971
    if args:
972
      msg = msg % args
973
    # then format the whole message
974
    if self.op.error_codes:
975
      msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
976
    else:
977
      if item:
978
        item = " " + item
979
      else:
980
        item = ""
981
      msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
982
    # and finally report it via the feedback_fn
983
    self._feedback_fn("  - %s" % msg)
984

    
985
  def _ErrorIf(self, cond, *args, **kwargs):
986
    """Log an error message if the passed condition is True.
987

988
    """
989
    cond = bool(cond) or self.op.debug_simulate_errors
990
    if cond:
991
      self._Error(*args, **kwargs)
992
    # do not mark the operation as failed for WARN cases only
993
    if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
994
      self.bad = self.bad or cond
995

    
996
  def _VerifyNode(self, nodeinfo, file_list, local_cksum,
997
                  node_result, master_files, drbd_map, vg_name):
998
    """Run multiple tests against a node.
999

1000
    Test list:
1001

1002
      - compares ganeti version
1003
      - checks vg existence and size > 20G
1004
      - checks config file checksum
1005
      - checks ssh to other nodes
1006

1007
    @type nodeinfo: L{objects.Node}
1008
    @param nodeinfo: the node to check
1009
    @param file_list: required list of files
1010
    @param local_cksum: dictionary of local files and their checksums
1011
    @param node_result: the results from the node
1012
    @param master_files: list of files that only masters should have
1013
    @param drbd_map: the useddrbd minors for this node, in
1014
        form of minor: (instance, must_exist) which correspond to instances
1015
        and their running status
1016
    @param vg_name: Ganeti Volume Group (result of self.cfg.GetVGName())
1017

1018
    """
1019
    node = nodeinfo.name
1020
    _ErrorIf = self._ErrorIf
1021

    
1022
    # main result, node_result should be a non-empty dict
1023
    test = not node_result or not isinstance(node_result, dict)
1024
    _ErrorIf(test, self.ENODERPC, node,
1025
                  "unable to verify node: no data returned")
1026
    if test:
1027
      return
1028

    
1029
    # compares ganeti version
1030
    local_version = constants.PROTOCOL_VERSION
1031
    remote_version = node_result.get('version', None)
1032
    test = not (remote_version and
1033
                isinstance(remote_version, (list, tuple)) and
1034
                len(remote_version) == 2)
1035
    _ErrorIf(test, self.ENODERPC, node,
1036
             "connection to node returned invalid data")
1037
    if test:
1038
      return
1039

    
1040
    test = local_version != remote_version[0]
1041
    _ErrorIf(test, self.ENODEVERSION, node,
1042
             "incompatible protocol versions: master %s,"
1043
             " node %s", local_version, remote_version[0])
1044
    if test:
1045
      return
1046

    
1047
    # node seems compatible, we can actually try to look into its results
1048

    
1049
    # full package version
1050
    self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
1051
                  self.ENODEVERSION, node,
1052
                  "software version mismatch: master %s, node %s",
1053
                  constants.RELEASE_VERSION, remote_version[1],
1054
                  code=self.ETYPE_WARNING)
1055

    
1056
    # checks vg existence and size > 20G
1057
    if vg_name is not None:
1058
      vglist = node_result.get(constants.NV_VGLIST, None)
1059
      test = not vglist
1060
      _ErrorIf(test, self.ENODELVM, node, "unable to check volume groups")
1061
      if not test:
1062
        vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
1063
                                              constants.MIN_VG_SIZE)
1064
        _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
1065

    
1066
    # checks config file checksum
1067

    
1068
    remote_cksum = node_result.get(constants.NV_FILELIST, None)
1069
    test = not isinstance(remote_cksum, dict)
1070
    _ErrorIf(test, self.ENODEFILECHECK, node,
1071
             "node hasn't returned file checksum data")
1072
    if not test:
1073
      for file_name in file_list:
1074
        node_is_mc = nodeinfo.master_candidate
1075
        must_have = (file_name not in master_files) or node_is_mc
1076
        # missing
1077
        test1 = file_name not in remote_cksum
1078
        # invalid checksum
1079
        test2 = not test1 and remote_cksum[file_name] != local_cksum[file_name]
1080
        # existing and good
1081
        test3 = not test1 and remote_cksum[file_name] == local_cksum[file_name]
1082
        _ErrorIf(test1 and must_have, self.ENODEFILECHECK, node,
1083
                 "file '%s' missing", file_name)
1084
        _ErrorIf(test2 and must_have, self.ENODEFILECHECK, node,
1085
                 "file '%s' has wrong checksum", file_name)
1086
        # not candidate and this is not a must-have file
1087
        _ErrorIf(test2 and not must_have, self.ENODEFILECHECK, node,
1088
                 "file '%s' should not exist on non master"
1089
                 " candidates (and the file is outdated)", file_name)
1090
        # all good, except non-master/non-must have combination
1091
        _ErrorIf(test3 and not must_have, self.ENODEFILECHECK, node,
1092
                 "file '%s' should not exist"
1093
                 " on non master candidates", file_name)
1094

    
1095
    # checks ssh to any
1096

    
1097
    test = constants.NV_NODELIST not in node_result
1098
    _ErrorIf(test, self.ENODESSH, node,
1099
             "node hasn't returned node ssh connectivity data")
1100
    if not test:
1101
      if node_result[constants.NV_NODELIST]:
1102
        for a_node, a_msg in node_result[constants.NV_NODELIST].items():
1103
          _ErrorIf(True, self.ENODESSH, node,
1104
                   "ssh communication with node '%s': %s", a_node, a_msg)
1105

    
1106
    test = constants.NV_NODENETTEST not in node_result
1107
    _ErrorIf(test, self.ENODENET, node,
1108
             "node hasn't returned node tcp connectivity data")
1109
    if not test:
1110
      if node_result[constants.NV_NODENETTEST]:
1111
        nlist = utils.NiceSort(node_result[constants.NV_NODENETTEST].keys())
1112
        for anode in nlist:
1113
          _ErrorIf(True, self.ENODENET, node,
1114
                   "tcp communication with node '%s': %s",
1115
                   anode, node_result[constants.NV_NODENETTEST][anode])
1116

    
1117
    hyp_result = node_result.get(constants.NV_HYPERVISOR, None)
1118
    if isinstance(hyp_result, dict):
1119
      for hv_name, hv_result in hyp_result.iteritems():
1120
        test = hv_result is not None
1121
        _ErrorIf(test, self.ENODEHV, node,
1122
                 "hypervisor %s verify failure: '%s'", hv_name, hv_result)
1123

    
1124
    # check used drbd list
1125
    if vg_name is not None:
1126
      used_minors = node_result.get(constants.NV_DRBDLIST, [])
1127
      test = not isinstance(used_minors, (tuple, list))
1128
      _ErrorIf(test, self.ENODEDRBD, node,
1129
               "cannot parse drbd status file: %s", str(used_minors))
1130
      if not test:
1131
        for minor, (iname, must_exist) in drbd_map.items():
1132
          test = minor not in used_minors and must_exist
1133
          _ErrorIf(test, self.ENODEDRBD, node,
1134
                   "drbd minor %d of instance %s is not active",
1135
                   minor, iname)
1136
        for minor in used_minors:
1137
          test = minor not in drbd_map
1138
          _ErrorIf(test, self.ENODEDRBD, node,
1139
                   "unallocated drbd minor %d is in use", minor)
1140
    test = node_result.get(constants.NV_NODESETUP,
1141
                           ["Missing NODESETUP results"])
1142
    _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s",
1143
             "; ".join(test))
1144

    
1145
    # check pv names
1146
    if vg_name is not None:
1147
      pvlist = node_result.get(constants.NV_PVLIST, None)
1148
      test = pvlist is None
1149
      _ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node")
1150
      if not test:
1151
        # check that ':' is not present in PV names, since it's a
1152
        # special character for lvcreate (denotes the range of PEs to
1153
        # use on the PV)
1154
        for size, pvname, owner_vg in pvlist:
1155
          test = ":" in pvname
1156
          _ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV"
1157
                   " '%s' of VG '%s'", pvname, owner_vg)
1158

    
1159
  def _VerifyInstance(self, instance, instanceconfig, node_vol_is,
1160
                      node_instance, n_offline):
1161
    """Verify an instance.
1162

1163
    This function checks to see if the required block devices are
1164
    available on the instance's node.
1165

1166
    """
1167
    _ErrorIf = self._ErrorIf
1168
    node_current = instanceconfig.primary_node
1169

    
1170
    node_vol_should = {}
1171
    instanceconfig.MapLVsByNode(node_vol_should)
1172

    
1173
    for node in node_vol_should:
1174
      if node in n_offline:
1175
        # ignore missing volumes on offline nodes
1176
        continue
1177
      for volume in node_vol_should[node]:
1178
        test = node not in node_vol_is or volume not in node_vol_is[node]
1179
        _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
1180
                 "volume %s missing on node %s", volume, node)
1181

    
1182
    if instanceconfig.admin_up:
1183
      test = ((node_current not in node_instance or
1184
               not instance in node_instance[node_current]) and
1185
              node_current not in n_offline)
1186
      _ErrorIf(test, self.EINSTANCEDOWN, instance,
1187
               "instance not running on its primary node %s",
1188
               node_current)
1189

    
1190
    for node in node_instance:
1191
      if (not node == node_current):
1192
        test = instance in node_instance[node]
1193
        _ErrorIf(test, self.EINSTANCEWRONGNODE, instance,
1194
                 "instance should not run on node %s", node)
1195

    
1196
  def _VerifyOrphanVolumes(self, node_vol_should, node_vol_is):
1197
    """Verify if there are any unknown volumes in the cluster.
1198

1199
    The .os, .swap and backup volumes are ignored. All other volumes are
1200
    reported as unknown.
1201

1202
    """
1203
    for node in node_vol_is:
1204
      for volume in node_vol_is[node]:
1205
        test = (node not in node_vol_should or
1206
                volume not in node_vol_should[node])
1207
        self._ErrorIf(test, self.ENODEORPHANLV, node,
1208
                      "volume %s is unknown", volume)
1209

    
1210
  def _VerifyOrphanInstances(self, instancelist, node_instance):
1211
    """Verify the list of running instances.
1212

1213
    This checks what instances are running but unknown to the cluster.
1214

1215
    """
1216
    for node in node_instance:
1217
      for o_inst in node_instance[node]:
1218
        test = o_inst not in instancelist
1219
        self._ErrorIf(test, self.ENODEORPHANINSTANCE, node,
1220
                      "instance %s on node %s should not exist", o_inst, node)
1221

    
1222
  def _VerifyNPlusOneMemory(self, node_info, instance_cfg):
1223
    """Verify N+1 Memory Resilience.
1224

1225
    Check that if one single node dies we can still start all the instances it
1226
    was primary for.
1227

1228
    """
1229
    for node, nodeinfo in node_info.iteritems():
1230
      # This code checks that every node which is now listed as secondary has
1231
      # enough memory to host all instances it is supposed to should a single
1232
      # other node in the cluster fail.
1233
      # FIXME: not ready for failover to an arbitrary node
1234
      # FIXME: does not support file-backed instances
1235
      # WARNING: we currently take into account down instances as well as up
1236
      # ones, considering that even if they're down someone might want to start
1237
      # them even in the event of a node failure.
1238
      for prinode, instances in nodeinfo['sinst-by-pnode'].iteritems():
1239
        needed_mem = 0
1240
        for instance in instances:
1241
          bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
1242
          if bep[constants.BE_AUTO_BALANCE]:
1243
            needed_mem += bep[constants.BE_MEMORY]
1244
        test = nodeinfo['mfree'] < needed_mem
1245
        self._ErrorIf(test, self.ENODEN1, node,
1246
                      "not enough memory on to accommodate"
1247
                      " failovers should peer node %s fail", prinode)
1248

    
1249
  def CheckPrereq(self):
1250
    """Check prerequisites.
1251

1252
    Transform the list of checks we're going to skip into a set and check that
1253
    all its members are valid.
1254

1255
    """
1256
    self.skip_set = frozenset(self.op.skip_checks)
1257
    if not constants.VERIFY_OPTIONAL_CHECKS.issuperset(self.skip_set):
1258
      raise errors.OpPrereqError("Invalid checks to be skipped specified",
1259
                                 errors.ECODE_INVAL)
1260

    
1261
  def BuildHooksEnv(self):
1262
    """Build hooks env.
1263

1264
    Cluster-Verify hooks just ran in the post phase and their failure makes
1265
    the output be logged in the verify output and the verification to fail.
1266

1267
    """
1268
    all_nodes = self.cfg.GetNodeList()
1269
    env = {
1270
      "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
1271
      }
1272
    for node in self.cfg.GetAllNodesInfo().values():
1273
      env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
1274

    
1275
    return env, [], all_nodes
1276

    
1277
  def Exec(self, feedback_fn):
1278
    """Verify integrity of cluster, performing various test on nodes.
1279

1280
    """
1281
    self.bad = False
1282
    _ErrorIf = self._ErrorIf
1283
    verbose = self.op.verbose
1284
    self._feedback_fn = feedback_fn
1285
    feedback_fn("* Verifying global settings")
1286
    for msg in self.cfg.VerifyConfig():
1287
      _ErrorIf(True, self.ECLUSTERCFG, None, msg)
1288

    
1289
    vg_name = self.cfg.GetVGName()
1290
    hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
1291
    nodelist = utils.NiceSort(self.cfg.GetNodeList())
1292
    nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
1293
    instancelist = utils.NiceSort(self.cfg.GetInstanceList())
1294
    instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
1295
                        for iname in instancelist)
1296
    i_non_redundant = [] # Non redundant instances
1297
    i_non_a_balanced = [] # Non auto-balanced instances
1298
    n_offline = [] # List of offline nodes
1299
    n_drained = [] # List of nodes being drained
1300
    node_volume = {}
1301
    node_instance = {}
1302
    node_info = {}
1303
    instance_cfg = {}
1304

    
1305
    # FIXME: verify OS list
1306
    # do local checksums
1307
    master_files = [constants.CLUSTER_CONF_FILE]
1308

    
1309
    file_names = ssconf.SimpleStore().GetFileList()
1310
    file_names.append(constants.SSL_CERT_FILE)
1311
    file_names.append(constants.RAPI_CERT_FILE)
1312
    file_names.extend(master_files)
1313

    
1314
    local_checksums = utils.FingerprintFiles(file_names)
1315

    
1316
    feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
1317
    node_verify_param = {
1318
      constants.NV_FILELIST: file_names,
1319
      constants.NV_NODELIST: [node.name for node in nodeinfo
1320
                              if not node.offline],
1321
      constants.NV_HYPERVISOR: hypervisors,
1322
      constants.NV_NODENETTEST: [(node.name, node.primary_ip,
1323
                                  node.secondary_ip) for node in nodeinfo
1324
                                 if not node.offline],
1325
      constants.NV_INSTANCELIST: hypervisors,
1326
      constants.NV_VERSION: None,
1327
      constants.NV_HVINFO: self.cfg.GetHypervisorType(),
1328
      constants.NV_NODESETUP: None,
1329
      }
1330
    if vg_name is not None:
1331
      node_verify_param[constants.NV_VGLIST] = None
1332
      node_verify_param[constants.NV_LVLIST] = vg_name
1333
      node_verify_param[constants.NV_PVLIST] = [vg_name]
1334
      node_verify_param[constants.NV_DRBDLIST] = None
1335
    all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
1336
                                           self.cfg.GetClusterName())
1337

    
1338
    cluster = self.cfg.GetClusterInfo()
1339
    master_node = self.cfg.GetMasterNode()
1340
    all_drbd_map = self.cfg.ComputeDRBDMap()
1341

    
1342
    feedback_fn("* Verifying node status")
1343
    for node_i in nodeinfo:
1344
      node = node_i.name
1345

    
1346
      if node_i.offline:
1347
        if verbose:
1348
          feedback_fn("* Skipping offline node %s" % (node,))
1349
        n_offline.append(node)
1350
        continue
1351

    
1352
      if node == master_node:
1353
        ntype = "master"
1354
      elif node_i.master_candidate:
1355
        ntype = "master candidate"
1356
      elif node_i.drained:
1357
        ntype = "drained"
1358
        n_drained.append(node)
1359
      else:
1360
        ntype = "regular"
1361
      if verbose:
1362
        feedback_fn("* Verifying node %s (%s)" % (node, ntype))
1363

    
1364
      msg = all_nvinfo[node].fail_msg
1365
      _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
1366
      if msg:
1367
        continue
1368

    
1369
      nresult = all_nvinfo[node].payload
1370
      node_drbd = {}
1371
      for minor, instance in all_drbd_map[node].items():
1372
        test = instance not in instanceinfo
1373
        _ErrorIf(test, self.ECLUSTERCFG, None,
1374
                 "ghost instance '%s' in temporary DRBD map", instance)
1375
          # ghost instance should not be running, but otherwise we
1376
          # don't give double warnings (both ghost instance and
1377
          # unallocated minor in use)
1378
        if test:
1379
          node_drbd[minor] = (instance, False)
1380
        else:
1381
          instance = instanceinfo[instance]
1382
          node_drbd[minor] = (instance.name, instance.admin_up)
1383
      self._VerifyNode(node_i, file_names, local_checksums,
1384
                       nresult, master_files, node_drbd, vg_name)
1385

    
1386
      lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
1387
      if vg_name is None:
1388
        node_volume[node] = {}
1389
      elif isinstance(lvdata, basestring):
1390
        _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
1391
                 utils.SafeEncode(lvdata))
1392
        node_volume[node] = {}
1393
      elif not isinstance(lvdata, dict):
1394
        _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
1395
        continue
1396
      else:
1397
        node_volume[node] = lvdata
1398

    
1399
      # node_instance
1400
      idata = nresult.get(constants.NV_INSTANCELIST, None)
1401
      test = not isinstance(idata, list)
1402
      _ErrorIf(test, self.ENODEHV, node,
1403
               "rpc call to node failed (instancelist)")
1404
      if test:
1405
        continue
1406

    
1407
      node_instance[node] = idata
1408

    
1409
      # node_info
1410
      nodeinfo = nresult.get(constants.NV_HVINFO, None)
1411
      test = not isinstance(nodeinfo, dict)
1412
      _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
1413
      if test:
1414
        continue
1415

    
1416
      try:
1417
        node_info[node] = {
1418
          "mfree": int(nodeinfo['memory_free']),
1419
          "pinst": [],
1420
          "sinst": [],
1421
          # dictionary holding all instances this node is secondary for,
1422
          # grouped by their primary node. Each key is a cluster node, and each
1423
          # value is a list of instances which have the key as primary and the
1424
          # current node as secondary.  this is handy to calculate N+1 memory
1425
          # availability if you can only failover from a primary to its
1426
          # secondary.
1427
          "sinst-by-pnode": {},
1428
        }
1429
        # FIXME: devise a free space model for file based instances as well
1430
        if vg_name is not None:
1431
          test = (constants.NV_VGLIST not in nresult or
1432
                  vg_name not in nresult[constants.NV_VGLIST])
1433
          _ErrorIf(test, self.ENODELVM, node,
1434
                   "node didn't return data for the volume group '%s'"
1435
                   " - it is either missing or broken", vg_name)
1436
          if test:
1437
            continue
1438
          node_info[node]["dfree"] = int(nresult[constants.NV_VGLIST][vg_name])
1439
      except (ValueError, KeyError):
1440
        _ErrorIf(True, self.ENODERPC, node,
1441
                 "node returned invalid nodeinfo, check lvm/hypervisor")
1442
        continue
1443

    
1444
    node_vol_should = {}
1445

    
1446
    feedback_fn("* Verifying instance status")
1447
    for instance in instancelist:
1448
      if verbose:
1449
        feedback_fn("* Verifying instance %s" % instance)
1450
      inst_config = instanceinfo[instance]
1451
      self._VerifyInstance(instance, inst_config, node_volume,
1452
                           node_instance, n_offline)
1453
      inst_nodes_offline = []
1454

    
1455
      inst_config.MapLVsByNode(node_vol_should)
1456

    
1457
      instance_cfg[instance] = inst_config
1458

    
1459
      pnode = inst_config.primary_node
1460
      _ErrorIf(pnode not in node_info and pnode not in n_offline,
1461
               self.ENODERPC, pnode, "instance %s, connection to"
1462
               " primary node failed", instance)
1463
      if pnode in node_info:
1464
        node_info[pnode]['pinst'].append(instance)
1465

    
1466
      if pnode in n_offline:
1467
        inst_nodes_offline.append(pnode)
1468

    
1469
      # If the instance is non-redundant we cannot survive losing its primary
1470
      # node, so we are not N+1 compliant. On the other hand we have no disk
1471
      # templates with more than one secondary so that situation is not well
1472
      # supported either.
1473
      # FIXME: does not support file-backed instances
1474
      if len(inst_config.secondary_nodes) == 0:
1475
        i_non_redundant.append(instance)
1476
      _ErrorIf(len(inst_config.secondary_nodes) > 1,
1477
               self.EINSTANCELAYOUT, instance,
1478
               "instance has multiple secondary nodes", code="WARNING")
1479

    
1480
      if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
1481
        i_non_a_balanced.append(instance)
1482

    
1483
      for snode in inst_config.secondary_nodes:
1484
        _ErrorIf(snode not in node_info and snode not in n_offline,
1485
                 self.ENODERPC, snode,
1486
                 "instance %s, connection to secondary node"
1487
                 "failed", instance)
1488

    
1489
        if snode in node_info:
1490
          node_info[snode]['sinst'].append(instance)
1491
          if pnode not in node_info[snode]['sinst-by-pnode']:
1492
            node_info[snode]['sinst-by-pnode'][pnode] = []
1493
          node_info[snode]['sinst-by-pnode'][pnode].append(instance)
1494

    
1495
        if snode in n_offline:
1496
          inst_nodes_offline.append(snode)
1497

    
1498
      # warn that the instance lives on offline nodes
1499
      _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
1500
               "instance lives on offline node(s) %s",
1501
               ", ".join(inst_nodes_offline))
1502

    
1503
    feedback_fn("* Verifying orphan volumes")
1504
    self._VerifyOrphanVolumes(node_vol_should, node_volume)
1505

    
1506
    feedback_fn("* Verifying remaining instances")
1507
    self._VerifyOrphanInstances(instancelist, node_instance)
1508

    
1509
    if constants.VERIFY_NPLUSONE_MEM not in self.skip_set:
1510
      feedback_fn("* Verifying N+1 Memory redundancy")
1511
      self._VerifyNPlusOneMemory(node_info, instance_cfg)
1512

    
1513
    feedback_fn("* Other Notes")
1514
    if i_non_redundant:
1515
      feedback_fn("  - NOTICE: %d non-redundant instance(s) found."
1516
                  % len(i_non_redundant))
1517

    
1518
    if i_non_a_balanced:
1519
      feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
1520
                  % len(i_non_a_balanced))
1521

    
1522
    if n_offline:
1523
      feedback_fn("  - NOTICE: %d offline node(s) found." % len(n_offline))
1524

    
1525
    if n_drained:
1526
      feedback_fn("  - NOTICE: %d drained node(s) found." % len(n_drained))
1527

    
1528
    return not self.bad
1529

    
1530
  def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
1531
    """Analyze the post-hooks' result
1532

1533
    This method analyses the hook result, handles it, and sends some
1534
    nicely-formatted feedback back to the user.
1535

1536
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
1537
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
1538
    @param hooks_results: the results of the multi-node hooks rpc call
1539
    @param feedback_fn: function used send feedback back to the caller
1540
    @param lu_result: previous Exec result
1541
    @return: the new Exec result, based on the previous result
1542
        and hook results
1543

1544
    """
1545
    # We only really run POST phase hooks, and are only interested in
1546
    # their results
1547
    if phase == constants.HOOKS_PHASE_POST:
1548
      # Used to change hooks' output to proper indentation
1549
      indent_re = re.compile('^', re.M)
1550
      feedback_fn("* Hooks Results")
1551
      assert hooks_results, "invalid result from hooks"
1552

    
1553
      for node_name in hooks_results:
1554
        show_node_header = True
1555
        res = hooks_results[node_name]
1556
        msg = res.fail_msg
1557
        test = msg and not res.offline
1558
        self._ErrorIf(test, self.ENODEHOOKS, node_name,
1559
                      "Communication failure in hooks execution: %s", msg)
1560
        if test:
1561
          # override manually lu_result here as _ErrorIf only
1562
          # overrides self.bad
1563
          lu_result = 1
1564
          continue
1565
        for script, hkr, output in res.payload:
1566
          test = hkr == constants.HKR_FAIL
1567
          self._ErrorIf(test, self.ENODEHOOKS, node_name,
1568
                        "Script %s failed, output:", script)
1569
          if test:
1570
            output = indent_re.sub('      ', output)
1571
            feedback_fn("%s" % output)
1572
            lu_result = 1
1573

    
1574
      return lu_result
1575

    
1576

    
1577
class LUVerifyDisks(NoHooksLU):
1578
  """Verifies the cluster disks status.
1579

1580
  """
1581
  _OP_REQP = []
1582
  REQ_BGL = False
1583

    
1584
  def ExpandNames(self):
1585
    self.needed_locks = {
1586
      locking.LEVEL_NODE: locking.ALL_SET,
1587
      locking.LEVEL_INSTANCE: locking.ALL_SET,
1588
    }
1589
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
1590

    
1591
  def CheckPrereq(self):
1592
    """Check prerequisites.
1593

1594
    This has no prerequisites.
1595

1596
    """
1597
    pass
1598

    
1599
  def Exec(self, feedback_fn):
1600
    """Verify integrity of cluster disks.
1601

1602
    @rtype: tuple of three items
1603
    @return: a tuple of (dict of node-to-node_error, list of instances
1604
        which need activate-disks, dict of instance: (node, volume) for
1605
        missing volumes
1606

1607
    """
1608
    result = res_nodes, res_instances, res_missing = {}, [], {}
1609

    
1610
    vg_name = self.cfg.GetVGName()
1611
    nodes = utils.NiceSort(self.cfg.GetNodeList())
1612
    instances = [self.cfg.GetInstanceInfo(name)
1613
                 for name in self.cfg.GetInstanceList()]
1614

    
1615
    nv_dict = {}
1616
    for inst in instances:
1617
      inst_lvs = {}
1618
      if (not inst.admin_up or
1619
          inst.disk_template not in constants.DTS_NET_MIRROR):
1620
        continue
1621
      inst.MapLVsByNode(inst_lvs)
1622
      # transform { iname: {node: [vol,],},} to {(node, vol): iname}
1623
      for node, vol_list in inst_lvs.iteritems():
1624
        for vol in vol_list:
1625
          nv_dict[(node, vol)] = inst
1626

    
1627
    if not nv_dict:
1628
      return result
1629

    
1630
    node_lvs = self.rpc.call_lv_list(nodes, vg_name)
1631

    
1632
    for node in nodes:
1633
      # node_volume
1634
      node_res = node_lvs[node]
1635
      if node_res.offline:
1636
        continue
1637
      msg = node_res.fail_msg
1638
      if msg:
1639
        logging.warning("Error enumerating LVs on node %s: %s", node, msg)
1640
        res_nodes[node] = msg
1641
        continue
1642

    
1643
      lvs = node_res.payload
1644
      for lv_name, (_, lv_inactive, lv_online) in lvs.items():
1645
        inst = nv_dict.pop((node, lv_name), None)
1646
        if (not lv_online and inst is not None
1647
            and inst.name not in res_instances):
1648
          res_instances.append(inst.name)
1649

    
1650
    # any leftover items in nv_dict are missing LVs, let's arrange the
1651
    # data better
1652
    for key, inst in nv_dict.iteritems():
1653
      if inst.name not in res_missing:
1654
        res_missing[inst.name] = []
1655
      res_missing[inst.name].append(key)
1656

    
1657
    return result
1658

    
1659

    
1660
class LURepairDiskSizes(NoHooksLU):
1661
  """Verifies the cluster disks sizes.
1662

1663
  """
1664
  _OP_REQP = ["instances"]
1665
  REQ_BGL = False
1666

    
1667
  def ExpandNames(self):
1668
    if not isinstance(self.op.instances, list):
1669
      raise errors.OpPrereqError("Invalid argument type 'instances'",
1670
                                 errors.ECODE_INVAL)
1671

    
1672
    if self.op.instances:
1673
      self.wanted_names = []
1674
      for name in self.op.instances:
1675
        full_name = self.cfg.ExpandInstanceName(name)
1676
        if full_name is None:
1677
          raise errors.OpPrereqError("Instance '%s' not known" % name,
1678
                                     errors.ECODE_NOENT)
1679
        self.wanted_names.append(full_name)
1680
      self.needed_locks = {
1681
        locking.LEVEL_NODE: [],
1682
        locking.LEVEL_INSTANCE: self.wanted_names,
1683
        }
1684
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
1685
    else:
1686
      self.wanted_names = None
1687
      self.needed_locks = {
1688
        locking.LEVEL_NODE: locking.ALL_SET,
1689
        locking.LEVEL_INSTANCE: locking.ALL_SET,
1690
        }
1691
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
1692

    
1693
  def DeclareLocks(self, level):
1694
    if level == locking.LEVEL_NODE and self.wanted_names is not None:
1695
      self._LockInstancesNodes(primary_only=True)
1696

    
1697
  def CheckPrereq(self):
1698
    """Check prerequisites.
1699

1700
    This only checks the optional instance list against the existing names.
1701

1702
    """
1703
    if self.wanted_names is None:
1704
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
1705

    
1706
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
1707
                             in self.wanted_names]
1708

    
1709
  def _EnsureChildSizes(self, disk):
1710
    """Ensure children of the disk have the needed disk size.
1711

1712
    This is valid mainly for DRBD8 and fixes an issue where the
1713
    children have smaller disk size.
1714

1715
    @param disk: an L{ganeti.objects.Disk} object
1716

1717
    """
1718
    if disk.dev_type == constants.LD_DRBD8:
1719
      assert disk.children, "Empty children for DRBD8?"
1720
      fchild = disk.children[0]
1721
      mismatch = fchild.size < disk.size
1722
      if mismatch:
1723
        self.LogInfo("Child disk has size %d, parent %d, fixing",
1724
                     fchild.size, disk.size)
1725
        fchild.size = disk.size
1726

    
1727
      # and we recurse on this child only, not on the metadev
1728
      return self._EnsureChildSizes(fchild) or mismatch
1729
    else:
1730
      return False
1731

    
1732
  def Exec(self, feedback_fn):
1733
    """Verify the size of cluster disks.
1734

1735
    """
1736
    # TODO: check child disks too
1737
    # TODO: check differences in size between primary/secondary nodes
1738
    per_node_disks = {}
1739
    for instance in self.wanted_instances:
1740
      pnode = instance.primary_node
1741
      if pnode not in per_node_disks:
1742
        per_node_disks[pnode] = []
1743
      for idx, disk in enumerate(instance.disks):
1744
        per_node_disks[pnode].append((instance, idx, disk))
1745

    
1746
    changed = []
1747
    for node, dskl in per_node_disks.items():
1748
      newl = [v[2].Copy() for v in dskl]
1749
      for dsk in newl:
1750
        self.cfg.SetDiskID(dsk, node)
1751
      result = self.rpc.call_blockdev_getsizes(node, newl)
1752
      if result.fail_msg:
1753
        self.LogWarning("Failure in blockdev_getsizes call to node"
1754
                        " %s, ignoring", node)
1755
        continue
1756
      if len(result.data) != len(dskl):
1757
        self.LogWarning("Invalid result from node %s, ignoring node results",
1758
                        node)
1759
        continue
1760
      for ((instance, idx, disk), size) in zip(dskl, result.data):
1761
        if size is None:
1762
          self.LogWarning("Disk %d of instance %s did not return size"
1763
                          " information, ignoring", idx, instance.name)
1764
          continue
1765
        if not isinstance(size, (int, long)):
1766
          self.LogWarning("Disk %d of instance %s did not return valid"
1767
                          " size information, ignoring", idx, instance.name)
1768
          continue
1769
        size = size >> 20
1770
        if size != disk.size:
1771
          self.LogInfo("Disk %d of instance %s has mismatched size,"
1772
                       " correcting: recorded %d, actual %d", idx,
1773
                       instance.name, disk.size, size)
1774
          disk.size = size
1775
          self.cfg.Update(instance, feedback_fn)
1776
          changed.append((instance.name, idx, size))
1777
        if self._EnsureChildSizes(disk):
1778
          self.cfg.Update(instance, feedback_fn)
1779
          changed.append((instance.name, idx, disk.size))
1780
    return changed
1781

    
1782

    
1783
class LURenameCluster(LogicalUnit):
1784
  """Rename the cluster.
1785

1786
  """
1787
  HPATH = "cluster-rename"
1788
  HTYPE = constants.HTYPE_CLUSTER
1789
  _OP_REQP = ["name"]
1790

    
1791
  def BuildHooksEnv(self):
1792
    """Build hooks env.
1793

1794
    """
1795
    env = {
1796
      "OP_TARGET": self.cfg.GetClusterName(),
1797
      "NEW_NAME": self.op.name,
1798
      }
1799
    mn = self.cfg.GetMasterNode()
1800
    return env, [mn], [mn]
1801

    
1802
  def CheckPrereq(self):
1803
    """Verify that the passed name is a valid one.
1804

1805
    """
1806
    hostname = utils.GetHostInfo(self.op.name)
1807

    
1808
    new_name = hostname.name
1809
    self.ip = new_ip = hostname.ip
1810
    old_name = self.cfg.GetClusterName()
1811
    old_ip = self.cfg.GetMasterIP()
1812
    if new_name == old_name and new_ip == old_ip:
1813
      raise errors.OpPrereqError("Neither the name nor the IP address of the"
1814
                                 " cluster has changed",
1815
                                 errors.ECODE_INVAL)
1816
    if new_ip != old_ip:
1817
      if utils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
1818
        raise errors.OpPrereqError("The given cluster IP address (%s) is"
1819
                                   " reachable on the network. Aborting." %
1820
                                   new_ip, errors.ECODE_NOTUNIQUE)
1821

    
1822
    self.op.name = new_name
1823

    
1824
  def Exec(self, feedback_fn):
1825
    """Rename the cluster.
1826

1827
    """
1828
    clustername = self.op.name
1829
    ip = self.ip
1830

    
1831
    # shutdown the master IP
1832
    master = self.cfg.GetMasterNode()
1833
    result = self.rpc.call_node_stop_master(master, False)
1834
    result.Raise("Could not disable the master role")
1835

    
1836
    try:
1837
      cluster = self.cfg.GetClusterInfo()
1838
      cluster.cluster_name = clustername
1839
      cluster.master_ip = ip
1840
      self.cfg.Update(cluster, feedback_fn)
1841

    
1842
      # update the known hosts file
1843
      ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
1844
      node_list = self.cfg.GetNodeList()
1845
      try:
1846
        node_list.remove(master)
1847
      except ValueError:
1848
        pass
1849
      result = self.rpc.call_upload_file(node_list,
1850
                                         constants.SSH_KNOWN_HOSTS_FILE)
1851
      for to_node, to_result in result.iteritems():
1852
        msg = to_result.fail_msg
1853
        if msg:
1854
          msg = ("Copy of file %s to node %s failed: %s" %
1855
                 (constants.SSH_KNOWN_HOSTS_FILE, to_node, msg))
1856
          self.proc.LogWarning(msg)
1857

    
1858
    finally:
1859
      result = self.rpc.call_node_start_master(master, False, False)
1860
      msg = result.fail_msg
1861
      if msg:
1862
        self.LogWarning("Could not re-enable the master role on"
1863
                        " the master, please restart manually: %s", msg)
1864

    
1865

    
1866
def _RecursiveCheckIfLVMBased(disk):
1867
  """Check if the given disk or its children are lvm-based.
1868

1869
  @type disk: L{objects.Disk}
1870
  @param disk: the disk to check
1871
  @rtype: boolean
1872
  @return: boolean indicating whether a LD_LV dev_type was found or not
1873

1874
  """
1875
  if disk.children:
1876
    for chdisk in disk.children:
1877
      if _RecursiveCheckIfLVMBased(chdisk):
1878
        return True
1879
  return disk.dev_type == constants.LD_LV
1880

    
1881

    
1882
class LUSetClusterParams(LogicalUnit):
1883
  """Change the parameters of the cluster.
1884

1885
  """
1886
  HPATH = "cluster-modify"
1887
  HTYPE = constants.HTYPE_CLUSTER
1888
  _OP_REQP = []
1889
  REQ_BGL = False
1890

    
1891
  def CheckArguments(self):
1892
    """Check parameters
1893

1894
    """
1895
    if not hasattr(self.op, "candidate_pool_size"):
1896
      self.op.candidate_pool_size = None
1897
    if self.op.candidate_pool_size is not None:
1898
      try:
1899
        self.op.candidate_pool_size = int(self.op.candidate_pool_size)
1900
      except (ValueError, TypeError), err:
1901
        raise errors.OpPrereqError("Invalid candidate_pool_size value: %s" %
1902
                                   str(err), errors.ECODE_INVAL)
1903
      if self.op.candidate_pool_size < 1:
1904
        raise errors.OpPrereqError("At least one master candidate needed",
1905
                                   errors.ECODE_INVAL)
1906

    
1907
  def ExpandNames(self):
1908
    # FIXME: in the future maybe other cluster params won't require checking on
1909
    # all nodes to be modified.
1910
    self.needed_locks = {
1911
      locking.LEVEL_NODE: locking.ALL_SET,
1912
    }
1913
    self.share_locks[locking.LEVEL_NODE] = 1
1914

    
1915
  def BuildHooksEnv(self):
1916
    """Build hooks env.
1917

1918
    """
1919
    env = {
1920
      "OP_TARGET": self.cfg.GetClusterName(),
1921
      "NEW_VG_NAME": self.op.vg_name,
1922
      }
1923
    mn = self.cfg.GetMasterNode()
1924
    return env, [mn], [mn]
1925

    
1926
  def CheckPrereq(self):
1927
    """Check prerequisites.
1928

1929
    This checks whether the given params don't conflict and
1930
    if the given volume group is valid.
1931

1932
    """
1933
    if self.op.vg_name is not None and not self.op.vg_name:
1934
      instances = self.cfg.GetAllInstancesInfo().values()
1935
      for inst in instances:
1936
        for disk in inst.disks:
1937
          if _RecursiveCheckIfLVMBased(disk):
1938
            raise errors.OpPrereqError("Cannot disable lvm storage while"
1939
                                       " lvm-based instances exist",
1940
                                       errors.ECODE_INVAL)
1941

    
1942
    node_list = self.acquired_locks[locking.LEVEL_NODE]
1943

    
1944
    # if vg_name not None, checks given volume group on all nodes
1945
    if self.op.vg_name:
1946
      vglist = self.rpc.call_vg_list(node_list)
1947
      for node in node_list:
1948
        msg = vglist[node].fail_msg
1949
        if msg:
1950
          # ignoring down node
1951
          self.LogWarning("Error while gathering data on node %s"
1952
                          " (ignoring node): %s", node, msg)
1953
          continue
1954
        vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
1955
                                              self.op.vg_name,
1956
                                              constants.MIN_VG_SIZE)
1957
        if vgstatus:
1958
          raise errors.OpPrereqError("Error on node '%s': %s" %
1959
                                     (node, vgstatus), errors.ECODE_ENVIRON)
1960

    
1961
    self.cluster = cluster = self.cfg.GetClusterInfo()
1962
    # validate params changes
1963
    if self.op.beparams:
1964
      utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
1965
      self.new_beparams = objects.FillDict(
1966
        cluster.beparams[constants.PP_DEFAULT], self.op.beparams)
1967

    
1968
    if self.op.nicparams:
1969
      utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
1970
      self.new_nicparams = objects.FillDict(
1971
        cluster.nicparams[constants.PP_DEFAULT], self.op.nicparams)
1972
      objects.NIC.CheckParameterSyntax(self.new_nicparams)
1973

    
1974
    # hypervisor list/parameters
1975
    self.new_hvparams = objects.FillDict(cluster.hvparams, {})
1976
    if self.op.hvparams:
1977
      if not isinstance(self.op.hvparams, dict):
1978
        raise errors.OpPrereqError("Invalid 'hvparams' parameter on input",
1979
                                   errors.ECODE_INVAL)
1980
      for hv_name, hv_dict in self.op.hvparams.items():
1981
        if hv_name not in self.new_hvparams:
1982
          self.new_hvparams[hv_name] = hv_dict
1983
        else:
1984
          self.new_hvparams[hv_name].update(hv_dict)
1985

    
1986
    if self.op.enabled_hypervisors is not None:
1987
      self.hv_list = self.op.enabled_hypervisors
1988
      if not self.hv_list:
1989
        raise errors.OpPrereqError("Enabled hypervisors list must contain at"
1990
                                   " least one member",
1991
                                   errors.ECODE_INVAL)
1992
      invalid_hvs = set(self.hv_list) - constants.HYPER_TYPES
1993
      if invalid_hvs:
1994
        raise errors.OpPrereqError("Enabled hypervisors contains invalid"
1995
                                   " entries: %s" % " ,".join(invalid_hvs),
1996
                                   errors.ECODE_INVAL)
1997
    else:
1998
      self.hv_list = cluster.enabled_hypervisors
1999

    
2000
    if self.op.hvparams or self.op.enabled_hypervisors is not None:
2001
      # either the enabled list has changed, or the parameters have, validate
2002
      for hv_name, hv_params in self.new_hvparams.items():
2003
        if ((self.op.hvparams and hv_name in self.op.hvparams) or
2004
            (self.op.enabled_hypervisors and
2005
             hv_name in self.op.enabled_hypervisors)):
2006
          # either this is a new hypervisor, or its parameters have changed
2007
          hv_class = hypervisor.GetHypervisor(hv_name)
2008
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2009
          hv_class.CheckParameterSyntax(hv_params)
2010
          _CheckHVParams(self, node_list, hv_name, hv_params)
2011

    
2012
  def Exec(self, feedback_fn):
2013
    """Change the parameters of the cluster.
2014

2015
    """
2016
    if self.op.vg_name is not None:
2017
      new_volume = self.op.vg_name
2018
      if not new_volume:
2019
        new_volume = None
2020
      if new_volume != self.cfg.GetVGName():
2021
        self.cfg.SetVGName(new_volume)
2022
      else:
2023
        feedback_fn("Cluster LVM configuration already in desired"
2024
                    " state, not changing")
2025
    if self.op.hvparams:
2026
      self.cluster.hvparams = self.new_hvparams
2027
    if self.op.enabled_hypervisors is not None:
2028
      self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
2029
    if self.op.beparams:
2030
      self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
2031
    if self.op.nicparams:
2032
      self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
2033

    
2034
    if self.op.candidate_pool_size is not None:
2035
      self.cluster.candidate_pool_size = self.op.candidate_pool_size
2036
      # we need to update the pool size here, otherwise the save will fail
2037
      _AdjustCandidatePool(self, [])
2038

    
2039
    self.cfg.Update(self.cluster, feedback_fn)
2040

    
2041

    
2042
def _RedistributeAncillaryFiles(lu, additional_nodes=None):
2043
  """Distribute additional files which are part of the cluster configuration.
2044

2045
  ConfigWriter takes care of distributing the config and ssconf files, but
2046
  there are more files which should be distributed to all nodes. This function
2047
  makes sure those are copied.
2048

2049
  @param lu: calling logical unit
2050
  @param additional_nodes: list of nodes not in the config to distribute to
2051

2052
  """
2053
  # 1. Gather target nodes
2054
  myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
2055
  dist_nodes = lu.cfg.GetNodeList()
2056
  if additional_nodes is not None:
2057
    dist_nodes.extend(additional_nodes)
2058
  if myself.name in dist_nodes:
2059
    dist_nodes.remove(myself.name)
2060

    
2061
  # 2. Gather files to distribute
2062
  dist_files = set([constants.ETC_HOSTS,
2063
                    constants.SSH_KNOWN_HOSTS_FILE,
2064
                    constants.RAPI_CERT_FILE,
2065
                    constants.RAPI_USERS_FILE,
2066
                    constants.HMAC_CLUSTER_KEY,
2067
                   ])
2068

    
2069
  enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
2070
  for hv_name in enabled_hypervisors:
2071
    hv_class = hypervisor.GetHypervisor(hv_name)
2072
    dist_files.update(hv_class.GetAncillaryFiles())
2073

    
2074
  # 3. Perform the files upload
2075
  for fname in dist_files:
2076
    if os.path.exists(fname):
2077
      result = lu.rpc.call_upload_file(dist_nodes, fname)
2078
      for to_node, to_result in result.items():
2079
        msg = to_result.fail_msg
2080
        if msg:
2081
          msg = ("Copy of file %s to node %s failed: %s" %
2082
                 (fname, to_node, msg))
2083
          lu.proc.LogWarning(msg)
2084

    
2085

    
2086
class LURedistributeConfig(NoHooksLU):
2087
  """Force the redistribution of cluster configuration.
2088

2089
  This is a very simple LU.
2090

2091
  """
2092
  _OP_REQP = []
2093
  REQ_BGL = False
2094

    
2095
  def ExpandNames(self):
2096
    self.needed_locks = {
2097
      locking.LEVEL_NODE: locking.ALL_SET,
2098
    }
2099
    self.share_locks[locking.LEVEL_NODE] = 1
2100

    
2101
  def CheckPrereq(self):
2102
    """Check prerequisites.
2103

2104
    """
2105

    
2106
  def Exec(self, feedback_fn):
2107
    """Redistribute the configuration.
2108

2109
    """
2110
    self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
2111
    _RedistributeAncillaryFiles(self)
2112

    
2113

    
2114
def _WaitForSync(lu, instance, oneshot=False, unlock=False):
2115
  """Sleep and poll for an instance's disk to sync.
2116

2117
  """
2118
  if not instance.disks:
2119
    return True
2120

    
2121
  if not oneshot:
2122
    lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
2123

    
2124
  node = instance.primary_node
2125

    
2126
  for dev in instance.disks:
2127
    lu.cfg.SetDiskID(dev, node)
2128

    
2129
  # TODO: Convert to utils.Retry
2130

    
2131
  retries = 0
2132
  degr_retries = 10 # in seconds, as we sleep 1 second each time
2133
  while True:
2134
    max_time = 0
2135
    done = True
2136
    cumul_degraded = False
2137
    rstats = lu.rpc.call_blockdev_getmirrorstatus(node, instance.disks)
2138
    msg = rstats.fail_msg
2139
    if msg:
2140
      lu.LogWarning("Can't get any data from node %s: %s", node, msg)
2141
      retries += 1
2142
      if retries >= 10:
2143
        raise errors.RemoteError("Can't contact node %s for mirror data,"
2144
                                 " aborting." % node)
2145
      time.sleep(6)
2146
      continue
2147
    rstats = rstats.payload
2148
    retries = 0
2149
    for i, mstat in enumerate(rstats):
2150
      if mstat is None:
2151
        lu.LogWarning("Can't compute data for node %s/%s",
2152
                           node, instance.disks[i].iv_name)
2153
        continue
2154

    
2155
      cumul_degraded = (cumul_degraded or
2156
                        (mstat.is_degraded and mstat.sync_percent is None))
2157
      if mstat.sync_percent is not None:
2158
        done = False
2159
        if mstat.estimated_time is not None:
2160
          rem_time = "%d estimated seconds remaining" % mstat.estimated_time
2161
          max_time = mstat.estimated_time
2162
        else:
2163
          rem_time = "no time estimate"
2164
        lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
2165
                        (instance.disks[i].iv_name, mstat.sync_percent,
2166
                         rem_time))
2167

    
2168
    # if we're done but degraded, let's do a few small retries, to
2169
    # make sure we see a stable and not transient situation; therefore
2170
    # we force restart of the loop
2171
    if (done or oneshot) and cumul_degraded and degr_retries > 0:
2172
      logging.info("Degraded disks found, %d retries left", degr_retries)
2173
      degr_retries -= 1
2174
      time.sleep(1)
2175
      continue
2176

    
2177
    if done or oneshot:
2178
      break
2179

    
2180
    time.sleep(min(60, max_time))
2181

    
2182
  if done:
2183
    lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
2184
  return not cumul_degraded
2185

    
2186

    
2187
def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
2188
  """Check that mirrors are not degraded.
2189

2190
  The ldisk parameter, if True, will change the test from the
2191
  is_degraded attribute (which represents overall non-ok status for
2192
  the device(s)) to the ldisk (representing the local storage status).
2193

2194
  """
2195
  lu.cfg.SetDiskID(dev, node)
2196

    
2197
  result = True
2198

    
2199
  if on_primary or dev.AssembleOnSecondary():
2200
    rstats = lu.rpc.call_blockdev_find(node, dev)
2201
    msg = rstats.fail_msg
2202
    if msg:
2203
      lu.LogWarning("Can't find disk on node %s: %s", node, msg)
2204
      result = False
2205
    elif not rstats.payload:
2206
      lu.LogWarning("Can't find disk on node %s", node)
2207
      result = False
2208
    else:
2209
      if ldisk:
2210
        result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
2211
      else:
2212
        result = result and not rstats.payload.is_degraded
2213

    
2214
  if dev.children:
2215
    for child in dev.children:
2216
      result = result and _CheckDiskConsistency(lu, child, node, on_primary)
2217

    
2218
  return result
2219

    
2220

    
2221
class LUDiagnoseOS(NoHooksLU):
2222
  """Logical unit for OS diagnose/query.
2223

2224
  """
2225
  _OP_REQP = ["output_fields", "names"]
2226
  REQ_BGL = False
2227
  _FIELDS_STATIC = utils.FieldSet()
2228
  _FIELDS_DYNAMIC = utils.FieldSet("name", "valid", "node_status", "variants")
2229
  # Fields that need calculation of global os validity
2230
  _FIELDS_NEEDVALID = frozenset(["valid", "variants"])
2231

    
2232
  def ExpandNames(self):
2233
    if self.op.names:
2234
      raise errors.OpPrereqError("Selective OS query not supported",
2235
                                 errors.ECODE_INVAL)
2236

    
2237
    _CheckOutputFields(static=self._FIELDS_STATIC,
2238
                       dynamic=self._FIELDS_DYNAMIC,
2239
                       selected=self.op.output_fields)
2240

    
2241
    # Lock all nodes, in shared mode
2242
    # Temporary removal of locks, should be reverted later
2243
    # TODO: reintroduce locks when they are lighter-weight
2244
    self.needed_locks = {}
2245
    #self.share_locks[locking.LEVEL_NODE] = 1
2246
    #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
2247

    
2248
  def CheckPrereq(self):
2249
    """Check prerequisites.
2250

2251
    """
2252

    
2253
  @staticmethod
2254
  def _DiagnoseByOS(node_list, rlist):
2255
    """Remaps a per-node return list into an a per-os per-node dictionary
2256

2257
    @param node_list: a list with the names of all nodes
2258
    @param rlist: a map with node names as keys and OS objects as values
2259

2260
    @rtype: dict
2261
    @return: a dictionary with osnames as keys and as value another map, with
2262
        nodes as keys and tuples of (path, status, diagnose) as values, eg::
2263

2264
          {"debian-etch": {"node1": [(/usr/lib/..., True, ""),
2265
                                     (/srv/..., False, "invalid api")],
2266
                           "node2": [(/srv/..., True, "")]}
2267
          }
2268

2269
    """
2270
    all_os = {}
2271
    # we build here the list of nodes that didn't fail the RPC (at RPC
2272
    # level), so that nodes with a non-responding node daemon don't
2273
    # make all OSes invalid
2274
    good_nodes = [node_name for node_name in rlist
2275
                  if not rlist[node_name].fail_msg]
2276
    for node_name, nr in rlist.items():
2277
      if nr.fail_msg or not nr.payload:
2278
        continue
2279
      for name, path, status, diagnose, variants in nr.payload:
2280
        if name not in all_os:
2281
          # build a list of nodes for this os containing empty lists
2282
          # for each node in node_list
2283
          all_os[name] = {}
2284
          for nname in good_nodes:
2285
            all_os[name][nname] = []
2286
        all_os[name][node_name].append((path, status, diagnose, variants))
2287
    return all_os
2288

    
2289
  def Exec(self, feedback_fn):
2290
    """Compute the list of OSes.
2291

2292
    """
2293
    valid_nodes = [node for node in self.cfg.GetOnlineNodeList()]
2294
    node_data = self.rpc.call_os_diagnose(valid_nodes)
2295
    pol = self._DiagnoseByOS(valid_nodes, node_data)
2296
    output = []
2297
    calc_valid = self._FIELDS_NEEDVALID.intersection(self.op.output_fields)
2298
    calc_variants = "variants" in self.op.output_fields
2299

    
2300
    for os_name, os_data in pol.items():
2301
      row = []
2302
      if calc_valid:
2303
        valid = True
2304
        variants = None
2305
        for osl in os_data.values():
2306
          valid = valid and osl and osl[0][1]
2307
          if not valid:
2308
            variants = None
2309
            break
2310
          if calc_variants:
2311
            node_variants = osl[0][3]
2312
            if variants is None:
2313
              variants = node_variants
2314
            else:
2315
              variants = [v for v in variants if v in node_variants]
2316

    
2317
      for field in self.op.output_fields:
2318
        if field == "name":
2319
          val = os_name
2320
        elif field == "valid":
2321
          val = valid
2322
        elif field == "node_status":
2323
          # this is just a copy of the dict
2324
          val = {}
2325
          for node_name, nos_list in os_data.items():
2326
            val[node_name] = nos_list
2327
        elif field == "variants":
2328
          val =  variants
2329
        else:
2330
          raise errors.ParameterError(field)
2331
        row.append(val)
2332
      output.append(row)
2333

    
2334
    return output
2335

    
2336

    
2337
class LURemoveNode(LogicalUnit):
2338
  """Logical unit for removing a node.
2339

2340
  """
2341
  HPATH = "node-remove"
2342
  HTYPE = constants.HTYPE_NODE
2343
  _OP_REQP = ["node_name"]
2344

    
2345
  def BuildHooksEnv(self):
2346
    """Build hooks env.
2347

2348
    This doesn't run on the target node in the pre phase as a failed
2349
    node would then be impossible to remove.
2350

2351
    """
2352
    env = {
2353
      "OP_TARGET": self.op.node_name,
2354
      "NODE_NAME": self.op.node_name,
2355
      }
2356
    all_nodes = self.cfg.GetNodeList()
2357
    if self.op.node_name in all_nodes:
2358
      all_nodes.remove(self.op.node_name)
2359
    return env, all_nodes, all_nodes
2360

    
2361
  def CheckPrereq(self):
2362
    """Check prerequisites.
2363

2364
    This checks:
2365
     - the node exists in the configuration
2366
     - it does not have primary or secondary instances
2367
     - it's not the master
2368

2369
    Any errors are signaled by raising errors.OpPrereqError.
2370

2371
    """
2372
    node = self.cfg.GetNodeInfo(self.cfg.ExpandNodeName(self.op.node_name))
2373
    if node is None:
2374
      raise errors.OpPrereqError("Node '%s' is unknown." % self.op.node_name,
2375
                                 errors.ECODE_NOENT)
2376

    
2377
    instance_list = self.cfg.GetInstanceList()
2378

    
2379
    masternode = self.cfg.GetMasterNode()
2380
    if node.name == masternode:
2381
      raise errors.OpPrereqError("Node is the master node,"
2382
                                 " you need to failover first.",
2383
                                 errors.ECODE_INVAL)
2384

    
2385
    for instance_name in instance_list:
2386
      instance = self.cfg.GetInstanceInfo(instance_name)
2387
      if node.name in instance.all_nodes:
2388
        raise errors.OpPrereqError("Instance %s is still running on the node,"
2389
                                   " please remove first." % instance_name,
2390
                                   errors.ECODE_INVAL)
2391
    self.op.node_name = node.name
2392
    self.node = node
2393

    
2394
  def Exec(self, feedback_fn):
2395
    """Removes the node from the cluster.
2396

2397
    """
2398
    node = self.node
2399
    logging.info("Stopping the node daemon and removing configs from node %s",
2400
                 node.name)
2401

    
2402
    modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
2403

    
2404
    # Promote nodes to master candidate as needed
2405
    _AdjustCandidatePool(self, exceptions=[node.name])
2406
    self.context.RemoveNode(node.name)
2407

    
2408
    # Run post hooks on the node before it's removed
2409
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
2410
    try:
2411
      h_results = hm.RunPhase(constants.HOOKS_PHASE_POST, [node.name])
2412
    except:
2413
      self.LogWarning("Errors occurred running hooks on %s" % node.name)
2414

    
2415
    result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
2416
    msg = result.fail_msg
2417
    if msg:
2418
      self.LogWarning("Errors encountered on the remote node while leaving"
2419
                      " the cluster: %s", msg)
2420

    
2421

    
2422
class LUQueryNodes(NoHooksLU):
2423
  """Logical unit for querying nodes.
2424

2425
  """
2426
  _OP_REQP = ["output_fields", "names", "use_locking"]
2427
  REQ_BGL = False
2428

    
2429
  _SIMPLE_FIELDS = ["name", "serial_no", "ctime", "mtime", "uuid",
2430
                    "master_candidate", "offline", "drained"]
2431

    
2432
  _FIELDS_DYNAMIC = utils.FieldSet(
2433
    "dtotal", "dfree",
2434
    "mtotal", "mnode", "mfree",
2435
    "bootid",
2436
    "ctotal", "cnodes", "csockets",
2437
    )
2438

    
2439
  _FIELDS_STATIC = utils.FieldSet(*[
2440
    "pinst_cnt", "sinst_cnt",
2441
    "pinst_list", "sinst_list",
2442
    "pip", "sip", "tags",
2443
    "master",
2444
    "role"] + _SIMPLE_FIELDS
2445
    )
2446

    
2447
  def ExpandNames(self):
2448
    _CheckOutputFields(static=self._FIELDS_STATIC,
2449
                       dynamic=self._FIELDS_DYNAMIC,
2450
                       selected=self.op.output_fields)
2451

    
2452
    self.needed_locks = {}
2453
    self.share_locks[locking.LEVEL_NODE] = 1
2454

    
2455
    if self.op.names:
2456
      self.wanted = _GetWantedNodes(self, self.op.names)
2457
    else:
2458
      self.wanted = locking.ALL_SET
2459

    
2460
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
2461
    self.do_locking = self.do_node_query and self.op.use_locking
2462
    if self.do_locking:
2463
      # if we don't request only static fields, we need to lock the nodes
2464
      self.needed_locks[locking.LEVEL_NODE] = self.wanted
2465

    
2466
  def CheckPrereq(self):
2467
    """Check prerequisites.
2468

2469
    """
2470
    # The validation of the node list is done in the _GetWantedNodes,
2471
    # if non empty, and if empty, there's no validation to do
2472
    pass
2473

    
2474
  def Exec(self, feedback_fn):
2475
    """Computes the list of nodes and their attributes.
2476

2477
    """
2478
    all_info = self.cfg.GetAllNodesInfo()
2479
    if self.do_locking:
2480
      nodenames = self.acquired_locks[locking.LEVEL_NODE]
2481
    elif self.wanted != locking.ALL_SET:
2482
      nodenames = self.wanted
2483
      missing = set(nodenames).difference(all_info.keys())
2484
      if missing:
2485
        raise errors.OpExecError(
2486
          "Some nodes were removed before retrieving their data: %s" % missing)
2487
    else:
2488
      nodenames = all_info.keys()
2489

    
2490
    nodenames = utils.NiceSort(nodenames)
2491
    nodelist = [all_info[name] for name in nodenames]
2492

    
2493
    # begin data gathering
2494

    
2495
    if self.do_node_query:
2496
      live_data = {}
2497
      node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
2498
                                          self.cfg.GetHypervisorType())
2499
      for name in nodenames:
2500
        nodeinfo = node_data[name]
2501
        if not nodeinfo.fail_msg and nodeinfo.payload:
2502
          nodeinfo = nodeinfo.payload
2503
          fn = utils.TryConvert
2504
          live_data[name] = {
2505
            "mtotal": fn(int, nodeinfo.get('memory_total', None)),
2506
            "mnode": fn(int, nodeinfo.get('memory_dom0', None)),
2507
            "mfree": fn(int, nodeinfo.get('memory_free', None)),
2508
            "dtotal": fn(int, nodeinfo.get('vg_size', None)),
2509
            "dfree": fn(int, nodeinfo.get('vg_free', None)),
2510
            "ctotal": fn(int, nodeinfo.get('cpu_total', None)),
2511
            "bootid": nodeinfo.get('bootid', None),
2512
            "cnodes": fn(int, nodeinfo.get('cpu_nodes', None)),
2513
            "csockets": fn(int, nodeinfo.get('cpu_sockets', None)),
2514
            }
2515
        else:
2516
          live_data[name] = {}
2517
    else:
2518
      live_data = dict.fromkeys(nodenames, {})
2519

    
2520
    node_to_primary = dict([(name, set()) for name in nodenames])
2521
    node_to_secondary = dict([(name, set()) for name in nodenames])
2522

    
2523
    inst_fields = frozenset(("pinst_cnt", "pinst_list",
2524
                             "sinst_cnt", "sinst_list"))
2525
    if inst_fields & frozenset(self.op.output_fields):
2526
      instancelist = self.cfg.GetInstanceList()
2527

    
2528
      for instance_name in instancelist:
2529
        inst = self.cfg.GetInstanceInfo(instance_name)
2530
        if inst.primary_node in node_to_primary:
2531
          node_to_primary[inst.primary_node].add(inst.name)
2532
        for secnode in inst.secondary_nodes:
2533
          if secnode in node_to_secondary:
2534
            node_to_secondary[secnode].add(inst.name)
2535

    
2536
    master_node = self.cfg.GetMasterNode()
2537

    
2538
    # end data gathering
2539

    
2540
    output = []
2541
    for node in nodelist:
2542
      node_output = []
2543
      for field in self.op.output_fields:
2544
        if field in self._SIMPLE_FIELDS:
2545
          val = getattr(node, field)
2546
        elif field == "pinst_list":
2547
          val = list(node_to_primary[node.name])
2548
        elif field == "sinst_list":
2549
          val = list(node_to_secondary[node.name])
2550
        elif field == "pinst_cnt":
2551
          val = len(node_to_primary[node.name])
2552
        elif field == "sinst_cnt":
2553
          val = len(node_to_secondary[node.name])
2554
        elif field == "pip":
2555
          val = node.primary_ip
2556
        elif field == "sip":
2557
          val = node.secondary_ip
2558
        elif field == "tags":
2559
          val = list(node.GetTags())
2560
        elif field == "master":
2561
          val = node.name == master_node
2562
        elif self._FIELDS_DYNAMIC.Matches(field):
2563
          val = live_data[node.name].get(field, None)
2564
        elif field == "role":
2565
          if node.name == master_node:
2566
            val = "M"
2567
          elif node.master_candidate:
2568
            val = "C"
2569
          elif node.drained:
2570
            val = "D"
2571
          elif node.offline:
2572
            val = "O"
2573
          else:
2574
            val = "R"
2575
        else:
2576
          raise errors.ParameterError(field)
2577
        node_output.append(val)
2578
      output.append(node_output)
2579

    
2580
    return output
2581

    
2582

    
2583
class LUQueryNodeVolumes(NoHooksLU):
2584
  """Logical unit for getting volumes on node(s).
2585

2586
  """
2587
  _OP_REQP = ["nodes", "output_fields"]
2588
  REQ_BGL = False
2589
  _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
2590
  _FIELDS_STATIC = utils.FieldSet("node")
2591

    
2592
  def ExpandNames(self):
2593
    _CheckOutputFields(static=self._FIELDS_STATIC,
2594
                       dynamic=self._FIELDS_DYNAMIC,
2595
                       selected=self.op.output_fields)
2596

    
2597
    self.needed_locks = {}
2598
    self.share_locks[locking.LEVEL_NODE] = 1
2599
    if not self.op.nodes:
2600
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
2601
    else:
2602
      self.needed_locks[locking.LEVEL_NODE] = \
2603
        _GetWantedNodes(self, self.op.nodes)
2604

    
2605
  def CheckPrereq(self):
2606
    """Check prerequisites.
2607

2608
    This checks that the fields required are valid output fields.
2609

2610
    """
2611
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
2612

    
2613
  def Exec(self, feedback_fn):
2614
    """Computes the list of nodes and their attributes.
2615

2616
    """
2617
    nodenames = self.nodes
2618
    volumes = self.rpc.call_node_volumes(nodenames)
2619

    
2620
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
2621
             in self.cfg.GetInstanceList()]
2622

    
2623
    lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
2624

    
2625
    output = []
2626
    for node in nodenames:
2627
      nresult = volumes[node]
2628
      if nresult.offline:
2629
        continue
2630
      msg = nresult.fail_msg
2631
      if msg:
2632
        self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
2633
        continue
2634

    
2635
      node_vols = nresult.payload[:]
2636
      node_vols.sort(key=lambda vol: vol['dev'])
2637

    
2638
      for vol in node_vols:
2639
        node_output = []
2640
        for field in self.op.output_fields:
2641
          if field == "node":
2642
            val = node
2643
          elif field == "phys":
2644
            val = vol['dev']
2645
          elif field == "vg":
2646
            val = vol['vg']
2647
          elif field == "name":
2648
            val = vol['name']
2649
          elif field == "size":
2650
            val = int(float(vol['size']))
2651
          elif field == "instance":
2652
            for inst in ilist:
2653
              if node not in lv_by_node[inst]:
2654
                continue
2655
              if vol['name'] in lv_by_node[inst][node]:
2656
                val = inst.name
2657
                break
2658
            else:
2659
              val = '-'
2660
          else:
2661
            raise errors.ParameterError(field)
2662
          node_output.append(str(val))
2663

    
2664
        output.append(node_output)
2665

    
2666
    return output
2667

    
2668

    
2669
class LUQueryNodeStorage(NoHooksLU):
2670
  """Logical unit for getting information on storage units on node(s).
2671

2672
  """
2673
  _OP_REQP = ["nodes", "storage_type", "output_fields"]
2674
  REQ_BGL = False
2675
  _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
2676

    
2677
  def ExpandNames(self):
2678
    storage_type = self.op.storage_type
2679

    
2680
    if storage_type not in constants.VALID_STORAGE_TYPES:
2681
      raise errors.OpPrereqError("Unknown storage type: %s" % storage_type,
2682
                                 errors.ECODE_INVAL)
2683

    
2684
    _CheckOutputFields(static=self._FIELDS_STATIC,
2685
                       dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
2686
                       selected=self.op.output_fields)
2687

    
2688
    self.needed_locks = {}
2689
    self.share_locks[locking.LEVEL_NODE] = 1
2690

    
2691
    if self.op.nodes:
2692
      self.needed_locks[locking.LEVEL_NODE] = \
2693
        _GetWantedNodes(self, self.op.nodes)
2694
    else:
2695
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
2696

    
2697
  def CheckPrereq(self):
2698
    """Check prerequisites.
2699

2700
    This checks that the fields required are valid output fields.
2701

2702
    """
2703
    self.op.name = getattr(self.op, "name", None)
2704

    
2705
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
2706

    
2707
  def Exec(self, feedback_fn):
2708
    """Computes the list of nodes and their attributes.
2709

2710
    """
2711
    # Always get name to sort by
2712
    if constants.SF_NAME in self.op.output_fields:
2713
      fields = self.op.output_fields[:]
2714
    else:
2715
      fields = [constants.SF_NAME] + self.op.output_fields
2716

    
2717
    # Never ask for node or type as it's only known to the LU
2718
    for extra in [constants.SF_NODE, constants.SF_TYPE]:
2719
      while extra in fields:
2720
        fields.remove(extra)
2721

    
2722
    field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
2723
    name_idx = field_idx[constants.SF_NAME]
2724

    
2725
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
2726
    data = self.rpc.call_storage_list(self.nodes,
2727
                                      self.op.storage_type, st_args,
2728
                                      self.op.name, fields)
2729

    
2730
    result = []
2731

    
2732
    for node in utils.NiceSort(self.nodes):
2733
      nresult = data[node]
2734
      if nresult.offline:
2735
        continue
2736

    
2737
      msg = nresult.fail_msg
2738
      if msg:
2739
        self.LogWarning("Can't get storage data from node %s: %s", node, msg)
2740
        continue
2741

    
2742
      rows = dict([(row[name_idx], row) for row in nresult.payload])
2743

    
2744
      for name in utils.NiceSort(rows.keys()):
2745
        row = rows[name]
2746

    
2747
        out = []
2748

    
2749
        for field in self.op.output_fields:
2750
          if field == constants.SF_NODE:
2751
            val = node
2752
          elif field == constants.SF_TYPE:
2753
            val = self.op.storage_type
2754
          elif field in field_idx:
2755
            val = row[field_idx[field]]
2756
          else:
2757
            raise errors.ParameterError(field)
2758

    
2759
          out.append(val)
2760

    
2761
        result.append(out)
2762

    
2763
    return result
2764

    
2765

    
2766
class LUModifyNodeStorage(NoHooksLU):
2767
  """Logical unit for modifying a storage volume on a node.
2768

2769
  """
2770
  _OP_REQP = ["node_name", "storage_type", "name", "changes"]
2771
  REQ_BGL = False
2772

    
2773
  def CheckArguments(self):
2774
    node_name = self.cfg.ExpandNodeName(self.op.node_name)
2775
    if node_name is None:
2776
      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name,
2777
                                 errors.ECODE_NOENT)
2778

    
2779
    self.op.node_name = node_name
2780

    
2781
    storage_type = self.op.storage_type
2782
    if storage_type not in constants.VALID_STORAGE_TYPES:
2783
      raise errors.OpPrereqError("Unknown storage type: %s" % storage_type,
2784
                                 errors.ECODE_INVAL)
2785

    
2786
  def ExpandNames(self):
2787
    self.needed_locks = {
2788
      locking.LEVEL_NODE: self.op.node_name,
2789
      }
2790

    
2791
  def CheckPrereq(self):
2792
    """Check prerequisites.
2793

2794
    """
2795
    storage_type = self.op.storage_type
2796

    
2797
    try:
2798
      modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
2799
    except KeyError:
2800
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
2801
                                 " modified" % storage_type,
2802
                                 errors.ECODE_INVAL)
2803

    
2804
    diff = set(self.op.changes.keys()) - modifiable
2805
    if diff:
2806
      raise errors.OpPrereqError("The following fields can not be modified for"
2807
                                 " storage units of type '%s': %r" %
2808
                                 (storage_type, list(diff)),
2809
                                 errors.ECODE_INVAL)
2810

    
2811
  def Exec(self, feedback_fn):
2812
    """Computes the list of nodes and their attributes.
2813

2814
    """
2815
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
2816
    result = self.rpc.call_storage_modify(self.op.node_name,
2817
                                          self.op.storage_type, st_args,
2818
                                          self.op.name, self.op.changes)
2819
    result.Raise("Failed to modify storage unit '%s' on %s" %
2820
                 (self.op.name, self.op.node_name))
2821

    
2822

    
2823
class LUAddNode(LogicalUnit):
2824
  """Logical unit for adding node to the cluster.
2825

2826
  """
2827
  HPATH = "node-add"
2828
  HTYPE = constants.HTYPE_NODE
2829
  _OP_REQP = ["node_name"]
2830

    
2831
  def BuildHooksEnv(self):
2832
    """Build hooks env.
2833

2834
    This will run on all nodes before, and on all nodes + the new node after.
2835

2836
    """
2837
    env = {
2838
      "OP_TARGET": self.op.node_name,
2839
      "NODE_NAME": self.op.node_name,
2840
      "NODE_PIP": self.op.primary_ip,
2841
      "NODE_SIP": self.op.secondary_ip,
2842
      }
2843
    nodes_0 = self.cfg.GetNodeList()
2844
    nodes_1 = nodes_0 + [self.op.node_name, ]
2845
    return env, nodes_0, nodes_1
2846

    
2847
  def CheckPrereq(self):
2848
    """Check prerequisites.
2849

2850
    This checks:
2851
     - the new node is not already in the config
2852
     - it is resolvable
2853
     - its parameters (single/dual homed) matches the cluster
2854

2855
    Any errors are signaled by raising errors.OpPrereqError.
2856

2857
    """
2858
    node_name = self.op.node_name
2859
    cfg = self.cfg
2860

    
2861
    dns_data = utils.GetHostInfo(node_name)
2862

    
2863
    node = dns_data.name
2864
    primary_ip = self.op.primary_ip = dns_data.ip
2865
    secondary_ip = getattr(self.op, "secondary_ip", None)
2866
    if secondary_ip is None:
2867
      secondary_ip = primary_ip
2868
    if not utils.IsValidIP(secondary_ip):
2869
      raise errors.OpPrereqError("Invalid secondary IP given",
2870
                                 errors.ECODE_INVAL)
2871
    self.op.secondary_ip = secondary_ip
2872

    
2873
    node_list = cfg.GetNodeList()
2874
    if not self.op.readd and node in node_list:
2875
      raise errors.OpPrereqError("Node %s is already in the configuration" %
2876
                                 node, errors.ECODE_EXISTS)
2877
    elif self.op.readd and node not in node_list:
2878
      raise errors.OpPrereqError("Node %s is not in the configuration" % node,
2879
                                 errors.ECODE_NOENT)
2880

    
2881
    for existing_node_name in node_list:
2882
      existing_node = cfg.GetNodeInfo(existing_node_name)
2883

    
2884
      if self.op.readd and node == existing_node_name:
2885
        if (existing_node.primary_ip != primary_ip or
2886
            existing_node.secondary_ip != secondary_ip):
2887
          raise errors.OpPrereqError("Readded node doesn't have the same IP"
2888
                                     " address configuration as before",
2889
                                     errors.ECODE_INVAL)
2890
        continue
2891

    
2892
      if (existing_node.primary_ip == primary_ip or
2893
          existing_node.secondary_ip == primary_ip or
2894
          existing_node.primary_ip == secondary_ip or
2895
          existing_node.secondary_ip == secondary_ip):
2896
        raise errors.OpPrereqError("New node ip address(es) conflict with"
2897
                                   " existing node %s" % existing_node.name,
2898
                                   errors.ECODE_NOTUNIQUE)
2899

    
2900
    # check that the type of the node (single versus dual homed) is the
2901
    # same as for the master
2902
    myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
2903
    master_singlehomed = myself.secondary_ip == myself.primary_ip
2904
    newbie_singlehomed = secondary_ip == primary_ip
2905
    if master_singlehomed != newbie_singlehomed:
2906
      if master_singlehomed:
2907
        raise errors.OpPrereqError("The master has no private ip but the"
2908
                                   " new node has one",
2909
                                   errors.ECODE_INVAL)
2910
      else:
2911
        raise errors.OpPrereqError("The master has a private ip but the"
2912
                                   " new node doesn't have one",
2913
                                   errors.ECODE_INVAL)
2914

    
2915
    # checks reachability
2916
    if not utils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
2917
      raise errors.OpPrereqError("Node not reachable by ping",
2918
                                 errors.ECODE_ENVIRON)
2919

    
2920
    if not newbie_singlehomed:
2921
      # check reachability from my secondary ip to newbie's secondary ip
2922
      if not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
2923
                           source=myself.secondary_ip):
2924
        raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
2925
                                   " based ping to noded port",
2926
                                   errors.ECODE_ENVIRON)
2927

    
2928
    if self.op.readd:
2929
      exceptions = [node]
2930
    else:
2931
      exceptions = []
2932

    
2933
    self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
2934

    
2935
    if self.op.readd:
2936
      self.new_node = self.cfg.GetNodeInfo(node)
2937
      assert self.new_node is not None, "Can't retrieve locked node %s" % node
2938
    else:
2939
      self.new_node = objects.Node(name=node,
2940
                                   primary_ip=primary_ip,
2941
                                   secondary_ip=secondary_ip,
2942
                                   master_candidate=self.master_candidate,
2943
                                   offline=False, drained=False)
2944

    
2945
  def Exec(self, feedback_fn):
2946
    """Adds the new node to the cluster.
2947

2948
    """
2949
    new_node = self.new_node
2950
    node = new_node.name
2951

    
2952
    # for re-adds, reset the offline/drained/master-candidate flags;
2953
    # we need to reset here, otherwise offline would prevent RPC calls
2954
    # later in the procedure; this also means that if the re-add
2955
    # fails, we are left with a non-offlined, broken node
2956
    if self.op.readd:
2957
      new_node.drained = new_node.offline = False
2958
      self.LogInfo("Readding a node, the offline/drained flags were reset")
2959
      # if we demote the node, we do cleanup later in the procedure
2960
      new_node.master_candidate = self.master_candidate
2961

    
2962
    # notify the user about any possible mc promotion
2963
    if new_node.master_candidate:
2964
      self.LogInfo("Node will be a master candidate")
2965

    
2966
    # check connectivity
2967
    result = self.rpc.call_version([node])[node]
2968
    result.Raise("Can't get version information from node %s" % node)
2969
    if constants.PROTOCOL_VERSION == result.payload:
2970
      logging.info("Communication to node %s fine, sw version %s match",
2971
                   node, result.payload)
2972
    else:
2973
      raise errors.OpExecError("Version mismatch master version %s,"
2974
                               " node version %s" %
2975
                               (constants.PROTOCOL_VERSION, result.payload))
2976

    
2977
    # setup ssh on node
2978
    if self.cfg.GetClusterInfo().modify_ssh_setup:
2979
      logging.info("Copy ssh key to node %s", node)
2980
      priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
2981
      keyarray = []
2982
      keyfiles = [constants.SSH_HOST_DSA_PRIV, constants.SSH_HOST_DSA_PUB,
2983
                  constants.SSH_HOST_RSA_PRIV, constants.SSH_HOST_RSA_PUB,
2984
                  priv_key, pub_key]
2985

    
2986
      for i in keyfiles:
2987
        keyarray.append(utils.ReadFile(i))
2988

    
2989
      result = self.rpc.call_node_add(node, keyarray[0], keyarray[1],
2990
                                      keyarray[2], keyarray[3], keyarray[4],
2991
                                      keyarray[5])
2992
      result.Raise("Cannot transfer ssh keys to the new node")
2993

    
2994
    # Add node to our /etc/hosts, and add key to known_hosts
2995
    if self.cfg.GetClusterInfo().modify_etc_hosts:
2996
      utils.AddHostToEtcHosts(new_node.name)
2997

    
2998
    if new_node.secondary_ip != new_node.primary_ip:
2999
      result = self.rpc.call_node_has_ip_address(new_node.name,
3000
                                                 new_node.secondary_ip)
3001
      result.Raise("Failure checking secondary ip on node %s" % new_node.name,
3002
                   prereq=True, ecode=errors.ECODE_ENVIRON)
3003
      if not result.payload:
3004
        raise errors.OpExecError("Node claims it doesn't have the secondary ip"
3005
                                 " you gave (%s). Please fix and re-run this"
3006
                                 " command." % new_node.secondary_ip)
3007

    
3008
    node_verify_list = [self.cfg.GetMasterNode()]
3009
    node_verify_param = {
3010
      constants.NV_NODELIST: [node],
3011
      # TODO: do a node-net-test as well?
3012
    }
3013

    
3014
    result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
3015
                                       self.cfg.GetClusterName())
3016
    for verifier in node_verify_list:
3017
      result[verifier].Raise("Cannot communicate with node %s" % verifier)
3018
      nl_payload = result[verifier].payload[constants.NV_NODELIST]
3019
      if nl_payload:
3020
        for failed in nl_payload:
3021
          feedback_fn("ssh/hostname verification failed"
3022
                      " (checking from %s): %s" %
3023
                      (verifier, nl_payload[failed]))
3024
        raise errors.OpExecError("ssh/hostname verification failed.")
3025

    
3026
    if self.op.readd:
3027
      _RedistributeAncillaryFiles(self)
3028
      self.context.ReaddNode(new_node)
3029
      # make sure we redistribute the config
3030
      self.cfg.Update(new_node, feedback_fn)
3031
      # and make sure the new node will not have old files around
3032
      if not new_node.master_candidate:
3033
        result = self.rpc.call_node_demote_from_mc(new_node.name)
3034
        msg = result.fail_msg
3035
        if msg:
3036
          self.LogWarning("Node failed to demote itself from master"
3037
                          " candidate status: %s" % msg)
3038
    else:
3039
      _RedistributeAncillaryFiles(self, additional_nodes=[node])
3040
      self.context.AddNode(new_node, self.proc.GetECId())
3041

    
3042

    
3043
class LUSetNodeParams(LogicalUnit):
3044
  """Modifies the parameters of a node.
3045

3046
  """
3047
  HPATH = "node-modify"
3048
  HTYPE = constants.HTYPE_NODE
3049
  _OP_REQP = ["node_name"]
3050
  REQ_BGL = False
3051

    
3052
  def CheckArguments(self):
3053
    node_name = self.cfg.ExpandNodeName(self.op.node_name)
3054
    if node_name is None:
3055
      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name,
3056
                                 errors.ECODE_INVAL)
3057
    self.op.node_name = node_name
3058
    _CheckBooleanOpField(self.op, 'master_candidate')
3059
    _CheckBooleanOpField(self.op, 'offline')
3060
    _CheckBooleanOpField(self.op, 'drained')
3061
    all_mods = [self.op.offline, self.op.master_candidate, self.op.drained]
3062
    if all_mods.count(None) == 3:
3063
      raise errors.OpPrereqError("Please pass at least one modification",
3064
                                 errors.ECODE_INVAL)
3065
    if all_mods.count(True) > 1:
3066
      raise errors.OpPrereqError("Can't set the node into more than one"
3067
                                 " state at the same time",
3068
                                 errors.ECODE_INVAL)
3069

    
3070
  def ExpandNames(self):
3071
    self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
3072

    
3073
  def BuildHooksEnv(self):
3074
    """Build hooks env.
3075

3076
    This runs on the master node.
3077

3078
    """
3079
    env = {
3080
      "OP_TARGET": self.op.node_name,
3081
      "MASTER_CANDIDATE": str(self.op.master_candidate),
3082
      "OFFLINE": str(self.op.offline),
3083
      "DRAINED": str(self.op.drained),
3084
      }
3085
    nl = [self.cfg.GetMasterNode(),
3086
          self.op.node_name]
3087
    return env, nl, nl
3088

    
3089
  def CheckPrereq(self):
3090
    """Check prerequisites.
3091

3092
    This only checks the instance list against the existing names.
3093

3094
    """
3095
    node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
3096

    
3097
    if (self.op.master_candidate is not None or
3098
        self.op.drained is not None or
3099
        self.op.offline is not None):
3100
      # we can't change the master's node flags
3101
      if self.op.node_name == self.cfg.GetMasterNode():
3102
        raise errors.OpPrereqError("The master role can be changed"
3103
                                   " only via masterfailover",
3104
                                   errors.ECODE_INVAL)
3105

    
3106
    # Boolean value that tells us whether we're offlining or draining the node
3107
    offline_or_drain = self.op.offline == True or self.op.drained == True
3108
    deoffline_or_drain = self.op.offline == False or self.op.drained == False
3109

    
3110
    if (node.master_candidate and
3111
        (self.op.master_candidate == False or offline_or_drain)):
3112
      cp_size = self.cfg.GetClusterInfo().candidate_pool_size
3113
      mc_now, mc_should, mc_max = self.cfg.GetMasterCandidateStats()
3114
      if mc_now <= cp_size:
3115
        msg = ("Not enough master candidates (desired"
3116
               " %d, new value will be %d)" % (cp_size, mc_now-1))
3117
        # Only allow forcing the operation if it's an offline/drain operation,
3118
        # and we could not possibly promote more nodes.
3119
        # FIXME: this can still lead to issues if in any way another node which
3120
        # could be promoted appears in the meantime.
3121
        if self.op.force and offline_or_drain and mc_should == mc_max:
3122
          self.LogWarning(msg)
3123
        else:
3124
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
3125

    
3126
    if (self.op.master_candidate == True and
3127
        ((node.offline and not self.op.offline == False) or
3128
         (node.drained and not self.op.drained == False))):
3129
      raise errors.OpPrereqError("Node '%s' is offline or drained, can't set"
3130
                                 " to master_candidate" % node.name,
3131
                                 errors.ECODE_INVAL)
3132

    
3133
    # If we're being deofflined/drained, we'll MC ourself if needed
3134
    if (deoffline_or_drain and not offline_or_drain and not
3135
        self.op.master_candidate == True):
3136
      self.op.master_candidate = _DecideSelfPromotion(self)
3137
      if self.op.master_candidate:
3138
        self.LogInfo("Autopromoting node to master candidate")
3139

    
3140
    return
3141

    
3142
  def Exec(self, feedback_fn):
3143
    """Modifies a node.
3144

3145
    """
3146
    node = self.node
3147

    
3148
    result = []
3149
    changed_mc = False
3150

    
3151
    if self.op.offline is not None:
3152
      node.offline = self.op.offline
3153
      result.append(("offline", str(self.op.offline)))
3154
      if self.op.offline == True:
3155
        if node.master_candidate:
3156
          node.master_candidate = False
3157
          changed_mc = True
3158
          result.append(("master_candidate", "auto-demotion due to offline"))
3159
        if node.drained:
3160
          node.drained = False
3161
          result.append(("drained", "clear drained status due to offline"))
3162

    
3163
    if self.op.master_candidate is not None:
3164
      node.master_candidate = self.op.master_candidate
3165
      changed_mc = True
3166
      result.append(("master_candidate", str(self.op.master_candidate)))
3167
      if self.op.master_candidate == False:
3168
        rrc = self.rpc.call_node_demote_from_mc(node.name)
3169
        msg = rrc.fail_msg
3170
        if msg:
3171
          self.LogWarning("Node failed to demote itself: %s" % msg)
3172

    
3173
    if self.op.drained is not None:
3174
      node.drained = self.op.drained
3175
      result.append(("drained", str(self.op.drained)))
3176
      if self.op.drained == True:
3177
        if node.master_candidate:
3178
          node.master_candidate = False
3179
          changed_mc = True
3180
          result.append(("master_candidate", "auto-demotion due to drain"))
3181
          rrc = self.rpc.call_node_demote_from_mc(node.name)
3182
          msg = rrc.fail_msg
3183
          if msg:
3184
            self.LogWarning("Node failed to demote itself: %s" % msg)
3185
        if node.offline:
3186
          node.offline = False
3187
          result.append(("offline", "clear offline status due to drain"))
3188

    
3189
    # this will trigger configuration file update, if needed
3190
    self.cfg.Update(node, feedback_fn)
3191
    # this will trigger job queue propagation or cleanup
3192
    if changed_mc:
3193
      self.context.ReaddNode(node)
3194

    
3195
    return result
3196

    
3197

    
3198
class LUPowercycleNode(NoHooksLU):
3199
  """Powercycles a node.
3200

3201
  """
3202
  _OP_REQP = ["node_name", "force"]
3203
  REQ_BGL = False
3204

    
3205
  def CheckArguments(self):
3206
    node_name = self.cfg.ExpandNodeName(self.op.node_name)
3207
    if node_name is None:
3208
      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name,
3209
                                 errors.ECODE_NOENT)
3210
    self.op.node_name = node_name
3211
    if node_name == self.cfg.GetMasterNode() and not self.op.force:
3212
      raise errors.OpPrereqError("The node is the master and the force"
3213
                                 " parameter was not set",
3214
                                 errors.ECODE_INVAL)
3215

    
3216
  def ExpandNames(self):
3217
    """Locking for PowercycleNode.
3218

3219
    This is a last-resort option and shouldn't block on other
3220
    jobs. Therefore, we grab no locks.
3221

3222
    """
3223
    self.needed_locks = {}
3224

    
3225
  def CheckPrereq(self):
3226
    """Check prerequisites.
3227

3228
    This LU has no prereqs.
3229

3230
    """
3231
    pass
3232

    
3233
  def Exec(self, feedback_fn):
3234
    """Reboots a node.
3235

3236
    """
3237
    result = self.rpc.call_node_powercycle(self.op.node_name,
3238
                                           self.cfg.GetHypervisorType())
3239
    result.Raise("Failed to schedule the reboot")
3240
    return result.payload
3241

    
3242

    
3243
class LUQueryClusterInfo(NoHooksLU):
3244
  """Query cluster configuration.
3245

3246
  """
3247
  _OP_REQP = []
3248
  REQ_BGL = False
3249

    
3250
  def ExpandNames(self):
3251
    self.needed_locks = {}
3252

    
3253
  def CheckPrereq(self):
3254
    """No prerequsites needed for this LU.
3255

3256
    """
3257
    pass
3258

    
3259
  def Exec(self, feedback_fn):
3260
    """Return cluster config.
3261

3262
    """
3263
    cluster = self.cfg.GetClusterInfo()
3264
    result = {
3265
      "software_version": constants.RELEASE_VERSION,
3266
      "protocol_version": constants.PROTOCOL_VERSION,
3267
      "config_version": constants.CONFIG_VERSION,
3268
      "os_api_version": max(constants.OS_API_VERSIONS),
3269
      "export_version": constants.EXPORT_VERSION,
3270
      "architecture": (platform.architecture()[0], platform.machine()),
3271
      "name": cluster.cluster_name,
3272
      "master": cluster.master_node,
3273
      "default_hypervisor": cluster.enabled_hypervisors[0],
3274
      "enabled_hypervisors": cluster.enabled_hypervisors,
3275
      "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
3276
                        for hypervisor_name in cluster.enabled_hypervisors]),
3277
      "beparams": cluster.beparams,
3278
      "nicparams": cluster.nicparams,
3279
      "candidate_pool_size": cluster.candidate_pool_size,
3280
      "master_netdev": cluster.master_netdev,
3281
      "volume_group_name": cluster.volume_group_name,
3282
      "file_storage_dir": cluster.file_storage_dir,
3283
      "ctime": cluster.ctime,
3284
      "mtime": cluster.mtime,
3285
      "uuid": cluster.uuid,
3286
      "tags": list(cluster.GetTags()),
3287
      }
3288

    
3289
    return result
3290

    
3291

    
3292
class LUQueryConfigValues(NoHooksLU):
3293
  """Return configuration values.
3294

3295
  """
3296
  _OP_REQP = []
3297
  REQ_BGL = False
3298
  _FIELDS_DYNAMIC = utils.FieldSet()
3299
  _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag",
3300
                                  "watcher_pause")
3301

    
3302
  def ExpandNames(self):
3303
    self.needed_locks = {}
3304

    
3305
    _CheckOutputFields(static=self._FIELDS_STATIC,
3306
                       dynamic=self._FIELDS_DYNAMIC,
3307
                       selected=self.op.output_fields)
3308

    
3309
  def CheckPrereq(self):
3310
    """No prerequisites.
3311

3312
    """
3313
    pass
3314

    
3315
  def Exec(self, feedback_fn):
3316
    """Dump a representation of the cluster config to the standard output.
3317

3318
    """
3319
    values = []
3320
    for field in self.op.output_fields:
3321
      if field == "cluster_name":
3322
        entry = self.cfg.GetClusterName()
3323
      elif field == "master_node":
3324
        entry = self.cfg.GetMasterNode()
3325
      elif field == "drain_flag":
3326
        entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
3327
      elif field == "watcher_pause":
3328
        return utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
3329
      else:
3330
        raise errors.ParameterError(field)
3331
      values.append(entry)
3332
    return values
3333

    
3334

    
3335
class LUActivateInstanceDisks(NoHooksLU):
3336
  """Bring up an instance's disks.
3337

3338
  """
3339
  _OP_REQP = ["instance_name"]
3340
  REQ_BGL = False
3341

    
3342
  def ExpandNames(self):
3343
    self._ExpandAndLockInstance()
3344
    self.needed_locks[locking.LEVEL_NODE] = []
3345
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3346

    
3347
  def DeclareLocks(self, level):
3348
    if level == locking.LEVEL_NODE:
3349
      self._LockInstancesNodes()
3350

    
3351
  def CheckPrereq(self):
3352
    """Check prerequisites.
3353

3354
    This checks that the instance is in the cluster.
3355

3356
    """
3357
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3358
    assert self.instance is not None, \
3359
      "Cannot retrieve locked instance %s" % self.op.instance_name
3360
    _CheckNodeOnline(self, self.instance.primary_node)
3361
    if not hasattr(self.op, "ignore_size"):
3362
      self.op.ignore_size = False
3363

    
3364
  def Exec(self, feedback_fn):
3365
    """Activate the disks.
3366

3367
    """
3368
    disks_ok, disks_info = \
3369
              _AssembleInstanceDisks(self, self.instance,
3370
                                     ignore_size=self.op.ignore_size)
3371
    if not disks_ok:
3372
      raise errors.OpExecError("Cannot activate block devices")
3373

    
3374
    return disks_info
3375

    
3376

    
3377
def _AssembleInstanceDisks(lu, instance, ignore_secondaries=False,
3378
                           ignore_size=False):
3379
  """Prepare the block devices for an instance.
3380

3381
  This sets up the block devices on all nodes.
3382

3383
  @type lu: L{LogicalUnit}
3384
  @param lu: the logical unit on whose behalf we execute
3385
  @type instance: L{objects.Instance}
3386
  @param instance: the instance for whose disks we assemble
3387
  @type ignore_secondaries: boolean
3388
  @param ignore_secondaries: if true, errors on secondary nodes
3389
      won't result in an error return from the function
3390
  @type ignore_size: boolean
3391
  @param ignore_size: if true, the current known size of the disk
3392
      will not be used during the disk activation, useful for cases
3393
      when the size is wrong
3394
  @return: False if the operation failed, otherwise a list of
3395
      (host, instance_visible_name, node_visible_name)
3396
      with the mapping from node devices to instance devices
3397

3398
  """
3399
  device_info = []
3400
  disks_ok = True
3401
  iname = instance.name
3402
  # With the two passes mechanism we try to reduce the window of
3403
  # opportunity for the race condition of switching DRBD to primary
3404
  # before handshaking occured, but we do not eliminate it
3405

    
3406
  # The proper fix would be to wait (with some limits) until the
3407
  # connection has been made and drbd transitions from WFConnection
3408
  # into any other network-connected state (Connected, SyncTarget,
3409
  # SyncSource, etc.)
3410

    
3411
  # 1st pass, assemble on all nodes in secondary mode
3412
  for inst_disk in instance.disks:
3413
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
3414
      if ignore_size:
3415
        node_disk = node_disk.Copy()
3416
        node_disk.UnsetSize()
3417
      lu.cfg.SetDiskID(node_disk, node)
3418
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
3419
      msg = result.fail_msg
3420
      if msg:
3421
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
3422
                           " (is_primary=False, pass=1): %s",
3423
                           inst_disk.iv_name, node, msg)
3424
        if not ignore_secondaries:
3425
          disks_ok = False
3426

    
3427
  # FIXME: race condition on drbd migration to primary
3428

    
3429
  # 2nd pass, do only the primary node
3430
  for inst_disk in instance.disks:
3431
    dev_path = None
3432

    
3433
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
3434
      if node != instance.primary_node:
3435
        continue
3436
      if ignore_size:
3437
        node_disk = node_disk.Copy()
3438
        node_disk.UnsetSize()
3439
      lu.cfg.SetDiskID(node_disk, node)
3440
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
3441
      msg = result.fail_msg
3442
      if msg:
3443
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
3444
                           " (is_primary=True, pass=2): %s",
3445
                           inst_disk.iv_name, node, msg)
3446
        disks_ok = False
3447
      else:
3448
        dev_path = result.payload
3449

    
3450
    device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
3451

    
3452
  # leave the disks configured for the primary node
3453
  # this is a workaround that would be fixed better by
3454
  # improving the logical/physical id handling
3455
  for disk in instance.disks:
3456
    lu.cfg.SetDiskID(disk, instance.primary_node)
3457

    
3458
  return disks_ok, device_info
3459

    
3460

    
3461
def _StartInstanceDisks(lu, instance, force):
3462
  """Start the disks of an instance.
3463

3464
  """
3465
  disks_ok, _ = _AssembleInstanceDisks(lu, instance,
3466
                                           ignore_secondaries=force)
3467
  if not disks_ok:
3468
    _ShutdownInstanceDisks(lu, instance)
3469
    if force is not None and not force:
3470
      lu.proc.LogWarning("", hint="If the message above refers to a"
3471
                         " secondary node,"
3472
                         " you can retry the operation using '--force'.")
3473
    raise errors.OpExecError("Disk consistency error")
3474

    
3475

    
3476
class LUDeactivateInstanceDisks(NoHooksLU):
3477
  """Shutdown an instance's disks.
3478

3479
  """
3480
  _OP_REQP = ["instance_name"]
3481
  REQ_BGL = False
3482

    
3483
  def ExpandNames(self):
3484
    self._ExpandAndLockInstance()
3485
    self.needed_locks[locking.LEVEL_NODE] = []
3486
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3487

    
3488
  def DeclareLocks(self, level):
3489
    if level == locking.LEVEL_NODE:
3490
      self._LockInstancesNodes()
3491

    
3492
  def CheckPrereq(self):
3493
    """Check prerequisites.
3494

3495
    This checks that the instance is in the cluster.
3496

3497
    """
3498
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3499
    assert self.instance is not None, \
3500
      "Cannot retrieve locked instance %s" % self.op.instance_name
3501

    
3502
  def Exec(self, feedback_fn):
3503
    """Deactivate the disks
3504

3505
    """
3506
    instance = self.instance
3507
    _SafeShutdownInstanceDisks(self, instance)
3508

    
3509

    
3510
def _SafeShutdownInstanceDisks(lu, instance):
3511
  """Shutdown block devices of an instance.
3512

3513
  This function checks if an instance is running, before calling
3514
  _ShutdownInstanceDisks.
3515

3516
  """
3517
  pnode = instance.primary_node
3518
  ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
3519
  ins_l.Raise("Can't contact node %s" % pnode)
3520

    
3521
  if instance.name in ins_l.payload:
3522
    raise errors.OpExecError("Instance is running, can't shutdown"
3523
                             " block devices.")
3524

    
3525
  _ShutdownInstanceDisks(lu, instance)
3526

    
3527

    
3528
def _ShutdownInstanceDisks(lu, instance, ignore_primary=False):
3529
  """Shutdown block devices of an instance.
3530

3531
  This does the shutdown on all nodes of the instance.
3532

3533
  If the ignore_primary is false, errors on the primary node are
3534
  ignored.
3535

3536
  """
3537
  all_result = True
3538
  for disk in instance.disks:
3539
    for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
3540
      lu.cfg.SetDiskID(top_disk, node)
3541
      result = lu.rpc.call_blockdev_shutdown(node, top_disk)
3542
      msg = result.fail_msg
3543
      if msg:
3544
        lu.LogWarning("Could not shutdown block device %s on node %s: %s",
3545
                      disk.iv_name, node, msg)
3546
        if not ignore_primary or node != instance.primary_node:
3547
          all_result = False
3548
  return all_result
3549

    
3550

    
3551
def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
3552
  """Checks if a node has enough free memory.
3553

3554
  This function check if a given node has the needed amount of free
3555
  memory. In case the node has less memory or we cannot get the
3556
  information from the node, this function raise an OpPrereqError
3557
  exception.
3558

3559
  @type lu: C{LogicalUnit}
3560
  @param lu: a logical unit from which we get configuration data
3561
  @type node: C{str}
3562
  @param node: the node to check
3563
  @type reason: C{str}
3564
  @param reason: string to use in the error message
3565
  @type requested: C{int}
3566
  @param requested: the amount of memory in MiB to check for
3567
  @type hypervisor_name: C{str}
3568
  @param hypervisor_name: the hypervisor to ask for memory stats
3569
  @raise errors.OpPrereqError: if the node doesn't have enough memory, or
3570
      we cannot check the node
3571

3572
  """
3573
  nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
3574
  nodeinfo[node].Raise("Can't get data from node %s" % node,
3575
                       prereq=True, ecode=errors.ECODE_ENVIRON)
3576
  free_mem = nodeinfo[node].payload.get('memory_free', None)
3577
  if not isinstance(free_mem, int):
3578
    raise errors.OpPrereqError("Can't compute free memory on node %s, result"
3579
                               " was '%s'" % (node, free_mem),
3580
                               errors.ECODE_ENVIRON)
3581
  if requested > free_mem:
3582
    raise errors.OpPrereqError("Not enough memory on node %s for %s:"
3583
                               " needed %s MiB, available %s MiB" %
3584
                               (node, reason, requested, free_mem),
3585
                               errors.ECODE_NORES)
3586

    
3587

    
3588
class LUStartupInstance(LogicalUnit):
3589
  """Starts an instance.
3590

3591
  """
3592
  HPATH = "instance-start"
3593
  HTYPE = constants.HTYPE_INSTANCE
3594
  _OP_REQP = ["instance_name", "force"]
3595
  REQ_BGL = False
3596

    
3597
  def ExpandNames(self):
3598
    self._ExpandAndLockInstance()
3599

    
3600
  def BuildHooksEnv(self):
3601
    """Build hooks env.
3602

3603
    This runs on master, primary and secondary nodes of the instance.
3604

3605
    """
3606
    env = {
3607
      "FORCE": self.op.force,
3608
      }
3609
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
3610
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3611
    return env, nl, nl
3612

    
3613
  def CheckPrereq(self):
3614
    """Check prerequisites.
3615

3616
    This checks that the instance is in the cluster.
3617

3618
    """
3619
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3620
    assert self.instance is not None, \
3621
      "Cannot retrieve locked instance %s" % self.op.instance_name
3622

    
3623
    # extra beparams
3624
    self.beparams = getattr(self.op, "beparams", {})
3625
    if self.beparams:
3626
      if not isinstance(self.beparams, dict):
3627
        raise errors.OpPrereqError("Invalid beparams passed: %s, expected"
3628
                                   " dict" % (type(self.beparams), ),
3629
                                   errors.ECODE_INVAL)
3630
      # fill the beparams dict
3631
      utils.ForceDictType(self.beparams, constants.BES_PARAMETER_TYPES)
3632
      self.op.beparams = self.beparams
3633

    
3634
    # extra hvparams
3635
    self.hvparams = getattr(self.op, "hvparams", {})
3636
    if self.hvparams:
3637
      if not isinstance(self.hvparams, dict):
3638
        raise errors.OpPrereqError("Invalid hvparams passed: %s, expected"
3639
                                   " dict" % (type(self.hvparams), ),
3640
                                   errors.ECODE_INVAL)
3641

    
3642
      # check hypervisor parameter syntax (locally)
3643
      cluster = self.cfg.GetClusterInfo()
3644
      utils.ForceDictType(self.hvparams, constants.HVS_PARAMETER_TYPES)
3645
      filled_hvp = objects.FillDict(cluster.hvparams[instance.hypervisor],
3646
                                    instance.hvparams)
3647
      filled_hvp.update(self.hvparams)
3648
      hv_type = hypervisor.GetHypervisor(instance.hypervisor)
3649
      hv_type.CheckParameterSyntax(filled_hvp)
3650
      _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
3651
      self.op.hvparams = self.hvparams
3652

    
3653
    _CheckNodeOnline(self, instance.primary_node)
3654

    
3655
    bep = self.cfg.GetClusterInfo().FillBE(instance)
3656
    # check bridges existence
3657
    _CheckInstanceBridgesExist(self, instance)
3658

    
3659
    remote_info = self.rpc.call_instance_info(instance.primary_node,
3660
                                              instance.name,
3661
                                              instance.hypervisor)
3662
    remote_info.Raise("Error checking node %s" % instance.primary_node,
3663
                      prereq=True, ecode=errors.ECODE_ENVIRON)
3664
    if not remote_info.payload: # not running already
3665
      _CheckNodeFreeMemory(self, instance.primary_node,
3666
                           "starting instance %s" % instance.name,
3667
                           bep[constants.BE_MEMORY], instance.hypervisor)
3668

    
3669
  def Exec(self, feedback_fn):
3670
    """Start the instance.
3671

3672
    """
3673
    instance = self.instance
3674
    force = self.op.force
3675

    
3676
    self.cfg.MarkInstanceUp(instance.name)
3677

    
3678
    node_current = instance.primary_node
3679

    
3680
    _StartInstanceDisks(self, instance, force)
3681

    
3682
    result = self.rpc.call_instance_start(node_current, instance,
3683
                                          self.hvparams, self.beparams)
3684
    msg = result.fail_msg
3685
    if msg:
3686
      _ShutdownInstanceDisks(self, instance)
3687
      raise errors.OpExecError("Could not start instance: %s" % msg)
3688

    
3689

    
3690
class LURebootInstance(LogicalUnit):
3691
  """Reboot an instance.
3692

3693
  """
3694
  HPATH = "instance-reboot"
3695
  HTYPE = constants.HTYPE_INSTANCE
3696
  _OP_REQP = ["instance_name", "ignore_secondaries", "reboot_type"]
3697
  REQ_BGL = False
3698

    
3699
  def CheckArguments(self):
3700
    """Check the arguments.
3701

3702
    """
3703
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
3704
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
3705

    
3706
  def ExpandNames(self):
3707
    if self.op.reboot_type not in [constants.INSTANCE_REBOOT_SOFT,
3708
                                   constants.INSTANCE_REBOOT_HARD,
3709
                                   constants.INSTANCE_REBOOT_FULL]:
3710
      raise errors.ParameterError("reboot type not in [%s, %s, %s]" %
3711
                                  (constants.INSTANCE_REBOOT_SOFT,
3712
                                   constants.INSTANCE_REBOOT_HARD,
3713
                                   constants.INSTANCE_REBOOT_FULL))
3714
    self._ExpandAndLockInstance()
3715

    
3716
  def BuildHooksEnv(self):
3717
    """Build hooks env.
3718

3719
    This runs on master, primary and secondary nodes of the instance.
3720

3721
    """
3722
    env = {
3723
      "IGNORE_SECONDARIES": self.op.ignore_secondaries,
3724
      "REBOOT_TYPE": self.op.reboot_type,
3725
      "SHUTDOWN_TIMEOUT": self.shutdown_timeout,
3726
      }
3727
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
3728
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3729
    return env, nl, nl
3730

    
3731
  def CheckPrereq(self):
3732
    """Check prerequisites.
3733

3734
    This checks that the instance is in the cluster.
3735

3736
    """
3737
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3738
    assert self.instance is not None, \
3739
      "Cannot retrieve locked instance %s" % self.op.instance_name
3740

    
3741
    _CheckNodeOnline(self, instance.primary_node)
3742

    
3743
    # check bridges existence
3744
    _CheckInstanceBridgesExist(self, instance)
3745

    
3746
  def Exec(self, feedback_fn):
3747
    """Reboot the instance.
3748

3749
    """
3750
    instance = self.instance
3751
    ignore_secondaries = self.op.ignore_secondaries
3752
    reboot_type = self.op.reboot_type
3753

    
3754
    node_current = instance.primary_node
3755

    
3756
    if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
3757
                       constants.INSTANCE_REBOOT_HARD]:
3758
      for disk in instance.disks:
3759
        self.cfg.SetDiskID(disk, node_current)
3760
      result = self.rpc.call_instance_reboot(node_current, instance,
3761
                                             reboot_type,
3762
                                             self.shutdown_timeout)
3763
      result.Raise("Could not reboot instance")
3764
    else:
3765
      result = self.rpc.call_instance_shutdown(node_current, instance,
3766
                                               self.shutdown_timeout)
3767
      result.Raise("Could not shutdown instance for full reboot")
3768
      _ShutdownInstanceDisks(self, instance)
3769
      _StartInstanceDisks(self, instance, ignore_secondaries)
3770
      result = self.rpc.call_instance_start(node_current, instance, None, None)
3771
      msg = result.fail_msg
3772
      if msg:
3773
        _ShutdownInstanceDisks(self, instance)
3774
        raise errors.OpExecError("Could not start instance for"
3775
                                 " full reboot: %s" % msg)
3776

    
3777
    self.cfg.MarkInstanceUp(instance.name)
3778

    
3779

    
3780
class LUShutdownInstance(LogicalUnit):
3781
  """Shutdown an instance.
3782

3783
  """
3784
  HPATH = "instance-stop"
3785
  HTYPE = constants.HTYPE_INSTANCE
3786
  _OP_REQP = ["instance_name"]
3787
  REQ_BGL = False
3788

    
3789
  def CheckArguments(self):
3790
    """Check the arguments.
3791

3792
    """
3793
    self.timeout = getattr(self.op, "timeout",
3794
                           constants.DEFAULT_SHUTDOWN_TIMEOUT)
3795

    
3796
  def ExpandNames(self):
3797
    self._ExpandAndLockInstance()
3798

    
3799
  def BuildHooksEnv(self):
3800
    """Build hooks env.
3801

3802
    This runs on master, primary and secondary nodes of the instance.
3803

3804
    """
3805
    env = _BuildInstanceHookEnvByObject(self, self.instance)
3806
    env["TIMEOUT"] = self.timeout
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 = 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
    _CheckNodeOnline(self, self.instance.primary_node)
3820

    
3821
  def Exec(self, feedback_fn):
3822
    """Shutdown the instance.
3823

3824
    """
3825
    instance = self.instance
3826
    node_current = instance.primary_node
3827
    timeout = self.timeout
3828
    self.cfg.MarkInstanceDown(instance.name)
3829
    result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
3830
    msg = result.fail_msg
3831
    if msg:
3832
      self.proc.LogWarning("Could not shutdown instance: %s" % msg)
3833

    
3834
    _ShutdownInstanceDisks(self, instance)
3835

    
3836

    
3837
class LUReinstallInstance(LogicalUnit):
3838
  """Reinstall an instance.
3839

3840
  """
3841
  HPATH = "instance-reinstall"
3842
  HTYPE = constants.HTYPE_INSTANCE
3843
  _OP_REQP = ["instance_name"]
3844
  REQ_BGL = False
3845

    
3846
  def ExpandNames(self):
3847
    self._ExpandAndLockInstance()
3848

    
3849
  def BuildHooksEnv(self):
3850
    """Build hooks env.
3851

3852
    This runs on master, primary and secondary nodes of the instance.
3853

3854
    """
3855
    env = _BuildInstanceHookEnvByObject(self, self.instance)
3856
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3857
    return env, nl, nl
3858

    
3859
  def CheckPrereq(self):
3860
    """Check prerequisites.
3861

3862
    This checks that the instance is in the cluster and is not running.
3863

3864
    """
3865
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3866
    assert instance is not None, \
3867
      "Cannot retrieve locked instance %s" % self.op.instance_name
3868
    _CheckNodeOnline(self, instance.primary_node)
3869

    
3870
    if instance.disk_template == constants.DT_DISKLESS:
3871
      raise errors.OpPrereqError("Instance '%s' has no disks" %
3872
                                 self.op.instance_name,
3873
                                 errors.ECODE_INVAL)
3874
    if instance.admin_up:
3875
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
3876
                                 self.op.instance_name,
3877
                                 errors.ECODE_STATE)
3878
    remote_info = self.rpc.call_instance_info(instance.primary_node,
3879
                                              instance.name,
3880
                                              instance.hypervisor)
3881
    remote_info.Raise("Error checking node %s" % instance.primary_node,
3882
                      prereq=True, ecode=errors.ECODE_ENVIRON)
3883
    if remote_info.payload:
3884
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
3885
                                 (self.op.instance_name,
3886
                                  instance.primary_node),
3887
                                 errors.ECODE_STATE)
3888

    
3889
    self.op.os_type = getattr(self.op, "os_type", None)
3890
    self.op.force_variant = getattr(self.op, "force_variant", False)
3891
    if self.op.os_type is not None:
3892
      # OS verification
3893
      pnode = self.cfg.GetNodeInfo(
3894
        self.cfg.ExpandNodeName(instance.primary_node))
3895
      if pnode is None:
3896
        raise errors.OpPrereqError("Primary node '%s' is unknown" %
3897
                                   self.op.pnode, errors.ECODE_NOENT)
3898
      result = self.rpc.call_os_get(pnode.name, self.op.os_type)
3899
      result.Raise("OS '%s' not in supported OS list for primary node %s" %
3900
                   (self.op.os_type, pnode.name),
3901
                   prereq=True, ecode=errors.ECODE_INVAL)
3902
      if not self.op.force_variant:
3903
        _CheckOSVariant(result.payload, self.op.os_type)
3904

    
3905
    self.instance = instance
3906

    
3907
  def Exec(self, feedback_fn):
3908
    """Reinstall the instance.
3909

3910
    """
3911
    inst = self.instance
3912

    
3913
    if self.op.os_type is not None:
3914
      feedback_fn("Changing OS to '%s'..." % self.op.os_type)
3915
      inst.os = self.op.os_type
3916
      self.cfg.Update(inst, feedback_fn)
3917

    
3918
    _StartInstanceDisks(self, inst, None)
3919
    try:
3920
      feedback_fn("Running the instance OS create scripts...")
3921
      result = self.rpc.call_instance_os_add(inst.primary_node, inst, True)
3922
      result.Raise("Could not install OS for instance %s on node %s" %
3923
                   (inst.name, inst.primary_node))
3924
    finally:
3925
      _ShutdownInstanceDisks(self, inst)
3926

    
3927

    
3928
class LURecreateInstanceDisks(LogicalUnit):
3929
  """Recreate an instance's missing disks.
3930

3931
  """
3932
  HPATH = "instance-recreate-disks"
3933
  HTYPE = constants.HTYPE_INSTANCE
3934
  _OP_REQP = ["instance_name", "disks"]
3935
  REQ_BGL = False
3936

    
3937
  def CheckArguments(self):
3938
    """Check the arguments.
3939

3940
    """
3941
    if not isinstance(self.op.disks, list):
3942
      raise errors.OpPrereqError("Invalid disks parameter", errors.ECODE_INVAL)
3943
    for item in self.op.disks:
3944
      if (not isinstance(item, int) or
3945
          item < 0):
3946
        raise errors.OpPrereqError("Invalid disk specification '%s'" %
3947
                                   str(item), errors.ECODE_INVAL)
3948

    
3949
  def ExpandNames(self):
3950
    self._ExpandAndLockInstance()
3951

    
3952
  def BuildHooksEnv(self):
3953
    """Build hooks env.
3954

3955
    This runs on master, primary and secondary nodes of the instance.
3956

3957
    """
3958
    env = _BuildInstanceHookEnvByObject(self, self.instance)
3959
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3960
    return env, nl, nl
3961

    
3962
  def CheckPrereq(self):
3963
    """Check prerequisites.
3964

3965
    This checks that the instance is in the cluster and is not running.
3966

3967
    """
3968
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3969
    assert instance is not None, \
3970
      "Cannot retrieve locked instance %s" % self.op.instance_name
3971
    _CheckNodeOnline(self, instance.primary_node)
3972

    
3973
    if instance.disk_template == constants.DT_DISKLESS:
3974
      raise errors.OpPrereqError("Instance '%s' has no disks" %
3975
                                 self.op.instance_name, errors.ECODE_INVAL)
3976
    if instance.admin_up:
3977
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
3978
                                 self.op.instance_name, errors.ECODE_STATE)
3979
    remote_info = self.rpc.call_instance_info(instance.primary_node,
3980
                                              instance.name,
3981
                                              instance.hypervisor)
3982
    remote_info.Raise("Error checking node %s" % instance.primary_node,
3983
                      prereq=True, ecode=errors.ECODE_ENVIRON)
3984
    if remote_info.payload:
3985
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
3986
                                 (self.op.instance_name,
3987
                                  instance.primary_node), errors.ECODE_STATE)
3988

    
3989
    if not self.op.disks:
3990
      self.op.disks = range(len(instance.disks))
3991
    else:
3992
      for idx in self.op.disks:
3993
        if idx >= len(instance.disks):
3994
          raise errors.OpPrereqError("Invalid disk index passed '%s'" % idx,
3995
                                     errors.ECODE_INVAL)
3996

    
3997
    self.instance = instance
3998

    
3999
  def Exec(self, feedback_fn):
4000
    """Recreate the disks.
4001

4002