Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib.py @ 54155f52

History | View | Annotate | Download (194.6 kB)

1
#
2
#
3

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

    
21

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

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

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

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

    
47

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

51
  Subclasses must follow these rules:
52
    - implement ExpandNames
53
    - implement CheckPrereq
54
    - implement Exec
55
    - implement BuildHooksEnv
56
    - redefine HPATH and HTYPE
57
    - optionally redefine their run requirements:
58
        REQ_MASTER: the LU needs to run on the master node
59
        REQ_BGL: the LU needs to hold the Big Ganeti Lock exclusively
60

61
  Note that all commands require root permissions.
62

63
  """
64
  HPATH = None
65
  HTYPE = None
66
  _OP_REQP = []
67
  REQ_MASTER = True
68
  REQ_BGL = True
69

    
70
  def __init__(self, processor, op, context, rpc):
71
    """Constructor for LogicalUnit.
72

73
    This needs to be overriden in derived classes in order to check op
74
    validity.
75

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

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

    
101
    if not self.cfg.IsCluster():
102
      raise errors.OpPrereqError("Cluster not initialized yet,"
103
                                 " use 'gnt-cluster init' first.")
104
    if self.REQ_MASTER:
105
      master = self.cfg.GetMasterNode()
106
      if master != utils.HostInfo().name:
107
        raise errors.OpPrereqError("Commands must be run on the master"
108
                                   " node %s" % master)
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 ExpandNames(self):
121
    """Expand names for this LU.
122

123
    This method is called before starting to execute the opcode, and it should
124
    update all the parameters of the opcode to their canonical form (e.g. a
125
    short node name must be fully expanded after this method has successfully
126
    completed). This way locking, hooks, logging, ecc. can work correctly.
127

128
    LUs which implement this method must also populate the self.needed_locks
129
    member, as a dict with lock levels as keys, and a list of needed lock names
130
    as values. Rules:
131
      - Use an empty dict if you don't need any lock
132
      - If you don't need any lock at a particular level omit that level
133
      - Don't put anything for the BGL level
134
      - If you want all locks at a level use locking.ALL_SET as a value
135

136
    If you need to share locks (rather than acquire them exclusively) at one
137
    level you can modify self.share_locks, setting a true value (usually 1) for
138
    that level. By default locks are not shared.
139

140
    Examples:
141
    # Acquire all nodes and one instance
142
    self.needed_locks = {
143
      locking.LEVEL_NODE: locking.ALL_SET,
144
      locking.LEVEL_INSTANCE: ['instance1.example.tld'],
145
    }
146
    # Acquire just two nodes
147
    self.needed_locks = {
148
      locking.LEVEL_NODE: ['node1.example.tld', 'node2.example.tld'],
149
    }
150
    # Acquire no locks
151
    self.needed_locks = {} # No, you can't leave it to the default value None
152

153
    """
154
    # The implementation of this method is mandatory only if the new LU is
155
    # concurrent, so that old LUs don't need to be changed all at the same
156
    # time.
157
    if self.REQ_BGL:
158
      self.needed_locks = {} # Exclusive LUs don't need locks.
159
    else:
160
      raise NotImplementedError
161

    
162
  def DeclareLocks(self, level):
163
    """Declare LU locking needs for a level
164

165
    While most LUs can just declare their locking needs at ExpandNames time,
166
    sometimes there's the need to calculate some locks after having acquired
167
    the ones before. This function is called just before acquiring locks at a
168
    particular level, but after acquiring the ones at lower levels, and permits
169
    such calculations. It can be used to modify self.needed_locks, and by
170
    default it does nothing.
171

172
    This function is only called if you have something already set in
173
    self.needed_locks for the level.
174

175
    @param level: Locking level which is going to be locked
176
    @type level: member of ganeti.locking.LEVELS
177

178
    """
179

    
180
  def CheckPrereq(self):
181
    """Check prerequisites for this LU.
182

183
    This method should check that the prerequisites for the execution
184
    of this LU are fulfilled. It can do internode communication, but
185
    it should be idempotent - no cluster or system changes are
186
    allowed.
187

188
    The method should raise errors.OpPrereqError in case something is
189
    not fulfilled. Its return value is ignored.
190

191
    This method should also update all the parameters of the opcode to
192
    their canonical form if it hasn't been done by ExpandNames before.
193

194
    """
195
    raise NotImplementedError
196

    
197
  def Exec(self, feedback_fn):
198
    """Execute the LU.
199

200
    This method should implement the actual work. It should raise
201
    errors.OpExecError for failures that are somewhat dealt with in
202
    code, or expected.
203

204
    """
205
    raise NotImplementedError
206

    
207
  def BuildHooksEnv(self):
208
    """Build hooks environment for this LU.
209

210
    This method should return a three-node tuple consisting of: a dict
211
    containing the environment that will be used for running the
212
    specific hook for this LU, a list of node names on which the hook
213
    should run before the execution, and a list of node names on which
214
    the hook should run after the execution.
215

216
    The keys of the dict must not have 'GANETI_' prefixed as this will
217
    be handled in the hooks runner. Also note additional keys will be
218
    added by the hooks runner. If the LU doesn't define any
219
    environment, an empty dict (and not None) should be returned.
220

221
    No nodes should be returned as an empty list (and not None).
222

223
    Note that if the HPATH for a LU class is None, this function will
224
    not be called.
225

226
    """
227
    raise NotImplementedError
228

    
229
  def HooksCallBack(self, phase, hook_results, feedback_fn, lu_result):
230
    """Notify the LU about the results of its hooks.
231

232
    This method is called every time a hooks phase is executed, and notifies
233
    the Logical Unit about the hooks' result. The LU can then use it to alter
234
    its result based on the hooks.  By default the method does nothing and the
235
    previous result is passed back unchanged but any LU can define it if it
236
    wants to use the local cluster hook-scripts somehow.
237

238
    Args:
239
      phase: the hooks phase that has just been run
240
      hooks_results: the results of the multi-node hooks rpc call
241
      feedback_fn: function to send feedback back to the caller
242
      lu_result: the previous result this LU had, or None in the PRE phase.
243

244
    """
245
    return lu_result
246

    
247
  def _ExpandAndLockInstance(self):
248
    """Helper function to expand and lock an instance.
249

250
    Many LUs that work on an instance take its name in self.op.instance_name
251
    and need to expand it and then declare the expanded name for locking. This
252
    function does it, and then updates self.op.instance_name to the expanded
253
    name. It also initializes needed_locks as a dict, if this hasn't been done
254
    before.
255

256
    """
257
    if self.needed_locks is None:
258
      self.needed_locks = {}
259
    else:
260
      assert locking.LEVEL_INSTANCE not in self.needed_locks, \
261
        "_ExpandAndLockInstance called with instance-level locks set"
262
    expanded_name = self.cfg.ExpandInstanceName(self.op.instance_name)
263
    if expanded_name is None:
264
      raise errors.OpPrereqError("Instance '%s' not known" %
265
                                  self.op.instance_name)
266
    self.needed_locks[locking.LEVEL_INSTANCE] = expanded_name
267
    self.op.instance_name = expanded_name
268

    
269
  def _LockInstancesNodes(self, primary_only=False):
270
    """Helper function to declare instances' nodes for locking.
271

272
    This function should be called after locking one or more instances to lock
273
    their nodes. Its effect is populating self.needed_locks[locking.LEVEL_NODE]
274
    with all primary or secondary nodes for instances already locked and
275
    present in self.needed_locks[locking.LEVEL_INSTANCE].
276

277
    It should be called from DeclareLocks, and for safety only works if
278
    self.recalculate_locks[locking.LEVEL_NODE] is set.
279

280
    In the future it may grow parameters to just lock some instance's nodes, or
281
    to just lock primaries or secondary nodes, if needed.
282

283
    If should be called in DeclareLocks in a way similar to:
284

285
    if level == locking.LEVEL_NODE:
286
      self._LockInstancesNodes()
287

288
    @type primary_only: boolean
289
    @param primary_only: only lock primary nodes of locked instances
290

291
    """
292
    assert locking.LEVEL_NODE in self.recalculate_locks, \
293
      "_LockInstancesNodes helper function called with no nodes to recalculate"
294

    
295
    # TODO: check if we're really been called with the instance locks held
296

    
297
    # For now we'll replace self.needed_locks[locking.LEVEL_NODE], but in the
298
    # future we might want to have different behaviors depending on the value
299
    # of self.recalculate_locks[locking.LEVEL_NODE]
300
    wanted_nodes = []
301
    for instance_name in self.acquired_locks[locking.LEVEL_INSTANCE]:
302
      instance = self.context.cfg.GetInstanceInfo(instance_name)
303
      wanted_nodes.append(instance.primary_node)
304
      if not primary_only:
305
        wanted_nodes.extend(instance.secondary_nodes)
306

    
307
    if self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_REPLACE:
308
      self.needed_locks[locking.LEVEL_NODE] = wanted_nodes
309
    elif self.recalculate_locks[locking.LEVEL_NODE] == constants.LOCKS_APPEND:
310
      self.needed_locks[locking.LEVEL_NODE].extend(wanted_nodes)
311

    
312
    del self.recalculate_locks[locking.LEVEL_NODE]
313

    
314

    
315
class NoHooksLU(LogicalUnit):
316
  """Simple LU which runs no hooks.
317

318
  This LU is intended as a parent for other LogicalUnits which will
319
  run no hooks, in order to reduce duplicate code.
320

321
  """
322
  HPATH = None
323
  HTYPE = None
324

    
325

    
326
class _FieldSet(object):
327
  """A simple field set.
328

329
  Among the features are:
330
    - checking if a string is among a list of static string or regex objects
331
    - checking if a whole list of string matches
332
    - returning the matching groups from a regex match
333

334
  Internally, all fields are held as regular expression objects.
335

336
  """
337
  def __init__(self, *items):
338
    self.items = [re.compile("^%s$" % value) for value in items]
339

    
340
  def Extend(self, other_set):
341
    """Extend the field set with the items from another one"""
342
    self.items.extend(other_set.items)
343

    
344
  def Matches(self, field):
345
    """Checks if a field matches the current set
346

347
    @type field: str
348
    @param field: the string to match
349
    @return: either False or a regular expression match object
350

351
    """
352
    for m in itertools.ifilter(None, (val.match(field) for val in self.items)):
353
      return m
354
    return False
355

    
356
  def NonMatching(self, items):
357
    """Returns the list of fields not matching the current set
358

359
    @type items: list
360
    @param items: the list of fields to check
361
    @rtype: list
362
    @return: list of non-matching fields
363

364
    """
365
    return [val for val in items if not self.Matches(val)]
366

    
367

    
368
def _GetWantedNodes(lu, nodes):
369
  """Returns list of checked and expanded node names.
370

371
  Args:
372
    nodes: List of nodes (strings) or None for all
373

374
  """
375
  if not isinstance(nodes, list):
376
    raise errors.OpPrereqError("Invalid argument type 'nodes'")
377

    
378
  if not nodes:
379
    raise errors.ProgrammerError("_GetWantedNodes should only be called with a"
380
      " non-empty list of nodes whose name is to be expanded.")
381

    
382
  wanted = []
383
  for name in nodes:
384
    node = lu.cfg.ExpandNodeName(name)
385
    if node is None:
386
      raise errors.OpPrereqError("No such node name '%s'" % name)
387
    wanted.append(node)
388

    
389
  return utils.NiceSort(wanted)
390

    
391

    
392
def _GetWantedInstances(lu, instances):
393
  """Returns list of checked and expanded instance names.
394

395
  Args:
396
    instances: List of instances (strings) or None for all
397

398
  """
399
  if not isinstance(instances, list):
400
    raise errors.OpPrereqError("Invalid argument type 'instances'")
401

    
402
  if instances:
403
    wanted = []
404

    
405
    for name in instances:
406
      instance = lu.cfg.ExpandInstanceName(name)
407
      if instance is None:
408
        raise errors.OpPrereqError("No such instance name '%s'" % name)
409
      wanted.append(instance)
410

    
411
  else:
412
    wanted = lu.cfg.GetInstanceList()
413
  return utils.NiceSort(wanted)
414

    
415

    
416
def _CheckOutputFields(static, dynamic, selected):
417
  """Checks whether all selected fields are valid.
418

419
  @type static: L{_FieldSet}
420
  @param static: static fields set
421
  @type dynamic: L{_FieldSet}
422
  @param dynamic: dynamic fields set
423

424
  """
425
  f = _FieldSet()
426
  f.Extend(static)
427
  f.Extend(dynamic)
428

    
429
  delta = f.NonMatching(selected)
430
  if delta:
431
    raise errors.OpPrereqError("Unknown output fields selected: %s"
432
                               % ",".join(delta))
433

    
434

    
435
def _BuildInstanceHookEnv(name, primary_node, secondary_nodes, os_type, status,
436
                          memory, vcpus, nics):
437
  """Builds instance related env variables for hooks from single variables.
438

439
  Args:
440
    secondary_nodes: List of secondary nodes as strings
441
  """
442
  env = {
443
    "OP_TARGET": name,
444
    "INSTANCE_NAME": name,
445
    "INSTANCE_PRIMARY": primary_node,
446
    "INSTANCE_SECONDARIES": " ".join(secondary_nodes),
447
    "INSTANCE_OS_TYPE": os_type,
448
    "INSTANCE_STATUS": status,
449
    "INSTANCE_MEMORY": memory,
450
    "INSTANCE_VCPUS": vcpus,
451
  }
452

    
453
  if nics:
454
    nic_count = len(nics)
455
    for idx, (ip, bridge, mac) in enumerate(nics):
456
      if ip is None:
457
        ip = ""
458
      env["INSTANCE_NIC%d_IP" % idx] = ip
459
      env["INSTANCE_NIC%d_BRIDGE" % idx] = bridge
460
      env["INSTANCE_NIC%d_HWADDR" % idx] = mac
461
  else:
462
    nic_count = 0
463

    
464
  env["INSTANCE_NIC_COUNT"] = nic_count
465

    
466
  return env
467

    
468

    
469
def _BuildInstanceHookEnvByObject(lu, instance, override=None):
470
  """Builds instance related env variables for hooks from an object.
471

472
  Args:
473
    instance: objects.Instance object of instance
474
    override: dict of values to override
475
  """
476
  bep = lu.cfg.GetClusterInfo().FillBE(instance)
477
  args = {
478
    'name': instance.name,
479
    'primary_node': instance.primary_node,
480
    'secondary_nodes': instance.secondary_nodes,
481
    'os_type': instance.os,
482
    'status': instance.os,
483
    'memory': bep[constants.BE_MEMORY],
484
    'vcpus': bep[constants.BE_VCPUS],
485
    'nics': [(nic.ip, nic.bridge, nic.mac) for nic in instance.nics],
486
  }
487
  if override:
488
    args.update(override)
489
  return _BuildInstanceHookEnv(**args)
490

    
491

    
492
def _CheckInstanceBridgesExist(lu, instance):
493
  """Check that the brigdes needed by an instance exist.
494

495
  """
496
  # check bridges existance
497
  brlist = [nic.bridge for nic in instance.nics]
498
  if not lu.rpc.call_bridges_exist(instance.primary_node, brlist):
499
    raise errors.OpPrereqError("one or more target bridges %s does not"
500
                               " exist on destination node '%s'" %
501
                               (brlist, instance.primary_node))
502

    
503

    
504
class LUDestroyCluster(NoHooksLU):
505
  """Logical unit for destroying the cluster.
506

507
  """
508
  _OP_REQP = []
509

    
510
  def CheckPrereq(self):
511
    """Check prerequisites.
512

513
    This checks whether the cluster is empty.
514

515
    Any errors are signalled by raising errors.OpPrereqError.
516

517
    """
518
    master = self.cfg.GetMasterNode()
519

    
520
    nodelist = self.cfg.GetNodeList()
521
    if len(nodelist) != 1 or nodelist[0] != master:
522
      raise errors.OpPrereqError("There are still %d node(s) in"
523
                                 " this cluster." % (len(nodelist) - 1))
524
    instancelist = self.cfg.GetInstanceList()
525
    if instancelist:
526
      raise errors.OpPrereqError("There are still %d instance(s) in"
527
                                 " this cluster." % len(instancelist))
528

    
529
  def Exec(self, feedback_fn):
530
    """Destroys the cluster.
531

532
    """
533
    master = self.cfg.GetMasterNode()
534
    if not self.rpc.call_node_stop_master(master, False):
535
      raise errors.OpExecError("Could not disable the master role")
536
    priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
537
    utils.CreateBackup(priv_key)
538
    utils.CreateBackup(pub_key)
539
    return master
540

    
541

    
542
class LUVerifyCluster(LogicalUnit):
543
  """Verifies the cluster status.
544

545
  """
546
  HPATH = "cluster-verify"
547
  HTYPE = constants.HTYPE_CLUSTER
548
  _OP_REQP = ["skip_checks"]
549
  REQ_BGL = False
550

    
551
  def ExpandNames(self):
552
    self.needed_locks = {
553
      locking.LEVEL_NODE: locking.ALL_SET,
554
      locking.LEVEL_INSTANCE: locking.ALL_SET,
555
    }
556
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
557

    
558
  def _VerifyNode(self, node, file_list, local_cksum, vglist, node_result,
559
                  remote_version, feedback_fn):
560
    """Run multiple tests against a node.
561

562
    Test list:
563
      - compares ganeti version
564
      - checks vg existance and size > 20G
565
      - checks config file checksum
566
      - checks ssh to other nodes
567

568
    Args:
569
      node: name of the node to check
570
      file_list: required list of files
571
      local_cksum: dictionary of local files and their checksums
572

573
    """
574
    # compares ganeti version
575
    local_version = constants.PROTOCOL_VERSION
576
    if not remote_version:
577
      feedback_fn("  - ERROR: connection to %s failed" % (node))
578
      return True
579

    
580
    if local_version != remote_version:
581
      feedback_fn("  - ERROR: sw version mismatch: master %s, node(%s) %s" %
582
                      (local_version, node, remote_version))
583
      return True
584

    
585
    # checks vg existance and size > 20G
586

    
587
    bad = False
588
    if not vglist:
589
      feedback_fn("  - ERROR: unable to check volume groups on node %s." %
590
                      (node,))
591
      bad = True
592
    else:
593
      vgstatus = utils.CheckVolumeGroupSize(vglist, self.cfg.GetVGName(),
594
                                            constants.MIN_VG_SIZE)
595
      if vgstatus:
596
        feedback_fn("  - ERROR: %s on node %s" % (vgstatus, node))
597
        bad = True
598

    
599
    if not node_result:
600
      feedback_fn("  - ERROR: unable to verify node %s." % (node,))
601
      return True
602

    
603
    # checks config file checksum
604
    # checks ssh to any
605

    
606
    if 'filelist' not in node_result:
607
      bad = True
608
      feedback_fn("  - ERROR: node hasn't returned file checksum data")
609
    else:
610
      remote_cksum = node_result['filelist']
611
      for file_name in file_list:
612
        if file_name not in remote_cksum:
613
          bad = True
614
          feedback_fn("  - ERROR: file '%s' missing" % file_name)
615
        elif remote_cksum[file_name] != local_cksum[file_name]:
616
          bad = True
617
          feedback_fn("  - ERROR: file '%s' has wrong checksum" % file_name)
618

    
619
    if 'nodelist' not in node_result:
620
      bad = True
621
      feedback_fn("  - ERROR: node hasn't returned node ssh connectivity data")
622
    else:
623
      if node_result['nodelist']:
624
        bad = True
625
        for node in node_result['nodelist']:
626
          feedback_fn("  - ERROR: ssh communication with node '%s': %s" %
627
                          (node, node_result['nodelist'][node]))
628
    if 'node-net-test' not in node_result:
629
      bad = True
630
      feedback_fn("  - ERROR: node hasn't returned node tcp connectivity data")
631
    else:
632
      if node_result['node-net-test']:
633
        bad = True
634
        nlist = utils.NiceSort(node_result['node-net-test'].keys())
635
        for node in nlist:
636
          feedback_fn("  - ERROR: tcp communication with node '%s': %s" %
637
                          (node, node_result['node-net-test'][node]))
638

    
639
    hyp_result = node_result.get('hypervisor', None)
640
    if isinstance(hyp_result, dict):
641
      for hv_name, hv_result in hyp_result.iteritems():
642
        if hv_result is not None:
643
          feedback_fn("  - ERROR: hypervisor %s verify failure: '%s'" %
644
                      (hv_name, hv_result))
645
    return bad
646

    
647
  def _VerifyInstance(self, instance, instanceconfig, node_vol_is,
648
                      node_instance, feedback_fn):
649
    """Verify an instance.
650

651
    This function checks to see if the required block devices are
652
    available on the instance's node.
653

654
    """
655
    bad = False
656

    
657
    node_current = instanceconfig.primary_node
658

    
659
    node_vol_should = {}
660
    instanceconfig.MapLVsByNode(node_vol_should)
661

    
662
    for node in node_vol_should:
663
      for volume in node_vol_should[node]:
664
        if node not in node_vol_is or volume not in node_vol_is[node]:
665
          feedback_fn("  - ERROR: volume %s missing on node %s" %
666
                          (volume, node))
667
          bad = True
668

    
669
    if not instanceconfig.status == 'down':
670
      if (node_current not in node_instance or
671
          not instance in node_instance[node_current]):
672
        feedback_fn("  - ERROR: instance %s not running on node %s" %
673
                        (instance, node_current))
674
        bad = True
675

    
676
    for node in node_instance:
677
      if (not node == node_current):
678
        if instance in node_instance[node]:
679
          feedback_fn("  - ERROR: instance %s should not run on node %s" %
680
                          (instance, node))
681
          bad = True
682

    
683
    return bad
684

    
685
  def _VerifyOrphanVolumes(self, node_vol_should, node_vol_is, feedback_fn):
686
    """Verify if there are any unknown volumes in the cluster.
687

688
    The .os, .swap and backup volumes are ignored. All other volumes are
689
    reported as unknown.
690

691
    """
692
    bad = False
693

    
694
    for node in node_vol_is:
695
      for volume in node_vol_is[node]:
696
        if node not in node_vol_should or volume not in node_vol_should[node]:
697
          feedback_fn("  - ERROR: volume %s on node %s should not exist" %
698
                      (volume, node))
699
          bad = True
700
    return bad
701

    
702
  def _VerifyOrphanInstances(self, instancelist, node_instance, feedback_fn):
703
    """Verify the list of running instances.
704

705
    This checks what instances are running but unknown to the cluster.
706

707
    """
708
    bad = False
709
    for node in node_instance:
710
      for runninginstance in node_instance[node]:
711
        if runninginstance not in instancelist:
712
          feedback_fn("  - ERROR: instance %s on node %s should not exist" %
713
                          (runninginstance, node))
714
          bad = True
715
    return bad
716

    
717
  def _VerifyNPlusOneMemory(self, node_info, instance_cfg, feedback_fn):
718
    """Verify N+1 Memory Resilience.
719

720
    Check that if one single node dies we can still start all the instances it
721
    was primary for.
722

723
    """
724
    bad = False
725

    
726
    for node, nodeinfo in node_info.iteritems():
727
      # This code checks that every node which is now listed as secondary has
728
      # enough memory to host all instances it is supposed to should a single
729
      # other node in the cluster fail.
730
      # FIXME: not ready for failover to an arbitrary node
731
      # FIXME: does not support file-backed instances
732
      # WARNING: we currently take into account down instances as well as up
733
      # ones, considering that even if they're down someone might want to start
734
      # them even in the event of a node failure.
735
      for prinode, instances in nodeinfo['sinst-by-pnode'].iteritems():
736
        needed_mem = 0
737
        for instance in instances:
738
          bep = self.cfg.GetClusterInfo().FillBE(instance_cfg[instance])
739
          if bep[constants.BE_AUTO_BALANCE]:
740
            needed_mem += bep[constants.BE_MEMORY]
741
        if nodeinfo['mfree'] < needed_mem:
742
          feedback_fn("  - ERROR: not enough memory on node %s to accomodate"
743
                      " failovers should node %s fail" % (node, prinode))
744
          bad = True
745
    return bad
746

    
747
  def CheckPrereq(self):
748
    """Check prerequisites.
749

750
    Transform the list of checks we're going to skip into a set and check that
751
    all its members are valid.
752

753
    """
754
    self.skip_set = frozenset(self.op.skip_checks)
755
    if not constants.VERIFY_OPTIONAL_CHECKS.issuperset(self.skip_set):
756
      raise errors.OpPrereqError("Invalid checks to be skipped specified")
757

    
758
  def BuildHooksEnv(self):
759
    """Build hooks env.
760

761
    Cluster-Verify hooks just rone in the post phase and their failure makes
762
    the output be logged in the verify output and the verification to fail.
763

764
    """
765
    all_nodes = self.cfg.GetNodeList()
766
    # TODO: populate the environment with useful information for verify hooks
767
    env = {}
768
    return env, [], all_nodes
769

    
770
  def Exec(self, feedback_fn):
771
    """Verify integrity of cluster, performing various test on nodes.
772

773
    """
774
    bad = False
775
    feedback_fn("* Verifying global settings")
776
    for msg in self.cfg.VerifyConfig():
777
      feedback_fn("  - ERROR: %s" % msg)
778

    
779
    vg_name = self.cfg.GetVGName()
780
    hypervisors = self.cfg.GetClusterInfo().enabled_hypervisors
781
    nodelist = utils.NiceSort(self.cfg.GetNodeList())
782
    nodeinfo = [self.cfg.GetNodeInfo(nname) for nname in nodelist]
783
    instancelist = utils.NiceSort(self.cfg.GetInstanceList())
784
    i_non_redundant = [] # Non redundant instances
785
    i_non_a_balanced = [] # Non auto-balanced instances
786
    node_volume = {}
787
    node_instance = {}
788
    node_info = {}
789
    instance_cfg = {}
790

    
791
    # FIXME: verify OS list
792
    # do local checksums
793
    file_names = []
794
    file_names.append(constants.SSL_CERT_FILE)
795
    file_names.append(constants.CLUSTER_CONF_FILE)
796
    local_checksums = utils.FingerprintFiles(file_names)
797

    
798
    feedback_fn("* Gathering data (%d nodes)" % len(nodelist))
799
    all_volumeinfo = self.rpc.call_volume_list(nodelist, vg_name)
800
    all_instanceinfo = self.rpc.call_instance_list(nodelist, hypervisors)
801
    all_vglist = self.rpc.call_vg_list(nodelist)
802
    node_verify_param = {
803
      'filelist': file_names,
804
      'nodelist': nodelist,
805
      'hypervisor': hypervisors,
806
      'node-net-test': [(node.name, node.primary_ip, node.secondary_ip)
807
                        for node in nodeinfo]
808
      }
809
    all_nvinfo = self.rpc.call_node_verify(nodelist, node_verify_param,
810
                                           self.cfg.GetClusterName())
811
    all_rversion = self.rpc.call_version(nodelist)
812
    all_ninfo = self.rpc.call_node_info(nodelist, self.cfg.GetVGName(),
813
                                        self.cfg.GetHypervisorType())
814

    
815
    cluster = self.cfg.GetClusterInfo()
816
    for node in nodelist:
817
      feedback_fn("* Verifying node %s" % node)
818
      result = self._VerifyNode(node, file_names, local_checksums,
819
                                all_vglist[node], all_nvinfo[node],
820
                                all_rversion[node], feedback_fn)
821
      bad = bad or result
822

    
823
      # node_volume
824
      volumeinfo = all_volumeinfo[node]
825

    
826
      if isinstance(volumeinfo, basestring):
827
        feedback_fn("  - ERROR: LVM problem on node %s: %s" %
828
                    (node, volumeinfo[-400:].encode('string_escape')))
829
        bad = True
830
        node_volume[node] = {}
831
      elif not isinstance(volumeinfo, dict):
832
        feedback_fn("  - ERROR: connection to %s failed" % (node,))
833
        bad = True
834
        continue
835
      else:
836
        node_volume[node] = volumeinfo
837

    
838
      # node_instance
839
      nodeinstance = all_instanceinfo[node]
840
      if type(nodeinstance) != list:
841
        feedback_fn("  - ERROR: connection to %s failed" % (node,))
842
        bad = True
843
        continue
844

    
845
      node_instance[node] = nodeinstance
846

    
847
      # node_info
848
      nodeinfo = all_ninfo[node]
849
      if not isinstance(nodeinfo, dict):
850
        feedback_fn("  - ERROR: connection to %s failed" % (node,))
851
        bad = True
852
        continue
853

    
854
      try:
855
        node_info[node] = {
856
          "mfree": int(nodeinfo['memory_free']),
857
          "dfree": int(nodeinfo['vg_free']),
858
          "pinst": [],
859
          "sinst": [],
860
          # dictionary holding all instances this node is secondary for,
861
          # grouped by their primary node. Each key is a cluster node, and each
862
          # value is a list of instances which have the key as primary and the
863
          # current node as secondary.  this is handy to calculate N+1 memory
864
          # availability if you can only failover from a primary to its
865
          # secondary.
866
          "sinst-by-pnode": {},
867
        }
868
      except ValueError:
869
        feedback_fn("  - ERROR: invalid value returned from node %s" % (node,))
870
        bad = True
871
        continue
872

    
873
    node_vol_should = {}
874

    
875
    for instance in instancelist:
876
      feedback_fn("* Verifying instance %s" % instance)
877
      inst_config = self.cfg.GetInstanceInfo(instance)
878
      result =  self._VerifyInstance(instance, inst_config, node_volume,
879
                                     node_instance, feedback_fn)
880
      bad = bad or result
881

    
882
      inst_config.MapLVsByNode(node_vol_should)
883

    
884
      instance_cfg[instance] = inst_config
885

    
886
      pnode = inst_config.primary_node
887
      if pnode in node_info:
888
        node_info[pnode]['pinst'].append(instance)
889
      else:
890
        feedback_fn("  - ERROR: instance %s, connection to primary node"
891
                    " %s failed" % (instance, pnode))
892
        bad = True
893

    
894
      # If the instance is non-redundant we cannot survive losing its primary
895
      # node, so we are not N+1 compliant. On the other hand we have no disk
896
      # templates with more than one secondary so that situation is not well
897
      # supported either.
898
      # FIXME: does not support file-backed instances
899
      if len(inst_config.secondary_nodes) == 0:
900
        i_non_redundant.append(instance)
901
      elif len(inst_config.secondary_nodes) > 1:
902
        feedback_fn("  - WARNING: multiple secondaries for instance %s"
903
                    % instance)
904

    
905
      if not cluster.FillBE(inst_config)[constants.BE_AUTO_BALANCE]:
906
        i_non_a_balanced.append(instance)
907

    
908
      for snode in inst_config.secondary_nodes:
909
        if snode in node_info:
910
          node_info[snode]['sinst'].append(instance)
911
          if pnode not in node_info[snode]['sinst-by-pnode']:
912
            node_info[snode]['sinst-by-pnode'][pnode] = []
913
          node_info[snode]['sinst-by-pnode'][pnode].append(instance)
914
        else:
915
          feedback_fn("  - ERROR: instance %s, connection to secondary node"
916
                      " %s failed" % (instance, snode))
917

    
918
    feedback_fn("* Verifying orphan volumes")
919
    result = self._VerifyOrphanVolumes(node_vol_should, node_volume,
920
                                       feedback_fn)
921
    bad = bad or result
922

    
923
    feedback_fn("* Verifying remaining instances")
924
    result = self._VerifyOrphanInstances(instancelist, node_instance,
925
                                         feedback_fn)
926
    bad = bad or result
927

    
928
    if constants.VERIFY_NPLUSONE_MEM not in self.skip_set:
929
      feedback_fn("* Verifying N+1 Memory redundancy")
930
      result = self._VerifyNPlusOneMemory(node_info, instance_cfg, feedback_fn)
931
      bad = bad or result
932

    
933
    feedback_fn("* Other Notes")
934
    if i_non_redundant:
935
      feedback_fn("  - NOTICE: %d non-redundant instance(s) found."
936
                  % len(i_non_redundant))
937

    
938
    if i_non_a_balanced:
939
      feedback_fn("  - NOTICE: %d non-auto-balanced instance(s) found."
940
                  % len(i_non_a_balanced))
941

    
942
    return not bad
943

    
944
  def HooksCallBack(self, phase, hooks_results, feedback_fn, lu_result):
945
    """Analize the post-hooks' result, handle it, and send some
946
    nicely-formatted feedback back to the user.
947

948
    Args:
949
      phase: the hooks phase that has just been run
950
      hooks_results: the results of the multi-node hooks rpc call
951
      feedback_fn: function to send feedback back to the caller
952
      lu_result: previous Exec result
953

954
    """
955
    # We only really run POST phase hooks, and are only interested in
956
    # their results
957
    if phase == constants.HOOKS_PHASE_POST:
958
      # Used to change hooks' output to proper indentation
959
      indent_re = re.compile('^', re.M)
960
      feedback_fn("* Hooks Results")
961
      if not hooks_results:
962
        feedback_fn("  - ERROR: general communication failure")
963
        lu_result = 1
964
      else:
965
        for node_name in hooks_results:
966
          show_node_header = True
967
          res = hooks_results[node_name]
968
          if res is False or not isinstance(res, list):
969
            feedback_fn("    Communication failure")
970
            lu_result = 1
971
            continue
972
          for script, hkr, output in res:
973
            if hkr == constants.HKR_FAIL:
974
              # The node header is only shown once, if there are
975
              # failing hooks on that node
976
              if show_node_header:
977
                feedback_fn("  Node %s:" % node_name)
978
                show_node_header = False
979
              feedback_fn("    ERROR: Script %s failed, output:" % script)
980
              output = indent_re.sub('      ', output)
981
              feedback_fn("%s" % output)
982
              lu_result = 1
983

    
984
      return lu_result
985

    
986

    
987
class LUVerifyDisks(NoHooksLU):
988
  """Verifies the cluster disks status.
989

990
  """
991
  _OP_REQP = []
992
  REQ_BGL = False
993

    
994
  def ExpandNames(self):
995
    self.needed_locks = {
996
      locking.LEVEL_NODE: locking.ALL_SET,
997
      locking.LEVEL_INSTANCE: locking.ALL_SET,
998
    }
999
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
1000

    
1001
  def CheckPrereq(self):
1002
    """Check prerequisites.
1003

1004
    This has no prerequisites.
1005

1006
    """
1007
    pass
1008

    
1009
  def Exec(self, feedback_fn):
1010
    """Verify integrity of cluster disks.
1011

1012
    """
1013
    result = res_nodes, res_nlvm, res_instances, res_missing = [], {}, [], {}
1014

    
1015
    vg_name = self.cfg.GetVGName()
1016
    nodes = utils.NiceSort(self.cfg.GetNodeList())
1017
    instances = [self.cfg.GetInstanceInfo(name)
1018
                 for name in self.cfg.GetInstanceList()]
1019

    
1020
    nv_dict = {}
1021
    for inst in instances:
1022
      inst_lvs = {}
1023
      if (inst.status != "up" or
1024
          inst.disk_template not in constants.DTS_NET_MIRROR):
1025
        continue
1026
      inst.MapLVsByNode(inst_lvs)
1027
      # transform { iname: {node: [vol,],},} to {(node, vol): iname}
1028
      for node, vol_list in inst_lvs.iteritems():
1029
        for vol in vol_list:
1030
          nv_dict[(node, vol)] = inst
1031

    
1032
    if not nv_dict:
1033
      return result
1034

    
1035
    node_lvs = self.rpc.call_volume_list(nodes, vg_name)
1036

    
1037
    to_act = set()
1038
    for node in nodes:
1039
      # node_volume
1040
      lvs = node_lvs[node]
1041

    
1042
      if isinstance(lvs, basestring):
1043
        logging.warning("Error enumerating LVs on node %s: %s", node, lvs)
1044
        res_nlvm[node] = lvs
1045
      elif not isinstance(lvs, dict):
1046
        logging.warning("Connection to node %s failed or invalid data"
1047
                        " returned", node)
1048
        res_nodes.append(node)
1049
        continue
1050

    
1051
      for lv_name, (_, lv_inactive, lv_online) in lvs.iteritems():
1052
        inst = nv_dict.pop((node, lv_name), None)
1053
        if (not lv_online and inst is not None
1054
            and inst.name not in res_instances):
1055
          res_instances.append(inst.name)
1056

    
1057
    # any leftover items in nv_dict are missing LVs, let's arrange the
1058
    # data better
1059
    for key, inst in nv_dict.iteritems():
1060
      if inst.name not in res_missing:
1061
        res_missing[inst.name] = []
1062
      res_missing[inst.name].append(key)
1063

    
1064
    return result
1065

    
1066

    
1067
class LURenameCluster(LogicalUnit):
1068
  """Rename the cluster.
1069

1070
  """
1071
  HPATH = "cluster-rename"
1072
  HTYPE = constants.HTYPE_CLUSTER
1073
  _OP_REQP = ["name"]
1074

    
1075
  def BuildHooksEnv(self):
1076
    """Build hooks env.
1077

1078
    """
1079
    env = {
1080
      "OP_TARGET": self.cfg.GetClusterName(),
1081
      "NEW_NAME": self.op.name,
1082
      }
1083
    mn = self.cfg.GetMasterNode()
1084
    return env, [mn], [mn]
1085

    
1086
  def CheckPrereq(self):
1087
    """Verify that the passed name is a valid one.
1088

1089
    """
1090
    hostname = utils.HostInfo(self.op.name)
1091

    
1092
    new_name = hostname.name
1093
    self.ip = new_ip = hostname.ip
1094
    old_name = self.cfg.GetClusterName()
1095
    old_ip = self.cfg.GetMasterIP()
1096
    if new_name == old_name and new_ip == old_ip:
1097
      raise errors.OpPrereqError("Neither the name nor the IP address of the"
1098
                                 " cluster has changed")
1099
    if new_ip != old_ip:
1100
      if utils.TcpPing(new_ip, constants.DEFAULT_NODED_PORT):
1101
        raise errors.OpPrereqError("The given cluster IP address (%s) is"
1102
                                   " reachable on the network. Aborting." %
1103
                                   new_ip)
1104

    
1105
    self.op.name = new_name
1106

    
1107
  def Exec(self, feedback_fn):
1108
    """Rename the cluster.
1109

1110
    """
1111
    clustername = self.op.name
1112
    ip = self.ip
1113

    
1114
    # shutdown the master IP
1115
    master = self.cfg.GetMasterNode()
1116
    if not self.rpc.call_node_stop_master(master, False):
1117
      raise errors.OpExecError("Could not disable the master role")
1118

    
1119
    try:
1120
      # modify the sstore
1121
      # TODO: sstore
1122
      ss.SetKey(ss.SS_MASTER_IP, ip)
1123
      ss.SetKey(ss.SS_CLUSTER_NAME, clustername)
1124

    
1125
      # Distribute updated ss config to all nodes
1126
      myself = self.cfg.GetNodeInfo(master)
1127
      dist_nodes = self.cfg.GetNodeList()
1128
      if myself.name in dist_nodes:
1129
        dist_nodes.remove(myself.name)
1130

    
1131
      logging.debug("Copying updated ssconf data to all nodes")
1132
      for keyname in [ss.SS_CLUSTER_NAME, ss.SS_MASTER_IP]:
1133
        fname = ss.KeyToFilename(keyname)
1134
        result = self.rpc.call_upload_file(dist_nodes, fname)
1135
        for to_node in dist_nodes:
1136
          if not result[to_node]:
1137
            self.LogWarning("Copy of file %s to node %s failed",
1138
                            fname, to_node)
1139
    finally:
1140
      if not self.rpc.call_node_start_master(master, False):
1141
        self.LogWarning("Could not re-enable the master role on"
1142
                        " the master, please restart manually.")
1143

    
1144

    
1145
def _RecursiveCheckIfLVMBased(disk):
1146
  """Check if the given disk or its children are lvm-based.
1147

1148
  Args:
1149
    disk: ganeti.objects.Disk object
1150

1151
  Returns:
1152
    boolean indicating whether a LD_LV dev_type was found or not
1153

1154
  """
1155
  if disk.children:
1156
    for chdisk in disk.children:
1157
      if _RecursiveCheckIfLVMBased(chdisk):
1158
        return True
1159
  return disk.dev_type == constants.LD_LV
1160

    
1161

    
1162
class LUSetClusterParams(LogicalUnit):
1163
  """Change the parameters of the cluster.
1164

1165
  """
1166
  HPATH = "cluster-modify"
1167
  HTYPE = constants.HTYPE_CLUSTER
1168
  _OP_REQP = []
1169
  REQ_BGL = False
1170

    
1171
  def ExpandNames(self):
1172
    # FIXME: in the future maybe other cluster params won't require checking on
1173
    # all nodes to be modified.
1174
    self.needed_locks = {
1175
      locking.LEVEL_NODE: locking.ALL_SET,
1176
    }
1177
    self.share_locks[locking.LEVEL_NODE] = 1
1178

    
1179
  def BuildHooksEnv(self):
1180
    """Build hooks env.
1181

1182
    """
1183
    env = {
1184
      "OP_TARGET": self.cfg.GetClusterName(),
1185
      "NEW_VG_NAME": self.op.vg_name,
1186
      }
1187
    mn = self.cfg.GetMasterNode()
1188
    return env, [mn], [mn]
1189

    
1190
  def CheckPrereq(self):
1191
    """Check prerequisites.
1192

1193
    This checks whether the given params don't conflict and
1194
    if the given volume group is valid.
1195

1196
    """
1197
    # FIXME: This only works because there is only one parameter that can be
1198
    # changed or removed.
1199
    if self.op.vg_name is not None and not self.op.vg_name:
1200
      instances = self.cfg.GetAllInstancesInfo().values()
1201
      for inst in instances:
1202
        for disk in inst.disks:
1203
          if _RecursiveCheckIfLVMBased(disk):
1204
            raise errors.OpPrereqError("Cannot disable lvm storage while"
1205
                                       " lvm-based instances exist")
1206

    
1207
    node_list = self.acquired_locks[locking.LEVEL_NODE]
1208

    
1209
    # if vg_name not None, checks given volume group on all nodes
1210
    if self.op.vg_name:
1211
      vglist = self.rpc.call_vg_list(node_list)
1212
      for node in node_list:
1213
        vgstatus = utils.CheckVolumeGroupSize(vglist[node], self.op.vg_name,
1214
                                              constants.MIN_VG_SIZE)
1215
        if vgstatus:
1216
          raise errors.OpPrereqError("Error on node '%s': %s" %
1217
                                     (node, vgstatus))
1218

    
1219
    self.cluster = cluster = self.cfg.GetClusterInfo()
1220
    # beparams changes do not need validation (we can't validate?),
1221
    # but we still process here
1222
    if self.op.beparams:
1223
      self.new_beparams = cluster.FillDict(
1224
        cluster.beparams[constants.BEGR_DEFAULT], self.op.beparams)
1225

    
1226
    # hypervisor list/parameters
1227
    self.new_hvparams = cluster.FillDict(cluster.hvparams, {})
1228
    if self.op.hvparams:
1229
      if not isinstance(self.op.hvparams, dict):
1230
        raise errors.OpPrereqError("Invalid 'hvparams' parameter on input")
1231
      for hv_name, hv_dict in self.op.hvparams.items():
1232
        if hv_name not in self.new_hvparams:
1233
          self.new_hvparams[hv_name] = hv_dict
1234
        else:
1235
          self.new_hvparams[hv_name].update(hv_dict)
1236

    
1237
    if self.op.enabled_hypervisors is not None:
1238
      self.hv_list = self.op.enabled_hypervisors
1239
    else:
1240
      self.hv_list = cluster.enabled_hypervisors
1241

    
1242
    if self.op.hvparams or self.op.enabled_hypervisors is not None:
1243
      # either the enabled list has changed, or the parameters have, validate
1244
      for hv_name, hv_params in self.new_hvparams.items():
1245
        if ((self.op.hvparams and hv_name in self.op.hvparams) or
1246
            (self.op.enabled_hypervisors and
1247
             hv_name in self.op.enabled_hypervisors)):
1248
          # either this is a new hypervisor, or its parameters have changed
1249
          hv_class = hypervisor.GetHypervisor(hv_name)
1250
          hv_class.CheckParameterSyntax(hv_params)
1251
          _CheckHVParams(self, node_list, hv_name, hv_params)
1252

    
1253
  def Exec(self, feedback_fn):
1254
    """Change the parameters of the cluster.
1255

1256
    """
1257
    if self.op.vg_name is not None:
1258
      if self.op.vg_name != self.cfg.GetVGName():
1259
        self.cfg.SetVGName(self.op.vg_name)
1260
      else:
1261
        feedback_fn("Cluster LVM configuration already in desired"
1262
                    " state, not changing")
1263
    if self.op.hvparams:
1264
      self.cluster.hvparams = self.new_hvparams
1265
    if self.op.enabled_hypervisors is not None:
1266
      self.cluster.enabled_hypervisors = self.op.enabled_hypervisors
1267
    if self.op.beparams:
1268
      self.cluster.beparams[constants.BEGR_DEFAULT] = self.new_beparams
1269
    self.cfg.Update(self.cluster)
1270

    
1271

    
1272
def _WaitForSync(lu, instance, oneshot=False, unlock=False):
1273
  """Sleep and poll for an instance's disk to sync.
1274

1275
  """
1276
  if not instance.disks:
1277
    return True
1278

    
1279
  if not oneshot:
1280
    lu.proc.LogInfo("Waiting for instance %s to sync disks." % instance.name)
1281

    
1282
  node = instance.primary_node
1283

    
1284
  for dev in instance.disks:
1285
    lu.cfg.SetDiskID(dev, node)
1286

    
1287
  retries = 0
1288
  while True:
1289
    max_time = 0
1290
    done = True
1291
    cumul_degraded = False
1292
    rstats = lu.rpc.call_blockdev_getmirrorstatus(node, instance.disks)
1293
    if not rstats:
1294
      lu.LogWarning("Can't get any data from node %s", node)
1295
      retries += 1
1296
      if retries >= 10:
1297
        raise errors.RemoteError("Can't contact node %s for mirror data,"
1298
                                 " aborting." % node)
1299
      time.sleep(6)
1300
      continue
1301
    retries = 0
1302
    for i in range(len(rstats)):
1303
      mstat = rstats[i]
1304
      if mstat is None:
1305
        lu.LogWarning("Can't compute data for node %s/%s",
1306
                           node, instance.disks[i].iv_name)
1307
        continue
1308
      # we ignore the ldisk parameter
1309
      perc_done, est_time, is_degraded, _ = mstat
1310
      cumul_degraded = cumul_degraded or (is_degraded and perc_done is None)
1311
      if perc_done is not None:
1312
        done = False
1313
        if est_time is not None:
1314
          rem_time = "%d estimated seconds remaining" % est_time
1315
          max_time = est_time
1316
        else:
1317
          rem_time = "no time estimate"
1318
        lu.proc.LogInfo("- device %s: %5.2f%% done, %s" %
1319
                        (instance.disks[i].iv_name, perc_done, rem_time))
1320
    if done or oneshot:
1321
      break
1322

    
1323
    time.sleep(min(60, max_time))
1324

    
1325
  if done:
1326
    lu.proc.LogInfo("Instance %s's disks are in sync." % instance.name)
1327
  return not cumul_degraded
1328

    
1329

    
1330
def _CheckDiskConsistency(lu, dev, node, on_primary, ldisk=False):
1331
  """Check that mirrors are not degraded.
1332

1333
  The ldisk parameter, if True, will change the test from the
1334
  is_degraded attribute (which represents overall non-ok status for
1335
  the device(s)) to the ldisk (representing the local storage status).
1336

1337
  """
1338
  lu.cfg.SetDiskID(dev, node)
1339
  if ldisk:
1340
    idx = 6
1341
  else:
1342
    idx = 5
1343

    
1344
  result = True
1345
  if on_primary or dev.AssembleOnSecondary():
1346
    rstats = lu.rpc.call_blockdev_find(node, dev)
1347
    if not rstats:
1348
      logging.warning("Node %s: disk degraded, not found or node down", node)
1349
      result = False
1350
    else:
1351
      result = result and (not rstats[idx])
1352
  if dev.children:
1353
    for child in dev.children:
1354
      result = result and _CheckDiskConsistency(lu, child, node, on_primary)
1355

    
1356
  return result
1357

    
1358

    
1359
class LUDiagnoseOS(NoHooksLU):
1360
  """Logical unit for OS diagnose/query.
1361

1362
  """
1363
  _OP_REQP = ["output_fields", "names"]
1364
  REQ_BGL = False
1365
  _FIELDS_STATIC = _FieldSet()
1366
  _FIELDS_DYNAMIC = _FieldSet("name", "valid", "node_status")
1367

    
1368
  def ExpandNames(self):
1369
    if self.op.names:
1370
      raise errors.OpPrereqError("Selective OS query not supported")
1371

    
1372
    _CheckOutputFields(static=self._FIELDS_STATIC,
1373
                       dynamic=self._FIELDS_DYNAMIC,
1374
                       selected=self.op.output_fields)
1375

    
1376
    # Lock all nodes, in shared mode
1377
    self.needed_locks = {}
1378
    self.share_locks[locking.LEVEL_NODE] = 1
1379
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
1380

    
1381
  def CheckPrereq(self):
1382
    """Check prerequisites.
1383

1384
    """
1385

    
1386
  @staticmethod
1387
  def _DiagnoseByOS(node_list, rlist):
1388
    """Remaps a per-node return list into an a per-os per-node dictionary
1389

1390
      Args:
1391
        node_list: a list with the names of all nodes
1392
        rlist: a map with node names as keys and OS objects as values
1393

1394
      Returns:
1395
        map: a map with osnames as keys and as value another map, with
1396
             nodes as
1397
             keys and list of OS objects as values
1398
             e.g. {"debian-etch": {"node1": [<object>,...],
1399
                                   "node2": [<object>,]}
1400
                  }
1401

1402
    """
1403
    all_os = {}
1404
    for node_name, nr in rlist.iteritems():
1405
      if not nr:
1406
        continue
1407
      for os_obj in nr:
1408
        if os_obj.name not in all_os:
1409
          # build a list of nodes for this os containing empty lists
1410
          # for each node in node_list
1411
          all_os[os_obj.name] = {}
1412
          for nname in node_list:
1413
            all_os[os_obj.name][nname] = []
1414
        all_os[os_obj.name][node_name].append(os_obj)
1415
    return all_os
1416

    
1417
  def Exec(self, feedback_fn):
1418
    """Compute the list of OSes.
1419

1420
    """
1421
    node_list = self.acquired_locks[locking.LEVEL_NODE]
1422
    node_data = self.rpc.call_os_diagnose(node_list)
1423
    if node_data == False:
1424
      raise errors.OpExecError("Can't gather the list of OSes")
1425
    pol = self._DiagnoseByOS(node_list, node_data)
1426
    output = []
1427
    for os_name, os_data in pol.iteritems():
1428
      row = []
1429
      for field in self.op.output_fields:
1430
        if field == "name":
1431
          val = os_name
1432
        elif field == "valid":
1433
          val = utils.all([osl and osl[0] for osl in os_data.values()])
1434
        elif field == "node_status":
1435
          val = {}
1436
          for node_name, nos_list in os_data.iteritems():
1437
            val[node_name] = [(v.status, v.path) for v in nos_list]
1438
        else:
1439
          raise errors.ParameterError(field)
1440
        row.append(val)
1441
      output.append(row)
1442

    
1443
    return output
1444

    
1445

    
1446
class LURemoveNode(LogicalUnit):
1447
  """Logical unit for removing a node.
1448

1449
  """
1450
  HPATH = "node-remove"
1451
  HTYPE = constants.HTYPE_NODE
1452
  _OP_REQP = ["node_name"]
1453

    
1454
  def BuildHooksEnv(self):
1455
    """Build hooks env.
1456

1457
    This doesn't run on the target node in the pre phase as a failed
1458
    node would then be impossible to remove.
1459

1460
    """
1461
    env = {
1462
      "OP_TARGET": self.op.node_name,
1463
      "NODE_NAME": self.op.node_name,
1464
      }
1465
    all_nodes = self.cfg.GetNodeList()
1466
    all_nodes.remove(self.op.node_name)
1467
    return env, all_nodes, all_nodes
1468

    
1469
  def CheckPrereq(self):
1470
    """Check prerequisites.
1471

1472
    This checks:
1473
     - the node exists in the configuration
1474
     - it does not have primary or secondary instances
1475
     - it's not the master
1476

1477
    Any errors are signalled by raising errors.OpPrereqError.
1478

1479
    """
1480
    node = self.cfg.GetNodeInfo(self.cfg.ExpandNodeName(self.op.node_name))
1481
    if node is None:
1482
      raise errors.OpPrereqError, ("Node '%s' is unknown." % self.op.node_name)
1483

    
1484
    instance_list = self.cfg.GetInstanceList()
1485

    
1486
    masternode = self.cfg.GetMasterNode()
1487
    if node.name == masternode:
1488
      raise errors.OpPrereqError("Node is the master node,"
1489
                                 " you need to failover first.")
1490

    
1491
    for instance_name in instance_list:
1492
      instance = self.cfg.GetInstanceInfo(instance_name)
1493
      if node.name == instance.primary_node:
1494
        raise errors.OpPrereqError("Instance %s still running on the node,"
1495
                                   " please remove first." % instance_name)
1496
      if node.name in instance.secondary_nodes:
1497
        raise errors.OpPrereqError("Instance %s has node as a secondary,"
1498
                                   " please remove first." % instance_name)
1499
    self.op.node_name = node.name
1500
    self.node = node
1501

    
1502
  def Exec(self, feedback_fn):
1503
    """Removes the node from the cluster.
1504

1505
    """
1506
    node = self.node
1507
    logging.info("Stopping the node daemon and removing configs from node %s",
1508
                 node.name)
1509

    
1510
    self.context.RemoveNode(node.name)
1511

    
1512
    self.rpc.call_node_leave_cluster(node.name)
1513

    
1514

    
1515
class LUQueryNodes(NoHooksLU):
1516
  """Logical unit for querying nodes.
1517

1518
  """
1519
  _OP_REQP = ["output_fields", "names"]
1520
  REQ_BGL = False
1521
  _FIELDS_DYNAMIC = _FieldSet(
1522
    "dtotal", "dfree",
1523
    "mtotal", "mnode", "mfree",
1524
    "bootid",
1525
    "ctotal",
1526
    )
1527

    
1528
  _FIELDS_STATIC = _FieldSet(
1529
    "name", "pinst_cnt", "sinst_cnt",
1530
    "pinst_list", "sinst_list",
1531
    "pip", "sip", "tags",
1532
    "serial_no",
1533
    )
1534

    
1535
  def ExpandNames(self):
1536
    _CheckOutputFields(static=self._FIELDS_STATIC,
1537
                       dynamic=self._FIELDS_DYNAMIC,
1538
                       selected=self.op.output_fields)
1539

    
1540
    self.needed_locks = {}
1541
    self.share_locks[locking.LEVEL_NODE] = 1
1542

    
1543
    if self.op.names:
1544
      self.wanted = _GetWantedNodes(self, self.op.names)
1545
    else:
1546
      self.wanted = locking.ALL_SET
1547

    
1548
    self.do_locking = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
1549
    if self.do_locking:
1550
      # if we don't request only static fields, we need to lock the nodes
1551
      self.needed_locks[locking.LEVEL_NODE] = self.wanted
1552

    
1553

    
1554
  def CheckPrereq(self):
1555
    """Check prerequisites.
1556

1557
    """
1558
    # The validation of the node list is done in the _GetWantedNodes,
1559
    # if non empty, and if empty, there's no validation to do
1560
    pass
1561

    
1562
  def Exec(self, feedback_fn):
1563
    """Computes the list of nodes and their attributes.
1564

1565
    """
1566
    all_info = self.cfg.GetAllNodesInfo()
1567
    if self.do_locking:
1568
      nodenames = self.acquired_locks[locking.LEVEL_NODE]
1569
    elif self.wanted != locking.ALL_SET:
1570
      nodenames = self.wanted
1571
      missing = set(nodenames).difference(all_info.keys())
1572
      if missing:
1573
        raise errors.OpExecError(
1574
          "Some nodes were removed before retrieving their data: %s" % missing)
1575
    else:
1576
      nodenames = all_info.keys()
1577

    
1578
    nodenames = utils.NiceSort(nodenames)
1579
    nodelist = [all_info[name] for name in nodenames]
1580

    
1581
    # begin data gathering
1582

    
1583
    if self.do_locking:
1584
      live_data = {}
1585
      node_data = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
1586
                                          self.cfg.GetHypervisorType())
1587
      for name in nodenames:
1588
        nodeinfo = node_data.get(name, None)
1589
        if nodeinfo:
1590
          live_data[name] = {
1591
            "mtotal": utils.TryConvert(int, nodeinfo['memory_total']),
1592
            "mnode": utils.TryConvert(int, nodeinfo['memory_dom0']),
1593
            "mfree": utils.TryConvert(int, nodeinfo['memory_free']),
1594
            "dtotal": utils.TryConvert(int, nodeinfo['vg_size']),
1595
            "dfree": utils.TryConvert(int, nodeinfo['vg_free']),
1596
            "ctotal": utils.TryConvert(int, nodeinfo['cpu_total']),
1597
            "bootid": nodeinfo['bootid'],
1598
            }
1599
        else:
1600
          live_data[name] = {}
1601
    else:
1602
      live_data = dict.fromkeys(nodenames, {})
1603

    
1604
    node_to_primary = dict([(name, set()) for name in nodenames])
1605
    node_to_secondary = dict([(name, set()) for name in nodenames])
1606

    
1607
    inst_fields = frozenset(("pinst_cnt", "pinst_list",
1608
                             "sinst_cnt", "sinst_list"))
1609
    if inst_fields & frozenset(self.op.output_fields):
1610
      instancelist = self.cfg.GetInstanceList()
1611

    
1612
      for instance_name in instancelist:
1613
        inst = self.cfg.GetInstanceInfo(instance_name)
1614
        if inst.primary_node in node_to_primary:
1615
          node_to_primary[inst.primary_node].add(inst.name)
1616
        for secnode in inst.secondary_nodes:
1617
          if secnode in node_to_secondary:
1618
            node_to_secondary[secnode].add(inst.name)
1619

    
1620
    # end data gathering
1621

    
1622
    output = []
1623
    for node in nodelist:
1624
      node_output = []
1625
      for field in self.op.output_fields:
1626
        if field == "name":
1627
          val = node.name
1628
        elif field == "pinst_list":
1629
          val = list(node_to_primary[node.name])
1630
        elif field == "sinst_list":
1631
          val = list(node_to_secondary[node.name])
1632
        elif field == "pinst_cnt":
1633
          val = len(node_to_primary[node.name])
1634
        elif field == "sinst_cnt":
1635
          val = len(node_to_secondary[node.name])
1636
        elif field == "pip":
1637
          val = node.primary_ip
1638
        elif field == "sip":
1639
          val = node.secondary_ip
1640
        elif field == "tags":
1641
          val = list(node.GetTags())
1642
        elif field == "serial_no":
1643
          val = node.serial_no
1644
        elif self._FIELDS_DYNAMIC.Matches(field):
1645
          val = live_data[node.name].get(field, None)
1646
        else:
1647
          raise errors.ParameterError(field)
1648
        node_output.append(val)
1649
      output.append(node_output)
1650

    
1651
    return output
1652

    
1653

    
1654
class LUQueryNodeVolumes(NoHooksLU):
1655
  """Logical unit for getting volumes on node(s).
1656

1657
  """
1658
  _OP_REQP = ["nodes", "output_fields"]
1659
  REQ_BGL = False
1660
  _FIELDS_DYNAMIC = _FieldSet("phys", "vg", "name", "size", "instance")
1661
  _FIELDS_STATIC = _FieldSet("node")
1662

    
1663
  def ExpandNames(self):
1664
    _CheckOutputFields(static=self._FIELDS_STATIC,
1665
                       dynamic=self._FIELDS_DYNAMIC,
1666
                       selected=self.op.output_fields)
1667

    
1668
    self.needed_locks = {}
1669
    self.share_locks[locking.LEVEL_NODE] = 1
1670
    if not self.op.nodes:
1671
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
1672
    else:
1673
      self.needed_locks[locking.LEVEL_NODE] = \
1674
        _GetWantedNodes(self, self.op.nodes)
1675

    
1676
  def CheckPrereq(self):
1677
    """Check prerequisites.
1678

1679
    This checks that the fields required are valid output fields.
1680

1681
    """
1682
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
1683

    
1684
  def Exec(self, feedback_fn):
1685
    """Computes the list of nodes and their attributes.
1686

1687
    """
1688
    nodenames = self.nodes
1689
    volumes = self.rpc.call_node_volumes(nodenames)
1690

    
1691
    ilist = [self.cfg.GetInstanceInfo(iname) for iname
1692
             in self.cfg.GetInstanceList()]
1693

    
1694
    lv_by_node = dict([(inst, inst.MapLVsByNode()) for inst in ilist])
1695

    
1696
    output = []
1697
    for node in nodenames:
1698
      if node not in volumes or not volumes[node]:
1699
        continue
1700

    
1701
      node_vols = volumes[node][:]
1702
      node_vols.sort(key=lambda vol: vol['dev'])
1703

    
1704
      for vol in node_vols:
1705
        node_output = []
1706
        for field in self.op.output_fields:
1707
          if field == "node":
1708
            val = node
1709
          elif field == "phys":
1710
            val = vol['dev']
1711
          elif field == "vg":
1712
            val = vol['vg']
1713
          elif field == "name":
1714
            val = vol['name']
1715
          elif field == "size":
1716
            val = int(float(vol['size']))
1717
          elif field == "instance":
1718
            for inst in ilist:
1719
              if node not in lv_by_node[inst]:
1720
                continue
1721
              if vol['name'] in lv_by_node[inst][node]:
1722
                val = inst.name
1723
                break
1724
            else:
1725
              val = '-'
1726
          else:
1727
            raise errors.ParameterError(field)
1728
          node_output.append(str(val))
1729

    
1730
        output.append(node_output)
1731

    
1732
    return output
1733

    
1734

    
1735
class LUAddNode(LogicalUnit):
1736
  """Logical unit for adding node to the cluster.
1737

1738
  """
1739
  HPATH = "node-add"
1740
  HTYPE = constants.HTYPE_NODE
1741
  _OP_REQP = ["node_name"]
1742

    
1743
  def BuildHooksEnv(self):
1744
    """Build hooks env.
1745

1746
    This will run on all nodes before, and on all nodes + the new node after.
1747

1748
    """
1749
    env = {
1750
      "OP_TARGET": self.op.node_name,
1751
      "NODE_NAME": self.op.node_name,
1752
      "NODE_PIP": self.op.primary_ip,
1753
      "NODE_SIP": self.op.secondary_ip,
1754
      }
1755
    nodes_0 = self.cfg.GetNodeList()
1756
    nodes_1 = nodes_0 + [self.op.node_name, ]
1757
    return env, nodes_0, nodes_1
1758

    
1759
  def CheckPrereq(self):
1760
    """Check prerequisites.
1761

1762
    This checks:
1763
     - the new node is not already in the config
1764
     - it is resolvable
1765
     - its parameters (single/dual homed) matches the cluster
1766

1767
    Any errors are signalled by raising errors.OpPrereqError.
1768

1769
    """
1770
    node_name = self.op.node_name
1771
    cfg = self.cfg
1772

    
1773
    dns_data = utils.HostInfo(node_name)
1774

    
1775
    node = dns_data.name
1776
    primary_ip = self.op.primary_ip = dns_data.ip
1777
    secondary_ip = getattr(self.op, "secondary_ip", None)
1778
    if secondary_ip is None:
1779
      secondary_ip = primary_ip
1780
    if not utils.IsValidIP(secondary_ip):
1781
      raise errors.OpPrereqError("Invalid secondary IP given")
1782
    self.op.secondary_ip = secondary_ip
1783

    
1784
    node_list = cfg.GetNodeList()
1785
    if not self.op.readd and node in node_list:
1786
      raise errors.OpPrereqError("Node %s is already in the configuration" %
1787
                                 node)
1788
    elif self.op.readd and node not in node_list:
1789
      raise errors.OpPrereqError("Node %s is not in the configuration" % node)
1790

    
1791
    for existing_node_name in node_list:
1792
      existing_node = cfg.GetNodeInfo(existing_node_name)
1793

    
1794
      if self.op.readd and node == existing_node_name:
1795
        if (existing_node.primary_ip != primary_ip or
1796
            existing_node.secondary_ip != secondary_ip):
1797
          raise errors.OpPrereqError("Readded node doesn't have the same IP"
1798
                                     " address configuration as before")
1799
        continue
1800

    
1801
      if (existing_node.primary_ip == primary_ip or
1802
          existing_node.secondary_ip == primary_ip or
1803
          existing_node.primary_ip == secondary_ip or
1804
          existing_node.secondary_ip == secondary_ip):
1805
        raise errors.OpPrereqError("New node ip address(es) conflict with"
1806
                                   " existing node %s" % existing_node.name)
1807

    
1808
    # check that the type of the node (single versus dual homed) is the
1809
    # same as for the master
1810
    myself = cfg.GetNodeInfo(self.cfg.GetMasterNode())
1811
    master_singlehomed = myself.secondary_ip == myself.primary_ip
1812
    newbie_singlehomed = secondary_ip == primary_ip
1813
    if master_singlehomed != newbie_singlehomed:
1814
      if master_singlehomed:
1815
        raise errors.OpPrereqError("The master has no private ip but the"
1816
                                   " new node has one")
1817
      else:
1818
        raise errors.OpPrereqError("The master has a private ip but the"
1819
                                   " new node doesn't have one")
1820

    
1821
    # checks reachablity
1822
    if not utils.TcpPing(primary_ip, constants.DEFAULT_NODED_PORT):
1823
      raise errors.OpPrereqError("Node not reachable by ping")
1824

    
1825
    if not newbie_singlehomed:
1826
      # check reachability from my secondary ip to newbie's secondary ip
1827
      if not utils.TcpPing(secondary_ip, constants.DEFAULT_NODED_PORT,
1828
                           source=myself.secondary_ip):
1829
        raise errors.OpPrereqError("Node secondary ip not reachable by TCP"
1830
                                   " based ping to noded port")
1831

    
1832
    self.new_node = objects.Node(name=node,
1833
                                 primary_ip=primary_ip,
1834
                                 secondary_ip=secondary_ip)
1835

    
1836
  def Exec(self, feedback_fn):
1837
    """Adds the new node to the cluster.
1838

1839
    """
1840
    new_node = self.new_node
1841
    node = new_node.name
1842

    
1843
    # check connectivity
1844
    result = self.rpc.call_version([node])[node]
1845
    if result:
1846
      if constants.PROTOCOL_VERSION == result:
1847
        logging.info("Communication to node %s fine, sw version %s match",
1848
                     node, result)
1849
      else:
1850
        raise errors.OpExecError("Version mismatch master version %s,"
1851
                                 " node version %s" %
1852
                                 (constants.PROTOCOL_VERSION, result))
1853
    else:
1854
      raise errors.OpExecError("Cannot get version from the new node")
1855

    
1856
    # setup ssh on node
1857
    logging.info("Copy ssh key to node %s", node)
1858
    priv_key, pub_key, _ = ssh.GetUserFiles(constants.GANETI_RUNAS)
1859
    keyarray = []
1860
    keyfiles = [constants.SSH_HOST_DSA_PRIV, constants.SSH_HOST_DSA_PUB,
1861
                constants.SSH_HOST_RSA_PRIV, constants.SSH_HOST_RSA_PUB,
1862
                priv_key, pub_key]
1863

    
1864
    for i in keyfiles:
1865
      f = open(i, 'r')
1866
      try:
1867
        keyarray.append(f.read())
1868
      finally:
1869
        f.close()
1870

    
1871
    result = self.rpc.call_node_add(node, keyarray[0], keyarray[1],
1872
                                    keyarray[2],
1873
                                    keyarray[3], keyarray[4], keyarray[5])
1874

    
1875
    if not result:
1876
      raise errors.OpExecError("Cannot transfer ssh keys to the new node")
1877

    
1878
    # Add node to our /etc/hosts, and add key to known_hosts
1879
    utils.AddHostToEtcHosts(new_node.name)
1880

    
1881
    if new_node.secondary_ip != new_node.primary_ip:
1882
      if not self.rpc.call_node_has_ip_address(new_node.name,
1883
                                               new_node.secondary_ip):
1884
        raise errors.OpExecError("Node claims it doesn't have the secondary ip"
1885
                                 " you gave (%s). Please fix and re-run this"
1886
                                 " command." % new_node.secondary_ip)
1887

    
1888
    node_verify_list = [self.cfg.GetMasterNode()]
1889
    node_verify_param = {
1890
      'nodelist': [node],
1891
      # TODO: do a node-net-test as well?
1892
    }
1893

    
1894
    result = self.rpc.call_node_verify(node_verify_list, node_verify_param,
1895
                                       self.cfg.GetClusterName())
1896
    for verifier in node_verify_list:
1897
      if not result[verifier]:
1898
        raise errors.OpExecError("Cannot communicate with %s's node daemon"
1899
                                 " for remote verification" % verifier)
1900
      if result[verifier]['nodelist']:
1901
        for failed in result[verifier]['nodelist']:
1902
          feedback_fn("ssh/hostname verification failed %s -> %s" %
1903
                      (verifier, result[verifier]['nodelist'][failed]))
1904
        raise errors.OpExecError("ssh/hostname verification failed.")
1905

    
1906
    # Distribute updated /etc/hosts and known_hosts to all nodes,
1907
    # including the node just added
1908
    myself = self.cfg.GetNodeInfo(self.cfg.GetMasterNode())
1909
    dist_nodes = self.cfg.GetNodeList()
1910
    if not self.op.readd:
1911
      dist_nodes.append(node)
1912
    if myself.name in dist_nodes:
1913
      dist_nodes.remove(myself.name)
1914

    
1915
    logging.debug("Copying hosts and known_hosts to all nodes")
1916
    for fname in (constants.ETC_HOSTS, constants.SSH_KNOWN_HOSTS_FILE):
1917
      result = self.rpc.call_upload_file(dist_nodes, fname)
1918
      for to_node in dist_nodes:
1919
        if not result[to_node]:
1920
          logging.error("Copy of file %s to node %s failed", fname, to_node)
1921

    
1922
    to_copy = []
1923
    if constants.HT_XEN_HVM in self.cfg.GetClusterInfo().enabled_hypervisors:
1924
      to_copy.append(constants.VNC_PASSWORD_FILE)
1925
    for fname in to_copy:
1926
      result = self.rpc.call_upload_file([node], fname)
1927
      if not result[node]:
1928
        logging.error("Could not copy file %s to node %s", fname, node)
1929

    
1930
    if self.op.readd:
1931
      self.context.ReaddNode(new_node)
1932
    else:
1933
      self.context.AddNode(new_node)
1934

    
1935

    
1936
class LUQueryClusterInfo(NoHooksLU):
1937
  """Query cluster configuration.
1938

1939
  """
1940
  _OP_REQP = []
1941
  REQ_MASTER = False
1942
  REQ_BGL = False
1943

    
1944
  def ExpandNames(self):
1945
    self.needed_locks = {}
1946

    
1947
  def CheckPrereq(self):
1948
    """No prerequsites needed for this LU.
1949

1950
    """
1951
    pass
1952

    
1953
  def Exec(self, feedback_fn):
1954
    """Return cluster config.
1955

1956
    """
1957
    cluster = self.cfg.GetClusterInfo()
1958
    result = {
1959
      "software_version": constants.RELEASE_VERSION,
1960
      "protocol_version": constants.PROTOCOL_VERSION,
1961
      "config_version": constants.CONFIG_VERSION,
1962
      "os_api_version": constants.OS_API_VERSION,
1963
      "export_version": constants.EXPORT_VERSION,
1964
      "architecture": (platform.architecture()[0], platform.machine()),
1965
      "name": cluster.cluster_name,
1966
      "master": cluster.master_node,
1967
      "default_hypervisor": cluster.default_hypervisor,
1968
      "enabled_hypervisors": cluster.enabled_hypervisors,
1969
      "hvparams": cluster.hvparams,
1970
      "beparams": cluster.beparams,
1971
      }
1972

    
1973
    return result
1974

    
1975

    
1976
class LUQueryConfigValues(NoHooksLU):
1977
  """Return configuration values.
1978

1979
  """
1980
  _OP_REQP = []
1981
  REQ_BGL = False
1982
  _FIELDS_DYNAMIC = _FieldSet()
1983
  _FIELDS_STATIC = _FieldSet("cluster_name", "master_node", "drain_flag")
1984

    
1985
  def ExpandNames(self):
1986
    self.needed_locks = {}
1987

    
1988
    _CheckOutputFields(static=self._FIELDS_STATIC,
1989
                       dynamic=self._FIELDS_DYNAMIC,
1990
                       selected=self.op.output_fields)
1991

    
1992
  def CheckPrereq(self):
1993
    """No prerequisites.
1994

1995
    """
1996
    pass
1997

    
1998
  def Exec(self, feedback_fn):
1999
    """Dump a representation of the cluster config to the standard output.
2000

2001
    """
2002
    values = []
2003
    for field in self.op.output_fields:
2004
      if field == "cluster_name":
2005
        entry = self.cfg.GetClusterName()
2006
      elif field == "master_node":
2007
        entry = self.cfg.GetMasterNode()
2008
      elif field == "drain_flag":
2009
        entry = os.path.exists(constants.JOB_QUEUE_DRAIN_FILE)
2010
      else:
2011
        raise errors.ParameterError(field)
2012
      values.append(entry)
2013
    return values
2014

    
2015

    
2016
class LUActivateInstanceDisks(NoHooksLU):
2017
  """Bring up an instance's disks.
2018

2019
  """
2020
  _OP_REQP = ["instance_name"]
2021
  REQ_BGL = False
2022

    
2023
  def ExpandNames(self):
2024
    self._ExpandAndLockInstance()
2025
    self.needed_locks[locking.LEVEL_NODE] = []
2026
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2027

    
2028
  def DeclareLocks(self, level):
2029
    if level == locking.LEVEL_NODE:
2030
      self._LockInstancesNodes()
2031

    
2032
  def CheckPrereq(self):
2033
    """Check prerequisites.
2034

2035
    This checks that the instance is in the cluster.
2036

2037
    """
2038
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2039
    assert self.instance is not None, \
2040
      "Cannot retrieve locked instance %s" % self.op.instance_name
2041

    
2042
  def Exec(self, feedback_fn):
2043
    """Activate the disks.
2044

2045
    """
2046
    disks_ok, disks_info = _AssembleInstanceDisks(self, self.instance)
2047
    if not disks_ok:
2048
      raise errors.OpExecError("Cannot activate block devices")
2049

    
2050
    return disks_info
2051

    
2052

    
2053
def _AssembleInstanceDisks(lu, instance, ignore_secondaries=False):
2054
  """Prepare the block devices for an instance.
2055

2056
  This sets up the block devices on all nodes.
2057

2058
  Args:
2059
    instance: a ganeti.objects.Instance object
2060
    ignore_secondaries: if true, errors on secondary nodes won't result
2061
                        in an error return from the function
2062

2063
  Returns:
2064
    false if the operation failed
2065
    list of (host, instance_visible_name, node_visible_name) if the operation
2066
         suceeded with the mapping from node devices to instance devices
2067
  """
2068
  device_info = []
2069
  disks_ok = True
2070
  iname = instance.name
2071
  # With the two passes mechanism we try to reduce the window of
2072
  # opportunity for the race condition of switching DRBD to primary
2073
  # before handshaking occured, but we do not eliminate it
2074

    
2075
  # The proper fix would be to wait (with some limits) until the
2076
  # connection has been made and drbd transitions from WFConnection
2077
  # into any other network-connected state (Connected, SyncTarget,
2078
  # SyncSource, etc.)
2079

    
2080
  # 1st pass, assemble on all nodes in secondary mode
2081
  for inst_disk in instance.disks:
2082
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
2083
      lu.cfg.SetDiskID(node_disk, node)
2084
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, False)
2085
      if not result:
2086
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
2087
                           " (is_primary=False, pass=1)",
2088
                           inst_disk.iv_name, node)
2089
        if not ignore_secondaries:
2090
          disks_ok = False
2091

    
2092
  # FIXME: race condition on drbd migration to primary
2093

    
2094
  # 2nd pass, do only the primary node
2095
  for inst_disk in instance.disks:
2096
    for node, node_disk in inst_disk.ComputeNodeTree(instance.primary_node):
2097
      if node != instance.primary_node:
2098
        continue
2099
      lu.cfg.SetDiskID(node_disk, node)
2100
      result = lu.rpc.call_blockdev_assemble(node, node_disk, iname, True)
2101
      if not result:
2102
        lu.proc.LogWarning("Could not prepare block device %s on node %s"
2103
                           " (is_primary=True, pass=2)",
2104
                           inst_disk.iv_name, node)
2105
        disks_ok = False
2106
    device_info.append((instance.primary_node, inst_disk.iv_name, result))
2107

    
2108
  # leave the disks configured for the primary node
2109
  # this is a workaround that would be fixed better by
2110
  # improving the logical/physical id handling
2111
  for disk in instance.disks:
2112
    lu.cfg.SetDiskID(disk, instance.primary_node)
2113

    
2114
  return disks_ok, device_info
2115

    
2116

    
2117
def _StartInstanceDisks(lu, instance, force):
2118
  """Start the disks of an instance.
2119

2120
  """
2121
  disks_ok, dummy = _AssembleInstanceDisks(lu, instance,
2122
                                           ignore_secondaries=force)
2123
  if not disks_ok:
2124
    _ShutdownInstanceDisks(lu, instance)
2125
    if force is not None and not force:
2126
      lu.proc.LogWarning("", hint="If the message above refers to a"
2127
                         " secondary node,"
2128
                         " you can retry the operation using '--force'.")
2129
    raise errors.OpExecError("Disk consistency error")
2130

    
2131

    
2132
class LUDeactivateInstanceDisks(NoHooksLU):
2133
  """Shutdown an instance's disks.
2134

2135
  """
2136
  _OP_REQP = ["instance_name"]
2137
  REQ_BGL = False
2138

    
2139
  def ExpandNames(self):
2140
    self._ExpandAndLockInstance()
2141
    self.needed_locks[locking.LEVEL_NODE] = []
2142
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2143

    
2144
  def DeclareLocks(self, level):
2145
    if level == locking.LEVEL_NODE:
2146
      self._LockInstancesNodes()
2147

    
2148
  def CheckPrereq(self):
2149
    """Check prerequisites.
2150

2151
    This checks that the instance is in the cluster.
2152

2153
    """
2154
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2155
    assert self.instance is not None, \
2156
      "Cannot retrieve locked instance %s" % self.op.instance_name
2157

    
2158
  def Exec(self, feedback_fn):
2159
    """Deactivate the disks
2160

2161
    """
2162
    instance = self.instance
2163
    _SafeShutdownInstanceDisks(self, instance)
2164

    
2165

    
2166
def _SafeShutdownInstanceDisks(lu, instance):
2167
  """Shutdown block devices of an instance.
2168

2169
  This function checks if an instance is running, before calling
2170
  _ShutdownInstanceDisks.
2171

2172
  """
2173
  ins_l = lu.rpc.call_instance_list([instance.primary_node],
2174
                                      [instance.hypervisor])
2175
  ins_l = ins_l[instance.primary_node]
2176
  if not type(ins_l) is list:
2177
    raise errors.OpExecError("Can't contact node '%s'" %
2178
                             instance.primary_node)
2179

    
2180
  if instance.name in ins_l:
2181
    raise errors.OpExecError("Instance is running, can't shutdown"
2182
                             " block devices.")
2183

    
2184
  _ShutdownInstanceDisks(lu, instance)
2185

    
2186

    
2187
def _ShutdownInstanceDisks(lu, instance, ignore_primary=False):
2188
  """Shutdown block devices of an instance.
2189

2190
  This does the shutdown on all nodes of the instance.
2191

2192
  If the ignore_primary is false, errors on the primary node are
2193
  ignored.
2194

2195
  """
2196
  result = True
2197
  for disk in instance.disks:
2198
    for node, top_disk in disk.ComputeNodeTree(instance.primary_node):
2199
      lu.cfg.SetDiskID(top_disk, node)
2200
      if not lu.rpc.call_blockdev_shutdown(node, top_disk):
2201
        logging.error("Could not shutdown block device %s on node %s",
2202
                      disk.iv_name, node)
2203
        if not ignore_primary or node != instance.primary_node:
2204
          result = False
2205
  return result
2206

    
2207

    
2208
def _CheckNodeFreeMemory(lu, node, reason, requested, hypervisor):
2209
  """Checks if a node has enough free memory.
2210

2211
  This function check if a given node has the needed amount of free
2212
  memory. In case the node has less memory or we cannot get the
2213
  information from the node, this function raise an OpPrereqError
2214
  exception.
2215

2216
  @type lu: C{LogicalUnit}
2217
  @param lu: a logical unit from which we get configuration data
2218
  @type node: C{str}
2219
  @param node: the node to check
2220
  @type reason: C{str}
2221
  @param reason: string to use in the error message
2222
  @type requested: C{int}
2223
  @param requested: the amount of memory in MiB to check for
2224
  @type hypervisor: C{str}
2225
  @param hypervisor: the hypervisor to ask for memory stats
2226
  @raise errors.OpPrereqError: if the node doesn't have enough memory, or
2227
      we cannot check the node
2228

2229
  """
2230
  nodeinfo = lu.rpc.call_node_info([node], lu.cfg.GetVGName(), hypervisor)
2231
  if not nodeinfo or not isinstance(nodeinfo, dict):
2232
    raise errors.OpPrereqError("Could not contact node %s for resource"
2233
                             " information" % (node,))
2234

    
2235
  free_mem = nodeinfo[node].get('memory_free')
2236
  if not isinstance(free_mem, int):
2237
    raise errors.OpPrereqError("Can't compute free memory on node %s, result"
2238
                             " was '%s'" % (node, free_mem))
2239
  if requested > free_mem:
2240
    raise errors.OpPrereqError("Not enough memory on node %s for %s:"
2241
                             " needed %s MiB, available %s MiB" %
2242
                             (node, reason, requested, free_mem))
2243

    
2244

    
2245
class LUStartupInstance(LogicalUnit):
2246
  """Starts an instance.
2247

2248
  """
2249
  HPATH = "instance-start"
2250
  HTYPE = constants.HTYPE_INSTANCE
2251
  _OP_REQP = ["instance_name", "force"]
2252
  REQ_BGL = False
2253

    
2254
  def ExpandNames(self):
2255
    self._ExpandAndLockInstance()
2256
    self.needed_locks[locking.LEVEL_NODE] = []
2257
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2258

    
2259
  def DeclareLocks(self, level):
2260
    if level == locking.LEVEL_NODE:
2261
      self._LockInstancesNodes()
2262

    
2263
  def BuildHooksEnv(self):
2264
    """Build hooks env.
2265

2266
    This runs on master, primary and secondary nodes of the instance.
2267

2268
    """
2269
    env = {
2270
      "FORCE": self.op.force,
2271
      }
2272
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2273
    nl = ([self.cfg.GetMasterNode(), self.instance.primary_node] +
2274
          list(self.instance.secondary_nodes))
2275
    return env, nl, nl
2276

    
2277
  def CheckPrereq(self):
2278
    """Check prerequisites.
2279

2280
    This checks that the instance is in the cluster.
2281

2282
    """
2283
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2284
    assert self.instance is not None, \
2285
      "Cannot retrieve locked instance %s" % self.op.instance_name
2286

    
2287
    bep = self.cfg.GetClusterInfo().FillBE(instance)
2288
    # check bridges existance
2289
    _CheckInstanceBridgesExist(self, instance)
2290

    
2291
    _CheckNodeFreeMemory(self, instance.primary_node,
2292
                         "starting instance %s" % instance.name,
2293
                         bep[constants.BE_MEMORY], instance.hypervisor)
2294

    
2295
  def Exec(self, feedback_fn):
2296
    """Start the instance.
2297

2298
    """
2299
    instance = self.instance
2300
    force = self.op.force
2301
    extra_args = getattr(self.op, "extra_args", "")
2302

    
2303
    self.cfg.MarkInstanceUp(instance.name)
2304

    
2305
    node_current = instance.primary_node
2306

    
2307
    _StartInstanceDisks(self, instance, force)
2308

    
2309
    if not self.rpc.call_instance_start(node_current, instance, extra_args):
2310
      _ShutdownInstanceDisks(self, instance)
2311
      raise errors.OpExecError("Could not start instance")
2312

    
2313

    
2314
class LURebootInstance(LogicalUnit):
2315
  """Reboot an instance.
2316

2317
  """
2318
  HPATH = "instance-reboot"
2319
  HTYPE = constants.HTYPE_INSTANCE
2320
  _OP_REQP = ["instance_name", "ignore_secondaries", "reboot_type"]
2321
  REQ_BGL = False
2322

    
2323
  def ExpandNames(self):
2324
    if self.op.reboot_type not in [constants.INSTANCE_REBOOT_SOFT,
2325
                                   constants.INSTANCE_REBOOT_HARD,
2326
                                   constants.INSTANCE_REBOOT_FULL]:
2327
      raise errors.ParameterError("reboot type not in [%s, %s, %s]" %
2328
                                  (constants.INSTANCE_REBOOT_SOFT,
2329
                                   constants.INSTANCE_REBOOT_HARD,
2330
                                   constants.INSTANCE_REBOOT_FULL))
2331
    self._ExpandAndLockInstance()
2332
    self.needed_locks[locking.LEVEL_NODE] = []
2333
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2334

    
2335
  def DeclareLocks(self, level):
2336
    if level == locking.LEVEL_NODE:
2337
      primary_only = not constants.INSTANCE_REBOOT_FULL
2338
      self._LockInstancesNodes(primary_only=primary_only)
2339

    
2340
  def BuildHooksEnv(self):
2341
    """Build hooks env.
2342

2343
    This runs on master, primary and secondary nodes of the instance.
2344

2345
    """
2346
    env = {
2347
      "IGNORE_SECONDARIES": self.op.ignore_secondaries,
2348
      }
2349
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2350
    nl = ([self.cfg.GetMasterNode(), self.instance.primary_node] +
2351
          list(self.instance.secondary_nodes))
2352
    return env, nl, nl
2353

    
2354
  def CheckPrereq(self):
2355
    """Check prerequisites.
2356

2357
    This checks that the instance is in the cluster.
2358

2359
    """
2360
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2361
    assert self.instance is not None, \
2362
      "Cannot retrieve locked instance %s" % self.op.instance_name
2363

    
2364
    # check bridges existance
2365
    _CheckInstanceBridgesExist(self, instance)
2366

    
2367
  def Exec(self, feedback_fn):
2368
    """Reboot the instance.
2369

2370
    """
2371
    instance = self.instance
2372
    ignore_secondaries = self.op.ignore_secondaries
2373
    reboot_type = self.op.reboot_type
2374
    extra_args = getattr(self.op, "extra_args", "")
2375

    
2376
    node_current = instance.primary_node
2377

    
2378
    if reboot_type in [constants.INSTANCE_REBOOT_SOFT,
2379
                       constants.INSTANCE_REBOOT_HARD]:
2380
      if not self.rpc.call_instance_reboot(node_current, instance,
2381
                                           reboot_type, extra_args):
2382
        raise errors.OpExecError("Could not reboot instance")
2383
    else:
2384
      if not self.rpc.call_instance_shutdown(node_current, instance):
2385
        raise errors.OpExecError("could not shutdown instance for full reboot")
2386
      _ShutdownInstanceDisks(self, instance)
2387
      _StartInstanceDisks(self, instance, ignore_secondaries)
2388
      if not self.rpc.call_instance_start(node_current, instance, extra_args):
2389
        _ShutdownInstanceDisks(self, instance)
2390
        raise errors.OpExecError("Could not start instance for full reboot")
2391

    
2392
    self.cfg.MarkInstanceUp(instance.name)
2393

    
2394

    
2395
class LUShutdownInstance(LogicalUnit):
2396
  """Shutdown an instance.
2397

2398
  """
2399
  HPATH = "instance-stop"
2400
  HTYPE = constants.HTYPE_INSTANCE
2401
  _OP_REQP = ["instance_name"]
2402
  REQ_BGL = False
2403

    
2404
  def ExpandNames(self):
2405
    self._ExpandAndLockInstance()
2406
    self.needed_locks[locking.LEVEL_NODE] = []
2407
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2408

    
2409
  def DeclareLocks(self, level):
2410
    if level == locking.LEVEL_NODE:
2411
      self._LockInstancesNodes()
2412

    
2413
  def BuildHooksEnv(self):
2414
    """Build hooks env.
2415

2416
    This runs on master, primary and secondary nodes of the instance.
2417

2418
    """
2419
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2420
    nl = ([self.cfg.GetMasterNode(), self.instance.primary_node] +
2421
          list(self.instance.secondary_nodes))
2422
    return env, nl, nl
2423

    
2424
  def CheckPrereq(self):
2425
    """Check prerequisites.
2426

2427
    This checks that the instance is in the cluster.
2428

2429
    """
2430
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2431
    assert self.instance is not None, \
2432
      "Cannot retrieve locked instance %s" % self.op.instance_name
2433

    
2434
  def Exec(self, feedback_fn):
2435
    """Shutdown the instance.
2436

2437
    """
2438
    instance = self.instance
2439
    node_current = instance.primary_node
2440
    self.cfg.MarkInstanceDown(instance.name)
2441
    if not self.rpc.call_instance_shutdown(node_current, instance):
2442
      self.proc.LogWarning("Could not shutdown instance")
2443

    
2444
    _ShutdownInstanceDisks(self, instance)
2445

    
2446

    
2447
class LUReinstallInstance(LogicalUnit):
2448
  """Reinstall an instance.
2449

2450
  """
2451
  HPATH = "instance-reinstall"
2452
  HTYPE = constants.HTYPE_INSTANCE
2453
  _OP_REQP = ["instance_name"]
2454
  REQ_BGL = False
2455

    
2456
  def ExpandNames(self):
2457
    self._ExpandAndLockInstance()
2458
    self.needed_locks[locking.LEVEL_NODE] = []
2459
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2460

    
2461
  def DeclareLocks(self, level):
2462
    if level == locking.LEVEL_NODE:
2463
      self._LockInstancesNodes()
2464

    
2465
  def BuildHooksEnv(self):
2466
    """Build hooks env.
2467

2468
    This runs on master, primary and secondary nodes of the instance.
2469

2470
    """
2471
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2472
    nl = ([self.cfg.GetMasterNode(), self.instance.primary_node] +
2473
          list(self.instance.secondary_nodes))
2474
    return env, nl, nl
2475

    
2476
  def CheckPrereq(self):
2477
    """Check prerequisites.
2478

2479
    This checks that the instance is in the cluster and is not running.
2480

2481
    """
2482
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2483
    assert instance is not None, \
2484
      "Cannot retrieve locked instance %s" % self.op.instance_name
2485

    
2486
    if instance.disk_template == constants.DT_DISKLESS:
2487
      raise errors.OpPrereqError("Instance '%s' has no disks" %
2488
                                 self.op.instance_name)
2489
    if instance.status != "down":
2490
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
2491
                                 self.op.instance_name)
2492
    remote_info = self.rpc.call_instance_info(instance.primary_node,
2493
                                              instance.name,
2494
                                              instance.hypervisor)
2495
    if remote_info:
2496
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
2497
                                 (self.op.instance_name,
2498
                                  instance.primary_node))
2499

    
2500
    self.op.os_type = getattr(self.op, "os_type", None)
2501
    if self.op.os_type is not None:
2502
      # OS verification
2503
      pnode = self.cfg.GetNodeInfo(
2504
        self.cfg.ExpandNodeName(instance.primary_node))
2505
      if pnode is None:
2506
        raise errors.OpPrereqError("Primary node '%s' is unknown" %
2507
                                   self.op.pnode)
2508
      os_obj = self.rpc.call_os_get(pnode.name, self.op.os_type)
2509
      if not os_obj:
2510
        raise errors.OpPrereqError("OS '%s' not in supported OS list for"
2511
                                   " primary node"  % self.op.os_type)
2512

    
2513
    self.instance = instance
2514

    
2515
  def Exec(self, feedback_fn):
2516
    """Reinstall the instance.
2517

2518
    """
2519
    inst = self.instance
2520

    
2521
    if self.op.os_type is not None:
2522
      feedback_fn("Changing OS to '%s'..." % self.op.os_type)
2523
      inst.os = self.op.os_type
2524
      self.cfg.Update(inst)
2525

    
2526
    _StartInstanceDisks(self, inst, None)
2527
    try:
2528
      feedback_fn("Running the instance OS create scripts...")
2529
      if not self.rpc.call_instance_os_add(inst.primary_node, inst):
2530
        raise errors.OpExecError("Could not install OS for instance %s"
2531
                                 " on node %s" %
2532
                                 (inst.name, inst.primary_node))
2533
    finally:
2534
      _ShutdownInstanceDisks(self, inst)
2535

    
2536

    
2537
class LURenameInstance(LogicalUnit):
2538
  """Rename an instance.
2539

2540
  """
2541
  HPATH = "instance-rename"
2542
  HTYPE = constants.HTYPE_INSTANCE
2543
  _OP_REQP = ["instance_name", "new_name"]
2544

    
2545
  def BuildHooksEnv(self):
2546
    """Build hooks env.
2547

2548
    This runs on master, primary and secondary nodes of the instance.
2549

2550
    """
2551
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2552
    env["INSTANCE_NEW_NAME"] = self.op.new_name
2553
    nl = ([self.cfg.GetMasterNode(), self.instance.primary_node] +
2554
          list(self.instance.secondary_nodes))
2555
    return env, nl, nl
2556

    
2557
  def CheckPrereq(self):
2558
    """Check prerequisites.
2559

2560
    This checks that the instance is in the cluster and is not running.
2561

2562
    """
2563
    instance = self.cfg.GetInstanceInfo(
2564
      self.cfg.ExpandInstanceName(self.op.instance_name))
2565
    if instance is None:
2566
      raise errors.OpPrereqError("Instance '%s' not known" %
2567
                                 self.op.instance_name)
2568
    if instance.status != "down":
2569
      raise errors.OpPrereqError("Instance '%s' is marked to be up" %
2570
                                 self.op.instance_name)
2571
    remote_info = self.rpc.call_instance_info(instance.primary_node,
2572
                                              instance.name,
2573
                                              instance.hypervisor)
2574
    if remote_info:
2575
      raise errors.OpPrereqError("Instance '%s' is running on the node %s" %
2576
                                 (self.op.instance_name,
2577
                                  instance.primary_node))
2578
    self.instance = instance
2579

    
2580
    # new name verification
2581
    name_info = utils.HostInfo(self.op.new_name)
2582

    
2583
    self.op.new_name = new_name = name_info.name
2584
    instance_list = self.cfg.GetInstanceList()
2585
    if new_name in instance_list:
2586
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
2587
                                 new_name)
2588

    
2589
    if not getattr(self.op, "ignore_ip", False):
2590
      if utils.TcpPing(name_info.ip, constants.DEFAULT_NODED_PORT):
2591
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
2592
                                   (name_info.ip, new_name))
2593

    
2594

    
2595
  def Exec(self, feedback_fn):
2596
    """Reinstall the instance.
2597

2598
    """
2599
    inst = self.instance
2600
    old_name = inst.name
2601

    
2602
    if inst.disk_template == constants.DT_FILE:
2603
      old_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
2604

    
2605
    self.cfg.RenameInstance(inst.name, self.op.new_name)
2606
    # Change the instance lock. This is definitely safe while we hold the BGL
2607
    self.context.glm.remove(locking.LEVEL_INSTANCE, old_name)
2608
    self.context.glm.add(locking.LEVEL_INSTANCE, self.op.new_name)
2609

    
2610
    # re-read the instance from the configuration after rename
2611
    inst = self.cfg.GetInstanceInfo(self.op.new_name)
2612

    
2613
    if inst.disk_template == constants.DT_FILE:
2614
      new_file_storage_dir = os.path.dirname(inst.disks[0].logical_id[1])
2615
      result = self.rpc.call_file_storage_dir_rename(inst.primary_node,
2616
                                                     old_file_storage_dir,
2617
                                                     new_file_storage_dir)
2618

    
2619
      if not result:
2620
        raise errors.OpExecError("Could not connect to node '%s' to rename"
2621
                                 " directory '%s' to '%s' (but the instance"
2622
                                 " has been renamed in Ganeti)" % (
2623
                                 inst.primary_node, old_file_storage_dir,
2624
                                 new_file_storage_dir))
2625

    
2626
      if not result[0]:
2627
        raise errors.OpExecError("Could not rename directory '%s' to '%s'"
2628
                                 " (but the instance has been renamed in"
2629
                                 " Ganeti)" % (old_file_storage_dir,
2630
                                               new_file_storage_dir))
2631

    
2632
    _StartInstanceDisks(self, inst, None)
2633
    try:
2634
      if not self.rpc.call_instance_run_rename(inst.primary_node, inst,
2635
                                               old_name):
2636
        msg = ("Could not run OS rename script for instance %s on node %s"
2637
               " (but the instance has been renamed in Ganeti)" %
2638
               (inst.name, inst.primary_node))
2639
        self.proc.LogWarning(msg)
2640
    finally:
2641
      _ShutdownInstanceDisks(self, inst)
2642

    
2643

    
2644
class LURemoveInstance(LogicalUnit):
2645
  """Remove an instance.
2646

2647
  """
2648
  HPATH = "instance-remove"
2649
  HTYPE = constants.HTYPE_INSTANCE
2650
  _OP_REQP = ["instance_name", "ignore_failures"]
2651
  REQ_BGL = False
2652

    
2653
  def ExpandNames(self):
2654
    self._ExpandAndLockInstance()
2655
    self.needed_locks[locking.LEVEL_NODE] = []
2656
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2657

    
2658
  def DeclareLocks(self, level):
2659
    if level == locking.LEVEL_NODE:
2660
      self._LockInstancesNodes()
2661

    
2662
  def BuildHooksEnv(self):
2663
    """Build hooks env.
2664

2665
    This runs on master, primary and secondary nodes of the instance.
2666

2667
    """
2668
    env = _BuildInstanceHookEnvByObject(self, self.instance)
2669
    nl = [self.cfg.GetMasterNode()]
2670
    return env, nl, nl
2671

    
2672
  def CheckPrereq(self):
2673
    """Check prerequisites.
2674

2675
    This checks that the instance is in the cluster.
2676

2677
    """
2678
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2679
    assert self.instance is not None, \
2680
      "Cannot retrieve locked instance %s" % self.op.instance_name
2681

    
2682
  def Exec(self, feedback_fn):
2683
    """Remove the instance.
2684

2685
    """
2686
    instance = self.instance
2687
    logging.info("Shutting down instance %s on node %s",
2688
                 instance.name, instance.primary_node)
2689

    
2690
    if not self.rpc.call_instance_shutdown(instance.primary_node, instance):
2691
      if self.op.ignore_failures:
2692
        feedback_fn("Warning: can't shutdown instance")
2693
      else:
2694
        raise errors.OpExecError("Could not shutdown instance %s on node %s" %
2695
                                 (instance.name, instance.primary_node))
2696

    
2697
    logging.info("Removing block devices for instance %s", instance.name)
2698

    
2699
    if not _RemoveDisks(self, instance):
2700
      if self.op.ignore_failures:
2701
        feedback_fn("Warning: can't remove instance's disks")
2702
      else:
2703
        raise errors.OpExecError("Can't remove instance's disks")
2704

    
2705
    logging.info("Removing instance %s out of cluster config", instance.name)
2706

    
2707
    self.cfg.RemoveInstance(instance.name)
2708
    self.remove_locks[locking.LEVEL_INSTANCE] = instance.name
2709

    
2710

    
2711
class LUQueryInstances(NoHooksLU):
2712
  """Logical unit for querying instances.
2713

2714
  """
2715
  _OP_REQP = ["output_fields", "names"]
2716
  REQ_BGL = False
2717
  _FIELDS_STATIC = _FieldSet(*["name", "os", "pnode", "snodes",
2718
                               "admin_state", "admin_ram",
2719
                               "disk_template", "ip", "mac", "bridge",
2720
                               "sda_size", "sdb_size", "vcpus", "tags",
2721
                               "network_port", "beparams",
2722
                               "(disk).(size)/([0-9]+)",
2723
                               "(disk).(sizes)",
2724
                               "(nic).(mac|ip|bridge)/([0-9]+)",
2725
                               "(nic).(macs|ips|bridges)",
2726
                               "(disk|nic).(count)",
2727
                               "serial_no", "hypervisor", "hvparams",] +
2728
                             ["hv/%s" % name
2729
                              for name in constants.HVS_PARAMETERS] +
2730
                             ["be/%s" % name
2731
                              for name in constants.BES_PARAMETERS])
2732
  _FIELDS_DYNAMIC = _FieldSet("oper_state", "oper_ram", "status")
2733

    
2734

    
2735
  def ExpandNames(self):
2736
    _CheckOutputFields(static=self._FIELDS_STATIC,
2737
                       dynamic=self._FIELDS_DYNAMIC,
2738
                       selected=self.op.output_fields)
2739

    
2740
    self.needed_locks = {}
2741
    self.share_locks[locking.LEVEL_INSTANCE] = 1
2742
    self.share_locks[locking.LEVEL_NODE] = 1
2743

    
2744
    if self.op.names:
2745
      self.wanted = _GetWantedInstances(self, self.op.names)
2746
    else:
2747
      self.wanted = locking.ALL_SET
2748

    
2749
    self.do_locking = self._FIELDS_STATIC.NonMatching(self.op.output_fields)
2750
    if self.do_locking:
2751
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted
2752
      self.needed_locks[locking.LEVEL_NODE] = []
2753
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2754

    
2755
  def DeclareLocks(self, level):
2756
    if level == locking.LEVEL_NODE and self.do_locking:
2757
      self._LockInstancesNodes()
2758

    
2759
  def CheckPrereq(self):
2760
    """Check prerequisites.
2761

2762
    """
2763
    pass
2764

    
2765
  def Exec(self, feedback_fn):
2766
    """Computes the list of nodes and their attributes.
2767

2768
    """
2769
    all_info = self.cfg.GetAllInstancesInfo()
2770
    if self.do_locking:
2771
      instance_names = self.acquired_locks[locking.LEVEL_INSTANCE]
2772
    elif self.wanted != locking.ALL_SET:
2773
      instance_names = self.wanted
2774
      missing = set(instance_names).difference(all_info.keys())
2775
      if missing:
2776
        raise errors.OpExecError(
2777
          "Some instances were removed before retrieving their data: %s"
2778
          % missing)
2779
    else:
2780
      instance_names = all_info.keys()
2781

    
2782
    instance_names = utils.NiceSort(instance_names)
2783
    instance_list = [all_info[iname] for iname in instance_names]
2784

    
2785
    # begin data gathering
2786

    
2787
    nodes = frozenset([inst.primary_node for inst in instance_list])
2788
    hv_list = list(set([inst.hypervisor for inst in instance_list]))
2789

    
2790
    bad_nodes = []
2791
    if self.do_locking:
2792
      live_data = {}
2793
      node_data = self.rpc.call_all_instances_info(nodes, hv_list)
2794
      for name in nodes:
2795
        result = node_data[name]
2796
        if result:
2797
          live_data.update(result)
2798
        elif result == False:
2799
          bad_nodes.append(name)
2800
        # else no instance is alive
2801
    else:
2802
      live_data = dict([(name, {}) for name in instance_names])
2803

    
2804
    # end data gathering
2805

    
2806
    HVPREFIX = "hv/"
2807
    BEPREFIX = "be/"
2808
    output = []
2809
    for instance in instance_list:
2810
      iout = []
2811
      i_hv = self.cfg.GetClusterInfo().FillHV(instance)
2812
      i_be = self.cfg.GetClusterInfo().FillBE(instance)
2813
      for field in self.op.output_fields:
2814
        st_match = self._FIELDS_STATIC.Matches(field)
2815
        if field == "name":
2816
          val = instance.name
2817
        elif field == "os":
2818
          val = instance.os
2819
        elif field == "pnode":
2820
          val = instance.primary_node
2821
        elif field == "snodes":
2822
          val = list(instance.secondary_nodes)
2823
        elif field == "admin_state":
2824
          val = (instance.status != "down")
2825
        elif field == "oper_state":
2826
          if instance.primary_node in bad_nodes:
2827
            val = None
2828
          else:
2829
            val = bool(live_data.get(instance.name))
2830
        elif field == "status":
2831
          if instance.primary_node in bad_nodes:
2832
            val = "ERROR_nodedown"
2833
          else:
2834
            running = bool(live_data.get(instance.name))
2835
            if running:
2836
              if instance.status != "down":
2837
                val = "running"
2838
              else:
2839
                val = "ERROR_up"
2840
            else:
2841
              if instance.status != "down":
2842
                val = "ERROR_down"
2843
              else:
2844
                val = "ADMIN_down"
2845
        elif field == "oper_ram":
2846
          if instance.primary_node in bad_nodes:
2847
            val = None
2848
          elif instance.name in live_data:
2849
            val = live_data[instance.name].get("memory", "?")
2850
          else:
2851
            val = "-"
2852
        elif field == "disk_template":
2853
          val = instance.disk_template
2854
        elif field == "ip":
2855
          val = instance.nics[0].ip
2856
        elif field == "bridge":
2857
          val = instance.nics[0].bridge
2858
        elif field == "mac":
2859
          val = instance.nics[0].mac
2860
        elif field == "sda_size" or field == "sdb_size":
2861
          disk = instance.FindDisk(field[:3])
2862
          if disk is None:
2863
            val = None
2864
          else:
2865
            val = disk.size
2866
        elif field == "tags":
2867
          val = list(instance.GetTags())
2868
        elif field == "serial_no":
2869
          val = instance.serial_no
2870
        elif field == "network_port":
2871
          val = instance.network_port
2872
        elif field == "hypervisor":
2873
          val = instance.hypervisor
2874
        elif field == "hvparams":
2875
          val = i_hv
2876
        elif (field.startswith(HVPREFIX) and
2877
              field[len(HVPREFIX):] in constants.HVS_PARAMETERS):
2878
          val = i_hv.get(field[len(HVPREFIX):], None)
2879
        elif field == "beparams":
2880
          val = i_be
2881
        elif (field.startswith(BEPREFIX) and
2882
              field[len(BEPREFIX):] in constants.BES_PARAMETERS):
2883
          val = i_be.get(field[len(BEPREFIX):], None)
2884
        elif st_match and st_match.groups():
2885
          # matches a variable list
2886
          st_groups = st_match.groups()
2887
          if st_groups and st_groups[0] == "disk":
2888
            if st_groups[1] == "count":
2889
              val = len(instance.disks)
2890
            elif st_groups[1] == "sizes":
2891
              val = [disk.size for disk in instance.disks]
2892
            elif st_groups[1] == "size":
2893
              disk_idx = int(st_groups[2])
2894
              if disk_idx >= len(instance.disks):
2895
                val = None
2896
              else:
2897
                val = instance.disks[disk_idx].size
2898
            else:
2899
              assert False, "Unhandled disk parameter"
2900
          elif st_groups[0] == "nic":
2901
            if st_groups[1] == "count":
2902
              val = len(instance.nics)
2903
            elif st_groups[1] == "macs":
2904
              val = [nic.mac for nic in instance.nics]
2905
            elif st_groups[1] == "ips":
2906
              val = [nic.ip for nic in instance.nics]
2907
            elif st_groups[1] == "bridges":
2908
              val = [nic.bridge for nic in instance.nics]
2909
            else:
2910
              # index-based item
2911
              nic_idx = int(st_groups[2])
2912
              if nic_idx >= len(instance.nics):
2913
                val = None
2914
              else:
2915
                if st_groups[1] == "mac":
2916
                  val = instance.nics[nic_idx].mac
2917
                elif st_groups[1] == "ip":
2918
                  val = instance.nics[nic_idx].ip
2919
                elif st_groups[1] == "bridge":
2920
                  val = instance.nics[nic_idx].bridge
2921
                else:
2922
                  assert False, "Unhandled NIC parameter"
2923
          else:
2924
            assert False, "Unhandled variable parameter"
2925
        else:
2926
          raise errors.ParameterError(field)
2927
        iout.append(val)
2928
      output.append(iout)
2929

    
2930
    return output
2931

    
2932

    
2933
class LUFailoverInstance(LogicalUnit):
2934
  """Failover an instance.
2935

2936
  """
2937
  HPATH = "instance-failover"
2938
  HTYPE = constants.HTYPE_INSTANCE
2939
  _OP_REQP = ["instance_name", "ignore_consistency"]
2940
  REQ_BGL = False
2941

    
2942
  def ExpandNames(self):
2943
    self._ExpandAndLockInstance()
2944
    self.needed_locks[locking.LEVEL_NODE] = []
2945
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
2946

    
2947
  def DeclareLocks(self, level):
2948
    if level == locking.LEVEL_NODE:
2949
      self._LockInstancesNodes()
2950

    
2951
  def BuildHooksEnv(self):
2952
    """Build hooks env.
2953

2954
    This runs on master, primary and secondary nodes of the instance.
2955

2956
    """
2957
    env = {
2958
      "IGNORE_CONSISTENCY": self.op.ignore_consistency,
2959
      }
2960
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
2961
    nl = [self.cfg.GetMasterNode()] + list(self.instance.secondary_nodes)
2962
    return env, nl, nl
2963

    
2964
  def CheckPrereq(self):
2965
    """Check prerequisites.
2966

2967
    This checks that the instance is in the cluster.
2968

2969
    """
2970
    self.instance = instance = self.cfg.GetInstanceInfo(self.op.instance_name)
2971
    assert self.instance is not None, \
2972
      "Cannot retrieve locked instance %s" % self.op.instance_name
2973

    
2974
    bep = self.cfg.GetClusterInfo().FillBE(instance)
2975
    if instance.disk_template not in constants.DTS_NET_MIRROR:
2976
      raise errors.OpPrereqError("Instance's disk layout is not"
2977
                                 " network mirrored, cannot failover.")
2978

    
2979
    secondary_nodes = instance.secondary_nodes
2980
    if not secondary_nodes:
2981
      raise errors.ProgrammerError("no secondary node but using "
2982
                                   "a mirrored disk template")
2983

    
2984
    target_node = secondary_nodes[0]
2985
    # check memory requirements on the secondary node
2986
    _CheckNodeFreeMemory(self, target_node, "failing over instance %s" %
2987
                         instance.name, bep[constants.BE_MEMORY],
2988
                         instance.hypervisor)
2989

    
2990
    # check bridge existance
2991
    brlist = [nic.bridge for nic in instance.nics]
2992
    if not self.rpc.call_bridges_exist(target_node, brlist):
2993
      raise errors.OpPrereqError("One or more target bridges %s does not"
2994
                                 " exist on destination node '%s'" %
2995
                                 (brlist, target_node))
2996

    
2997
  def Exec(self, feedback_fn):
2998
    """Failover an instance.
2999

3000
    The failover is done by shutting it down on its present node and
3001
    starting it on the secondary.
3002

3003
    """
3004
    instance = self.instance
3005

    
3006
    source_node = instance.primary_node
3007
    target_node = instance.secondary_nodes[0]
3008

    
3009
    feedback_fn("* checking disk consistency between source and target")
3010
    for dev in instance.disks:
3011
      # for drbd, these are drbd over lvm
3012
      if not _CheckDiskConsistency(self, dev, target_node, False):
3013
        if instance.status == "up" and not self.op.ignore_consistency:
3014
          raise errors.OpExecError("Disk %s is degraded on target node,"
3015
                                   " aborting failover." % dev.iv_name)
3016

    
3017
    feedback_fn("* shutting down instance on source node")
3018
    logging.info("Shutting down instance %s on node %s",
3019
                 instance.name, source_node)
3020

    
3021
    if not self.rpc.call_instance_shutdown(source_node, instance):
3022
      if self.op.ignore_consistency:
3023
        self.proc.LogWarning("Could not shutdown instance %s on node %s."
3024
                             " Proceeding"
3025
                             " anyway. Please make sure node %s is down",
3026
                             instance.name, source_node, source_node)
3027
      else:
3028
        raise errors.OpExecError("Could not shutdown instance %s on node %s" %
3029
                                 (instance.name, source_node))
3030

    
3031
    feedback_fn("* deactivating the instance's disks on source node")
3032
    if not _ShutdownInstanceDisks(self, instance, ignore_primary=True):
3033
      raise errors.OpExecError("Can't shut down the instance's disks.")
3034

    
3035
    instance.primary_node = target_node
3036
    # distribute new instance config to the other nodes
3037
    self.cfg.Update(instance)
3038

    
3039
    # Only start the instance if it's marked as up
3040
    if instance.status == "up":
3041
      feedback_fn("* activating the instance's disks on target node")
3042
      logging.info("Starting instance %s on node %s",
3043
                   instance.name, target_node)
3044

    
3045
      disks_ok, dummy = _AssembleInstanceDisks(self, instance,
3046
                                               ignore_secondaries=True)
3047
      if not disks_ok:
3048
        _ShutdownInstanceDisks(self, instance)
3049
        raise errors.OpExecError("Can't activate the instance's disks")
3050

    
3051
      feedback_fn("* starting the instance on the target node")
3052
      if not self.rpc.call_instance_start(target_node, instance, None):
3053
        _ShutdownInstanceDisks(self, instance)
3054
        raise errors.OpExecError("Could not start instance %s on node %s." %
3055
                                 (instance.name, target_node))
3056

    
3057

    
3058
def _CreateBlockDevOnPrimary(lu, node, instance, device, info):
3059
  """Create a tree of block devices on the primary node.
3060

3061
  This always creates all devices.
3062

3063
  """
3064
  if device.children:
3065
    for child in device.children:
3066
      if not _CreateBlockDevOnPrimary(lu, node, instance, child, info):
3067
        return False
3068

    
3069
  lu.cfg.SetDiskID(device, node)
3070
  new_id = lu.rpc.call_blockdev_create(node, device, device.size,
3071
                                       instance.name, True, info)
3072
  if not new_id:
3073
    return False
3074
  if device.physical_id is None:
3075
    device.physical_id = new_id
3076
  return True
3077

    
3078

    
3079
def _CreateBlockDevOnSecondary(lu, node, instance, device, force, info):
3080
  """Create a tree of block devices on a secondary node.
3081

3082
  If this device type has to be created on secondaries, create it and
3083
  all its children.
3084

3085
  If not, just recurse to children keeping the same 'force' value.
3086

3087
  """
3088
  if device.CreateOnSecondary():
3089
    force = True
3090
  if device.children:
3091
    for child in device.children:
3092
      if not _CreateBlockDevOnSecondary(lu, node, instance,
3093
                                        child, force, info):
3094
        return False
3095

    
3096
  if not force:
3097
    return True
3098
  lu.cfg.SetDiskID(device, node)
3099
  new_id = lu.rpc.call_blockdev_create(node, device, device.size,
3100
                                       instance.name, False, info)
3101
  if not new_id:
3102
    return False
3103
  if device.physical_id is None:
3104
    device.physical_id = new_id
3105
  return True
3106

    
3107

    
3108
def _GenerateUniqueNames(lu, exts):
3109
  """Generate a suitable LV name.
3110

3111
  This will generate a logical volume name for the given instance.
3112

3113
  """
3114
  results = []
3115
  for val in exts:
3116
    new_id = lu.cfg.GenerateUniqueID()
3117
    results.append("%s%s" % (new_id, val))
3118
  return results
3119

    
3120

    
3121
def _GenerateDRBD8Branch(lu, primary, secondary, size, names, iv_name,
3122
                         p_minor, s_minor):
3123
  """Generate a drbd8 device complete with its children.
3124

3125
  """
3126
  port = lu.cfg.AllocatePort()
3127
  vgname = lu.cfg.GetVGName()
3128
  shared_secret = lu.cfg.GenerateDRBDSecret()
3129
  dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
3130
                          logical_id=(vgname, names[0]))
3131
  dev_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
3132
                          logical_id=(vgname, names[1]))
3133
  drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
3134
                          logical_id=(primary, secondary, port,
3135
                                      p_minor, s_minor,
3136
                                      shared_secret),
3137
                          children=[dev_data, dev_meta],
3138
                          iv_name=iv_name)
3139
  return drbd_dev
3140

    
3141

    
3142
def _GenerateDiskTemplate(lu, template_name,
3143
                          instance_name, primary_node,
3144
                          secondary_nodes, disk_info,
3145
                          file_storage_dir, file_driver):
3146
  """Generate the entire disk layout for a given template type.
3147

3148
  """
3149
  #TODO: compute space requirements
3150

    
3151
  vgname = lu.cfg.GetVGName()
3152
  disk_count = len(disk_info)
3153
  disks = []
3154
  if template_name == constants.DT_DISKLESS:
3155
    pass
3156
  elif template_name == constants.DT_PLAIN:
3157
    if len(secondary_nodes) != 0:
3158
      raise errors.ProgrammerError("Wrong template configuration")
3159

    
3160
    names = _GenerateUniqueNames(lu, [".disk%d" % i
3161
                                      for i in range(disk_count)])
3162
    for idx, disk in enumerate(disk_info):
3163
      disk_dev = objects.Disk(dev_type=constants.LD_LV, size=disk["size"],
3164
                              logical_id=(vgname, names[idx]),
3165
                              iv_name = "disk/%d" % idx)
3166
      disks.append(disk_dev)
3167
  elif template_name == constants.DT_DRBD8:
3168
    if len(secondary_nodes) != 1:
3169
      raise errors.ProgrammerError("Wrong template configuration")
3170
    remote_node = secondary_nodes[0]
3171
    minors = lu.cfg.AllocateDRBDMinor(
3172
      [primary_node, remote_node] * len(disk_info), instance_name)
3173

    
3174
    names = _GenerateUniqueNames(lu,
3175
                                 [".disk%d_%s" % (i, s)
3176
                                  for i in range(disk_count)
3177
                                  for s in ("data", "meta")
3178
                                  ])
3179
    for idx, disk in enumerate(disk_info):
3180
      disk_dev = _GenerateDRBD8Branch(lu, primary_node, remote_node,
3181
                                      disk["size"], names[idx*2:idx*2+2],
3182
                                      "disk/%d" % idx,
3183
                                      minors[idx*2], minors[idx*2+1])
3184
      disks.append(disk_dev)
3185
  elif template_name == constants.DT_FILE:
3186
    if len(secondary_nodes) != 0:
3187
      raise errors.ProgrammerError("Wrong template configuration")
3188

    
3189
    for idx, disk in enumerate(disk_info):
3190

    
3191
      disk_dev = objects.Disk(dev_type=constants.LD_FILE, size=disk["size"],
3192
                              iv_name="disk/%d" % idx,
3193
                              logical_id=(file_driver,
3194
                                          "%s/disk%d" % (file_storage_dir,
3195
                                                         idx)))
3196
      disks.append(disk_dev)
3197
  else:
3198
    raise errors.ProgrammerError("Invalid disk template '%s'" % template_name)
3199
  return disks
3200

    
3201

    
3202
def _GetInstanceInfoText(instance):
3203
  """Compute that text that should be added to the disk's metadata.
3204

3205
  """
3206
  return "originstname+%s" % instance.name
3207

    
3208

    
3209
def _CreateDisks(lu, instance):
3210
  """Create all disks for an instance.
3211

3212
  This abstracts away some work from AddInstance.
3213

3214
  Args:
3215
    instance: the instance object
3216

3217
  Returns:
3218
    True or False showing the success of the creation process
3219

3220
  """
3221
  info = _GetInstanceInfoText(instance)
3222

    
3223
  if instance.disk_template == constants.DT_FILE:
3224
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
3225
    result = lu.rpc.call_file_storage_dir_create(instance.primary_node,
3226
                                                 file_storage_dir)
3227

    
3228
    if not result:
3229
      logging.error("Could not connect to node '%s'", instance.primary_node)
3230
      return False
3231

    
3232
    if not result[0]:
3233
      logging.error("Failed to create directory '%s'", file_storage_dir)
3234
      return False
3235

    
3236
  for device in instance.disks:
3237
    logging.info("Creating volume %s for instance %s",
3238
                 device.iv_name, instance.name)
3239
    #HARDCODE
3240
    for secondary_node in instance.secondary_nodes:
3241
      if not _CreateBlockDevOnSecondary(lu, secondary_node, instance,
3242
                                        device, False, info):
3243
        logging.error("Failed to create volume %s (%s) on secondary node %s!",
3244
                      device.iv_name, device, secondary_node)
3245
        return False
3246
    #HARDCODE
3247
    if not _CreateBlockDevOnPrimary(lu, instance.primary_node,
3248
                                    instance, device, info):
3249
      logging.error("Failed to create volume %s on primary!", device.iv_name)
3250
      return False
3251

    
3252
  return True
3253

    
3254

    
3255
def _RemoveDisks(lu, instance):
3256
  """Remove all disks for an instance.
3257

3258
  This abstracts away some work from `AddInstance()` and
3259
  `RemoveInstance()`. Note that in case some of the devices couldn't
3260
  be removed, the removal will continue with the other ones (compare
3261
  with `_CreateDisks()`).
3262

3263
  Args:
3264
    instance: the instance object
3265

3266
  Returns:
3267
    True or False showing the success of the removal proces
3268

3269
  """
3270
  logging.info("Removing block devices for instance %s", instance.name)
3271

    
3272
  result = True
3273
  for device in instance.disks:
3274
    for node, disk in device.ComputeNodeTree(instance.primary_node):
3275
      lu.cfg.SetDiskID(disk, node)
3276
      if not lu.rpc.call_blockdev_remove(node, disk):
3277
        lu.proc.LogWarning("Could not remove block device %s on node %s,"
3278
                           " continuing anyway", device.iv_name, node)
3279
        result = False
3280

    
3281
  if instance.disk_template == constants.DT_FILE:
3282
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
3283
    if not lu.rpc.call_file_storage_dir_remove(instance.primary_node,
3284
                                               file_storage_dir):
3285
      logging.error("Could not remove directory '%s'", file_storage_dir)
3286
      result = False
3287

    
3288
  return result
3289

    
3290

    
3291
def _ComputeDiskSize(disk_template, disks):
3292
  """Compute disk size requirements in the volume group
3293

3294
  This is currently hard-coded for the two-drive layout.
3295

3296
  """
3297
  # Required free disk space as a function of disk and swap space
3298
  req_size_dict = {
3299
    constants.DT_DISKLESS: None,
3300
    constants.DT_PLAIN: sum(d["size"] for d in disks),
3301
    # 128 MB are added for drbd metadata for each disk
3302
    constants.DT_DRBD8: sum(d["size"] + 128 for d in disks),
3303
    constants.DT_FILE: None,
3304
  }
3305

    
3306
  if disk_template not in req_size_dict:
3307
    raise errors.ProgrammerError("Disk template '%s' size requirement"
3308
                                 " is unknown" %  disk_template)
3309

    
3310
  return req_size_dict[disk_template]
3311

    
3312

    
3313
def _CheckHVParams(lu, nodenames, hvname, hvparams):
3314
  """Hypervisor parameter validation.
3315

3316
  This function abstract the hypervisor parameter validation to be
3317
  used in both instance create and instance modify.
3318

3319
  @type lu: L{LogicalUnit}
3320
  @param lu: the logical unit for which we check
3321
  @type nodenames: list
3322
  @param nodenames: the list of nodes on which we should check
3323
  @type hvname: string
3324
  @param hvname: the name of the hypervisor we should use
3325
  @type hvparams: dict
3326
  @param hvparams: the parameters which we need to check
3327
  @raise errors.OpPrereqError: if the parameters are not valid
3328

3329
  """
3330
  hvinfo = lu.rpc.call_hypervisor_validate_params(nodenames,
3331
                                                  hvname,
3332
                                                  hvparams)
3333
  for node in nodenames:
3334
    info = hvinfo.get(node, None)
3335
    if not info or not isinstance(info, (tuple, list)):
3336
      raise errors.OpPrereqError("Cannot get current information"
3337
                                 " from node '%s' (%s)" % (node, info))
3338
    if not info[0]:
3339
      raise errors.OpPrereqError("Hypervisor parameter validation failed:"
3340
                                 " %s" % info[1])
3341

    
3342

    
3343
class LUCreateInstance(LogicalUnit):
3344
  """Create an instance.
3345

3346
  """
3347
  HPATH = "instance-add"
3348
  HTYPE = constants.HTYPE_INSTANCE
3349
  _OP_REQP = ["instance_name", "disks", "disk_template",
3350
              "mode", "start",
3351
              "wait_for_sync", "ip_check", "nics",
3352
              "hvparams", "beparams"]
3353
  REQ_BGL = False
3354

    
3355
  def _ExpandNode(self, node):
3356
    """Expands and checks one node name.
3357

3358
    """
3359
    node_full = self.cfg.ExpandNodeName(node)
3360
    if node_full is None:
3361
      raise errors.OpPrereqError("Unknown node %s" % node)
3362
    return node_full
3363

    
3364
  def ExpandNames(self):
3365
    """ExpandNames for CreateInstance.
3366

3367
    Figure out the right locks for instance creation.
3368

3369
    """
3370
    self.needed_locks = {}
3371

    
3372
    # set optional parameters to none if they don't exist
3373
    for attr in ["pnode", "snode", "iallocator", "hypervisor"]:
3374
      if not hasattr(self.op, attr):
3375
        setattr(self.op, attr, None)
3376

    
3377
    # cheap checks, mostly valid constants given
3378

    
3379
    # verify creation mode
3380
    if self.op.mode not in (constants.INSTANCE_CREATE,
3381
                            constants.INSTANCE_IMPORT):
3382
      raise errors.OpPrereqError("Invalid instance creation mode '%s'" %
3383
                                 self.op.mode)
3384

    
3385
    # disk template and mirror node verification
3386
    if self.op.disk_template not in constants.DISK_TEMPLATES:
3387
      raise errors.OpPrereqError("Invalid disk template name")
3388

    
3389
    if self.op.hypervisor is None:
3390
      self.op.hypervisor = self.cfg.GetHypervisorType()
3391

    
3392
    cluster = self.cfg.GetClusterInfo()
3393
    enabled_hvs = cluster.enabled_hypervisors
3394
    if self.op.hypervisor not in enabled_hvs:
3395
      raise errors.OpPrereqError("Selected hypervisor (%s) not enabled in the"
3396
                                 " cluster (%s)" % (self.op.hypervisor,
3397
                                  ",".join(enabled_hvs)))
3398

    
3399
    # check hypervisor parameter syntax (locally)
3400

    
3401
    filled_hvp = cluster.FillDict(cluster.hvparams[self.op.hypervisor],
3402
                                  self.op.hvparams)
3403
    hv_type = hypervisor.GetHypervisor(self.op.hypervisor)
3404
    hv_type.CheckParameterSyntax(filled_hvp)
3405

    
3406
    # fill and remember the beparams dict
3407
    self.be_full = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
3408
                                    self.op.beparams)
3409

    
3410
    #### instance parameters check
3411

    
3412
    # instance name verification
3413
    hostname1 = utils.HostInfo(self.op.instance_name)
3414
    self.op.instance_name = instance_name = hostname1.name
3415

    
3416
    # this is just a preventive check, but someone might still add this
3417
    # instance in the meantime, and creation will fail at lock-add time
3418
    if instance_name in self.cfg.GetInstanceList():
3419
      raise errors.OpPrereqError("Instance '%s' is already in the cluster" %
3420
                                 instance_name)
3421

    
3422
    self.add_locks[locking.LEVEL_INSTANCE] = instance_name
3423

    
3424
    # NIC buildup
3425
    self.nics = []
3426
    for nic in self.op.nics:
3427
      # ip validity checks
3428
      ip = nic.get("ip", None)
3429
      if ip is None or ip.lower() == "none":
3430
        nic_ip = None
3431
      elif ip.lower() == constants.VALUE_AUTO:
3432
        nic_ip = hostname1.ip
3433
      else:
3434
        if not utils.IsValidIP(ip):
3435
          raise errors.OpPrereqError("Given IP address '%s' doesn't look"
3436
                                     " like a valid IP" % ip)
3437
        nic_ip = ip
3438

    
3439
      # MAC address verification
3440
      mac = nic.get("mac", constants.VALUE_AUTO)
3441
      if mac not in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
3442
        if not utils.IsValidMac(mac.lower()):
3443
          raise errors.OpPrereqError("Invalid MAC address specified: %s" %
3444
                                     mac)
3445
      # bridge verification
3446
      bridge = nic.get("bridge", self.cfg.GetDefBridge())
3447
      self.nics.append(objects.NIC(mac=mac, ip=nic_ip, bridge=bridge))
3448

    
3449
    # disk checks/pre-build
3450
    self.disks = []
3451
    for disk in self.op.disks:
3452
      mode = disk.get("mode", constants.DISK_RDWR)
3453
      if mode not in constants.DISK_ACCESS_SET:
3454
        raise errors.OpPrereqError("Invalid disk access mode '%s'" %
3455
                                   mode)
3456
      size = disk.get("size", None)
3457
      if size is None:
3458
        raise errors.OpPrereqError("Missing disk size")
3459
      try:
3460
        size = int(size)
3461
      except ValueError:
3462
        raise errors.OpPrereqError("Invalid disk size '%s'" % size)
3463
      self.disks.append({"size": size, "mode": mode})
3464

    
3465
    # used in CheckPrereq for ip ping check
3466
    self.check_ip = hostname1.ip
3467

    
3468
    # file storage checks
3469
    if (self.op.file_driver and
3470
        not self.op.file_driver in constants.FILE_DRIVER):
3471
      raise errors.OpPrereqError("Invalid file driver name '%s'" %
3472
                                 self.op.file_driver)
3473

    
3474
    if self.op.file_storage_dir and os.path.isabs(self.op.file_storage_dir):
3475
      raise errors.OpPrereqError("File storage directory path not absolute")
3476

    
3477
    ### Node/iallocator related checks
3478
    if [self.op.iallocator, self.op.pnode].count(None) != 1:
3479
      raise errors.OpPrereqError("One and only one of iallocator and primary"
3480
                                 " node must be given")
3481

    
3482
    if self.op.iallocator:
3483
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3484
    else:
3485
      self.op.pnode = self._ExpandNode(self.op.pnode)
3486
      nodelist = [self.op.pnode]
3487
      if self.op.snode is not None:
3488
        self.op.snode = self._ExpandNode(self.op.snode)
3489
        nodelist.append(self.op.snode)
3490
      self.needed_locks[locking.LEVEL_NODE] = nodelist
3491

    
3492
    # in case of import lock the source node too
3493
    if self.op.mode == constants.INSTANCE_IMPORT:
3494
      src_node = getattr(self.op, "src_node", None)
3495
      src_path = getattr(self.op, "src_path", None)
3496

    
3497
      if src_node is None or src_path is None:
3498
        raise errors.OpPrereqError("Importing an instance requires source"
3499
                                   " node and path options")
3500

    
3501
      if not os.path.isabs(src_path):
3502
        raise errors.OpPrereqError("The source path must be absolute")
3503

    
3504
      self.op.src_node = src_node = self._ExpandNode(src_node)
3505
      if self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET:
3506
        self.needed_locks[locking.LEVEL_NODE].append(src_node)
3507

    
3508
    else: # INSTANCE_CREATE
3509
      if getattr(self.op, "os_type", None) is None:
3510
        raise errors.OpPrereqError("No guest OS specified")
3511

    
3512
  def _RunAllocator(self):
3513
    """Run the allocator based on input opcode.
3514

3515
    """
3516
    nics = [n.ToDict() for n in self.nics]
3517
    ial = IAllocator(self,
3518
                     mode=constants.IALLOCATOR_MODE_ALLOC,
3519
                     name=self.op.instance_name,
3520
                     disk_template=self.op.disk_template,
3521
                     tags=[],
3522
                     os=self.op.os_type,
3523
                     vcpus=self.be_full[constants.BE_VCPUS],
3524
                     mem_size=self.be_full[constants.BE_MEMORY],
3525
                     disks=self.disks,
3526
                     nics=nics,
3527
                     )
3528

    
3529
    ial.Run(self.op.iallocator)
3530

    
3531
    if not ial.success:
3532
      raise errors.OpPrereqError("Can't compute nodes using"
3533
                                 " iallocator '%s': %s" % (self.op.iallocator,
3534
                                                           ial.info))
3535
    if len(ial.nodes) != ial.required_nodes:
3536
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
3537
                                 " of nodes (%s), required %s" %
3538
                                 (self.op.iallocator, len(ial.nodes),
3539
                                  ial.required_nodes))
3540
    self.op.pnode = ial.nodes[0]
3541
    self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
3542
                 self.op.instance_name, self.op.iallocator,
3543
                 ", ".join(ial.nodes))
3544
    if ial.required_nodes == 2:
3545
      self.op.snode = ial.nodes[1]
3546

    
3547
  def BuildHooksEnv(self):
3548
    """Build hooks env.
3549

3550
    This runs on master, primary and secondary nodes of the instance.
3551

3552
    """
3553
    env = {
3554
      "INSTANCE_DISK_TEMPLATE": self.op.disk_template,
3555
      "INSTANCE_DISK_SIZE": ",".join(str(d["size"]) for d in self.disks),
3556
      "INSTANCE_ADD_MODE": self.op.mode,
3557
      }
3558
    if self.op.mode == constants.INSTANCE_IMPORT:
3559
      env["INSTANCE_SRC_NODE"] = self.op.src_node
3560
      env["INSTANCE_SRC_PATH"] = self.op.src_path
3561
      env["INSTANCE_SRC_IMAGES"] = self.src_images
3562

    
3563
    env.update(_BuildInstanceHookEnv(name=self.op.instance_name,
3564
      primary_node=self.op.pnode,
3565
      secondary_nodes=self.secondaries,
3566
      status=self.instance_status,
3567
      os_type=self.op.os_type,
3568
      memory=self.be_full[constants.BE_MEMORY],
3569
      vcpus=self.be_full[constants.BE_VCPUS],
3570
      nics=[(n.ip, n.bridge, n.mac) for n in self.nics],
3571
    ))
3572

    
3573
    nl = ([self.cfg.GetMasterNode(), self.op.pnode] +
3574
          self.secondaries)
3575
    return env, nl, nl
3576

    
3577

    
3578
  def CheckPrereq(self):
3579
    """Check prerequisites.
3580

3581
    """
3582
    if (not self.cfg.GetVGName() and
3583
        self.op.disk_template not in constants.DTS_NOT_LVM):
3584
      raise errors.OpPrereqError("Cluster does not support lvm-based"
3585
                                 " instances")
3586

    
3587

    
3588
    if self.op.mode == constants.INSTANCE_IMPORT:
3589
      src_node = self.op.src_node
3590
      src_path = self.op.src_path
3591

    
3592
      export_info = self.rpc.call_export_info(src_node, src_path)
3593

    
3594
      if not export_info:
3595
        raise errors.OpPrereqError("No export found in dir %s" % src_path)
3596

    
3597
      if not export_info.has_section(constants.INISECT_EXP):
3598
        raise errors.ProgrammerError("Corrupted export config")
3599

    
3600
      ei_version = export_info.get(constants.INISECT_EXP, 'version')
3601
      if (int(ei_version) != constants.EXPORT_VERSION):
3602
        raise errors.OpPrereqError("Wrong export version %s (wanted %d)" %
3603
                                   (ei_version, constants.EXPORT_VERSION))
3604

    
3605
      # Check that the new instance doesn't have less disks than the export
3606
      instance_disks = len(self.disks)
3607
      export_disks = export_info.getint(constants.INISECT_INS, 'disk_count')
3608
      if instance_disks < export_disks:
3609
        raise errors.OpPrereqError("Not enough disks to import."
3610
                                   " (instance: %d, export: %d)" %
3611
                                   (2, export_disks))
3612

    
3613
      self.op.os_type = export_info.get(constants.INISECT_EXP, 'os')
3614
      disk_images = []
3615
      for idx in range(export_disks):
3616
        option = 'disk%d_dump' % idx
3617
        if export_info.has_option(constants.INISECT_INS, option):
3618
          # FIXME: are the old os-es, disk sizes, etc. useful?
3619
          export_name = export_info.get(constants.INISECT_INS, option)
3620
          image = os.path.join(src_path, export_name)
3621
          disk_images.append(image)
3622
        else:
3623
          disk_images.append(False)
3624

    
3625
      self.src_images = disk_images
3626

    
3627
      if self.op.mac == constants.VALUE_AUTO:
3628
        old_name = export_info.get(constants.INISECT_INS, 'name')
3629
        if self.op.instance_name == old_name:
3630
          # FIXME: adjust every nic, when we'll be able to create instances
3631
          # with more than one
3632
          if int(export_info.get(constants.INISECT_INS, 'nic_count')) >= 1:
3633
            self.op.mac = export_info.get(constants.INISECT_INS, 'nic_0_mac')
3634

    
3635
    # ip ping checks (we use the same ip that was resolved in ExpandNames)
3636

    
3637
    if self.op.start and not self.op.ip_check:
3638
      raise errors.OpPrereqError("Cannot ignore IP address conflicts when"
3639
                                 " adding an instance in start mode")
3640

    
3641
    if self.op.ip_check:
3642
      if utils.TcpPing(self.check_ip, constants.DEFAULT_NODED_PORT):
3643
        raise errors.OpPrereqError("IP %s of instance %s already in use" %
3644
                                   (self.check_ip, self.op.instance_name))
3645

    
3646
    #### allocator run
3647

    
3648
    if self.op.iallocator is not None:
3649
      self._RunAllocator()
3650

    
3651
    #### node related checks
3652

    
3653
    # check primary node
3654
    self.pnode = pnode = self.cfg.GetNodeInfo(self.op.pnode)
3655
    assert self.pnode is not None, \
3656
      "Cannot retrieve locked node %s" % self.op.pnode
3657
    self.secondaries = []
3658

    
3659
    # mirror node verification
3660
    if self.op.disk_template in constants.DTS_NET_MIRROR:
3661
      if self.op.snode is None:
3662
        raise errors.OpPrereqError("The networked disk templates need"
3663
                                   " a mirror node")
3664
      if self.op.snode == pnode.name:
3665
        raise errors.OpPrereqError("The secondary node cannot be"
3666
                                   " the primary node.")
3667
      self.secondaries.append(self.op.snode)
3668

    
3669
    nodenames = [pnode.name] + self.secondaries
3670

    
3671
    req_size = _ComputeDiskSize(self.op.disk_template,
3672
                                self.disks)
3673

    
3674
    # Check lv size requirements
3675
    if req_size is not None:
3676
      nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
3677
                                         self.op.hypervisor)
3678
      for node in nodenames:
3679
        info = nodeinfo.get(node, None)
3680
        if not info:
3681
          raise errors.OpPrereqError("Cannot get current information"
3682
                                     " from node '%s'" % node)
3683
        vg_free = info.get('vg_free', None)
3684
        if not isinstance(vg_free, int):
3685
          raise errors.OpPrereqError("Can't compute free disk space on"
3686
                                     " node %s" % node)
3687
        if req_size > info['vg_free']:
3688
          raise errors.OpPrereqError("Not enough disk space on target node %s."
3689
                                     " %d MB available, %d MB required" %
3690
                                     (node, info['vg_free'], req_size))
3691

    
3692
    _CheckHVParams(self, nodenames, self.op.hypervisor, self.op.hvparams)
3693

    
3694
    # os verification
3695
    os_obj = self.rpc.call_os_get(pnode.name, self.op.os_type)
3696
    if not os_obj:
3697
      raise errors.OpPrereqError("OS '%s' not in supported os list for"
3698
                                 " primary node"  % self.op.os_type)
3699

    
3700
    # bridge check on primary node
3701
    bridges = [n.bridge for n in self.nics]
3702
    if not self.rpc.call_bridges_exist(self.pnode.name, bridges):
3703
      raise errors.OpPrereqError("one of the target bridges '%s' does not"
3704
                                 " exist on"
3705
                                 " destination node '%s'" %
3706
                                 (",".join(bridges), pnode.name))
3707

    
3708
    # memory check on primary node
3709
    if self.op.start:
3710
      _CheckNodeFreeMemory(self, self.pnode.name,
3711
                           "creating instance %s" % self.op.instance_name,
3712
                           self.be_full[constants.BE_MEMORY],
3713
                           self.op.hypervisor)
3714

    
3715
    if self.op.start:
3716
      self.instance_status = 'up'
3717
    else:
3718
      self.instance_status = 'down'
3719

    
3720
  def Exec(self, feedback_fn):
3721
    """Create and add the instance to the cluster.
3722

3723
    """
3724
    instance = self.op.instance_name
3725
    pnode_name = self.pnode.name
3726

    
3727
    for nic in self.nics:
3728
      if nic.mac in (constants.VALUE_AUTO, constants.VALUE_GENERATE):
3729
        nic.mac = self.cfg.GenerateMAC()
3730

    
3731
    ht_kind = self.op.hypervisor
3732
    if ht_kind in constants.HTS_REQ_PORT:
3733
      network_port = self.cfg.AllocatePort()
3734
    else:
3735
      network_port = None
3736

    
3737
    ##if self.op.vnc_bind_address is None:
3738
    ##  self.op.vnc_bind_address = constants.VNC_DEFAULT_BIND_ADDRESS
3739

    
3740
    # this is needed because os.path.join does not accept None arguments
3741
    if self.op.file_storage_dir is None:
3742
      string_file_storage_dir = ""
3743
    else:
3744
      string_file_storage_dir = self.op.file_storage_dir
3745

    
3746
    # build the full file storage dir path
3747
    file_storage_dir = os.path.normpath(os.path.join(
3748
                                        self.cfg.GetFileStorageDir(),
3749
                                        string_file_storage_dir, instance))
3750

    
3751

    
3752
    disks = _GenerateDiskTemplate(self,
3753
                                  self.op.disk_template,
3754
                                  instance, pnode_name,
3755
                                  self.secondaries,
3756
                                  self.disks,
3757
                                  file_storage_dir,
3758
                                  self.op.file_driver)
3759

    
3760
    iobj = objects.Instance(name=instance, os=self.op.os_type,
3761
                            primary_node=pnode_name,
3762
                            nics=self.nics, disks=disks,
3763
                            disk_template=self.op.disk_template,
3764
                            status=self.instance_status,
3765
                            network_port=network_port,
3766
                            beparams=self.op.beparams,
3767
                            hvparams=self.op.hvparams,
3768
                            hypervisor=self.op.hypervisor,
3769
                            )
3770

    
3771
    feedback_fn("* creating instance disks...")
3772
    if not _CreateDisks(self, iobj):
3773
      _RemoveDisks(self, iobj)
3774
      self.cfg.ReleaseDRBDMinors(instance)
3775
      raise errors.OpExecError("Device creation failed, reverting...")
3776

    
3777
    feedback_fn("adding instance %s to cluster config" % instance)
3778

    
3779
    self.cfg.AddInstance(iobj)
3780
    # Declare that we don't want to remove the instance lock anymore, as we've
3781
    # added the instance to the config
3782
    del self.remove_locks[locking.LEVEL_INSTANCE]
3783
    # Remove the temp. assignements for the instance's drbds
3784
    self.cfg.ReleaseDRBDMinors(instance)
3785

    
3786
    if self.op.wait_for_sync:
3787
      disk_abort = not _WaitForSync(self, iobj)
3788
    elif iobj.disk_template in constants.DTS_NET_MIRROR:
3789
      # make sure the disks are not degraded (still sync-ing is ok)
3790
      time.sleep(15)
3791
      feedback_fn("* checking mirrors status")
3792
      disk_abort = not _WaitForSync(self, iobj, oneshot=True)
3793
    else:
3794
      disk_abort = False
3795

    
3796
    if disk_abort:
3797
      _RemoveDisks(self, iobj)
3798
      self.cfg.RemoveInstance(iobj.name)
3799
      # Make sure the instance lock gets removed
3800
      self.remove_locks[locking.LEVEL_INSTANCE] = iobj.name
3801
      raise errors.OpExecError("There are some degraded disks for"
3802
                               " this instance")
3803

    
3804
    feedback_fn("creating os for instance %s on node %s" %
3805
                (instance, pnode_name))
3806

    
3807
    if iobj.disk_template != constants.DT_DISKLESS:
3808
      if self.op.mode == constants.INSTANCE_CREATE:
3809
        feedback_fn("* running the instance OS create scripts...")
3810
        if not self.rpc.call_instance_os_add(pnode_name, iobj):
3811
          raise errors.OpExecError("could not add os for instance %s"
3812
                                   " on node %s" %
3813
                                   (instance, pnode_name))
3814

    
3815
      elif self.op.mode == constants.INSTANCE_IMPORT:
3816
        feedback_fn("* running the instance OS import scripts...")
3817
        src_node = self.op.src_node
3818
        src_images = self.src_images
3819
        cluster_name = self.cfg.GetClusterName()
3820
        import_result = self.rpc.call_instance_os_import(pnode_name, iobj,
3821
                                                         src_node, src_images,
3822
                                                         cluster_name)
3823
        for idx, result in enumerate(import_result):
3824
          if not result:
3825
            self.LogWarning("Could not image %s for on instance %s, disk %d,"
3826
                            " on node %s" % (src_images[idx], instance, idx,
3827
                                             pnode_name))
3828
      else:
3829
        # also checked in the prereq part
3830
        raise errors.ProgrammerError("Unknown OS initialization mode '%s'"
3831
                                     % self.op.mode)
3832

    
3833
    if self.op.start:
3834
      logging.info("Starting instance %s on node %s", instance, pnode_name)
3835
      feedback_fn("* starting instance...")
3836
      if not self.rpc.call_instance_start(pnode_name, iobj, None):
3837
        raise errors.OpExecError("Could not start instance")
3838

    
3839

    
3840
class LUConnectConsole(NoHooksLU):
3841
  """Connect to an instance's console.
3842

3843
  This is somewhat special in that it returns the command line that
3844
  you need to run on the master node in order to connect to the
3845
  console.
3846

3847
  """
3848
  _OP_REQP = ["instance_name"]
3849
  REQ_BGL = False
3850

    
3851
  def ExpandNames(self):
3852
    self._ExpandAndLockInstance()
3853

    
3854
  def CheckPrereq(self):
3855
    """Check prerequisites.
3856

3857
    This checks that the instance is in the cluster.
3858

3859
    """
3860
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3861
    assert self.instance is not None, \
3862
      "Cannot retrieve locked instance %s" % self.op.instance_name
3863

    
3864
  def Exec(self, feedback_fn):
3865
    """Connect to the console of an instance
3866

3867
    """
3868
    instance = self.instance
3869
    node = instance.primary_node
3870

    
3871
    node_insts = self.rpc.call_instance_list([node],
3872
                                             [instance.hypervisor])[node]
3873
    if node_insts is False:
3874
      raise errors.OpExecError("Can't connect to node %s." % node)
3875

    
3876
    if instance.name not in node_insts:
3877
      raise errors.OpExecError("Instance %s is not running." % instance.name)
3878

    
3879
    logging.debug("Connecting to console of %s on %s", instance.name, node)
3880

    
3881
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
3882
    console_cmd = hyper.GetShellCommandForConsole(instance)
3883

    
3884
    # build ssh cmdline
3885
    return self.ssh.BuildCmd(node, "root", console_cmd, batch=True, tty=True)
3886

    
3887

    
3888
class LUReplaceDisks(LogicalUnit):
3889
  """Replace the disks of an instance.
3890

3891
  """
3892
  HPATH = "mirrors-replace"
3893
  HTYPE = constants.HTYPE_INSTANCE
3894
  _OP_REQP = ["instance_name", "mode", "disks"]
3895
  REQ_BGL = False
3896

    
3897
  def ExpandNames(self):
3898
    self._ExpandAndLockInstance()
3899

    
3900
    if not hasattr(self.op, "remote_node"):
3901
      self.op.remote_node = None
3902

    
3903
    ia_name = getattr(self.op, "iallocator", None)
3904
    if ia_name is not None:
3905
      if self.op.remote_node is not None:
3906
        raise errors.OpPrereqError("Give either the iallocator or the new"
3907
                                   " secondary, not both")
3908
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
3909
    elif self.op.remote_node is not None:
3910
      remote_node = self.cfg.ExpandNodeName(self.op.remote_node)
3911
      if remote_node is None:
3912
        raise errors.OpPrereqError("Node '%s' not known" %
3913
                                   self.op.remote_node)
3914
      self.op.remote_node = remote_node
3915
      self.needed_locks[locking.LEVEL_NODE] = [remote_node]
3916
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
3917
    else:
3918
      self.needed_locks[locking.LEVEL_NODE] = []
3919
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
3920

    
3921
  def DeclareLocks(self, level):
3922
    # If we're not already locking all nodes in the set we have to declare the
3923
    # instance's primary/secondary nodes.
3924
    if (level == locking.LEVEL_NODE and
3925
        self.needed_locks[locking.LEVEL_NODE] is not locking.ALL_SET):
3926
      self._LockInstancesNodes()
3927

    
3928
  def _RunAllocator(self):
3929
    """Compute a new secondary node using an IAllocator.
3930

3931
    """
3932
    ial = IAllocator(self,
3933
                     mode=constants.IALLOCATOR_MODE_RELOC,
3934
                     name=self.op.instance_name,
3935
                     relocate_from=[self.sec_node])
3936

    
3937
    ial.Run(self.op.iallocator)
3938

    
3939
    if not ial.success:
3940
      raise errors.OpPrereqError("Can't compute nodes using"
3941
                                 " iallocator '%s': %s" % (self.op.iallocator,
3942
                                                           ial.info))
3943
    if len(ial.nodes) != ial.required_nodes:
3944
      raise errors.OpPrereqError("iallocator '%s' returned invalid number"
3945
                                 " of nodes (%s), required %s" %
3946
                                 (len(ial.nodes), ial.required_nodes))
3947
    self.op.remote_node = ial.nodes[0]
3948
    self.LogInfo("Selected new secondary for the instance: %s",
3949
                 self.op.remote_node)
3950

    
3951
  def BuildHooksEnv(self):
3952
    """Build hooks env.
3953

3954
    This runs on the master, the primary and all the secondaries.
3955

3956
    """
3957
    env = {
3958
      "MODE": self.op.mode,
3959
      "NEW_SECONDARY": self.op.remote_node,
3960
      "OLD_SECONDARY": self.instance.secondary_nodes[0],
3961
      }
3962
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
3963
    nl = [
3964
      self.cfg.GetMasterNode(),
3965
      self.instance.primary_node,
3966
      ]
3967
    if self.op.remote_node is not None:
3968
      nl.append(self.op.remote_node)
3969
    return env, nl, nl
3970

    
3971
  def CheckPrereq(self):
3972
    """Check prerequisites.
3973

3974
    This checks that the instance is in the cluster.
3975

3976
    """
3977
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
3978
    assert instance is not None, \
3979
      "Cannot retrieve locked instance %s" % self.op.instance_name
3980
    self.instance = instance
3981

    
3982
    if instance.disk_template not in constants.DTS_NET_MIRROR:
3983
      raise errors.OpPrereqError("Instance's disk layout is not"
3984
                                 " network mirrored.")
3985

    
3986
    if len(instance.secondary_nodes) != 1:
3987
      raise errors.OpPrereqError("The instance has a strange layout,"
3988
                                 " expected one secondary but found %d" %
3989
                                 len(instance.secondary_nodes))
3990

    
3991
    self.sec_node = instance.secondary_nodes[0]
3992

    
3993
    ia_name = getattr(self.op, "iallocator", None)
3994
    if ia_name is not None:
3995
      self._RunAllocator()
3996

    
3997
    remote_node = self.op.remote_node
3998
    if remote_node is not None:
3999
      self.remote_node_info = self.cfg.GetNodeInfo(remote_node)
4000
      assert self.remote_node_info is not None, \
4001
        "Cannot retrieve locked node %s" % remote_node
4002
    else:
4003
      self.remote_node_info = None
4004
    if remote_node == instance.primary_node:
4005
      raise errors.OpPrereqError("The specified node is the primary node of"
4006
                                 " the instance.")
4007
    elif remote_node == self.sec_node:
4008
      if self.op.mode == constants.REPLACE_DISK_SEC:
4009
        # this is for DRBD8, where we can't execute the same mode of
4010
        # replacement as for drbd7 (no different port allocated)
4011
        raise errors.OpPrereqError("Same secondary given, cannot execute"
4012
                                   " replacement")
4013
    if instance.disk_template == constants.DT_DRBD8:
4014
      if (self.op.mode == constants.REPLACE_DISK_ALL and
4015
          remote_node is not None):
4016
        # switch to replace secondary mode
4017
        self.op.mode = constants.REPLACE_DISK_SEC
4018

    
4019
      if self.op.mode == constants.REPLACE_DISK_ALL:
4020
        raise errors.OpPrereqError("Template 'drbd' only allows primary or"
4021
                                   " secondary disk replacement, not"
4022
                                   " both at once")
4023
      elif self.op.mode == constants.REPLACE_DISK_PRI:
4024
        if remote_node is not None:
4025
          raise errors.OpPrereqError("Template 'drbd' does not allow changing"
4026
                                     " the secondary while doing a primary"
4027
                                     " node disk replacement")
4028
        self.tgt_node = instance.primary_node
4029
        self.oth_node = instance.secondary_nodes[0]
4030
      elif self.op.mode == constants.REPLACE_DISK_SEC:
4031
        self.new_node = remote_node # this can be None, in which case
4032
                                    # we don't change the secondary
4033
        self.tgt_node = instance.secondary_nodes[0]
4034
        self.oth_node = instance.primary_node
4035
      else:
4036
        raise errors.ProgrammerError("Unhandled disk replace mode")
4037

    
4038
    if not self.op.disks:
4039
      self.op.disks = range(len(instance.disks))
4040

    
4041
    for disk_idx in self.op.disks:
4042
      if disk_idx < 0 or disk_idx >= len(instance.disks):
4043
        raise errors.OpPrereqError("Disk '%s' not found for instance '%s'" %
4044
                                   (name, instance.name))
4045

    
4046
  def _ExecD8DiskOnly(self, feedback_fn):
4047
    """Replace a disk on the primary or secondary for dbrd8.
4048

4049
    The algorithm for replace is quite complicated:
4050
      - for each disk to be replaced:
4051
        - create new LVs on the target node with unique names
4052
        - detach old LVs from the drbd device
4053
        - rename old LVs to name_replaced.<time_t>
4054
        - rename new LVs to old LVs
4055
        - attach the new LVs (with the old names now) to the drbd device
4056
      - wait for sync across all devices
4057
      - for each modified disk:
4058
        - remove old LVs (which have the name name_replaces.<time_t>)
4059

4060
    Failures are not very well handled.
4061

4062
    """
4063
    steps_total = 6
4064
    warning, info = (self.proc.LogWarning, self.proc.LogInfo)
4065
    instance = self.instance
4066
    iv_names = {}
4067
    vgname = self.cfg.GetVGName()
4068
    # start of work
4069
    cfg = self.cfg
4070
    tgt_node = self.tgt_node
4071
    oth_node = self.oth_node
4072

    
4073
    # Step: check device activation
4074
    self.proc.LogStep(1, steps_total, "check device existence")
4075
    info("checking volume groups")
4076
    my_vg = cfg.GetVGName()
4077
    results = self.rpc.call_vg_list([oth_node, tgt_node])
4078
    if not results:
4079
      raise errors.OpExecError("Can't list volume groups on the nodes")
4080
    for node in oth_node, tgt_node:
4081
      res = results.get(node, False)
4082
      if not res or my_vg not in res:
4083
        raise errors.OpExecError("Volume group '%s' not found on %s" %
4084
                                 (my_vg, node))
4085
    for idx, dev in enumerate(instance.disks):
4086
      if idx not in self.op.disks:
4087
        continue
4088
      for node in tgt_node, oth_node:
4089
        info("checking disk/%d on %s" % (idx, node))
4090
        cfg.SetDiskID(dev, node)
4091
        if not self.rpc.call_blockdev_find(node, dev):
4092
          raise errors.OpExecError("Can't find disk/%d on node %s" %
4093
                                   (idx, node))
4094

    
4095
    # Step: check other node consistency
4096
    self.proc.LogStep(2, steps_total, "check peer consistency")
4097
    for idx, dev in enumerate(instance.disks):
4098
      if idx not in self.op.disks:
4099
        continue
4100
      info("checking disk/%d consistency on %s" % (idx, oth_node))
4101
      if not _CheckDiskConsistency(self, dev, oth_node,
4102
                                   oth_node==instance.primary_node):
4103
        raise errors.OpExecError("Peer node (%s) has degraded storage, unsafe"
4104
                                 " to replace disks on this node (%s)" %
4105
                                 (oth_node, tgt_node))
4106

    
4107
    # Step: create new storage
4108
    self.proc.LogStep(3, steps_total, "allocate new storage")
4109
    for idx, dev in enumerate(instance.disks):
4110
      if idx not in self.op.disks:
4111
        continue
4112
      size = dev.size
4113
      cfg.SetDiskID(dev, tgt_node)
4114
      lv_names = [".disk%d_%s" % (idx, suf)
4115
                  for suf in ["data", "meta"]]
4116
      names = _GenerateUniqueNames(self, lv_names)
4117
      lv_data = objects.Disk(dev_type=constants.LD_LV, size=size,
4118
                             logical_id=(vgname, names[0]))
4119
      lv_meta = objects.Disk(dev_type=constants.LD_LV, size=128,
4120
                             logical_id=(vgname, names[1]))
4121
      new_lvs = [lv_data, lv_meta]
4122
      old_lvs = dev.children
4123
      iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
4124
      info("creating new local storage on %s for %s" %
4125
           (tgt_node, dev.iv_name))
4126
      # since we *always* want to create this LV, we use the
4127
      # _Create...OnPrimary (which forces the creation), even if we
4128
      # are talking about the secondary node
4129
      for new_lv in new_lvs:
4130
        if not _CreateBlockDevOnPrimary(self, tgt_node, instance, new_lv,
4131
                                        _GetInstanceInfoText(instance)):
4132
          raise errors.OpExecError("Failed to create new LV named '%s' on"
4133
                                   " node '%s'" %
4134
                                   (new_lv.logical_id[1], tgt_node))
4135

    
4136
    # Step: for each lv, detach+rename*2+attach
4137
    self.proc.LogStep(4, steps_total, "change drbd configuration")
4138
    for dev, old_lvs, new_lvs in iv_names.itervalues():
4139
      info("detaching %s drbd from local storage" % dev.iv_name)
4140
      if not self.rpc.call_blockdev_removechildren(tgt_node, dev, old_lvs):
4141
        raise errors.OpExecError("Can't detach drbd from local storage on node"
4142
                                 " %s for device %s" % (tgt_node, dev.iv_name))
4143
      #dev.children = []
4144
      #cfg.Update(instance)
4145

    
4146
      # ok, we created the new LVs, so now we know we have the needed
4147
      # storage; as such, we proceed on the target node to rename
4148
      # old_lv to _old, and new_lv to old_lv; note that we rename LVs
4149
      # using the assumption that logical_id == physical_id (which in
4150
      # turn is the unique_id on that node)
4151

    
4152
      # FIXME(iustin): use a better name for the replaced LVs
4153
      temp_suffix = int(time.time())
4154
      ren_fn = lambda d, suff: (d.physical_id[0],
4155
                                d.physical_id[1] + "_replaced-%s" % suff)
4156
      # build the rename list based on what LVs exist on the node
4157
      rlist = []
4158
      for to_ren in old_lvs:
4159
        find_res = self.rpc.call_blockdev_find(tgt_node, to_ren)
4160
        if find_res is not None: # device exists
4161
          rlist.append((to_ren, ren_fn(to_ren, temp_suffix)))
4162

    
4163
      info("renaming the old LVs on the target node")
4164
      if not self.rpc.call_blockdev_rename(tgt_node, rlist):
4165
        raise errors.OpExecError("Can't rename old LVs on node %s" % tgt_node)
4166
      # now we rename the new LVs to the old LVs
4167
      info("renaming the new LVs on the target node")
4168
      rlist = [(new, old.physical_id) for old, new in zip(old_lvs, new_lvs)]
4169
      if not self.rpc.call_blockdev_rename(tgt_node, rlist):
4170
        raise errors.OpExecError("Can't rename new LVs on node %s" % tgt_node)
4171

    
4172
      for old, new in zip(old_lvs, new_lvs):
4173
        new.logical_id = old.logical_id
4174
        cfg.SetDiskID(new, tgt_node)
4175

    
4176
      for disk in old_lvs:
4177
        disk.logical_id = ren_fn(disk, temp_suffix)
4178
        cfg.SetDiskID(disk, tgt_node)
4179

    
4180
      # now that the new lvs have the old name, we can add them to the device
4181
      info("adding new mirror component on %s" % tgt_node)
4182
      if not self.rpc.call_blockdev_addchildren(tgt_node, dev, new_lvs):
4183
        for new_lv in new_lvs:
4184
          if not self.rpc.call_blockdev_remove(tgt_node, new_lv):
4185
            warning("Can't rollback device %s", hint="manually cleanup unused"
4186
                    " logical volumes")
4187
        raise errors.OpExecError("Can't add local storage to drbd")
4188

    
4189
      dev.children = new_lvs
4190
      cfg.Update(instance)
4191

    
4192
    # Step: wait for sync
4193

    
4194
    # this can fail as the old devices are degraded and _WaitForSync
4195
    # does a combined result over all disks, so we don't check its
4196
    # return value
4197
    self.proc.LogStep(5, steps_total, "sync devices")
4198
    _WaitForSync(self, instance, unlock=True)
4199

    
4200
    # so check manually all the devices
4201
    for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
4202
      cfg.SetDiskID(dev, instance.primary_node)
4203
      is_degr = self.rpc.call_blockdev_find(instance.primary_node, dev)[5]
4204
      if is_degr:
4205
        raise errors.OpExecError("DRBD device %s is degraded!" % name)
4206

    
4207
    # Step: remove old storage
4208
    self.proc.LogStep(6, steps_total, "removing old storage")
4209
    for name, (dev, old_lvs, new_lvs) in iv_names.iteritems():
4210
      info("remove logical volumes for %s" % name)
4211
      for lv in old_lvs:
4212
        cfg.SetDiskID(lv, tgt_node)
4213
        if not self.rpc.call_blockdev_remove(tgt_node, lv):
4214
          warning("Can't remove old LV", hint="manually remove unused LVs")
4215
          continue
4216

    
4217
  def _ExecD8Secondary(self, feedback_fn):
4218
    """Replace the secondary node for drbd8.
4219

4220
    The algorithm for replace is quite complicated:
4221
      - for all disks of the instance:
4222
        - create new LVs on the new node with same names
4223
        - shutdown the drbd device on the old secondary
4224
        - disconnect the drbd network on the primary
4225
        - create the drbd device on the new secondary
4226
        - network attach the drbd on the primary, using an artifice:
4227
          the drbd code for Attach() will connect to the network if it
4228
          finds a device which is connected to the good local disks but
4229
          not network enabled
4230
      - wait for sync across all devices
4231
      - remove all disks from the old secondary
4232

4233
    Failures are not very well handled.
4234

4235
    """
4236
    steps_total = 6
4237
    warning, info = (self.proc.LogWarning, self.proc.LogInfo)
4238
    instance = self.instance
4239
    iv_names = {}
4240
    vgname = self.cfg.GetVGName()
4241
    # start of work
4242
    cfg = self.cfg
4243
    old_node = self.tgt_node
4244
    new_node = self.new_node
4245
    pri_node = instance.primary_node
4246

    
4247
    # Step: check device activation
4248
    self.proc.LogStep(1, steps_total, "check device existence")
4249
    info("checking volume groups")
4250
    my_vg = cfg.GetVGName()
4251
    results = self.rpc.call_vg_list([pri_node, new_node])
4252
    if not results:
4253
      raise errors.OpExecError("Can't list volume groups on the nodes")
4254
    for node in pri_node, new_node:
4255
      res = results.get(node, False)
4256
      if not res or my_vg not in res:
4257
        raise errors.OpExecError("Volume group '%s' not found on %s" %
4258
                                 (my_vg, node))
4259
    for dev in instance.disks:
4260
      if not dev.iv_name in self.op.disks:
4261
        continue
4262
      info("checking %s on %s" % (dev.iv_name, pri_node))
4263
      cfg.SetDiskID(dev, pri_node)
4264
      if not self.rpc.call_blockdev_find(pri_node, dev):
4265
        raise errors.OpExecError("Can't find device %s on node %s" %
4266
                                 (dev.iv_name, pri_node))
4267

    
4268
    # Step: check other node consistency
4269
    self.proc.LogStep(2, steps_total, "check peer consistency")
4270
    for dev in instance.disks:
4271
      if not dev.iv_name in self.op.disks:
4272
        continue
4273
      info("checking %s consistency on %s" % (dev.iv_name, pri_node))
4274
      if not _CheckDiskConsistency(self, dev, pri_node, True, ldisk=True):
4275
        raise errors.OpExecError("Primary node (%s) has degraded storage,"
4276
                                 " unsafe to replace the secondary" %
4277
                                 pri_node)
4278

    
4279
    # Step: create new storage
4280
    self.proc.LogStep(3, steps_total, "allocate new storage")
4281
    for dev in instance.disks:
4282
      size = dev.size
4283
      info("adding new local storage on %s for %s" % (new_node, dev.iv_name))
4284
      # since we *always* want to create this LV, we use the
4285
      # _Create...OnPrimary (which forces the creation), even if we
4286
      # are talking about the secondary node
4287
      for new_lv in dev.children:
4288
        if not _CreateBlockDevOnPrimary(self, new_node, instance, new_lv,
4289
                                        _GetInstanceInfoText(instance)):
4290
          raise errors.OpExecError("Failed to create new LV named '%s' on"
4291
                                   " node '%s'" %
4292
                                   (new_lv.logical_id[1], new_node))
4293

    
4294

    
4295
    # Step 4: dbrd minors and drbd setups changes
4296
    # after this, we must manually remove the drbd minors on both the
4297
    # error and the success paths
4298
    minors = cfg.AllocateDRBDMinor([new_node for dev in instance.disks],
4299
                                   instance.name)
4300
    logging.debug("Allocated minors %s" % (minors,))
4301
    self.proc.LogStep(4, steps_total, "changing drbd configuration")
4302
    for dev, new_minor in zip(instance.disks, minors):
4303
      size = dev.size
4304
      info("activating a new drbd on %s for %s" % (new_node, dev.iv_name))
4305
      # create new devices on new_node
4306
      if pri_node == dev.logical_id[0]:
4307
        new_logical_id = (pri_node, new_node,
4308
                          dev.logical_id[2], dev.logical_id[3], new_minor,
4309
                          dev.logical_id[5])
4310
      else:
4311
        new_logical_id = (new_node, pri_node,
4312
                          dev.logical_id[2], new_minor, dev.logical_id[4],
4313
                          dev.logical_id[5])
4314
      iv_names[dev.iv_name] = (dev, dev.children, new_logical_id)
4315
      logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
4316
                    new_logical_id)
4317
      new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
4318
                              logical_id=new_logical_id,
4319
                              children=dev.children)
4320
      if not _CreateBlockDevOnSecondary(self, new_node, instance,
4321
                                        new_drbd, False,
4322
                                        _GetInstanceInfoText(instance)):
4323
        self.cfg.ReleaseDRBDMinors(instance.name)
4324
        raise errors.OpExecError("Failed to create new DRBD on"
4325
                                 " node '%s'" % new_node)
4326

    
4327
    for dev in instance.disks:
4328
      # we have new devices, shutdown the drbd on the old secondary
4329
      info("shutting down drbd for %s on old node" % dev.iv_name)
4330
      cfg.SetDiskID(dev, old_node)
4331
      if not self.rpc.call_blockdev_shutdown(old_node, dev):
4332
        warning("Failed to shutdown drbd for %s on old node" % dev.iv_name,
4333
                hint="Please cleanup this device manually as soon as possible")
4334

    
4335
    info("detaching primary drbds from the network (=> standalone)")
4336
    done = 0
4337
    for dev in instance.disks:
4338
      cfg.SetDiskID(dev, pri_node)
4339
      # set the network part of the physical (unique in bdev terms) id
4340
      # to None, meaning detach from network
4341
      dev.physical_id = (None, None, None, None) + dev.physical_id[4:]
4342
      # and 'find' the device, which will 'fix' it to match the
4343
      # standalone state
4344
      if self.rpc.call_blockdev_find(pri_node, dev):
4345
        done += 1
4346
      else:
4347
        warning("Failed to detach drbd %s from network, unusual case" %
4348
                dev.iv_name)
4349

    
4350
    if not done:
4351
      # no detaches succeeded (very unlikely)
4352
      self.cfg.ReleaseDRBDMinors(instance.name)
4353
      raise errors.OpExecError("Can't detach at least one DRBD from old node")
4354

    
4355
    # if we managed to detach at least one, we update all the disks of
4356
    # the instance to point to the new secondary
4357
    info("updating instance configuration")
4358
    for dev, _, new_logical_id in iv_names.itervalues():
4359
      dev.logical_id = new_logical_id
4360
      cfg.SetDiskID(dev, pri_node)
4361
    cfg.Update(instance)
4362
    # we can remove now the temp minors as now the new values are
4363
    # written to the config file (and therefore stable)
4364
    self.cfg.ReleaseDRBDMinors(instance.name)
4365

    
4366
    # and now perform the drbd attach
4367
    info("attaching primary drbds to new secondary (standalone => connected)")
4368
    failures = []
4369
    for dev in instance.disks:
4370
      info("attaching primary drbd for %s to new secondary node" % dev.iv_name)
4371
      # since the attach is smart, it's enough to 'find' the device,
4372
      # it will automatically activate the network, if the physical_id
4373
      # is correct
4374
      cfg.SetDiskID(dev, pri_node)
4375
      logging.debug("Disk to attach: %s", dev)
4376
      if not self.rpc.call_blockdev_find(pri_node, dev):
4377
        warning("can't attach drbd %s to new secondary!" % dev.iv_name,
4378
                "please do a gnt-instance info to see the status of disks")
4379

    
4380
    # this can fail as the old devices are degraded and _WaitForSync
4381
    # does a combined result over all disks, so we don't check its
4382
    # return value
4383
    self.proc.LogStep(5, steps_total, "sync devices")
4384
    _WaitForSync(self, instance, unlock=True)
4385

    
4386
    # so check manually all the devices
4387
    for name, (dev, old_lvs, _) in iv_names.iteritems():
4388
      cfg.SetDiskID(dev, pri_node)
4389
      is_degr = self.rpc.call_blockdev_find(pri_node, dev)[5]
4390
      if is_degr:
4391
        raise errors.OpExecError("DRBD device %s is degraded!" % name)
4392

    
4393
    self.proc.LogStep(6, steps_total, "removing old storage")
4394
    for name, (dev, old_lvs, _) in iv_names.iteritems():
4395
      info("remove logical volumes for %s" % name)
4396
      for lv in old_lvs:
4397
        cfg.SetDiskID(lv, old_node)
4398
        if not self.rpc.call_blockdev_remove(old_node, lv):
4399
          warning("Can't remove LV on old secondary",
4400
                  hint="Cleanup stale volumes by hand")
4401

    
4402
  def Exec(self, feedback_fn):
4403
    """Execute disk replacement.
4404

4405
    This dispatches the disk replacement to the appropriate handler.
4406

4407
    """
4408
    instance = self.instance
4409

    
4410
    # Activate the instance disks if we're replacing them on a down instance
4411
    if instance.status == "down":
4412
      _StartInstanceDisks(self, instance, True)
4413

    
4414
    if instance.disk_template == constants.DT_DRBD8:
4415
      if self.op.remote_node is None:
4416
        fn = self._ExecD8DiskOnly
4417
      else:
4418
        fn = self._ExecD8Secondary
4419
    else:
4420
      raise errors.ProgrammerError("Unhandled disk replacement case")
4421

    
4422
    ret = fn(feedback_fn)
4423

    
4424
    # Deactivate the instance disks if we're replacing them on a down instance
4425
    if instance.status == "down":
4426
      _SafeShutdownInstanceDisks(self, instance)
4427

    
4428
    return ret
4429

    
4430

    
4431
class LUGrowDisk(LogicalUnit):
4432
  """Grow a disk of an instance.
4433

4434
  """
4435
  HPATH = "disk-grow"
4436
  HTYPE = constants.HTYPE_INSTANCE
4437
  _OP_REQP = ["instance_name", "disk", "amount", "wait_for_sync"]
4438
  REQ_BGL = False
4439

    
4440
  def ExpandNames(self):
4441
    self._ExpandAndLockInstance()
4442
    self.needed_locks[locking.LEVEL_NODE] = []
4443
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4444

    
4445
  def DeclareLocks(self, level):
4446
    if level == locking.LEVEL_NODE:
4447
      self._LockInstancesNodes()
4448

    
4449
  def BuildHooksEnv(self):
4450
    """Build hooks env.
4451

4452
    This runs on the master, the primary and all the secondaries.
4453

4454
    """
4455
    env = {
4456
      "DISK": self.op.disk,
4457
      "AMOUNT": self.op.amount,
4458
      }
4459
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4460
    nl = [
4461
      self.cfg.GetMasterNode(),
4462
      self.instance.primary_node,
4463
      ]
4464
    return env, nl, nl
4465

    
4466
  def CheckPrereq(self):
4467
    """Check prerequisites.
4468

4469
    This checks that the instance is in the cluster.
4470

4471
    """
4472
    instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4473
    assert instance is not None, \
4474
      "Cannot retrieve locked instance %s" % self.op.instance_name
4475

    
4476
    self.instance = instance
4477

    
4478
    if instance.disk_template not in (constants.DT_PLAIN, constants.DT_DRBD8):
4479
      raise errors.OpPrereqError("Instance's disk layout does not support"
4480
                                 " growing.")
4481

    
4482
    if instance.FindDisk(self.op.disk) is None:
4483
      raise errors.OpPrereqError("Disk '%s' not found for instance '%s'" %
4484
                                 (self.op.disk, instance.name))
4485

    
4486
    nodenames = [instance.primary_node] + list(instance.secondary_nodes)
4487
    nodeinfo = self.rpc.call_node_info(nodenames, self.cfg.GetVGName(),
4488
                                       instance.hypervisor)
4489
    for node in nodenames:
4490
      info = nodeinfo.get(node, None)
4491
      if not info:
4492
        raise errors.OpPrereqError("Cannot get current information"
4493
                                   " from node '%s'" % node)
4494
      vg_free = info.get('vg_free', None)
4495
      if not isinstance(vg_free, int):
4496
        raise errors.OpPrereqError("Can't compute free disk space on"
4497
                                   " node %s" % node)
4498
      if self.op.amount > info['vg_free']:
4499
        raise errors.OpPrereqError("Not enough disk space on target node %s:"
4500
                                   " %d MiB available, %d MiB required" %
4501
                                   (node, info['vg_free'], self.op.amount))
4502

    
4503
  def Exec(self, feedback_fn):
4504
    """Execute disk grow.
4505

4506
    """
4507
    instance = self.instance
4508
    disk = instance.FindDisk(self.op.disk)
4509
    for node in (instance.secondary_nodes + (instance.primary_node,)):
4510
      self.cfg.SetDiskID(disk, node)
4511
      result = self.rpc.call_blockdev_grow(node, disk, self.op.amount)
4512
      if (not result or not isinstance(result, (list, tuple)) or
4513
          len(result) != 2):
4514
        raise errors.OpExecError("grow request failed to node %s" % node)
4515
      elif not result[0]:
4516
        raise errors.OpExecError("grow request failed to node %s: %s" %
4517
                                 (node, result[1]))
4518
    disk.RecordGrow(self.op.amount)
4519
    self.cfg.Update(instance)
4520
    if self.op.wait_for_sync:
4521
      disk_abort = not _WaitForSync(self, instance)
4522
      if disk_abort:
4523
        self.proc.LogWarning("Warning: disk sync-ing has not returned a good"
4524
                             " status.\nPlease check the instance.")
4525

    
4526

    
4527
class LUQueryInstanceData(NoHooksLU):
4528
  """Query runtime instance data.
4529

4530
  """
4531
  _OP_REQP = ["instances", "static"]
4532
  REQ_BGL = False
4533

    
4534
  def ExpandNames(self):
4535
    self.needed_locks = {}
4536
    self.share_locks = dict(((i, 1) for i in locking.LEVELS))
4537

    
4538
    if not isinstance(self.op.instances, list):
4539
      raise errors.OpPrereqError("Invalid argument type 'instances'")
4540

    
4541
    if self.op.instances:
4542
      self.wanted_names = []
4543
      for name in self.op.instances:
4544
        full_name = self.cfg.ExpandInstanceName(name)
4545
        if full_name is None:
4546
          raise errors.OpPrereqError("Instance '%s' not known" %
4547
                                     self.op.instance_name)
4548
        self.wanted_names.append(full_name)
4549
      self.needed_locks[locking.LEVEL_INSTANCE] = self.wanted_names
4550
    else:
4551
      self.wanted_names = None
4552
      self.needed_locks[locking.LEVEL_INSTANCE] = locking.ALL_SET
4553

    
4554
    self.needed_locks[locking.LEVEL_NODE] = []
4555
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4556

    
4557
  def DeclareLocks(self, level):
4558
    if level == locking.LEVEL_NODE:
4559
      self._LockInstancesNodes()
4560

    
4561
  def CheckPrereq(self):
4562
    """Check prerequisites.
4563

4564
    This only checks the optional instance list against the existing names.
4565

4566
    """
4567
    if self.wanted_names is None:
4568
      self.wanted_names = self.acquired_locks[locking.LEVEL_INSTANCE]
4569

    
4570
    self.wanted_instances = [self.cfg.GetInstanceInfo(name) for name
4571
                             in self.wanted_names]
4572
    return
4573

    
4574
  def _ComputeDiskStatus(self, instance, snode, dev):
4575
    """Compute block device status.
4576

4577
    """
4578
    static = self.op.static
4579
    if not static:
4580
      self.cfg.SetDiskID(dev, instance.primary_node)
4581
      dev_pstatus = self.rpc.call_blockdev_find(instance.primary_node, dev)
4582
    else:
4583
      dev_pstatus = None
4584

    
4585
    if dev.dev_type in constants.LDS_DRBD:
4586
      # we change the snode then (otherwise we use the one passed in)
4587
      if dev.logical_id[0] == instance.primary_node:
4588
        snode = dev.logical_id[1]
4589
      else:
4590
        snode = dev.logical_id[0]
4591

    
4592
    if snode and not static:
4593
      self.cfg.SetDiskID(dev, snode)
4594
      dev_sstatus = self.rpc.call_blockdev_find(snode, dev)
4595
    else:
4596
      dev_sstatus = None
4597

    
4598
    if dev.children:
4599
      dev_children = [self._ComputeDiskStatus(instance, snode, child)
4600
                      for child in dev.children]
4601
    else:
4602
      dev_children = []
4603

    
4604
    data = {
4605
      "iv_name": dev.iv_name,
4606
      "dev_type": dev.dev_type,
4607
      "logical_id": dev.logical_id,
4608
      "physical_id": dev.physical_id,
4609
      "pstatus": dev_pstatus,
4610
      "sstatus": dev_sstatus,
4611
      "children": dev_children,
4612
      }
4613

    
4614
    return data
4615

    
4616
  def Exec(self, feedback_fn):
4617
    """Gather and return data"""
4618
    result = {}
4619

    
4620
    cluster = self.cfg.GetClusterInfo()
4621

    
4622
    for instance in self.wanted_instances:
4623
      if not self.op.static:
4624
        remote_info = self.rpc.call_instance_info(instance.primary_node,
4625
                                                  instance.name,
4626
                                                  instance.hypervisor)
4627
        if remote_info and "state" in remote_info:
4628
          remote_state = "up"
4629
        else:
4630
          remote_state = "down"
4631
      else:
4632
        remote_state = None
4633
      if instance.status == "down":
4634
        config_state = "down"
4635
      else:
4636
        config_state = "up"
4637

    
4638
      disks = [self._ComputeDiskStatus(instance, None, device)
4639
               for device in instance.disks]
4640

    
4641
      idict = {
4642
        "name": instance.name,
4643
        "config_state": config_state,
4644
        "run_state": remote_state,
4645
        "pnode": instance.primary_node,
4646
        "snodes": instance.secondary_nodes,
4647
        "os": instance.os,
4648
        "nics": [(nic.mac, nic.ip, nic.bridge) for nic in instance.nics],
4649
        "disks": disks,
4650
        "hypervisor": instance.hypervisor,
4651
        "network_port": instance.network_port,
4652
        "hv_instance": instance.hvparams,
4653
        "hv_actual": cluster.FillHV(instance),
4654
        "be_instance": instance.beparams,
4655
        "be_actual": cluster.FillBE(instance),
4656
        }
4657

    
4658
      result[instance.name] = idict
4659

    
4660
    return result
4661

    
4662

    
4663
class LUSetInstanceParams(LogicalUnit):
4664
  """Modifies an instances's parameters.
4665

4666
  """
4667
  HPATH = "instance-modify"
4668
  HTYPE = constants.HTYPE_INSTANCE
4669
  _OP_REQP = ["instance_name", "hvparams"]
4670
  REQ_BGL = False
4671

    
4672
  def ExpandNames(self):
4673
    self._ExpandAndLockInstance()
4674
    self.needed_locks[locking.LEVEL_NODE] = []
4675
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
4676

    
4677

    
4678
  def DeclareLocks(self, level):
4679
    if level == locking.LEVEL_NODE:
4680
      self._LockInstancesNodes()
4681

    
4682
  def BuildHooksEnv(self):
4683
    """Build hooks env.
4684

4685
    This runs on the master, primary and secondaries.
4686

4687
    """
4688
    args = dict()
4689
    if constants.BE_MEMORY in self.be_new:
4690
      args['memory'] = self.be_new[constants.BE_MEMORY]
4691
    if constants.BE_VCPUS in self.be_new:
4692
      args['vcpus'] = self.be_new[constants.BE_VCPUS]
4693
    if self.do_ip or self.do_bridge or self.mac:
4694
      if self.do_ip:
4695
        ip = self.ip
4696
      else:
4697
        ip = self.instance.nics[0].ip
4698
      if self.bridge:
4699
        bridge = self.bridge
4700
      else:
4701
        bridge = self.instance.nics[0].bridge
4702
      if self.mac:
4703
        mac = self.mac
4704
      else:
4705
        mac = self.instance.nics[0].mac
4706
      args['nics'] = [(ip, bridge, mac)]
4707
    env = _BuildInstanceHookEnvByObject(self, self.instance, override=args)
4708
    nl = [self.cfg.GetMasterNode(),
4709
          self.instance.primary_node] + list(self.instance.secondary_nodes)
4710
    return env, nl, nl
4711

    
4712
  def CheckPrereq(self):
4713
    """Check prerequisites.
4714

4715
    This only checks the instance list against the existing names.
4716

4717
    """
4718
    # FIXME: all the parameters could be checked before, in ExpandNames, or in
4719
    # a separate CheckArguments function, if we implement one, so the operation
4720
    # can be aborted without waiting for any lock, should it have an error...
4721
    self.ip = getattr(self.op, "ip", None)
4722
    self.mac = getattr(self.op, "mac", None)
4723
    self.bridge = getattr(self.op, "bridge", None)
4724
    self.kernel_path = getattr(self.op, "kernel_path", None)
4725
    self.initrd_path = getattr(self.op, "initrd_path", None)
4726
    self.force = getattr(self.op, "force", None)
4727
    all_parms = [self.ip, self.bridge, self.mac]
4728
    if (all_parms.count(None) == len(all_parms) and
4729
        not self.op.hvparams and
4730
        not self.op.beparams):
4731
      raise errors.OpPrereqError("No changes submitted")
4732
    for item in (constants.BE_MEMORY, constants.BE_VCPUS):
4733
      val = self.op.beparams.get(item, None)
4734
      if val is not None:
4735
        try:
4736
          val = int(val)
4737
        except ValueError, err:
4738
          raise errors.OpPrereqError("Invalid %s size: %s" % (item, str(err)))
4739
        self.op.beparams[item] = val
4740
    if self.ip is not None:
4741
      self.do_ip = True
4742
      if self.ip.lower() == "none":
4743
        self.ip = None
4744
      else:
4745
        if not utils.IsValidIP(self.ip):
4746
          raise errors.OpPrereqError("Invalid IP address '%s'." % self.ip)
4747
    else:
4748
      self.do_ip = False
4749
    self.do_bridge = (self.bridge is not None)
4750
    if self.mac is not None:
4751
      if self.cfg.IsMacInUse(self.mac):
4752
        raise errors.OpPrereqError('MAC address %s already in use in cluster' %
4753
                                   self.mac)
4754
      if not utils.IsValidMac(self.mac):
4755
        raise errors.OpPrereqError('Invalid MAC address %s' % self.mac)
4756

    
4757
    # checking the new params on the primary/secondary nodes
4758

    
4759
    instance = self.instance = self.cfg.GetInstanceInfo(self.op.instance_name)
4760
    assert self.instance is not None, \
4761
      "Cannot retrieve locked instance %s" % self.op.instance_name
4762
    pnode = self.instance.primary_node
4763
    nodelist = [pnode]
4764
    nodelist.extend(instance.secondary_nodes)
4765

    
4766
    # hvparams processing
4767
    if self.op.hvparams:
4768
      i_hvdict = copy.deepcopy(instance.hvparams)
4769
      for key, val in self.op.hvparams.iteritems():
4770
        if val is None:
4771
          try:
4772
            del i_hvdict[key]
4773
          except KeyError:
4774
            pass
4775
        else:
4776
          i_hvdict[key] = val
4777
      cluster = self.cfg.GetClusterInfo()
4778
      hv_new = cluster.FillDict(cluster.hvparams[instance.hypervisor],
4779
                                i_hvdict)
4780
      # local check
4781
      hypervisor.GetHypervisor(
4782
        instance.hypervisor).CheckParameterSyntax(hv_new)
4783
      _CheckHVParams(self, nodelist, instance.hypervisor, hv_new)
4784
      self.hv_new = hv_new # the new actual values
4785
      self.hv_inst = i_hvdict # the new dict (without defaults)
4786
    else:
4787
      self.hv_new = self.hv_inst = {}
4788

    
4789
    # beparams processing
4790
    if self.op.beparams:
4791
      i_bedict = copy.deepcopy(instance.beparams)
4792
      for key, val in self.op.beparams.iteritems():
4793
        if val is None:
4794
          try:
4795
            del i_bedict[key]
4796
          except KeyError:
4797
            pass
4798
        else:
4799
          i_bedict[key] = val
4800
      cluster = self.cfg.GetClusterInfo()
4801
      be_new = cluster.FillDict(cluster.beparams[constants.BEGR_DEFAULT],
4802
                                i_bedict)
4803
      self.be_new = be_new # the new actual values
4804
      self.be_inst = i_bedict # the new dict (without defaults)
4805
    else:
4806
      self.hv_new = self.hv_inst = {}
4807

    
4808
    self.warn = []
4809

    
4810
    if constants.BE_MEMORY in self.op.beparams and not self.force:
4811
      mem_check_list = [pnode]
4812
      if be_new[constants.BE_AUTO_BALANCE]:
4813
        # either we changed auto_balance to yes or it was from before
4814
        mem_check_list.extend(instance.secondary_nodes)
4815
      instance_info = self.rpc.call_instance_info(pnode, instance.name,
4816
                                                  instance.hypervisor)
4817
      nodeinfo = self.rpc.call_node_info(mem_check_list, self.cfg.GetVGName(),
4818
                                         instance.hypervisor)
4819

    
4820
      if pnode not in nodeinfo or not isinstance(nodeinfo[pnode], dict):
4821
        # Assume the primary node is unreachable and go ahead
4822
        self.warn.append("Can't get info from primary node %s" % pnode)
4823
      else:
4824
        if instance_info:
4825
          current_mem = instance_info['memory']
4826
        else:
4827
          # Assume instance not running
4828
          # (there is a slight race condition here, but it's not very probable,
4829
          # and we have no other way to check)
4830
          current_mem = 0
4831
        miss_mem = (be_new[constants.BE_MEMORY] - current_mem -
4832
                    nodeinfo[pnode]['memory_free'])
4833
        if miss_mem > 0:
4834
          raise errors.OpPrereqError("This change will prevent the instance"
4835
                                     " from starting, due to %d MB of memory"
4836
                                     " missing on its primary node" % miss_mem)
4837

    
4838
      if be_new[constants.BE_AUTO_BALANCE]:
4839
        for node in instance.secondary_nodes:
4840
          if node not in nodeinfo or not isinstance(nodeinfo[node], dict):
4841
            self.warn.append("Can't get info from secondary node %s" % node)
4842
          elif be_new[constants.BE_MEMORY] > nodeinfo[node]['memory_free']:
4843
            self.warn.append("Not enough memory to failover instance to"
4844
                             " secondary node %s" % node)
4845

    
4846
    return
4847

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

4851
    All parameters take effect only at the next restart of the instance.
4852
    """
4853
    # Process here the warnings from CheckPrereq, as we don't have a
4854
    # feedback_fn there.
4855
    for warn in self.warn:
4856
      feedback_fn("WARNING: %s" % warn)
4857

    
4858
    result = []
4859
    instance = self.instance
4860
    if self.do_ip:
4861
      instance.nics[0].ip = self.ip
4862
      result.append(("ip", self.ip))
4863
    if self.bridge:
4864
      instance.nics[0].bridge = self.bridge
4865
      result.append(("bridge", self.bridge))
4866
    if self.mac:
4867
      instance.nics[0].mac = self.mac
4868
      result.append(("mac", self.mac))
4869
    if self.op.hvparams:
4870
      instance.hvparams = self.hv_new
4871
      for key, val in self.op.hvparams.iteritems():
4872
        result.append(("hv/%s" % key, val))
4873
    if self.op.beparams:
4874
      instance.beparams = self.be_inst
4875
      for key, val in self.op.beparams.iteritems():
4876
        result.append(("be/%s" % key, val))
4877

    
4878
    self.cfg.Update(instance)
4879

    
4880
    return result
4881

    
4882

    
4883
class LUQueryExports(NoHooksLU):
4884
  """Query the exports list
4885

4886
  """
4887
  _OP_REQP = ['nodes']
4888
  REQ_BGL = False
4889

    
4890
  def ExpandNames(self):
4891
    self.needed_locks = {}
4892
    self.share_locks[locking.LEVEL_NODE] = 1
4893
    if not self.op.nodes:
4894
      self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4895
    else:
4896
      self.needed_locks[locking.LEVEL_NODE] = \
4897
        _GetWantedNodes(self, self.op.nodes)
4898

    
4899
  def CheckPrereq(self):
4900
    """Check prerequisites.
4901

4902
    """
4903
    self.nodes = self.acquired_locks[locking.LEVEL_NODE]
4904

    
4905
  def Exec(self, feedback_fn):
4906
    """Compute the list of all the exported system images.
4907

4908
    Returns:
4909
      a dictionary with the structure node->(export-list)
4910
      where export-list is a list of the instances exported on
4911
      that node.
4912

4913
    """
4914
    return self.rpc.call_export_list(self.nodes)
4915

    
4916

    
4917
class LUExportInstance(LogicalUnit):
4918
  """Export an instance to an image in the cluster.
4919

4920
  """
4921
  HPATH = "instance-export"
4922
  HTYPE = constants.HTYPE_INSTANCE
4923
  _OP_REQP = ["instance_name", "target_node", "shutdown"]
4924
  REQ_BGL = False
4925

    
4926
  def ExpandNames(self):
4927
    self._ExpandAndLockInstance()
4928
    # FIXME: lock only instance primary and destination node
4929
    #
4930
    # Sad but true, for now we have do lock all nodes, as we don't know where
4931
    # the previous export might be, and and in this LU we search for it and
4932
    # remove it from its current node. In the future we could fix this by:
4933
    #  - making a tasklet to search (share-lock all), then create the new one,
4934
    #    then one to remove, after
4935
    #  - removing the removal operation altoghether
4936
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
4937

    
4938
  def DeclareLocks(self, level):
4939
    """Last minute lock declaration."""
4940
    # All nodes are locked anyway, so nothing to do here.
4941

    
4942
  def BuildHooksEnv(self):
4943
    """Build hooks env.
4944

4945
    This will run on the master, primary node and target node.
4946

4947
    """
4948
    env = {
4949
      "EXPORT_NODE": self.op.target_node,
4950
      "EXPORT_DO_SHUTDOWN": self.op.shutdown,
4951
      }
4952
    env.update(_BuildInstanceHookEnvByObject(self, self.instance))
4953
    nl = [self.cfg.GetMasterNode(), self.instance.primary_node,
4954
          self.op.target_node]
4955
    return env, nl, nl
4956

    
4957
  def CheckPrereq(self):
4958
    """Check prerequisites.
4959

4960
    This checks that the instance and node names are valid.
4961

4962
    """
4963
    instance_name = self.op.instance_name
4964
    self.instance = self.cfg.GetInstanceInfo(instance_name)
4965
    assert self.instance is not None, \
4966
          "Cannot retrieve locked instance %s" % self.op.instance_name
4967

    
4968
    self.dst_node = self.cfg.GetNodeInfo(
4969
      self.cfg.ExpandNodeName(self.op.target_node))
4970

    
4971
    assert self.dst_node is not None, \
4972
          "Cannot retrieve locked node %s" % self.op.target_node
4973

    
4974
    # instance disk type verification
4975
    for disk in self.instance.disks:
4976
      if disk.dev_type == constants.LD_FILE:
4977
        raise errors.OpPrereqError("Export not supported for instances with"
4978
                                   " file-based disks")
4979

    
4980
  def Exec(self, feedback_fn):
4981
    """Export an instance to an image in the cluster.
4982

4983
    """
4984
    instance = self.instance
4985
    dst_node = self.dst_node
4986
    src_node = instance.primary_node
4987
    if self.op.shutdown:
4988
      # shutdown the instance, but not the disks
4989
      if not self.rpc.call_instance_shutdown(src_node, instance):
4990
        raise errors.OpExecError("Could not shutdown instance %s on node %s" %
4991
                                 (instance.name, src_node))
4992

    
4993
    vgname = self.cfg.GetVGName()
4994

    
4995
    snap_disks = []
4996

    
4997
    try:
4998
      for disk in instance.disks:
4999
        # new_dev_name will be a snapshot of an lvm leaf of the one we passed
5000
        new_dev_name = self.rpc.call_blockdev_snapshot(src_node, disk)
5001

    
5002
        if not new_dev_name:
5003
          self.LogWarning("Could not snapshot block device %s on node %s",
5004
                          disk.logical_id[1], src_node)
5005
          snap_disks.append(False)
5006
        else:
5007
          new_dev = objects.Disk(dev_type=constants.LD_LV, size=disk.size,
5008
                                 logical_id=(vgname, new_dev_name),
5009
                                 physical_id=(vgname, new_dev_name),
5010
                                 iv_name=disk.iv_name)
5011
          snap_disks.append(new_dev)
5012

    
5013
    finally:
5014
      if self.op.shutdown and instance.status == "up":
5015
        if not self.rpc.call_instance_start(src_node, instance, None):
5016
          _ShutdownInstanceDisks(self, instance)
5017
          raise errors.OpExecError("Could not start instance")
5018

    
5019
    # TODO: check for size
5020

    
5021
    cluster_name = self.cfg.GetClusterName()
5022
    for idx, dev in enumerate(snap_disks):
5023
      if dev:
5024
        if not self.rpc.call_snapshot_export(src_node, dev, dst_node.name,
5025
                                             instance, cluster_name, idx):
5026
          self.LogWarning("Could not export block device %s from node %s to"
5027
                          " node %s", dev.logical_id[1], src_node,
5028
                          dst_node.name)
5029
        if not self.rpc.call_blockdev_remove(src_node, dev):
5030
          self.LogWarning("Could not remove snapshot block device %s from node"
5031
                          " %s", dev.logical_id[1], src_node)
5032

    
5033
    if not self.rpc.call_finalize_export(dst_node.name, instance, snap_disks):
5034
      self.LogWarning("Could not finalize export for instance %s on node %s",
5035
                      instance.name, dst_node.name)
5036

    
5037
    nodelist = self.cfg.GetNodeList()
5038
    nodelist.remove(dst_node.name)
5039

    
5040
    # on one-node clusters nodelist will be empty after the removal
5041
    # if we proceed the backup would be removed because OpQueryExports
5042
    # substitutes an empty list with the full cluster node list.
5043
    if nodelist:
5044
      exportlist = self.rpc.call_export_list(nodelist)
5045
      for node in exportlist:
5046
        if instance.name in exportlist[node]:
5047
          if not self.rpc.call_export_remove(node, instance.name):
5048
            self.LogWarning("Could not remove older export for instance %s"
5049
                            " on node %s", instance.name, node)
5050

    
5051

    
5052
class LURemoveExport(NoHooksLU):
5053
  """Remove exports related to the named instance.
5054

5055
  """
5056
  _OP_REQP = ["instance_name"]
5057
  REQ_BGL = False
5058

    
5059
  def ExpandNames(self):
5060
    self.needed_locks = {}
5061
    # We need all nodes to be locked in order for RemoveExport to work, but we
5062
    # don't need to lock the instance itself, as nothing will happen to it (and
5063
    # we can remove exports also for a removed instance)
5064
    self.needed_locks[locking.LEVEL_NODE] = locking.ALL_SET
5065

    
5066
  def CheckPrereq(self):
5067
    """Check prerequisites.
5068
    """
5069
    pass
5070

    
5071
  def Exec(self, feedback_fn):
5072
    """Remove any export.
5073

5074
    """
5075
    instance_name = self.cfg.ExpandInstanceName(self.op.instance_name)
5076
    # If the instance was not found we'll try with the name that was passed in.
5077
    # This will only work if it was an FQDN, though.
5078
    fqdn_warn = False
5079
    if not instance_name:
5080
      fqdn_warn = True
5081
      instance_name = self.op.instance_name
5082

    
5083
    exportlist = self.rpc.call_export_list(self.acquired_locks[
5084
      locking.LEVEL_NODE])
5085
    found = False
5086
    for node in exportlist:
5087
      if instance_name in exportlist[node]:
5088
        found = True
5089
        if not self.rpc.call_export_remove(node, instance_name):
5090
          logging.error("Could not remove export for instance %s"
5091
                        " on node %s", instance_name, node)
5092

    
5093
    if fqdn_warn and not found:
5094
      feedback_fn("Export not found. If trying to remove an export belonging"
5095
                  " to a deleted instance please use its Fully Qualified"
5096
                  " Domain Name.")
5097

    
5098

    
5099
class TagsLU(NoHooksLU):
5100
  """Generic tags LU.
5101

5102
  This is an abstract class which is the parent of all the other tags LUs.
5103

5104
  """
5105

    
5106
  def ExpandNames(self):
5107
    self.needed_locks = {}
5108
    if self.op.kind == constants.TAG_NODE:
5109
      name = self.cfg.ExpandNodeName(self.op.name)
5110
      if name is None:
5111
        raise errors.OpPrereqError("Invalid node name (%s)" %
5112
                                   (self.op.name,))
5113
      self.op.name = name
5114
      self.needed_locks[locking.LEVEL_NODE] = name
5115
    elif self.op.kind == constants.TAG_INSTANCE:
5116
      name = self.cfg.ExpandInstanceName(self.op.name)
5117
      if name is None:
5118
        raise errors.OpPrereqError("Invalid instance name (%s)" %
5119
                                   (self.op.name,))
5120
      self.op.name = name
5121
      self.needed_locks[locking.LEVEL_INSTANCE] = name
5122

    
5123
  def CheckPrereq(self):
5124
    """Check prerequisites.
5125

5126
    """
5127
    if self.op.kind == constants.TAG_CLUSTER:
5128
      self.target = self.cfg.GetClusterInfo()
5129
    elif self.op.kind == constants.TAG_NODE:
5130
      self.target = self.cfg.GetNodeInfo(self.op.name)
5131
    elif self.op.kind == constants.TAG_INSTANCE:
5132
      self.target = self.cfg.GetInstanceInfo(self.op.name)
5133
    else:
5134
      raise errors.OpPrereqError("Wrong tag type requested (%s)" %
5135
                                 str(self.op.kind))
5136

    
5137

    
5138
class LUGetTags(TagsLU):
5139
  """Returns the tags of a given object.
5140

5141
  """
5142
  _OP_REQP = ["kind", "name"]
5143
  REQ_BGL = False
5144

    
5145
  def Exec(self, feedback_fn):
5146
    """Returns the tag list.
5147

5148
    """
5149
    return list(self.target.GetTags())
5150

    
5151

    
5152
class LUSearchTags(NoHooksLU):
5153
  """Searches the tags for a given pattern.
5154

5155
  """
5156
  _OP_REQP = ["pattern"]
5157
  REQ_BGL = False
5158

    
5159
  def ExpandNames(self):
5160
    self.needed_locks = {}
5161

    
5162
  def CheckPrereq(self):
5163
    """Check prerequisites.
5164

5165
    This checks the pattern passed for validity by compiling it.
5166

5167
    """
5168
    try:
5169
      self.re = re.compile(self.op.pattern)
5170
    except re.error, err:
5171
      raise errors.OpPrereqError("Invalid search pattern '%s': %s" %
5172
                                 (self.op.pattern, err))
5173

    
5174
  def Exec(self, feedback_fn):
5175
    """Returns the tag list.
5176

5177
    """
5178
    cfg = self.cfg
5179
    tgts = [("/cluster", cfg.GetClusterInfo())]
5180
    ilist = cfg.GetAllInstancesInfo().values()
5181
    tgts.extend([("/instances/%s" % i.name, i) for i in ilist])
5182
    nlist = cfg.GetAllNodesInfo().values()
5183
    tgts.extend([("/nodes/%s" % n.name, n) for n in nlist])
5184
    results = []
5185
    for path, target in tgts:
5186
      for tag in target.GetTags():
5187
        if self.re.search(tag):
5188
          results.append((path, tag))
5189
    return results
5190

    
5191

    
5192
class LUAddTags(TagsLU):
5193
  """Sets a tag on a given object.
5194

5195
  """
5196
  _OP_REQP = ["kind", "name", "tags"]
5197
  REQ_BGL = False
5198

    
5199
  def CheckPrereq(self):
5200
    """Check prerequisites.
5201

5202
    This checks the type and length of the tag name and value.
5203

5204
    """
5205
    TagsLU.CheckPrereq(self)
5206
    for tag in self.op.tags:
5207
      objects.TaggableObject.ValidateTag(tag)
5208

    
5209
  def Exec(self, feedback_fn):
5210
    """Sets the tag.
5211

5212
    """
5213
    try:
5214
      for tag in self.op.tags:
5215
        self.target.AddTag(tag)
5216
    except errors.TagError, err:
5217
      raise errors.OpExecError("Error while setting tag: %s" % str(err))
5218
    try:
5219
      self.cfg.Update(self.target)
5220
    except errors.ConfigurationError:
5221
      raise errors.OpRetryError("There has been a modification to the"
5222
                                " config file and the operation has been"
5223
                                " aborted. Please retry.")
5224

    
5225

    
5226
class LUDelTags(TagsLU):
5227
  """Delete a list of tags from a given object.
5228

5229
  """
5230
  _OP_REQP = ["kind", "name", "tags"]
5231
  REQ_BGL = False
5232

    
5233
  def CheckPrereq(self):
5234
    """Check prerequisites.
5235

5236
    This checks that we have the given tag.
5237

5238
    """
5239
    TagsLU.CheckPrereq(self)
5240
    for tag in self.op.tags:
5241
      objects.TaggableObject.ValidateTag(tag)
5242
    del_tags = frozenset(self.op.tags)
5243
    cur_tags = self.target.GetTags()
5244
    if not del_tags <= cur_tags:
5245
      diff_tags = del_tags - cur_tags
5246
      diff_names = ["'%s'" % tag for tag in diff_tags]
5247
      diff_names.sort()
5248
      raise errors.OpPrereqError("Tag(s) %s not found" %
5249
                                 (",".join(diff_names)))
5250

    
5251
  def Exec(self, feedback_fn):
5252
    """Remove the tag from the object.
5253

5254
    """
5255
    for tag in self.op.tags:
5256
      self.target.RemoveTag(tag)
5257
    try:
5258
      self.cfg.Update(self.target)
5259
    except errors.ConfigurationError:
5260
      raise errors.OpRetryError("There has been a modification to the"
5261
                                " config file and the operation has been"
5262
                                " aborted. Please retry.")
5263

    
5264

    
5265
class LUTestDelay(NoHooksLU):
5266
  """Sleep for a specified amount of time.
5267

5268
  This LU sleeps on the master and/or nodes for a specified amount of
5269
  time.
5270

5271
  """
5272
  _OP_REQP = ["duration", "on_master", "on_nodes"]
5273
  REQ_BGL = False
5274

    
5275
  def ExpandNames(self):
5276
    """Expand names and set required locks.
5277

5278
    This expands the node list, if any.
5279

5280
    """
5281
    self.needed_locks = {}
5282
    if self.op.on_nodes:
5283
      # _GetWantedNodes can be used here, but is not always appropriate to use
5284
      # this way in ExpandNames. Check LogicalUnit.ExpandNames docstring for
5285
      # more information.
5286
      self.op.on_nodes = _GetWantedNodes(self, self.op.on_nodes)
5287
      self.needed_locks[locking.LEVEL_NODE] = self.op.on_nodes
5288

    
5289
  def CheckPrereq(self):
5290
    """Check prerequisites.
5291

5292
    """
5293

    
5294
  def Exec(self, feedback_fn):
5295
    """Do the actual sleep.
5296

5297
    """
5298
    if self.op.on_master:
5299
      if not utils.TestDelay(self.op.duration):
5300
        raise errors.OpExecError("Error during master delay test")
5301
    if self.op.on_nodes:
5302
      result = self.rpc.call_test_delay(self.op.on_nodes, self.op.duration)
5303
      if not result:
5304
        raise errors.OpExecError("Complete failure from rpc call")
5305
      for node, node_result in result.items():
5306
        if not node_result:
5307
          raise errors.OpExecError("Failure during rpc call to node %s,"
5308
                                   " result: %s" % (node, node_result))
5309

    
5310

    
5311
class IAllocator(object):
5312
  """IAllocator framework.
5313

5314
  An IAllocator instance has three sets of attributes:
5315
    - cfg that is needed to query the cluster
5316
    - input data (all members of the _KEYS class attribute are required)
5317
    - four buffer attributes (in|out_data|text), that represent the
5318
      input (to the external script) in text and data structure format,
5319
      and the output from it, again in two formats
5320
    - the result variables from the script (success, info, nodes) for
5321
      easy usage
5322

5323
  """
5324
  _ALLO_KEYS = [
5325
    "mem_size", "disks", "disk_template",
5326
    "os", "tags", "nics", "vcpus",
5327
    ]
5328
  _RELO_KEYS = [
5329
    "relocate_from",
5330
    ]
5331

    
5332
  def __init__(self, lu, mode, name, **kwargs):
5333
    self.lu = lu
5334
    # init buffer variables
5335
    self.in_text = self.out_text = self.in_data = self.out_data = None
5336
    # init all input fields so that pylint is happy
5337
    self.mode = mode
5338
    self.name = name
5339
    self.mem_size = self.disks = self.disk_template = None
5340
    self.os = self.tags = self.nics = self.vcpus = None
5341
    self.relocate_from = None
5342
    # computed fields
5343
    self.required_nodes = None
5344
    # init result fields
5345
    self.success = self.info = self.nodes = None
5346
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
5347
      keyset = self._ALLO_KEYS
5348
    elif self.mode == constants.IALLOCATOR_MODE_RELOC:
5349
      keyset = self._RELO_KEYS
5350
    else:
5351
      raise errors.ProgrammerError("Unknown mode '%s' passed to the"
5352
                                   " IAllocator" % self.mode)
5353
    for key in kwargs:
5354
      if key not in keyset:
5355
        raise errors.ProgrammerError("Invalid input parameter '%s' to"
5356
                                     " IAllocator" % key)
5357
      setattr(self, key, kwargs[key])
5358
    for key in keyset:
5359
      if key not in kwargs:
5360
        raise errors.ProgrammerError("Missing input parameter '%s' to"
5361
                                     " IAllocator" % key)
5362
    self._BuildInputData()
5363

    
5364
  def _ComputeClusterData(self):
5365
    """Compute the generic allocator input data.
5366

5367
    This is the data that is independent of the actual operation.
5368

5369
    """
5370
    cfg = self.lu.cfg
5371
    cluster_info = cfg.GetClusterInfo()
5372
    # cluster data
5373
    data = {
5374
      "version": 1,
5375
      "cluster_name": cfg.GetClusterName(),
5376
      "cluster_tags": list(cluster_info.GetTags()),
5377
      "enable_hypervisors": list(cluster_info.enabled_hypervisors),
5378
      # we don't have job IDs
5379
      }
5380

    
5381
    i_list = []
5382
    cluster = self.cfg.GetClusterInfo()
5383
    for iname in cfg.GetInstanceList():
5384
      i_obj = cfg.GetInstanceInfo(iname)
5385
      i_list.append((i_obj, cluster.FillBE(i_obj)))
5386

    
5387
    # node data
5388
    node_results = {}
5389
    node_list = cfg.GetNodeList()
5390
    # FIXME: here we have only one hypervisor information, but
5391
    # instance can belong to different hypervisors
5392
    node_data = self.lu.rpc.call_node_info(node_list, cfg.GetVGName(),
5393
                                           cfg.GetHypervisorType())
5394
    for nname in node_list:
5395
      ninfo = cfg.GetNodeInfo(nname)
5396
      if nname not in node_data or not isinstance(node_data[nname], dict):
5397
        raise errors.OpExecError("Can't get data for node %s" % nname)
5398
      remote_info = node_data[nname]
5399
      for attr in ['memory_total', 'memory_free', 'memory_dom0',
5400
                   'vg_size', 'vg_free', 'cpu_total']:
5401
        if attr not in remote_info:
5402
          raise errors.OpExecError("Node '%s' didn't return attribute '%s'" %
5403
                                   (nname, attr))
5404
        try:
5405
          remote_info[attr] = int(remote_info[attr])
5406
        except ValueError, err:
5407
          raise errors.OpExecError("Node '%s' returned invalid value for '%s':"
5408
                                   " %s" % (nname, attr, str(err)))
5409
      # compute memory used by primary instances
5410
      i_p_mem = i_p_up_mem = 0
5411
      for iinfo, beinfo in i_list:
5412
        if iinfo.primary_node == nname:
5413
          i_p_mem += beinfo[constants.BE_MEMORY]
5414
          if iinfo.status == "up":
5415
            i_p_up_mem += beinfo[constants.BE_MEMORY]
5416

    
5417
      # compute memory used by instances
5418
      pnr = {
5419
        "tags": list(ninfo.GetTags()),
5420
        "total_memory": remote_info['memory_total'],
5421
        "reserved_memory": remote_info['memory_dom0'],
5422
        "free_memory": remote_info['memory_free'],
5423
        "i_pri_memory": i_p_mem,
5424
        "i_pri_up_memory": i_p_up_mem,
5425
        "total_disk": remote_info['vg_size'],
5426
        "free_disk": remote_info['vg_free'],
5427
        "primary_ip": ninfo.primary_ip,
5428
        "secondary_ip": ninfo.secondary_ip,
5429
        "total_cpus": remote_info['cpu_total'],
5430
        }
5431
      node_results[nname] = pnr
5432
    data["nodes"] = node_results
5433

    
5434
    # instance data
5435
    instance_data = {}
5436
    for iinfo, beinfo in i_list:
5437
      nic_data = [{"mac": n.mac, "ip": n.ip, "bridge": n.bridge}
5438
                  for n in iinfo.nics]
5439
      pir = {
5440
        "tags": list(iinfo.GetTags()),
5441
        "should_run": iinfo.status == "up",
5442
        "vcpus": beinfo[constants.BE_VCPUS],
5443
        "memory": beinfo[constants.BE_MEMORY],
5444
        "os": iinfo.os,
5445
        "nodes": [iinfo.primary_node] + list(iinfo.secondary_nodes),
5446
        "nics": nic_data,
5447
        "disks": [{"size": dsk.size, "mode": "w"} for dsk in iinfo.disks],
5448
        "disk_template": iinfo.disk_template,
5449
        "hypervisor": iinfo.hypervisor,
5450
        }
5451
      instance_data[iinfo.name] = pir
5452

    
5453
    data["instances"] = instance_data
5454

    
5455
    self.in_data = data
5456

    
5457
  def _AddNewInstance(self):
5458
    """Add new instance data to allocator structure.
5459

5460
    This in combination with _AllocatorGetClusterData will create the
5461
    correct structure needed as input for the allocator.
5462

5463
    The checks for the completeness of the opcode must have already been
5464
    done.
5465

5466
    """
5467
    data = self.in_data
5468
    if len(self.disks) != 2:
5469
      raise errors.OpExecError("Only two-disk configurations supported")
5470

    
5471
    disk_space = _ComputeDiskSize(self.disk_template,
5472
                                  self.disks[0]["size"], self.disks[1]["size"])
5473

    
5474
    if self.disk_template in constants.DTS_NET_MIRROR:
5475
      self.required_nodes = 2
5476
    else:
5477
      self.required_nodes = 1
5478
    request = {
5479
      "type": "allocate",
5480
      "name": self.name,
5481
      "disk_template": self.disk_template,
5482
      "tags": self.tags,
5483
      "os": self.os,
5484
      "vcpus": self.vcpus,
5485
      "memory": self.mem_size,
5486
      "disks": self.disks,
5487
      "disk_space_total": disk_space,
5488
      "nics": self.nics,
5489
      "required_nodes": self.required_nodes,
5490
      }
5491
    data["request"] = request
5492

    
5493
  def _AddRelocateInstance(self):
5494
    """Add relocate instance data to allocator structure.
5495

5496
    This in combination with _IAllocatorGetClusterData will create the
5497
    correct structure needed as input for the allocator.
5498

5499
    The checks for the completeness of the opcode must have already been
5500
    done.
5501

5502
    """
5503
    instance = self.lu.cfg.GetInstanceInfo(self.name)
5504
    if instance is None:
5505
      raise errors.ProgrammerError("Unknown instance '%s' passed to"
5506
                                   " IAllocator" % self.name)
5507

    
5508
    if instance.disk_template not in constants.DTS_NET_MIRROR:
5509
      raise errors.OpPrereqError("Can't relocate non-mirrored instances")
5510

    
5511
    if len(instance.secondary_nodes) != 1:
5512
      raise errors.OpPrereqError("Instance has not exactly one secondary node")
5513

    
5514
    self.required_nodes = 1
5515

    
5516
    disk_space = _ComputeDiskSize(instance.disk_template,
5517
                                  instance.disks[0].size,
5518
                                  instance.disks[1].size)
5519

    
5520
    request = {
5521
      "type": "relocate",
5522
      "name": self.name,
5523
      "disk_space_total": disk_space,
5524
      "required_nodes": self.required_nodes,
5525
      "relocate_from": self.relocate_from,
5526
      }
5527
    self.in_data["request"] = request
5528

    
5529
  def _BuildInputData(self):
5530
    """Build input data structures.
5531

5532
    """
5533
    self._ComputeClusterData()
5534

    
5535
    if self.mode == constants.IALLOCATOR_MODE_ALLOC:
5536
      self._AddNewInstance()
5537
    else:
5538
      self._AddRelocateInstance()
5539

    
5540
    self.in_text = serializer.Dump(self.in_data)
5541

    
5542
  def Run(self, name, validate=True, call_fn=None):
5543
    """Run an instance allocator and return the results.
5544

5545
    """
5546
    if call_fn is None:
5547
      call_fn = self.lu.rpc.call_iallocator_runner
5548
    data = self.in_text
5549

    
5550
    result = call_fn(self.lu.cfg.GetMasterNode(), name, self.in_text)
5551

    
5552
    if not isinstance(result, (list, tuple)) or len(result) != 4:
5553
      raise errors.OpExecError("Invalid result from master iallocator runner")
5554

    
5555
    rcode, stdout, stderr, fail = result
5556

    
5557
    if rcode == constants.IARUN_NOTFOUND:
5558
      raise errors.OpExecError("Can't find allocator '%s'" % name)
5559
    elif rcode == constants.IARUN_FAILURE:
5560
      raise errors.OpExecError("Instance allocator call failed: %s,"
5561
                               " output: %s" % (fail, stdout+stderr))
5562
    self.out_text = stdout
5563
    if validate:
5564
      self._ValidateResult()
5565

    
5566
  def _ValidateResult(self):
5567
    """Process the allocator results.
5568

5569
    This will process and if successful save the result in
5570
    self.out_data and the other parameters.
5571

5572
    """
5573
    try:
5574
      rdict = serializer.Load(self.out_text)
5575
    except Exception, err:
5576
      raise errors.OpExecError("Can't parse iallocator results: %s" % str(err))
5577

    
5578
    if not isinstance(rdict, dict):
5579
      raise errors.OpExecError("Can't parse iallocator results: not a dict")
5580

    
5581
    for key in "success", "info", "nodes":
5582
      if key not in rdict:
5583
        raise errors.OpExecError("Can't parse iallocator results:"
5584
                                 " missing key '%s'" % key)
5585
      setattr(self, key, rdict[key])
5586

    
5587
    if not isinstance(rdict["nodes"], list):
5588
      raise errors.OpExecError("Can't parse iallocator results: 'nodes' key"
5589
                               " is not a list")
5590
    self.out_data = rdict
5591

    
5592

    
5593
class LUTestAllocator(NoHooksLU):
5594
  """Run allocator tests.
5595

5596
  This LU runs the allocator tests
5597

5598
  """
5599
  _OP_REQP = ["direction", "mode", "name"]
5600

    
5601
  def CheckPrereq(self):
5602
    """Check prerequisites.
5603

5604
    This checks the opcode parameters depending on the director and mode test.
5605

5606
    """
5607
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
5608
      for attr in ["name", "mem_size", "disks", "disk_template",
5609
                   "os", "tags", "nics", "vcpus"]:
5610
        if not hasattr(self.op, attr):
5611
          raise errors.OpPrereqError("Missing attribute '%s' on opcode input" %
5612
                                     attr)
5613
      iname = self.cfg.ExpandInstanceName(self.op.name)
5614
      if iname is not None:
5615
        raise errors.OpPrereqError("Instance '%s' already in the cluster" %
5616
                                   iname)
5617
      if not isinstance(self.op.nics, list):
5618
        raise errors.OpPrereqError("Invalid parameter 'nics'")
5619
      for row in self.op.nics:
5620
        if (not isinstance(row, dict) or
5621
            "mac" not in row or
5622
            "ip" not in row or
5623
            "bridge" not in row):
5624
          raise errors.OpPrereqError("Invalid contents of the"
5625
                                     " 'nics' parameter")
5626
      if not isinstance(self.op.disks, list):
5627
        raise errors.OpPrereqError("Invalid parameter 'disks'")
5628
      if len(self.op.disks) != 2:
5629
        raise errors.OpPrereqError("Only two-disk configurations supported")
5630
      for row in self.op.disks:
5631
        if (not isinstance(row, dict) or
5632
            "size" not in row or
5633
            not isinstance(row["size"], int) or
5634
            "mode" not in row or
5635
            row["mode"] not in ['r', 'w']):
5636
          raise errors.OpPrereqError("Invalid contents of the"
5637
                                     " 'disks' parameter")
5638
    elif self.op.mode == constants.IALLOCATOR_MODE_RELOC:
5639
      if not hasattr(self.op, "name"):
5640
        raise errors.OpPrereqError("Missing attribute 'name' on opcode input")
5641
      fname = self.cfg.ExpandInstanceName(self.op.name)
5642
      if fname is None:
5643
        raise errors.OpPrereqError("Instance '%s' not found for relocation" %
5644
                                   self.op.name)
5645
      self.op.name = fname
5646
      self.relocate_from = self.cfg.GetInstanceInfo(fname).secondary_nodes
5647
    else:
5648
      raise errors.OpPrereqError("Invalid test allocator mode '%s'" %
5649
                                 self.op.mode)
5650

    
5651
    if self.op.direction == constants.IALLOCATOR_DIR_OUT:
5652
      if not hasattr(self.op, "allocator") or self.op.allocator is None:
5653
        raise errors.OpPrereqError("Missing allocator name")
5654
    elif self.op.direction != constants.IALLOCATOR_DIR_IN:
5655
      raise errors.OpPrereqError("Wrong allocator test '%s'" %
5656
                                 self.op.direction)
5657

    
5658
  def Exec(self, feedback_fn):
5659
    """Run the allocator test.
5660

5661
    """
5662
    if self.op.mode == constants.IALLOCATOR_MODE_ALLOC:
5663
      ial = IAllocator(self,
5664
                       mode=self.op.mode,
5665
                       name=self.op.name,
5666
                       mem_size=self.op.mem_size,
5667
                       disks=self.op.disks,
5668
                       disk_template=self.op.disk_template,
5669
                       os=self.op.os,
5670
                       tags=self.op.tags,
5671
                       nics=self.op.nics,
5672
                       vcpus=self.op.vcpus,
5673
                       )
5674
    else:
5675
      ial = IAllocator(self,
5676
                       mode=self.op.mode,
5677
                       name=self.op.name,
5678
                       relocate_from=list(self.relocate_from),
5679
                       )
5680

    
5681
    if self.op.direction == constants.IALLOCATOR_DIR_IN:
5682
      result = ial.in_text
5683
    else:
5684
      ial.Run(self.op.allocator, validate=False)
5685
      result = ial.out_text
5686
    return result