Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ dd9e9f9c

History | View | Annotate | Download (309.7 kB)

1
#
2
#
3

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

    
21

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

    
24
# pylint: disable-msg=W0201
25

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

    
29
import os
30
import os.path
31
import time
32
import re
33
import platform
34
import logging
35
import copy
36

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

    
47

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

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

60
  Note that all commands require root permissions.
61

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

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

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

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

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

    
99
    # Tasklets
100
    self.tasklets = None
101

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

    
108
    self.CheckArguments()
109

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

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

    
118
  ssh = property(fget=__GetSSH)
119

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

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

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

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

135
    """
136
    pass
137

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

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

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

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

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

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

163
    Examples::
164

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

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

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

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

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

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

202
    """
203

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

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

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

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

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

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

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

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

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

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

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

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

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

261
    """
262
    raise NotImplementedError
263

    
264
  def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
265
    """Notify the LU about the results of its hooks.
266

267
    This method is called every time a hooks phase is executed, and notifies
268
    the Logical Unit about the hooks' result. The LU can then use it to alter
269
    its result based on the hooks.  By default the method does nothing and the
270
    previous result is passed back unchanged but any LU can define it if it
271
    wants to use the local cluster hook-scripts somehow.
272

273
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
274
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
275
    @param hook_results: the results of the multi-node hooks rpc call
276
    @param feedback_fn: function used send feedback back to the caller
277
    @param lu_result: the previous Exec result this LU had, or None
278
        in the PRE phase
279
    @return: the new Exec result, based on the previous result
280
        and hook results
281

282
    """
283
    # API must be kept, thus we ignore the unused argument and could
284
    # be a function warnings
285
    # pylint: disable-msg=W0613,R0201
286
    return lu_result
287

    
288
  def _ExpandAndLockInstance(self):
289
    """Helper function to expand and lock an instance.
290

291
    Many LUs that work on an instance take its name in self.op.instance_name
292
    and need to expand it and then declare the expanded name for locking. This
293
    function does it, and then updates self.op.instance_name to the expanded
294
    name. It also initializes needed_locks as a dict, if this hasn't been done
295
    before.
296

297
    """
298
    if self.needed_locks is None:
299
      self.needed_locks = {}
300
    else:
301
      assert locking.LEVEL_INSTANCE not in self.needed_locks, \
302
        "_ExpandAndLockInstance called with instance-level locks set"
303
    expanded_name = self.cfg.ExpandInstanceName(self.op.instance_name)
304
    if expanded_name is None:
305
      raise errors.OpPrereqError("Instance '%s' not known" %
306
                                 self.op.instance_name, errors.ECODE_NOENT)
307
    self.needed_locks[locking.LEVEL_INSTANCE] = expanded_name
308
    self.op.instance_name = expanded_name
309

    
310
  def _LockInstancesNodes(self, primary_only=False):
311
    """Helper function to declare instances' nodes for locking.
312

313
    This function should be called after locking one or more instances to lock
314
    their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
315
    with all primary or secondary nodes for instances already locked and
316
    present in self.needed_locks[locking.LEVEL_INSTANCE].
317

318
    It should be called from DeclareLocks, and for safety only works if
319
    self.recalculate_locks[locking.LEVEL_NODE] is set.
320

321
    In the future it may grow parameters to just lock some instance's nodes, or
322
    to just lock primaries or secondary nodes, if needed.
323

324
    If should be called in DeclareLocks in a way similar to::
325

326
      if level == locking.LEVEL_NODE:
327
        self._LockInstancesNodes()
328

329
    @type primary_only: boolean
330
    @param primary_only: only lock primary nodes of locked instances
331

332
    """
333
    assert locking.LEVEL_NODE in self.recalculate_locks, \
334
      "_LockInstancesNodes helper function called with no nodes to recalculate"
335

    
336
    # TODO: check if we're really been called with the instance locks held
337

    
338
    # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
339
    # future we might want to have different behaviors depending on the value
340
    # of self.recalculate_locks[locking.LEVEL_NODE]
341
    wanted_nodes = []
342
    for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
343
      instance = self.context.cfg.GetInstanceInfo(instance_name)
344
      wanted_nodes.append(instance.primary_node)
345
      if not primary_only:
346
        wanted_nodes.extend(instance.secondary_nodes)
347

    
348
    if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE:
349
      self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
350
    elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND:
351
      self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
352

    
353
    del self.recalculate_locks[locking.LEVEL_NODE]
354

    
355

    
356
class NoHooksLU(LogicalUnit): # pylint: disable-msg=W0223
357
  """Simple LU which runs no hooks.
358

359
  This LU is intended as a parent for other LogicalUnits which will
360
  run no hooks, in order to reduce duplicate code.
361

362
  """
363
  HPATH = None
364
  HTYPE = None
365

    
366
  def BuildHooksEnv(self):
367
    """Empty BuildHooksEnv for NoHooksLu.
368

369
    This just raises an error.
370

371
    """
372
    assert False, "BuildHooksEnv called for NoHooksLUs"
373

    
374

    
375
class Tasklet:
376
  """Tasklet base class.
377

378
  Tasklets are subcomponents for LUs. LUs can consist entirely of tasklets or
379
  they can mix legacy code with tasklets. Locking needs to be done in the LU,
380
  tasklets know nothing about locks.
381

382
  Subclasses must follow these rules:
383
    - Implement CheckPrereq
384
    - Implement Exec
385

386
  """
387
  def __init__(self, lu):
388
    self.lu = lu
389

    
390
    # Shortcuts
391
    self.cfg = lu.cfg
392
    self.rpc = lu.rpc
393

    
394
  def CheckPrereq(self):
395
    """Check prerequisites for this tasklets.
396

397
    This method should check whether the prerequisites for the execution of
398
    this tasklet are fulfilled. It can do internode communication, but it
399
    should be idempotent - no cluster or system changes are allowed.
400

401
    The method should raise errors.OpPrereqError in case something is not
402
    fulfilled. Its return value is ignored.
403

404
    This method should also update all parameters to their canonical form if it
405
    hasn't been done before.
406

407
    """
408
    raise NotImplementedError
409

    
410
  def Exec(self, feedback_fn):
411
    """Execute the tasklet.
412

413
    This method should implement the actual work. It should raise
414
    errors.OpExecError for failures that are somewhat dealt with in code, or
415
    expected.
416

417
    """
418
    raise NotImplementedError
419

    
420

    
421
def _GetWantedNodes(lu, nodes):
422
  """Returns list of checked and expanded node names.
423

424
  @type lu: L{LogicalUnit}
425
  @param lu: the logical unit on whose behalf we execute
426
  @type nodes: list
427
  @param nodes: list of node names or None for all nodes
428
  @rtype: list
429
  @return: the list of nodes, sorted
430
  @raise errors.OpProgrammerError: if the nodes parameter is wrong type
431

432
  """
433
  if not isinstance(nodes, list):
434
    raise errors.OpPrereqError("Invalid argument type 'nodes'",
435
                               errors.ECODE_INVAL)
436

    
437
  if not nodes:
438
    raise errors.ProgrammerError("_GetWantedNodes should only be called with a"
439
      " non-empty list of nodes whose name is to be expanded.")
440

    
441
  wanted = []
442
  for name in nodes:
443
    node = lu.cfg.ExpandNodeName(name)
444
    if node is None:
445
      raise errors.OpPrereqError("No such node name '%s'" % name,
446
                                 errors.ECODE_NOENT)
447
    wanted.append(node)
448

    
449
  return utils.NiceSort(wanted)
450

    
451

    
452
def _GetWantedInstances(lu, instances):
453
  """Returns list of checked and expanded instance names.
454

455
  @type lu: L{LogicalUnit}
456
  @param lu: the logical unit on whose behalf we execute
457
  @type instances: list
458
  @param instances: list of instance names or None for all instances
459
  @rtype: list
460
  @return: the list of instances, sorted
461
  @raise errors.OpPrereqError: if the instances parameter is wrong type
462
  @raise errors.OpPrereqError: if any of the passed instances is not found
463

464
  """
465
  if not isinstance(instances, list):
466
    raise errors.OpPrereqError("Invalid argument type 'instances'",
467
                               errors.ECODE_INVAL)
468

    
469
  if instances:
470
    wanted = []
471

    
472
    for name in instances:
473
      instance = lu.cfg.ExpandInstanceName(name)
474
      if instance is None:
475
        raise errors.OpPrereqError("No such instance name '%s'" % name,
476
                                   errors.ECODE_NOENT)
477
      wanted.append(instance)
478

    
479
  else:
480
    wanted = utils.NiceSort(lu.cfg.GetInstanceList())
481
  return wanted
482

    
483

    
484
def _CheckOutputFields(static, dynamic, selected):
485
  """Checks whether all selected fields are valid.
486

487
  @type static: L{utils.FieldSet}
488
  @param static: static fields set
489
  @type dynamic: L{utils.FieldSet}
490
  @param dynamic: dynamic fields set
491

492
  """
493
  f = utils.FieldSet()
494
  f.Extend(static)
495
  f.Extend(dynamic)
496

    
497
  delta = f.NonMatching(selected)
498
  if delta:
499
    raise errors.OpPrereqError("Unknown output fields selected: %s"
500
                               % ",".join(delta), errors.ECODE_INVAL)
501

    
502

    
503
def _CheckBooleanOpField(op, name):
504
  """Validates boolean opcode parameters.
505

506
  This will ensure that an opcode parameter is either a boolean value,
507
  or None (but that it always exists).
508

509
  """
510
  val = getattr(op, name, None)
511
  if not (val is None or isinstance(val, bool)):
512
    raise errors.OpPrereqError("Invalid boolean parameter '%s' (%s)" %
513
                               (name, str(val)), errors.ECODE_INVAL)
514
  setattr(op, name, val)
515

    
516

    
517
def _CheckGlobalHvParams(params):
518
  """Validates that given hypervisor params are not global ones.
519

520
  This will ensure that instances don't get customised versions of
521
  global params.
522

523
  """
524
  used_globals = constants.HVC_GLOBALS.intersection(params)
525
  if used_globals:
526
    msg = ("The following hypervisor parameters are global and cannot"
527
           " be customized at instance level, please modify them at"
528
           " cluster level: %s" % utils.CommaJoin(used_globals))
529
    raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
530

    
531

    
532
def _CheckNodeOnline(lu, node):
533
  """Ensure that a given node is online.
534

535
  @param lu: the LU on behalf of which we make the check
536
  @param node: the node to check
537
  @raise errors.OpPrereqError: if the node is offline
538

539
  """
540
  if lu.cfg.GetNodeInfo(node).offline:
541
    raise errors.OpPrereqError("Can't use offline node %s" % node,
542
                               errors.ECODE_INVAL)
543

    
544

    
545
def _CheckNodeNotDrained(lu, node):
546
  """Ensure that a given node is not drained.
547

548
  @param lu: the LU on behalf of which we make the check
549
  @param node: the node to check
550
  @raise errors.OpPrereqError: if the node is drained
551

552
  """
553
  if lu.cfg.GetNodeInfo(node).drained:
554
    raise errors.OpPrereqError("Can't use drained node %s" % node,
555
                               errors.ECODE_INVAL)
556

    
557

    
558
def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
559
                          memory, vcpus, nics, disk_template, disks,
560
                          bep, hvp, hypervisor_name):
561
  """Builds instance related env variables for hooks
562

563
  This builds the hook environment from individual variables.
564

565
  @type name: string
566
  @param name: the name of the instance
567
  @type primary_node: string
568
  @param primary_node: the name of the instance's primary node
569
  @type secondary_nodes: list
570
  @param secondary_nodes: list of secondary nodes as strings
571
  @type os_type: string
572
  @param os_type: the name of the instance's OS
573
  @type status: boolean
574
  @param status: the should_run status of the instance
575
  @type memory: string
576
  @param memory: the memory size of the instance
577
  @type vcpus: string
578
  @param vcpus: the count of VCPUs the instance has
579
  @type nics: list
580
  @param nics: list of tuples (ip, mac, mode, link) representing
581
      the NICs the instance has
582
  @type disk_template: string
583
  @param disk_template: the disk template of the instance
584
  @type disks: list
585
  @param disks: the list of (size, mode) pairs
586
  @type bep: dict
587
  @param bep: the backend parameters for the instance
588
  @type hvp: dict
589
  @param hvp: the hypervisor parameters for the instance
590
  @type hypervisor_name: string
591
  @param hypervisor_name: the hypervisor for the instance
592
  @rtype: dict
593
  @return: the hook environment for this instance
594

595
  """
596
  if status:
597
    str_status = "up"
598
  else:
599
    str_status = "down"
600
  env = {
601
    "OP_TARGET": name,
602
    "INSTANCE_NAME": name,
603
    "INSTANCE_PRIMARY": primary_node,
604
    "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
605
    "INSTANCE_OS_TYPE": os_type,
606
    "INSTANCE_STATUS": str_status,
607
    "INSTANCE_MEMORY": memory,
608
    "INSTANCE_VCPUS": vcpus,
609
    "INSTANCE_DISK_TEMPLATE": disk_template,
610
    "INSTANCE_HYPERVISOR": hypervisor_name,
611
  }
612

    
613
  if nics:
614
    nic_count = len(nics)
615
    for idx, (ip, mac, mode, link) in enumerate(nics):
616
      if ip is None:
617
        ip = ""
618
      env["INSTANCE_NIC%d_IP" % idx] = ip
619
      env["INSTANCE_NIC%d_MAC" % idx] = mac
620
      env["INSTANCE_NIC%d_MODE" % idx] = mode
621
      env["INSTANCE_NIC%d_LINK" % idx] = link
622
      if mode == constants.NIC_MODE_BRIDGED:
623
        env["INSTANCE_NIC%d_BRIDGE" % idx] = link
624
  else:
625
    nic_count = 0
626

    
627
  env["INSTANCE_NIC_COUNT"] = nic_count
628

    
629
  if disks:
630
    disk_count = len(disks)
631
    for idx, (size, mode) in enumerate(disks):
632
      env["INSTANCE_DISK%d_SIZE" % idx] = size
633
      env["INSTANCE_DISK%d_MODE" % idx] = mode
634
  else:
635
    disk_count = 0
636

    
637
  env["INSTANCE_DISK_COUNT"] = disk_count
638

    
639
  for source, kind in [(bep, "BE"), (hvp, "HV")]:
640
    for key, value in source.items():
641
      env["INSTANCE_%s_%s" % (kind, key)] = value
642

    
643
  return env
644

    
645

    
646
def _NICListToTuple(lu, nics):
647
  """Build a list of nic information tuples.
648

649
  This list is suitable to be passed to _BuildInstanceHookEnv or as a return
650
  value in LUQueryInstanceData.
651

652
  @type lu:  L{LogicalUnit}
653
  @param lu: the logical unit on whose behalf we execute
654
  @type nics: list of L{objects.NIC}
655
  @param nics: list of nics to convert to hooks tuples
656

657
  """
658
  hooks_nics = []
659
  c_nicparams = lu.cfg.GetClusterInfo().nicparams[constants.PP_DEFAULT]
660
  for nic in nics:
661
    ip = nic.ip
662
    mac = nic.mac
663
    filled_params = objects.FillDict(c_nicparams, nic.nicparams)
664
    mode = filled_params[constants.NIC_MODE]
665
    link = filled_params[constants.NIC_LINK]
666
    hooks_nics.append((ip, mac, mode, link))
667
  return hooks_nics
668

    
669

    
670
def _BuildInstanceHookEnvByObject(lu, instance, override=None):
671
  """Builds instance related env variables for hooks from an object.
672

673
  @type lu: L{LogicalUnit}
674
  @param lu: the logical unit on whose behalf we execute
675
  @type instance: L{objects.Instance}
676
  @param instance: the instance for which we should build the
677
      environment
678
  @type override: dict
679
  @param override: dictionary with key/values that will override
680
      our values
681
  @rtype: dict
682
  @return: the hook environment dictionary
683

684
  """
685
  cluster = lu.cfg.GetClusterInfo()
686
  bep = cluster.FillBE(instance)
687
  hvp = cluster.FillHV(instance)
688
  args = {
689
    'name': instance.name,
690
    'primary_node': instance.primary_node,
691
    'secondary_nodes': instance.secondary_nodes,
692
    'os_type': instance.os,
693
    'status': instance.admin_up,
694
    'memory': bep[constants.BE_MEMORY],
695
    'vcpus': bep[constants.BE_VCPUS],
696
    'nics': _NICListToTuple(lu, instance.nics),
697
    'disk_template': instance.disk_template,
698
    'disks': [(disk.size, disk.mode) for disk in instance.disks],
699
    'bep': bep,
700
    'hvp': hvp,
701
    'hypervisor_name': instance.hypervisor,
702
  }
703
  if override:
704
    args.update(override)
705
  return _BuildInstanceHookEnv(**args) # pylint: disable-msg=W0142
706

    
707

    
708
def _AdjustCandidatePool(lu, exceptions):
709
  """Adjust the candidate pool after node operations.
710

711
  """
712
  mod_list = lu.cfg.MaintainCandidatePool(exceptions)
713
  if mod_list:
714
    lu.LogInfo("Promoted nodes to master candidate role: %s",
715
               utils.CommaJoin(node.name for node in mod_list))
716
    for name in mod_list:
717
      lu.context.ReaddNode(name)
718
  mc_now, mc_max, _ = lu.cfg.GetMasterCandidateStats(exceptions)
719
  if mc_now > mc_max:
720
    lu.LogInfo("Note: more nodes are candidates (%d) than desired (%d)" %
721
               (mc_now, mc_max))
722

    
723

    
724
def _DecideSelfPromotion(lu, exceptions=None):
725
  """Decide whether I should promote myself as a master candidate.
726

727
  """
728
  cp_size = lu.cfg.GetClusterInfo().candidate_pool_size
729
  mc_now, mc_should, _ = lu.cfg.GetMasterCandidateStats(exceptions)
730
  # the new node will increase mc_max with one, so:
731
  mc_should = min(mc_should + 1, cp_size)
732
  return mc_now < mc_should
733

    
734

    
735
def _CheckNicsBridgesExist(lu, target_nics, target_node,
736
                               profile=constants.PP_DEFAULT):
737
  """Check that the brigdes needed by a list of nics exist.
738

739
  """
740
  c_nicparams = lu.cfg.GetClusterInfo().nicparams[profile]
741
  paramslist = [objects.FillDict(c_nicparams, nic.nicparams)
742
                for nic in target_nics]
743
  brlist = [params[constants.NIC_LINK] for params in paramslist
744
            if params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED]
745
  if brlist:
746
    result = lu.rpc.call_bridges_exist(target_node, brlist)
747
    result.Raise("Error checking bridges on destination node '%s'" %
748
                 target_node, prereq=True, ecode=errors.ECODE_ENVIRON)
749

    
750

    
751
def _CheckInstanceBridgesExist(lu, instance, node=None):
752
  """Check that the brigdes needed by an instance exist.
753

754
  """
755
  if node is None:
756
    node = instance.primary_node
757
  _CheckNicsBridgesExist(lu, instance.nics, node)
758

    
759

    
760
def _CheckOSVariant(os_obj, name):
761
  """Check whether an OS name conforms to the os variants specification.
762

763
  @type os_obj: L{objects.OS}
764
  @param os_obj: OS object to check
765
  @type name: string
766
  @param name: OS name passed by the user, to check for validity
767

768
  """
769
  if not os_obj.supported_variants:
770
    return
771
  try:
772
    variant = name.split("+", 1)[1]
773
  except IndexError:
774
    raise errors.OpPrereqError("OS name must include a variant",
775
                               errors.ECODE_INVAL)
776

    
777
  if variant not in os_obj.supported_variants:
778
    raise errors.OpPrereqError("Unsupported OS variant", errors.ECODE_INVAL)
779

    
780

    
781
def _GetNodeInstancesInner(cfg, fn):
782
  return [i for i in cfg.GetAllInstancesInfo().values() if fn(i)]
783

    
784

    
785
def _GetNodeInstances(cfg, node_name):
786
  """Returns a list of all primary and secondary instances on a node.
787

788
  """
789

    
790
  return _GetNodeInstancesInner(cfg, lambda inst: node_name in inst.all_nodes)
791

    
792

    
793
def _GetNodePrimaryInstances(cfg, node_name):
794
  """Returns primary instances on a node.
795

796
  """
797
  return _GetNodeInstancesInner(cfg,
798
                                lambda inst: node_name == inst.primary_node)
799

    
800

    
801
def _GetNodeSecondaryInstances(cfg, node_name):
802
  """Returns secondary instances on a node.
803

804
  """
805
  return _GetNodeInstancesInner(cfg,
806
                                lambda inst: node_name in inst.secondary_nodes)
807

    
808

    
809
def _GetStorageTypeArgs(cfg, storage_type):
810
  """Returns the arguments for a storage type.
811

812
  """
813
  # Special case for file storage
814
  if storage_type == constants.ST_FILE:
815
    # storage.FileStorage wants a list of storage directories
816
    return [[cfg.GetFileStorageDir()]]
817

    
818
  return []
819

    
820

    
821
def _FindFaultyInstanceDisks(cfg, rpc, instance, node_name, prereq):
822
  faulty = []
823

    
824
  for dev in instance.disks:
825
    cfg.SetDiskID(dev, node_name)
826

    
827
  result = rpc.call_blockdev_getmirrorstatus(node_name, instance.disks)
828
  result.Raise("Failed to get disk status from node %s" % node_name,
829
               prereq=prereq, ecode=errors.ECODE_ENVIRON)
830

    
831
  for idx, bdev_status in enumerate(result.payload):
832
    if bdev_status and bdev_status.ldisk_status == constants.LDS_FAULTY:
833
      faulty.append(idx)
834

    
835
  return faulty
836

    
837

    
838
class LUPostInitCluster(LogicalUnit):
839
  """Logical unit for running hooks after cluster initialization.
840

841
  """
842
  HPATH = "cluster-init"
843
  HTYPE = constants.HTYPE_CLUSTER
844
  _OP_REQP = []
845

    
846
  def BuildHooksEnv(self):
847
    """Build hooks env.
848

849
    """
850
    env = {"OP_TARGET": self.cfg.GetClusterName()}
851
    mn = self.cfg.GetMasterNode()
852
    return env, [], [mn]
853

    
854
  def CheckPrereq(self):
855
    """No prerequisites to check.
856

857
    """
858
    return True
859

    
860
  def Exec(self, feedback_fn):
861
    """Nothing to do.
862

863
    """
864
    return True
865

    
866

    
867
class LUDestroyCluster(LogicalUnit):
868
  """Logical unit for destroying the cluster.
869

870
  """
871
  HPATH = "cluster-destroy"
872
  HTYPE = constants.HTYPE_CLUSTER
873
  _OP_REQP = []
874

    
875
  def BuildHooksEnv(self):
876
    """Build hooks env.
877

878
    """
879
    env = {"OP_TARGET": self.cfg.GetClusterName()}
880
    return env, [], []
881

    
882
  def CheckPrereq(self):
883
    """Check prerequisites.
884

885
    This checks whether the cluster is empty.
886

887
    Any errors are signaled by raising errors.OpPrereqError.
888

889
    """
890
    master = self.cfg.GetMasterNode()
891

    
892
    nodelist = self.cfg.GetNodeList()
893
    if len(nodelist) != 1 or nodelist[0] != master:
894
      raise errors.OpPrereqError("There are still %d node(s) in"
895
                                 " this cluster." % (len(nodelist) - 1),
896
                                 errors.ECODE_INVAL)
897
    instancelist = self.cfg.GetInstanceList()
898
    if instancelist:
899
      raise errors.OpPrereqError("There are still %d instance(s) in"
900
                                 " this cluster." % len(instancelist),
901
                                 errors.ECODE_INVAL)
902

    
903
  def Exec(self, feedback_fn):
904
    """Destroys the cluster.
905

906
    """
907
    master = self.cfg.GetMasterNode()
908
    modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
909

    
910
    # Run post hooks on master node before it's removed
911
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
912
    try:
913
      hm.RunPhase(constants.HOOKS_PHASE_POST, [master])
914
    except:
915
      # pylint: disable-msg=W0702
916
      self.LogWarning("Errors occurred running hooks on %s" % master)
917

    
918
    result = self.rpc.call_node_stop_master(master, False)
919
    result.Raise("Could not disable the master role")
920

    
921
    if modify_ssh_setup:
922
      priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
923
      utils.CreateBackup(priv_key)
924
      utils.CreateBackup(pub_key)
925

    
926
    return master
927

    
928

    
929
class LUVerifyCluster(LogicalUnit):
930
  """Verifies the cluster status.
931

932
  """
933
  HPATH = "cluster-verify"
934
  HTYPE = constants.HTYPE_CLUSTER
935
  _OP_REQP = ["skip_checks", "verbose", "error_codes", "debug_simulate_errors"]
936
  REQ_BGL = False
937

    
938
  TCLUSTER = "cluster"
939
  TNODE = "node"
940
  TINSTANCE = "instance"
941

    
942
  ECLUSTERCFG = (TCLUSTER, "ECLUSTERCFG")
943
  EINSTANCEBADNODE = (TINSTANCE, "EINSTANCEBADNODE")
944
  EINSTANCEDOWN = (TINSTANCE, "EINSTANCEDOWN")
945
  EINSTANCELAYOUT = (TINSTANCE, "EINSTANCELAYOUT")
946
  EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
947
  EINSTANCEMISSINGDISK = (TINSTANCE, "EINSTANCEMISSINGDISK")
948
  EINSTANCEWRONGNODE = (TINSTANCE, "EINSTANCEWRONGNODE")
949
  ENODEDRBD = (TNODE, "ENODEDRBD")
950
  ENODEFILECHECK = (TNODE, "ENODEFILECHECK")
951
  ENODEHOOKS = (TNODE, "ENODEHOOKS")
952
  ENODEHV = (TNODE, "ENODEHV")
953
  ENODELVM = (TNODE, "ENODELVM")
954
  ENODEN1 = (TNODE, "ENODEN1")
955
  ENODENET = (TNODE, "ENODENET")
956
  ENODEORPHANINSTANCE = (TNODE, "ENODEORPHANINSTANCE")
957
  ENODEORPHANLV = (TNODE, "ENODEORPHANLV")
958
  ENODERPC = (TNODE, "ENODERPC")
959
  ENODESSH = (TNODE, "ENODESSH")
960
  ENODEVERSION = (TNODE, "ENODEVERSION")
961
  ENODESETUP = (TNODE, "ENODESETUP")
962
  ENODETIME = (TNODE, "ENODETIME")
963

    
964
  ETYPE_FIELD = "code"
965
  ETYPE_ERROR = "ERROR"
966
  ETYPE_WARNING = "WARNING"
967

    
968
  def ExpandNames(self):
969
    self.needed_locks = {
970
      locking.LEVEL_NODE: locking.ALL_SET,
971
      locking.LEVEL_INSTANCE: locking.ALL_SET,
972
    }
973
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
974

    
975
  def _Error(self, ecode, item, msg, *args, **kwargs):
976
    """Format an error message.
977

978
    Based on the opcode's error_codes parameter, either format a
979
    parseable error code, or a simpler error string.
980

981
    This must be called only from Exec and functions called from Exec.
982

983
    """
984
    ltype = kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR)
985
    itype, etxt = ecode
986
    # first complete the msg
987
    if args:
988
      msg = msg % args
989
    # then format the whole message
990
    if self.op.error_codes:
991
      msg = "%s:%s:%s:%s:%s" % (ltype, etxt, itype, item, msg)
992
    else:
993
      if item:
994
        item = " " + item
995
      else:
996
        item = ""
997
      msg = "%s: %s%s: %s" % (ltype, itype, item, msg)
998
    # and finally report it via the feedback_fn
999
    self._feedback_fn("  - %s" % msg)
1000

    
1001
  def _ErrorIf(self, cond, *args, **kwargs):
1002
    """Log an error message if the passed condition is True.
1003

1004
    """
1005
    cond = bool(cond) or self.op.debug_simulate_errors
1006
    if cond:
1007
      self._Error(*args, **kwargs)
1008
    # do not mark the operation as failed for WARN cases only
1009
    if kwargs.get(self.ETYPE_FIELD, self.ETYPE_ERROR) == self.ETYPE_ERROR:
1010
      self.bad = self.bad or cond
1011

    
1012
  def _VerifyNode(self, nodeinfo, file_list, local_cksum,
1013
                  node_result, master_files, drbd_map, vg_name):
1014
    """Run multiple tests against a node.
1015

1016
    Test list:
1017

1018
      - compares ganeti version
1019
      - checks vg existence and size > 20G
1020
      - checks config file checksum
1021
      - checks ssh to other nodes
1022

1023
    @type nodeinfo: L{objects.Node}
1024
    @param nodeinfo: the node to check
1025
    @param file_list: required list of files
1026
    @param local_cksum: dictionary of local files and their checksums
1027
    @param node_result: the results from the node
1028
    @param master_files: list of files that only masters should have
1029
    @param drbd_map: the useddrbd minors for this node, in
1030
        form of minor: (instance, must_exist) which correspond to instances
1031
        and their running status
1032
    @param vg_name: Ganeti Volume Group (result of self.cfg.GetVGName())
1033

1034
    """
1035
    node = nodeinfo.name
1036
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1037

    
1038
    # main result, node_result should be a non-empty dict
1039
    test = not node_result or not isinstance(node_result, dict)
1040
    _ErrorIf(test, self.ENODERPC, node,
1041
                  "unable to verify node: no data returned")
1042
    if test:
1043
      return
1044

    
1045
    # compares ganeti version
1046
    local_version = constants.PROTOCOL_VERSION
1047
    remote_version = node_result.get('version', None)
1048
    test = not (remote_version and
1049
                isinstance(remote_version, (list, tuple)) and
1050
                len(remote_version) == 2)
1051
    _ErrorIf(test, self.ENODERPC, node,
1052
             "connection to node returned invalid data")
1053
    if test:
1054
      return
1055

    
1056
    test = local_version != remote_version[0]
1057
    _ErrorIf(test, self.ENODEVERSION, node,
1058
             "incompatible protocol versions: master %s,"
1059
             " node %s", local_version, remote_version[0])
1060
    if test:
1061
      return
1062

    
1063
    # node seems compatible, we can actually try to look into its results
1064

    
1065
    # full package version
1066
    self._ErrorIf(constants.RELEASE_VERSION != remote_version[1],
1067
                  self.ENODEVERSION, node,
1068
                  "software version mismatch: master %s, node %s",
1069
                  constants.RELEASE_VERSION, remote_version[1],
1070
                  code=self.ETYPE_WARNING)
1071

    
1072
    # checks vg existence and size > 20G
1073
    if vg_name is not None:
1074
      vglist = node_result.get(constants.NV_VGLIST, None)
1075
      test = not vglist
1076
      _ErrorIf(test, self.ENODELVM, node, "unable to check volume groups")
1077
      if not test:
1078
        vgstatus = utils.CheckVolumeGroupSize(vglist, vg_name,
1079
                                              constants.MIN_VG_SIZE)
1080
        _ErrorIf(vgstatus, self.ENODELVM, node, vgstatus)
1081

    
1082
    # checks config file checksum
1083

    
1084
    remote_cksum = node_result.get(constants.NV_FILELIST, None)
1085
    test = not isinstance(remote_cksum, dict)
1086
    _ErrorIf(test, self.ENODEFILECHECK, node,
1087
             "node hasn't returned file checksum data")
1088
    if not test:
1089
      for file_name in file_list:
1090
        node_is_mc = nodeinfo.master_candidate
1091
        must_have = (file_name not in master_files) or node_is_mc
1092
        # missing
1093
        test1 = file_name not in remote_cksum
1094
        # invalid checksum
1095
        test2 = not test1 and remote_cksum[file_name] != local_cksum[file_name]
1096
        # existing and good
1097
        test3 = not test1 and remote_cksum[file_name] == local_cksum[file_name]
1098
        _ErrorIf(test1 and must_have, self.ENODEFILECHECK, node,
1099
                 "file '%s' missing", file_name)
1100
        _ErrorIf(test2 and must_have, self.ENODEFILECHECK, node,
1101
                 "file '%s' has wrong checksum", file_name)
1102
        # not candidate and this is not a must-have file
1103
        _ErrorIf(test2 and not must_have, self.ENODEFILECHECK, node,
1104
                 "file '%s' should not exist on non master"
1105
                 " candidates (and the file is outdated)", file_name)
1106
        # all good, except non-master/non-must have combination
1107
        _ErrorIf(test3 and not must_have, self.ENODEFILECHECK, node,
1108
                 "file '%s' should not exist"
1109
                 " on non master candidates", file_name)
1110

    
1111
    # checks ssh to any
1112

    
1113
    test = constants.NV_NODELIST not in node_result
1114
    _ErrorIf(test, self.ENODESSH, node,
1115
             "node hasn't returned node ssh connectivity data")
1116
    if not test:
1117
      if node_result[constants.NV_NODELIST]:
1118
        for a_node, a_msg in node_result[constants.NV_NODELIST].items():
1119
          _ErrorIf(True, self.ENODESSH, node,
1120
                   "ssh communication with node '%s': %s", a_node, a_msg)
1121

    
1122
    test = constants.NV_NODENETTEST not in node_result
1123
    _ErrorIf(test, self.ENODENET, node,
1124
             "node hasn't returned node tcp connectivity data")
1125
    if not test:
1126
      if node_result[constants.NV_NODENETTEST]:
1127
        nlist = utils.NiceSort(node_result[constants.NV_NODENETTEST].keys())
1128
        for anode in nlist:
1129
          _ErrorIf(True, self.ENODENET, node,
1130
                   "tcp communication with node '%s': %s",
1131
                   anode, node_result[constants.NV_NODENETTEST][anode])
1132

    
1133
    hyp_result = node_result.get(constants.NV_HYPERVISOR, None)
1134
    if isinstance(hyp_result, dict):
1135
      for hv_name, hv_result in hyp_result.iteritems():
1136
        test = hv_result is not None
1137
        _ErrorIf(test, self.ENODEHV, node,
1138
                 "hypervisor %s verify failure: '%s'", hv_name, hv_result)
1139

    
1140
    # check used drbd list
1141
    if vg_name is not None:
1142
      used_minors = node_result.get(constants.NV_DRBDLIST, [])
1143
      test = not isinstance(used_minors, (tuple, list))
1144
      _ErrorIf(test, self.ENODEDRBD, node,
1145
               "cannot parse drbd status file: %s", str(used_minors))
1146
      if not test:
1147
        for minor, (iname, must_exist) in drbd_map.items():
1148
          test = minor not in used_minors and must_exist
1149
          _ErrorIf(test, self.ENODEDRBD, node,
1150
                   "drbd minor %d of instance %s is not active",
1151
                   minor, iname)
1152
        for minor in used_minors:
1153
          test = minor not in drbd_map
1154
          _ErrorIf(test, self.ENODEDRBD, node,
1155
                   "unallocated drbd minor %d is in use", minor)
1156
    test = node_result.get(constants.NV_NODESETUP,
1157
                           ["Missing NODESETUP results"])
1158
    _ErrorIf(test, self.ENODESETUP, node, "node setup error: %s",
1159
             "; ".join(test))
1160

    
1161
    # check pv names
1162
    if vg_name is not None:
1163
      pvlist = node_result.get(constants.NV_PVLIST, None)
1164
      test = pvlist is None
1165
      _ErrorIf(test, self.ENODELVM, node, "Can't get PV list from node")
1166
      if not test:
1167
        # check that ':' is not present in PV names, since it's a
1168
        # special character for lvcreate (denotes the range of PEs to
1169
        # use on the PV)
1170
        for _, pvname, owner_vg in pvlist:
1171
          test = ":" in pvname
1172
          _ErrorIf(test, self.ENODELVM, node, "Invalid character ':' in PV"
1173
                   " '%s' of VG '%s'", pvname, owner_vg)
1174

    
1175
  def _VerifyInstance(self, instance, instanceconfig, node_vol_is,
1176
                      node_instance, n_offline):
1177
    """Verify an instance.
1178

1179
    This function checks to see if the required block devices are
1180
    available on the instance's node.
1181

1182
    """
1183
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1184
    node_current = instanceconfig.primary_node
1185

    
1186
    node_vol_should = {}
1187
    instanceconfig.MapLVsByNode(node_vol_should)
1188

    
1189
    for node in node_vol_should:
1190
      if node in n_offline:
1191
        # ignore missing volumes on offline nodes
1192
        continue
1193
      for volume in node_vol_should[node]:
1194
        test = node not in node_vol_is or volume not in node_vol_is[node]
1195
        _ErrorIf(test, self.EINSTANCEMISSINGDISK, instance,
1196
                 "volume %s missing on node %s", volume, node)
1197

    
1198
    if instanceconfig.admin_up:
1199
      test = ((node_current not in node_instance or
1200
               not instance in node_instance[node_current]) and
1201
              node_current not in n_offline)
1202
      _ErrorIf(test, self.EINSTANCEDOWN, instance,
1203
               "instance not running on its primary node %s",
1204
               node_current)
1205

    
1206
    for node in node_instance:
1207
      if (not node == node_current):
1208
        test = instance in node_instance[node]
1209
        _ErrorIf(test, self.EINSTANCEWRONGNODE, instance,
1210
                 "instance should not run on node %s", node)
1211

    
1212
  def _VerifyOrphanVolumes(self, node_vol_should, node_vol_is):
1213
    """Verify if there are any unknown volumes in the cluster.
1214

1215
    The .os, .swap and backup volumes are ignored. All other volumes are
1216
    reported as unknown.
1217

1218
    """
1219
    for node in node_vol_is:
1220
      for volume in node_vol_is[node]:
1221
        test = (node not in node_vol_should or
1222
                volume not in node_vol_should[node])
1223
        self._ErrorIf(test, self.ENODEORPHANLV, node,
1224
                      "volume %s is unknown", volume)
1225

    
1226
  def _VerifyOrphanInstances(self, instancelist, node_instance):
1227
    """Verify the list of running instances.
1228

1229
    This checks what instances are running but unknown to the cluster.
1230

1231
    """
1232
    for node in node_instance:
1233
      for o_inst in node_instance[node]:
1234
        test = o_inst not in instancelist
1235
        self._ErrorIf(test, self.ENODEORPHANINSTANCE, node,
1236
                      "instance %s on node %s should not exist", o_inst, node)
1237

    
1238
  def _VerifyNPlusOneMemory(self, node_info, instance_cfg):
1239
    """Verify N+1 Memory Resilience.
1240

1241
    Check that if one single node dies we can still start all the instances it
1242
    was primary for.
1243

1244
    """
1245
    for node, nodeinfo in node_info.iteritems():
1246
      # This code checks that every node which is now listed as secondary has
1247
      # enough memory to host all instances it is supposed to should a single
1248
      # other node in the cluster fail.
1249
      # FIXME: not ready for failover to an arbitrary node
1250
      # FIXME: does not support file-backed instances
1251
      # WARNING: we currently take into account down instances as well as up
1252
      # ones, considering that even if they're down someone might want to start
1253
      # them even in the event of a node failure.
1254
      for prinode, instances in nodeinfo['sinst-by-pnode'].iteritems():
1255
        needed_mem = 0
1256
        for instance in instances:
1257
          bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
1258
          if bep[constants.BE_AUTO_BALANCE]:
1259
            needed_mem += bep[constants.BE_MEMORY]
1260
        test = nodeinfo['mfree'] < needed_mem
1261
        self._ErrorIf(test, self.ENODEN1, node,
1262
                      "not enough memory on to accommodate"
1263
                      " failovers should peer node %s fail", prinode)
1264

    
1265
  def CheckPrereq(self):
1266
    """Check prerequisites.
1267

1268
    Transform the list of checks we're going to skip into a set and check that
1269
    all its members are valid.
1270

1271
    """
1272
    self.skip_set = frozenset(self.op.skip_checks)
1273
    if not constants.VERIFY_OPTIONAL_CHECKS.issuperset(self.skip_set):
1274
      raise errors.OpPrereqError("Invalid checks to be skipped specified",
1275
                                 errors.ECODE_INVAL)
1276

    
1277
  def BuildHooksEnv(self):
1278
    """Build hooks env.
1279

1280
    Cluster-Verify hooks just ran in the post phase and their failure makes
1281
    the output be logged in the verify output and the verification to fail.
1282

1283
    """
1284
    all_nodes = self.cfg.GetNodeList()
1285
    env = {
1286
      "CLUSTER_TAGS": " ".join(self.cfg.GetClusterInfo().GetTags())
1287
      }
1288
    for node in self.cfg.GetAllNodesInfo().values():
1289
      env["NODE_TAGS_%s" % node.name] = " ".join(node.GetTags())
1290

    
1291
    return env, [], all_nodes
1292

    
1293
  def Exec(self, feedback_fn):
1294
    """Verify integrity of cluster, performing various test on nodes.
1295

1296
    """
1297
    self.bad = False
1298
    _ErrorIf = self._ErrorIf # pylint: disable-msg=C0103
1299
    verbose = self.op.verbose
1300
    self._feedback_fn = feedback_fn
1301
    feedback_fn("* Verifying global settings")
1302
    for msg in self.cfg.VerifyConfig():
1303
      _ErrorIf(True, self.ECLUSTERCFG, None, msg)
1304

    
1305
    vg_name = self.cfg.GetVGName()
1306
    hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
1307
    nodelist = utils.NiceSort(self.cfg.GetNodeList())
1308
    nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
1309
    instancelist = utils.NiceSort(self.cfg.GetInstanceList())
1310
    instanceinfo = dict((iname, self.cfg.GetInstanceInfo(iname))
1311
                        for iname in instancelist)
1312
    i_non_redundant = [] # Non redundant instances
1313
    i_non_a_balanced = [] # Non auto-balanced instances
1314
    n_offline = [] # List of offline nodes
1315
    n_drained = [] # List of nodes being drained
1316
    node_volume = {}
1317
    node_instance = {}
1318
    node_info = {}
1319
    instance_cfg = {}
1320

    
1321
    # FIXME: verify OS list
1322
    # do local checksums
1323
    master_files = [constants.CLUSTER_CONF_FILE]
1324

    
1325
    file_names = ssconf.SimpleStore().GetFileList()
1326
    file_names.append(constants.SSL_CERT_FILE)
1327
    file_names.append(constants.RAPI_CERT_FILE)
1328
    file_names.extend(master_files)
1329

    
1330
    local_checksums = utils.FingerprintFiles(file_names)
1331

    
1332
    feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
1333
    node_verify_param = {
1334
      constants.NV_FILELIST: file_names,
1335
      constants.NV_NODELIST: [node.name for node in nodeinfo
1336
                              if not node.offline],
1337
      constants.NV_HYPERVISOR: hypervisors,
1338
      constants.NV_NODENETTEST: [(node.name, node.primary_ip,
1339
                                  node.secondary_ip) for node in nodeinfo
1340
                                 if not node.offline],
1341
      constants.NV_INSTANCELIST: hypervisors,
1342
      constants.NV_VERSION: None,
1343
      constants.NV_HVINFO: self.cfg.GetHypervisorType(),
1344
      constants.NV_NODESETUP: None,
1345
      constants.NV_TIME: None,
1346
      }
1347

    
1348
    if vg_name is not None:
1349
      node_verify_param[constants.NV_VGLIST] = None
1350
      node_verify_param[constants.NV_LVLIST] = vg_name
1351
      node_verify_param[constants.NV_PVLIST] = [vg_name]
1352
      node_verify_param[constants.NV_DRBDLIST] = None
1353

    
1354
    # Due to the way our RPC system works, exact response times cannot be
1355
    # guaranteed (e.g. a broken node could run into a timeout). By keeping the
1356
    # time before and after executing the request, we can at least have a time
1357
    # window.
1358
    nvinfo_starttime = time.time()
1359
    all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
1360
                                           self.cfg.GetClusterName())
1361
    nvinfo_endtime = time.time()
1362

    
1363
    cluster = self.cfg.GetClusterInfo()
1364
    master_node = self.cfg.GetMasterNode()
1365
    all_drbd_map = self.cfg.ComputeDRBDMap()
1366

    
1367
    feedback_fn("* Verifying node status")
1368
    for node_i in nodeinfo:
1369
      node = node_i.name
1370

    
1371
      if node_i.offline:
1372
        if verbose:
1373
          feedback_fn("* Skipping offline node %s" % (node,))
1374
        n_offline.append(node)
1375
        continue
1376

    
1377
      if node == master_node:
1378
        ntype = "master"
1379
      elif node_i.master_candidate:
1380
        ntype = "master candidate"
1381
      elif node_i.drained:
1382
        ntype = "drained"
1383
        n_drained.append(node)
1384
      else:
1385
        ntype = "regular"
1386
      if verbose:
1387
        feedback_fn("* Verifying node %s (%s)" % (node, ntype))
1388

    
1389
      msg = all_nvinfo[node].fail_msg
1390
      _ErrorIf(msg, self.ENODERPC, node, "while contacting node: %s", msg)
1391
      if msg:
1392
        continue
1393

    
1394
      nresult = all_nvinfo[node].payload
1395
      node_drbd = {}
1396
      for minor, instance in all_drbd_map[node].items():
1397
        test = instance not in instanceinfo
1398
        _ErrorIf(test, self.ECLUSTERCFG, None,
1399
                 "ghost instance '%s' in temporary DRBD map", instance)
1400
          # ghost instance should not be running, but otherwise we
1401
          # don't give double warnings (both ghost instance and
1402
          # unallocated minor in use)
1403
        if test:
1404
          node_drbd[minor] = (instance, False)
1405
        else:
1406
          instance = instanceinfo[instance]
1407
          node_drbd[minor] = (instance.name, instance.admin_up)
1408

    
1409
      self._VerifyNode(node_i, file_names, local_checksums,
1410
                       nresult, master_files, node_drbd, vg_name)
1411

    
1412
      lvdata = nresult.get(constants.NV_LVLIST, "Missing LV data")
1413
      if vg_name is None:
1414
        node_volume[node] = {}
1415
      elif isinstance(lvdata, basestring):
1416
        _ErrorIf(True, self.ENODELVM, node, "LVM problem on node: %s",
1417
                 utils.SafeEncode(lvdata))
1418
        node_volume[node] = {}
1419
      elif not isinstance(lvdata, dict):
1420
        _ErrorIf(True, self.ENODELVM, node, "rpc call to node failed (lvlist)")
1421
        continue
1422
      else:
1423
        node_volume[node] = lvdata
1424

    
1425
      # node_instance
1426
      idata = nresult.get(constants.NV_INSTANCELIST, None)
1427
      test = not isinstance(idata, list)
1428
      _ErrorIf(test, self.ENODEHV, node,
1429
               "rpc call to node failed (instancelist)")
1430
      if test:
1431
        continue
1432

    
1433
      node_instance[node] = idata
1434

    
1435
      # node_info
1436
      nodeinfo = nresult.get(constants.NV_HVINFO, None)
1437
      test = not isinstance(nodeinfo, dict)
1438
      _ErrorIf(test, self.ENODEHV, node, "rpc call to node failed (hvinfo)")
1439
      if test:
1440
        continue
1441

    
1442
      # Node time
1443
      ntime = nresult.get(constants.NV_TIME, None)
1444
      try:
1445
        ntime_merged = utils.MergeTime(ntime)
1446
      except (ValueError, TypeError):
1447
        _ErrorIf(test, self.ENODETIME, node, "Node returned invalid time")
1448

    
1449
      if ntime_merged < (nvinfo_starttime - constants.NODE_MAX_CLOCK_SKEW):
1450
        ntime_diff = abs(nvinfo_starttime - ntime_merged)
1451
      elif ntime_merged > (nvinfo_endtime + constants.NODE_MAX_CLOCK_SKEW):
1452
        ntime_diff = abs(ntime_merged - nvinfo_endtime)
1453
      else:
1454
        ntime_diff = None
1455

    
1456
      _ErrorIf(ntime_diff is not None, self.ENODETIME, node,
1457
               "Node time diverges by at least %0.1fs from master node time",
1458
               ntime_diff)
1459

    
1460
      if ntime_diff is not None:
1461
        continue
1462

    
1463
      try:
1464
        node_info[node] = {
1465
          "mfree": int(nodeinfo['memory_free']),
1466
          "pinst": [],
1467
          "sinst": [],
1468
          # dictionary holding all instances this node is secondary for,
1469
          # grouped by their primary node. Each key is a cluster node, and each
1470
          # value is a list of instances which have the key as primary and the
1471
          # current node as secondary.  this is handy to calculate N+1 memory
1472
          # availability if you can only failover from a primary to its
1473
          # secondary.
1474
          "sinst-by-pnode": {},
1475
        }
1476
        # FIXME: devise a free space model for file based instances as well
1477
        if vg_name is not None:
1478
          test = (constants.NV_VGLIST not in nresult or
1479
                  vg_name not in nresult[constants.NV_VGLIST])
1480
          _ErrorIf(test, self.ENODELVM, node,
1481
                   "node didn't return data for the volume group '%s'"
1482
                   " - it is either missing or broken", vg_name)
1483
          if test:
1484
            continue
1485
          node_info[node]["dfree"] = int(nresult[constants.NV_VGLIST][vg_name])
1486
      except (ValueError, KeyError):
1487
        _ErrorIf(True, self.ENODERPC, node,
1488
                 "node returned invalid nodeinfo, check lvm/hypervisor")
1489
        continue
1490

    
1491
    node_vol_should = {}
1492

    
1493
    feedback_fn("* Verifying instance status")
1494
    for instance in instancelist:
1495
      if verbose:
1496
        feedback_fn("* Verifying instance %s" % instance)
1497
      inst_config = instanceinfo[instance]
1498
      self._VerifyInstance(instance, inst_config, node_volume,
1499
                           node_instance, n_offline)
1500
      inst_nodes_offline = []
1501

    
1502
      inst_config.MapLVsByNode(node_vol_should)
1503

    
1504
      instance_cfg[instance] = inst_config
1505

    
1506
      pnode = inst_config.primary_node
1507
      _ErrorIf(pnode not in node_info and pnode not in n_offline,
1508
               self.ENODERPC, pnode, "instance %s, connection to"
1509
               " primary node failed", instance)
1510
      if pnode in node_info:
1511
        node_info[pnode]['pinst'].append(instance)
1512

    
1513
      if pnode in n_offline:
1514
        inst_nodes_offline.append(pnode)
1515

    
1516
      # If the instance is non-redundant we cannot survive losing its primary
1517
      # node, so we are not N+1 compliant. On the other hand we have no disk
1518
      # templates with more than one secondary so that situation is not well
1519
      # supported either.
1520
      # FIXME: does not support file-backed instances
1521
      if len(inst_config.secondary_nodes) == 0:
1522
        i_non_redundant.append(instance)
1523
      _ErrorIf(len(inst_config.secondary_nodes) > 1,
1524
               self.EINSTANCELAYOUT, instance,
1525
               "instance has multiple secondary nodes", code="WARNING")
1526

    
1527
      if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
1528
        i_non_a_balanced.append(instance)
1529

    
1530
      for snode in inst_config.secondary_nodes:
1531
        _ErrorIf(snode not in node_info and snode not in n_offline,
1532
                 self.ENODERPC, snode,
1533
                 "instance %s, connection to secondary node"
1534
                 "failed", instance)
1535

    
1536
        if snode in node_info:
1537
          node_info[snode]['sinst'].append(instance)
1538
          if pnode not in node_info[snode]['sinst-by-pnode']:
1539
            node_info[snode]['sinst-by-pnode'][pnode] = []
1540
          node_info[snode]['sinst-by-pnode'][pnode].append(instance)
1541

    
1542
        if snode in n_offline:
1543
          inst_nodes_offline.append(snode)
1544

    
1545
      # warn that the instance lives on offline nodes
1546
      _ErrorIf(inst_nodes_offline, self.EINSTANCEBADNODE, instance,
1547
               "instance lives on offline node(s) %s",
1548
               utils.CommaJoin(inst_nodes_offline))
1549

    
1550
    feedback_fn("* Verifying orphan volumes")
1551
    self._VerifyOrphanVolumes(node_vol_should, node_volume)
1552

    
1553
    feedback_fn("* Verifying remaining instances")
1554
    self._VerifyOrphanInstances(instancelist, node_instance)
1555

    
1556
    if constants.VERIFY_NPLUSONE_MEM not in self.skip_set:
1557
      feedback_fn("* Verifying N+1 Memory redundancy")
1558
      self._VerifyNPlusOneMemory(node_info, instance_cfg)
1559

    
1560
    feedback_fn("* Other Notes")
1561
    if i_non_redundant:
1562
      feedback_fn("  - NOTICE: %d non-redundant instance(s) found."
1563
                  % len(i_non_redundant))
1564

    
1565
    if i_non_a_balanced:
1566
      feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
1567
                  % len(i_non_a_balanced))
1568

    
1569
    if n_offline:
1570
      feedback_fn("  - NOTICE: %d offline node(s) found." % len(n_offline))
1571

    
1572
    if n_drained:
1573
      feedback_fn("  - NOTICE: %d drained node(s) found." % len(n_drained))
1574

    
1575
    return not self.bad
1576

    
1577
  def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
1578
    """Analyze the post-hooks' result
1579

1580
    This method analyses the hook result, handles it, and sends some
1581
    nicely-formatted feedback back to the user.
1582

1583
    @param phase: one of L{constants.HOOKS_PHASE_POST} or
1584
        L{constants.HOOKS_PHASE_PRE}; it denotes the hooks phase
1585
    @param hooks_results: the results of the multi-node hooks rpc call
1586
    @param feedback_fn: function used send feedback back to the caller
1587
    @param lu_result: previous Exec result
1588
    @return: the new Exec result, based on the previous result
1589
        and hook results
1590

1591
    """
1592
    # We only really run POST phase hooks, and are only interested in
1593
    # their results
1594
    if phase == constants.HOOKS_PHASE_POST:
1595
      # Used to change hooks' output to proper indentation
1596
      indent_re = re.compile('^', re.M)
1597
      feedback_fn("* Hooks Results")
1598
      assert hooks_results, "invalid result from hooks"
1599

    
1600
      for node_name in hooks_results:
1601
        res = hooks_results[node_name]
1602
        msg = res.fail_msg
1603
        test = msg and not res.offline
1604
        self._ErrorIf(test, self.ENODEHOOKS, node_name,
1605
                      "Communication failure in hooks execution: %s", msg)
1606
        if res.offline or msg:
1607
          # No need to investigate payload if node is offline or gave an error.
1608
          # override manually lu_result here as _ErrorIf only
1609
          # overrides self.bad
1610
          lu_result = 1
1611
          continue
1612
        for script, hkr, output in res.payload:
1613
          test = hkr == constants.HKR_FAIL
1614
          self._ErrorIf(test, self.ENODEHOOKS, node_name,
1615
                        "Script %s failed, output:", script)
1616
          if test:
1617
            output = indent_re.sub('      ', output)
1618
            feedback_fn("%s" % output)
1619
            lu_result = 1
1620

    
1621
      return lu_result
1622

    
1623

    
1624
class LUVerifyDisks(NoHooksLU):
1625
  """Verifies the cluster disks status.
1626

1627
  """
1628
  _OP_REQP = []
1629
  REQ_BGL = False
1630

    
1631
  def ExpandNames(self):
1632
    self.needed_locks = {
1633
      locking.LEVEL_NODE: locking.ALL_SET,
1634
      locking.LEVEL_INSTANCE: locking.ALL_SET,
1635
    }
1636
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
1637

    
1638
  def CheckPrereq(self):
1639
    """Check prerequisites.
1640

1641
    This has no prerequisites.
1642

1643
    """
1644
    pass
1645

    
1646
  def Exec(self, feedback_fn):
1647
    """Verify integrity of cluster disks.
1648

1649
    @rtype: tuple of three items
1650
    @return: a tuple of (dict of node-to-node_error, list of instances
1651
        which need activate-disks, dict of instance: (node, volume) for
1652
        missing volumes
1653

1654
    """
1655
    result = res_nodes, res_instances, res_missing = {}, [], {}
1656

    
1657
    vg_name = self.cfg.GetVGName()
1658
    nodes = utils.NiceSort(self.cfg.GetNodeList())
1659
    instances = [self.cfg.GetInstanceInfo(name)
1660
                 for name in self.cfg.GetInstanceList()]
1661

    
1662
    nv_dict = {}
1663
    for inst in instances:
1664
      inst_lvs = {}
1665
      if (not inst.admin_up or
1666
          inst.disk_template not in constants.DTS_NET_MIRROR):
1667
        continue
1668
      inst.MapLVsByNode(inst_lvs)
1669
      # transform { iname: {node: [vol,],},} to {(node, vol): iname}
1670
      for node, vol_list in inst_lvs.iteritems():
1671
        for vol in vol_list:
1672
          nv_dict[(node, vol)] = inst
1673

    
1674
    if not nv_dict:
1675
      return result
1676

    
1677
    node_lvs = self.rpc.call_lv_list(nodes, vg_name)
1678

    
1679
    for node in nodes:
1680
      # node_volume
1681
      node_res = node_lvs[node]
1682
      if node_res.offline:
1683
        continue
1684
      msg = node_res.fail_msg
1685
      if msg:
1686
        logging.warning("Error enumerating LVs on node %s: %s", node, msg)
1687
        res_nodes[node] = msg
1688
        continue
1689

    
1690
      lvs = node_res.payload
1691
      for lv_name, (_, _, lv_online) in lvs.items():
1692
        inst = nv_dict.pop((node, lv_name), None)
1693
        if (not lv_online and inst is not None
1694
            and inst.name not in res_instances):
1695
          res_instances.append(inst.name)
1696

    
1697
    # any leftover items in nv_dict are missing LVs, let's arrange the
1698
    # data better
1699
    for key, inst in nv_dict.iteritems():
1700
      if inst.name not in res_missing:
1701
        res_missing[inst.name] = []
1702
      res_missing[inst.name].append(key)
1703

    
1704
    return result
1705

    
1706

    
1707
class LURepairDiskSizes(NoHooksLU):
1708
  """Verifies the cluster disks sizes.
1709

1710
  """
1711
  _OP_REQP = ["instances"]
1712
  REQ_BGL = False
1713

    
1714
  def ExpandNames(self):
1715
    if not isinstance(self.op.instances, list):
1716
      raise errors.OpPrereqError("Invalid argument type 'instances'",
1717
                                 errors.ECODE_INVAL)
1718

    
1719
    if self.op.instances:
1720
      self.wanted_names = []
1721
      for name in self.op.instances:
1722
        full_name = self.cfg.ExpandInstanceName(name)
1723
        if full_name is None:
1724
          raise errors.OpPrereqError("Instance '%s' not known" % name,
1725
                                     errors.ECODE_NOENT)
1726
        self.wanted_names.append(full_name)
1727
      self.needed_locks = {
1728
        locking.LEVEL_NODE: [],
1729
        locking.LEVEL_INSTANCE: self.wanted_names,
1730
        }
1731
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
1732
    else:
1733
      self.wanted_names = None
1734
      self.needed_locks = {
1735
        locking.LEVEL_NODE: locking.ALL_SET,
1736
        locking.LEVEL_INSTANCE: locking.ALL_SET,
1737
        }
1738
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
1739

    
1740
  def DeclareLocks(self, level):
1741
    if level == locking.LEVEL_NODE and self.wanted_names is not None:
1742
      self._LockInstancesNodes(primary_only=True)
1743

    
1744
  def CheckPrereq(self):
1745
    """Check prerequisites.
1746

1747
    This only checks the optional instance list against the existing names.
1748

1749
    """
1750
    if self.wanted_names is None:
1751
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
1752

    
1753
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
1754
                             in self.wanted_names]
1755

    
1756
  def _EnsureChildSizes(self, disk):
1757
    """Ensure children of the disk have the needed disk size.
1758

1759
    This is valid mainly for DRBD8 and fixes an issue where the
1760
    children have smaller disk size.
1761

1762
    @param disk: an L{ganeti.objects.Disk} object
1763

1764
    """
1765
    if disk.dev_type == constants.LD_DRBD8:
1766
      assert disk.children, "Empty children for DRBD8?"
1767
      fchild = disk.children[0]
1768
      mismatch = fchild.size < disk.size
1769
      if mismatch:
1770
        self.LogInfo("Child disk has size %d, parent %d, fixing",
1771
                     fchild.size, disk.size)
1772
        fchild.size = disk.size
1773

    
1774
      # and we recurse on this child only, not on the metadev
1775
      return self._EnsureChildSizes(fchild) or mismatch
1776
    else:
1777
      return False
1778

    
1779
  def Exec(self, feedback_fn):
1780
    """Verify the size of cluster disks.
1781

1782
    """
1783
    # TODO: check child disks too
1784
    # TODO: check differences in size between primary/secondary nodes
1785
    per_node_disks = {}
1786
    for instance in self.wanted_instances:
1787
      pnode = instance.primary_node
1788
      if pnode not in per_node_disks:
1789
        per_node_disks[pnode] = []
1790
      for idx, disk in enumerate(instance.disks):
1791
        per_node_disks[pnode].append((instance, idx, disk))
1792

    
1793
    changed = []
1794
    for node, dskl in per_node_disks.items():
1795
      newl = [v[2].Copy() for v in dskl]
1796
      for dsk in newl:
1797
        self.cfg.SetDiskID(dsk, node)
1798
      result = self.rpc.call_blockdev_getsizes(node, newl)
1799
      if result.fail_msg:
1800
        self.LogWarning("Failure in blockdev_getsizes call to node"
1801
                        " %s, ignoring", node)
1802
        continue
1803
      if len(result.data) != len(dskl):
1804
        self.LogWarning("Invalid result from node %s, ignoring node results",
1805
                        node)
1806
        continue
1807
      for ((instance, idx, disk), size) in zip(dskl, result.data):
1808
        if size is None:
1809
          self.LogWarning("Disk %d of instance %s did not return size"
1810
                          " information, ignoring", idx, instance.name)
1811
          continue
1812
        if not isinstance(size, (int, long)):
1813
          self.LogWarning("Disk %d of instance %s did not return valid"
1814
                          " size information, ignoring", idx, instance.name)
1815
          continue
1816
        size = size >> 20
1817
        if size != disk.size:
1818
          self.LogInfo("Disk %d of instance %s has mismatched size,"
1819
                       " correcting: recorded %d, actual %d", idx,
1820
                       instance.name, disk.size, size)
1821
          disk.size = size
1822
          self.cfg.Update(instance, feedback_fn)
1823
          changed.append((instance.name, idx, size))
1824
        if self._EnsureChildSizes(disk):
1825
          self.cfg.Update(instance, feedback_fn)
1826
          changed.append((instance.name, idx, disk.size))
1827
    return changed
1828

    
1829

    
1830
class LURenameCluster(LogicalUnit):
1831
  """Rename the cluster.
1832

1833
  """
1834
  HPATH = "cluster-rename"
1835
  HTYPE = constants.HTYPE_CLUSTER
1836
  _OP_REQP = ["name"]
1837

    
1838
  def BuildHooksEnv(self):
1839
    """Build hooks env.
1840

1841
    """
1842
    env = {
1843
      "OP_TARGET": self.cfg.GetClusterName(),
1844
      "NEW_NAME": self.op.name,
1845
      }
1846
    mn = self.cfg.GetMasterNode()
1847
    all_nodes = self.cfg.GetNodeList()
1848
    return env, [mn], all_nodes
1849

    
1850
  def CheckPrereq(self):
1851
    """Verify that the passed name is a valid one.
1852

1853
    """
1854
    hostname = utils.GetHostInfo(self.op.name)
1855

    
1856
    new_name = hostname.name
1857
    self.ip = new_ip = hostname.ip
1858
    old_name = self.cfg.GetClusterName()
1859
    old_ip = self.cfg.GetMasterIP()
1860
    if new_name == old_name and new_ip == old_ip:
1861
      raise errors.OpPrereqError("Neither the name nor the IP address of the"
1862
                                 " cluster has changed",
1863
                                 errors.ECODE_INVAL)
1864
    if new_ip != old_ip:
1865
      if utils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
1866
        raise errors.OpPrereqError("The given cluster IP address (%s) is"
1867
                                   " reachable on the network. Aborting." %
1868
                                   new_ip, errors.ECODE_NOTUNIQUE)
1869

    
1870
    self.op.name = new_name
1871

    
1872
  def Exec(self, feedback_fn):
1873
    """Rename the cluster.
1874

1875
    """
1876
    clustername = self.op.name
1877
    ip = self.ip
1878

    
1879
    # shutdown the master IP
1880
    master = self.cfg.GetMasterNode()
1881
    result = self.rpc.call_node_stop_master(master, False)
1882
    result.Raise("Could not disable the master role")
1883

    
1884
    try:
1885
      cluster = self.cfg.GetClusterInfo()
1886
      cluster.cluster_name = clustername
1887
      cluster.master_ip = ip
1888
      self.cfg.Update(cluster, feedback_fn)
1889

    
1890
      # update the known hosts file
1891
      ssh.WriteKnownHostsFile(self.cfg, constants.SSH_KNOWN_HOSTS_FILE)
1892
      node_list = self.cfg.GetNodeList()
1893
      try:
1894
        node_list.remove(master)
1895
      except ValueError:
1896
        pass
1897
      result = self.rpc.call_upload_file(node_list,
1898
                                         constants.SSH_KNOWN_HOSTS_FILE)
1899
      for to_node, to_result in result.iteritems():
1900
        msg = to_result.fail_msg
1901
        if msg:
1902
          msg = ("Copy of file %s to node %s failed: %s" %
1903
                 (constants.SSH_KNOWN_HOSTS_FILE, to_node, msg))
1904
          self.proc.LogWarning(msg)
1905

    
1906
    finally:
1907
      result = self.rpc.call_node_start_master(master, False, False)
1908
      msg = result.fail_msg
1909
      if msg:
1910
        self.LogWarning("Could not re-enable the master role on"
1911
                        " the master, please restart manually: %s", msg)
1912

    
1913

    
1914
def _RecursiveCheckIfLVMBased(disk):
1915
  """Check if the given disk or its children are lvm-based.
1916

1917
  @type disk: L{objects.Disk}
1918
  @param disk: the disk to check
1919
  @rtype: boolean
1920
  @return: boolean indicating whether a LD_LV dev_type was found or not
1921

1922
  """
1923
  if disk.children:
1924
    for chdisk in disk.children:
1925
      if _RecursiveCheckIfLVMBased(chdisk):
1926
        return True
1927
  return disk.dev_type == constants.LD_LV
1928

    
1929

    
1930
class LUSetClusterParams(LogicalUnit):
1931
  """Change the parameters of the cluster.
1932

1933
  """
1934
  HPATH = "cluster-modify"
1935
  HTYPE = constants.HTYPE_CLUSTER
1936
  _OP_REQP = []
1937
  REQ_BGL = False
1938

    
1939
  def CheckArguments(self):
1940
    """Check parameters
1941

1942
    """
1943
    if not hasattr(self.op, "candidate_pool_size"):
1944
      self.op.candidate_pool_size = None
1945
    if self.op.candidate_pool_size is not None:
1946
      try:
1947
        self.op.candidate_pool_size = int(self.op.candidate_pool_size)
1948
      except (ValueError, TypeError), err:
1949
        raise errors.OpPrereqError("Invalid candidate_pool_size value: %s" %
1950
                                   str(err), errors.ECODE_INVAL)
1951
      if self.op.candidate_pool_size < 1:
1952
        raise errors.OpPrereqError("At least one master candidate needed",
1953
                                   errors.ECODE_INVAL)
1954

    
1955
  def ExpandNames(self):
1956
    # FIXME: in the future maybe other cluster params won't require checking on
1957
    # all nodes to be modified.
1958
    self.needed_locks = {
1959
      locking.LEVEL_NODE: locking.ALL_SET,
1960
    }
1961
    self.share_locks[locking.LEVEL_NODE] = 1
1962

    
1963
  def BuildHooksEnv(self):
1964
    """Build hooks env.
1965

1966
    """
1967
    env = {
1968
      "OP_TARGET": self.cfg.GetClusterName(),
1969
      "NEW_VG_NAME": self.op.vg_name,
1970
      }
1971
    mn = self.cfg.GetMasterNode()
1972
    return env, [mn], [mn]
1973

    
1974
  def CheckPrereq(self):
1975
    """Check prerequisites.
1976

1977
    This checks whether the given params don't conflict and
1978
    if the given volume group is valid.
1979

1980
    """
1981
    if self.op.vg_name is not None and not self.op.vg_name:
1982
      instances = self.cfg.GetAllInstancesInfo().values()
1983
      for inst in instances:
1984
        for disk in inst.disks:
1985
          if _RecursiveCheckIfLVMBased(disk):
1986
            raise errors.OpPrereqError("Cannot disable lvm storage while"
1987
                                       " lvm-based instances exist",
1988
                                       errors.ECODE_INVAL)
1989

    
1990
    node_list = self.acquired_locks[locking.LEVEL_NODE]
1991

    
1992
    # if vg_name not None, checks given volume group on all nodes
1993
    if self.op.vg_name:
1994
      vglist = self.rpc.call_vg_list(node_list)
1995
      for node in node_list:
1996
        msg = vglist[node].fail_msg
1997
        if msg:
1998
          # ignoring down node
1999
          self.LogWarning("Error while gathering data on node %s"
2000
                          " (ignoring node): %s", node, msg)
2001
          continue
2002
        vgstatus = utils.CheckVolumeGroupSize(vglist[node].payload,
2003
                                              self.op.vg_name,
2004
                                              constants.MIN_VG_SIZE)
2005
        if vgstatus:
2006
          raise errors.OpPrereqError("Error on node '%s': %s" %
2007
                                     (node, vgstatus), errors.ECODE_ENVIRON)
2008

    
2009
    self.cluster = cluster = self.cfg.GetClusterInfo()
2010
    # validate params changes
2011
    if self.op.beparams:
2012
      utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
2013
      self.new_beparams = objects.FillDict(
2014
        cluster.beparams[constants.PP_DEFAULT], self.op.beparams)
2015

    
2016
    if self.op.nicparams:
2017
      utils.ForceDictType(self.op.nicparams, constants.NICS_PARAMETER_TYPES)
2018
      self.new_nicparams = objects.FillDict(
2019
        cluster.nicparams[constants.PP_DEFAULT], self.op.nicparams)
2020
      objects.NIC.CheckParameterSyntax(self.new_nicparams)
2021
      nic_errors = []
2022

    
2023
      # check all instances for consistency
2024
      for instance in self.cfg.GetAllInstancesInfo().values():
2025
        for nic_idx, nic in enumerate(instance.nics):
2026
          params_copy = copy.deepcopy(nic.nicparams)
2027
          params_filled = objects.FillDict(self.new_nicparams, params_copy)
2028

    
2029
          # check parameter syntax
2030
          try:
2031
            objects.NIC.CheckParameterSyntax(params_filled)
2032
          except errors.ConfigurationError, err:
2033
            nic_errors.append("Instance %s, nic/%d: %s" %
2034
                              (instance.name, nic_idx, err))
2035

    
2036
          # if we're moving instances to routed, check that they have an ip
2037
          target_mode = params_filled[constants.NIC_MODE]
2038
          if target_mode == constants.NIC_MODE_ROUTED and not nic.ip:
2039
            nic_errors.append("Instance %s, nic/%d: routed nick with no ip" %
2040
                              (instance.name, nic_idx))
2041
      if nic_errors:
2042
        raise errors.OpPrereqError("Cannot apply the change, errors:\n%s" %
2043
                                   "\n".join(nic_errors))
2044

    
2045
    # hypervisor list/parameters
2046
    self.new_hvparams = objects.FillDict(cluster.hvparams, {})
2047
    if self.op.hvparams:
2048
      if not isinstance(self.op.hvparams, dict):
2049
        raise errors.OpPrereqError("Invalid 'hvparams' parameter on input",
2050
                                   errors.ECODE_INVAL)
2051
      for hv_name, hv_dict in self.op.hvparams.items():
2052
        if hv_name not in self.new_hvparams:
2053
          self.new_hvparams[hv_name] = hv_dict
2054
        else:
2055
          self.new_hvparams[hv_name].update(hv_dict)
2056

    
2057
    if self.op.enabled_hypervisors is not None:
2058
      self.hv_list = self.op.enabled_hypervisors
2059
      if not self.hv_list:
2060
        raise errors.OpPrereqError("Enabled hypervisors list must contain at"
2061
                                   " least one member",
2062
                                   errors.ECODE_INVAL)
2063
      invalid_hvs = set(self.hv_list) - constants.HYPER_TYPES
2064
      if invalid_hvs:
2065
        raise errors.OpPrereqError("Enabled hypervisors contains invalid"
2066
                                   " entries: %s" %
2067
                                   utils.CommaJoin(invalid_hvs),
2068
                                   errors.ECODE_INVAL)
2069
    else:
2070
      self.hv_list = cluster.enabled_hypervisors
2071

    
2072
    if self.op.hvparams or self.op.enabled_hypervisors is not None:
2073
      # either the enabled list has changed, or the parameters have, validate
2074
      for hv_name, hv_params in self.new_hvparams.items():
2075
        if ((self.op.hvparams and hv_name in self.op.hvparams) or
2076
            (self.op.enabled_hypervisors and
2077
             hv_name in self.op.enabled_hypervisors)):
2078
          # either this is a new hypervisor, or its parameters have changed
2079
          hv_class = hypervisor.GetHypervisor(hv_name)
2080
          utils.ForceDictType(hv_params, constants.HVS_PARAMETER_TYPES)
2081
          hv_class.CheckParameterSyntax(hv_params)
2082
          _CheckHVParams(self, node_list, hv_name, hv_params)
2083

    
2084
  def Exec(self, feedback_fn):
2085
    """Change the parameters of the cluster.
2086

2087
    """
2088
    if self.op.vg_name is not None:
2089
      new_volume = self.op.vg_name
2090
      if not new_volume:
2091
        new_volume = None
2092
      if new_volume != self.cfg.GetVGName():
2093
        self.cfg.SetVGName(new_volume)
2094
      else:
2095
        feedback_fn("Cluster LVM configuration already in desired"
2096
                    " state, not changing")
2097
    if self.op.hvparams:
2098
      self.cluster.hvparams = self.new_hvparams
2099
    if self.op.enabled_hypervisors is not None:
2100
      self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
2101
    if self.op.beparams:
2102
      self.cluster.beparams[constants.PP_DEFAULT] = self.new_beparams
2103
    if self.op.nicparams:
2104
      self.cluster.nicparams[constants.PP_DEFAULT] = self.new_nicparams
2105

    
2106
    if self.op.candidate_pool_size is not None:
2107
      self.cluster.candidate_pool_size = self.op.candidate_pool_size
2108
      # we need to update the pool size here, otherwise the save will fail
2109
      _AdjustCandidatePool(self, [])
2110

    
2111
    self.cfg.Update(self.cluster, feedback_fn)
2112

    
2113

    
2114
def _RedistributeAncillaryFiles(lu, additional_nodes=None):
2115
  """Distribute additional files which are part of the cluster configuration.
2116

2117
  ConfigWriter takes care of distributing the config and ssconf files, but
2118
  there are more files which should be distributed to all nodes. This function
2119
  makes sure those are copied.
2120

2121
  @param lu: calling logical unit
2122
  @param additional_nodes: list of nodes not in the config to distribute to
2123

2124
  """
2125
  # 1. Gather target nodes
2126
  myself = lu.cfg.GetNodeInfo(lu.cfg.GetMasterNode())
2127
  dist_nodes = lu.cfg.GetNodeList()
2128
  if additional_nodes is not None:
2129
    dist_nodes.extend(additional_nodes)
2130
  if myself.name in dist_nodes:
2131
    dist_nodes.remove(myself.name)
2132

    
2133
  # 2. Gather files to distribute
2134
  dist_files = set([constants.ETC_HOSTS,
2135
                    constants.SSH_KNOWN_HOSTS_FILE,
2136
                    constants.RAPI_CERT_FILE,
2137
                    constants.RAPI_USERS_FILE,
2138
                    constants.HMAC_CLUSTER_KEY,
2139
                   ])
2140

    
2141
  enabled_hypervisors = lu.cfg.GetClusterInfo().enabled_hypervisors
2142
  for hv_name in enabled_hypervisors:
2143
    hv_class = hypervisor.GetHypervisor(hv_name)
2144
    dist_files.update(hv_class.GetAncillaryFiles())
2145

    
2146
  # 3. Perform the files upload
2147
  for fname in dist_files:
2148
    if os.path.exists(fname):
2149
      result = lu.rpc.call_upload_file(dist_nodes, fname)
2150
      for to_node, to_result in result.items():
2151
        msg = to_result.fail_msg
2152
        if msg:
2153
          msg = ("Copy of file %s to node %s failed: %s" %
2154
                 (fname, to_node, msg))
2155
          lu.proc.LogWarning(msg)
2156

    
2157

    
2158
class LURedistributeConfig(NoHooksLU):
2159
  """Force the redistribution of cluster configuration.
2160

2161
  This is a very simple LU.
2162

2163
  """
2164
  _OP_REQP = []
2165
  REQ_BGL = False
2166

    
2167
  def ExpandNames(self):
2168
    self.needed_locks = {
2169
      locking.LEVEL_NODE: locking.ALL_SET,
2170
    }
2171
    self.share_locks[locking.LEVEL_NODE] = 1
2172

    
2173
  def CheckPrereq(self):
2174
    """Check prerequisites.
2175

2176
    """
2177

    
2178
  def Exec(self, feedback_fn):
2179
    """Redistribute the configuration.
2180

2181
    """
2182
    self.cfg.Update(self.cfg.GetClusterInfo(), feedback_fn)
2183
    _RedistributeAncillaryFiles(self)
2184

    
2185

    
2186
def _WaitForSync(lu, instance, oneshot=False):
2187
  """Sleep and poll for an instance's disk to sync.
2188

2189
  """
2190
  if not instance.disks:
2191
    return True
2192

    
2193
  if not oneshot:
2194
    lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
2195

    
2196
  node = instance.primary_node
2197

    
2198
  for dev in instance.disks:
2199
    lu.cfg.SetDiskID(dev, node)
2200

    
2201
  # TODO: Convert to utils.Retry
2202

    
2203
  retries = 0
2204
  degr_retries = 10 # in seconds, as we sleep 1 second each time
2205
  while True:
2206
    max_time = 0
2207
    done = True
2208
    cumul_degraded = False
2209
    rstats = lu.rpc.call_blockdev_getmirrorstatus(node, instance.disks)
2210
    msg = rstats.fail_msg
2211
    if msg:
2212
      lu.LogWarning("Can't get any data from node %s: %s", node, msg)
2213
      retries += 1
2214
      if retries >= 10:
2215
        raise errors.RemoteError("Can't contact node %s for mirror data,"
2216
                                 " aborting." % node)
2217
      time.sleep(6)
2218
      continue
2219
    rstats = rstats.payload
2220
    retries = 0
2221
    for i, mstat in enumerate(rstats):
2222
      if mstat is None:
2223
        lu.LogWarning("Can't compute data for node %s/%s",
2224
                           node, instance.disks[i].iv_name)
2225
        continue
2226

    
2227
      cumul_degraded = (cumul_degraded or
2228
                        (mstat.is_degraded and mstat.sync_percent is None))
2229
      if mstat.sync_percent is not None:
2230
        done = False
2231
        if mstat.estimated_time is not None:
2232
          rem_time = "%d estimated seconds remaining" % mstat.estimated_time
2233
          max_time = mstat.estimated_time
2234
        else:
2235
          rem_time = "no time estimate"
2236
        lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
2237
                        (instance.disks[i].iv_name, mstat.sync_percent,
2238
                         rem_time))
2239

    
2240
    # if we're done but degraded, let's do a few small retries, to
2241
    # make sure we see a stable and not transient situation; therefore
2242
    # we force restart of the loop
2243
    if (done or oneshot) and cumul_degraded and degr_retries > 0:
2244
      logging.info("Degraded disks found, %d retries left", degr_retries)
2245
      degr_retries -= 1
2246
      time.sleep(1)
2247
      continue
2248

    
2249
    if done or oneshot:
2250
      break
2251

    
2252
    time.sleep(min(60, max_time))
2253

    
2254
  if done:
2255
    lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
2256
  return not cumul_degraded
2257

    
2258

    
2259
def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
2260
  """Check that mirrors are not degraded.
2261

2262
  The ldisk parameter, if True, will change the test from the
2263
  is_degraded attribute (which represents overall non-ok status for
2264
  the device(s)) to the ldisk (representing the local storage status).
2265

2266
  """
2267
  lu.cfg.SetDiskID(dev, node)
2268

    
2269
  result = True
2270

    
2271
  if on_primary or dev.AssembleOnSecondary():
2272
    rstats = lu.rpc.call_blockdev_find(node, dev)
2273
    msg = rstats.fail_msg
2274
    if msg:
2275
      lu.LogWarning("Can't find disk on node %s: %s", node, msg)
2276
      result = False
2277
    elif not rstats.payload:
2278
      lu.LogWarning("Can't find disk on node %s", node)
2279
      result = False
2280
    else:
2281
      if ldisk:
2282
        result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
2283
      else:
2284
        result = result and not rstats.payload.is_degraded
2285

    
2286
  if dev.children:
2287
    for child in dev.children:
2288
      result = result and _CheckDiskConsistency(lu, child, node, on_primary)
2289

    
2290
  return result
2291

    
2292

    
2293
class LUDiagnoseOS(NoHooksLU):
2294
  """Logical unit for OS diagnose/query.
2295

2296
  """
2297
  _OP_REQP = ["output_fields", "names"]
2298
  REQ_BGL = False
2299
  _FIELDS_STATIC = utils.FieldSet()
2300
  _FIELDS_DYNAMIC = utils.FieldSet("name", "valid", "node_status", "variants")
2301
  # Fields that need calculation of global os validity
2302
  _FIELDS_NEEDVALID = frozenset(["valid", "variants"])
2303

    
2304
  def ExpandNames(self):
2305
    if self.op.names:
2306
      raise errors.OpPrereqError("Selective OS query not supported",
2307
                                 errors.ECODE_INVAL)
2308

    
2309
    _CheckOutputFields(static=self._FIELDS_STATIC,
2310
                       dynamic=self._FIELDS_DYNAMIC,
2311
                       selected=self.op.output_fields)
2312

    
2313
    # Lock all nodes, in shared mode
2314
    # Temporary removal of locks, should be reverted later
2315
    # TODO: reintroduce locks when they are lighter-weight
2316
    self.needed_locks = {}
2317
    #self.share_locks[locking.LEVEL_NODE] = 1
2318
    #self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
2319

    
2320
  def CheckPrereq(self):
2321
    """Check prerequisites.
2322

2323
    """
2324

    
2325
  @staticmethod
2326
  def _DiagnoseByOS(rlist):
2327
    """Remaps a per-node return list into an a per-os per-node dictionary
2328

2329
    @param rlist: a map with node names as keys and OS objects as values
2330

2331
    @rtype: dict
2332
    @return: a dictionary with osnames as keys and as value another map, with
2333
        nodes as keys and tuples of (path, status, diagnose) as values, eg::
2334

2335
          {"debian-etch": {"node1": [(/usr/lib/..., True, ""),
2336
                                     (/srv/..., False, "invalid api")],
2337
                           "node2": [(/srv/..., True, "")]}
2338
          }
2339

2340
    """
2341
    all_os = {}
2342
    # we build here the list of nodes that didn't fail the RPC (at RPC
2343
    # level), so that nodes with a non-responding node daemon don't
2344
    # make all OSes invalid
2345
    good_nodes = [node_name for node_name in rlist
2346
                  if not rlist[node_name].fail_msg]
2347
    for node_name, nr in rlist.items():
2348
      if nr.fail_msg or not nr.payload:
2349
        continue
2350
      for name, path, status, diagnose, variants in nr.payload:
2351
        if name not in all_os:
2352
          # build a list of nodes for this os containing empty lists
2353
          # for each node in node_list
2354
          all_os[name] = {}
2355
          for nname in good_nodes:
2356
            all_os[name][nname] = []
2357
        all_os[name][node_name].append((path, status, diagnose, variants))
2358
    return all_os
2359

    
2360
  def Exec(self, feedback_fn):
2361
    """Compute the list of OSes.
2362

2363
    """
2364
    valid_nodes = [node for node in self.cfg.GetOnlineNodeList()]
2365
    node_data = self.rpc.call_os_diagnose(valid_nodes)
2366
    pol = self._DiagnoseByOS(node_data)
2367
    output = []
2368
    calc_valid = self._FIELDS_NEEDVALID.intersection(self.op.output_fields)
2369
    calc_variants = "variants" in self.op.output_fields
2370

    
2371
    for os_name, os_data in pol.items():
2372
      row = []
2373
      if calc_valid:
2374
        valid = True
2375
        variants = None
2376
        for osl in os_data.values():
2377
          valid = valid and osl and osl[0][1]
2378
          if not valid:
2379
            variants = None
2380
            break
2381
          if calc_variants:
2382
            node_variants = osl[0][3]
2383
            if variants is None:
2384
              variants = node_variants
2385
            else:
2386
              variants = [v for v in variants if v in node_variants]
2387

    
2388
      for field in self.op.output_fields:
2389
        if field == "name":
2390
          val = os_name
2391
        elif field == "valid":
2392
          val = valid
2393
        elif field == "node_status":
2394
          # this is just a copy of the dict
2395
          val = {}
2396
          for node_name, nos_list in os_data.items():
2397
            val[node_name] = nos_list
2398
        elif field == "variants":
2399
          val =  variants
2400
        else:
2401
          raise errors.ParameterError(field)
2402
        row.append(val)
2403
      output.append(row)
2404

    
2405
    return output
2406

    
2407

    
2408
class LURemoveNode(LogicalUnit):
2409
  """Logical unit for removing a node.
2410

2411
  """
2412
  HPATH = "node-remove"
2413
  HTYPE = constants.HTYPE_NODE
2414
  _OP_REQP = ["node_name"]
2415

    
2416
  def BuildHooksEnv(self):
2417
    """Build hooks env.
2418

2419
    This doesn't run on the target node in the pre phase as a failed
2420
    node would then be impossible to remove.
2421

2422
    """
2423
    env = {
2424
      "OP_TARGET": self.op.node_name,
2425
      "NODE_NAME": self.op.node_name,
2426
      }
2427
    all_nodes = self.cfg.GetNodeList()
2428
    if self.op.node_name in all_nodes:
2429
      all_nodes.remove(self.op.node_name)
2430
    return env, all_nodes, all_nodes
2431

    
2432
  def CheckPrereq(self):
2433
    """Check prerequisites.
2434

2435
    This checks:
2436
     - the node exists in the configuration
2437
     - it does not have primary or secondary instances
2438
     - it's not the master
2439

2440
    Any errors are signaled by raising errors.OpPrereqError.
2441

2442
    """
2443
    node = self.cfg.GetNodeInfo(self.cfg.ExpandNodeName(self.op.node_name))
2444
    if node is None:
2445
      raise errors.OpPrereqError("Node '%s' is unknown." % self.op.node_name,
2446
                                 errors.ECODE_NOENT)
2447

    
2448
    instance_list = self.cfg.GetInstanceList()
2449

    
2450
    masternode = self.cfg.GetMasterNode()
2451
    if node.name == masternode:
2452
      raise errors.OpPrereqError("Node is the master node,"
2453
                                 " you need to failover first.",
2454
                                 errors.ECODE_INVAL)
2455

    
2456
    for instance_name in instance_list:
2457
      instance = self.cfg.GetInstanceInfo(instance_name)
2458
      if node.name in instance.all_nodes:
2459
        raise errors.OpPrereqError("Instance %s is still running on the node,"
2460
                                   " please remove first." % instance_name,
2461
                                   errors.ECODE_INVAL)
2462
    self.op.node_name = node.name
2463
    self.node = node
2464

    
2465
  def Exec(self, feedback_fn):
2466
    """Removes the node from the cluster.
2467

2468
    """
2469
    node = self.node
2470
    logging.info("Stopping the node daemon and removing configs from node %s",
2471
                 node.name)
2472

    
2473
    modify_ssh_setup = self.cfg.GetClusterInfo().modify_ssh_setup
2474

    
2475
    # Promote nodes to master candidate as needed
2476
    _AdjustCandidatePool(self, exceptions=[node.name])
2477
    self.context.RemoveNode(node.name)
2478

    
2479
    # Run post hooks on the node before it's removed
2480
    hm = self.proc.hmclass(self.rpc.call_hooks_runner, self)
2481
    try:
2482
      hm.RunPhase(constants.HOOKS_PHASE_POST, [node.name])
2483
    except:
2484
      # pylint: disable-msg=W0702
2485
      self.LogWarning("Errors occurred running hooks on %s" % node.name)
2486

    
2487
    result = self.rpc.call_node_leave_cluster(node.name, modify_ssh_setup)
2488
    msg = result.fail_msg
2489
    if msg:
2490
      self.LogWarning("Errors encountered on the remote node while leaving"
2491
                      " the cluster: %s", msg)
2492

    
2493

    
2494
class LUQueryNodes(NoHooksLU):
2495
  """Logical unit for querying nodes.
2496

2497
  """
2498
  # pylint: disable-msg=W0142
2499
  _OP_REQP = ["output_fields", "names", "use_locking"]
2500
  REQ_BGL = False
2501

    
2502
  _SIMPLE_FIELDS = ["name", "serial_no", "ctime", "mtime", "uuid",
2503
                    "master_candidate", "offline", "drained"]
2504

    
2505
  _FIELDS_DYNAMIC = utils.FieldSet(
2506
    "dtotal", "dfree",
2507
    "mtotal", "mnode", "mfree",
2508
    "bootid",
2509
    "ctotal", "cnodes", "csockets",
2510
    )
2511

    
2512
  _FIELDS_STATIC = utils.FieldSet(*[
2513
    "pinst_cnt", "sinst_cnt",
2514
    "pinst_list", "sinst_list",
2515
    "pip", "sip", "tags",
2516
    "master",
2517
    "role"] + _SIMPLE_FIELDS
2518
    )
2519

    
2520
  def ExpandNames(self):
2521
    _CheckOutputFields(static=self._FIELDS_STATIC,
2522
                       dynamic=self._FIELDS_DYNAMIC,
2523
                       selected=self.op.output_fields)
2524

    
2525
    self.needed_locks = {}
2526
    self.share_locks[locking.LEVEL_NODE] = 1
2527

    
2528
    if self.op.names:
2529
      self.wanted = _GetWantedNodes(self, self.op.names)
2530
    else:
2531
      self.wanted = locking.ALL_SET
2532

    
2533
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
2534
    self.do_locking = self.do_node_query and self.op.use_locking
2535
    if self.do_locking:
2536
      # if we don't request only static fields, we need to lock the nodes
2537
      self.needed_locks[locking.LEVEL_NODE] = self.wanted
2538

    
2539
  def CheckPrereq(self):
2540
    """Check prerequisites.
2541

2542
    """
2543
    # The validation of the node list is done in the _GetWantedNodes,
2544
    # if non empty, and if empty, there's no validation to do
2545
    pass
2546

    
2547
  def Exec(self, feedback_fn):
2548
    """Computes the list of nodes and their attributes.
2549

2550
    """
2551
    all_info = self.cfg.GetAllNodesInfo()
2552
    if self.do_locking:
2553
      nodenames = self.acquired_locks[locking.LEVEL_NODE]
2554
    elif self.wanted != locking.ALL_SET:
2555
      nodenames = self.wanted
2556
      missing = set(nodenames).difference(all_info.keys())
2557
      if missing:
2558
        raise errors.OpExecError(
2559
          "Some nodes were removed before retrieving their data: %s" % missing)
2560
    else:
2561
      nodenames = all_info.keys()
2562

    
2563
    nodenames = utils.NiceSort(nodenames)
2564
    nodelist = [all_info[name] for name in nodenames]
2565

    
2566
    # begin data gathering
2567

    
2568
    if self.do_node_query:
2569
      live_data = {}
2570
      node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
2571
                                          self.cfg.GetHypervisorType())
2572
      for name in nodenames:
2573
        nodeinfo = node_data[name]
2574
        if not nodeinfo.fail_msg and nodeinfo.payload:
2575
          nodeinfo = nodeinfo.payload
2576
          fn = utils.TryConvert
2577
          live_data[name] = {
2578
            "mtotal": fn(int, nodeinfo.get('memory_total', None)),
2579
            "mnode": fn(int, nodeinfo.get('memory_dom0', None)),
2580
            "mfree": fn(int, nodeinfo.get('memory_free', None)),
2581
            "dtotal": fn(int, nodeinfo.get('vg_size', None)),
2582
            "dfree": fn(int, nodeinfo.get('vg_free', None)),
2583
            "ctotal": fn(int, nodeinfo.get('cpu_total', None)),
2584
            "bootid": nodeinfo.get('bootid', None),
2585
            "cnodes": fn(int, nodeinfo.get('cpu_nodes', None)),
2586
            "csockets": fn(int, nodeinfo.get('cpu_sockets', None)),
2587
            }
2588
        else:
2589
          live_data[name] = {}
2590
    else:
2591
      live_data = dict.fromkeys(nodenames, {})
2592

    
2593
    node_to_primary = dict([(name, set()) for name in nodenames])
2594
    node_to_secondary = dict([(name, set()) for name in nodenames])
2595

    
2596
    inst_fields = frozenset(("pinst_cnt", "pinst_list",
2597
                             "sinst_cnt", "sinst_list"))
2598
    if inst_fields & frozenset(self.op.output_fields):
2599
      inst_data = self.cfg.GetAllInstancesInfo()
2600

    
2601
      for inst in inst_data.values():
2602
        if inst.primary_node in node_to_primary:
2603
          node_to_primary[inst.primary_node].add(inst.name)
2604
        for secnode in inst.secondary_nodes:
2605
          if secnode in node_to_secondary:
2606
            node_to_secondary[secnode].add(inst.name)
2607

    
2608
    master_node = self.cfg.GetMasterNode()
2609

    
2610
    # end data gathering
2611

    
2612
    output = []
2613
    for node in nodelist:
2614
      node_output = []
2615
      for field in self.op.output_fields:
2616
        if field in self._SIMPLE_FIELDS:
2617
          val = getattr(node, field)
2618
        elif field == "pinst_list":
2619
          val = list(node_to_primary[node.name])
2620
        elif field == "sinst_list":
2621
          val = list(node_to_secondary[node.name])
2622
        elif field == "pinst_cnt":
2623
          val = len(node_to_primary[node.name])
2624
        elif field == "sinst_cnt":
2625
          val = len(node_to_secondary[node.name])
2626
        elif field == "pip":
2627
          val = node.primary_ip
2628
        elif field == "sip":
2629
          val = node.secondary_ip
2630
        elif field == "tags":
2631
          val = list(node.GetTags())
2632
        elif field == "master":
2633
          val = node.name == master_node
2634
        elif self._FIELDS_DYNAMIC.Matches(field):
2635
          val = live_data[node.name].get(field, None)
2636
        elif field == "role":
2637
          if node.name == master_node:
2638
            val = "M"
2639
          elif node.master_candidate:
2640
            val = "C"
2641
          elif node.drained:
2642
            val = "D"
2643
          elif node.offline:
2644
            val = "O"
2645
          else:
2646
            val = "R"
2647
        else:
2648
          raise errors.ParameterError(field)
2649
        node_output.append(val)
2650
      output.append(node_output)
2651

    
2652
    return output
2653

    
2654

    
2655
class LUQueryNodeVolumes(NoHooksLU):
2656
  """Logical unit for getting volumes on node(s).
2657

2658
  """
2659
  _OP_REQP = ["nodes", "output_fields"]
2660
  REQ_BGL = False
2661
  _FIELDS_DYNAMIC = utils.FieldSet("phys", "vg", "name", "size", "instance")
2662
  _FIELDS_STATIC = utils.FieldSet("node")
2663

    
2664
  def ExpandNames(self):
2665
    _CheckOutputFields(static=self._FIELDS_STATIC,
2666
                       dynamic=self._FIELDS_DYNAMIC,
2667
                       selected=self.op.output_fields)
2668

    
2669
    self.needed_locks = {}
2670
    self.share_locks[locking.LEVEL_NODE] = 1
2671
    if not self.op.nodes:
2672
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
2673
    else:
2674
      self.needed_locks[locking.LEVEL_NODE] = \
2675
        _GetWantedNodes(self, self.op.nodes)
2676

    
2677
  def CheckPrereq(self):
2678
    """Check prerequisites.
2679

2680
    This checks that the fields required are valid output fields.
2681

2682
    """
2683
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
2684

    
2685
  def Exec(self, feedback_fn):
2686
    """Computes the list of nodes and their attributes.
2687

2688
    """
2689
    nodenames = self.nodes
2690
    volumes = self.rpc.call_node_volumes(nodenames)
2691

    
2692
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
2693
             in self.cfg.GetInstanceList()]
2694

    
2695
    lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
2696

    
2697
    output = []
2698
    for node in nodenames:
2699
      nresult = volumes[node]
2700
      if nresult.offline:
2701
        continue
2702
      msg = nresult.fail_msg
2703
      if msg:
2704
        self.LogWarning("Can't compute volume data on node %s: %s", node, msg)
2705
        continue
2706

    
2707
      node_vols = nresult.payload[:]
2708
      node_vols.sort(key=lambda vol: vol['dev'])
2709

    
2710
      for vol in node_vols:
2711
        node_output = []
2712
        for field in self.op.output_fields:
2713
          if field == "node":
2714
            val = node
2715
          elif field == "phys":
2716
            val = vol['dev']
2717
          elif field == "vg":
2718
            val = vol['vg']
2719
          elif field == "name":
2720
            val = vol['name']
2721
          elif field == "size":
2722
            val = int(float(vol['size']))
2723
          elif field == "instance":
2724
            for inst in ilist:
2725
              if node not in lv_by_node[inst]:
2726
                continue
2727
              if vol['name'] in lv_by_node[inst][node]:
2728
                val = inst.name
2729
                break
2730
            else:
2731
              val = '-'
2732
          else:
2733
            raise errors.ParameterError(field)
2734
          node_output.append(str(val))
2735

    
2736
        output.append(node_output)
2737

    
2738
    return output
2739

    
2740

    
2741
class LUQueryNodeStorage(NoHooksLU):
2742
  """Logical unit for getting information on storage units on node(s).
2743

2744
  """
2745
  _OP_REQP = ["nodes", "storage_type", "output_fields"]
2746
  REQ_BGL = False
2747
  _FIELDS_STATIC = utils.FieldSet(constants.SF_NODE)
2748

    
2749
  def ExpandNames(self):
2750
    storage_type = self.op.storage_type
2751

    
2752
    if storage_type not in constants.VALID_STORAGE_TYPES:
2753
      raise errors.OpPrereqError("Unknown storage type: %s" % storage_type,
2754
                                 errors.ECODE_INVAL)
2755

    
2756
    _CheckOutputFields(static=self._FIELDS_STATIC,
2757
                       dynamic=utils.FieldSet(*constants.VALID_STORAGE_FIELDS),
2758
                       selected=self.op.output_fields)
2759

    
2760
    self.needed_locks = {}
2761
    self.share_locks[locking.LEVEL_NODE] = 1
2762

    
2763
    if self.op.nodes:
2764
      self.needed_locks[locking.LEVEL_NODE] = \
2765
        _GetWantedNodes(self, self.op.nodes)
2766
    else:
2767
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
2768

    
2769
  def CheckPrereq(self):
2770
    """Check prerequisites.
2771

2772
    This checks that the fields required are valid output fields.
2773

2774
    """
2775
    self.op.name = getattr(self.op, "name", None)
2776

    
2777
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
2778

    
2779
  def Exec(self, feedback_fn):
2780
    """Computes the list of nodes and their attributes.
2781

2782
    """
2783
    # Always get name to sort by
2784
    if constants.SF_NAME in self.op.output_fields:
2785
      fields = self.op.output_fields[:]
2786
    else:
2787
      fields = [constants.SF_NAME] + self.op.output_fields
2788

    
2789
    # Never ask for node or type as it's only known to the LU
2790
    for extra in [constants.SF_NODE, constants.SF_TYPE]:
2791
      while extra in fields:
2792
        fields.remove(extra)
2793

    
2794
    field_idx = dict([(name, idx) for (idx, name) in enumerate(fields)])
2795
    name_idx = field_idx[constants.SF_NAME]
2796

    
2797
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
2798
    data = self.rpc.call_storage_list(self.nodes,
2799
                                      self.op.storage_type, st_args,
2800
                                      self.op.name, fields)
2801

    
2802
    result = []
2803

    
2804
    for node in utils.NiceSort(self.nodes):
2805
      nresult = data[node]
2806
      if nresult.offline:
2807
        continue
2808

    
2809
      msg = nresult.fail_msg
2810
      if msg:
2811
        self.LogWarning("Can't get storage data from node %s: %s", node, msg)
2812
        continue
2813

    
2814
      rows = dict([(row[name_idx], row) for row in nresult.payload])
2815

    
2816
      for name in utils.NiceSort(rows.keys()):
2817
        row = rows[name]
2818

    
2819
        out = []
2820

    
2821
        for field in self.op.output_fields:
2822
          if field == constants.SF_NODE:
2823
            val = node
2824
          elif field == constants.SF_TYPE:
2825
            val = self.op.storage_type
2826
          elif field in field_idx:
2827
            val = row[field_idx[field]]
2828
          else:
2829
            raise errors.ParameterError(field)
2830

    
2831
          out.append(val)
2832

    
2833
        result.append(out)
2834

    
2835
    return result
2836

    
2837

    
2838
class LUModifyNodeStorage(NoHooksLU):
2839
  """Logical unit for modifying a storage volume on a node.
2840

2841
  """
2842
  _OP_REQP = ["node_name", "storage_type", "name", "changes"]
2843
  REQ_BGL = False
2844

    
2845
  def CheckArguments(self):
2846
    node_name = self.cfg.ExpandNodeName(self.op.node_name)
2847
    if node_name is None:
2848
      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name,
2849
                                 errors.ECODE_NOENT)
2850

    
2851
    self.op.node_name = node_name
2852

    
2853
    storage_type = self.op.storage_type
2854
    if storage_type not in constants.VALID_STORAGE_TYPES:
2855
      raise errors.OpPrereqError("Unknown storage type: %s" % storage_type,
2856
                                 errors.ECODE_INVAL)
2857

    
2858
  def ExpandNames(self):
2859
    self.needed_locks = {
2860
      locking.LEVEL_NODE: self.op.node_name,
2861
      }
2862

    
2863
  def CheckPrereq(self):
2864
    """Check prerequisites.
2865

2866
    """
2867
    storage_type = self.op.storage_type
2868

    
2869
    try:
2870
      modifiable = constants.MODIFIABLE_STORAGE_FIELDS[storage_type]
2871
    except KeyError:
2872
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
2873
                                 " modified" % storage_type,
2874
                                 errors.ECODE_INVAL)
2875

    
2876
    diff = set(self.op.changes.keys()) - modifiable
2877
    if diff:
2878
      raise errors.OpPrereqError("The following fields can not be modified for"
2879
                                 " storage units of type '%s': %r" %
2880
                                 (storage_type, list(diff)),
2881
                                 errors.ECODE_INVAL)
2882

    
2883
  def Exec(self, feedback_fn):
2884
    """Computes the list of nodes and their attributes.
2885

2886
    """
2887
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
2888
    result = self.rpc.call_storage_modify(self.op.node_name,
2889
                                          self.op.storage_type, st_args,
2890
                                          self.op.name, self.op.changes)
2891
    result.Raise("Failed to modify storage unit '%s' on %s" %
2892
                 (self.op.name, self.op.node_name))
2893

    
2894

    
2895
class LUAddNode(LogicalUnit):
2896
  """Logical unit for adding node to the cluster.
2897

2898
  """
2899
  HPATH = "node-add"
2900
  HTYPE = constants.HTYPE_NODE
2901
  _OP_REQP = ["node_name"]
2902

    
2903
  def BuildHooksEnv(self):
2904
    """Build hooks env.
2905

2906
    This will run on all nodes before, and on all nodes + the new node after.
2907

2908
    """
2909
    env = {
2910
      "OP_TARGET": self.op.node_name,
2911
      "NODE_NAME": self.op.node_name,
2912
      "NODE_PIP": self.op.primary_ip,
2913
      "NODE_SIP": self.op.secondary_ip,
2914
      }
2915
    nodes_0 = self.cfg.GetNodeList()
2916
    nodes_1 = nodes_0 + [self.op.node_name, ]
2917
    return env, nodes_0, nodes_1
2918

    
2919
  def CheckPrereq(self):
2920
    """Check prerequisites.
2921

2922
    This checks:
2923
     - the new node is not already in the config
2924
     - it is resolvable
2925
     - its parameters (single/dual homed) matches the cluster
2926

2927
    Any errors are signaled by raising errors.OpPrereqError.
2928

2929
    """
2930
    node_name = self.op.node_name
2931
    cfg = self.cfg
2932

    
2933
    dns_data = utils.GetHostInfo(node_name)
2934

    
2935
    node = dns_data.name
2936
    primary_ip = self.op.primary_ip = dns_data.ip
2937
    secondary_ip = getattr(self.op, "secondary_ip", None)
2938
    if secondary_ip is None:
2939
      secondary_ip = primary_ip
2940
    if not utils.IsValidIP(secondary_ip):
2941
      raise errors.OpPrereqError("Invalid secondary IP given",
2942
                                 errors.ECODE_INVAL)
2943
    self.op.secondary_ip = secondary_ip
2944

    
2945
    node_list = cfg.GetNodeList()
2946
    if not self.op.readd and node in node_list:
2947
      raise errors.OpPrereqError("Node %s is already in the configuration" %
2948
                                 node, errors.ECODE_EXISTS)
2949
    elif self.op.readd and node not in node_list:
2950
      raise errors.OpPrereqError("Node %s is not in the configuration" % node,
2951
                                 errors.ECODE_NOENT)
2952

    
2953
    for existing_node_name in node_list:
2954
      existing_node = cfg.GetNodeInfo(existing_node_name)
2955

    
2956
      if self.op.readd and node == existing_node_name:
2957
        if (existing_node.primary_ip != primary_ip or
2958
            existing_node.secondary_ip != secondary_ip):
2959
          raise errors.OpPrereqError("Readded node doesn't have the same IP"
2960
                                     " address configuration as before",
2961
                                     errors.ECODE_INVAL)
2962
        continue
2963

    
2964
      if (existing_node.primary_ip == primary_ip or
2965
          existing_node.secondary_ip == primary_ip or
2966
          existing_node.primary_ip == secondary_ip or
2967
          existing_node.secondary_ip == secondary_ip):
2968
        raise errors.OpPrereqError("New node ip address(es) conflict with"
2969
                                   " existing node %s" % existing_node.name,
2970
                                   errors.ECODE_NOTUNIQUE)
2971

    
2972
    # check that the type of the node (single versus dual homed) is the
2973
    # same as for the master
2974
    myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
2975
    master_singlehomed = myself.secondary_ip == myself.primary_ip
2976
    newbie_singlehomed = secondary_ip == primary_ip
2977
    if master_singlehomed != newbie_singlehomed:
2978
      if master_singlehomed:
2979
        raise errors.OpPrereqError("The master has no private ip but the"
2980
                                   " new node has one",
2981
                                   errors.ECODE_INVAL)
2982
      else:
2983
        raise errors.OpPrereqError("The master has a private ip but the"
2984
                                   " new node doesn't have one",
2985
                                   errors.ECODE_INVAL)
2986

    
2987
    # checks reachability
2988
    if not utils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
2989
      raise errors.OpPrereqError("Node not reachable by ping",
2990
                                 errors.ECODE_ENVIRON)
2991

    
2992
    if not newbie_singlehomed:
2993
      # check reachability from my secondary ip to newbie's secondary ip
2994
      if not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
2995
                           source=myself.secondary_ip):
2996
        raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
2997
                                   " based ping to noded port",
2998
                                   errors.ECODE_ENVIRON)
2999

    
3000
    if self.op.readd:
3001
      exceptions = [node]
3002
    else:
3003
      exceptions = []
3004

    
3005
    self.master_candidate = _DecideSelfPromotion(self, exceptions=exceptions)
3006

    
3007
    if self.op.readd:
3008
      self.new_node = self.cfg.GetNodeInfo(node)
3009
      assert self.new_node is not None, "Can't retrieve locked node %s" % node
3010
    else:
3011
      self.new_node = objects.Node(name=node,
3012
                                   primary_ip=primary_ip,
3013
                                   secondary_ip=secondary_ip,
3014
                                   master_candidate=self.master_candidate,
3015
                                   offline=False, drained=False)
3016

    
3017
  def Exec(self, feedback_fn):
3018
    """Adds the new node to the cluster.
3019

3020
    """
3021
    new_node = self.new_node
3022
    node = new_node.name
3023

    
3024
    # for re-adds, reset the offline/drained/master-candidate flags;
3025
    # we need to reset here, otherwise offline would prevent RPC calls
3026
    # later in the procedure; this also means that if the re-add
3027
    # fails, we are left with a non-offlined, broken node
3028
    if self.op.readd:
3029
      new_node.drained = new_node.offline = False # pylint: disable-msg=W0201
3030
      self.LogInfo("Readding a node, the offline/drained flags were reset")
3031
      # if we demote the node, we do cleanup later in the procedure
3032
      new_node.master_candidate = self.master_candidate
3033

    
3034
    # notify the user about any possible mc promotion
3035
    if new_node.master_candidate:
3036
      self.LogInfo("Node will be a master candidate")
3037

    
3038
    # check connectivity
3039
    result = self.rpc.call_version([node])[node]
3040
    result.Raise("Can't get version information from node %s" % node)
3041
    if constants.PROTOCOL_VERSION == result.payload:
3042
      logging.info("Communication to node %s fine, sw version %s match",
3043
                   node, result.payload)
3044
    else:
3045
      raise errors.OpExecError("Version mismatch master version %s,"
3046
                               " node version %s" %
3047
                               (constants.PROTOCOL_VERSION, result.payload))
3048

    
3049
    # setup ssh on node
3050
    if self.cfg.GetClusterInfo().modify_ssh_setup:
3051
      logging.info("Copy ssh key to node %s", node)
3052
      priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
3053
      keyarray = []
3054
      keyfiles = [constants.SSH_HOST_DSA_PRIV, constants.SSH_HOST_DSA_PUB,
3055
                  constants.SSH_HOST_RSA_PRIV, constants.SSH_HOST_RSA_PUB,
3056
                  priv_key, pub_key]
3057

    
3058
      for i in keyfiles:
3059
        keyarray.append(utils.ReadFile(i))
3060

    
3061
      result = self.rpc.call_node_add(node, keyarray[0], keyarray[1],
3062
                                      keyarray[2], keyarray[3], keyarray[4],
3063
                                      keyarray[5])
3064
      result.Raise("Cannot transfer ssh keys to the new node")
3065

    
3066
    # Add node to our /etc/hosts, and add key to known_hosts
3067
    if self.cfg.GetClusterInfo().modify_etc_hosts:
3068
      utils.AddHostToEtcHosts(new_node.name)
3069

    
3070
    if new_node.secondary_ip != new_node.primary_ip:
3071
      result = self.rpc.call_node_has_ip_address(new_node.name,
3072
                                                 new_node.secondary_ip)
3073
      result.Raise("Failure checking secondary ip on node %s" % new_node.name,
3074
                   prereq=True, ecode=errors.ECODE_ENVIRON)
3075
      if not result.payload:
3076
        raise errors.OpExecError("Node claims it doesn't have the secondary ip"
3077
                                 " you gave (%s). Please fix and re-run this"
3078
                                 " command." % new_node.secondary_ip)
3079

    
3080
    node_verify_list = [self.cfg.GetMasterNode()]
3081
    node_verify_param = {
3082
      constants.NV_NODELIST: [node],
3083
      # TODO: do a node-net-test as well?
3084
    }
3085

    
3086
    result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
3087
                                       self.cfg.GetClusterName())
3088
    for verifier in node_verify_list:
3089
      result[verifier].Raise("Cannot communicate with node %s" % verifier)
3090
      nl_payload = result[verifier].payload[constants.NV_NODELIST]
3091
      if nl_payload:
3092
        for failed in nl_payload:
3093
          feedback_fn("ssh/hostname verification failed"
3094
                      " (checking from %s): %s" %
3095
                      (verifier, nl_payload[failed]))
3096
        raise errors.OpExecError("ssh/hostname verification failed.")
3097

    
3098
    if self.op.readd:
3099
      _RedistributeAncillaryFiles(self)
3100
      self.context.ReaddNode(new_node)
3101
      # make sure we redistribute the config
3102
      self.cfg.Update(new_node, feedback_fn)
3103
      # and make sure the new node will not have old files around
3104
      if not new_node.master_candidate:
3105
        result = self.rpc.call_node_demote_from_mc(new_node.name)
3106
        msg = result.fail_msg
3107
        if msg:
3108
          self.LogWarning("Node failed to demote itself from master"
3109
                          " candidate status: %s" % msg)
3110
    else:
3111
      _RedistributeAncillaryFiles(self, additional_nodes=[node])
3112
      self.context.AddNode(new_node, self.proc.GetECId())
3113

    
3114

    
3115
class LUSetNodeParams(LogicalUnit):
3116
  """Modifies the parameters of a node.
3117

3118
  """
3119
  HPATH = "node-modify"
3120
  HTYPE = constants.HTYPE_NODE
3121
  _OP_REQP = ["node_name"]
3122
  REQ_BGL = False
3123

    
3124
  def CheckArguments(self):
3125
    node_name = self.cfg.ExpandNodeName(self.op.node_name)
3126
    if node_name is None:
3127
      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name,
3128
                                 errors.ECODE_INVAL)
3129
    self.op.node_name = node_name
3130
    _CheckBooleanOpField(self.op, 'master_candidate')
3131
    _CheckBooleanOpField(self.op, 'offline')
3132
    _CheckBooleanOpField(self.op, 'drained')
3133
    all_mods = [self.op.offline, self.op.master_candidate, self.op.drained]
3134
    if all_mods.count(None) == 3:
3135
      raise errors.OpPrereqError("Please pass at least one modification",
3136
                                 errors.ECODE_INVAL)
3137
    if all_mods.count(True) > 1:
3138
      raise errors.OpPrereqError("Can't set the node into more than one"
3139
                                 " state at the same time",
3140
                                 errors.ECODE_INVAL)
3141

    
3142
  def ExpandNames(self):
3143
    self.needed_locks = {locking.LEVEL_NODE: self.op.node_name}
3144

    
3145
  def BuildHooksEnv(self):
3146
    """Build hooks env.
3147

3148
    This runs on the master node.
3149

3150
    """
3151
    env = {
3152
      "OP_TARGET": self.op.node_name,
3153
      "MASTER_CANDIDATE": str(self.op.master_candidate),
3154
      "OFFLINE": str(self.op.offline),
3155
      "DRAINED": str(self.op.drained),
3156
      }
3157
    nl = [self.cfg.GetMasterNode(),
3158
          self.op.node_name]
3159
    return env, nl, nl
3160

    
3161
  def CheckPrereq(self):
3162
    """Check prerequisites.
3163

3164
    This only checks the instance list against the existing names.
3165

3166
    """
3167
    node = self.node = self.cfg.GetNodeInfo(self.op.node_name)
3168

    
3169
    if (self.op.master_candidate is not None or
3170
        self.op.drained is not None or
3171
        self.op.offline is not None):
3172
      # we can't change the master's node flags
3173
      if self.op.node_name == self.cfg.GetMasterNode():
3174
        raise errors.OpPrereqError("The master role can be changed"
3175
                                   " only via masterfailover",
3176
                                   errors.ECODE_INVAL)
3177

    
3178
    # Boolean value that tells us whether we're offlining or draining the node
3179
    offline_or_drain = self.op.offline == True or self.op.drained == True
3180
    deoffline_or_drain = self.op.offline == False or self.op.drained == False
3181

    
3182
    if (node.master_candidate and
3183
        (self.op.master_candidate == False or offline_or_drain)):
3184
      cp_size = self.cfg.GetClusterInfo().candidate_pool_size
3185
      mc_now, mc_should, mc_max = self.cfg.GetMasterCandidateStats()
3186
      if mc_now <= cp_size:
3187
        msg = ("Not enough master candidates (desired"
3188
               " %d, new value will be %d)" % (cp_size, mc_now-1))
3189
        # Only allow forcing the operation if it's an offline/drain operation,
3190
        # and we could not possibly promote more nodes.
3191
        # FIXME: this can still lead to issues if in any way another node which
3192
        # could be promoted appears in the meantime.
3193
        if self.op.force and offline_or_drain and mc_should == mc_max:
3194
          self.LogWarning(msg)
3195
        else:
3196
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
3197

    
3198
    if (self.op.master_candidate == True and
3199
        ((node.offline and not self.op.offline == False) or
3200
         (node.drained and not self.op.drained == False))):
3201
      raise errors.OpPrereqError("Node '%s' is offline or drained, can't set"
3202
                                 " to master_candidate" % node.name,
3203
                                 errors.ECODE_INVAL)
3204

    
3205
    # If we're being deofflined/drained, we'll MC ourself if needed
3206
    if (deoffline_or_drain and not offline_or_drain and not
3207
        self.op.master_candidate == True and not node.master_candidate):
3208
      self.op.master_candidate = _DecideSelfPromotion(self)
3209
      if self.op.master_candidate:
3210
        self.LogInfo("Autopromoting node to master candidate")
3211

    
3212
    return
3213

    
3214
  def Exec(self, feedback_fn):
3215
    """Modifies a node.
3216

3217
    """
3218
    node = self.node
3219

    
3220
    result = []
3221
    changed_mc = False
3222

    
3223
    if self.op.offline is not None:
3224
      node.offline = self.op.offline
3225
      result.append(("offline", str(self.op.offline)))
3226
      if self.op.offline == True:
3227
        if node.master_candidate:
3228
          node.master_candidate = False
3229
          changed_mc = True
3230
          result.append(("master_candidate", "auto-demotion due to offline"))
3231
        if node.drained:
3232
          node.drained = False
3233
          result.append(("drained", "clear drained status due to offline"))
3234

    
3235
    if self.op.master_candidate is not None:
3236
      node.master_candidate = self.op.master_candidate
3237
      changed_mc = True
3238
      result.append(("master_candidate", str(self.op.master_candidate)))
3239
      if self.op.master_candidate == False:
3240
        rrc = self.rpc.call_node_demote_from_mc(node.name)
3241
        msg = rrc.fail_msg
3242
        if msg:
3243
          self.LogWarning("Node failed to demote itself: %s" % msg)
3244

    
3245
    if self.op.drained is not None:
3246
      node.drained = self.op.drained
3247
      result.append(("drained", str(self.op.drained)))
3248
      if self.op.drained == True:
3249
        if node.master_candidate:
3250
          node.master_candidate = False
3251
          changed_mc = True
3252
          result.append(("master_candidate", "auto-demotion due to drain"))
3253
          rrc = self.rpc.call_node_demote_from_mc(node.name)
3254
          msg = rrc.fail_msg
3255
          if msg:
3256
            self.LogWarning("Node failed to demote itself: %s" % msg)
3257
        if node.offline:
3258
          node.offline = False
3259
          result.append(("offline", "clear offline status due to drain"))
3260

    
3261
    # this will trigger configuration file update, if needed
3262
    self.cfg.Update(node, feedback_fn)
3263
    # this will trigger job queue propagation or cleanup
3264
    if changed_mc:
3265
      self.context.ReaddNode(node)
3266

    
3267
    return result
3268

    
3269

    
3270
class LUPowercycleNode(NoHooksLU):
3271
  """Powercycles a node.
3272

3273
  """
3274
  _OP_REQP = ["node_name", "force"]
3275
  REQ_BGL = False
3276

    
3277
  def CheckArguments(self):
3278
    node_name = self.cfg.ExpandNodeName(self.op.node_name)
3279
    if node_name is None:
3280
      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name,
3281
                                 errors.ECODE_NOENT)
3282
    self.op.node_name = node_name
3283
    if node_name == self.cfg.GetMasterNode() and not self.op.force:
3284
      raise errors.OpPrereqError("The node is the master and the force"
3285
                                 " parameter was not set",
3286
                                 errors.ECODE_INVAL)
3287

    
3288
  def ExpandNames(self):
3289
    """Locking for PowercycleNode.
3290

3291
    This is a last-resort option and shouldn't block on other
3292
    jobs. Therefore, we grab no locks.
3293

3294
    """
3295
    self.needed_locks = {}
3296

    
3297
  def CheckPrereq(self):
3298
    """Check prerequisites.
3299

3300
    This LU has no prereqs.
3301

3302
    """
3303
    pass
3304

    
3305
  def Exec(self, feedback_fn):
3306
    """Reboots a node.
3307

3308
    """
3309
    result = self.rpc.call_node_powercycle(self.op.node_name,
3310
                                           self.cfg.GetHypervisorType())
3311
    result.Raise("Failed to schedule the reboot")
3312
    return result.payload
3313

    
3314

    
3315
class LUQueryClusterInfo(NoHooksLU):
3316
  """Query cluster configuration.
3317

3318
  """
3319
  _OP_REQP = []
3320
  REQ_BGL = False
3321

    
3322
  def ExpandNames(self):
3323
    self.needed_locks = {}
3324

    
3325
  def CheckPrereq(self):
3326
    """No prerequsites needed for this LU.
3327

3328
    """
3329
    pass
3330

    
3331
  def Exec(self, feedback_fn):
3332
    """Return cluster config.
3333

3334
    """
3335
    cluster = self.cfg.GetClusterInfo()
3336
    result = {
3337
      "software_version": constants.RELEASE_VERSION,
3338
      "protocol_version": constants.PROTOCOL_VERSION,
3339
      "config_version": constants.CONFIG_VERSION,
3340
      "os_api_version": max(constants.OS_API_VERSIONS),
3341
      "export_version": constants.EXPORT_VERSION,
3342
      "architecture": (platform.architecture()[0], platform.machine()),
3343
      "name": cluster.cluster_name,
3344
      "master": cluster.master_node,
3345
      "default_hypervisor": cluster.enabled_hypervisors[0],
3346
      "enabled_hypervisors": cluster.enabled_hypervisors,
3347
      "hvparams": dict([(hypervisor_name, cluster.hvparams[hypervisor_name])
3348
                        for hypervisor_name in cluster.enabled_hypervisors]),
3349
      "beparams": cluster.beparams,
3350
      "nicparams": cluster.nicparams,
3351
      "candidate_pool_size": cluster.candidate_pool_size,
3352
      "master_netdev": cluster.master_netdev,
3353
      "volume_group_name": cluster.volume_group_name,
3354
      "file_storage_dir": cluster.file_storage_dir,
3355
      "ctime": cluster.ctime,
3356
      "mtime": cluster.mtime,
3357
      "uuid": cluster.uuid,
3358
      "tags": list(cluster.GetTags()),
3359
      }
3360

    
3361
    return result
3362

    
3363

    
3364
class LUQueryConfigValues(NoHooksLU):
3365
  """Return configuration values.
3366

3367
  """
3368
  _OP_REQP = []
3369
  REQ_BGL = False
3370
  _FIELDS_DYNAMIC = utils.FieldSet()
3371
  _FIELDS_STATIC = utils.FieldSet("cluster_name", "master_node", "drain_flag",
3372
                                  "watcher_pause")
3373

    
3374
  def ExpandNames(self):
3375
    self.needed_locks = {}
3376

    
3377
    _CheckOutputFields(static=self._FIELDS_STATIC,
3378
                       dynamic=self._FIELDS_DYNAMIC,
3379
                       selected=self.op.output_fields)
3380

    
3381
  def CheckPrereq(self):
3382
    """No prerequisites.
3383

3384
    """
3385
    pass
3386

    
3387
  def Exec(self, feedback_fn):
3388
    """Dump a representation of the cluster config to the standard output.
3389

3390
    """
3391
    values = []
3392
    for field in self.op.output_fields:
3393
      if field == "cluster_name":
3394
        entry = self.cfg.GetClusterName()
3395
      elif field == "master_node":
3396
        entry = self.cfg.GetMasterNode()
3397
      elif field == "drain_flag":
3398
        entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
3399
      elif field == "watcher_pause":
3400
        return utils.ReadWatcherPauseFile(constants.WATCHER_PAUSEFILE)
3401
      else:
3402
        raise errors.ParameterError(field)
3403
      values.append(entry)
3404
    return values
3405

    
3406

    
3407
class LUActivateInstanceDisks(NoHooksLU):
3408
  """Bring up an instance's disks.
3409

3410
  """
3411
  _OP_REQP = ["instance_name"]
3412
  REQ_BGL = False
3413

    
3414
  def ExpandNames(self):
3415
    self._ExpandAndLockInstance()
3416
    self.needed_locks[locking.LEVEL_NODE] = []
3417
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3418

    
3419
  def DeclareLocks(self, level):
3420
    if level == locking.LEVEL_NODE:
3421
      self._LockInstancesNodes()
3422

    
3423
  def CheckPrereq(self):
3424
    """Check prerequisites.
3425

3426
    This checks that the instance is in the cluster.
3427

3428
    """
3429
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3430
    assert self.instance is not None, \
3431
      "Cannot retrieve locked instance %s" % self.op.instance_name
3432
    _CheckNodeOnline(self, self.instance.primary_node)
3433
    if not hasattr(self.op, "ignore_size"):
3434
      self.op.ignore_size = False
3435

    
3436
  def Exec(self, feedback_fn):
3437
    """Activate the disks.
3438

3439
    """
3440
    disks_ok, disks_info = \
3441
              _AssembleInstanceDisks(self, self.instance,
3442
                                     ignore_size=self.op.ignore_size)
3443
    if not disks_ok:
3444
      raise errors.OpExecError("Cannot activate block devices")
3445

    
3446
    return disks_info
3447

    
3448

    
3449
def _AssembleInstanceDisks(lu, instance, ignore_secondaries=False,
3450
                           ignore_size=False):
3451
  """Prepare the block devices for an instance.
3452

3453
  This sets up the block devices on all nodes.
3454

3455
  @type lu: L{LogicalUnit}
3456
  @param lu: the logical unit on whose behalf we execute
3457
  @type instance: L{objects.Instance}
3458
  @param instance: the instance for whose disks we assemble
3459
  @type ignore_secondaries: boolean
3460
  @param ignore_secondaries: if true, errors on secondary nodes
3461
      won't result in an error return from the function
3462
  @type ignore_size: boolean
3463
  @param ignore_size: if true, the current known size of the disk
3464
      will not be used during the disk activation, useful for cases
3465
      when the size is wrong
3466
  @return: False if the operation failed, otherwise a list of
3467
      (host, instance_visible_name, node_visible_name)
3468
      with the mapping from node devices to instance devices
3469

3470
  """
3471
  device_info = []
3472
  disks_ok = True
3473
  iname = instance.name
3474
  # With the two passes mechanism we try to reduce the window of
3475
  # opportunity for the race condition of switching DRBD to primary
3476
  # before handshaking occured, but we do not eliminate it
3477

    
3478
  # The proper fix would be to wait (with some limits) until the
3479
  # connection has been made and drbd transitions from WFConnection
3480
  # into any other network-connected state (Connected, SyncTarget,
3481
  # SyncSource, etc.)
3482

    
3483
  # 1st pass, assemble on all nodes in secondary mode
3484
  for inst_disk in instance.disks:
3485
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
3486
      if ignore_size:
3487
        node_disk = node_disk.Copy()
3488
        node_disk.UnsetSize()
3489
      lu.cfg.SetDiskID(node_disk, node)
3490
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
3491
      msg = result.fail_msg
3492
      if msg:
3493
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
3494
                           " (is_primary=False, pass=1): %s",
3495
                           inst_disk.iv_name, node, msg)
3496
        if not ignore_secondaries:
3497
          disks_ok = False
3498

    
3499
  # FIXME: race condition on drbd migration to primary
3500

    
3501
  # 2nd pass, do only the primary node
3502
  for inst_disk in instance.disks:
3503
    dev_path = None
3504

    
3505
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
3506
      if node != instance.primary_node:
3507
        continue
3508
      if ignore_size:
3509
        node_disk = node_disk.Copy()
3510
        node_disk.UnsetSize()
3511
      lu.cfg.SetDiskID(node_disk, node)
3512
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
3513
      msg = result.fail_msg
3514
      if msg:
3515
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
3516
                           " (is_primary=True, pass=2): %s",
3517
                           inst_disk.iv_name, node, msg)
3518
        disks_ok = False
3519
      else:
3520
        dev_path = result.payload
3521

    
3522
    device_info.append((instance.primary_node, inst_disk.iv_name, dev_path))
3523

    
3524
  # leave the disks configured for the primary node
3525
  # this is a workaround that would be fixed better by
3526
  # improving the logical/physical id handling
3527
  for disk in instance.disks:
3528
    lu.cfg.SetDiskID(disk, instance.primary_node)
3529

    
3530
  return disks_ok, device_info
3531

    
3532

    
3533
def _StartInstanceDisks(lu, instance, force):
3534
  """Start the disks of an instance.
3535

3536
  """
3537
  disks_ok, _ = _AssembleInstanceDisks(lu, instance,
3538
                                           ignore_secondaries=force)
3539
  if not disks_ok:
3540
    _ShutdownInstanceDisks(lu, instance)
3541
    if force is not None and not force:
3542
      lu.proc.LogWarning("", hint="If the message above refers to a"
3543
                         " secondary node,"
3544
                         " you can retry the operation using '--force'.")
3545
    raise errors.OpExecError("Disk consistency error")
3546

    
3547

    
3548
class LUDeactivateInstanceDisks(NoHooksLU):
3549
  """Shutdown an instance's disks.
3550

3551
  """
3552
  _OP_REQP = ["instance_name"]
3553
  REQ_BGL = False
3554

    
3555
  def ExpandNames(self):
3556
    self._ExpandAndLockInstance()
3557
    self.needed_locks[locking.LEVEL_NODE] = []
3558
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3559

    
3560
  def DeclareLocks(self, level):
3561
    if level == locking.LEVEL_NODE:
3562
      self._LockInstancesNodes()
3563

    
3564
  def CheckPrereq(self):
3565
    """Check prerequisites.
3566

3567
    This checks that the instance is in the cluster.
3568

3569
    """
3570
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3571
    assert self.instance is not None, \
3572
      "Cannot retrieve locked instance %s" % self.op.instance_name
3573

    
3574
  def Exec(self, feedback_fn):
3575
    """Deactivate the disks
3576

3577
    """
3578
    instance = self.instance
3579
    _SafeShutdownInstanceDisks(self, instance)
3580

    
3581

    
3582
def _SafeShutdownInstanceDisks(lu, instance):
3583
  """Shutdown block devices of an instance.
3584

3585
  This function checks if an instance is running, before calling
3586
  _ShutdownInstanceDisks.
3587

3588
  """
3589
  pnode = instance.primary_node
3590
  ins_l = lu.rpc.call_instance_list([pnode], [instance.hypervisor])[pnode]
3591
  ins_l.Raise("Can't contact node %s" % pnode)
3592

    
3593
  if instance.name in ins_l.payload:
3594
    raise errors.OpExecError("Instance is running, can't shutdown"
3595
                             " block devices.")
3596

    
3597
  _ShutdownInstanceDisks(lu, instance)
3598

    
3599

    
3600
def _ShutdownInstanceDisks(lu, instance, ignore_primary=False):
3601
  """Shutdown block devices of an instance.
3602

3603
  This does the shutdown on all nodes of the instance.
3604

3605
  If the ignore_primary is false, errors on the primary node are
3606
  ignored.
3607

3608
  """
3609
  all_result = True
3610
  for disk in instance.disks:
3611
    for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
3612
      lu.cfg.SetDiskID(top_disk, node)
3613
      result = lu.rpc.call_blockdev_shutdown(node, top_disk)
3614
      msg = result.fail_msg
3615
      if msg:
3616
        lu.LogWarning("Could not shutdown block device %s on node %s: %s",
3617
                      disk.iv_name, node, msg)
3618
        if not ignore_primary or node != instance.primary_node:
3619
          all_result = False
3620
  return all_result
3621

    
3622

    
3623
def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor_name):
3624
  """Checks if a node has enough free memory.
3625

3626
  This function check if a given node has the needed amount of free
3627
  memory. In case the node has less memory or we cannot get the
3628
  information from the node, this function raise an OpPrereqError
3629
  exception.
3630

3631
  @type lu: C{LogicalUnit}
3632
  @param lu: a logical unit from which we get configuration data
3633
  @type node: C{str}
3634
  @param node: the node to check
3635
  @type reason: C{str}
3636
  @param reason: string to use in the error message
3637
  @type requested: C{int}
3638
  @param requested: the amount of memory in MiB to check for
3639
  @type hypervisor_name: C{str}
3640
  @param hypervisor_name: the hypervisor to ask for memory stats
3641
  @raise errors.OpPrereqError: if the node doesn't have enough memory, or
3642
      we cannot check the node
3643

3644
  """
3645
  nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor_name)
3646
  nodeinfo[node].Raise("Can't get data from node %s" % node,
3647
                       prereq=True, ecode=errors.ECODE_ENVIRON)
3648
  free_mem = nodeinfo[node].payload.get('memory_free', None)
3649
  if not isinstance(free_mem, int):
3650
    raise errors.OpPrereqError("Can't compute free memory on node %s, result"
3651
                               " was '%s'" % (node, free_mem),
3652
                               errors.ECODE_ENVIRON)
3653
  if requested > free_mem:
3654
    raise errors.OpPrereqError("Not enough memory on node %s for %s:"
3655
                               " needed %s MiB, available %s MiB" %
3656
                               (node, reason, requested, free_mem),
3657
                               errors.ECODE_NORES)
3658

    
3659

    
3660
class LUStartupInstance(LogicalUnit):
3661
  """Starts an instance.
3662

3663
  """
3664
  HPATH = "instance-start"
3665
  HTYPE = constants.HTYPE_INSTANCE
3666
  _OP_REQP = ["instance_name", "force"]
3667
  REQ_BGL = False
3668

    
3669
  def ExpandNames(self):
3670
    self._ExpandAndLockInstance()
3671

    
3672
  def BuildHooksEnv(self):
3673
    """Build hooks env.
3674

3675
    This runs on master, primary and secondary nodes of the instance.
3676

3677
    """
3678
    env = {
3679
      "FORCE": self.op.force,
3680
      }
3681
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
3682
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3683
    return env, nl, nl
3684

    
3685
  def CheckPrereq(self):
3686
    """Check prerequisites.
3687

3688
    This checks that the instance is in the cluster.
3689

3690
    """
3691
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3692
    assert self.instance is not None, \
3693
      "Cannot retrieve locked instance %s" % self.op.instance_name
3694

    
3695
    # extra beparams
3696
    self.beparams = getattr(self.op, "beparams", {})
3697
    if self.beparams:
3698
      if not isinstance(self.beparams, dict):
3699
        raise errors.OpPrereqError("Invalid beparams passed: %s, expected"
3700
                                   " dict" % (type(self.beparams), ),
3701
                                   errors.ECODE_INVAL)
3702
      # fill the beparams dict
3703
      utils.ForceDictType(self.beparams, constants.BES_PARAMETER_TYPES)
3704
      self.op.beparams = self.beparams
3705

    
3706
    # extra hvparams
3707
    self.hvparams = getattr(self.op, "hvparams", {})
3708
    if self.hvparams:
3709
      if not isinstance(self.hvparams, dict):
3710
        raise errors.OpPrereqError("Invalid hvparams passed: %s, expected"
3711
                                   " dict" % (type(self.hvparams), ),
3712
                                   errors.ECODE_INVAL)
3713

    
3714
      # check hypervisor parameter syntax (locally)
3715
      cluster = self.cfg.GetClusterInfo()
3716
      utils.ForceDictType(self.hvparams, constants.HVS_PARAMETER_TYPES)
3717
      filled_hvp = objects.FillDict(cluster.hvparams[instance.hypervisor],
3718
                                    instance.hvparams)
3719
      filled_hvp.update(self.hvparams)
3720
      hv_type = hypervisor.GetHypervisor(instance.hypervisor)
3721
      hv_type.CheckParameterSyntax(filled_hvp)
3722
      _CheckHVParams(self, instance.all_nodes, instance.hypervisor, filled_hvp)
3723
      self.op.hvparams = self.hvparams
3724

    
3725
    _CheckNodeOnline(self, instance.primary_node)
3726

    
3727
    bep = self.cfg.GetClusterInfo().FillBE(instance)
3728
    # check bridges existence
3729
    _CheckInstanceBridgesExist(self, instance)
3730

    
3731
    remote_info = self.rpc.call_instance_info(instance.primary_node,
3732
                                              instance.name,
3733
                                              instance.hypervisor)
3734
    remote_info.Raise("Error checking node %s" % instance.primary_node,
3735
                      prereq=True, ecode=errors.ECODE_ENVIRON)
3736
    if not remote_info.payload: # not running already
3737
      _CheckNodeFreeMemory(self, instance.primary_node,
3738
                           "starting instance %s" % instance.name,
3739
                           bep[constants.BE_MEMORY], instance.hypervisor)
3740

    
3741
  def Exec(self, feedback_fn):
3742
    """Start the instance.
3743

3744
    """
3745
    instance = self.instance
3746
    force = self.op.force
3747

    
3748
    self.cfg.MarkInstanceUp(instance.name)
3749

    
3750
    node_current = instance.primary_node
3751

    
3752
    _StartInstanceDisks(self, instance, force)
3753

    
3754
    result = self.rpc.call_instance_start(node_current, instance,
3755
                                          self.hvparams, self.beparams)
3756
    msg = result.fail_msg
3757
    if msg:
3758
      _ShutdownInstanceDisks(self, instance)
3759
      raise errors.OpExecError("Could not start instance: %s" % msg)
3760

    
3761

    
3762
class LURebootInstance(LogicalUnit):
3763
  """Reboot an instance.
3764

3765
  """
3766
  HPATH = "instance-reboot"
3767
  HTYPE = constants.HTYPE_INSTANCE
3768
  _OP_REQP = ["instance_name", "ignore_secondaries", "reboot_type"]
3769
  REQ_BGL = False
3770

    
3771
  def CheckArguments(self):
3772
    """Check the arguments.
3773

3774
    """
3775
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
3776
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
3777

    
3778
  def ExpandNames(self):
3779
    if self.op.reboot_type not in [constants.INSTANCE_REBOOT_SOFT,
3780
                                   constants.INSTANCE_REBOOT_HARD,
3781
                                   constants.INSTANCE_REBOOT_FULL]:
3782
      raise errors.ParameterError("reboot type not in [%s, %s, %s]" %
3783
                                  (constants.INSTANCE_REBOOT_SOFT,
3784
                                   constants.INSTANCE_REBOOT_HARD,
3785
                                   constants.INSTANCE_REBOOT_FULL))
3786
    self._ExpandAndLockInstance()
3787

    
3788
  def BuildHooksEnv(self):
3789
    """Build hooks env.
3790

3791
    This runs on master, primary and secondary nodes of the instance.
3792

3793
    """
3794
    env = {
3795
      "IGNORE_SECONDARIES": self.op.ignore_secondaries,
3796
      "REBOOT_TYPE": self.op.reboot_type,
3797
      "SHUTDOWN_TIMEOUT": self.shutdown_timeout,
3798
      }
3799
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
3800
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3801
    return env, nl, nl
3802

    
3803
  def CheckPrereq(self):
3804
    """Check prerequisites.
3805

3806
    This checks that the instance is in the cluster.
3807

3808
    """
3809
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3810
    assert self.instance is not None, \
3811
      "Cannot retrieve locked instance %s" % self.op.instance_name
3812

    
3813
    _CheckNodeOnline(self, instance.primary_node)
3814

    
3815
    # check bridges existence
3816
    _CheckInstanceBridgesExist(self, instance)
3817

    
3818
  def Exec(self, feedback_fn):
3819
    """Reboot the instance.
3820

3821
    """
3822
    instance = self.instance
3823
    ignore_secondaries = self.op.ignore_secondaries
3824
    reboot_type = self.op.reboot_type
3825

    
3826
    node_current = instance.primary_node
3827

    
3828
    if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
3829
                       constants.INSTANCE_REBOOT_HARD]:
3830
      for disk in instance.disks:
3831
        self.cfg.SetDiskID(disk, node_current)
3832
      result = self.rpc.call_instance_reboot(node_current, instance,
3833
                                             reboot_type,
3834
                                             self.shutdown_timeout)
3835
      result.Raise("Could not reboot instance")
3836
    else:
3837
      result = self.rpc.call_instance_shutdown(node_current, instance,
3838
                                               self.shutdown_timeout)
3839
      result.Raise("Could not shutdown instance for full reboot")
3840
      _ShutdownInstanceDisks(self, instance)
3841
      _StartInstanceDisks(self, instance, ignore_secondaries)
3842
      result = self.rpc.call_instance_start(node_current, instance, None, None)
3843
      msg = result.fail_msg
3844
      if msg:
3845
        _ShutdownInstanceDisks(self, instance)
3846
        raise errors.OpExecError("Could not start instance for"
3847
                                 " full reboot: %s" % msg)
3848

    
3849
    self.cfg.MarkInstanceUp(instance.name)
3850

    
3851

    
3852
class LUShutdownInstance(LogicalUnit):
3853
  """Shutdown an instance.
3854

3855
  """
3856
  HPATH = "instance-stop"
3857
  HTYPE = constants.HTYPE_INSTANCE
3858
  _OP_REQP = ["instance_name"]
3859
  REQ_BGL = False
3860

    
3861
  def CheckArguments(self):
3862
    """Check the arguments.
3863

3864
    """
3865
    self.timeout = getattr(self.op, "timeout",
3866
                           constants.DEFAULT_SHUTDOWN_TIMEOUT)
3867

    
3868
  def ExpandNames(self):
3869
    self._ExpandAndLockInstance()
3870

    
3871
  def BuildHooksEnv(self):
3872
    """Build hooks env.
3873

3874
    This runs on master, primary and secondary nodes of the instance.
3875

3876
    """
3877
    env = _BuildInstanceHookEnvByObject(self, self.instance)
3878
    env["TIMEOUT"] = self.timeout
3879
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3880
    return env, nl, nl
3881

    
3882
  def CheckPrereq(self):
3883
    """Check prerequisites.
3884

3885
    This checks that the instance is in the cluster.
3886

3887
    """
3888
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3889
    assert self.instance is not None, \
3890
      "Cannot retrieve locked instance %s" % self.op.instance_name
3891
    _CheckNodeOnline(self, self.instance.primary_node)
3892

    
3893
  def Exec(self, feedback_fn):
3894
    """Shutdown the instance.
3895

3896
    """
3897
    instance = self.instance
3898
    node_current = instance.primary_node
3899
    timeout = self.timeout
3900
    self.cfg.MarkInstanceDown(instance.name)
3901
    result = self.rpc.call_instance_shutdown(node_current, instance, timeout)
3902
    msg = result.fail_msg
3903
    if msg:
3904
      self.proc.LogWarning("Could not shutdown instance: %s" % msg)
3905

    
3906
    _ShutdownInstanceDisks(self, instance)
3907

    
3908

    
3909
class LUReinstallInstance(LogicalUnit):
3910
  """Reinstall an instance.
3911

3912
  """
3913
  HPATH = "instance-reinstall"
3914
  HTYPE = constants.HTYPE_INSTANCE
3915
  _OP_REQP = ["instance_name"]
3916
  REQ_BGL = False
3917

    
3918
  def ExpandNames(self):
3919
    self._ExpandAndLockInstance()
3920

    
3921
  def BuildHooksEnv(self):
3922
    """Build hooks env.
3923

3924
    This runs on master, primary and secondary nodes of the instance.
3925

3926
    """
3927
    env = _BuildInstanceHookEnvByObject(self, self.instance)
3928
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
3929
    return env, nl, nl
3930

    
3931
  def CheckPrereq(self):
3932
    """Check prerequisites.
3933

3934
    This checks that the instance is in the cluster and is not running.
3935

3936
    """
3937
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3938
    assert instance is not None, \
3939
      "Cannot retrieve locked instance %s" % self.op.instance_name
3940
    _CheckNodeOnline(self, instance.primary_node)
3941

    
3942
    if instance.disk_template == constants.DT_DISKLESS:
3943
      raise errors.OpPrereqError("Instance '%s' has no disks" %
3944
                                 self.op.instance_name,
3945
                                 errors.ECODE_INVAL)
3946
    if instance.admin_up:
3947
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
3948
                                 self.op.instance_name,
3949
                                 errors.ECODE_STATE)
3950
    remote_info = self.rpc.call_instance_info(instance.primary_node,
3951
                                              instance.name,
3952
                                              instance.hypervisor)
3953
    remote_info.Raise("Error checking node %s" % instance.primary_node,
3954
                      prereq=True, ecode=errors.ECODE_ENVIRON)
3955
    if remote_info.payload:
3956
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
3957
                                 (self.op.instance_name,
3958
                                  instance.primary_node),
3959
                                 errors.ECODE_STATE)
3960

    
3961
    self.op.os_type = getattr(self.op, "os_type", None)
3962
    self.op.force_variant = getattr(self.op, "force_variant", False)
3963
    if self.op.os_type is not None:
3964
      # OS verification
3965
      pnode = self.cfg.GetNodeInfo(
3966
        self.cfg.ExpandNodeName(instance.primary_node))
3967
      if pnode is None:
3968
        raise errors.OpPrereqError("Primary node '%s' is unknown" %
3969
                                   self.op.pnode, errors.ECODE_NOENT)
3970
      result = self.rpc.call_os_get(pnode.name, self.op.os_type)
3971
      result.Raise("OS '%s' not in supported OS list for primary node %s" %
3972
                   (self.op.os_type, pnode.name),
3973
                   prereq=True, ecode=errors.ECODE_INVAL)
3974
      if not self.op.force_variant:
3975
        _CheckOSVariant(result.payload, self.op.os_type)
3976

    
3977
    self.instance = instance
3978

    
3979
  def Exec(self, feedback_fn):
3980
    """Reinstall the instance.
3981

3982
    """
3983
    inst = self.instance
3984

    
3985
    if self.op.os_type is not None:
3986
      feedback_fn("Changing OS to '%s'..." % self.op.os_type)
3987
      inst.os = self.op.os_type
3988
      self.cfg.Update(inst, feedback_fn)
3989

    
3990
    _StartInstanceDisks(self, inst, None)
3991
    try:
3992
      feedback_fn("Running the instance OS create scripts...")
3993
      result = self.rpc.call_instance_os_add(inst.primary_node, inst, True)
3994
      result.Raise("Could not install OS for instance %s on node %s" %
3995
                   (inst.name, inst.primary_node))
3996
    finally:
3997
      _ShutdownInstanceDisks(self, inst)
3998

    
3999

    
4000
class LURecreateInstanceDisks(LogicalUnit):
4001
  """Recreate an instance's missing disks.
4002

4003
  """
4004
  HPATH = "instance-recreate-disks"
4005
  HTYPE = constants.HTYPE_INSTANCE
4006
  _OP_REQP = ["instance_name", "disks"]
4007
  REQ_BGL = False
4008

    
4009
  def CheckArguments(self):
4010
    """Check the arguments.
4011

4012
    """
4013
    if not isinstance(self.op.disks, list):
4014
      raise errors.OpPrereqError("Invalid disks parameter", errors.ECODE_INVAL)
4015
    for item in self.op.disks:
4016
      if (not isinstance(item, int) or
4017
          item < 0):
4018
        raise errors.OpPrereqError("Invalid disk specification '%s'" %
4019
                                   str(item), errors.ECODE_INVAL)
4020

    
4021
  def ExpandNames(self):
4022
    self._ExpandAndLockInstance()
4023

    
4024
  def BuildHooksEnv(self):
4025
    """Build hooks env.
4026

4027
    This runs on master, primary and secondary nodes of the instance.
4028

4029
    """
4030
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4031
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4032
    return env, nl, nl
4033

    
4034
  def CheckPrereq(self):
4035
    """Check prerequisites.
4036

4037
    This checks that the instance is in the cluster and is not running.
4038

4039
    """
4040
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4041
    assert instance is not None, \
4042
      "Cannot retrieve locked instance %s" % self.op.instance_name
4043
    _CheckNodeOnline(self, instance.primary_node)
4044

    
4045
    if instance.disk_template == constants.DT_DISKLESS:
4046
      raise errors.OpPrereqError("Instance '%s' has no disks" %
4047
                                 self.op.instance_name, errors.ECODE_INVAL)
4048
    if instance.admin_up:
4049
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
4050
                                 self.op.instance_name, errors.ECODE_STATE)
4051
    remote_info = self.rpc.call_instance_info(instance.primary_node,
4052
                                              instance.name,
4053
                                              instance.hypervisor)
4054
    remote_info.Raise("Error checking node %s" % instance.primary_node,
4055
                      prereq=True, ecode=errors.ECODE_ENVIRON)
4056
    if remote_info.payload:
4057
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
4058
                                 (self.op.instance_name,
4059
                                  instance.primary_node), errors.ECODE_STATE)
4060

    
4061
    if not self.op.disks:
4062
      self.op.disks = range(len(instance.disks))
4063
    else:
4064
      for idx in self.op.disks:
4065
        if idx >= len(instance.disks):
4066
          raise errors.OpPrereqError("Invalid disk index passed '%s'" % idx,
4067
                                     errors.ECODE_INVAL)
4068

    
4069
    self.instance = instance
4070

    
4071
  def Exec(self, feedback_fn):
4072
    """Recreate the disks.
4073

4074
    """
4075
    to_skip = []
4076
    for idx, _ in enumerate(self.instance.disks):
4077
      if idx not in self.op.disks: # disk idx has not been passed in
4078
        to_skip.append(idx)
4079
        continue
4080

    
4081
    _CreateDisks(self, self.instance, to_skip=to_skip)
4082

    
4083

    
4084
class LURenameInstance(LogicalUnit):
4085
  """Rename an instance.
4086

4087
  """
4088
  HPATH = "instance-rename"
4089
  HTYPE = constants.HTYPE_INSTANCE
4090
  _OP_REQP = ["instance_name", "new_name"]
4091

    
4092
  def BuildHooksEnv(self):
4093
    """Build hooks env.
4094

4095
    This runs on master, primary and secondary nodes of the instance.
4096

4097
    """
4098
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4099
    env["INSTANCE_NEW_NAME"] = self.op.new_name
4100
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
4101
    return env, nl, nl
4102

    
4103
  def CheckPrereq(self):
4104
    """Check prerequisites.
4105

4106
    This checks that the instance is in the cluster and is not running.
4107

4108
    """
4109
    instance = self.cfg.GetInstanceInfo(
4110
      self.cfg.ExpandInstanceName(self.op.instance_name))
4111
    if instance is None:
4112
      raise errors.OpPrereqError("Instance '%s' not known" %
4113
                                 self.op.instance_name, errors.ECODE_NOENT)
4114
    _CheckNodeOnline(self, instance.primary_node)
4115

    
4116
    if instance.admin_up:
4117
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
4118
                                 self.op.instance_name, errors.ECODE_STATE)
4119
    remote_info = self.rpc.call_instance_info(instance.primary_node,
4120
                                              instance.name,
4121
                                              instance.hypervisor)
4122
    remote_info.Raise("Error checking node %s" % instance.primary_node,
4123
                      prereq=True, ecode=errors.ECODE_ENVIRON)
4124
    if remote_info.payload:
4125
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
4126
                                 (self.op.instance_name,
4127
                                  instance.primary_node), errors.ECODE_STATE)
4128
    self.instance = instance
4129

    
4130
    # new name verification
4131
    name_info = utils.GetHostInfo(self.op.new_name)
4132

    
4133
    self.op.new_name = new_name = name_info.name
4134
    instance_list = self.cfg.GetInstanceList()
4135
    if new_name in instance_list:
4136
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
4137
                                 new_name, errors.ECODE_EXISTS)
4138

    
4139
    if not getattr(self.op, "ignore_ip", False):
4140
      if utils.TcpPing(name_info.ip, constants.DEFAULT_NODED_PORT):
4141
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
4142
                                   (name_info.ip, new_name),
4143
                                   errors.ECODE_NOTUNIQUE)
4144

    
4145

    
4146
  def Exec(self, feedback_fn):
4147
    """Reinstall the instance.
4148

4149
    """
4150
    inst = self.instance
4151
    old_name = inst.name
4152

    
4153
    if inst.disk_template == constants.DT_FILE:
4154
      old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
4155

    
4156
    self.cfg.RenameInstance(inst.name, self.op.new_name)
4157
    # Change the instance lock. This is definitely safe while we hold the BGL
4158
    self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
4159
    self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
4160

    
4161
    # re-read the instance from the configuration after rename
4162
    inst = self.cfg.GetInstanceInfo(self.op.new_name)
4163

    
4164
    if inst.disk_template == constants.DT_FILE:
4165
      new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
4166
      result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
4167
                                                     old_file_storage_dir,
4168
                                                     new_file_storage_dir)
4169
      result.Raise("Could not rename on node %s directory '%s' to '%s'"
4170
                   " (but the instance has been renamed in Ganeti)" %
4171
                   (inst.primary_node, old_file_storage_dir,
4172
                    new_file_storage_dir))
4173

    
4174
    _StartInstanceDisks(self, inst, None)
4175
    try:
4176
      result = self.rpc.call_instance_run_rename(inst.primary_node, inst,
4177
                                                 old_name)
4178
      msg = result.fail_msg
4179
      if msg:
4180
        msg = ("Could not run OS rename script for instance %s on node %s"
4181
               " (but the instance has been renamed in Ganeti): %s" %
4182
               (inst.name, inst.primary_node, msg))
4183
        self.proc.LogWarning(msg)
4184
    finally:
4185
      _ShutdownInstanceDisks(self, inst)
4186

    
4187

    
4188
class LURemoveInstance(LogicalUnit):
4189
  """Remove an instance.
4190

4191
  """
4192
  HPATH = "instance-remove"
4193
  HTYPE = constants.HTYPE_INSTANCE
4194
  _OP_REQP = ["instance_name", "ignore_failures"]
4195
  REQ_BGL = False
4196

    
4197
  def CheckArguments(self):
4198
    """Check the arguments.
4199

4200
    """
4201
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
4202
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
4203

    
4204
  def ExpandNames(self):
4205
    self._ExpandAndLockInstance()
4206
    self.needed_locks[locking.LEVEL_NODE] = []
4207
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4208

    
4209
  def DeclareLocks(self, level):
4210
    if level == locking.LEVEL_NODE:
4211
      self._LockInstancesNodes()
4212

    
4213
  def BuildHooksEnv(self):
4214
    """Build hooks env.
4215

4216
    This runs on master, primary and secondary nodes of the instance.
4217

4218
    """
4219
    env = _BuildInstanceHookEnvByObject(self, self.instance)
4220
    env["SHUTDOWN_TIMEOUT"] = self.shutdown_timeout
4221
    nl = [self.cfg.GetMasterNode()]
4222
    return env, nl, nl
4223

    
4224
  def CheckPrereq(self):
4225
    """Check prerequisites.
4226

4227
    This checks that the instance is in the cluster.
4228

4229
    """
4230
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4231
    assert self.instance is not None, \
4232
      "Cannot retrieve locked instance %s" % self.op.instance_name
4233

    
4234
  def Exec(self, feedback_fn):
4235
    """Remove the instance.
4236

4237
    """
4238
    instance = self.instance
4239
    logging.info("Shutting down instance %s on node %s",
4240
                 instance.name, instance.primary_node)
4241

    
4242
    result = self.rpc.call_instance_shutdown(instance.primary_node, instance,
4243
                                             self.shutdown_timeout)
4244
    msg = result.fail_msg
4245
    if msg:
4246
      if self.op.ignore_failures:
4247
        feedback_fn("Warning: can't shutdown instance: %s" % msg)
4248
      else:
4249
        raise errors.OpExecError("Could not shutdown instance %s on"
4250
                                 " node %s: %s" %
4251
                                 (instance.name, instance.primary_node, msg))
4252

    
4253
    logging.info("Removing block devices for instance %s", instance.name)
4254

    
4255
    if not _RemoveDisks(self, instance):
4256
      if self.op.ignore_failures:
4257
        feedback_fn("Warning: can't remove instance's disks")
4258
      else:
4259
        raise errors.OpExecError("Can't remove instance's disks")
4260

    
4261
    logging.info("Removing instance %s out of cluster config", instance.name)
4262

    
4263
    self.cfg.RemoveInstance(instance.name)
4264
    self.remove_locks[locking.LEVEL_INSTANCE] = instance.name
4265

    
4266

    
4267
class LUQueryInstances(NoHooksLU):
4268
  """Logical unit for querying instances.
4269

4270
  """
4271
  # pylint: disable-msg=W0142
4272
  _OP_REQP = ["output_fields", "names", "use_locking"]
4273
  REQ_BGL = False
4274
  _SIMPLE_FIELDS = ["name", "os", "network_port", "hypervisor",
4275
                    "serial_no", "ctime", "mtime", "uuid"]
4276
  _FIELDS_STATIC = utils.FieldSet(*["name", "os", "pnode", "snodes",
4277
                                    "admin_state",
4278
                                    "disk_template", "ip", "mac", "bridge",
4279
                                    "nic_mode", "nic_link",
4280
                                    "sda_size", "sdb_size", "vcpus", "tags",
4281
                                    "network_port", "beparams",
4282
                                    r"(disk)\.(size)/([0-9]+)",
4283
                                    r"(disk)\.(sizes)", "disk_usage",
4284
                                    r"(nic)\.(mac|ip|mode|link)/([0-9]+)",
4285
                                    r"(nic)\.(bridge)/([0-9]+)",
4286
                                    r"(nic)\.(macs|ips|modes|links|bridges)",
4287
                                    r"(disk|nic)\.(count)",
4288
                                    "hvparams",
4289
                                    ] + _SIMPLE_FIELDS +
4290
                                  ["hv/%s" % name
4291
                                   for name in constants.HVS_PARAMETERS
4292
                                   if name not in constants.HVC_GLOBALS] +
4293
                                  ["be/%s" % name
4294
                                   for name in constants.BES_PARAMETERS])
4295
  _FIELDS_DYNAMIC = utils.FieldSet("oper_state", "oper_ram", "status")
4296

    
4297

    
4298
  def ExpandNames(self):
4299
    _CheckOutputFields(static=self._FIELDS_STATIC,
4300
                       dynamic=self._FIELDS_DYNAMIC,
4301
                       selected=self.op.output_fields)
4302

    
4303
    self.needed_locks = {}
4304
    self.share_locks[locking.LEVEL_INSTANCE] = 1
4305
    self.share_locks[locking.LEVEL_NODE] = 1
4306

    
4307
    if self.op.names:
4308
      self.wanted = _GetWantedInstances(self, self.op.names)
4309
    else:
4310
      self.wanted = locking.ALL_SET
4311

    
4312
    self.do_node_query = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
4313
    self.do_locking = self.do_node_query and self.op.use_locking
4314
    if self.do_locking:
4315
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
4316
      self.needed_locks[locking.LEVEL_NODE] = []
4317
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4318

    
4319
  def DeclareLocks(self, level):
4320
    if level == locking.LEVEL_NODE and self.do_locking:
4321
      self._LockInstancesNodes()
4322

    
4323
  def CheckPrereq(self):
4324
    """Check prerequisites.
4325

4326
    """
4327
    pass
4328

    
4329
  def Exec(self, feedback_fn):
4330
    """Computes the list of nodes and their attributes.
4331

4332
    """
4333
    # pylint: disable-msg=R0912
4334
    # way too many branches here
4335
    all_info = self.cfg.GetAllInstancesInfo()
4336
    if self.wanted == locking.ALL_SET:
4337
      # caller didn't specify instance names, so ordering is not important
4338
      if self.do_locking:
4339
        instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
4340
      else:
4341
        instance_names = all_info.keys()
4342
      instance_names = utils.NiceSort(instance_names)
4343
    else:
4344
      # caller did specify names, so we must keep the ordering
4345
      if self.do_locking:
4346
        tgt_set = self.acquired_locks[locking.LEVEL_INSTANCE]
4347
      else:
4348
        tgt_set = all_info.keys()
4349
      missing = set(self.wanted).difference(tgt_set)
4350
      if missing:
4351
        raise errors.OpExecError("Some instances were removed before"
4352
                                 " retrieving their data: %s" % missing)
4353
      instance_names = self.wanted
4354

    
4355
    instance_list = [all_info[iname] for iname in instance_names]
4356

    
4357
    # begin data gathering
4358

    
4359
    nodes = frozenset([inst.primary_node for inst in instance_list])
4360
    hv_list = list(set([inst.hypervisor for inst in instance_list]))
4361

    
4362
    bad_nodes = []
4363
    off_nodes = []
4364
    if self.do_node_query:
4365
      live_data = {}
4366
      node_data = self.rpc.call_all_instances_info(nodes, hv_list)
4367
      for name in nodes:
4368
        result = node_data[name]
4369
        if result.offline:
4370
          # offline nodes will be in both lists
4371
          off_nodes.append(name)
4372
        if result.fail_msg:
4373
          bad_nodes.append(name)
4374
        else:
4375
          if result.payload:
4376
            live_data.update(result.payload)
4377
          # else no instance is alive
4378
    else:
4379
      live_data = dict([(name, {}) for name in instance_names])
4380

    
4381
    # end data gathering
4382

    
4383
    HVPREFIX = "hv/"
4384
    BEPREFIX = "be/"
4385
    output = []
4386
    cluster = self.cfg.GetClusterInfo()
4387
    for instance in instance_list:
4388
      iout = []
4389
      i_hv = cluster.FillHV(instance, skip_globals=True)
4390
      i_be = cluster.FillBE(instance)
4391
      i_nicp = [objects.FillDict(cluster.nicparams[constants.PP_DEFAULT],
4392
                                 nic.nicparams) for nic in instance.nics]
4393
      for field in self.op.output_fields:
4394
        st_match = self._FIELDS_STATIC.Matches(field)
4395
        if field in self._SIMPLE_FIELDS:
4396
          val = getattr(instance, field)
4397
        elif field == "pnode":
4398
          val = instance.primary_node
4399
        elif field == "snodes":
4400
          val = list(instance.secondary_nodes)
4401
        elif field == "admin_state":
4402
          val = instance.admin_up
4403
        elif field == "oper_state":
4404
          if instance.primary_node in bad_nodes:
4405
            val = None
4406
          else:
4407
            val = bool(live_data.get(instance.name))
4408
        elif field == "status":
4409
          if instance.primary_node in off_nodes:
4410
            val = "ERROR_nodeoffline"
4411
          elif instance.primary_node in bad_nodes:
4412
            val = "ERROR_nodedown"
4413
          else:
4414
            running = bool(live_data.get(instance.name))
4415
            if running:
4416
              if instance.admin_up:
4417
                val = "running"
4418
              else:
4419
                val = "ERROR_up"
4420
            else:
4421
              if instance.admin_up:
4422
                val = "ERROR_down"
4423
              else:
4424
                val = "ADMIN_down"
4425
        elif field == "oper_ram":
4426
          if instance.primary_node in bad_nodes:
4427
            val = None
4428
          elif instance.name in live_data:
4429
            val = live_data[instance.name].get("memory", "?")
4430
          else:
4431
            val = "-"
4432
        elif field == "vcpus":
4433
          val = i_be[constants.BE_VCPUS]
4434
        elif field == "disk_template":
4435
          val = instance.disk_template
4436
        elif field == "ip":
4437
          if instance.nics:
4438
            val = instance.nics[0].ip
4439
          else:
4440
            val = None
4441
        elif field == "nic_mode":
4442
          if instance.nics:
4443
            val = i_nicp[0][constants.NIC_MODE]
4444
          else:
4445
            val = None
4446
        elif field == "nic_link":
4447
          if instance.nics:
4448
            val = i_nicp[0][constants.NIC_LINK]
4449
          else:
4450
            val = None
4451
        elif field == "bridge":
4452
          if (instance.nics and
4453
              i_nicp[0][constants.NIC_MODE] == constants.NIC_MODE_BRIDGED):
4454
            val = i_nicp[0][constants.NIC_LINK]
4455
          else:
4456
            val = None
4457
        elif field == "mac":
4458
          if instance.nics:
4459
            val = instance.nics[0].mac
4460
          else:
4461
            val = None
4462
        elif field == "sda_size" or field == "sdb_size":
4463
          idx = ord(field[2]) - ord('a')
4464
          try:
4465
            val = instance.FindDisk(idx).size
4466
          except errors.OpPrereqError:
4467
            val = None
4468
        elif field == "disk_usage": # total disk usage per node
4469
          disk_sizes = [{'size': disk.size} for disk in instance.disks]
4470
          val = _ComputeDiskSize(instance.disk_template, disk_sizes)
4471
        elif field == "tags":
4472
          val = list(instance.GetTags())
4473
        elif field == "hvparams":
4474
          val = i_hv
4475
        elif (field.startswith(HVPREFIX) and
4476
              field[len(HVPREFIX):] in constants.HVS_PARAMETERS and
4477
              field[len(HVPREFIX):] not in constants.HVC_GLOBALS):
4478
          val = i_hv.get(field[len(HVPREFIX):], None)
4479
        elif field == "beparams":
4480
          val = i_be
4481
        elif (field.startswith(BEPREFIX) and
4482
              field[len(BEPREFIX):] in constants.BES_PARAMETERS):
4483
          val = i_be.get(field[len(BEPREFIX):], None)
4484
        elif st_match and st_match.groups():
4485
          # matches a variable list
4486
          st_groups = st_match.groups()
4487
          if st_groups and st_groups[0] == "disk":
4488
            if st_groups[1] == "count":
4489
              val = len(instance.disks)
4490
            elif st_groups[1] == "sizes":
4491
              val = [disk.size for disk in instance.disks]
4492
            elif st_groups[1] == "size":
4493
              try:
4494
                val = instance.FindDisk(st_groups[2]).size
4495
              except errors.OpPrereqError:
4496
                val = None
4497
            else:
4498
              assert False, "Unhandled disk parameter"
4499
          elif st_groups[0] == "nic":
4500
            if st_groups[1] == "count":
4501
              val = len(instance.nics)
4502
            elif st_groups[1] == "macs":
4503
              val = [nic.mac for nic in instance.nics]
4504
            elif st_groups[1] == "ips":
4505
              val = [nic.ip for nic in instance.nics]
4506
            elif st_groups[1] == "modes":
4507
              val = [nicp[constants.NIC_MODE] for nicp in i_nicp]
4508
            elif st_groups[1] == "links":
4509
              val = [nicp[constants.NIC_LINK] for nicp in i_nicp]
4510
            elif st_groups[1] == "bridges":
4511
              val = []
4512
              for nicp in i_nicp:
4513
                if nicp[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
4514
                  val.append(nicp[constants.NIC_LINK])
4515
                else:
4516
                  val.append(None)
4517
            else:
4518
              # index-based item
4519
              nic_idx = int(st_groups[2])
4520
              if nic_idx >= len(instance.nics):
4521
                val = None
4522
              else:
4523
                if st_groups[1] == "mac":
4524
                  val = instance.nics[nic_idx].mac
4525
                elif st_groups[1] == "ip":
4526
                  val = instance.nics[nic_idx].ip
4527
                elif st_groups[1] == "mode":
4528
                  val = i_nicp[nic_idx][constants.NIC_MODE]
4529
                elif st_groups[1] == "link":
4530
                  val = i_nicp[nic_idx][constants.NIC_LINK]
4531
                elif st_groups[1] == "bridge":
4532
                  nic_mode = i_nicp[nic_idx][constants.NIC_MODE]
4533
                  if nic_mode == constants.NIC_MODE_BRIDGED:
4534
                    val = i_nicp[nic_idx][constants.NIC_LINK]
4535
                  else:
4536
                    val = None
4537
                else:
4538
                  assert False, "Unhandled NIC parameter"
4539
          else:
4540
            assert False, ("Declared but unhandled variable parameter '%s'" %
4541
                           field)
4542
        else:
4543
          assert False, "Declared but unhandled parameter '%s'" % field
4544
        iout.append(val)
4545
      output.append(iout)
4546

    
4547
    return output
4548

    
4549

    
4550
class LUFailoverInstance(LogicalUnit):
4551
  """Failover an instance.
4552

4553
  """
4554
  HPATH = "instance-failover"
4555
  HTYPE = constants.HTYPE_INSTANCE
4556
  _OP_REQP = ["instance_name", "ignore_consistency"]
4557
  REQ_BGL = False
4558

    
4559
  def CheckArguments(self):
4560
    """Check the arguments.
4561

4562
    """
4563
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
4564
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
4565

    
4566
  def ExpandNames(self):
4567
    self._ExpandAndLockInstance()
4568
    self.needed_locks[locking.LEVEL_NODE] = []
4569
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4570

    
4571
  def DeclareLocks(self, level):
4572
    if level == locking.LEVEL_NODE:
4573
      self._LockInstancesNodes()
4574

    
4575
  def BuildHooksEnv(self):
4576
    """Build hooks env.
4577

4578
    This runs on master, primary and secondary nodes of the instance.
4579

4580
    """
4581
    env = {
4582
      "IGNORE_CONSISTENCY": self.op.ignore_consistency,
4583
      "SHUTDOWN_TIMEOUT": self.shutdown_timeout,
4584
      }
4585
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4586
    nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
4587
    return env, nl, nl
4588

    
4589
  def CheckPrereq(self):
4590
    """Check prerequisites.
4591

4592
    This checks that the instance is in the cluster.
4593

4594
    """
4595
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4596
    assert self.instance is not None, \
4597
      "Cannot retrieve locked instance %s" % self.op.instance_name
4598

    
4599
    bep = self.cfg.GetClusterInfo().FillBE(instance)
4600
    if instance.disk_template not in constants.DTS_NET_MIRROR:
4601
      raise errors.OpPrereqError("Instance's disk layout is not"
4602
                                 " network mirrored, cannot failover.",
4603
                                 errors.ECODE_STATE)
4604

    
4605
    secondary_nodes = instance.secondary_nodes
4606
    if not secondary_nodes:
4607
      raise errors.ProgrammerError("no secondary node but using "
4608
                                   "a mirrored disk template")
4609

    
4610
    target_node = secondary_nodes[0]
4611
    _CheckNodeOnline(self, target_node)
4612
    _CheckNodeNotDrained(self, target_node)
4613
    if instance.admin_up:
4614
      # check memory requirements on the secondary node
4615
      _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
4616
                           instance.name, bep[constants.BE_MEMORY],
4617
                           instance.hypervisor)
4618
    else:
4619
      self.LogInfo("Not checking memory on the secondary node as"
4620
                   " instance will not be started")
4621

    
4622
    # check bridge existance
4623
    _CheckInstanceBridgesExist(self, instance, node=target_node)
4624

    
4625
  def Exec(self, feedback_fn):
4626
    """Failover an instance.
4627

4628
    The failover is done by shutting it down on its present node and
4629
    starting it on the secondary.
4630

4631
    """
4632
    instance = self.instance
4633

    
4634
    source_node = instance.primary_node
4635
    target_node = instance.secondary_nodes[0]
4636

    
4637
    if instance.admin_up:
4638
      feedback_fn("* checking disk consistency between source and target")
4639
      for dev in instance.disks:
4640
        # for drbd, these are drbd over lvm
4641
        if not _CheckDiskConsistency(self, dev, target_node, False):
4642
          if not self.op.ignore_consistency:
4643
            raise errors.OpExecError("Disk %s is degraded on target node,"
4644
                                     " aborting failover." % dev.iv_name)
4645
    else:
4646
      feedback_fn("* not checking disk consistency as instance is not running")
4647

    
4648
    feedback_fn("* shutting down instance on source node")
4649
    logging.info("Shutting down instance %s on node %s",
4650
                 instance.name, source_node)
4651

    
4652
    result = self.rpc.call_instance_shutdown(source_node, instance,
4653
                                             self.shutdown_timeout)
4654
    msg = result.fail_msg
4655
    if msg:
4656
      if self.op.ignore_consistency:
4657
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
4658
                             " Proceeding anyway. Please make sure node"
4659
                             " %s is down. Error details: %s",
4660
                             instance.name, source_node, source_node, msg)
4661
      else:
4662
        raise errors.OpExecError("Could not shutdown instance %s on"
4663
                                 " node %s: %s" %
4664
                                 (instance.name, source_node, msg))
4665

    
4666
    feedback_fn("* deactivating the instance's disks on source node")
4667
    if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
4668
      raise errors.OpExecError("Can't shut down the instance's disks.")
4669

    
4670
    instance.primary_node = target_node
4671
    # distribute new instance config to the other nodes
4672
    self.cfg.Update(instance, feedback_fn)
4673

    
4674
    # Only start the instance if it's marked as up
4675
    if instance.admin_up:
4676
      feedback_fn("* activating the instance's disks on target node")
4677
      logging.info("Starting instance %s on node %s",
4678
                   instance.name, target_node)
4679

    
4680
      disks_ok, _ = _AssembleInstanceDisks(self, instance,
4681
                                               ignore_secondaries=True)
4682
      if not disks_ok:
4683
        _ShutdownInstanceDisks(self, instance)
4684
        raise errors.OpExecError("Can't activate the instance's disks")
4685

    
4686
      feedback_fn("* starting the instance on the target node")
4687
      result = self.rpc.call_instance_start(target_node, instance, None, None)
4688
      msg = result.fail_msg
4689
      if msg:
4690
        _ShutdownInstanceDisks(self, instance)
4691
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
4692
                                 (instance.name, target_node, msg))
4693

    
4694

    
4695
class LUMigrateInstance(LogicalUnit):
4696
  """Migrate an instance.
4697

4698
  This is migration without shutting down, compared to the failover,
4699
  which is done with shutdown.
4700

4701
  """
4702
  HPATH = "instance-migrate"
4703
  HTYPE = constants.HTYPE_INSTANCE
4704
  _OP_REQP = ["instance_name", "live", "cleanup"]
4705

    
4706
  REQ_BGL = False
4707

    
4708
  def ExpandNames(self):
4709
    self._ExpandAndLockInstance()
4710

    
4711
    self.needed_locks[locking.LEVEL_NODE] = []
4712
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4713

    
4714
    self._migrater = TLMigrateInstance(self, self.op.instance_name,
4715
                                       self.op.live, self.op.cleanup)
4716
    self.tasklets = [self._migrater]
4717

    
4718
  def DeclareLocks(self, level):
4719
    if level == locking.LEVEL_NODE:
4720
      self._LockInstancesNodes()
4721

    
4722
  def BuildHooksEnv(self):
4723
    """Build hooks env.
4724

4725
    This runs on master, primary and secondary nodes of the instance.
4726

4727
    """
4728
    instance = self._migrater.instance
4729
    env = _BuildInstanceHookEnvByObject(self, instance)
4730
    env["MIGRATE_LIVE"] = self.op.live
4731
    env["MIGRATE_CLEANUP"] = self.op.cleanup
4732
    nl = [self.cfg.GetMasterNode()] + list(instance.secondary_nodes)
4733
    return env, nl, nl
4734

    
4735

    
4736
class LUMoveInstance(LogicalUnit):
4737
  """Move an instance by data-copying.
4738

4739
  """
4740
  HPATH = "instance-move"
4741
  HTYPE = constants.HTYPE_INSTANCE
4742
  _OP_REQP = ["instance_name", "target_node"]
4743
  REQ_BGL = False
4744

    
4745
  def CheckArguments(self):
4746
    """Check the arguments.
4747

4748
    """
4749
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
4750
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
4751

    
4752
  def ExpandNames(self):
4753
    self._ExpandAndLockInstance()
4754
    target_node = self.cfg.ExpandNodeName(self.op.target_node)
4755
    if target_node is None:
4756
      raise errors.OpPrereqError("Node '%s' not known" %
4757
                                  self.op.target_node, errors.ECODE_NOENT)
4758
    self.op.target_node = target_node
4759
    self.needed_locks[locking.LEVEL_NODE] = [target_node]
4760
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
4761

    
4762
  def DeclareLocks(self, level):
4763
    if level == locking.LEVEL_NODE:
4764
      self._LockInstancesNodes(primary_only=True)
4765

    
4766
  def BuildHooksEnv(self):
4767
    """Build hooks env.
4768

4769
    This runs on master, primary and secondary nodes of the instance.
4770

4771
    """
4772
    env = {
4773
      "TARGET_NODE": self.op.target_node,
4774
      "SHUTDOWN_TIMEOUT": self.shutdown_timeout,
4775
      }
4776
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4777
    nl = [self.cfg.GetMasterNode()] + [self.instance.primary_node,
4778
                                       self.op.target_node]
4779
    return env, nl, nl
4780

    
4781
  def CheckPrereq(self):
4782
    """Check prerequisites.
4783

4784
    This checks that the instance is in the cluster.
4785

4786
    """
4787
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4788
    assert self.instance is not None, \
4789
      "Cannot retrieve locked instance %s" % self.op.instance_name
4790

    
4791
    node = self.cfg.GetNodeInfo(self.op.target_node)
4792
    assert node is not None, \
4793
      "Cannot retrieve locked node %s" % self.op.target_node
4794

    
4795
    self.target_node = target_node = node.name
4796

    
4797
    if target_node == instance.primary_node:
4798
      raise errors.OpPrereqError("Instance %s is already on the node %s" %
4799
                                 (instance.name, target_node),
4800
                                 errors.ECODE_STATE)
4801

    
4802
    bep = self.cfg.GetClusterInfo().FillBE(instance)
4803

    
4804
    for idx, dsk in enumerate(instance.disks):
4805
      if dsk.dev_type not in (constants.LD_LV, constants.LD_FILE):
4806
        raise errors.OpPrereqError("Instance disk %d has a complex layout,"
4807
                                   " cannot copy" % idx, errors.ECODE_STATE)
4808

    
4809
    _CheckNodeOnline(self, target_node)
4810
    _CheckNodeNotDrained(self, target_node)
4811

    
4812
    if instance.admin_up:
4813
      # check memory requirements on the secondary node
4814
      _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
4815
                           instance.name, bep[constants.BE_MEMORY],
4816
                           instance.hypervisor)
4817
    else:
4818
      self.LogInfo("Not checking memory on the secondary node as"
4819
                   " instance will not be started")
4820

    
4821
    # check bridge existance
4822
    _CheckInstanceBridgesExist(self, instance, node=target_node)
4823

    
4824
  def Exec(self, feedback_fn):
4825
    """Move an instance.
4826

4827
    The move is done by shutting it down on its present node, copying
4828
    the data over (slow) and starting it on the new node.
4829

4830
    """
4831
    instance = self.instance
4832

    
4833
    source_node = instance.primary_node
4834
    target_node = self.target_node
4835

    
4836
    self.LogInfo("Shutting down instance %s on source node %s",
4837
                 instance.name, source_node)
4838

    
4839
    result = self.rpc.call_instance_shutdown(source_node, instance,
4840
                                             self.shutdown_timeout)
4841
    msg = result.fail_msg
4842
    if msg:
4843
      if self.op.ignore_consistency:
4844
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
4845
                             " Proceeding anyway. Please make sure node"
4846
                             " %s is down. Error details: %s",
4847
                             instance.name, source_node, source_node, msg)
4848
      else:
4849
        raise errors.OpExecError("Could not shutdown instance %s on"
4850
                                 " node %s: %s" %
4851
                                 (instance.name, source_node, msg))
4852

    
4853
    # create the target disks
4854
    try:
4855
      _CreateDisks(self, instance, target_node=target_node)
4856
    except errors.OpExecError:
4857
      self.LogWarning("Device creation failed, reverting...")
4858
      try:
4859
        _RemoveDisks(self, instance, target_node=target_node)
4860
      finally:
4861
        self.cfg.ReleaseDRBDMinors(instance.name)
4862
        raise
4863

    
4864
    cluster_name = self.cfg.GetClusterInfo().cluster_name
4865

    
4866
    errs = []
4867
    # activate, get path, copy the data over
4868
    for idx, disk in enumerate(instance.disks):
4869
      self.LogInfo("Copying data for disk %d", idx)
4870
      result = self.rpc.call_blockdev_assemble(target_node, disk,
4871
                                               instance.name, True)
4872
      if result.fail_msg:
4873
        self.LogWarning("Can't assemble newly created disk %d: %s",
4874
                        idx, result.fail_msg)
4875
        errs.append(result.fail_msg)
4876
        break
4877
      dev_path = result.payload
4878
      result = self.rpc.call_blockdev_export(source_node, disk,
4879
                                             target_node, dev_path,
4880
                                             cluster_name)
4881
      if result.fail_msg:
4882
        self.LogWarning("Can't copy data over for disk %d: %s",
4883
                        idx, result.fail_msg)
4884
        errs.append(result.fail_msg)
4885
        break
4886

    
4887
    if errs:
4888
      self.LogWarning("Some disks failed to copy, aborting")
4889
      try:
4890
        _RemoveDisks(self, instance, target_node=target_node)
4891
      finally:
4892
        self.cfg.ReleaseDRBDMinors(instance.name)
4893
        raise errors.OpExecError("Errors during disk copy: %s" %
4894
                                 (",".join(errs),))
4895

    
4896
    instance.primary_node = target_node
4897
    self.cfg.Update(instance, feedback_fn)
4898

    
4899
    self.LogInfo("Removing the disks on the original node")
4900
    _RemoveDisks(self, instance, target_node=source_node)
4901

    
4902
    # Only start the instance if it's marked as up
4903
    if instance.admin_up:
4904
      self.LogInfo("Starting instance %s on node %s",
4905
                   instance.name, target_node)
4906

    
4907
      disks_ok, _ = _AssembleInstanceDisks(self, instance,
4908
                                           ignore_secondaries=True)
4909
      if not disks_ok:
4910
        _ShutdownInstanceDisks(self, instance)
4911
        raise errors.OpExecError("Can't activate the instance's disks")
4912

    
4913
      result = self.rpc.call_instance_start(target_node, instance, None, None)
4914
      msg = result.fail_msg
4915
      if msg:
4916
        _ShutdownInstanceDisks(self, instance)
4917
        raise errors.OpExecError("Could not start instance %s on node %s: %s" %
4918
                                 (instance.name, target_node, msg))
4919

    
4920

    
4921
class LUMigrateNode(LogicalUnit):
4922
  """Migrate all instances from a node.
4923

4924
  """
4925
  HPATH = "node-migrate"
4926
  HTYPE = constants.HTYPE_NODE
4927
  _OP_REQP = ["node_name", "live"]
4928
  REQ_BGL = False
4929

    
4930
  def ExpandNames(self):
4931
    self.op.node_name = self.cfg.ExpandNodeName(self.op.node_name)
4932
    if self.op.node_name is None:
4933
      raise errors.OpPrereqError("Node '%s' not known" % self.op.node_name,
4934
                                 errors.ECODE_NOENT)
4935

    
4936
    self.needed_locks = {
4937
      locking.LEVEL_NODE: [self.op.node_name],
4938
      }
4939

    
4940
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
4941

    
4942
    # Create tasklets for migrating instances for all instances on this node
4943
    names = []
4944
    tasklets = []
4945

    
4946
    for inst in _GetNodePrimaryInstances(self.cfg, self.op.node_name):
4947
      logging.debug("Migrating instance %s", inst.name)
4948
      names.append(inst.name)
4949

    
4950
      tasklets.append(TLMigrateInstance(self, inst.name, self.op.live, False))
4951

    
4952
    self.tasklets = tasklets
4953

    
4954
    # Declare instance locks
4955
    self.needed_locks[locking.LEVEL_INSTANCE] = names
4956

    
4957
  def DeclareLocks(self, level):
4958
    if level == locking.LEVEL_NODE:
4959
      self._LockInstancesNodes()
4960

    
4961
  def BuildHooksEnv(self):
4962
    """Build hooks env.
4963

4964
    This runs on the master, the primary and all the secondaries.
4965

4966
    """
4967
    env = {
4968
      "NODE_NAME": self.op.node_name,
4969
      }
4970

    
4971
    nl = [self.cfg.GetMasterNode()]
4972

    
4973
    return (env, nl, nl)
4974

    
4975

    
4976
class TLMigrateInstance(Tasklet):
4977
  def __init__(self, lu, instance_name, live, cleanup):
4978
    """Initializes this class.
4979

4980
    """
4981
    Tasklet.__init__(self, lu)
4982

    
4983
    # Parameters
4984
    self.instance_name = instance_name
4985
    self.live = live
4986
    self.cleanup = cleanup
4987

    
4988
  def CheckPrereq(self):
4989
    """Check prerequisites.
4990

4991
    This checks that the instance is in the cluster.
4992

4993
    """
4994
    instance = self.cfg.GetInstanceInfo(
4995
      self.cfg.ExpandInstanceName(self.instance_name))
4996
    if instance is None:
4997
      raise errors.OpPrereqError("Instance '%s' not known" %
4998
                                 self.instance_name, errors.ECODE_NOENT)
4999

    
5000
    if instance.disk_template != constants.DT_DRBD8:
5001
      raise errors.OpPrereqError("Instance's disk layout is not"
5002
                                 " drbd8, cannot migrate.", errors.ECODE_STATE)
5003

    
5004
    secondary_nodes = instance.secondary_nodes
5005
    if not secondary_nodes:
5006
      raise errors.ConfigurationError("No secondary node but using"
5007
                                      " drbd8 disk template")
5008

    
5009
    i_be = self.cfg.GetClusterInfo().FillBE(instance)
5010

    
5011
    target_node = secondary_nodes[0]
5012
    # check memory requirements on the secondary node
5013
    _CheckNodeFreeMemory(self, target_node, "migrating instance %s" %
5014
                         instance.name, i_be[constants.BE_MEMORY],
5015
                         instance.hypervisor)
5016

    
5017
    # check bridge existance
5018
    _CheckInstanceBridgesExist(self, instance, node=target_node)
5019

    
5020
    if not self.cleanup:
5021
      _CheckNodeNotDrained(self, target_node)
5022
      result = self.rpc.call_instance_migratable(instance.primary_node,
5023
                                                 instance)
5024
      result.Raise("Can't migrate, please use failover",
5025
                   prereq=True, ecode=errors.ECODE_STATE)
5026

    
5027
    self.instance = instance
5028

    
5029
  def _WaitUntilSync(self):
5030
    """Poll with custom rpc for disk sync.
5031

5032
    This uses our own step-based rpc call.
5033

5034
    """
5035
    self.feedback_fn("* wait until resync is done")
5036
    all_done = False
5037
    while not all_done:
5038
      all_done = True
5039
      result = self.rpc.call_drbd_wait_sync(self.all_nodes,
5040
                                            self.nodes_ip,
5041
                                            self.instance.disks)
5042
      min_percent = 100
5043
      for node, nres in result.items():
5044
        nres.Raise("Cannot resync disks on node %s" % node)
5045
        node_done, node_percent = nres.payload
5046
        all_done = all_done and node_done
5047
        if node_percent is not None:
5048
          min_percent = min(min_percent, node_percent)
5049
      if not all_done:
5050
        if min_percent < 100:
5051
          self.feedback_fn("   - progress: %.1f%%" % min_percent)
5052
        time.sleep(2)
5053

    
5054
  def _EnsureSecondary(self, node):
5055
    """Demote a node to secondary.
5056

5057
    """
5058
    self.feedback_fn("* switching node %s to secondary mode" % node)
5059

    
5060
    for dev in self.instance.disks:
5061
      self.cfg.SetDiskID(dev, node)
5062

    
5063
    result = self.rpc.call_blockdev_close(node, self.instance.name,
5064
                                          self.instance.disks)
5065
    result.Raise("Cannot change disk to secondary on node %s" % node)
5066

    
5067
  def _GoStandalone(self):
5068
    """Disconnect from the network.
5069

5070
    """
5071
    self.feedback_fn("* changing into standalone mode")
5072
    result = self.rpc.call_drbd_disconnect_net(self.all_nodes, self.nodes_ip,
5073
                                               self.instance.disks)
5074
    for node, nres in result.items():
5075
      nres.Raise("Cannot disconnect disks node %s" % node)
5076

    
5077
  def _GoReconnect(self, multimaster):
5078
    """Reconnect to the network.
5079

5080
    """
5081
    if multimaster:
5082
      msg = "dual-master"
5083
    else:
5084
      msg = "single-master"
5085
    self.feedback_fn("* changing disks into %s mode" % msg)
5086
    result = self.rpc.call_drbd_attach_net(self.all_nodes, self.nodes_ip,
5087
                                           self.instance.disks,
5088
                                           self.instance.name, multimaster)
5089
    for node, nres in result.items():
5090
      nres.Raise("Cannot change disks config on node %s" % node)
5091

    
5092
  def _ExecCleanup(self):
5093
    """Try to cleanup after a failed migration.
5094

5095
    The cleanup is done by:
5096
      - check that the instance is running only on one node
5097
        (and update the config if needed)
5098
      - change disks on its secondary node to secondary
5099
      - wait until disks are fully synchronized
5100
      - disconnect from the network
5101
      - change disks into single-master mode
5102
      - wait again until disks are fully synchronized
5103

5104
    """
5105
    instance = self.instance
5106
    target_node = self.target_node
5107
    source_node = self.source_node
5108

    
5109
    # check running on only one node
5110
    self.feedback_fn("* checking where the instance actually runs"
5111
                     " (if this hangs, the hypervisor might be in"
5112
                     " a bad state)")
5113
    ins_l = self.rpc.call_instance_list(self.all_nodes, [instance.hypervisor])
5114
    for node, result in ins_l.items():
5115
      result.Raise("Can't contact node %s" % node)
5116

    
5117
    runningon_source = instance.name in ins_l[source_node].payload
5118
    runningon_target = instance.name in ins_l[target_node].payload
5119

    
5120
    if runningon_source and runningon_target:
5121
      raise errors.OpExecError("Instance seems to be running on two nodes,"
5122
                               " or the hypervisor is confused. You will have"
5123
                               " to ensure manually that it runs only on one"
5124
                               " and restart this operation.")
5125

    
5126
    if not (runningon_source or runningon_target):
5127
      raise errors.OpExecError("Instance does not seem to be running at all."
5128
                               " In this case, it's safer to repair by"
5129
                               " running 'gnt-instance stop' to ensure disk"
5130
                               " shutdown, and then restarting it.")
5131

    
5132
    if runningon_target:
5133
      # the migration has actually succeeded, we need to update the config
5134
      self.feedback_fn("* instance running on secondary node (%s),"
5135
                       " updating config" % target_node)
5136
      instance.primary_node = target_node
5137
      self.cfg.Update(instance, self.feedback_fn)
5138
      demoted_node = source_node
5139
    else:
5140
      self.feedback_fn("* instance confirmed to be running on its"
5141
                       " primary node (%s)" % source_node)
5142
      demoted_node = target_node
5143

    
5144
    self._EnsureSecondary(demoted_node)
5145
    try:
5146
      self._WaitUntilSync()
5147
    except errors.OpExecError:
5148
      # we ignore here errors, since if the device is standalone, it
5149
      # won't be able to sync
5150
      pass
5151
    self._GoStandalone()
5152
    self._GoReconnect(False)
5153
    self._WaitUntilSync()
5154

    
5155
    self.feedback_fn("* done")
5156

    
5157
  def _RevertDiskStatus(self):
5158
    """Try to revert the disk status after a failed migration.
5159

5160
    """
5161
    target_node = self.target_node
5162
    try:
5163
      self._EnsureSecondary(target_node)
5164
      self._GoStandalone()
5165
      self._GoReconnect(False)
5166
      self._WaitUntilSync()
5167
    except errors.OpExecError, err:
5168
      self.lu.LogWarning("Migration failed and I can't reconnect the"
5169
                         " drives: error '%s'\n"
5170
                         "Please look and recover the instance status" %
5171
                         str(err))
5172

    
5173
  def _AbortMigration(self):
5174
    """Call the hypervisor code to abort a started migration.
5175

5176
    """
5177
    instance = self.instance
5178
    target_node = self.target_node
5179
    migration_info = self.migration_info
5180

    
5181
    abort_result = self.rpc.call_finalize_migration(target_node,
5182
                                                    instance,
5183
                                                    migration_info,
5184
                                                    False)
5185
    abort_msg = abort_result.fail_msg
5186
    if abort_msg:
5187
      logging.error("Aborting migration failed on target node %s: %s",
5188
                    target_node, abort_msg)
5189
      # Don't raise an exception here, as we stil have to try to revert the
5190
      # disk status, even if this step failed.
5191

    
5192
  def _ExecMigration(self):
5193
    """Migrate an instance.
5194

5195
    The migrate is done by:
5196
      - change the disks into dual-master mode
5197
      - wait until disks are fully synchronized again
5198
      - migrate the instance
5199
      - change disks on the new secondary node (the old primary) to secondary
5200
      - wait until disks are fully synchronized
5201
      - change disks into single-master mode
5202

5203
    """
5204
    instance = self.instance
5205
    target_node = self.target_node
5206
    source_node = self.source_node
5207

    
5208
    self.feedback_fn("* checking disk consistency between source and target")
5209
    for dev in instance.disks:
5210
      if not _CheckDiskConsistency(self, dev, target_node, False):
5211
        raise errors.OpExecError("Disk %s is degraded or not fully"
5212
                                 " synchronized on target node,"
5213
                                 " aborting migrate." % dev.iv_name)
5214

    
5215
    # First get the migration information from the remote node
5216
    result = self.rpc.call_migration_info(source_node, instance)
5217
    msg = result.fail_msg
5218
    if msg:
5219
      log_err = ("Failed fetching source migration information from %s: %s" %
5220
                 (source_node, msg))
5221
      logging.error(log_err)
5222
      raise errors.OpExecError(log_err)
5223

    
5224
    self.migration_info = migration_info = result.payload
5225

    
5226
    # Then switch the disks to master/master mode
5227
    self._EnsureSecondary(target_node)
5228
    self._GoStandalone()
5229
    self._GoReconnect(True)
5230
    self._WaitUntilSync()
5231

    
5232
    self.feedback_fn("* preparing %s to accept the instance" % target_node)
5233
    result = self.rpc.call_accept_instance(target_node,
5234
                                           instance,
5235
                                           migration_info,
5236
                                           self.nodes_ip[target_node])
5237

    
5238
    msg = result.fail_msg
5239
    if msg:
5240
      logging.error("Instance pre-migration failed, trying to revert"
5241
                    " disk status: %s", msg)
5242
      self.feedback_fn("Pre-migration failed, aborting")
5243
      self._AbortMigration()
5244
      self._RevertDiskStatus()
5245
      raise errors.OpExecError("Could not pre-migrate instance %s: %s" %
5246
                               (instance.name, msg))
5247

    
5248
    self.feedback_fn("* migrating instance to %s" % target_node)
5249
    time.sleep(10)
5250
    result = self.rpc.call_instance_migrate(source_node, instance,
5251
                                            self.nodes_ip[target_node],
5252
                                            self.live)
5253
    msg = result.fail_msg
5254
    if msg:
5255
      logging.error("Instance migration failed, trying to revert"
5256
                    " disk status: %s", msg)
5257
      self.feedback_fn("Migration failed, aborting")
5258
      self._AbortMigration()
5259
      self._RevertDiskStatus()
5260
      raise errors.OpExecError("Could not migrate instance %s: %s" %
5261
                               (instance.name, msg))
5262
    time.sleep(10)
5263

    
5264
    instance.primary_node = target_node
5265
    # distribute new instance config to the other nodes
5266
    self.cfg.Update(instance, self.feedback_fn)
5267

    
5268
    result = self.rpc.call_finalize_migration(target_node,
5269
                                              instance,
5270
                                              migration_info,
5271
                                              True)
5272
    msg = result.fail_msg
5273
    if msg:
5274
      logging.error("Instance migration succeeded, but finalization failed:"
5275
                    " %s", msg)
5276
      raise errors.OpExecError("Could not finalize instance migration: %s" %
5277
                               msg)
5278

    
5279
    self._EnsureSecondary(source_node)
5280
    self._WaitUntilSync()
5281
    self._GoStandalone()
5282
    self._GoReconnect(False)
5283
    self._WaitUntilSync()
5284

    
5285
    self.feedback_fn("* done")
5286

    
5287
  def Exec(self, feedback_fn):
5288
    """Perform the migration.
5289

5290
    """
5291
    feedback_fn("Migrating instance %s" % self.instance.name)
5292

    
5293
    self.feedback_fn = feedback_fn
5294

    
5295
    self.source_node = self.instance.primary_node
5296
    self.target_node = self.instance.secondary_nodes[0]
5297
    self.all_nodes = [self.source_node, self.target_node]
5298
    self.nodes_ip = {
5299
      self.source_node: self.cfg.GetNodeInfo(self.source_node).secondary_ip,
5300
      self.target_node: self.cfg.GetNodeInfo(self.target_node).secondary_ip,
5301
      }
5302

    
5303
    if self.cleanup:
5304
      return self._ExecCleanup()
5305
    else:
5306
      return self._ExecMigration()
5307

    
5308

    
5309
def _CreateBlockDev(lu, node, instance, device, force_create,
5310
                    info, force_open):
5311
  """Create a tree of block devices on a given node.
5312

5313
  If this device type has to be created on secondaries, create it and
5314
  all its children.
5315

5316
  If not, just recurse to children keeping the same 'force' value.
5317

5318
  @param lu: the lu on whose behalf we execute
5319
  @param node: the node on which to create the device
5320
  @type instance: L{objects.Instance}
5321
  @param instance: the instance which owns the device
5322
  @type device: L{objects.Disk}
5323
  @param device: the device to create
5324
  @type force_create: boolean
5325
  @param force_create: whether to force creation of this device; this
5326
      will be change to True whenever we find a device which has
5327
      CreateOnSecondary() attribute
5328
  @param info: the extra 'metadata' we should attach to the device
5329
      (this will be represented as a LVM tag)
5330
  @type force_open: boolean
5331
  @param force_open: this parameter will be passes to the
5332
      L{backend.BlockdevCreate} function where it specifies
5333
      whether we run on primary or not, and it affects both
5334
      the child assembly and the device own Open() execution
5335

5336
  """
5337
  if device.CreateOnSecondary():
5338
    force_create = True
5339

    
5340
  if device.children:
5341
    for child in device.children:
5342
      _CreateBlockDev(lu, node, instance, child, force_create,
5343
                      info, force_open)
5344

    
5345
  if not force_create:
5346
    return
5347

    
5348
  _CreateSingleBlockDev(lu, node, instance, device, info, force_open)
5349

    
5350

    
5351
def _CreateSingleBlockDev(lu, node, instance, device, info, force_open):
5352
  """Create a single block device on a given node.
5353

5354
  This will not recurse over children of the device, so they must be
5355
  created in advance.
5356

5357
  @param lu: the lu on whose behalf we execute
5358
  @param node: the node on which to create the device
5359
  @type instance: L{objects.Instance}
5360
  @param instance: the instance which owns the device
5361
  @type device: L{objects.Disk}
5362
  @param device: the device to create
5363
  @param info: the extra 'metadata' we should attach to the device
5364
      (this will be represented as a LVM tag)
5365
  @type force_open: boolean
5366
  @param force_open: this parameter will be passes to the
5367
      L{backend.BlockdevCreate} function where it specifies
5368
      whether we run on primary or not, and it affects both
5369
      the child assembly and the device own Open() execution
5370

5371
  """
5372
  lu.cfg.SetDiskID(device, node)
5373
  result = lu.rpc.call_blockdev_create(node, device, device.size,
5374
                                       instance.name, force_open, info)
5375
  result.Raise("Can't create block device %s on"
5376
               " node %s for instance %s" % (device, node, instance.name))
5377
  if device.physical_id is None:
5378
    device.physical_id = result.payload
5379

    
5380

    
5381
def _GenerateUniqueNames(lu, exts):
5382
  """Generate a suitable LV name.
5383

5384
  This will generate a logical volume name for the given instance.
5385

5386
  """
5387
  results = []
5388
  for val in exts:
5389
    new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
5390
    results.append("%s%s" % (new_id, val))
5391
  return results
5392

    
5393

    
5394
def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
5395
                         p_minor, s_minor):
5396
  """Generate a drbd8 device complete with its children.
5397

5398
  """
5399
  port = lu.cfg.AllocatePort()
5400
  vgname = lu.cfg.GetVGName()
5401
  shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
5402
  dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
5403
                          logical_id=(vgname, names[0]))
5404
  dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
5405
                          logical_id=(vgname, names[1]))
5406
  drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
5407
                          logical_id=(primary, secondary, port,
5408
                                      p_minor, s_minor,
5409
                                      shared_secret),
5410
                          children=[dev_data, dev_meta],
5411
                          iv_name=iv_name)
5412
  return drbd_dev
5413

    
5414

    
5415
def _GenerateDiskTemplate(lu, template_name,
5416
                          instance_name, primary_node,
5417
                          secondary_nodes, disk_info,
5418
                          file_storage_dir, file_driver,
5419
                          base_index):
5420
  """Generate the entire disk layout for a given template type.
5421

5422
  """
5423
  #TODO: compute space requirements
5424

    
5425
  vgname = lu.cfg.GetVGName()
5426
  disk_count = len(disk_info)
5427
  disks = []
5428
  if template_name == constants.DT_DISKLESS:
5429
    pass
5430
  elif template_name == constants.DT_PLAIN:
5431
    if len(secondary_nodes) != 0:
5432
      raise errors.ProgrammerError("Wrong template configuration")
5433

    
5434
    names = _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
5435
                                      for i in range(disk_count)])
5436
    for idx, disk in enumerate(disk_info):
5437
      disk_index = idx + base_index
5438
      disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
5439
                              logical_id=(vgname, names[idx]),
5440
                              iv_name="disk/%d" % disk_index,
5441
                              mode=disk["mode"])
5442
      disks.append(disk_dev)
5443
  elif template_name == constants.DT_DRBD8:
5444
    if len(secondary_nodes) != 1:
5445
      raise errors.ProgrammerError("Wrong template configuration")
5446
    remote_node = secondary_nodes[0]
5447
    minors = lu.cfg.AllocateDRBDMinor(
5448
      [primary_node, remote_node] * len(disk_info), instance_name)
5449

    
5450
    names = []
5451
    for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
5452
                                               for i in range(disk_count)]):
5453
      names.append(lv_prefix + "_data")
5454
      names.append(lv_prefix + "_meta")
5455
    for idx, disk in enumerate(disk_info):
5456
      disk_index = idx + base_index
5457
      disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
5458
                                      disk["size"], names[idx*2:idx*2+2],
5459
                                      "disk/%d" % disk_index,
5460
                                      minors[idx*2], minors[idx*2+1])
5461
      disk_dev.mode = disk["mode"]
5462
      disks.append(disk_dev)
5463
  elif template_name == constants.DT_FILE:
5464
    if len(secondary_nodes) != 0:
5465
      raise errors.ProgrammerError("Wrong template configuration")
5466

    
5467
    for idx, disk in enumerate(disk_info):
5468
      disk_index = idx + base_index
5469
      disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
5470
                              iv_name="disk/%d" % disk_index,
5471
                              logical_id=(file_driver,
5472
                                          "%s/disk%d" % (file_storage_dir,
5473
                                                         disk_index)),
5474
                              mode=disk["mode"])
5475
      disks.append(disk_dev)
5476
  else:
5477
    raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
5478
  return disks
5479

    
5480

    
5481
def _GetInstanceInfoText(instance):
5482
  """Compute that text that should be added to the disk's metadata.
5483

5484
  """
5485
  return "originstname+%s" % instance.name
5486

    
5487

    
5488
def _CreateDisks(lu, instance, to_skip=None, target_node=None):
5489
  """Create all disks for an instance.
5490

5491
  This abstracts away some work from AddInstance.
5492

5493
  @type lu: L{LogicalUnit}
5494
  @param lu: the logical unit on whose behalf we execute
5495
  @type instance: L{objects.Instance}
5496
  @param instance: the instance whose disks we should create
5497
  @type to_skip: list
5498
  @param to_skip: list of indices to skip
5499
  @type target_node: string
5500
  @param target_node: if passed, overrides the target node for creation
5501
  @rtype: boolean
5502
  @return: the success of the creation
5503

5504
  """
5505
  info = _GetInstanceInfoText(instance)
5506
  if target_node is None:
5507
    pnode = instance.primary_node
5508
    all_nodes = instance.all_nodes
5509
  else:
5510
    pnode = target_node
5511
    all_nodes = [pnode]
5512

    
5513
  if instance.disk_template == constants.DT_FILE:
5514
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
5515
    result = lu.rpc.call_file_storage_dir_create(pnode, file_storage_dir)
5516

    
5517
    result.Raise("Failed to create directory '%s' on"
5518
                 " node %s" % (file_storage_dir, pnode))
5519

    
5520
  # Note: this needs to be kept in sync with adding of disks in
5521
  # LUSetInstanceParams
5522
  for idx, device in enumerate(instance.disks):
5523
    if to_skip and idx in to_skip:
5524
      continue
5525
    logging.info("Creating volume %s for instance %s",
5526
                 device.iv_name, instance.name)
5527
    #HARDCODE
5528
    for node in all_nodes:
5529
      f_create = node == pnode
5530
      _CreateBlockDev(lu, node, instance, device, f_create, info, f_create)
5531

    
5532

    
5533
def _RemoveDisks(lu, instance, target_node=None):
5534
  """Remove all disks for an instance.
5535

5536
  This abstracts away some work from `AddInstance()` and
5537
  `RemoveInstance()`. Note that in case some of the devices couldn't
5538
  be removed, the removal will continue with the other ones (compare
5539
  with `_CreateDisks()`).
5540

5541
  @type lu: L{LogicalUnit}
5542
  @param lu: the logical unit on whose behalf we execute
5543
  @type instance: L{objects.Instance}
5544
  @param instance: the instance whose disks we should remove
5545
  @type target_node: string
5546
  @param target_node: used to override the node on which to remove the disks
5547
  @rtype: boolean
5548
  @return: the success of the removal
5549

5550
  """
5551
  logging.info("Removing block devices for instance %s", instance.name)
5552

    
5553
  all_result = True
5554
  for device in instance.disks:
5555
    if target_node:
5556
      edata = [(target_node, device)]
5557
    else:
5558
      edata = device.ComputeNodeTree(instance.primary_node)
5559
    for node, disk in edata:
5560
      lu.cfg.SetDiskID(disk, node)
5561
      msg = lu.rpc.call_blockdev_remove(node, disk).fail_msg
5562
      if msg:
5563
        lu.LogWarning("Could not remove block device %s on node %s,"
5564
                      " continuing anyway: %s", device.iv_name, node, msg)
5565
        all_result = False
5566

    
5567
  if instance.disk_template == constants.DT_FILE:
5568
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
5569
    if target_node:
5570
      tgt = target_node
5571
    else:
5572
      tgt = instance.primary_node
5573
    result = lu.rpc.call_file_storage_dir_remove(tgt, file_storage_dir)
5574
    if result.fail_msg:
5575
      lu.LogWarning("Could not remove directory '%s' on node %s: %s",
5576
                    file_storage_dir, instance.primary_node, result.fail_msg)
5577
      all_result = False
5578

    
5579
  return all_result
5580

    
5581

    
5582
def _ComputeDiskSize(disk_template, disks):
5583
  """Compute disk size requirements in the volume group
5584

5585
  """
5586
  # Required free disk space as a function of disk and swap space
5587
  req_size_dict = {
5588
    constants.DT_DISKLESS: None,
5589
    constants.DT_PLAIN: sum(d["size"] for d in disks),
5590
    # 128 MB are added for drbd metadata for each disk
5591
    constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
5592
    constants.DT_FILE: None,
5593
  }
5594

    
5595
  if disk_template not in req_size_dict:
5596
    raise errors.ProgrammerError("Disk template '%s' size requirement"
5597
                                 " is unknown" %  disk_template)
5598

    
5599
  return req_size_dict[disk_template]
5600

    
5601

    
5602
def _CheckHVParams(lu, nodenames, hvname, hvparams):
5603
  """Hypervisor parameter validation.
5604

5605
  This function abstract the hypervisor parameter validation to be
5606
  used in both instance create and instance modify.
5607

5608
  @type lu: L{LogicalUnit}
5609
  @param lu: the logical unit for which we check
5610
  @type nodenames: list
5611
  @param nodenames: the list of nodes on which we should check
5612
  @type hvname: string
5613
  @param hvname: the name of the hypervisor we should use
5614
  @type hvparams: dict
5615
  @param hvparams: the parameters which we need to check
5616
  @raise errors.OpPrereqError: if the parameters are not valid
5617

5618
  """
5619
  hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
5620
                                                  hvname,
5621
                                                  hvparams)
5622
  for node in nodenames:
5623
    info = hvinfo[node]
5624
    if info.offline:
5625
      continue
5626
    info.Raise("Hypervisor parameter validation failed on node %s" % node)
5627

    
5628

    
5629
class LUCreateInstance(LogicalUnit):
5630
  """Create an instance.
5631

5632
  """
5633
  HPATH = "instance-add"
5634
  HTYPE = constants.HTYPE_INSTANCE
5635
  _OP_REQP = ["instance_name", "disks", "disk_template",
5636
              "mode", "start",
5637
              "wait_for_sync", "ip_check", "nics",
5638
              "hvparams", "beparams"]
5639
  REQ_BGL = False
5640

    
5641
  def CheckArguments(self):
5642
    """Check arguments.
5643

5644
    """
5645
    # do not require name_check to ease forward/backward compatibility
5646
    # for tools
5647
    if not hasattr(self.op, "name_check"):
5648
      self.op.name_check = True
5649
    if self.op.ip_check and not self.op.name_check:
5650
      # TODO: make the ip check more flexible and not depend on the name check
5651
      raise errors.OpPrereqError("Cannot do ip checks without a name check",
5652
                                 errors.ECODE_INVAL)
5653

    
5654
  def _ExpandNode(self, node):
5655
    """Expands and checks one node name.
5656

5657
    """
5658
    node_full = self.cfg.ExpandNodeName(node)
5659
    if node_full is None:
5660
      raise errors.OpPrereqError("Unknown node %s" % node, errors.ECODE_NOENT)
5661
    return node_full
5662

    
5663
  def ExpandNames(self):
5664
    """ExpandNames for CreateInstance.
5665

5666
    Figure out the right locks for instance creation.
5667

5668
    """
5669
    self.needed_locks = {}
5670

    
5671
    # set optional parameters to none if they don't exist
5672
    for attr in ["pnode", "snode", "iallocator", "hypervisor"]:
5673
      if not hasattr(self.op, attr):
5674
        setattr(self.op, attr, None)
5675

    
5676
    # cheap checks, mostly valid constants given
5677

    
5678
    # verify creation mode
5679
    if self.op.mode not in (constants.INSTANCE_CREATE,
5680
                            constants.INSTANCE_IMPORT):
5681
      raise errors.OpPrereqError("Invalid instance creation mode '%s'" %
5682
                                 self.op.mode, errors.ECODE_INVAL)
5683

    
5684
    # disk template and mirror node verification
5685
    if self.op.disk_template not in constants.DISK_TEMPLATES:
5686
      raise errors.OpPrereqError("Invalid disk template name",
5687
                                 errors.ECODE_INVAL)
5688

    
5689
    if self.op.hypervisor is None:
5690
      self.op.hypervisor = self.cfg.GetHypervisorType()
5691

    
5692
    cluster = self.cfg.GetClusterInfo()
5693
    enabled_hvs = cluster.enabled_hypervisors
5694
    if self.op.hypervisor not in enabled_hvs:
5695
      raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
5696
                                 " cluster (%s)" % (self.op.hypervisor,
5697
                                  ",".join(enabled_hvs)),
5698
                                 errors.ECODE_STATE)
5699

    
5700
    # check hypervisor parameter syntax (locally)
5701
    utils.ForceDictType(self.op.hvparams, constants.HVS_PARAMETER_TYPES)
5702
    filled_hvp = objects.FillDict(cluster.hvparams[self.op.hypervisor],
5703
                                  self.op.hvparams)
5704
    hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
5705
    hv_type.CheckParameterSyntax(filled_hvp)
5706
    self.hv_full = filled_hvp
5707
    # check that we don't specify global parameters on an instance
5708
    _CheckGlobalHvParams(self.op.hvparams)
5709

    
5710
    # fill and remember the beparams dict
5711
    utils.ForceDictType(self.op.beparams, constants.BES_PARAMETER_TYPES)
5712
    self.be_full = objects.FillDict(cluster.beparams[constants.PP_DEFAULT],
5713
                                    self.op.beparams)
5714

    
5715
    #### instance parameters check
5716

    
5717
    # instance name verification
5718
    if self.op.name_check:
5719
      hostname1 = utils.GetHostInfo(self.op.instance_name)
5720
      self.op.instance_name = instance_name = hostname1.name
5721
      # used in CheckPrereq for ip ping check
5722
      self.check_ip = hostname1.ip
5723
    else:
5724
      instance_name = self.op.instance_name
5725
      self.check_ip = None
5726

    
5727
    # this is just a preventive check, but someone might still add this
5728
    # instance in the meantime, and creation will fail at lock-add time
5729
    if instance_name in self.cfg.GetInstanceList():
5730
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
5731
                                 instance_name, errors.ECODE_EXISTS)
5732

    
5733
    self.add_locks[locking.LEVEL_INSTANCE] = instance_name
5734

    
5735
    # NIC buildup
5736
    self.nics = []
5737
    for idx, nic in enumerate(self.op.nics):
5738
      nic_mode_req = nic.get("mode", None)
5739
      nic_mode = nic_mode_req
5740
      if nic_mode is None:
5741
        nic_mode = cluster.nicparams[constants.PP_DEFAULT][constants.NIC_MODE]
5742

    
5743
      # in routed mode, for the first nic, the default ip is 'auto'
5744
      if nic_mode == constants.NIC_MODE_ROUTED and idx == 0:
5745
        default_ip_mode = constants.VALUE_AUTO
5746
      else:
5747
        default_ip_mode = constants.VALUE_NONE
5748

    
5749
      # ip validity checks
5750
      ip = nic.get("ip", default_ip_mode)
5751
      if ip is None or ip.lower() == constants.VALUE_NONE:
5752
        nic_ip = None
5753
      elif ip.lower() == constants.VALUE_AUTO:
5754
        if not self.op.name_check:
5755
          raise errors.OpPrereqError("IP address set to auto but name checks"
5756
                                     " have been skipped. Aborting.",
5757
                                     errors.ECODE_INVAL)
5758
        nic_ip = hostname1.ip
5759
      else:
5760
        if not utils.IsValidIP(ip):
5761
          raise errors.OpPrereqError("Given IP address '%s' doesn't look"
5762
                                     " like a valid IP" % ip,
5763
                                     errors.ECODE_INVAL)
5764
        nic_ip = ip
5765

    
5766
      # TODO: check the ip address for uniqueness
5767
      if nic_mode == constants.NIC_MODE_ROUTED and not nic_ip:
5768
        raise errors.OpPrereqError("Routed nic mode requires an ip address",
5769
                                   errors.ECODE_INVAL)
5770

    
5771
      # MAC address verification
5772
      mac = nic.get("mac", constants.VALUE_AUTO)
5773
      if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
5774
        mac = utils.NormalizeAndValidateMac(mac)
5775

    
5776
        try:
5777
          self.cfg.ReserveMAC(mac, self.proc.GetECId())
5778
        except errors.ReservationError:
5779
          raise errors.OpPrereqError("MAC address %s already in use"
5780
                                     " in cluster" % mac,
5781
                                     errors.ECODE_NOTUNIQUE)
5782

    
5783
      # bridge verification
5784
      bridge = nic.get("bridge", None)
5785
      link = nic.get("link", None)
5786
      if bridge and link:
5787
        raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
5788
                                   " at the same time", errors.ECODE_INVAL)
5789
      elif bridge and nic_mode == constants.NIC_MODE_ROUTED:
5790
        raise errors.OpPrereqError("Cannot pass 'bridge' on a routed nic",
5791
                                   errors.ECODE_INVAL)
5792
      elif bridge:
5793
        link = bridge
5794

    
5795
      nicparams = {}
5796
      if nic_mode_req:
5797
        nicparams[constants.NIC_MODE] = nic_mode_req
5798
      if link:
5799
        nicparams[constants.NIC_LINK] = link
5800

    
5801
      check_params = objects.FillDict(cluster.nicparams[constants.PP_DEFAULT],
5802
                                      nicparams)
5803
      objects.NIC.CheckParameterSyntax(check_params)
5804
      self.nics.append(objects.NIC(mac=mac, ip=nic_ip, nicparams=nicparams))
5805

    
5806
    # disk checks/pre-build
5807
    self.disks = []
5808
    for disk in self.op.disks:
5809
      mode = disk.get("mode", constants.DISK_RDWR)
5810
      if mode not in constants.DISK_ACCESS_SET:
5811
        raise errors.OpPrereqError("Invalid disk access mode '%s'" %
5812
                                   mode, errors.ECODE_INVAL)
5813
      size = disk.get("size", None)
5814
      if size is None:
5815
        raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
5816
      try:
5817
        size = int(size)
5818
      except ValueError:
5819
        raise errors.OpPrereqError("Invalid disk size '%s'" % size,
5820
                                   errors.ECODE_INVAL)
5821
      self.disks.append({"size": size, "mode": mode})
5822

    
5823
    # file storage checks
5824
    if (self.op.file_driver and
5825
        not self.op.file_driver in constants.FILE_DRIVER):
5826
      raise errors.OpPrereqError("Invalid file driver name '%s'" %
5827
                                 self.op.file_driver, errors.ECODE_INVAL)
5828

    
5829
    if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
5830
      raise errors.OpPrereqError("File storage directory path not absolute",
5831
                                 errors.ECODE_INVAL)
5832

    
5833
    ### Node/iallocator related checks
5834
    if [self.op.iallocator, self.op.pnode].count(None) != 1:
5835
      raise errors.OpPrereqError("One and only one of iallocator and primary"
5836
                                 " node must be given",
5837
                                 errors.ECODE_INVAL)
5838

    
5839
    if self.op.iallocator:
5840
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5841
    else:
5842
      self.op.pnode = self._ExpandNode(self.op.pnode)
5843
      nodelist = [self.op.pnode]
5844
      if self.op.snode is not None:
5845
        self.op.snode = self._ExpandNode(self.op.snode)
5846
        nodelist.append(self.op.snode)
5847
      self.needed_locks[locking.LEVEL_NODE] = nodelist
5848

    
5849
    # in case of import lock the source node too
5850
    if self.op.mode == constants.INSTANCE_IMPORT:
5851
      src_node = getattr(self.op, "src_node", None)
5852
      src_path = getattr(self.op, "src_path", None)
5853

    
5854
      if src_path is None:
5855
        self.op.src_path = src_path = self.op.instance_name
5856

    
5857
      if src_node is None:
5858
        self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5859
        self.op.src_node = None
5860
        if os.path.isabs(src_path):
5861
          raise errors.OpPrereqError("Importing an instance from an absolute"
5862
                                     " path requires a source node option.",
5863
                                     errors.ECODE_INVAL)
5864
      else:
5865
        self.op.src_node = src_node = self._ExpandNode(src_node)
5866
        if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
5867
          self.needed_locks[locking.LEVEL_NODE].append(src_node)
5868
        if not os.path.isabs(src_path):
5869
          self.op.src_path = src_path = \
5870
            os.path.join(constants.EXPORT_DIR, src_path)
5871

    
5872
      # On import force_variant must be True, because if we forced it at
5873
      # initial install, our only chance when importing it back is that it
5874
      # works again!
5875
      self.op.force_variant = True
5876

    
5877
    else: # INSTANCE_CREATE
5878
      if getattr(self.op, "os_type", None) is None:
5879
        raise errors.OpPrereqError("No guest OS specified",
5880
                                   errors.ECODE_INVAL)
5881
      self.op.force_variant = getattr(self.op, "force_variant", False)
5882

    
5883
  def _RunAllocator(self):
5884
    """Run the allocator based on input opcode.
5885

5886
    """
5887
    nics = [n.ToDict() for n in self.nics]
5888
    ial = IAllocator(self.cfg, self.rpc,
5889
                     mode=constants.IALLOCATOR_MODE_ALLOC,
5890
                     name=self.op.instance_name,
5891
                     disk_template=self.op.disk_template,
5892
                     tags=[],
5893
                     os=self.op.os_type,
5894
                     vcpus=self.be_full[constants.BE_VCPUS],
5895
                     mem_size=self.be_full[constants.BE_MEMORY],
5896
                     disks=self.disks,
5897
                     nics=nics,
5898
                     hypervisor=self.op.hypervisor,
5899
                     )
5900

    
5901
    ial.Run(self.op.iallocator)
5902

    
5903
    if not ial.success:
5904
      raise errors.OpPrereqError("Can't compute nodes using"
5905
                                 " iallocator '%s': %s" %
5906
                                 (self.op.iallocator, ial.info),
5907
                                 errors.ECODE_NORES)
5908
    if len(ial.nodes) != ial.required_nodes:
5909
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
5910
                                 " of nodes (%s), required %s" %
5911
                                 (self.op.iallocator, len(ial.nodes),
5912
                                  ial.required_nodes), errors.ECODE_FAULT)
5913
    self.op.pnode = ial.nodes[0]
5914
    self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
5915
                 self.op.instance_name, self.op.iallocator,
5916
                 utils.CommaJoin(ial.nodes))
5917
    if ial.required_nodes == 2:
5918
      self.op.snode = ial.nodes[1]
5919

    
5920
  def BuildHooksEnv(self):
5921
    """Build hooks env.
5922

5923
    This runs on master, primary and secondary nodes of the instance.
5924

5925
    """
5926
    env = {
5927
      "ADD_MODE": self.op.mode,
5928
      }
5929
    if self.op.mode == constants.INSTANCE_IMPORT:
5930
      env["SRC_NODE"] = self.op.src_node
5931
      env["SRC_PATH"] = self.op.src_path
5932
      env["SRC_IMAGES"] = self.src_images
5933

    
5934
    env.update(_BuildInstanceHookEnv(
5935
      name=self.op.instance_name,
5936
      primary_node=self.op.pnode,
5937
      secondary_nodes=self.secondaries,
5938
      status=self.op.start,
5939
      os_type=self.op.os_type,
5940
      memory=self.be_full[constants.BE_MEMORY],
5941
      vcpus=self.be_full[constants.BE_VCPUS],
5942
      nics=_NICListToTuple(self, self.nics),
5943
      disk_template=self.op.disk_template,
5944
      disks=[(d["size"], d["mode"]) for d in self.disks],
5945
      bep=self.be_full,
5946
      hvp=self.hv_full,
5947
      hypervisor_name=self.op.hypervisor,
5948
    ))
5949

    
5950
    nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
5951
          self.secondaries)
5952
    return env, nl, nl
5953

    
5954

    
5955
  def CheckPrereq(self):
5956
    """Check prerequisites.
5957

5958
    """
5959
    if (not self.cfg.GetVGName() and
5960
        self.op.disk_template not in constants.DTS_NOT_LVM):
5961
      raise errors.OpPrereqError("Cluster does not support lvm-based"
5962
                                 " instances", errors.ECODE_STATE)
5963

    
5964
    if self.op.mode == constants.INSTANCE_IMPORT:
5965
      src_node = self.op.src_node
5966
      src_path = self.op.src_path
5967

    
5968
      if src_node is None:
5969
        locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
5970
        exp_list = self.rpc.call_export_list(locked_nodes)
5971
        found = False
5972
        for node in exp_list:
5973
          if exp_list[node].fail_msg:
5974
            continue
5975
          if src_path in exp_list[node].payload:
5976
            found = True
5977
            self.op.src_node = src_node = node
5978
            self.op.src_path = src_path = os.path.join(constants.EXPORT_DIR,
5979
                                                       src_path)
5980
            break
5981
        if not found:
5982
          raise errors.OpPrereqError("No export found for relative path %s" %
5983
                                      src_path, errors.ECODE_INVAL)
5984

    
5985
      _CheckNodeOnline(self, src_node)
5986
      result = self.rpc.call_export_info(src_node, src_path)
5987
      result.Raise("No export or invalid export found in dir %s" % src_path)
5988

    
5989
      export_info = objects.SerializableConfigParser.Loads(str(result.payload))
5990
      if not export_info.has_section(constants.INISECT_EXP):
5991
        raise errors.ProgrammerError("Corrupted export config",
5992
                                     errors.ECODE_ENVIRON)
5993

    
5994
      ei_version = export_info.get(constants.INISECT_EXP, 'version')
5995
      if (int(ei_version) != constants.EXPORT_VERSION):
5996
        raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
5997
                                   (ei_version, constants.EXPORT_VERSION),
5998
                                   errors.ECODE_ENVIRON)
5999

    
6000
      # Check that the new instance doesn't have less disks than the export
6001
      instance_disks = len(self.disks)
6002
      export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
6003
      if instance_disks < export_disks:
6004
        raise errors.OpPrereqError("Not enough disks to import."
6005
                                   " (instance: %d, export: %d)" %
6006
                                   (instance_disks, export_disks),
6007
                                   errors.ECODE_INVAL)
6008

    
6009
      self.op.os_type = export_info.get(constants.INISECT_EXP, 'os')
6010
      disk_images = []
6011
      for idx in range(export_disks):
6012
        option = 'disk%d_dump' % idx
6013
        if export_info.has_option(constants.INISECT_INS, option):
6014
          # FIXME: are the old os-es, disk sizes, etc. useful?
6015
          export_name = export_info.get(constants.INISECT_INS, option)
6016
          image = os.path.join(src_path, export_name)
6017
          disk_images.append(image)
6018
        else:
6019
          disk_images.append(False)
6020

    
6021
      self.src_images = disk_images
6022

    
6023
      old_name = export_info.get(constants.INISECT_INS, 'name')
6024
      # FIXME: int() here could throw a ValueError on broken exports
6025
      exp_nic_count = int(export_info.get(constants.INISECT_INS, 'nic_count'))
6026
      if self.op.instance_name == old_name:
6027
        for idx, nic in enumerate(self.nics):
6028
          if nic.mac == constants.VALUE_AUTO and exp_nic_count >= idx:
6029
            nic_mac_ini = 'nic%d_mac' % idx
6030
            nic.mac = export_info.get(constants.INISECT_INS, nic_mac_ini)
6031

    
6032
    # ENDIF: self.op.mode == constants.INSTANCE_IMPORT
6033

    
6034
    # ip ping checks (we use the same ip that was resolved in ExpandNames)
6035
    if self.op.ip_check:
6036
      if utils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
6037
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
6038
                                   (self.check_ip, self.op.instance_name),
6039
                                   errors.ECODE_NOTUNIQUE)
6040

    
6041
    #### mac address generation
6042
    # By generating here the mac address both the allocator and the hooks get
6043
    # the real final mac address rather than the 'auto' or 'generate' value.
6044
    # There is a race condition between the generation and the instance object
6045
    # creation, which means that we know the mac is valid now, but we're not
6046
    # sure it will be when we actually add the instance. If things go bad
6047
    # adding the instance will abort because of a duplicate mac, and the
6048
    # creation job will fail.
6049
    for nic in self.nics:
6050
      if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
6051
        nic.mac = self.cfg.GenerateMAC(self.proc.GetECId())
6052

    
6053
    #### allocator run
6054

    
6055
    if self.op.iallocator is not None:
6056
      self._RunAllocator()
6057

    
6058
    #### node related checks
6059

    
6060
    # check primary node
6061
    self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
6062
    assert self.pnode is not None, \
6063
      "Cannot retrieve locked node %s" % self.op.pnode
6064
    if pnode.offline:
6065
      raise errors.OpPrereqError("Cannot use offline primary node '%s'" %
6066
                                 pnode.name, errors.ECODE_STATE)
6067
    if pnode.drained:
6068
      raise errors.OpPrereqError("Cannot use drained primary node '%s'" %
6069
                                 pnode.name, errors.ECODE_STATE)
6070

    
6071
    self.secondaries = []
6072

    
6073
    # mirror node verification
6074
    if self.op.disk_template in constants.DTS_NET_MIRROR:
6075
      if self.op.snode is None:
6076
        raise errors.OpPrereqError("The networked disk templates need"
6077
                                   " a mirror node", errors.ECODE_INVAL)
6078
      if self.op.snode == pnode.name:
6079
        raise errors.OpPrereqError("The secondary node cannot be the"
6080
                                   " primary node.", errors.ECODE_INVAL)
6081
      _CheckNodeOnline(self, self.op.snode)
6082
      _CheckNodeNotDrained(self, self.op.snode)
6083
      self.secondaries.append(self.op.snode)
6084

    
6085
    nodenames = [pnode.name] + self.secondaries
6086

    
6087
    req_size = _ComputeDiskSize(self.op.disk_template,
6088
                                self.disks)
6089

    
6090
    # Check lv size requirements
6091
    if req_size is not None:
6092
      nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
6093
                                         self.op.hypervisor)
6094
      for node in nodenames:
6095
        info = nodeinfo[node]
6096
        info.Raise("Cannot get current information from node %s" % node)
6097
        info = info.payload
6098
        vg_free = info.get('vg_free', None)
6099
        if not isinstance(vg_free, int):
6100
          raise errors.OpPrereqError("Can't compute free disk space on"
6101
                                     " node %s" % node, errors.ECODE_ENVIRON)
6102
        if req_size > vg_free:
6103
          raise errors.OpPrereqError("Not enough disk space on target node %s."
6104
                                     " %d MB available, %d MB required" %
6105
                                     (node, vg_free, req_size),
6106
                                     errors.ECODE_NORES)
6107

    
6108
    _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
6109

    
6110
    # os verification
6111
    result = self.rpc.call_os_get(pnode.name, self.op.os_type)
6112
    result.Raise("OS '%s' not in supported os list for primary node %s" %
6113
                 (self.op.os_type, pnode.name),
6114
                 prereq=True, ecode=errors.ECODE_INVAL)
6115
    if not self.op.force_variant:
6116
      _CheckOSVariant(result.payload, self.op.os_type)
6117

    
6118
    _CheckNicsBridgesExist(self, self.nics, self.pnode.name)
6119

    
6120
    # memory check on primary node
6121
    if self.op.start:
6122
      _CheckNodeFreeMemory(self, self.pnode.name,
6123
                           "creating instance %s" % self.op.instance_name,
6124
                           self.be_full[constants.BE_MEMORY],
6125
                           self.op.hypervisor)
6126

    
6127
    self.dry_run_result = list(nodenames)
6128

    
6129
  def Exec(self, feedback_fn):
6130
    """Create and add the instance to the cluster.
6131

6132
    """
6133
    instance = self.op.instance_name
6134
    pnode_name = self.pnode.name
6135

    
6136
    ht_kind = self.op.hypervisor
6137
    if ht_kind in constants.HTS_REQ_PORT:
6138
      network_port = self.cfg.AllocatePort()
6139
    else:
6140
      network_port = None
6141

    
6142
    ##if self.op.vnc_bind_address is None:
6143
    ##  self.op.vnc_bind_address = constants.VNC_DEFAULT_BIND_ADDRESS
6144

    
6145
    # this is needed because os.path.join does not accept None arguments
6146
    if self.op.file_storage_dir is None:
6147
      string_file_storage_dir = ""
6148
    else:
6149
      string_file_storage_dir = self.op.file_storage_dir
6150

    
6151
    # build the full file storage dir path
6152
    file_storage_dir = os.path.normpath(os.path.join(
6153
                                        self.cfg.GetFileStorageDir(),
6154
                                        string_file_storage_dir, instance))
6155

    
6156

    
6157
    disks = _GenerateDiskTemplate(self,
6158
                                  self.op.disk_template,
6159
                                  instance, pnode_name,
6160
                                  self.secondaries,
6161
                                  self.disks,
6162
                                  file_storage_dir,
6163
                                  self.op.file_driver,
6164
                                  0)
6165

    
6166
    iobj = objects.Instance(name=instance, os=self.op.os_type,
6167
                            primary_node=pnode_name,
6168
                            nics=self.nics, disks=disks,
6169
                            disk_template=self.op.disk_template,
6170
                            admin_up=False,
6171
                            network_port=network_port,
6172
                            beparams=self.op.beparams,
6173
                            hvparams=self.op.hvparams,
6174
                            hypervisor=self.op.hypervisor,
6175
                            )
6176

    
6177
    feedback_fn("* creating instance disks...")
6178
    try:
6179
      _CreateDisks(self, iobj)
6180
    except errors.OpExecError:
6181
      self.LogWarning("Device creation failed, reverting...")
6182
      try:
6183
        _RemoveDisks(self, iobj)
6184
      finally:
6185
        self.cfg.ReleaseDRBDMinors(instance)
6186
        raise
6187

    
6188
    feedback_fn("adding instance %s to cluster config" % instance)
6189

    
6190
    self.cfg.AddInstance(iobj, self.proc.GetECId())
6191

    
6192
    # Declare that we don't want to remove the instance lock anymore, as we've
6193
    # added the instance to the config
6194
    del self.remove_locks[locking.LEVEL_INSTANCE]
6195
    # Unlock all the nodes
6196
    if self.op.mode == constants.INSTANCE_IMPORT:
6197
      nodes_keep = [self.op.src_node]
6198
      nodes_release = [node for node in self.acquired_locks[locking.LEVEL_NODE]
6199
                       if node != self.op.src_node]
6200
      self.context.glm.release(locking.LEVEL_NODE, nodes_release)
6201
      self.acquired_locks[locking.LEVEL_NODE] = nodes_keep
6202
    else:
6203
      self.context.glm.release(locking.LEVEL_NODE)
6204
      del self.acquired_locks[locking.LEVEL_NODE]
6205

    
6206
    if self.op.wait_for_sync:
6207
      disk_abort = not _WaitForSync(self, iobj)
6208
    elif iobj.disk_template in constants.DTS_NET_MIRROR:
6209
      # make sure the disks are not degraded (still sync-ing is ok)
6210
      time.sleep(15)
6211
      feedback_fn("* checking mirrors status")
6212
      disk_abort = not _WaitForSync(self, iobj, oneshot=True)
6213
    else:
6214
      disk_abort = False
6215

    
6216
    if disk_abort:
6217
      _RemoveDisks(self, iobj)
6218
      self.cfg.RemoveInstance(iobj.name)
6219
      # Make sure the instance lock gets removed
6220
      self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
6221
      raise errors.OpExecError("There are some degraded disks for"
6222
                               " this instance")
6223

    
6224
    feedback_fn("creating os for instance %s on node %s" %
6225
                (instance, pnode_name))
6226

    
6227
    if iobj.disk_template != constants.DT_DISKLESS:
6228
      if self.op.mode == constants.INSTANCE_CREATE:
6229
        feedback_fn("* running the instance OS create scripts...")
6230
        result = self.rpc.call_instance_os_add(pnode_name, iobj, False)
6231
        result.Raise("Could not add os for instance %s"
6232
                     " on node %s" % (instance, pnode_name))
6233

    
6234
      elif self.op.mode == constants.INSTANCE_IMPORT:
6235
        feedback_fn("* running the instance OS import scripts...")
6236
        src_node = self.op.src_node
6237
        src_images = self.src_images
6238
        cluster_name = self.cfg.GetClusterName()
6239
        import_result = self.rpc.call_instance_os_import(pnode_name, iobj,
6240
                                                         src_node, src_images,
6241
                                                         cluster_name)
6242
        msg = import_result.fail_msg
6243
        if msg:
6244
          self.LogWarning("Error while importing the disk images for instance"
6245
                          " %s on node %s: %s" % (instance, pnode_name, msg))
6246
      else:
6247
        # also checked in the prereq part
6248
        raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
6249
                                     % self.op.mode)
6250

    
6251
    if self.op.start:
6252
      iobj.admin_up = True
6253
      self.cfg.Update(iobj, feedback_fn)
6254
      logging.info("Starting instance %s on node %s", instance, pnode_name)
6255
      feedback_fn("* starting instance...")
6256
      result = self.rpc.call_instance_start(pnode_name, iobj, None, None)
6257
      result.Raise("Could not start instance")
6258

    
6259
    return list(iobj.all_nodes)
6260

    
6261

    
6262
class LUConnectConsole(NoHooksLU):
6263
  """Connect to an instance's console.
6264

6265
  This is somewhat special in that it returns the command line that
6266
  you need to run on the master node in order to connect to the
6267
  console.
6268

6269
  """
6270
  _OP_REQP = ["instance_name"]
6271
  REQ_BGL = False
6272

    
6273
  def ExpandNames(self):
6274
    self._ExpandAndLockInstance()
6275

    
6276
  def CheckPrereq(self):
6277
    """Check prerequisites.
6278

6279
    This checks that the instance is in the cluster.
6280

6281
    """
6282
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
6283
    assert self.instance is not None, \
6284
      "Cannot retrieve locked instance %s" % self.op.instance_name
6285
    _CheckNodeOnline(self, self.instance.primary_node)
6286

    
6287
  def Exec(self, feedback_fn):
6288
    """Connect to the console of an instance
6289

6290
    """
6291
    instance = self.instance
6292
    node = instance.primary_node
6293

    
6294
    node_insts = self.rpc.call_instance_list([node],
6295
                                             [instance.hypervisor])[node]
6296
    node_insts.Raise("Can't get node information from %s" % node)
6297

    
6298
    if instance.name not in node_insts.payload:
6299
      raise errors.OpExecError("Instance %s is not running." % instance.name)
6300

    
6301
    logging.debug("Connecting to console of %s on %s", instance.name, node)
6302

    
6303
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
6304
    cluster = self.cfg.GetClusterInfo()
6305
    # beparams and hvparams are passed separately, to avoid editing the
6306
    # instance and then saving the defaults in the instance itself.
6307
    hvparams = cluster.FillHV(instance)
6308
    beparams = cluster.FillBE(instance)
6309
    console_cmd = hyper.GetShellCommandForConsole(instance, hvparams, beparams)
6310

    
6311
    # build ssh cmdline
6312
    return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
6313

    
6314

    
6315
class LUReplaceDisks(LogicalUnit):
6316
  """Replace the disks of an instance.
6317

6318
  """
6319
  HPATH = "mirrors-replace"
6320
  HTYPE = constants.HTYPE_INSTANCE
6321
  _OP_REQP = ["instance_name", "mode", "disks"]
6322
  REQ_BGL = False
6323

    
6324
  def CheckArguments(self):
6325
    if not hasattr(self.op, "remote_node"):
6326
      self.op.remote_node = None
6327
    if not hasattr(self.op, "iallocator"):
6328
      self.op.iallocator = None
6329

    
6330
    TLReplaceDisks.CheckArguments(self.op.mode, self.op.remote_node,
6331
                                  self.op.iallocator)
6332

    
6333
  def ExpandNames(self):
6334
    self._ExpandAndLockInstance()
6335

    
6336
    if self.op.iallocator is not None:
6337
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6338

    
6339
    elif self.op.remote_node is not None:
6340
      remote_node = self.cfg.ExpandNodeName(self.op.remote_node)
6341
      if remote_node is None:
6342
        raise errors.OpPrereqError("Node '%s' not known" %
6343
                                   self.op.remote_node, errors.ECODE_NOENT)
6344

    
6345
      self.op.remote_node = remote_node
6346

    
6347
      # Warning: do not remove the locking of the new secondary here
6348
      # unless DRBD8.AddChildren is changed to work in parallel;
6349
      # currently it doesn't since parallel invocations of
6350
      # FindUnusedMinor will conflict
6351
      self.needed_locks[locking.LEVEL_NODE] = [remote_node]
6352
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
6353

    
6354
    else:
6355
      self.needed_locks[locking.LEVEL_NODE] = []
6356
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
6357

    
6358
    self.replacer = TLReplaceDisks(self, self.op.instance_name, self.op.mode,
6359
                                   self.op.iallocator, self.op.remote_node,
6360
                                   self.op.disks)
6361

    
6362
    self.tasklets = [self.replacer]
6363

    
6364
  def DeclareLocks(self, level):
6365
    # If we're not already locking all nodes in the set we have to declare the
6366
    # instance's primary/secondary nodes.
6367
    if (level == locking.LEVEL_NODE and
6368
        self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
6369
      self._LockInstancesNodes()
6370

    
6371
  def BuildHooksEnv(self):
6372
    """Build hooks env.
6373

6374
    This runs on the master, the primary and all the secondaries.
6375

6376
    """
6377
    instance = self.replacer.instance
6378
    env = {
6379
      "MODE": self.op.mode,
6380
      "NEW_SECONDARY": self.op.remote_node,
6381
      "OLD_SECONDARY": instance.secondary_nodes[0],
6382
      }
6383
    env.update(_BuildInstanceHookEnvByObject(self, instance))
6384
    nl = [
6385
      self.cfg.GetMasterNode(),
6386
      instance.primary_node,
6387
      ]
6388
    if self.op.remote_node is not None:
6389
      nl.append(self.op.remote_node)
6390
    return env, nl, nl
6391

    
6392

    
6393
class LUEvacuateNode(LogicalUnit):
6394
  """Relocate the secondary instances from a node.
6395

6396
  """
6397
  HPATH = "node-evacuate"
6398
  HTYPE = constants.HTYPE_NODE
6399
  _OP_REQP = ["node_name"]
6400
  REQ_BGL = False
6401

    
6402
  def CheckArguments(self):
6403
    if not hasattr(self.op, "remote_node"):
6404
      self.op.remote_node = None
6405
    if not hasattr(self.op, "iallocator"):
6406
      self.op.iallocator = None
6407

    
6408
    TLReplaceDisks.CheckArguments(constants.REPLACE_DISK_CHG,
6409
                                  self.op.remote_node,
6410
                                  self.op.iallocator)
6411

    
6412
  def ExpandNames(self):
6413
    self.op.node_name = self.cfg.ExpandNodeName(self.op.node_name)
6414
    if self.op.node_name is None:
6415
      raise errors.OpPrereqError("Node '%s' not known" % self.op.node_name,
6416
                                 errors.ECODE_NOENT)
6417

    
6418
    self.needed_locks = {}
6419

    
6420
    # Declare node locks
6421
    if self.op.iallocator is not None:
6422
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
6423

    
6424
    elif self.op.remote_node is not None:
6425
      remote_node = self.cfg.ExpandNodeName(self.op.remote_node)
6426
      if remote_node is None:
6427
        raise errors.OpPrereqError("Node '%s' not known" %
6428
                                   self.op.remote_node, errors.ECODE_NOENT)
6429

    
6430
      self.op.remote_node = remote_node
6431

    
6432
      # Warning: do not remove the locking of the new secondary here
6433
      # unless DRBD8.AddChildren is changed to work in parallel;
6434
      # currently it doesn't since parallel invocations of
6435
      # FindUnusedMinor will conflict
6436
      self.needed_locks[locking.LEVEL_NODE] = [remote_node]
6437
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
6438

    
6439
    else:
6440
      raise errors.OpPrereqError("Invalid parameters", errors.ECODE_INVAL)
6441

    
6442
    # Create tasklets for replacing disks for all secondary instances on this
6443
    # node
6444
    names = []
6445
    tasklets = []
6446

    
6447
    for inst in _GetNodeSecondaryInstances(self.cfg, self.op.node_name):
6448
      logging.debug("Replacing disks for instance %s", inst.name)
6449
      names.append(inst.name)
6450

    
6451
      replacer = TLReplaceDisks(self, inst.name, constants.REPLACE_DISK_CHG,
6452
                                self.op.iallocator, self.op.remote_node, [])
6453
      tasklets.append(replacer)
6454

    
6455
    self.tasklets = tasklets
6456
    self.instance_names = names
6457

    
6458
    # Declare instance locks
6459
    self.needed_locks[locking.LEVEL_INSTANCE] = self.instance_names
6460

    
6461
  def DeclareLocks(self, level):
6462
    # If we're not already locking all nodes in the set we have to declare the
6463
    # instance's primary/secondary nodes.
6464
    if (level == locking.LEVEL_NODE and
6465
        self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
6466
      self._LockInstancesNodes()
6467

    
6468
  def BuildHooksEnv(self):
6469
    """Build hooks env.
6470

6471
    This runs on the master, the primary and all the secondaries.
6472

6473
    """
6474
    env = {
6475
      "NODE_NAME": self.op.node_name,
6476
      }
6477

    
6478
    nl = [self.cfg.GetMasterNode()]
6479

    
6480
    if self.op.remote_node is not None:
6481
      env["NEW_SECONDARY"] = self.op.remote_node
6482
      nl.append(self.op.remote_node)
6483

    
6484
    return (env, nl, nl)
6485

    
6486

    
6487
class TLReplaceDisks(Tasklet):
6488
  """Replaces disks for an instance.
6489

6490
  Note: Locking is not within the scope of this class.
6491

6492
  """
6493
  def __init__(self, lu, instance_name, mode, iallocator_name, remote_node,
6494
               disks):
6495
    """Initializes this class.
6496

6497
    """
6498
    Tasklet.__init__(self, lu)
6499

    
6500
    # Parameters
6501
    self.instance_name = instance_name
6502
    self.mode = mode
6503
    self.iallocator_name = iallocator_name
6504
    self.remote_node = remote_node
6505
    self.disks = disks
6506

    
6507
    # Runtime data
6508
    self.instance = None
6509
    self.new_node = None
6510
    self.target_node = None
6511
    self.other_node = None
6512
    self.remote_node_info = None
6513
    self.node_secondary_ip = None
6514

    
6515
  @staticmethod
6516
  def CheckArguments(mode, remote_node, iallocator):
6517
    """Helper function for users of this class.
6518

6519
    """
6520
    # check for valid parameter combination
6521
    if mode == constants.REPLACE_DISK_CHG:
6522
      if remote_node is None and iallocator is None:
6523
        raise errors.OpPrereqError("When changing the secondary either an"
6524
                                   " iallocator script must be used or the"
6525
                                   " new node given", errors.ECODE_INVAL)
6526

    
6527
      if remote_node is not None and iallocator is not None:
6528
        raise errors.OpPrereqError("Give either the iallocator or the new"
6529
                                   " secondary, not both", errors.ECODE_INVAL)
6530

    
6531
    elif remote_node is not None or iallocator is not None:
6532
      # Not replacing the secondary
6533
      raise errors.OpPrereqError("The iallocator and new node options can"
6534
                                 " only be used when changing the"
6535
                                 " secondary node", errors.ECODE_INVAL)
6536

    
6537
  @staticmethod
6538
  def _RunAllocator(lu, iallocator_name, instance_name, relocate_from):
6539
    """Compute a new secondary node using an IAllocator.
6540

6541
    """
6542
    ial = IAllocator(lu.cfg, lu.rpc,
6543
                     mode=constants.IALLOCATOR_MODE_RELOC,
6544
                     name=instance_name,
6545
                     relocate_from=relocate_from)
6546

    
6547
    ial.Run(iallocator_name)
6548

    
6549
    if not ial.success:
6550
      raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
6551
                                 " %s" % (iallocator_name, ial.info),
6552
                                 errors.ECODE_NORES)
6553

    
6554
    if len(ial.nodes) != ial.required_nodes:
6555
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
6556
                                 " of nodes (%s), required %s" %
6557
                                 (iallocator_name,
6558
                                  len(ial.nodes), ial.required_nodes),
6559
                                 errors.ECODE_FAULT)
6560

    
6561
    remote_node_name = ial.nodes[0]
6562

    
6563
    lu.LogInfo("Selected new secondary for instance '%s': %s",
6564
               instance_name, remote_node_name)
6565

    
6566
    return remote_node_name
6567

    
6568
  def _FindFaultyDisks(self, node_name):
6569
    return _FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
6570
                                    node_name, True)
6571

    
6572
  def CheckPrereq(self):
6573
    """Check prerequisites.
6574

6575
    This checks that the instance is in the cluster.
6576

6577
    """
6578
    self.instance = instance = self.cfg.GetInstanceInfo(self.instance_name)
6579
    assert instance is not None, \
6580
      "Cannot retrieve locked instance %s" % self.instance_name
6581

    
6582
    if instance.disk_template != constants.DT_DRBD8:
6583
      raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
6584
                                 " instances", errors.ECODE_INVAL)
6585

    
6586
    if len(instance.secondary_nodes) != 1:
6587
      raise errors.OpPrereqError("The instance has a strange layout,"
6588
                                 " expected one secondary but found %d" %
6589
                                 len(instance.secondary_nodes),
6590
                                 errors.ECODE_FAULT)
6591

    
6592
    secondary_node = instance.secondary_nodes[0]
6593

    
6594
    if self.iallocator_name is None:
6595
      remote_node = self.remote_node
6596
    else:
6597
      remote_node = self._RunAllocator(self.lu, self.iallocator_name,
6598
                                       instance.name, instance.secondary_nodes)
6599

    
6600
    if remote_node is not None:
6601
      self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
6602
      assert self.remote_node_info is not None, \
6603
        "Cannot retrieve locked node %s" % remote_node
6604
    else:
6605
      self.remote_node_info = None
6606

    
6607
    if remote_node == self.instance.primary_node:
6608
      raise errors.OpPrereqError("The specified node is the primary node of"
6609
                                 " the instance.", errors.ECODE_INVAL)
6610

    
6611
    if remote_node == secondary_node:
6612
      raise errors.OpPrereqError("The specified node is already the"
6613
                                 " secondary node of the instance.",
6614
                                 errors.ECODE_INVAL)
6615

    
6616
    if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
6617
                                    constants.REPLACE_DISK_CHG):
6618
      raise errors.OpPrereqError("Cannot specify disks to be replaced",
6619
                                 errors.ECODE_INVAL)
6620

    
6621
    if self.mode == constants.REPLACE_DISK_AUTO:
6622
      faulty_primary = self._FindFaultyDisks(instance.primary_node)
6623
      faulty_secondary = self._FindFaultyDisks(secondary_node)
6624

    
6625
      if faulty_primary and faulty_secondary:
6626
        raise errors.OpPrereqError("Instance %s has faulty disks on more than"
6627
                                   " one node and can not be repaired"
6628
                                   " automatically" % self.instance_name,
6629
                                   errors.ECODE_STATE)
6630

    
6631
      if faulty_primary:
6632
        self.disks = faulty_primary
6633
        self.target_node = instance.primary_node
6634
        self.other_node = secondary_node
6635
        check_nodes = [self.target_node, self.other_node]
6636
      elif faulty_secondary:
6637
        self.disks = faulty_secondary
6638
        self.target_node = secondary_node
6639
        self.other_node = instance.primary_node
6640
        check_nodes = [self.target_node, self.other_node]
6641
      else:
6642
        self.disks = []
6643
        check_nodes = []
6644

    
6645
    else:
6646
      # Non-automatic modes
6647
      if self.mode == constants.REPLACE_DISK_PRI:
6648
        self.target_node = instance.primary_node
6649
        self.other_node = secondary_node
6650
        check_nodes = [self.target_node, self.other_node]
6651

    
6652
      elif self.mode == constants.REPLACE_DISK_SEC:
6653
        self.target_node = secondary_node
6654
        self.other_node = instance.primary_node
6655
        check_nodes = [self.target_node, self.other_node]
6656

    
6657
      elif self.mode == constants.REPLACE_DISK_CHG:
6658
        self.new_node = remote_node
6659
        self.other_node = instance.primary_node
6660
        self.target_node = secondary_node
6661
        check_nodes = [self.new_node, self.other_node]
6662

    
6663
        _CheckNodeNotDrained(self.lu, remote_node)
6664

    
6665
      else:
6666
        raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
6667
                                     self.mode)
6668

    
6669
      # If not specified all disks should be replaced
6670
      if not self.disks:
6671
        self.disks = range(len(self.instance.disks))
6672

    
6673
    for node in check_nodes:
6674
      _CheckNodeOnline(self.lu, node)
6675

    
6676
    # Check whether disks are valid
6677
    for disk_idx in self.disks:
6678
      instance.FindDisk(disk_idx)
6679

    
6680
    # Get secondary node IP addresses
6681
    node_2nd_ip = {}
6682

    
6683
    for node_name in [self.target_node, self.other_node, self.new_node]:
6684
      if node_name is not None:
6685
        node_2nd_ip[node_name] = self.cfg.GetNodeInfo(node_name).secondary_ip
6686

    
6687
    self.node_secondary_ip = node_2nd_ip
6688

    
6689
  def Exec(self, feedback_fn):
6690
    """Execute disk replacement.
6691

6692
    This dispatches the disk replacement to the appropriate handler.
6693

6694
    """
6695
    if not self.disks:
6696
      feedback_fn("No disks need replacement")
6697
      return
6698

    
6699
    feedback_fn("Replacing disk(s) %s for %s" %
6700
                (utils.CommaJoin(self.disks), self.instance.name))
6701

    
6702
    activate_disks = (not self.instance.admin_up)
6703

    
6704
    # Activate the instance disks if we're replacing them on a down instance
6705
    if activate_disks:
6706
      _StartInstanceDisks(self.lu, self.instance, True)
6707

    
6708
    try:
6709
      # Should we replace the secondary node?
6710
      if self.new_node is not None:
6711
        fn = self._ExecDrbd8Secondary
6712
      else:
6713
        fn = self._ExecDrbd8DiskOnly
6714

    
6715
      return fn(feedback_fn)
6716

    
6717
    finally:
6718
      # Deactivate the instance disks if we're replacing them on a
6719
      # down instance
6720
      if activate_disks:
6721
        _SafeShutdownInstanceDisks(self.lu, self.instance)
6722

    
6723
  def _CheckVolumeGroup(self, nodes):
6724
    self.lu.LogInfo("Checking volume groups")
6725

    
6726
    vgname = self.cfg.GetVGName()
6727

    
6728
    # Make sure volume group exists on all involved nodes
6729
    results = self.rpc.call_vg_list(nodes)
6730
    if not results:
6731
      raise errors.OpExecError("Can't list volume groups on the nodes")
6732

    
6733
    for node in nodes:
6734
      res = results[node]
6735
      res.Raise("Error checking node %s" % node)
6736
      if vgname not in res.payload:
6737
        raise errors.OpExecError("Volume group '%s' not found on node %s" %
6738
                                 (vgname, node))
6739

    
6740
  def _CheckDisksExistence(self, nodes):
6741
    # Check disk existence
6742
    for idx, dev in enumerate(self.instance.disks):
6743
      if idx not in self.disks:
6744
        continue
6745

    
6746
      for node in nodes:
6747
        self.lu.LogInfo("Checking disk/%d on %s" % (idx, node))
6748
        self.cfg.SetDiskID(dev, node)
6749

    
6750
        result = self.rpc.call_blockdev_find(node, dev)
6751

    
6752
        msg = result.fail_msg
6753
        if msg or not result.payload:
6754
          if not msg:
6755
            msg = "disk not found"
6756
          raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
6757
                                   (idx, node, msg))
6758

    
6759
  def _CheckDisksConsistency(self, node_name, on_primary, ldisk):
6760
    for idx, dev in enumerate(self.instance.disks):
6761
      if idx not in self.disks:
6762
        continue
6763

    
6764
      self.lu.LogInfo("Checking disk/%d consistency on node %s" %
6765
                      (idx, node_name))
6766

    
6767
      if not _CheckDiskConsistency(self.lu, dev, node_name, on_primary,
6768
                                   ldisk=ldisk):
6769
        raise errors.OpExecError("Node %s has degraded storage, unsafe to"
6770
                                 " replace disks for instance %s" %
6771
                                 (node_name, self.instance.name))
6772

    
6773
  def _CreateNewStorage(self, node_name):
6774
    vgname = self.cfg.GetVGName()
6775
    iv_names = {}
6776

    
6777
    for idx, dev in enumerate(self.instance.disks):
6778
      if idx not in self.disks:
6779
        continue
6780

    
6781
      self.lu.LogInfo("Adding storage on %s for disk/%d" % (node_name, idx))
6782

    
6783
      self.cfg.SetDiskID(dev, node_name)
6784

    
6785
      lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
6786
      names = _GenerateUniqueNames(self.lu, lv_names)
6787

    
6788
      lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
6789
                             logical_id=(vgname, names[0]))
6790
      lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
6791
                             logical_id=(vgname, names[1]))
6792

    
6793
      new_lvs = [lv_data, lv_meta]
6794
      old_lvs = dev.children
6795
      iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
6796

    
6797
      # we pass force_create=True to force the LVM creation
6798
      for new_lv in new_lvs:
6799
        _CreateBlockDev(self.lu, node_name, self.instance, new_lv, True,
6800
                        _GetInstanceInfoText(self.instance), False)
6801

    
6802
    return iv_names
6803

    
6804
  def _CheckDevices(self, node_name, iv_names):
6805
    for name, (dev, _, _) in iv_names.iteritems():
6806
      self.cfg.SetDiskID(dev, node_name)
6807

    
6808
      result = self.rpc.call_blockdev_find(node_name, dev)
6809

    
6810
      msg = result.fail_msg
6811
      if msg or not result.payload:
6812
        if not msg:
6813
          msg = "disk not found"
6814
        raise errors.OpExecError("Can't find DRBD device %s: %s" %
6815
                                 (name, msg))
6816

    
6817
      if result.payload.is_degraded:
6818
        raise errors.OpExecError("DRBD device %s is degraded!" % name)
6819

    
6820
  def _RemoveOldStorage(self, node_name, iv_names):
6821
    for name, (_, old_lvs, _) in iv_names.iteritems():
6822
      self.lu.LogInfo("Remove logical volumes for %s" % name)
6823

    
6824
      for lv in old_lvs:
6825
        self.cfg.SetDiskID(lv, node_name)
6826

    
6827
        msg = self.rpc.call_blockdev_remove(node_name, lv).fail_msg
6828
        if msg:
6829
          self.lu.LogWarning("Can't remove old LV: %s" % msg,
6830
                             hint="remove unused LVs manually")
6831

    
6832
  def _ExecDrbd8DiskOnly(self, feedback_fn):
6833
    """Replace a disk on the primary or secondary for DRBD 8.
6834

6835
    The algorithm for replace is quite complicated:
6836

6837
      1. for each disk to be replaced:
6838

6839
        1. create new LVs on the target node with unique names
6840
        1. detach old LVs from the drbd device
6841
        1. rename old LVs to name_replaced.<time_t>
6842
        1. rename new LVs to old LVs
6843
        1. attach the new LVs (with the old names now) to the drbd device
6844

6845
      1. wait for sync across all devices
6846

6847
      1. for each modified disk:
6848

6849
        1. remove old LVs (which have the name name_replaces.<time_t>)
6850

6851
    Failures are not very well handled.
6852

6853
    """
6854
    steps_total = 6
6855

    
6856
    # Step: check device activation
6857
    self.lu.LogStep(1, steps_total, "Check device existence")
6858
    self._CheckDisksExistence([self.other_node, self.target_node])
6859
    self._CheckVolumeGroup([self.target_node, self.other_node])
6860

    
6861
    # Step: check other node consistency
6862
    self.lu.LogStep(2, steps_total, "Check peer consistency")
6863
    self._CheckDisksConsistency(self.other_node,
6864
                                self.other_node == self.instance.primary_node,
6865
                                False)
6866

    
6867
    # Step: create new storage
6868
    self.lu.LogStep(3, steps_total, "Allocate new storage")
6869
    iv_names = self._CreateNewStorage(self.target_node)
6870

    
6871
    # Step: for each lv, detach+rename*2+attach
6872
    self.lu.LogStep(4, steps_total, "Changing drbd configuration")
6873
    for dev, old_lvs, new_lvs in iv_names.itervalues():
6874
      self.lu.LogInfo("Detaching %s drbd from local storage" % dev.iv_name)
6875

    
6876
      result = self.rpc.call_blockdev_removechildren(self.target_node, dev,
6877
                                                     old_lvs)
6878
      result.Raise("Can't detach drbd from local storage on node"
6879
                   " %s for device %s" % (self.target_node, dev.iv_name))
6880
      #dev.children = []
6881
      #cfg.Update(instance)
6882

    
6883
      # ok, we created the new LVs, so now we know we have the needed
6884
      # storage; as such, we proceed on the target node to rename
6885
      # old_lv to _old, and new_lv to old_lv; note that we rename LVs
6886
      # using the assumption that logical_id == physical_id (which in
6887
      # turn is the unique_id on that node)
6888

    
6889
      # FIXME(iustin): use a better name for the replaced LVs
6890
      temp_suffix = int(time.time())
6891
      ren_fn = lambda d, suff: (d.physical_id[0],
6892
                                d.physical_id[1] + "_replaced-%s" % suff)
6893

    
6894
      # Build the rename list based on what LVs exist on the node
6895
      rename_old_to_new = []
6896
      for to_ren in old_lvs:
6897
        result = self.rpc.call_blockdev_find(self.target_node, to_ren)
6898
        if not result.fail_msg and result.payload:
6899
          # device exists
6900
          rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
6901

    
6902
      self.lu.LogInfo("Renaming the old LVs on the target node")
6903
      result = self.rpc.call_blockdev_rename(self.target_node,
6904
                                             rename_old_to_new)
6905
      result.Raise("Can't rename old LVs on node %s" % self.target_node)
6906

    
6907
      # Now we rename the new LVs to the old LVs
6908
      self.lu.LogInfo("Renaming the new LVs on the target node")
6909
      rename_new_to_old = [(new, old.physical_id)
6910
                           for old, new in zip(old_lvs, new_lvs)]
6911
      result = self.rpc.call_blockdev_rename(self.target_node,
6912
                                             rename_new_to_old)
6913
      result.Raise("Can't rename new LVs on node %s" % self.target_node)
6914

    
6915
      for old, new in zip(old_lvs, new_lvs):
6916
        new.logical_id = old.logical_id
6917
        self.cfg.SetDiskID(new, self.target_node)
6918

    
6919
      for disk in old_lvs:
6920
        disk.logical_id = ren_fn(disk, temp_suffix)
6921
        self.cfg.SetDiskID(disk, self.target_node)
6922

    
6923
      # Now that the new lvs have the old name, we can add them to the device
6924
      self.lu.LogInfo("Adding new mirror component on %s" % self.target_node)
6925
      result = self.rpc.call_blockdev_addchildren(self.target_node, dev,
6926
                                                  new_lvs)
6927
      msg = result.fail_msg
6928
      if msg:
6929
        for new_lv in new_lvs:
6930
          msg2 = self.rpc.call_blockdev_remove(self.target_node,
6931
                                               new_lv).fail_msg
6932
          if msg2:
6933
            self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
6934
                               hint=("cleanup manually the unused logical"
6935
                                     "volumes"))
6936
        raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
6937

    
6938
      dev.children = new_lvs
6939

    
6940
      self.cfg.Update(self.instance, feedback_fn)
6941

    
6942
    # Wait for sync
6943
    # This can fail as the old devices are degraded and _WaitForSync
6944
    # does a combined result over all disks, so we don't check its return value
6945
    self.lu.LogStep(5, steps_total, "Sync devices")
6946
    _WaitForSync(self.lu, self.instance)
6947

    
6948
    # Check all devices manually
6949
    self._CheckDevices(self.instance.primary_node, iv_names)
6950

    
6951
    # Step: remove old storage
6952
    self.lu.LogStep(6, steps_total, "Removing old storage")
6953
    self._RemoveOldStorage(self.target_node, iv_names)
6954

    
6955
  def _ExecDrbd8Secondary(self, feedback_fn):
6956
    """Replace the secondary node for DRBD 8.
6957

6958
    The algorithm for replace is quite complicated:
6959
      - for all disks of the instance:
6960
        - create new LVs on the new node with same names
6961
        - shutdown the drbd device on the old secondary
6962
        - disconnect the drbd network on the primary
6963
        - create the drbd device on the new secondary
6964
        - network attach the drbd on the primary, using an artifice:
6965
          the drbd code for Attach() will connect to the network if it
6966
          finds a device which is connected to the good local disks but
6967
          not network enabled
6968
      - wait for sync across all devices
6969
      - remove all disks from the old secondary
6970

6971
    Failures are not very well handled.
6972

6973
    """
6974
    steps_total = 6
6975

    
6976
    # Step: check device activation
6977
    self.lu.LogStep(1, steps_total, "Check device existence")
6978
    self._CheckDisksExistence([self.instance.primary_node])
6979
    self._CheckVolumeGroup([self.instance.primary_node])
6980

    
6981
    # Step: check other node consistency
6982
    self.lu.LogStep(2, steps_total, "Check peer consistency")
6983
    self._CheckDisksConsistency(self.instance.primary_node, True, True)
6984

    
6985
    # Step: create new storage
6986
    self.lu.LogStep(3, steps_total, "Allocate new storage")
6987
    for idx, dev in enumerate(self.instance.disks):
6988
      self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
6989
                      (self.new_node, idx))
6990
      # we pass force_create=True to force LVM creation
6991
      for new_lv in dev.children:
6992
        _CreateBlockDev(self.lu, self.new_node, self.instance, new_lv, True,
6993
                        _GetInstanceInfoText(self.instance), False)
6994

    
6995
    # Step 4: dbrd minors and drbd setups changes
6996
    # after this, we must manually remove the drbd minors on both the
6997
    # error and the success paths
6998
    self.lu.LogStep(4, steps_total, "Changing drbd configuration")
6999
    minors = self.cfg.AllocateDRBDMinor([self.new_node
7000
                                         for dev in self.instance.disks],
7001
                                        self.instance.name)
7002
    logging.debug("Allocated minors %r", minors)
7003

    
7004
    iv_names = {}
7005
    for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
7006
      self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
7007
                      (self.new_node, idx))
7008
      # create new devices on new_node; note that we create two IDs:
7009
      # one without port, so the drbd will be activated without
7010
      # networking information on the new node at this stage, and one
7011
      # with network, for the latter activation in step 4
7012
      (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
7013
      if self.instance.primary_node == o_node1:
7014
        p_minor = o_minor1
7015
      else:
7016
        assert self.instance.primary_node == o_node2, "Three-node instance?"
7017
        p_minor = o_minor2
7018

    
7019
      new_alone_id = (self.instance.primary_node, self.new_node, None,
7020
                      p_minor, new_minor, o_secret)
7021
      new_net_id = (self.instance.primary_node, self.new_node, o_port,
7022
                    p_minor, new_minor, o_secret)
7023

    
7024
      iv_names[idx] = (dev, dev.children, new_net_id)
7025
      logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
7026
                    new_net_id)
7027
      new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
7028
                              logical_id=new_alone_id,
7029
                              children=dev.children,
7030
                              size=dev.size)
7031
      try:
7032
        _CreateSingleBlockDev(self.lu, self.new_node, self.instance, new_drbd,
7033
                              _GetInstanceInfoText(self.instance), False)
7034
      except errors.GenericError:
7035
        self.cfg.ReleaseDRBDMinors(self.instance.name)
7036
        raise
7037

    
7038
    # We have new devices, shutdown the drbd on the old secondary
7039
    for idx, dev in enumerate(self.instance.disks):
7040
      self.lu.LogInfo("Shutting down drbd for disk/%d on old node" % idx)
7041
      self.cfg.SetDiskID(dev, self.target_node)
7042
      msg = self.rpc.call_blockdev_shutdown(self.target_node, dev).fail_msg
7043
      if msg:
7044
        self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
7045
                           "node: %s" % (idx, msg),
7046
                           hint=("Please cleanup this device manually as"
7047
                                 " soon as possible"))
7048

    
7049
    self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
7050
    result = self.rpc.call_drbd_disconnect_net([self.instance.primary_node],
7051
                                               self.node_secondary_ip,
7052
                                               self.instance.disks)\
7053
                                              [self.instance.primary_node]
7054

    
7055
    msg = result.fail_msg
7056
    if msg:
7057
      # detaches didn't succeed (unlikely)
7058
      self.cfg.ReleaseDRBDMinors(self.instance.name)
7059
      raise errors.OpExecError("Can't detach the disks from the network on"
7060
                               " old node: %s" % (msg,))
7061

    
7062
    # if we managed to detach at least one, we update all the disks of
7063
    # the instance to point to the new secondary
7064
    self.lu.LogInfo("Updating instance configuration")
7065
    for dev, _, new_logical_id in iv_names.itervalues():
7066
      dev.logical_id = new_logical_id
7067
      self.cfg.SetDiskID(dev, self.instance.primary_node)
7068

    
7069
    self.cfg.Update(self.instance, feedback_fn)
7070

    
7071
    # and now perform the drbd attach
7072
    self.lu.LogInfo("Attaching primary drbds to new secondary"
7073
                    " (standalone => connected)")
7074
    result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
7075
                                            self.new_node],
7076
                                           self.node_secondary_ip,
7077
                                           self.instance.disks,
7078
                                           self.instance.name,
7079
                                           False)
7080
    for to_node, to_result in result.items():
7081
      msg = to_result.fail_msg
7082
      if msg:
7083
        self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
7084
                           to_node, msg,
7085
                           hint=("please do a gnt-instance info to see the"
7086
                                 " status of disks"))
7087

    
7088
    # Wait for sync
7089
    # This can fail as the old devices are degraded and _WaitForSync
7090
    # does a combined result over all disks, so we don't check its return value
7091
    self.lu.LogStep(5, steps_total, "Sync devices")
7092
    _WaitForSync(self.lu, self.instance)
7093

    
7094
    # Check all devices manually
7095
    self._CheckDevices(self.instance.primary_node, iv_names)
7096

    
7097
    # Step: remove old storage
7098
    self.lu.LogStep(6, steps_total, "Removing old storage")
7099
    self._RemoveOldStorage(self.target_node, iv_names)
7100

    
7101

    
7102
class LURepairNodeStorage(NoHooksLU):
7103
  """Repairs the volume group on a node.
7104

7105
  """
7106
  _OP_REQP = ["node_name"]
7107
  REQ_BGL = False
7108

    
7109
  def CheckArguments(self):
7110
    node_name = self.cfg.ExpandNodeName(self.op.node_name)
7111
    if node_name is None:
7112
      raise errors.OpPrereqError("Invalid node name '%s'" % self.op.node_name,
7113
                                 errors.ECODE_NOENT)
7114

    
7115
    self.op.node_name = node_name
7116

    
7117
  def ExpandNames(self):
7118
    self.needed_locks = {
7119
      locking.LEVEL_NODE: [self.op.node_name],
7120
      }
7121

    
7122
  def _CheckFaultyDisks(self, instance, node_name):
7123
    """Ensure faulty disks abort the opcode or at least warn."""
7124
    try:
7125
      if _FindFaultyInstanceDisks(self.cfg, self.rpc, instance,
7126
                                  node_name, True):
7127
        raise errors.OpPrereqError("Instance '%s' has faulty disks on"
7128
                                   " node '%s'" % (instance.name, node_name),
7129
                                   errors.ECODE_STATE)
7130
    except errors.OpPrereqError, err:
7131
      if self.op.ignore_consistency:
7132
        self.proc.LogWarning(str(err.args[0]))
7133
      else:
7134
        raise
7135

    
7136
  def CheckPrereq(self):
7137
    """Check prerequisites.
7138

7139
    """
7140
    storage_type = self.op.storage_type
7141

    
7142
    if (constants.SO_FIX_CONSISTENCY not in
7143
        constants.VALID_STORAGE_OPERATIONS.get(storage_type, [])):
7144
      raise errors.OpPrereqError("Storage units of type '%s' can not be"
7145
                                 " repaired" % storage_type,
7146
                                 errors.ECODE_INVAL)
7147

    
7148
    # Check whether any instance on this node has faulty disks
7149
    for inst in _GetNodeInstances(self.cfg, self.op.node_name):
7150
      if not inst.admin_up:
7151
        continue
7152
      check_nodes = set(inst.all_nodes)
7153
      check_nodes.discard(self.op.node_name)
7154
      for inst_node_name in check_nodes:
7155
        self._CheckFaultyDisks(inst, inst_node_name)
7156

    
7157
  def Exec(self, feedback_fn):
7158
    feedback_fn("Repairing storage unit '%s' on %s ..." %
7159
                (self.op.name, self.op.node_name))
7160

    
7161
    st_args = _GetStorageTypeArgs(self.cfg, self.op.storage_type)
7162
    result = self.rpc.call_storage_execute(self.op.node_name,
7163
                                           self.op.storage_type, st_args,
7164
                                           self.op.name,
7165
                                           constants.SO_FIX_CONSISTENCY)
7166
    result.Raise("Failed to repair storage unit '%s' on %s" %
7167
                 (self.op.name, self.op.node_name))
7168

    
7169

    
7170
class LUGrowDisk(LogicalUnit):
7171
  """Grow a disk of an instance.
7172

7173
  """
7174
  HPATH = "disk-grow"
7175
  HTYPE = constants.HTYPE_INSTANCE
7176
  _OP_REQP = ["instance_name", "disk", "amount", "wait_for_sync"]
7177
  REQ_BGL = False
7178

    
7179
  def ExpandNames(self):
7180
    self._ExpandAndLockInstance()
7181
    self.needed_locks[locking.LEVEL_NODE] = []
7182
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7183

    
7184
  def DeclareLocks(self, level):
7185
    if level == locking.LEVEL_NODE:
7186
      self._LockInstancesNodes()
7187

    
7188
  def BuildHooksEnv(self):
7189
    """Build hooks env.
7190

7191
    This runs on the master, the primary and all the secondaries.
7192

7193
    """
7194
    env = {
7195
      "DISK": self.op.disk,
7196
      "AMOUNT": self.op.amount,
7197
      }
7198
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
7199
    nl = [
7200
      self.cfg.GetMasterNode(),
7201
      self.instance.primary_node,
7202
      ]
7203
    return env, nl, nl
7204

    
7205
  def CheckPrereq(self):
7206
    """Check prerequisites.
7207

7208
    This checks that the instance is in the cluster.
7209

7210
    """
7211
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7212
    assert instance is not None, \
7213
      "Cannot retrieve locked instance %s" % self.op.instance_name
7214
    nodenames = list(instance.all_nodes)
7215
    for node in nodenames:
7216
      _CheckNodeOnline(self, node)
7217

    
7218

    
7219
    self.instance = instance
7220

    
7221
    if instance.disk_template not in (constants.DT_PLAIN, constants.DT_DRBD8):
7222
      raise errors.OpPrereqError("Instance's disk layout does not support"
7223
                                 " growing.", errors.ECODE_INVAL)
7224

    
7225
    self.disk = instance.FindDisk(self.op.disk)
7226

    
7227
    nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
7228
                                       instance.hypervisor)
7229
    for node in nodenames:
7230
      info = nodeinfo[node]
7231
      info.Raise("Cannot get current information from node %s" % node)
7232
      vg_free = info.payload.get('vg_free', None)
7233
      if not isinstance(vg_free, int):
7234
        raise errors.OpPrereqError("Can't compute free disk space on"
7235
                                   " node %s" % node, errors.ECODE_ENVIRON)
7236
      if self.op.amount > vg_free:
7237
        raise errors.OpPrereqError("Not enough disk space on target node %s:"
7238
                                   " %d MiB available, %d MiB required" %
7239
                                   (node, vg_free, self.op.amount),
7240
                                   errors.ECODE_NORES)
7241

    
7242
  def Exec(self, feedback_fn):
7243
    """Execute disk grow.
7244

7245
    """
7246
    instance = self.instance
7247
    disk = self.disk
7248
    for node in instance.all_nodes:
7249
      self.cfg.SetDiskID(disk, node)
7250
      result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
7251
      result.Raise("Grow request failed to node %s" % node)
7252

    
7253
      # TODO: Rewrite code to work properly
7254
      # DRBD goes into sync mode for a short amount of time after executing the
7255
      # "resize" command. DRBD 8.x below version 8.0.13 contains a bug whereby
7256
      # calling "resize" in sync mode fails. Sleeping for a short amount of
7257
      # time is a work-around.
7258
      time.sleep(5)
7259

    
7260
    disk.RecordGrow(self.op.amount)
7261
    self.cfg.Update(instance, feedback_fn)
7262
    if self.op.wait_for_sync:
7263
      disk_abort = not _WaitForSync(self, instance)
7264
      if disk_abort:
7265
        self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
7266
                             " status.\nPlease check the instance.")
7267

    
7268

    
7269
class LUQueryInstanceData(NoHooksLU):
7270
  """Query runtime instance data.
7271

7272
  """
7273
  _OP_REQP = ["instances", "static"]
7274
  REQ_BGL = False
7275

    
7276
  def ExpandNames(self):
7277
    self.needed_locks = {}
7278
    self.share_locks = dict.fromkeys(locking.LEVELS, 1)
7279

    
7280
    if not isinstance(self.op.instances, list):
7281
      raise errors.OpPrereqError("Invalid argument type 'instances'",
7282
                                 errors.ECODE_INVAL)
7283

    
7284
    if self.op.instances:
7285
      self.wanted_names = []
7286
      for name in self.op.instances:
7287
        full_name = self.cfg.ExpandInstanceName(name)
7288
        if full_name is None:
7289
          raise errors.OpPrereqError("Instance '%s' not known" % name,
7290
                                     errors.ECODE_NOENT)
7291
        self.wanted_names.append(full_name)
7292
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
7293
    else:
7294
      self.wanted_names = None
7295
      self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
7296

    
7297
    self.needed_locks[locking.LEVEL_NODE] = []
7298
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7299

    
7300
  def DeclareLocks(self, level):
7301
    if level == locking.LEVEL_NODE:
7302
      self._LockInstancesNodes()
7303

    
7304
  def CheckPrereq(self):
7305
    """Check prerequisites.
7306

7307
    This only checks the optional instance list against the existing names.
7308

7309
    """
7310
    if self.wanted_names is None:
7311
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
7312

    
7313
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
7314
                             in self.wanted_names]
7315
    return
7316

    
7317
  def _ComputeBlockdevStatus(self, node, instance_name, dev):
7318
    """Returns the status of a block device
7319

7320
    """
7321
    if self.op.static or not node:
7322
      return None
7323

    
7324
    self.cfg.SetDiskID(dev, node)
7325

    
7326
    result = self.rpc.call_blockdev_find(node, dev)
7327
    if result.offline:
7328
      return None
7329

    
7330
    result.Raise("Can't compute disk status for %s" % instance_name)
7331

    
7332
    status = result.payload
7333
    if status is None:
7334
      return None
7335

    
7336
    return (status.dev_path, status.major, status.minor,
7337
            status.sync_percent, status.estimated_time,
7338
            status.is_degraded, status.ldisk_status)
7339

    
7340
  def _ComputeDiskStatus(self, instance, snode, dev):
7341
    """Compute block device status.
7342

7343
    """
7344
    if dev.dev_type in constants.LDS_DRBD:
7345
      # we change the snode then (otherwise we use the one passed in)
7346
      if dev.logical_id[0] == instance.primary_node:
7347
        snode = dev.logical_id[1]
7348
      else:
7349
        snode = dev.logical_id[0]
7350

    
7351
    dev_pstatus = self._ComputeBlockdevStatus(instance.primary_node,
7352
                                              instance.name, dev)
7353
    dev_sstatus = self._ComputeBlockdevStatus(snode, instance.name, dev)
7354

    
7355
    if dev.children:
7356
      dev_children = [self._ComputeDiskStatus(instance, snode, child)
7357
                      for child in dev.children]
7358
    else:
7359
      dev_children = []
7360

    
7361
    data = {
7362
      "iv_name": dev.iv_name,
7363
      "dev_type": dev.dev_type,
7364
      "logical_id": dev.logical_id,
7365
      "physical_id": dev.physical_id,
7366
      "pstatus": dev_pstatus,
7367
      "sstatus": dev_sstatus,
7368
      "children": dev_children,
7369
      "mode": dev.mode,
7370
      "size": dev.size,
7371
      }
7372

    
7373
    return data
7374

    
7375
  def Exec(self, feedback_fn):
7376
    """Gather and return data"""
7377
    result = {}
7378

    
7379
    cluster = self.cfg.GetClusterInfo()
7380

    
7381
    for instance in self.wanted_instances:
7382
      if not self.op.static:
7383
        remote_info = self.rpc.call_instance_info(instance.primary_node,
7384
                                                  instance.name,
7385
                                                  instance.hypervisor)
7386
        remote_info.Raise("Error checking node %s" % instance.primary_node)
7387
        remote_info = remote_info.payload
7388
        if remote_info and "state" in remote_info:
7389
          remote_state = "up"
7390
        else:
7391
          remote_state = "down"
7392
      else:
7393
        remote_state = None
7394
      if instance.admin_up:
7395
        config_state = "up"
7396
      else:
7397
        config_state = "down"
7398

    
7399
      disks = [self._ComputeDiskStatus(instance, None, device)
7400
               for device in instance.disks]
7401

    
7402
      idict = {
7403
        "name": instance.name,
7404
        "config_state": config_state,
7405
        "run_state": remote_state,
7406
        "pnode": instance.primary_node,
7407
        "snodes": instance.secondary_nodes,
7408
        "os": instance.os,
7409
        # this happens to be the same format used for hooks
7410
        "nics": _NICListToTuple(self, instance.nics),
7411
        "disks": disks,
7412
        "hypervisor": instance.hypervisor,
7413
        "network_port": instance.network_port,
7414
        "hv_instance": instance.hvparams,
7415
        "hv_actual": cluster.FillHV(instance, skip_globals=True),
7416
        "be_instance": instance.beparams,
7417
        "be_actual": cluster.FillBE(instance),
7418
        "serial_no": instance.serial_no,
7419
        "mtime": instance.mtime,
7420
        "ctime": instance.ctime,
7421
        "uuid": instance.uuid,
7422
        }
7423

    
7424
      result[instance.name] = idict
7425

    
7426
    return result
7427

    
7428

    
7429
class LUSetInstanceParams(LogicalUnit):
7430
  """Modifies an instances's parameters.
7431

7432
  """
7433
  HPATH = "instance-modify"
7434
  HTYPE = constants.HTYPE_INSTANCE
7435
  _OP_REQP = ["instance_name"]
7436
  REQ_BGL = False
7437

    
7438
  def CheckArguments(self):
7439
    if not hasattr(self.op, 'nics'):
7440
      self.op.nics = []
7441
    if not hasattr(self.op, 'disks'):
7442
      self.op.disks = []
7443
    if not hasattr(self.op, 'beparams'):
7444
      self.op.beparams = {}
7445
    if not hasattr(self.op, 'hvparams'):
7446
      self.op.hvparams = {}
7447
    self.op.force = getattr(self.op, "force", False)
7448
    if not (self.op.nics or self.op.disks or
7449
            self.op.hvparams or self.op.beparams):
7450
      raise errors.OpPrereqError("No changes submitted", errors.ECODE_INVAL)
7451

    
7452
    if self.op.hvparams:
7453
      _CheckGlobalHvParams(self.op.hvparams)
7454

    
7455
    # Disk validation
7456
    disk_addremove = 0
7457
    for disk_op, disk_dict in self.op.disks:
7458
      if disk_op == constants.DDM_REMOVE:
7459
        disk_addremove += 1
7460
        continue
7461
      elif disk_op == constants.DDM_ADD:
7462
        disk_addremove += 1
7463
      else:
7464
        if not isinstance(disk_op, int):
7465
          raise errors.OpPrereqError("Invalid disk index", errors.ECODE_INVAL)
7466
        if not isinstance(disk_dict, dict):
7467
          msg = "Invalid disk value: expected dict, got '%s'" % disk_dict
7468
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
7469

    
7470
      if disk_op == constants.DDM_ADD:
7471
        mode = disk_dict.setdefault('mode', constants.DISK_RDWR)
7472
        if mode not in constants.DISK_ACCESS_SET:
7473
          raise errors.OpPrereqError("Invalid disk access mode '%s'" % mode,
7474
                                     errors.ECODE_INVAL)
7475
        size = disk_dict.get('size', None)
7476
        if size is None:
7477
          raise errors.OpPrereqError("Required disk parameter size missing",
7478
                                     errors.ECODE_INVAL)
7479
        try:
7480
          size = int(size)
7481
        except ValueError, err:
7482
          raise errors.OpPrereqError("Invalid disk size parameter: %s" %
7483
                                     str(err), errors.ECODE_INVAL)
7484
        disk_dict['size'] = size
7485
      else:
7486
        # modification of disk
7487
        if 'size' in disk_dict:
7488
          raise errors.OpPrereqError("Disk size change not possible, use"
7489
                                     " grow-disk", errors.ECODE_INVAL)
7490

    
7491
    if disk_addremove > 1:
7492
      raise errors.OpPrereqError("Only one disk add or remove operation"
7493
                                 " supported at a time", errors.ECODE_INVAL)
7494

    
7495
    # NIC validation
7496
    nic_addremove = 0
7497
    for nic_op, nic_dict in self.op.nics:
7498
      if nic_op == constants.DDM_REMOVE:
7499
        nic_addremove += 1
7500
        continue
7501
      elif nic_op == constants.DDM_ADD:
7502
        nic_addremove += 1
7503
      else:
7504
        if not isinstance(nic_op, int):
7505
          raise errors.OpPrereqError("Invalid nic index", errors.ECODE_INVAL)
7506
        if not isinstance(nic_dict, dict):
7507
          msg = "Invalid nic value: expected dict, got '%s'" % nic_dict
7508
          raise errors.OpPrereqError(msg, errors.ECODE_INVAL)
7509

    
7510
      # nic_dict should be a dict
7511
      nic_ip = nic_dict.get('ip', None)
7512
      if nic_ip is not None:
7513
        if nic_ip.lower() == constants.VALUE_NONE:
7514
          nic_dict['ip'] = None
7515
        else:
7516
          if not utils.IsValidIP(nic_ip):
7517
            raise errors.OpPrereqError("Invalid IP address '%s'" % nic_ip,
7518
                                       errors.ECODE_INVAL)
7519

    
7520
      nic_bridge = nic_dict.get('bridge', None)
7521
      nic_link = nic_dict.get('link', None)
7522
      if nic_bridge and nic_link:
7523
        raise errors.OpPrereqError("Cannot pass 'bridge' and 'link'"
7524
                                   " at the same time", errors.ECODE_INVAL)
7525
      elif nic_bridge and nic_bridge.lower() == constants.VALUE_NONE:
7526
        nic_dict['bridge'] = None
7527
      elif nic_link and nic_link.lower() == constants.VALUE_NONE:
7528
        nic_dict['link'] = None
7529

    
7530
      if nic_op == constants.DDM_ADD:
7531
        nic_mac = nic_dict.get('mac', None)
7532
        if nic_mac is None:
7533
          nic_dict['mac'] = constants.VALUE_AUTO
7534

    
7535
      if 'mac' in nic_dict:
7536
        nic_mac = nic_dict['mac']
7537
        if nic_mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7538
          nic_mac = utils.NormalizeAndValidateMac(nic_mac)
7539

    
7540
        if nic_op != constants.DDM_ADD and nic_mac == constants.VALUE_AUTO:
7541
          raise errors.OpPrereqError("'auto' is not a valid MAC address when"
7542
                                     " modifying an existing nic",
7543
                                     errors.ECODE_INVAL)
7544

    
7545
    if nic_addremove > 1:
7546
      raise errors.OpPrereqError("Only one NIC add or remove operation"
7547
                                 " supported at a time", errors.ECODE_INVAL)
7548

    
7549
  def ExpandNames(self):
7550
    self._ExpandAndLockInstance()
7551
    self.needed_locks[locking.LEVEL_NODE] = []
7552
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
7553

    
7554
  def DeclareLocks(self, level):
7555
    if level == locking.LEVEL_NODE:
7556
      self._LockInstancesNodes()
7557

    
7558
  def BuildHooksEnv(self):
7559
    """Build hooks env.
7560

7561
    This runs on the master, primary and secondaries.
7562

7563
    """
7564
    args = dict()
7565
    if constants.BE_MEMORY in self.be_new:
7566
      args['memory'] = self.be_new[constants.BE_MEMORY]
7567
    if constants.BE_VCPUS in self.be_new:
7568
      args['vcpus'] = self.be_new[constants.BE_VCPUS]
7569
    # TODO: export disk changes. Note: _BuildInstanceHookEnv* don't export disk
7570
    # information at all.
7571
    if self.op.nics:
7572
      args['nics'] = []
7573
      nic_override = dict(self.op.nics)
7574
      c_nicparams = self.cluster.nicparams[constants.PP_DEFAULT]
7575
      for idx, nic in enumerate(self.instance.nics):
7576
        if idx in nic_override:
7577
          this_nic_override = nic_override[idx]
7578
        else:
7579
          this_nic_override = {}
7580
        if 'ip' in this_nic_override:
7581
          ip = this_nic_override['ip']
7582
        else:
7583
          ip = nic.ip
7584
        if 'mac' in this_nic_override:
7585
          mac = this_nic_override['mac']
7586
        else:
7587
          mac = nic.mac
7588
        if idx in self.nic_pnew:
7589
          nicparams = self.nic_pnew[idx]
7590
        else:
7591
          nicparams = objects.FillDict(c_nicparams, nic.nicparams)
7592
        mode = nicparams[constants.NIC_MODE]
7593
        link = nicparams[constants.NIC_LINK]
7594
        args['nics'].append((ip, mac, mode, link))
7595
      if constants.DDM_ADD in nic_override:
7596
        ip = nic_override[constants.DDM_ADD].get('ip', None)
7597
        mac = nic_override[constants.DDM_ADD]['mac']
7598
        nicparams = self.nic_pnew[constants.DDM_ADD]
7599
        mode = nicparams[constants.NIC_MODE]
7600
        link = nicparams[constants.NIC_LINK]
7601
        args['nics'].append((ip, mac, mode, link))
7602
      elif constants.DDM_REMOVE in nic_override:
7603
        del args['nics'][-1]
7604

    
7605
    env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
7606
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
7607
    return env, nl, nl
7608

    
7609
  @staticmethod
7610
  def _GetUpdatedParams(old_params, update_dict,
7611
                        default_values, parameter_types):
7612
    """Return the new params dict for the given params.
7613

7614
    @type old_params: dict
7615
    @param old_params: old parameters
7616
    @type update_dict: dict
7617
    @param update_dict: dict containing new parameter values,
7618
                        or constants.VALUE_DEFAULT to reset the
7619
                        parameter to its default value
7620
    @type default_values: dict
7621
    @param default_values: default values for the filled parameters
7622
    @type parameter_types: dict
7623
    @param parameter_types: dict mapping target dict keys to types
7624
                            in constants.ENFORCEABLE_TYPES
7625
    @rtype: (dict, dict)
7626
    @return: (new_parameters, filled_parameters)
7627

7628
    """
7629
    params_copy = copy.deepcopy(old_params)
7630
    for key, val in update_dict.iteritems():
7631
      if val == constants.VALUE_DEFAULT:
7632
        try:
7633
          del params_copy[key]
7634
        except KeyError:
7635
          pass
7636
      else:
7637
        params_copy[key] = val
7638
    utils.ForceDictType(params_copy, parameter_types)
7639
    params_filled = objects.FillDict(default_values, params_copy)
7640
    return (params_copy, params_filled)
7641

    
7642
  def CheckPrereq(self):
7643
    """Check prerequisites.
7644

7645
    This only checks the instance list against the existing names.
7646

7647
    """
7648
    self.force = self.op.force
7649

    
7650
    # checking the new params on the primary/secondary nodes
7651

    
7652
    instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
7653
    cluster = self.cluster = self.cfg.GetClusterInfo()
7654
    assert self.instance is not None, \
7655
      "Cannot retrieve locked instance %s" % self.op.instance_name
7656
    pnode = instance.primary_node
7657
    nodelist = list(instance.all_nodes)
7658

    
7659
    # hvparams processing
7660
    if self.op.hvparams:
7661
      i_hvdict, hv_new = self._GetUpdatedParams(
7662
                             instance.hvparams, self.op.hvparams,
7663
                             cluster.hvparams[instance.hypervisor],
7664
                             constants.HVS_PARAMETER_TYPES)
7665
      # local check
7666
      hypervisor.GetHypervisor(
7667
        instance.hypervisor).CheckParameterSyntax(hv_new)
7668
      _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
7669
      self.hv_new = hv_new # the new actual values
7670
      self.hv_inst = i_hvdict # the new dict (without defaults)
7671
    else:
7672
      self.hv_new = self.hv_inst = {}
7673

    
7674
    # beparams processing
7675
    if self.op.beparams:
7676
      i_bedict, be_new = self._GetUpdatedParams(
7677
                             instance.beparams, self.op.beparams,
7678
                             cluster.beparams[constants.PP_DEFAULT],
7679
                             constants.BES_PARAMETER_TYPES)
7680
      self.be_new = be_new # the new actual values
7681
      self.be_inst = i_bedict # the new dict (without defaults)
7682
    else:
7683
      self.be_new = self.be_inst = {}
7684

    
7685
    self.warn = []
7686

    
7687
    if constants.BE_MEMORY in self.op.beparams and not self.force:
7688
      mem_check_list = [pnode]
7689
      if be_new[constants.BE_AUTO_BALANCE]:
7690
        # either we changed auto_balance to yes or it was from before
7691
        mem_check_list.extend(instance.secondary_nodes)
7692
      instance_info = self.rpc.call_instance_info(pnode, instance.name,
7693
                                                  instance.hypervisor)
7694
      nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
7695
                                         instance.hypervisor)
7696
      pninfo = nodeinfo[pnode]
7697
      msg = pninfo.fail_msg
7698
      if msg:
7699
        # Assume the primary node is unreachable and go ahead
7700
        self.warn.append("Can't get info from primary node %s: %s" %
7701
                         (pnode,  msg))
7702
      elif not isinstance(pninfo.payload.get('memory_free', None), int):
7703
        self.warn.append("Node data from primary node %s doesn't contain"
7704
                         " free memory information" % pnode)
7705
      elif instance_info.fail_msg:
7706
        self.warn.append("Can't get instance runtime information: %s" %
7707
                        instance_info.fail_msg)
7708
      else:
7709
        if instance_info.payload:
7710
          current_mem = int(instance_info.payload['memory'])
7711
        else:
7712
          # Assume instance not running
7713
          # (there is a slight race condition here, but it's not very probable,
7714
          # and we have no other way to check)
7715
          current_mem = 0
7716
        miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
7717
                    pninfo.payload['memory_free'])
7718
        if miss_mem > 0:
7719
          raise errors.OpPrereqError("This change will prevent the instance"
7720
                                     " from starting, due to %d MB of memory"
7721
                                     " missing on its primary node" % miss_mem,
7722
                                     errors.ECODE_NORES)
7723

    
7724
      if be_new[constants.BE_AUTO_BALANCE]:
7725
        for node, nres in nodeinfo.items():
7726
          if node not in instance.secondary_nodes:
7727
            continue
7728
          msg = nres.fail_msg
7729
          if msg:
7730
            self.warn.append("Can't get info from secondary node %s: %s" %
7731
                             (node, msg))
7732
          elif not isinstance(nres.payload.get('memory_free', None), int):
7733
            self.warn.append("Secondary node %s didn't return free"
7734
                             " memory information" % node)
7735
          elif be_new[constants.BE_MEMORY] > nres.payload['memory_free']:
7736
            self.warn.append("Not enough memory to failover instance to"
7737
                             " secondary node %s" % node)
7738

    
7739
    # NIC processing
7740
    self.nic_pnew = {}
7741
    self.nic_pinst = {}
7742
    for nic_op, nic_dict in self.op.nics:
7743
      if nic_op == constants.DDM_REMOVE:
7744
        if not instance.nics:
7745
          raise errors.OpPrereqError("Instance has no NICs, cannot remove",
7746
                                     errors.ECODE_INVAL)
7747
        continue
7748
      if nic_op != constants.DDM_ADD:
7749
        # an existing nic
7750
        if not instance.nics:
7751
          raise errors.OpPrereqError("Invalid NIC index %s, instance has"
7752
                                     " no NICs" % nic_op,
7753
                                     errors.ECODE_INVAL)
7754
        if nic_op < 0 or nic_op >= len(instance.nics):
7755
          raise errors.OpPrereqError("Invalid NIC index %s, valid values"
7756
                                     " are 0 to %d" %
7757
                                     (nic_op, len(instance.nics) - 1),
7758
                                     errors.ECODE_INVAL)
7759
        old_nic_params = instance.nics[nic_op].nicparams
7760
        old_nic_ip = instance.nics[nic_op].ip
7761
      else:
7762
        old_nic_params = {}
7763
        old_nic_ip = None
7764

    
7765
      update_params_dict = dict([(key, nic_dict[key])
7766
                                 for key in constants.NICS_PARAMETERS
7767
                                 if key in nic_dict])
7768

    
7769
      if 'bridge' in nic_dict:
7770
        update_params_dict[constants.NIC_LINK] = nic_dict['bridge']
7771

    
7772
      new_nic_params, new_filled_nic_params = \
7773
          self._GetUpdatedParams(old_nic_params, update_params_dict,
7774
                                 cluster.nicparams[constants.PP_DEFAULT],
7775
                                 constants.NICS_PARAMETER_TYPES)
7776
      objects.NIC.CheckParameterSyntax(new_filled_nic_params)
7777
      self.nic_pinst[nic_op] = new_nic_params
7778
      self.nic_pnew[nic_op] = new_filled_nic_params
7779
      new_nic_mode = new_filled_nic_params[constants.NIC_MODE]
7780

    
7781
      if new_nic_mode == constants.NIC_MODE_BRIDGED:
7782
        nic_bridge = new_filled_nic_params[constants.NIC_LINK]
7783
        msg = self.rpc.call_bridges_exist(pnode, [nic_bridge]).fail_msg
7784
        if msg:
7785
          msg = "Error checking bridges on node %s: %s" % (pnode, msg)
7786
          if self.force:
7787
            self.warn.append(msg)
7788
          else:
7789
            raise errors.OpPrereqError(msg, errors.ECODE_ENVIRON)
7790
      if new_nic_mode == constants.NIC_MODE_ROUTED:
7791
        if 'ip' in nic_dict:
7792
          nic_ip = nic_dict['ip']
7793
        else:
7794
          nic_ip = old_nic_ip
7795
        if nic_ip is None:
7796
          raise errors.OpPrereqError('Cannot set the nic ip to None'
7797
                                     ' on a routed nic', errors.ECODE_INVAL)
7798
      if 'mac' in nic_dict:
7799
        nic_mac = nic_dict['mac']
7800
        if nic_mac is None:
7801
          raise errors.OpPrereqError('Cannot set the nic mac to None',
7802
                                     errors.ECODE_INVAL)
7803
        elif nic_mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
7804
          # otherwise generate the mac
7805
          nic_dict['mac'] = self.cfg.GenerateMAC(self.proc.GetECId())
7806
        else:
7807
          # or validate/reserve the current one
7808
          try:
7809
            self.cfg.ReserveMAC(nic_mac, self.proc.GetECId())
7810
          except errors.ReservationError:
7811
            raise errors.OpPrereqError("MAC address %s already in use"
7812
                                       " in cluster" % nic_mac,
7813
                                       errors.ECODE_NOTUNIQUE)
7814

    
7815
    # DISK processing
7816
    if self.op.disks and instance.disk_template == constants.DT_DISKLESS:
7817
      raise errors.OpPrereqError("Disk operations not supported for"
7818
                                 " diskless instances",
7819
                                 errors.ECODE_INVAL)
7820
    for disk_op, _ in self.op.disks:
7821
      if disk_op == constants.DDM_REMOVE:
7822
        if len(instance.disks) == 1:
7823
          raise errors.OpPrereqError("Cannot remove the last disk of"
7824
                                     " an instance",
7825
                                     errors.ECODE_INVAL)
7826
        ins_l = self.rpc.call_instance_list([pnode], [instance.hypervisor])
7827
        ins_l = ins_l[pnode]
7828
        msg = ins_l.fail_msg
7829
        if msg:
7830
          raise errors.OpPrereqError("Can't contact node %s: %s" %
7831
                                     (pnode, msg), errors.ECODE_ENVIRON)
7832
        if instance.name in ins_l.payload:
7833
          raise errors.OpPrereqError("Instance is running, can't remove"
7834
                                     " disks.", errors.ECODE_STATE)
7835

    
7836
      if (disk_op == constants.DDM_ADD and
7837
          len(instance.nics) >= constants.MAX_DISKS):
7838
        raise errors.OpPrereqError("Instance has too many disks (%d), cannot"
7839
                                   " add more" % constants.MAX_DISKS,
7840
                                   errors.ECODE_STATE)
7841
      if disk_op not in (constants.DDM_ADD, constants.DDM_REMOVE):
7842
        # an existing disk
7843
        if disk_op < 0 or disk_op >= len(instance.disks):
7844
          raise errors.OpPrereqError("Invalid disk index %s, valid values"
7845
                                     " are 0 to %d" %
7846
                                     (disk_op, len(instance.disks)),
7847
                                     errors.ECODE_INVAL)
7848

    
7849
    return
7850

    
7851
  def Exec(self, feedback_fn):
7852
    """Modifies an instance.
7853

7854
    All parameters take effect only at the next restart of the instance.
7855

7856
    """
7857
    # Process here the warnings from CheckPrereq, as we don't have a
7858
    # feedback_fn there.
7859
    for warn in self.warn:
7860
      feedback_fn("WARNING: %s" % warn)
7861

    
7862
    result = []
7863
    instance = self.instance
7864
    # disk changes
7865
    for disk_op, disk_dict in self.op.disks:
7866
      if disk_op == constants.DDM_REMOVE:
7867
        # remove the last disk
7868
        device = instance.disks.pop()
7869
        device_idx = len(instance.disks)
7870
        for node, disk in device.ComputeNodeTree(instance.primary_node):
7871
          self.cfg.SetDiskID(disk, node)
7872
          msg = self.rpc.call_blockdev_remove(node, disk).fail_msg
7873
          if msg:
7874
            self.LogWarning("Could not remove disk/%d on node %s: %s,"
7875
                            " continuing anyway", device_idx, node, msg)
7876
        result.append(("disk/%d" % device_idx, "remove"))
7877
      elif disk_op == constants.DDM_ADD:
7878
        # add a new disk
7879
        if instance.disk_template == constants.DT_FILE:
7880
          file_driver, file_path = instance.disks[0].logical_id
7881
          file_path = os.path.dirname(file_path)
7882
        else:
7883
          file_driver = file_path = None
7884
        disk_idx_base = len(instance.disks)
7885
        new_disk = _GenerateDiskTemplate(self,
7886
                                         instance.disk_template,
7887
                                         instance.name, instance.primary_node,
7888
                                         instance.secondary_nodes,
7889
                                         [disk_dict],
7890
                                         file_path,
7891
                                         file_driver,
7892
                                         disk_idx_base)[0]
7893
        instance.disks.append(new_disk)
7894
        info = _GetInstanceInfoText(instance)
7895

    
7896
        logging.info("Creating volume %s for instance %s",
7897
                     new_disk.iv_name, instance.name)
7898
        # Note: this needs to be kept in sync with _CreateDisks
7899
        #HARDCODE
7900
        for node in instance.all_nodes:
7901
          f_create = node == instance.primary_node
7902
          try:
7903
            _CreateBlockDev(self, node, instance, new_disk,
7904
                            f_create, info, f_create)
7905
          except errors.OpExecError, err:
7906
            self.LogWarning("Failed to create volume %s (%s) on"
7907
                            " node %s: %s",
7908
                            new_disk.iv_name, new_disk, node, err)
7909
        result.append(("disk/%d" % disk_idx_base, "add:size=%s,mode=%s" %
7910
                       (new_disk.size, new_disk.mode)))
7911
      else:
7912
        # change a given disk
7913
        instance.disks[disk_op].mode = disk_dict['mode']
7914
        result.append(("disk.mode/%d" % disk_op, disk_dict['mode']))
7915
    # NIC changes
7916
    for nic_op, nic_dict in self.op.nics:
7917
      if nic_op == constants.DDM_REMOVE:
7918
        # remove the last nic
7919
        del instance.nics[-1]
7920
        result.append(("nic.%d" % len(instance.nics), "remove"))
7921
      elif nic_op == constants.DDM_ADD:
7922
        # mac and bridge should be set, by now
7923
        mac = nic_dict['mac']
7924
        ip = nic_dict.get('ip', None)
7925
        nicparams = self.nic_pinst[constants.DDM_ADD]
7926
        new_nic = objects.NIC(mac=mac, ip=ip, nicparams=nicparams)
7927
        instance.nics.append(new_nic)
7928
        result.append(("nic.%d" % (len(instance.nics) - 1),
7929
                       "add:mac=%s,ip=%s,mode=%s,link=%s" %
7930
                       (new_nic.mac, new_nic.ip,
7931
                        self.nic_pnew[constants.DDM_ADD][constants.NIC_MODE],
7932
                        self.nic_pnew[constants.DDM_ADD][constants.NIC_LINK]
7933
                       )))
7934
      else:
7935
        for key in 'mac', 'ip':
7936
          if key in nic_dict:
7937
            setattr(instance.nics[nic_op], key, nic_dict[key])
7938
        if nic_op in self.nic_pinst:
7939
          instance.nics[nic_op].nicparams = self.nic_pinst[nic_op]
7940
        for key, val in nic_dict.iteritems():
7941
          result.append(("nic.%s/%d" % (key, nic_op), val))
7942

    
7943
    # hvparams changes
7944
    if self.op.hvparams:
7945
      instance.hvparams = self.hv_inst
7946
      for key, val in self.op.hvparams.iteritems():
7947
        result.append(("hv/%s" % key, val))
7948

    
7949
    # beparams changes
7950
    if self.op.beparams:
7951
      instance.beparams = self.be_inst
7952
      for key, val in self.op.beparams.iteritems():
7953
        result.append(("be/%s" % key, val))
7954

    
7955
    self.cfg.Update(instance, feedback_fn)
7956

    
7957
    return result
7958

    
7959

    
7960
class LUQueryExports(NoHooksLU):
7961
  """Query the exports list
7962

7963
  """
7964
  _OP_REQP = ['nodes']
7965
  REQ_BGL = False
7966

    
7967
  def ExpandNames(self):
7968
    self.needed_locks = {}
7969
    self.share_locks[locking.LEVEL_NODE] = 1
7970
    if not self.op.nodes:
7971
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
7972
    else:
7973
      self.needed_locks[locking.LEVEL_NODE] = \
7974
        _GetWantedNodes(self, self.op.nodes)
7975

    
7976
  def CheckPrereq(self):
7977
    """Check prerequisites.
7978

7979
    """
7980
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
7981

    
7982
  def Exec(self, feedback_fn):
7983
    """Compute the list of all the exported system images.
7984

7985
    @rtype: dict
7986
    @return: a dictionary with the structure node->(export-list)
7987
        where export-list is a list of the instances exported on
7988
        that node.
7989

7990
    """
7991
    rpcresult = self.rpc.call_export_list(self.nodes)
7992
    result = {}
7993
    for node in rpcresult:
7994
      if rpcresult[node].fail_msg:
7995
        result[node] = False
7996
      else:
7997
        result[node] = rpcresult[node].payload
7998

    
7999
    return result
8000

    
8001

    
8002
class LUExportInstance(LogicalUnit):
8003
  """Export an instance to an image in the cluster.
8004

8005
  """
8006
  HPATH = "instance-export"
8007
  HTYPE = constants.HTYPE_INSTANCE
8008
  _OP_REQP = ["instance_name", "target_node", "shutdown"]
8009
  REQ_BGL = False
8010

    
8011
  def CheckArguments(self):
8012
    """Check the arguments.
8013

8014
    """
8015
    self.shutdown_timeout = getattr(self.op, "shutdown_timeout",
8016
                                    constants.DEFAULT_SHUTDOWN_TIMEOUT)
8017

    
8018
  def ExpandNames(self):
8019
    self._ExpandAndLockInstance()
8020
    # FIXME: lock only instance primary and destination node
8021
    #
8022
    # Sad but true, for now we have do lock all nodes, as we don't know where
8023
    # the previous export might be, and and in this LU we search for it and
8024
    # remove it from its current node. In the future we could fix this by:
8025
    #  - making a tasklet to search (share-lock all), then create the new one,
8026
    #    then one to remove, after
8027
    #  - removing the removal operation altogether
8028
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
8029

    
8030
  def DeclareLocks(self, level):
8031
    """Last minute lock declaration."""
8032
    # All nodes are locked anyway, so nothing to do here.
8033

    
8034
  def BuildHooksEnv(self):
8035
    """Build hooks env.
8036

8037
    This will run on the master, primary node and target node.
8038

8039
    """
8040
    env = {
8041
      "EXPORT_NODE": self.op.target_node,
8042
      "EXPORT_DO_SHUTDOWN": self.op.shutdown,
8043
      "SHUTDOWN_TIMEOUT": self.shutdown_timeout,
8044
      }
8045
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
8046
    nl = [self.cfg.GetMasterNode(), self.instance.primary_node,
8047
          self.op.target_node]
8048
    return env, nl, nl
8049

    
8050
  def CheckPrereq(self):
8051
    """Check prerequisites.
8052

8053
    This checks that the instance and node names are valid.
8054

8055
    """
8056
    instance_name = self.op.instance_name
8057
    self.instance = self.cfg.GetInstanceInfo(instance_name)
8058
    assert self.instance is not None, \
8059
          "Cannot retrieve locked instance %s" % self.op.instance_name
8060
    _CheckNodeOnline(self, self.instance.primary_node)
8061

    
8062
    self.dst_node = self.cfg.GetNodeInfo(
8063
      self.cfg.ExpandNodeName(self.op.target_node))
8064

    
8065
    if self.dst_node is None:
8066
      # This is wrong node name, not a non-locked node
8067
      raise errors.OpPrereqError("Wrong node name %s" % self.op.target_node,
8068
                                 errors.ECODE_NOENT)
8069
    _CheckNodeOnline(self, self.dst_node.name)
8070
    _CheckNodeNotDrained(self, self.dst_node.name)
8071

    
8072
    # instance disk type verification
8073
    for disk in self.instance.disks:
8074
      if disk.dev_type == constants.LD_FILE:
8075
        raise errors.OpPrereqError("Export not supported for instances with"
8076
                                   " file-based disks", errors.ECODE_INVAL)
8077

    
8078
  def Exec(self, feedback_fn):
8079
    """Export an instance to an image in the cluster.
8080

8081
    """
8082
    instance = self.instance
8083
    dst_node = self.dst_node
8084
    src_node = instance.primary_node
8085

    
8086
    if self.op.shutdown:
8087
      # shutdown the instance, but not the disks
8088
      feedback_fn("Shutting down instance %s" % instance.name)
8089
      result = self.rpc.call_instance_shutdown(src_node, instance,
8090
                                               self.shutdown_timeout)
8091
      result.Raise("Could not shutdown instance %s on"
8092
                   " node %s" % (instance.name, src_node))
8093

    
8094
    vgname = self.cfg.GetVGName()
8095

    
8096
    snap_disks = []
8097

    
8098
    # set the disks ID correctly since call_instance_start needs the
8099
    # correct drbd minor to create the symlinks
8100
    for disk in instance.disks:
8101
      self.cfg.SetDiskID(disk, src_node)
8102

    
8103
    activate_disks = (not instance.admin_up)
8104

    
8105
    if activate_disks:
8106
      # Activate the instance disks if we'exporting a stopped instance
8107
      feedback_fn("Activating disks for %s" % instance.name)
8108
      _StartInstanceDisks(self, instance, None)
8109

    
8110
    try:
8111
      # per-disk results
8112
      dresults = []
8113
      try:
8114
        for idx, disk in enumerate(instance.disks):
8115
          feedback_fn("Creating a snapshot of disk/%s on node %s" %
8116
                      (idx, src_node))
8117

    
8118
          # result.payload will be a snapshot of an lvm leaf of the one we
8119
          # passed
8120
          result = self.rpc.call_blockdev_snapshot(src_node, disk)
8121
          msg = result.fail_msg
8122
          if msg:
8123
            self.LogWarning("Could not snapshot disk/%s on node %s: %s",
8124
                            idx, src_node, msg)
8125
            snap_disks.append(False)
8126
          else:
8127
            disk_id = (vgname, result.payload)
8128
            new_dev = objects.Disk(dev_type=constants.LD_LV, size=disk.size,
8129
                                   logical_id=disk_id, physical_id=disk_id,
8130
                                   iv_name=disk.iv_name)
8131
            snap_disks.append(new_dev)
8132

    
8133
      finally:
8134
        if self.op.shutdown and instance.admin_up:
8135
          feedback_fn("Starting instance %s" % instance.name)
8136
          result = self.rpc.call_instance_start(src_node, instance, None, None)
8137
          msg = result.fail_msg
8138
          if msg:
8139
            _ShutdownInstanceDisks(self, instance)
8140
            raise errors.OpExecError("Could not start instance: %s" % msg)
8141

    
8142
      # TODO: check for size
8143

    
8144
      cluster_name = self.cfg.GetClusterName()
8145
      for idx, dev in enumerate(snap_disks):
8146
        feedback_fn("Exporting snapshot %s from %s to %s" %
8147
                    (idx, src_node, dst_node.name))
8148
        if dev:
8149
          result = self.rpc.call_snapshot_export(src_node, dev, dst_node.name,
8150
                                                 instance, cluster_name, idx)
8151
          msg = result.fail_msg
8152
          if msg:
8153
            self.LogWarning("Could not export disk/%s from node %s to"
8154
                            " node %s: %s", idx, src_node, dst_node.name, msg)
8155
            dresults.append(False)
8156
          else:
8157
            dresults.append(True)
8158
          msg = self.rpc.call_blockdev_remove(src_node, dev).fail_msg
8159
          if msg:
8160
            self.LogWarning("Could not remove snapshot for disk/%d from node"
8161
                            " %s: %s", idx, src_node, msg)
8162
        else:
8163
          dresults.append(False)
8164

    
8165
      feedback_fn("Finalizing export on %s" % dst_node.name)
8166
      result = self.rpc.call_finalize_export(dst_node.name, instance,
8167
                                             snap_disks)
8168
      fin_resu = True
8169
      msg = result.fail_msg
8170
      if msg:
8171
        self.LogWarning("Could not finalize export for instance %s"
8172
                        " on node %s: %s", instance.name, dst_node.name, msg)
8173
        fin_resu = False
8174

    
8175
    finally:
8176
      if activate_disks:
8177
        feedback_fn("Deactivating disks for %s" % instance.name)
8178
        _ShutdownInstanceDisks(self, instance)
8179

    
8180
    nodelist = self.cfg.GetNodeList()
8181
    nodelist.remove(dst_node.name)
8182

    
8183
    # on one-node clusters nodelist will be empty after the removal
8184
    # if we proceed the backup would be removed because OpQueryExports
8185
    # substitutes an empty list with the full cluster node list.
8186
    iname = instance.name
8187
    if nodelist:
8188
      feedback_fn("Removing old exports for instance %s" % iname)
8189
      exportlist = self.rpc.call_export_list(nodelist)
8190
      for node in exportlist:
8191
        if exportlist[node].fail_msg:
8192
          continue
8193
        if iname in exportlist[node].payload:
8194
          msg = self.rpc.call_export_remove(node, iname).fail_msg
8195
          if msg:
8196
            self.LogWarning("Could not remove older export for instance %s"
8197
                            " on node %s: %s", iname, node, msg)
8198
    return fin_resu, dresults
8199

    
8200

    
8201
class LURemoveExport(NoHooksLU):
8202
  """Remove exports related to the named instance.
8203

8204
  """
8205
  _OP_REQP = ["instance_name"]
8206
  REQ_BGL = False
8207

    
8208
  def ExpandNames(self):
8209
    self.needed_locks = {}
8210
    # We need all nodes to be locked in order for RemoveExport to work, but we
8211
    # don't need to lock the instance itself, as nothing will happen to it (and
8212
    # we can remove exports also for a removed instance)
8213
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
8214

    
8215
  def CheckPrereq(self):
8216
    """Check prerequisites.
8217
    """
8218
    pass
8219

    
8220
  def Exec(self, feedback_fn):
8221
    """Remove any export.
8222

8223
    """
8224
    instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
8225
    # If the instance was not found we'll try with the name that was passed in.
8226
    # This will only work if it was an FQDN, though.
8227
    fqdn_warn = False
8228
    if not instance_name:
8229
      fqdn_warn = True
8230
      instance_name = self.op.instance_name
8231

    
8232
    locked_nodes = self.acquired_locks[locking.LEVEL_NODE]
8233
    exportlist = self.rpc.call_export_list(locked_nodes)
8234
    found = False
8235
    for node in exportlist:
8236
      msg = exportlist[node].fail_msg
8237
      if msg:
8238
        self.LogWarning("Failed to query node %s (continuing): %s", node, msg)
8239
        continue
8240
      if instance_name in exportlist[node].payload:
8241
        found = True
8242
        result = self.rpc.call_export_remove(node, instance_name)
8243
        msg = result.fail_msg
8244
        if msg:
8245
          logging.error("Could not remove export for instance %s"
8246
                        " on node %s: %s", instance_name, node, msg)
8247

    
8248
    if fqdn_warn and not found:
8249
      feedback_fn("Export not found. If trying to remove an export belonging"
8250
                  " to a deleted instance please use its Fully Qualified"
8251
                  " Domain Name.")
8252

    
8253

    
8254
class TagsLU(NoHooksLU): # pylint: disable-msg=W0223
8255
  """Generic tags LU.
8256

8257
  This is an abstract class which is the parent of all the other tags LUs.
8258

8259
  """
8260

    
8261
  def ExpandNames(self):
8262
    self.needed_locks = {}
8263
    if self.op.kind == constants.TAG_NODE:
8264
      name = self.cfg.ExpandNodeName(self.op.name)
8265
      if name is None:
8266
        raise errors.OpPrereqError("Invalid node name (%s)" %
8267
                                   (self.op.name,), errors.ECODE_NOENT)
8268
      self.op.name = name
8269
      self.needed_locks[locking.LEVEL_NODE] = name
8270
    elif self.op.kind == constants.TAG_INSTANCE:
8271
      name = self.cfg.ExpandInstanceName(self.op.name)
8272
      if name is None:
8273
        raise errors.OpPrereqError("Invalid instance name (%s)" %
8274
                                   (self.op.name,), errors.ECODE_NOENT)
8275
      self.op.name = name
8276
      self.needed_locks[locking.LEVEL_INSTANCE] = name
8277

    
8278
  def CheckPrereq(self):
8279
    """Check prerequisites.
8280

8281
    """
8282
    if self.op.kind == constants.TAG_CLUSTER:
8283
      self.target = self.cfg.GetClusterInfo()
8284
    elif self.op.kind == constants.TAG_NODE:
8285
      self.target = self.cfg.GetNodeInfo(self.op.name)
8286
    elif self.op.kind == constants.TAG_INSTANCE:
8287
      self.target = self.cfg.GetInstanceInfo(self.op.name)
8288
    else:
8289
      raise errors.OpPrereqError("Wrong tag type requested (%s)" %
8290
                                 str(self.op.kind), errors.ECODE_INVAL)
8291

    
8292

    
8293
class LUGetTags(TagsLU):
8294
  """Returns the tags of a given object.
8295

8296
  """
8297
  _OP_REQP = ["kind", "name"]
8298
  REQ_BGL = False
8299

    
8300
  def Exec(self, feedback_fn):
8301
    """Returns the tag list.
8302

8303
    """
8304
    return list(self.target.GetTags())
8305

    
8306

    
8307
class LUSearchTags(NoHooksLU):
8308
  """Searches the tags for a given pattern.
8309

8310
  """
8311
  _OP_REQP = ["pattern"]
8312
  REQ_BGL = False
8313

    
8314
  def ExpandNames(self):
8315
    self.needed_locks = {}
8316

    
8317
  def CheckPrereq(self):
8318
    """Check prerequisites.
8319

8320
    This checks the pattern passed for validity by compiling it.
8321

8322
    """
8323
    try:
8324
      self.re = re.compile(self.op.pattern)
8325
    except re.error, err:
8326
      raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
8327
                                 (self.op.pattern, err), errors.ECODE_INVAL)
8328

    
8329
  def Exec(self, feedback_fn):
8330
    """Returns the tag list.
8331

8332
    """
8333
    cfg = self.cfg
8334
    tgts = [("/cluster", cfg.GetClusterInfo())]
8335
    ilist = cfg.GetAllInstancesInfo().values()
8336
    tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
8337
    nlist = cfg.GetAllNodesInfo().values()
8338
    tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
8339
    results = []
8340
    for path, target in tgts:
8341
      for tag in target.GetTags():
8342
        if self.re.search(tag):
8343
          results.append((path, tag))
8344
    return results
8345

    
8346

    
8347
class LUAddTags(TagsLU):
8348
  """Sets a tag on a given object.
8349

8350
  """
8351
  _OP_REQP = ["kind", "name", "tags"]
8352
  REQ_BGL = False
8353

    
8354
  def CheckPrereq(self):
8355
    """Check prerequisites.
8356

8357
    This checks the type and length of the tag name and value.
8358

8359
    """
8360
    TagsLU.CheckPrereq(self)
8361
    for tag in self.op.tags:
8362
      objects.TaggableObject.ValidateTag(tag)
8363

    
8364
  def Exec(self, feedback_fn):
8365
    """Sets the tag.
8366

8367
    """
8368
    try:
8369
      for tag in self.op.tags:
8370
        self.target.AddTag(tag)
8371
    except errors.TagError, err:
8372
      raise errors.OpExecError("Error while setting tag: %s" % str(err))
8373
    self.cfg.Update(self.target, feedback_fn)
8374

    
8375

    
8376
class LUDelTags(TagsLU):
8377
  """Delete a list of tags from a given object.
8378

8379
  """
8380
  _OP_REQP = ["kind", "name", "tags"]
8381
  REQ_BGL = False
8382

    
8383
  def CheckPrereq(self):
8384
    """Check prerequisites.
8385

8386
    This checks that we have the given tag.
8387

8388
    """
8389
    TagsLU.CheckPrereq(self)
8390
    for tag in self.op.tags:
8391
      objects.TaggableObject.ValidateTag(tag)
8392
    del_tags = frozenset(self.op.tags)
8393
    cur_tags = self.target.GetTags()
8394
    if not del_tags <= cur_tags:
8395
      diff_tags = del_tags - cur_tags
8396
      diff_names = ["'%s'" % tag for tag in diff_tags]
8397
      diff_names.sort()
8398
      raise errors.OpPrereqError("Tag(s) %s not found" %
8399
                                 (",".join(diff_names)), errors.ECODE_NOENT)
8400

    
8401
  def Exec(self, feedback_fn):
8402
    """Remove the tag from the object.
8403

8404
    """
8405
    for tag in self.op.tags:
8406
      self.target.RemoveTag(tag)
8407
    self.cfg.Update(self.target, feedback_fn)
8408

    
8409

    
8410
class LUTestDelay(NoHooksLU):
8411
  """Sleep for a specified amount of time.
8412

8413
  This LU sleeps on the master and/or nodes for a specified amount of
8414
  time.
8415

8416
  """
8417
  _OP_REQP = ["duration", "on_master", "on_nodes"]
8418
  REQ_BGL = False
8419

    
8420
  def ExpandNames(self):
8421
    """Expand names and set required locks.
8422

8423
    This expands the node list, if any.
8424

8425
    """
8426
    self.needed_locks = {}
8427
    if self.op.on_nodes:
8428
      # _GetWantedNodes can be used here, but is not always appropriate to use
8429
      # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
8430
      # more information.
8431
      self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
8432
      self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
8433

    
8434
  def CheckPrereq(self):
8435
    """Check prerequisites.
8436

8437
    """
8438

    
8439
  def Exec(self, feedback_fn):
8440
    """Do the actual sleep.
8441

8442
    """
8443
    if self.op.on_master:
8444
      if not utils.TestDelay(self.op.duration):
8445
        raise errors.OpExecError("Error during master delay test")
8446
    if self.op.on_nodes:
8447
      result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
8448
      for node, node_result in result.items():
8449
        node_result.Raise("Failure during rpc call to node %s" % node)
8450

    
8451

    
8452
class IAllocator(object):
8453
  """IAllocator framework.
8454

8455
  An IAllocator instance has three sets of attributes:
8456
    - cfg that is needed to query the cluster
8457
    - input data (all members of the _KEYS class attribute are required)
8458
    - four buffer attributes (in|out_data|text), that represent the
8459
      input (to the external script) in text and data structure format,
8460
      and the output from it, again in two formats
8461
    - the result variables from the script (success, info, nodes) for
8462
      easy usage
8463

8464
  """
8465
  # pylint: disable-msg=R0902
8466
  # lots of instance attributes
8467
  _ALLO_KEYS = [
8468
    "mem_size", "disks", "disk_template",
8469
    "os", "tags", "nics", "vcpus", "hypervisor",
8470
    ]
8471
  _RELO_KEYS = [
8472
    "relocate_from",
8473
    ]
8474

    
8475
  def __init__(self, cfg, rpc, mode, name, **kwargs):
8476
    self.cfg = cfg
8477
    self.rpc = rpc
8478
    # init buffer variables
8479
    self.in_text = self.out_text = self.in_data = self.out_data = None
8480
    # init all input fields so that pylint is happy
8481
    self.mode = mode
8482
    self.name = name
8483
    self.mem_size = self.disks = self.disk_template = None
8484
    self.os = self.tags = self.nics = self.vcpus = None
8485
    self.hypervisor = None
8486
    self.relocate_from = None
8487
    # computed fields
8488
    self.required_nodes = None
8489
    # init result fields
8490
    self.success = self.info = self.nodes = None
8491
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
8492
      keyset = self._ALLO_KEYS
8493
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
8494
      keyset = self._RELO_KEYS
8495
    else:
8496
      raise errors.ProgrammerError("Unknown mode '%s' passed to the"
8497
                                   " IAllocator" % self.mode)
8498
    for key in kwargs:
8499
      if key not in keyset:
8500
        raise errors.ProgrammerError("Invalid input parameter '%s' to"
8501
                                     " IAllocator" % key)
8502
      setattr(self, key, kwargs[key])
8503
    for key in keyset:
8504
      if key not in kwargs:
8505
        raise errors.ProgrammerError("Missing input parameter '%s' to"
8506
                                     " IAllocator" % key)
8507
    self._BuildInputData()
8508

    
8509
  def _ComputeClusterData(self):
8510
    """Compute the generic allocator input data.
8511

8512
    This is the data that is independent of the actual operation.
8513

8514
    """
8515
    cfg = self.cfg
8516
    cluster_info = cfg.GetClusterInfo()
8517
    # cluster data
8518
    data = {
8519
      "version": constants.IALLOCATOR_VERSION,
8520
      "cluster_name": cfg.GetClusterName(),
8521
      "cluster_tags": list(cluster_info.GetTags()),
8522
      "enabled_hypervisors": list(cluster_info.enabled_hypervisors),
8523
      # we don't have job IDs
8524
      }
8525
    iinfo = cfg.GetAllInstancesInfo().values()
8526
    i_list = [(inst, cluster_info.FillBE(inst)) for inst in iinfo]
8527

    
8528
    # node data
8529
    node_results = {}
8530
    node_list = cfg.GetNodeList()
8531

    
8532
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
8533
      hypervisor_name = self.hypervisor
8534
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
8535
      hypervisor_name = cfg.GetInstanceInfo(self.name).hypervisor
8536

    
8537
    node_data = self.rpc.call_node_info(node_list, cfg.GetVGName(),
8538
                                        hypervisor_name)
8539
    node_iinfo = \
8540
      self.rpc.call_all_instances_info(node_list,
8541
                                       cluster_info.enabled_hypervisors)
8542
    for nname, nresult in node_data.items():
8543
      # first fill in static (config-based) values
8544
      ninfo = cfg.GetNodeInfo(nname)
8545
      pnr = {
8546
        "tags": list(ninfo.GetTags()),
8547
        "primary_ip": ninfo.primary_ip,
8548
        "secondary_ip": ninfo.secondary_ip,
8549
        "offline": ninfo.offline,
8550
        "drained": ninfo.drained,
8551
        "master_candidate": ninfo.master_candidate,
8552
        }
8553

    
8554
      if not (ninfo.offline or ninfo.drained):
8555
        nresult.Raise("Can't get data for node %s" % nname)
8556
        node_iinfo[nname].Raise("Can't get node instance info from node %s" %
8557
                                nname)
8558
        remote_info = nresult.payload
8559

    
8560
        for attr in ['memory_total', 'memory_free', 'memory_dom0',
8561
                     'vg_size', 'vg_free', 'cpu_total']:
8562
          if attr not in remote_info:
8563
            raise errors.OpExecError("Node '%s' didn't return attribute"
8564
                                     " '%s'" % (nname, attr))
8565
          if not isinstance(remote_info[attr], int):
8566
            raise errors.OpExecError("Node '%s' returned invalid value"
8567
                                     " for '%s': %s" %
8568
                                     (nname, attr, remote_info[attr]))
8569
        # compute memory used by primary instances
8570
        i_p_mem = i_p_up_mem = 0
8571
        for iinfo, beinfo in i_list:
8572
          if iinfo.primary_node == nname:
8573
            i_p_mem += beinfo[constants.BE_MEMORY]
8574
            if iinfo.name not in node_iinfo[nname].payload:
8575
              i_used_mem = 0
8576
            else:
8577
              i_used_mem = int(node_iinfo[nname].payload[iinfo.name]['memory'])
8578
            i_mem_diff = beinfo[constants.BE_MEMORY] - i_used_mem
8579
            remote_info['memory_free'] -= max(0, i_mem_diff)
8580

    
8581
            if iinfo.admin_up:
8582
              i_p_up_mem += beinfo[constants.BE_MEMORY]
8583

    
8584
        # compute memory used by instances
8585
        pnr_dyn = {
8586
          "total_memory": remote_info['memory_total'],
8587
          "reserved_memory": remote_info['memory_dom0'],
8588
          "free_memory": remote_info['memory_free'],
8589
          "total_disk": remote_info['vg_size'],
8590
          "free_disk": remote_info['vg_free'],
8591
          "total_cpus": remote_info['cpu_total'],
8592
          "i_pri_memory": i_p_mem,
8593
          "i_pri_up_memory": i_p_up_mem,
8594
          }
8595
        pnr.update(pnr_dyn)
8596

    
8597
      node_results[nname] = pnr
8598
    data["nodes"] = node_results
8599

    
8600
    # instance data
8601
    instance_data = {}
8602
    for iinfo, beinfo in i_list:
8603
      nic_data = []
8604
      for nic in iinfo.nics:
8605
        filled_params = objects.FillDict(
8606
            cluster_info.nicparams[constants.PP_DEFAULT],
8607
            nic.nicparams)
8608
        nic_dict = {"mac": nic.mac,
8609
                    "ip": nic.ip,
8610
                    "mode": filled_params[constants.NIC_MODE],
8611
                    "link": filled_params[constants.NIC_LINK],
8612
                   }
8613
        if filled_params[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
8614
          nic_dict["bridge"] = filled_params[constants.NIC_LINK]
8615
        nic_data.append(nic_dict)
8616
      pir = {
8617
        "tags": list(iinfo.GetTags()),
8618
        "admin_up": iinfo.admin_up,
8619
        "vcpus": beinfo[constants.BE_VCPUS],
8620
        "memory": beinfo[constants.BE_MEMORY],
8621
        "os": iinfo.os,
8622
        "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
8623
        "nics": nic_data,
8624
        "disks": [{"size": dsk.size, "mode": dsk.mode} for dsk in iinfo.disks],
8625
        "disk_template": iinfo.disk_template,
8626
        "hypervisor": iinfo.hypervisor,
8627
        }
8628
      pir["disk_space_total"] = _ComputeDiskSize(iinfo.disk_template,
8629
                                                 pir["disks"])
8630
      instance_data[iinfo.name] = pir
8631

    
8632
    data["instances"] = instance_data
8633

    
8634
    self.in_data = data
8635

    
8636
  def _AddNewInstance(self):
8637
    """Add new instance data to allocator structure.
8638

8639
    This in combination with _AllocatorGetClusterData will create the
8640
    correct structure needed as input for the allocator.
8641

8642
    The checks for the completeness of the opcode must have already been
8643
    done.
8644

8645
    """
8646
    data = self.in_data
8647

    
8648
    disk_space = _ComputeDiskSize(self.disk_template, self.disks)
8649

    
8650
    if self.disk_template in constants.DTS_NET_MIRROR:
8651
      self.required_nodes = 2
8652
    else:
8653
      self.required_nodes = 1
8654
    request = {
8655
      "type": "allocate",
8656
      "name": self.name,
8657
      "disk_template": self.disk_template,
8658
      "tags": self.tags,
8659
      "os": self.os,
8660
      "vcpus": self.vcpus,
8661
      "memory": self.mem_size,
8662
      "disks": self.disks,
8663
      "disk_space_total": disk_space,
8664
      "nics": self.nics,
8665
      "required_nodes": self.required_nodes,
8666
      }
8667
    data["request"] = request
8668

    
8669
  def _AddRelocateInstance(self):
8670
    """Add relocate instance data to allocator structure.
8671

8672
    This in combination with _IAllocatorGetClusterData will create the
8673
    correct structure needed as input for the allocator.
8674

8675
    The checks for the completeness of the opcode must have already been
8676
    done.
8677

8678
    """
8679
    instance = self.cfg.GetInstanceInfo(self.name)
8680
    if instance is None:
8681
      raise errors.ProgrammerError("Unknown instance '%s' passed to"
8682
                                   " IAllocator" % self.name)
8683

    
8684
    if instance.disk_template not in constants.DTS_NET_MIRROR:
8685
      raise errors.OpPrereqError("Can't relocate non-mirrored instances",
8686
                                 errors.ECODE_INVAL)
8687

    
8688
    if len(instance.secondary_nodes) != 1:
8689
      raise errors.OpPrereqError("Instance has not exactly one secondary node",
8690
                                 errors.ECODE_STATE)
8691

    
8692
    self.required_nodes = 1
8693
    disk_sizes = [{'size': disk.size} for disk in instance.disks]
8694
    disk_space = _ComputeDiskSize(instance.disk_template, disk_sizes)
8695

    
8696
    request = {
8697
      "type": "relocate",
8698
      "name": self.name,
8699
      "disk_space_total": disk_space,
8700
      "required_nodes": self.required_nodes,
8701
      "relocate_from": self.relocate_from,
8702
      }
8703
    self.in_data["request"] = request
8704

    
8705
  def _BuildInputData(self):
8706
    """Build input data structures.
8707

8708
    """
8709
    self._ComputeClusterData()
8710

    
8711
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
8712
      self._AddNewInstance()
8713
    else:
8714
      self._AddRelocateInstance()
8715

    
8716
    self.in_text = serializer.Dump(self.in_data)
8717

    
8718
  def Run(self, name, validate=True, call_fn=None):
8719
    """Run an instance allocator and return the results.
8720

8721
    """
8722
    if call_fn is None:
8723
      call_fn = self.rpc.call_iallocator_runner
8724

    
8725
    result = call_fn(self.cfg.GetMasterNode(), name, self.in_text)
8726
    result.Raise("Failure while running the iallocator script")
8727

    
8728
    self.out_text = result.payload
8729
    if validate:
8730
      self._ValidateResult()
8731

    
8732
  def _ValidateResult(self):
8733
    """Process the allocator results.
8734

8735
    This will process and if successful save the result in
8736
    self.out_data and the other parameters.
8737

8738
    """
8739
    try:
8740
      rdict = serializer.Load(self.out_text)
8741
    except Exception, err:
8742
      raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
8743

    
8744
    if not isinstance(rdict, dict):
8745
      raise errors.OpExecError("Can't parse iallocator results: not a dict")
8746

    
8747
    for key in "success", "info", "nodes":
8748
      if key not in rdict:
8749
        raise errors.OpExecError("Can't parse iallocator results:"
8750
                                 " missing key '%s'" % key)
8751
      setattr(self, key, rdict[key])
8752

    
8753
    if not isinstance(rdict["nodes"], list):
8754
      raise errors.OpExecError("Can't parse iallocator results: 'nodes' key"
8755
                               " is not a list")
8756
    self.out_data = rdict
8757

    
8758

    
8759
class LUTestAllocator(NoHooksLU):
8760
  """Run allocator tests.
8761

8762
  This LU runs the allocator tests
8763

8764
  """
8765
  _OP_REQP = ["direction", "mode", "name"]
8766

    
8767
  def CheckPrereq(self):
8768
    """Check prerequisites.
8769

8770
    This checks the opcode parameters depending on the director and mode test.
8771

8772
    """
8773
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
8774
      for attr in ["name", "mem_size", "disks", "disk_template",
8775
                   "os", "tags", "nics", "vcpus"]:
8776
        if not hasattr(self.op, attr):
8777
          raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
8778
                                     attr, errors.ECODE_INVAL)
8779
      iname = self.cfg.ExpandInstanceName(self.op.name)
8780
      if iname is not None:
8781
        raise errors.OpPrereqError("Instance '%s' already in the cluster" %
8782
                                   iname, errors.ECODE_EXISTS)
8783
      if not isinstance(self.op.nics, list):
8784
        raise errors.OpPrereqError("Invalid parameter 'nics'",
8785
                                   errors.ECODE_INVAL)
8786
      for row in self.op.nics:
8787
        if (not isinstance(row, dict) or
8788
            "mac" not in row or
8789
            "ip" not in row or
8790
            "bridge" not in row):
8791
          raise errors.OpPrereqError("Invalid contents of the 'nics'"
8792
                                     " parameter", errors.ECODE_INVAL)
8793
      if not isinstance(self.op.disks, list):
8794
        raise errors.OpPrereqError("Invalid parameter 'disks'",
8795
                                   errors.ECODE_INVAL)
8796
      for row in self.op.disks:
8797
        if (not isinstance(row, dict) or
8798
            "size" not in row or
8799
            not isinstance(row["size"], int) or
8800
            "mode" not in row or
8801
            row["mode"] not in ['r', 'w']):
8802
          raise errors.OpPrereqError("Invalid contents of the 'disks'"
8803
                                     " parameter", errors.ECODE_INVAL)
8804
      if not hasattr(self.op, "hypervisor") or self.op.hypervisor is None:
8805
        self.op.hypervisor = self.cfg.GetHypervisorType()
8806
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
8807
      if not hasattr(self.op, "name"):
8808
        raise errors.OpPrereqError("Missing attribute 'name' on opcode input",
8809
                                   errors.ECODE_INVAL)
8810
      fname = self.cfg.ExpandInstanceName(self.op.name)
8811
      if fname is None:
8812
        raise errors.OpPrereqError("Instance '%s' not found for relocation" %
8813
                                   self.op.name, errors.ECODE_NOENT)
8814
      self.op.name = fname
8815
      self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
8816
    else:
8817
      raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
8818
                                 self.op.mode, errors.ECODE_INVAL)
8819

    
8820
    if self.op.direction == constants.IALLOCATOR_DIR_OUT:
8821
      if not hasattr(self.op, "allocator") or self.op.allocator is None:
8822
        raise errors.OpPrereqError("Missing allocator name",
8823
                                   errors.ECODE_INVAL)
8824
    elif self.op.direction != constants.IALLOCATOR_DIR_IN:
8825
      raise errors.OpPrereqError("Wrong allocator test '%s'" %
8826
                                 self.op.direction, errors.ECODE_INVAL)
8827

    
8828
  def Exec(self, feedback_fn):
8829
    """Run the allocator test.
8830

8831
    """
8832
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
8833
      ial = IAllocator(self.cfg, self.rpc,
8834
                       mode=self.op.mode,
8835
                       name=self.op.name,
8836
                       mem_size=self.op.mem_size,
8837
                       disks=self.op.disks,
8838
                       disk_template=self.op.disk_template,
8839
                       os=self.op.os,
8840
                       tags=self.op.tags,
8841
                       nics=self.op.nics,
8842
                       vcpus=self.op.vcpus,
8843
                       hypervisor=self.op.hypervisor,
8844
                       )
8845
    else:
8846
      ial = IAllocator(self.cfg, self.rpc,
8847
                       mode=self.op.mode,
8848
                       name=self.op.name,
8849
                       relocate_from=list(self.relocate_from),
8850
                       )
8851

    
8852
    if self.op.direction == constants.IALLOCATOR_DIR_IN:
8853
      result = ial.in_text
8854
    else:
8855
      ial.Run(self.op.allocator, validate=False)
8856
      result = ial.out_text
8857
    return result