Statistics
| Branch: | Tag: | Revision:

root / lib / config.py @ 1f1d3bf2

History | View | Annotate | Download (84.7 kB)

1
#
2
#
3

    
4
# Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011, 2012 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
"""Configuration management for Ganeti
23

24
This module provides the interface to the Ganeti cluster configuration.
25

26
The configuration data is stored on every node but is updated on the master
27
only. After each update, the master distributes the data to the other nodes.
28

29
Currently, the data storage format is JSON. YAML was slow and consuming too
30
much memory.
31

32
"""
33

    
34
# pylint: disable=R0904
35
# R0904: Too many public methods
36

    
37
import os
38
import random
39
import logging
40
import time
41
import itertools
42

    
43
from ganeti import errors
44
from ganeti import locking
45
from ganeti import utils
46
from ganeti import constants
47
from ganeti import rpc
48
from ganeti import objects
49
from ganeti import serializer
50
from ganeti import uidpool
51
from ganeti import netutils
52
from ganeti import runtime
53
from ganeti import pathutils
54
from ganeti import network
55

    
56

    
57
_config_lock = locking.SharedLock("ConfigWriter")
58

    
59
# job id used for resource management at config upgrade time
60
_UPGRADE_CONFIG_JID = "jid-cfg-upgrade"
61

    
62

    
63
def _ValidateConfig(data):
64
  """Verifies that a configuration objects looks valid.
65

66
  This only verifies the version of the configuration.
67

68
  @raise errors.ConfigurationError: if the version differs from what
69
      we expect
70

71
  """
72
  if data.version != constants.CONFIG_VERSION:
73
    raise errors.ConfigVersionMismatch(constants.CONFIG_VERSION, data.version)
74

    
75

    
76
class TemporaryReservationManager:
77
  """A temporary resource reservation manager.
78

79
  This is used to reserve resources in a job, before using them, making sure
80
  other jobs cannot get them in the meantime.
81

82
  """
83
  def __init__(self):
84
    self._ec_reserved = {}
85

    
86
  def Reserved(self, resource):
87
    for holder_reserved in self._ec_reserved.values():
88
      if resource in holder_reserved:
89
        return True
90
    return False
91

    
92
  def Reserve(self, ec_id, resource):
93
    if self.Reserved(resource):
94
      raise errors.ReservationError("Duplicate reservation for resource '%s'"
95
                                    % str(resource))
96
    if ec_id not in self._ec_reserved:
97
      self._ec_reserved[ec_id] = set([resource])
98
    else:
99
      self._ec_reserved[ec_id].add(resource)
100

    
101
  def DropECReservations(self, ec_id):
102
    if ec_id in self._ec_reserved:
103
      del self._ec_reserved[ec_id]
104

    
105
  def GetReserved(self):
106
    all_reserved = set()
107
    for holder_reserved in self._ec_reserved.values():
108
      all_reserved.update(holder_reserved)
109
    return all_reserved
110

    
111
  def GetECReserved(self, ec_id):
112
    """ Used when you want to retrieve all reservations for a specific
113
        execution context. E.g when commiting reserved IPs for a specific
114
        network.
115

116
    """
117
    ec_reserved = set()
118
    if ec_id in self._ec_reserved:
119
      ec_reserved.update(self._ec_reserved[ec_id])
120
    return ec_reserved
121

    
122
  def Generate(self, existing, generate_one_fn, ec_id):
123
    """Generate a new resource of this type
124

125
    """
126
    assert callable(generate_one_fn)
127

    
128
    all_elems = self.GetReserved()
129
    all_elems.update(existing)
130
    retries = 64
131
    while retries > 0:
132
      new_resource = generate_one_fn()
133
      if new_resource is not None and new_resource not in all_elems:
134
        break
135
    else:
136
      raise errors.ConfigurationError("Not able generate new resource"
137
                                      " (last tried: %s)" % new_resource)
138
    self.Reserve(ec_id, new_resource)
139
    return new_resource
140

    
141

    
142
def _MatchNameComponentIgnoreCase(short_name, names):
143
  """Wrapper around L{utils.text.MatchNameComponent}.
144

145
  """
146
  return utils.MatchNameComponent(short_name, names, case_sensitive=False)
147

    
148

    
149
def _CheckInstanceDiskIvNames(disks):
150
  """Checks if instance's disks' C{iv_name} attributes are in order.
151

152
  @type disks: list of L{objects.Disk}
153
  @param disks: List of disks
154
  @rtype: list of tuples; (int, string, string)
155
  @return: List of wrongly named disks, each tuple contains disk index,
156
    expected and actual name
157

158
  """
159
  result = []
160

    
161
  for (idx, disk) in enumerate(disks):
162
    exp_iv_name = "disk/%s" % idx
163
    if disk.iv_name != exp_iv_name:
164
      result.append((idx, exp_iv_name, disk.iv_name))
165

    
166
  return result
167

    
168

    
169
class ConfigWriter:
170
  """The interface to the cluster configuration.
171

172
  @ivar _temporary_lvs: reservation manager for temporary LVs
173
  @ivar _all_rms: a list of all temporary reservation managers
174

175
  """
176
  def __init__(self, cfg_file=None, offline=False, _getents=runtime.GetEnts,
177
               accept_foreign=False):
178
    self.write_count = 0
179
    self._lock = _config_lock
180
    self._config_data = None
181
    self._offline = offline
182
    if cfg_file is None:
183
      self._cfg_file = pathutils.CLUSTER_CONF_FILE
184
    else:
185
      self._cfg_file = cfg_file
186
    self._getents = _getents
187
    self._temporary_ids = TemporaryReservationManager()
188
    self._temporary_drbds = {}
189
    self._temporary_macs = TemporaryReservationManager()
190
    self._temporary_secrets = TemporaryReservationManager()
191
    self._temporary_lvs = TemporaryReservationManager()
192
    self._temporary_ips = TemporaryReservationManager()
193
    self._all_rms = [self._temporary_ids, self._temporary_macs,
194
                     self._temporary_secrets, self._temporary_lvs,
195
                     self._temporary_ips]
196
    # Note: in order to prevent errors when resolving our name in
197
    # _DistributeConfig, we compute it here once and reuse it; it's
198
    # better to raise an error before starting to modify the config
199
    # file than after it was modified
200
    self._my_hostname = netutils.Hostname.GetSysName()
201
    self._last_cluster_serial = -1
202
    self._cfg_id = None
203
    self._context = None
204
    self._OpenConfig(accept_foreign)
205

    
206
  def _GetRpc(self, address_list):
207
    """Returns RPC runner for configuration.
208

209
    """
210
    return rpc.ConfigRunner(self._context, address_list)
211

    
212
  def SetContext(self, context):
213
    """Sets Ganeti context.
214

215
    """
216
    self._context = context
217

    
218
  # this method needs to be static, so that we can call it on the class
219
  @staticmethod
220
  def IsCluster():
221
    """Check if the cluster is configured.
222

223
    """
224
    return os.path.exists(pathutils.CLUSTER_CONF_FILE)
225

    
226
  @locking.ssynchronized(_config_lock, shared=1)
227
  def GetNdParams(self, node):
228
    """Get the node params populated with cluster defaults.
229

230
    @type node: L{objects.Node}
231
    @param node: The node we want to know the params for
232
    @return: A dict with the filled in node params
233

234
    """
235
    nodegroup = self._UnlockedGetNodeGroup(node.group)
236
    return self._config_data.cluster.FillND(node, nodegroup)
237

    
238
  @locking.ssynchronized(_config_lock, shared=1)
239
  def GetInstanceDiskParams(self, instance):
240
    """Get the disk params populated with inherit chain.
241

242
    @type instance: L{objects.Instance}
243
    @param instance: The instance we want to know the params for
244
    @return: A dict with the filled in disk params
245

246
    """
247
    node = self._UnlockedGetNodeInfo(instance.primary_node)
248
    nodegroup = self._UnlockedGetNodeGroup(node.group)
249
    return self._UnlockedGetGroupDiskParams(nodegroup)
250

    
251
  @locking.ssynchronized(_config_lock, shared=1)
252
  def GetGroupDiskParams(self, group):
253
    """Get the disk params populated with inherit chain.
254

255
    @type group: L{objects.NodeGroup}
256
    @param group: The group we want to know the params for
257
    @return: A dict with the filled in disk params
258

259
    """
260
    return self._UnlockedGetGroupDiskParams(group)
261

    
262
  def _UnlockedGetGroupDiskParams(self, group):
263
    """Get the disk params populated with inherit chain down to node-group.
264

265
    @type group: L{objects.NodeGroup}
266
    @param group: The group we want to know the params for
267
    @return: A dict with the filled in disk params
268

269
    """
270
    return self._config_data.cluster.SimpleFillDP(group.diskparams)
271

    
272
  def _UnlockedGetNetworkMACPrefix(self, net):
273
    """Return the network mac prefix if it exists or the cluster level default.
274

275
    """
276
    prefix = None
277
    if net:
278
      net_uuid = self._UnlockedLookupNetwork(net)
279
      if net_uuid:
280
        nobj = self._UnlockedGetNetwork(net_uuid)
281
        if nobj.mac_prefix:
282
          prefix = nobj.mac_prefix
283

    
284
    return prefix
285

    
286
  def _GenerateOneMAC(self, prefix=None):
287
    """Return a function that randomly generates a MAC suffic
288
       and appends it to the given prefix. If prefix is not given get
289
       the cluster level default.
290

291
    """
292
    if not prefix:
293
      prefix = self._config_data.cluster.mac_prefix
294

    
295
    def GenMac():
296
      byte1 = random.randrange(0, 256)
297
      byte2 = random.randrange(0, 256)
298
      byte3 = random.randrange(0, 256)
299
      mac = "%s:%02x:%02x:%02x" % (prefix, byte1, byte2, byte3)
300
      return mac
301

    
302
    return GenMac
303

    
304
  @locking.ssynchronized(_config_lock, shared=1)
305
  def GenerateMAC(self, net, ec_id):
306
    """Generate a MAC for an instance.
307

308
    This should check the current instances for duplicates.
309

310
    """
311
    existing = self._AllMACs()
312
    prefix = self._UnlockedGetNetworkMACPrefix(net)
313
    gen_mac = self._GenerateOneMAC(prefix)
314
    return self._temporary_ids.Generate(existing, gen_mac, ec_id)
315

    
316
  @locking.ssynchronized(_config_lock, shared=1)
317
  def ReserveMAC(self, mac, ec_id):
318
    """Reserve a MAC for an instance.
319

320
    This only checks instances managed by this cluster, it does not
321
    check for potential collisions elsewhere.
322

323
    """
324
    all_macs = self._AllMACs()
325
    if mac in all_macs:
326
      raise errors.ReservationError("mac already in use")
327
    else:
328
      self._temporary_macs.Reserve(ec_id, mac)
329

    
330
  def _UnlockedCommitTemporaryIps(self, ec_id):
331
    """Commit all reserved IP address to their respective pools
332

333
    """
334
    for action, address, net_uuid in self._temporary_ips.GetECReserved(ec_id):
335
      self._UnlockedCommitIp(action, net_uuid, address)
336

    
337
  def _UnlockedCommitIp(self, action, net_uuid, address):
338
    """Commit a reserved IP address to an IP pool.
339

340
    The IP address is taken from the network's IP pool and marked as reserved.
341

342
    """
343
    nobj = self._UnlockedGetNetwork(net_uuid)
344
    pool = network.AddressPool(nobj)
345
    if action == constants.RESERVE_ACTION:
346
      pool.Reserve(address)
347
    elif action == constants.RELEASE_ACTION:
348
      pool.Release(address)
349

    
350
  def _UnlockedReleaseIp(self, net_uuid, address, ec_id):
351
    """Give a specific IP address back to an IP pool.
352

353
    The IP address is returned to the IP pool designated by pool_id and marked
354
    as reserved.
355

356
    """
357
    self._temporary_ips.Reserve(ec_id,
358
                                (constants.RELEASE_ACTION, address, net_uuid))
359

    
360
  @locking.ssynchronized(_config_lock, shared=1)
361
  def ReleaseIp(self, net, address, ec_id):
362
    """Give a specified IP address back to an IP pool.
363

364
    This is just a wrapper around _UnlockedReleaseIp.
365

366
    """
367
    net_uuid = self._UnlockedLookupNetwork(net)
368
    if net_uuid:
369
      self._UnlockedReleaseIp(net_uuid, address, ec_id)
370

    
371
  @locking.ssynchronized(_config_lock, shared=1)
372
  def GenerateIp(self, net, ec_id):
373
    """Find a free IPv4 address for an instance.
374

375
    """
376
    net_uuid = self._UnlockedLookupNetwork(net)
377
    nobj = self._UnlockedGetNetwork(net_uuid)
378
    pool = network.AddressPool(nobj)
379

    
380
    def gen_one():
381
      try:
382
        ip = pool.GenerateFree()
383
      except errors.AddressPoolError:
384
        raise errors.ReservationError("Cannot generate IP. Network is full")
385
      return (constants.RESERVE_ACTION, ip, net_uuid)
386

    
387
    _, address, _ = self._temporary_ips.Generate([], gen_one, ec_id)
388
    return address
389

    
390
  def _UnlockedReserveIp(self, net_uuid, address, ec_id):
391
    """Reserve a given IPv4 address for use by an instance.
392

393
    """
394
    nobj = self._UnlockedGetNetwork(net_uuid)
395
    pool = network.AddressPool(nobj)
396
    try:
397
      isreserved = pool.IsReserved(address)
398
    except errors.AddressPoolError:
399
      raise errors.ReservationError("IP address not in network")
400
    if isreserved:
401
      raise errors.ReservationError("IP address already in use")
402

    
403
    return self._temporary_ips.Reserve(ec_id,
404
                                       (constants.RESERVE_ACTION,
405
                                        address, net_uuid))
406

    
407
  @locking.ssynchronized(_config_lock, shared=1)
408
  def ReserveIp(self, net, address, ec_id):
409
    """Reserve a given IPv4 address for use by an instance.
410

411
    """
412
    net_uuid = self._UnlockedLookupNetwork(net)
413
    if net_uuid:
414
      return self._UnlockedReserveIp(net_uuid, address, ec_id)
415

    
416
  @locking.ssynchronized(_config_lock, shared=1)
417
  def ReserveLV(self, lv_name, ec_id):
418
    """Reserve an VG/LV pair for an instance.
419

420
    @type lv_name: string
421
    @param lv_name: the logical volume name to reserve
422

423
    """
424
    all_lvs = self._AllLVs()
425
    if lv_name in all_lvs:
426
      raise errors.ReservationError("LV already in use")
427
    else:
428
      self._temporary_lvs.Reserve(ec_id, lv_name)
429

    
430
  @locking.ssynchronized(_config_lock, shared=1)
431
  def GenerateDRBDSecret(self, ec_id):
432
    """Generate a DRBD secret.
433

434
    This checks the current disks for duplicates.
435

436
    """
437
    return self._temporary_secrets.Generate(self._AllDRBDSecrets(),
438
                                            utils.GenerateSecret,
439
                                            ec_id)
440

    
441
  def _AllLVs(self):
442
    """Compute the list of all LVs.
443

444
    """
445
    lvnames = set()
446
    for instance in self._config_data.instances.values():
447
      node_data = instance.MapLVsByNode()
448
      for lv_list in node_data.values():
449
        lvnames.update(lv_list)
450
    return lvnames
451

    
452
  def _AllIDs(self, include_temporary):
453
    """Compute the list of all UUIDs and names we have.
454

455
    @type include_temporary: boolean
456
    @param include_temporary: whether to include the _temporary_ids set
457
    @rtype: set
458
    @return: a set of IDs
459

460
    """
461
    existing = set()
462
    if include_temporary:
463
      existing.update(self._temporary_ids.GetReserved())
464
    existing.update(self._AllLVs())
465
    existing.update(self._config_data.instances.keys())
466
    existing.update(self._config_data.nodes.keys())
467
    existing.update([i.uuid for i in self._AllUUIDObjects() if i.uuid])
468
    return existing
469

    
470
  def _GenerateUniqueID(self, ec_id):
471
    """Generate an unique UUID.
472

473
    This checks the current node, instances and disk names for
474
    duplicates.
475

476
    @rtype: string
477
    @return: the unique id
478

479
    """
480
    existing = self._AllIDs(include_temporary=False)
481
    return self._temporary_ids.Generate(existing, utils.NewUUID, ec_id)
482

    
483
  @locking.ssynchronized(_config_lock, shared=1)
484
  def GenerateUniqueID(self, ec_id):
485
    """Generate an unique ID.
486

487
    This is just a wrapper over the unlocked version.
488

489
    @type ec_id: string
490
    @param ec_id: unique id for the job to reserve the id to
491

492
    """
493
    return self._GenerateUniqueID(ec_id)
494

    
495
  def _AllMACs(self):
496
    """Return all MACs present in the config.
497

498
    @rtype: list
499
    @return: the list of all MACs
500

501
    """
502
    result = []
503
    for instance in self._config_data.instances.values():
504
      for nic in instance.nics:
505
        result.append(nic.mac)
506

    
507
    return result
508

    
509
  def _AllDRBDSecrets(self):
510
    """Return all DRBD secrets present in the config.
511

512
    @rtype: list
513
    @return: the list of all DRBD secrets
514

515
    """
516
    def helper(disk, result):
517
      """Recursively gather secrets from this disk."""
518
      if disk.dev_type == constants.DT_DRBD8:
519
        result.append(disk.logical_id[5])
520
      if disk.children:
521
        for child in disk.children:
522
          helper(child, result)
523

    
524
    result = []
525
    for instance in self._config_data.instances.values():
526
      for disk in instance.disks:
527
        helper(disk, result)
528

    
529
    return result
530

    
531
  def _CheckDiskIDs(self, disk, l_ids, p_ids):
532
    """Compute duplicate disk IDs
533

534
    @type disk: L{objects.Disk}
535
    @param disk: the disk at which to start searching
536
    @type l_ids: list
537
    @param l_ids: list of current logical ids
538
    @type p_ids: list
539
    @param p_ids: list of current physical ids
540
    @rtype: list
541
    @return: a list of error messages
542

543
    """
544
    result = []
545
    if disk.logical_id is not None:
546
      if disk.logical_id in l_ids:
547
        result.append("duplicate logical id %s" % str(disk.logical_id))
548
      else:
549
        l_ids.append(disk.logical_id)
550
    if disk.physical_id is not None:
551
      if disk.physical_id in p_ids:
552
        result.append("duplicate physical id %s" % str(disk.physical_id))
553
      else:
554
        p_ids.append(disk.physical_id)
555

    
556
    if disk.children:
557
      for child in disk.children:
558
        result.extend(self._CheckDiskIDs(child, l_ids, p_ids))
559
    return result
560

    
561
  def _UnlockedVerifyConfig(self):
562
    """Verify function.
563

564
    @rtype: list
565
    @return: a list of error messages; a non-empty list signifies
566
        configuration errors
567

568
    """
569
    # pylint: disable=R0914
570
    result = []
571
    seen_macs = []
572
    ports = {}
573
    data = self._config_data
574
    cluster = data.cluster
575
    seen_lids = []
576
    seen_pids = []
577

    
578
    # global cluster checks
579
    if not cluster.enabled_hypervisors:
580
      result.append("enabled hypervisors list doesn't have any entries")
581
    invalid_hvs = set(cluster.enabled_hypervisors) - constants.HYPER_TYPES
582
    if invalid_hvs:
583
      result.append("enabled hypervisors contains invalid entries: %s" %
584
                    invalid_hvs)
585
    missing_hvp = (set(cluster.enabled_hypervisors) -
586
                   set(cluster.hvparams.keys()))
587
    if missing_hvp:
588
      result.append("hypervisor parameters missing for the enabled"
589
                    " hypervisor(s) %s" % utils.CommaJoin(missing_hvp))
590

    
591
    if cluster.master_node not in data.nodes:
592
      result.append("cluster has invalid primary node '%s'" %
593
                    cluster.master_node)
594

    
595
    def _helper(owner, attr, value, template):
596
      try:
597
        utils.ForceDictType(value, template)
598
      except errors.GenericError, err:
599
        result.append("%s has invalid %s: %s" % (owner, attr, err))
600

    
601
    def _helper_nic(owner, params):
602
      try:
603
        objects.NIC.CheckParameterSyntax(params)
604
      except errors.ConfigurationError, err:
605
        result.append("%s has invalid nicparams: %s" % (owner, err))
606

    
607
    def _helper_ipolicy(owner, params, check_std):
608
      try:
609
        objects.InstancePolicy.CheckParameterSyntax(params, check_std)
610
      except errors.ConfigurationError, err:
611
        result.append("%s has invalid instance policy: %s" % (owner, err))
612

    
613
    def _helper_ispecs(owner, params):
614
      for key, value in params.items():
615
        if key in constants.IPOLICY_ISPECS:
616
          fullkey = "ipolicy/" + key
617
          _helper(owner, fullkey, value, constants.ISPECS_PARAMETER_TYPES)
618
        else:
619
          # FIXME: assuming list type
620
          if key in constants.IPOLICY_PARAMETERS:
621
            exp_type = float
622
          else:
623
            exp_type = list
624
          if not isinstance(value, exp_type):
625
            result.append("%s has invalid instance policy: for %s,"
626
                          " expecting %s, got %s" %
627
                          (owner, key, exp_type.__name__, type(value)))
628

    
629
    # check cluster parameters
630
    _helper("cluster", "beparams", cluster.SimpleFillBE({}),
631
            constants.BES_PARAMETER_TYPES)
632
    _helper("cluster", "nicparams", cluster.SimpleFillNIC({}),
633
            constants.NICS_PARAMETER_TYPES)
634
    _helper_nic("cluster", cluster.SimpleFillNIC({}))
635
    _helper("cluster", "ndparams", cluster.SimpleFillND({}),
636
            constants.NDS_PARAMETER_TYPES)
637
    _helper_ipolicy("cluster", cluster.SimpleFillIPolicy({}), True)
638
    _helper_ispecs("cluster", cluster.SimpleFillIPolicy({}))
639

    
640
    # per-instance checks
641
    for instance_name in data.instances:
642
      instance = data.instances[instance_name]
643
      if instance.name != instance_name:
644
        result.append("instance '%s' is indexed by wrong name '%s'" %
645
                      (instance.name, instance_name))
646
      if instance.primary_node not in data.nodes:
647
        result.append("instance '%s' has invalid primary node '%s'" %
648
                      (instance_name, instance.primary_node))
649
      for snode in instance.secondary_nodes:
650
        if snode not in data.nodes:
651
          result.append("instance '%s' has invalid secondary node '%s'" %
652
                        (instance_name, snode))
653
      for idx, nic in enumerate(instance.nics):
654
        if nic.mac in seen_macs:
655
          result.append("instance '%s' has NIC %d mac %s duplicate" %
656
                        (instance_name, idx, nic.mac))
657
        else:
658
          seen_macs.append(nic.mac)
659
        if nic.nicparams:
660
          filled = cluster.SimpleFillNIC(nic.nicparams)
661
          owner = "instance %s nic %d" % (instance.name, idx)
662
          _helper(owner, "nicparams",
663
                  filled, constants.NICS_PARAMETER_TYPES)
664
          _helper_nic(owner, filled)
665

    
666
      # parameter checks
667
      if instance.beparams:
668
        _helper("instance %s" % instance.name, "beparams",
669
                cluster.FillBE(instance), constants.BES_PARAMETER_TYPES)
670

    
671
      # gather the drbd ports for duplicate checks
672
      for (idx, dsk) in enumerate(instance.disks):
673
        if dsk.dev_type in constants.LDS_DRBD:
674
          tcp_port = dsk.logical_id[2]
675
          if tcp_port not in ports:
676
            ports[tcp_port] = []
677
          ports[tcp_port].append((instance.name, "drbd disk %s" % idx))
678
      # gather network port reservation
679
      net_port = getattr(instance, "network_port", None)
680
      if net_port is not None:
681
        if net_port not in ports:
682
          ports[net_port] = []
683
        ports[net_port].append((instance.name, "network port"))
684

    
685
      # instance disk verify
686
      for idx, disk in enumerate(instance.disks):
687
        result.extend(["instance '%s' disk %d error: %s" %
688
                       (instance.name, idx, msg) for msg in disk.Verify()])
689
        result.extend(self._CheckDiskIDs(disk, seen_lids, seen_pids))
690

    
691
      wrong_names = _CheckInstanceDiskIvNames(instance.disks)
692
      if wrong_names:
693
        tmp = "; ".join(("name of disk %s should be '%s', but is '%s'" %
694
                         (idx, exp_name, actual_name))
695
                        for (idx, exp_name, actual_name) in wrong_names)
696

    
697
        result.append("Instance '%s' has wrongly named disks: %s" %
698
                      (instance.name, tmp))
699

    
700
    # cluster-wide pool of free ports
701
    for free_port in cluster.tcpudp_port_pool:
702
      if free_port not in ports:
703
        ports[free_port] = []
704
      ports[free_port].append(("cluster", "port marked as free"))
705

    
706
    # compute tcp/udp duplicate ports
707
    keys = ports.keys()
708
    keys.sort()
709
    for pnum in keys:
710
      pdata = ports[pnum]
711
      if len(pdata) > 1:
712
        txt = utils.CommaJoin(["%s/%s" % val for val in pdata])
713
        result.append("tcp/udp port %s has duplicates: %s" % (pnum, txt))
714

    
715
    # highest used tcp port check
716
    if keys:
717
      if keys[-1] > cluster.highest_used_port:
718
        result.append("Highest used port mismatch, saved %s, computed %s" %
719
                      (cluster.highest_used_port, keys[-1]))
720

    
721
    if not data.nodes[cluster.master_node].master_candidate:
722
      result.append("Master node is not a master candidate")
723

    
724
    # master candidate checks
725
    mc_now, mc_max, _ = self._UnlockedGetMasterCandidateStats()
726
    if mc_now < mc_max:
727
      result.append("Not enough master candidates: actual %d, target %d" %
728
                    (mc_now, mc_max))
729

    
730
    # node checks
731
    for node_name, node in data.nodes.items():
732
      if node.name != node_name:
733
        result.append("Node '%s' is indexed by wrong name '%s'" %
734
                      (node.name, node_name))
735
      if [node.master_candidate, node.drained, node.offline].count(True) > 1:
736
        result.append("Node %s state is invalid: master_candidate=%s,"
737
                      " drain=%s, offline=%s" %
738
                      (node.name, node.master_candidate, node.drained,
739
                       node.offline))
740
      if node.group not in data.nodegroups:
741
        result.append("Node '%s' has invalid group '%s'" %
742
                      (node.name, node.group))
743
      else:
744
        _helper("node %s" % node.name, "ndparams",
745
                cluster.FillND(node, data.nodegroups[node.group]),
746
                constants.NDS_PARAMETER_TYPES)
747

    
748
    # nodegroups checks
749
    nodegroups_names = set()
750
    for nodegroup_uuid in data.nodegroups:
751
      nodegroup = data.nodegroups[nodegroup_uuid]
752
      if nodegroup.uuid != nodegroup_uuid:
753
        result.append("node group '%s' (uuid: '%s') indexed by wrong uuid '%s'"
754
                      % (nodegroup.name, nodegroup.uuid, nodegroup_uuid))
755
      if utils.UUID_RE.match(nodegroup.name.lower()):
756
        result.append("node group '%s' (uuid: '%s') has uuid-like name" %
757
                      (nodegroup.name, nodegroup.uuid))
758
      if nodegroup.name in nodegroups_names:
759
        result.append("duplicate node group name '%s'" % nodegroup.name)
760
      else:
761
        nodegroups_names.add(nodegroup.name)
762
      group_name = "group %s" % nodegroup.name
763
      _helper_ipolicy(group_name, cluster.SimpleFillIPolicy(nodegroup.ipolicy),
764
                      False)
765
      _helper_ispecs(group_name, cluster.SimpleFillIPolicy(nodegroup.ipolicy))
766
      if nodegroup.ndparams:
767
        _helper(group_name, "ndparams",
768
                cluster.SimpleFillND(nodegroup.ndparams),
769
                constants.NDS_PARAMETER_TYPES)
770

    
771
    # drbd minors check
772
    _, duplicates = self._UnlockedComputeDRBDMap()
773
    for node, minor, instance_a, instance_b in duplicates:
774
      result.append("DRBD minor %d on node %s is assigned twice to instances"
775
                    " %s and %s" % (minor, node, instance_a, instance_b))
776

    
777
    # IP checks
778
    default_nicparams = cluster.nicparams[constants.PP_DEFAULT]
779
    ips = {}
780

    
781
    def _AddIpAddress(ip, name):
782
      ips.setdefault(ip, []).append(name)
783

    
784
    _AddIpAddress(cluster.master_ip, "cluster_ip")
785

    
786
    for node in data.nodes.values():
787
      _AddIpAddress(node.primary_ip, "node:%s/primary" % node.name)
788
      if node.secondary_ip != node.primary_ip:
789
        _AddIpAddress(node.secondary_ip, "node:%s/secondary" % node.name)
790

    
791
    for instance in data.instances.values():
792
      for idx, nic in enumerate(instance.nics):
793
        if nic.ip is None:
794
          continue
795

    
796
        nicparams = objects.FillDict(default_nicparams, nic.nicparams)
797
        nic_mode = nicparams[constants.NIC_MODE]
798
        nic_link = nicparams[constants.NIC_LINK]
799

    
800
        if nic_mode == constants.NIC_MODE_BRIDGED:
801
          link = "bridge:%s" % nic_link
802
        elif nic_mode == constants.NIC_MODE_ROUTED:
803
          link = "route:%s" % nic_link
804
        else:
805
          raise errors.ProgrammerError("NIC mode '%s' not handled" % nic_mode)
806

    
807
        _AddIpAddress("%s/%s/%s" % (link, nic.ip, nic.network),
808
                      "instance:%s/nic:%d" % (instance.name, idx))
809

    
810
    for ip, owners in ips.items():
811
      if len(owners) > 1:
812
        result.append("IP address %s is used by multiple owners: %s" %
813
                      (ip, utils.CommaJoin(owners)))
814

    
815
    return result
816

    
817
  @locking.ssynchronized(_config_lock, shared=1)
818
  def VerifyConfig(self):
819
    """Verify function.
820

821
    This is just a wrapper over L{_UnlockedVerifyConfig}.
822

823
    @rtype: list
824
    @return: a list of error messages; a non-empty list signifies
825
        configuration errors
826

827
    """
828
    return self._UnlockedVerifyConfig()
829

    
830
  def _UnlockedSetDiskID(self, disk, node_name):
831
    """Convert the unique ID to the ID needed on the target nodes.
832

833
    This is used only for drbd, which needs ip/port configuration.
834

835
    The routine descends down and updates its children also, because
836
    this helps when the only the top device is passed to the remote
837
    node.
838

839
    This function is for internal use, when the config lock is already held.
840

841
    """
842
    if disk.children:
843
      for child in disk.children:
844
        self._UnlockedSetDiskID(child, node_name)
845

    
846
    if disk.logical_id is None and disk.physical_id is not None:
847
      return
848
    if disk.dev_type == constants.LD_DRBD8:
849
      pnode, snode, port, pminor, sminor, secret = disk.logical_id
850
      if node_name not in (pnode, snode):
851
        raise errors.ConfigurationError("DRBD device not knowing node %s" %
852
                                        node_name)
853
      pnode_info = self._UnlockedGetNodeInfo(pnode)
854
      snode_info = self._UnlockedGetNodeInfo(snode)
855
      if pnode_info is None or snode_info is None:
856
        raise errors.ConfigurationError("Can't find primary or secondary node"
857
                                        " for %s" % str(disk))
858
      p_data = (pnode_info.secondary_ip, port)
859
      s_data = (snode_info.secondary_ip, port)
860
      if pnode == node_name:
861
        disk.physical_id = p_data + s_data + (pminor, secret)
862
      else: # it must be secondary, we tested above
863
        disk.physical_id = s_data + p_data + (sminor, secret)
864
    else:
865
      disk.physical_id = disk.logical_id
866
    return
867

    
868
  @locking.ssynchronized(_config_lock)
869
  def SetDiskID(self, disk, node_name):
870
    """Convert the unique ID to the ID needed on the target nodes.
871

872
    This is used only for drbd, which needs ip/port configuration.
873

874
    The routine descends down and updates its children also, because
875
    this helps when the only the top device is passed to the remote
876
    node.
877

878
    """
879
    return self._UnlockedSetDiskID(disk, node_name)
880

    
881
  @locking.ssynchronized(_config_lock)
882
  def AddTcpUdpPort(self, port):
883
    """Adds a new port to the available port pool.
884

885
    @warning: this method does not "flush" the configuration (via
886
        L{_WriteConfig}); callers should do that themselves once the
887
        configuration is stable
888

889
    """
890
    if not isinstance(port, int):
891
      raise errors.ProgrammerError("Invalid type passed for port")
892

    
893
    self._config_data.cluster.tcpudp_port_pool.add(port)
894

    
895
  @locking.ssynchronized(_config_lock, shared=1)
896
  def GetPortList(self):
897
    """Returns a copy of the current port list.
898

899
    """
900
    return self._config_data.cluster.tcpudp_port_pool.copy()
901

    
902
  @locking.ssynchronized(_config_lock)
903
  def AllocatePort(self):
904
    """Allocate a port.
905

906
    The port will be taken from the available port pool or from the
907
    default port range (and in this case we increase
908
    highest_used_port).
909

910
    """
911
    # If there are TCP/IP ports configured, we use them first.
912
    if self._config_data.cluster.tcpudp_port_pool:
913
      port = self._config_data.cluster.tcpudp_port_pool.pop()
914
    else:
915
      port = self._config_data.cluster.highest_used_port + 1
916
      if port >= constants.LAST_DRBD_PORT:
917
        raise errors.ConfigurationError("The highest used port is greater"
918
                                        " than %s. Aborting." %
919
                                        constants.LAST_DRBD_PORT)
920
      self._config_data.cluster.highest_used_port = port
921

    
922
    self._WriteConfig()
923
    return port
924

    
925
  def _UnlockedComputeDRBDMap(self):
926
    """Compute the used DRBD minor/nodes.
927

928
    @rtype: (dict, list)
929
    @return: dictionary of node_name: dict of minor: instance_name;
930
        the returned dict will have all the nodes in it (even if with
931
        an empty list), and a list of duplicates; if the duplicates
932
        list is not empty, the configuration is corrupted and its caller
933
        should raise an exception
934

935
    """
936
    def _AppendUsedPorts(instance_name, disk, used):
937
      duplicates = []
938
      if disk.dev_type == constants.LD_DRBD8 and len(disk.logical_id) >= 5:
939
        node_a, node_b, _, minor_a, minor_b = disk.logical_id[:5]
940
        for node, port in ((node_a, minor_a), (node_b, minor_b)):
941
          assert node in used, ("Node '%s' of instance '%s' not found"
942
                                " in node list" % (node, instance_name))
943
          if port in used[node]:
944
            duplicates.append((node, port, instance_name, used[node][port]))
945
          else:
946
            used[node][port] = instance_name
947
      if disk.children:
948
        for child in disk.children:
949
          duplicates.extend(_AppendUsedPorts(instance_name, child, used))
950
      return duplicates
951

    
952
    duplicates = []
953
    my_dict = dict((node, {}) for node in self._config_data.nodes)
954
    for instance in self._config_data.instances.itervalues():
955
      for disk in instance.disks:
956
        duplicates.extend(_AppendUsedPorts(instance.name, disk, my_dict))
957
    for (node, minor), instance in self._temporary_drbds.iteritems():
958
      if minor in my_dict[node] and my_dict[node][minor] != instance:
959
        duplicates.append((node, minor, instance, my_dict[node][minor]))
960
      else:
961
        my_dict[node][minor] = instance
962
    return my_dict, duplicates
963

    
964
  @locking.ssynchronized(_config_lock)
965
  def ComputeDRBDMap(self):
966
    """Compute the used DRBD minor/nodes.
967

968
    This is just a wrapper over L{_UnlockedComputeDRBDMap}.
969

970
    @return: dictionary of node_name: dict of minor: instance_name;
971
        the returned dict will have all the nodes in it (even if with
972
        an empty list).
973

974
    """
975
    d_map, duplicates = self._UnlockedComputeDRBDMap()
976
    if duplicates:
977
      raise errors.ConfigurationError("Duplicate DRBD ports detected: %s" %
978
                                      str(duplicates))
979
    return d_map
980

    
981
  @locking.ssynchronized(_config_lock)
982
  def AllocateDRBDMinor(self, nodes, instance):
983
    """Allocate a drbd minor.
984

985
    The free minor will be automatically computed from the existing
986
    devices. A node can be given multiple times in order to allocate
987
    multiple minors. The result is the list of minors, in the same
988
    order as the passed nodes.
989

990
    @type instance: string
991
    @param instance: the instance for which we allocate minors
992

993
    """
994
    assert isinstance(instance, basestring), \
995
           "Invalid argument '%s' passed to AllocateDRBDMinor" % instance
996

    
997
    d_map, duplicates = self._UnlockedComputeDRBDMap()
998
    if duplicates:
999
      raise errors.ConfigurationError("Duplicate DRBD ports detected: %s" %
1000
                                      str(duplicates))
1001
    result = []
1002
    for nname in nodes:
1003
      ndata = d_map[nname]
1004
      if not ndata:
1005
        # no minors used, we can start at 0
1006
        result.append(0)
1007
        ndata[0] = instance
1008
        self._temporary_drbds[(nname, 0)] = instance
1009
        continue
1010
      keys = ndata.keys()
1011
      keys.sort()
1012
      ffree = utils.FirstFree(keys)
1013
      if ffree is None:
1014
        # return the next minor
1015
        # TODO: implement high-limit check
1016
        minor = keys[-1] + 1
1017
      else:
1018
        minor = ffree
1019
      # double-check minor against current instances
1020
      assert minor not in d_map[nname], \
1021
             ("Attempt to reuse allocated DRBD minor %d on node %s,"
1022
              " already allocated to instance %s" %
1023
              (minor, nname, d_map[nname][minor]))
1024
      ndata[minor] = instance
1025
      # double-check minor against reservation
1026
      r_key = (nname, minor)
1027
      assert r_key not in self._temporary_drbds, \
1028
             ("Attempt to reuse reserved DRBD minor %d on node %s,"
1029
              " reserved for instance %s" %
1030
              (minor, nname, self._temporary_drbds[r_key]))
1031
      self._temporary_drbds[r_key] = instance
1032
      result.append(minor)
1033
    logging.debug("Request to allocate drbd minors, input: %s, returning %s",
1034
                  nodes, result)
1035
    return result
1036

    
1037
  def _UnlockedReleaseDRBDMinors(self, instance):
1038
    """Release temporary drbd minors allocated for a given instance.
1039

1040
    @type instance: string
1041
    @param instance: the instance for which temporary minors should be
1042
                     released
1043

1044
    """
1045
    assert isinstance(instance, basestring), \
1046
           "Invalid argument passed to ReleaseDRBDMinors"
1047
    for key, name in self._temporary_drbds.items():
1048
      if name == instance:
1049
        del self._temporary_drbds[key]
1050

    
1051
  @locking.ssynchronized(_config_lock)
1052
  def ReleaseDRBDMinors(self, instance):
1053
    """Release temporary drbd minors allocated for a given instance.
1054

1055
    This should be called on the error paths, on the success paths
1056
    it's automatically called by the ConfigWriter add and update
1057
    functions.
1058

1059
    This function is just a wrapper over L{_UnlockedReleaseDRBDMinors}.
1060

1061
    @type instance: string
1062
    @param instance: the instance for which temporary minors should be
1063
                     released
1064

1065
    """
1066
    self._UnlockedReleaseDRBDMinors(instance)
1067

    
1068
  @locking.ssynchronized(_config_lock, shared=1)
1069
  def GetConfigVersion(self):
1070
    """Get the configuration version.
1071

1072
    @return: Config version
1073

1074
    """
1075
    return self._config_data.version
1076

    
1077
  @locking.ssynchronized(_config_lock, shared=1)
1078
  def GetClusterName(self):
1079
    """Get cluster name.
1080

1081
    @return: Cluster name
1082

1083
    """
1084
    return self._config_data.cluster.cluster_name
1085

    
1086
  @locking.ssynchronized(_config_lock, shared=1)
1087
  def GetMasterNode(self):
1088
    """Get the hostname of the master node for this cluster.
1089

1090
    @return: Master hostname
1091

1092
    """
1093
    return self._config_data.cluster.master_node
1094

    
1095
  @locking.ssynchronized(_config_lock, shared=1)
1096
  def GetMasterIP(self):
1097
    """Get the IP of the master node for this cluster.
1098

1099
    @return: Master IP
1100

1101
    """
1102
    return self._config_data.cluster.master_ip
1103

    
1104
  @locking.ssynchronized(_config_lock, shared=1)
1105
  def GetMasterNetdev(self):
1106
    """Get the master network device for this cluster.
1107

1108
    """
1109
    return self._config_data.cluster.master_netdev
1110

    
1111
  @locking.ssynchronized(_config_lock, shared=1)
1112
  def GetMasterNetmask(self):
1113
    """Get the netmask of the master node for this cluster.
1114

1115
    """
1116
    return self._config_data.cluster.master_netmask
1117

    
1118
  @locking.ssynchronized(_config_lock, shared=1)
1119
  def GetUseExternalMipScript(self):
1120
    """Get flag representing whether to use the external master IP setup script.
1121

1122
    """
1123
    return self._config_data.cluster.use_external_mip_script
1124

    
1125
  @locking.ssynchronized(_config_lock, shared=1)
1126
  def GetFileStorageDir(self):
1127
    """Get the file storage dir for this cluster.
1128

1129
    """
1130
    return self._config_data.cluster.file_storage_dir
1131

    
1132
  @locking.ssynchronized(_config_lock, shared=1)
1133
  def GetSharedFileStorageDir(self):
1134
    """Get the shared file storage dir for this cluster.
1135

1136
    """
1137
    return self._config_data.cluster.shared_file_storage_dir
1138

    
1139
  @locking.ssynchronized(_config_lock, shared=1)
1140
  def GetHypervisorType(self):
1141
    """Get the hypervisor type for this cluster.
1142

1143
    """
1144
    return self._config_data.cluster.enabled_hypervisors[0]
1145

    
1146
  @locking.ssynchronized(_config_lock, shared=1)
1147
  def GetHostKey(self):
1148
    """Return the rsa hostkey from the config.
1149

1150
    @rtype: string
1151
    @return: the rsa hostkey
1152

1153
    """
1154
    return self._config_data.cluster.rsahostkeypub
1155

    
1156
  @locking.ssynchronized(_config_lock, shared=1)
1157
  def GetDefaultIAllocator(self):
1158
    """Get the default instance allocator for this cluster.
1159

1160
    """
1161
    return self._config_data.cluster.default_iallocator
1162

    
1163
  @locking.ssynchronized(_config_lock, shared=1)
1164
  def GetPrimaryIPFamily(self):
1165
    """Get cluster primary ip family.
1166

1167
    @return: primary ip family
1168

1169
    """
1170
    return self._config_data.cluster.primary_ip_family
1171

    
1172
  @locking.ssynchronized(_config_lock, shared=1)
1173
  def GetMasterNetworkParameters(self):
1174
    """Get network parameters of the master node.
1175

1176
    @rtype: L{object.MasterNetworkParameters}
1177
    @return: network parameters of the master node
1178

1179
    """
1180
    cluster = self._config_data.cluster
1181
    result = objects.MasterNetworkParameters(
1182
      name=cluster.master_node, ip=cluster.master_ip,
1183
      netmask=cluster.master_netmask, netdev=cluster.master_netdev,
1184
      ip_family=cluster.primary_ip_family)
1185

    
1186
    return result
1187

    
1188
  @locking.ssynchronized(_config_lock)
1189
  def AddNodeGroup(self, group, ec_id, check_uuid=True):
1190
    """Add a node group to the configuration.
1191

1192
    This method calls group.UpgradeConfig() to fill any missing attributes
1193
    according to their default values.
1194

1195
    @type group: L{objects.NodeGroup}
1196
    @param group: the NodeGroup object to add
1197
    @type ec_id: string
1198
    @param ec_id: unique id for the job to use when creating a missing UUID
1199
    @type check_uuid: bool
1200
    @param check_uuid: add an UUID to the group if it doesn't have one or, if
1201
                       it does, ensure that it does not exist in the
1202
                       configuration already
1203

1204
    """
1205
    self._UnlockedAddNodeGroup(group, ec_id, check_uuid)
1206
    self._WriteConfig()
1207

    
1208
  def _UnlockedAddNodeGroup(self, group, ec_id, check_uuid):
1209
    """Add a node group to the configuration.
1210

1211
    """
1212
    logging.info("Adding node group %s to configuration", group.name)
1213

    
1214
    # Some code might need to add a node group with a pre-populated UUID
1215
    # generated with ConfigWriter.GenerateUniqueID(). We allow them to bypass
1216
    # the "does this UUID" exist already check.
1217
    if check_uuid:
1218
      self._EnsureUUID(group, ec_id)
1219

    
1220
    try:
1221
      existing_uuid = self._UnlockedLookupNodeGroup(group.name)
1222
    except errors.OpPrereqError:
1223
      pass
1224
    else:
1225
      raise errors.OpPrereqError("Desired group name '%s' already exists as a"
1226
                                 " node group (UUID: %s)" %
1227
                                 (group.name, existing_uuid),
1228
                                 errors.ECODE_EXISTS)
1229

    
1230
    group.serial_no = 1
1231
    group.ctime = group.mtime = time.time()
1232
    group.UpgradeConfig()
1233

    
1234
    self._config_data.nodegroups[group.uuid] = group
1235
    self._config_data.cluster.serial_no += 1
1236

    
1237
  @locking.ssynchronized(_config_lock)
1238
  def RemoveNodeGroup(self, group_uuid):
1239
    """Remove a node group from the configuration.
1240

1241
    @type group_uuid: string
1242
    @param group_uuid: the UUID of the node group to remove
1243

1244
    """
1245
    logging.info("Removing node group %s from configuration", group_uuid)
1246

    
1247
    if group_uuid not in self._config_data.nodegroups:
1248
      raise errors.ConfigurationError("Unknown node group '%s'" % group_uuid)
1249

    
1250
    assert len(self._config_data.nodegroups) != 1, \
1251
            "Group '%s' is the only group, cannot be removed" % group_uuid
1252

    
1253
    del self._config_data.nodegroups[group_uuid]
1254
    self._config_data.cluster.serial_no += 1
1255
    self._WriteConfig()
1256

    
1257
  def _UnlockedLookupNodeGroup(self, target):
1258
    """Lookup a node group's UUID.
1259

1260
    @type target: string or None
1261
    @param target: group name or UUID or None to look for the default
1262
    @rtype: string
1263
    @return: nodegroup UUID
1264
    @raises errors.OpPrereqError: when the target group cannot be found
1265

1266
    """
1267
    if target is None:
1268
      if len(self._config_data.nodegroups) != 1:
1269
        raise errors.OpPrereqError("More than one node group exists. Target"
1270
                                   " group must be specified explicitly.")
1271
      else:
1272
        return self._config_data.nodegroups.keys()[0]
1273
    if target in self._config_data.nodegroups:
1274
      return target
1275
    for nodegroup in self._config_data.nodegroups.values():
1276
      if nodegroup.name == target:
1277
        return nodegroup.uuid
1278
    raise errors.OpPrereqError("Node group '%s' not found" % target,
1279
                               errors.ECODE_NOENT)
1280

    
1281
  @locking.ssynchronized(_config_lock, shared=1)
1282
  def LookupNodeGroup(self, target):
1283
    """Lookup a node group's UUID.
1284

1285
    This function is just a wrapper over L{_UnlockedLookupNodeGroup}.
1286

1287
    @type target: string or None
1288
    @param target: group name or UUID or None to look for the default
1289
    @rtype: string
1290
    @return: nodegroup UUID
1291

1292
    """
1293
    return self._UnlockedLookupNodeGroup(target)
1294

    
1295
  def _UnlockedGetNodeGroup(self, uuid):
1296
    """Lookup a node group.
1297

1298
    @type uuid: string
1299
    @param uuid: group UUID
1300
    @rtype: L{objects.NodeGroup} or None
1301
    @return: nodegroup object, or None if not found
1302

1303
    """
1304
    if uuid not in self._config_data.nodegroups:
1305
      return None
1306

    
1307
    return self._config_data.nodegroups[uuid]
1308

    
1309
  @locking.ssynchronized(_config_lock, shared=1)
1310
  def GetNodeGroup(self, uuid):
1311
    """Lookup a node group.
1312

1313
    @type uuid: string
1314
    @param uuid: group UUID
1315
    @rtype: L{objects.NodeGroup} or None
1316
    @return: nodegroup object, or None if not found
1317

1318
    """
1319
    return self._UnlockedGetNodeGroup(uuid)
1320

    
1321
  @locking.ssynchronized(_config_lock, shared=1)
1322
  def GetAllNodeGroupsInfo(self):
1323
    """Get the configuration of all node groups.
1324

1325
    """
1326
    return dict(self._config_data.nodegroups)
1327

    
1328
  @locking.ssynchronized(_config_lock, shared=1)
1329
  def GetNodeGroupList(self):
1330
    """Get a list of node groups.
1331

1332
    """
1333
    return self._config_data.nodegroups.keys()
1334

    
1335
  @locking.ssynchronized(_config_lock, shared=1)
1336
  def GetNodeGroupMembersByNodes(self, nodes):
1337
    """Get nodes which are member in the same nodegroups as the given nodes.
1338

1339
    """
1340
    ngfn = lambda node_name: self._UnlockedGetNodeInfo(node_name).group
1341
    return frozenset(member_name
1342
                     for node_name in nodes
1343
                     for member_name in
1344
                       self._UnlockedGetNodeGroup(ngfn(node_name)).members)
1345

    
1346
  @locking.ssynchronized(_config_lock, shared=1)
1347
  def GetMultiNodeGroupInfo(self, group_uuids):
1348
    """Get the configuration of multiple node groups.
1349

1350
    @param group_uuids: List of node group UUIDs
1351
    @rtype: list
1352
    @return: List of tuples of (group_uuid, group_info)
1353

1354
    """
1355
    return [(uuid, self._UnlockedGetNodeGroup(uuid)) for uuid in group_uuids]
1356

    
1357
  @locking.ssynchronized(_config_lock)
1358
  def AddInstance(self, instance, ec_id):
1359
    """Add an instance to the config.
1360

1361
    This should be used after creating a new instance.
1362

1363
    @type instance: L{objects.Instance}
1364
    @param instance: the instance object
1365

1366
    """
1367
    if not isinstance(instance, objects.Instance):
1368
      raise errors.ProgrammerError("Invalid type passed to AddInstance")
1369

    
1370
    if instance.disk_template != constants.DT_DISKLESS:
1371
      all_lvs = instance.MapLVsByNode()
1372
      logging.info("Instance '%s' DISK_LAYOUT: %s", instance.name, all_lvs)
1373

    
1374
    all_macs = self._AllMACs()
1375
    for nic in instance.nics:
1376
      if nic.mac in all_macs:
1377
        raise errors.ConfigurationError("Cannot add instance %s:"
1378
                                        " MAC address '%s' already in use." %
1379
                                        (instance.name, nic.mac))
1380

    
1381
    self._EnsureUUID(instance, ec_id)
1382

    
1383
    instance.serial_no = 1
1384
    instance.ctime = instance.mtime = time.time()
1385
    self._config_data.instances[instance.name] = instance
1386
    self._config_data.cluster.serial_no += 1
1387
    self._UnlockedReleaseDRBDMinors(instance.name)
1388
    self._UnlockedCommitTemporaryIps(ec_id)
1389
    self._WriteConfig()
1390

    
1391
  def _EnsureUUID(self, item, ec_id):
1392
    """Ensures a given object has a valid UUID.
1393

1394
    @param item: the instance or node to be checked
1395
    @param ec_id: the execution context id for the uuid reservation
1396

1397
    """
1398
    if not item.uuid:
1399
      item.uuid = self._GenerateUniqueID(ec_id)
1400
    elif item.uuid in self._AllIDs(include_temporary=True):
1401
      raise errors.ConfigurationError("Cannot add '%s': UUID %s already"
1402
                                      " in use" % (item.name, item.uuid))
1403

    
1404
  def _SetInstanceStatus(self, instance_name, status):
1405
    """Set the instance's status to a given value.
1406

1407
    """
1408
    assert status in constants.ADMINST_ALL, \
1409
           "Invalid status '%s' passed to SetInstanceStatus" % (status,)
1410

    
1411
    if instance_name not in self._config_data.instances:
1412
      raise errors.ConfigurationError("Unknown instance '%s'" %
1413
                                      instance_name)
1414
    instance = self._config_data.instances[instance_name]
1415
    if instance.admin_state != status:
1416
      instance.admin_state = status
1417
      instance.serial_no += 1
1418
      instance.mtime = time.time()
1419
      self._WriteConfig()
1420

    
1421
  @locking.ssynchronized(_config_lock)
1422
  def MarkInstanceUp(self, instance_name):
1423
    """Mark the instance status to up in the config.
1424

1425
    """
1426
    self._SetInstanceStatus(instance_name, constants.ADMINST_UP)
1427

    
1428
  @locking.ssynchronized(_config_lock)
1429
  def MarkInstanceOffline(self, instance_name):
1430
    """Mark the instance status to down in the config.
1431

1432
    """
1433
    self._SetInstanceStatus(instance_name, constants.ADMINST_OFFLINE)
1434

    
1435
  @locking.ssynchronized(_config_lock)
1436
  def RemoveInstance(self, instance_name):
1437
    """Remove the instance from the configuration.
1438

1439
    """
1440
    if instance_name not in self._config_data.instances:
1441
      raise errors.ConfigurationError("Unknown instance '%s'" % instance_name)
1442

    
1443
    # If a network port has been allocated to the instance,
1444
    # return it to the pool of free ports.
1445
    inst = self._config_data.instances[instance_name]
1446
    network_port = getattr(inst, "network_port", None)
1447
    if network_port is not None:
1448
      self._config_data.cluster.tcpudp_port_pool.add(network_port)
1449

    
1450
    instance = self._UnlockedGetInstanceInfo(instance_name)
1451

    
1452
    for nic in instance.nics:
1453
      if nic.network is not None and nic.ip is not None:
1454
        net_uuid = self._UnlockedLookupNetwork(nic.network)
1455
        if net_uuid:
1456
          # Return all IP addresses to the respective address pools
1457
          self._UnlockedCommitIp(constants.RELEASE_ACTION, net_uuid, nic.ip)
1458

    
1459
    del self._config_data.instances[instance_name]
1460
    self._config_data.cluster.serial_no += 1
1461
    self._WriteConfig()
1462

    
1463
  @locking.ssynchronized(_config_lock)
1464
  def RenameInstance(self, old_name, new_name):
1465
    """Rename an instance.
1466

1467
    This needs to be done in ConfigWriter and not by RemoveInstance
1468
    combined with AddInstance as only we can guarantee an atomic
1469
    rename.
1470

1471
    """
1472
    if old_name not in self._config_data.instances:
1473
      raise errors.ConfigurationError("Unknown instance '%s'" % old_name)
1474

    
1475
    # Operate on a copy to not loose instance object in case of a failure
1476
    inst = self._config_data.instances[old_name].Copy()
1477
    inst.name = new_name
1478

    
1479
    for (idx, disk) in enumerate(inst.disks):
1480
      if disk.dev_type == constants.LD_FILE:
1481
        # rename the file paths in logical and physical id
1482
        file_storage_dir = os.path.dirname(os.path.dirname(disk.logical_id[1]))
1483
        disk.logical_id = (disk.logical_id[0],
1484
                           utils.PathJoin(file_storage_dir, inst.name,
1485
                                          "disk%s" % idx))
1486
        disk.physical_id = disk.logical_id
1487

    
1488
    # Actually replace instance object
1489
    del self._config_data.instances[old_name]
1490
    self._config_data.instances[inst.name] = inst
1491

    
1492
    # Force update of ssconf files
1493
    self._config_data.cluster.serial_no += 1
1494

    
1495
    self._WriteConfig()
1496

    
1497
  @locking.ssynchronized(_config_lock)
1498
  def MarkInstanceDown(self, instance_name):
1499
    """Mark the status of an instance to down in the configuration.
1500

1501
    """
1502
    self._SetInstanceStatus(instance_name, constants.ADMINST_DOWN)
1503

    
1504
  def _UnlockedGetInstanceList(self):
1505
    """Get the list of instances.
1506

1507
    This function is for internal use, when the config lock is already held.
1508

1509
    """
1510
    return self._config_data.instances.keys()
1511

    
1512
  @locking.ssynchronized(_config_lock, shared=1)
1513
  def GetInstanceList(self):
1514
    """Get the list of instances.
1515

1516
    @return: array of instances, ex. ['instance2.example.com',
1517
        'instance1.example.com']
1518

1519
    """
1520
    return self._UnlockedGetInstanceList()
1521

    
1522
  def ExpandInstanceName(self, short_name):
1523
    """Attempt to expand an incomplete instance name.
1524

1525
    """
1526
    # Locking is done in L{ConfigWriter.GetInstanceList}
1527
    return _MatchNameComponentIgnoreCase(short_name, self.GetInstanceList())
1528

    
1529
  def _UnlockedGetInstanceInfo(self, instance_name):
1530
    """Returns information about an instance.
1531

1532
    This function is for internal use, when the config lock is already held.
1533

1534
    """
1535
    if instance_name not in self._config_data.instances:
1536
      return None
1537

    
1538
    return self._config_data.instances[instance_name]
1539

    
1540
  @locking.ssynchronized(_config_lock, shared=1)
1541
  def GetInstanceInfo(self, instance_name):
1542
    """Returns information about an instance.
1543

1544
    It takes the information from the configuration file. Other information of
1545
    an instance are taken from the live systems.
1546

1547
    @param instance_name: name of the instance, e.g.
1548
        I{instance1.example.com}
1549

1550
    @rtype: L{objects.Instance}
1551
    @return: the instance object
1552

1553
    """
1554
    return self._UnlockedGetInstanceInfo(instance_name)
1555

    
1556
  @locking.ssynchronized(_config_lock, shared=1)
1557
  def GetInstanceNodeGroups(self, instance_name, primary_only=False):
1558
    """Returns set of node group UUIDs for instance's nodes.
1559

1560
    @rtype: frozenset
1561

1562
    """
1563
    instance = self._UnlockedGetInstanceInfo(instance_name)
1564
    if not instance:
1565
      raise errors.ConfigurationError("Unknown instance '%s'" % instance_name)
1566

    
1567
    if primary_only:
1568
      nodes = [instance.primary_node]
1569
    else:
1570
      nodes = instance.all_nodes
1571

    
1572
    return frozenset(self._UnlockedGetNodeInfo(node_name).group
1573
                     for node_name in nodes)
1574

    
1575
  @locking.ssynchronized(_config_lock, shared=1)
1576
  def GetMultiInstanceInfo(self, instances):
1577
    """Get the configuration of multiple instances.
1578

1579
    @param instances: list of instance names
1580
    @rtype: list
1581
    @return: list of tuples (instance, instance_info), where
1582
        instance_info is what would GetInstanceInfo return for the
1583
        node, while keeping the original order
1584

1585
    """
1586
    return [(name, self._UnlockedGetInstanceInfo(name)) for name in instances]
1587

    
1588
  @locking.ssynchronized(_config_lock, shared=1)
1589
  def GetAllInstancesInfo(self):
1590
    """Get the configuration of all instances.
1591

1592
    @rtype: dict
1593
    @return: dict of (instance, instance_info), where instance_info is what
1594
              would GetInstanceInfo return for the node
1595

1596
    """
1597
    my_dict = dict([(instance, self._UnlockedGetInstanceInfo(instance))
1598
                    for instance in self._UnlockedGetInstanceList()])
1599
    return my_dict
1600

    
1601
  @locking.ssynchronized(_config_lock, shared=1)
1602
  def GetInstancesInfoByFilter(self, filter_fn):
1603
    """Get instance configuration with a filter.
1604

1605
    @type filter_fn: callable
1606
    @param filter_fn: Filter function receiving instance object as parameter,
1607
      returning boolean. Important: this function is called while the
1608
      configuration locks is held. It must not do any complex work or call
1609
      functions potentially leading to a deadlock. Ideally it doesn't call any
1610
      other functions and just compares instance attributes.
1611

1612
    """
1613
    return dict((name, inst)
1614
                for (name, inst) in self._config_data.instances.items()
1615
                if filter_fn(inst))
1616

    
1617
  @locking.ssynchronized(_config_lock)
1618
  def AddNode(self, node, ec_id):
1619
    """Add a node to the configuration.
1620

1621
    @type node: L{objects.Node}
1622
    @param node: a Node instance
1623

1624
    """
1625
    logging.info("Adding node %s to configuration", node.name)
1626

    
1627
    self._EnsureUUID(node, ec_id)
1628

    
1629
    node.serial_no = 1
1630
    node.ctime = node.mtime = time.time()
1631
    self._UnlockedAddNodeToGroup(node.name, node.group)
1632
    self._config_data.nodes[node.name] = node
1633
    self._config_data.cluster.serial_no += 1
1634
    self._WriteConfig()
1635

    
1636
  @locking.ssynchronized(_config_lock)
1637
  def RemoveNode(self, node_name):
1638
    """Remove a node from the configuration.
1639

1640
    """
1641
    logging.info("Removing node %s from configuration", node_name)
1642

    
1643
    if node_name not in self._config_data.nodes:
1644
      raise errors.ConfigurationError("Unknown node '%s'" % node_name)
1645

    
1646
    self._UnlockedRemoveNodeFromGroup(self._config_data.nodes[node_name])
1647
    del self._config_data.nodes[node_name]
1648
    self._config_data.cluster.serial_no += 1
1649
    self._WriteConfig()
1650

    
1651
  def ExpandNodeName(self, short_name):
1652
    """Attempt to expand an incomplete node name.
1653

1654
    """
1655
    # Locking is done in L{ConfigWriter.GetNodeList}
1656
    return _MatchNameComponentIgnoreCase(short_name, self.GetNodeList())
1657

    
1658
  def _UnlockedGetNodeInfo(self, node_name):
1659
    """Get the configuration of a node, as stored in the config.
1660

1661
    This function is for internal use, when the config lock is already
1662
    held.
1663

1664
    @param node_name: the node name, e.g. I{node1.example.com}
1665

1666
    @rtype: L{objects.Node}
1667
    @return: the node object
1668

1669
    """
1670
    if node_name not in self._config_data.nodes:
1671
      return None
1672

    
1673
    return self._config_data.nodes[node_name]
1674

    
1675
  @locking.ssynchronized(_config_lock, shared=1)
1676
  def GetNodeInfo(self, node_name):
1677
    """Get the configuration of a node, as stored in the config.
1678

1679
    This is just a locked wrapper over L{_UnlockedGetNodeInfo}.
1680

1681
    @param node_name: the node name, e.g. I{node1.example.com}
1682

1683
    @rtype: L{objects.Node}
1684
    @return: the node object
1685

1686
    """
1687
    return self._UnlockedGetNodeInfo(node_name)
1688

    
1689
  @locking.ssynchronized(_config_lock, shared=1)
1690
  def GetNodeInstances(self, node_name):
1691
    """Get the instances of a node, as stored in the config.
1692

1693
    @param node_name: the node name, e.g. I{node1.example.com}
1694

1695
    @rtype: (list, list)
1696
    @return: a tuple with two lists: the primary and the secondary instances
1697

1698
    """
1699
    pri = []
1700
    sec = []
1701
    for inst in self._config_data.instances.values():
1702
      if inst.primary_node == node_name:
1703
        pri.append(inst.name)
1704
      if node_name in inst.secondary_nodes:
1705
        sec.append(inst.name)
1706
    return (pri, sec)
1707

    
1708
  @locking.ssynchronized(_config_lock, shared=1)
1709
  def GetNodeGroupInstances(self, uuid, primary_only=False):
1710
    """Get the instances of a node group.
1711

1712
    @param uuid: Node group UUID
1713
    @param primary_only: Whether to only consider primary nodes
1714
    @rtype: frozenset
1715
    @return: List of instance names in node group
1716

1717
    """
1718
    if primary_only:
1719
      nodes_fn = lambda inst: [inst.primary_node]
1720
    else:
1721
      nodes_fn = lambda inst: inst.all_nodes
1722

    
1723
    return frozenset(inst.name
1724
                     for inst in self._config_data.instances.values()
1725
                     for node_name in nodes_fn(inst)
1726
                     if self._UnlockedGetNodeInfo(node_name).group == uuid)
1727

    
1728
  def _UnlockedGetNodeList(self):
1729
    """Return the list of nodes which are in the configuration.
1730

1731
    This function is for internal use, when the config lock is already
1732
    held.
1733

1734
    @rtype: list
1735

1736
    """
1737
    return self._config_data.nodes.keys()
1738

    
1739
  @locking.ssynchronized(_config_lock, shared=1)
1740
  def GetNodeList(self):
1741
    """Return the list of nodes which are in the configuration.
1742

1743
    """
1744
    return self._UnlockedGetNodeList()
1745

    
1746
  def _UnlockedGetOnlineNodeList(self):
1747
    """Return the list of nodes which are online.
1748

1749
    """
1750
    all_nodes = [self._UnlockedGetNodeInfo(node)
1751
                 for node in self._UnlockedGetNodeList()]
1752
    return [node.name for node in all_nodes if not node.offline]
1753

    
1754
  @locking.ssynchronized(_config_lock, shared=1)
1755
  def GetOnlineNodeList(self):
1756
    """Return the list of nodes which are online.
1757

1758
    """
1759
    return self._UnlockedGetOnlineNodeList()
1760

    
1761
  @locking.ssynchronized(_config_lock, shared=1)
1762
  def GetVmCapableNodeList(self):
1763
    """Return the list of nodes which are not vm capable.
1764

1765
    """
1766
    all_nodes = [self._UnlockedGetNodeInfo(node)
1767
                 for node in self._UnlockedGetNodeList()]
1768
    return [node.name for node in all_nodes if node.vm_capable]
1769

    
1770
  @locking.ssynchronized(_config_lock, shared=1)
1771
  def GetNonVmCapableNodeList(self):
1772
    """Return the list of nodes which are not vm capable.
1773

1774
    """
1775
    all_nodes = [self._UnlockedGetNodeInfo(node)
1776
                 for node in self._UnlockedGetNodeList()]
1777
    return [node.name for node in all_nodes if not node.vm_capable]
1778

    
1779
  @locking.ssynchronized(_config_lock, shared=1)
1780
  def GetMultiNodeInfo(self, nodes):
1781
    """Get the configuration of multiple nodes.
1782

1783
    @param nodes: list of node names
1784
    @rtype: list
1785
    @return: list of tuples of (node, node_info), where node_info is
1786
        what would GetNodeInfo return for the node, in the original
1787
        order
1788

1789
    """
1790
    return [(name, self._UnlockedGetNodeInfo(name)) for name in nodes]
1791

    
1792
  @locking.ssynchronized(_config_lock, shared=1)
1793
  def GetAllNodesInfo(self):
1794
    """Get the configuration of all nodes.
1795

1796
    @rtype: dict
1797
    @return: dict of (node, node_info), where node_info is what
1798
              would GetNodeInfo return for the node
1799

1800
    """
1801
    return self._UnlockedGetAllNodesInfo()
1802

    
1803
  def _UnlockedGetAllNodesInfo(self):
1804
    """Gets configuration of all nodes.
1805

1806
    @note: See L{GetAllNodesInfo}
1807

1808
    """
1809
    return dict([(node, self._UnlockedGetNodeInfo(node))
1810
                 for node in self._UnlockedGetNodeList()])
1811

    
1812
  @locking.ssynchronized(_config_lock, shared=1)
1813
  def GetNodeGroupsFromNodes(self, nodes):
1814
    """Returns groups for a list of nodes.
1815

1816
    @type nodes: list of string
1817
    @param nodes: List of node names
1818
    @rtype: frozenset
1819

1820
    """
1821
    return frozenset(self._UnlockedGetNodeInfo(name).group for name in nodes)
1822

    
1823
  def _UnlockedGetMasterCandidateStats(self, exceptions=None):
1824
    """Get the number of current and maximum desired and possible candidates.
1825

1826
    @type exceptions: list
1827
    @param exceptions: if passed, list of nodes that should be ignored
1828
    @rtype: tuple
1829
    @return: tuple of (current, desired and possible, possible)
1830

1831
    """
1832
    mc_now = mc_should = mc_max = 0
1833
    for node in self._config_data.nodes.values():
1834
      if exceptions and node.name in exceptions:
1835
        continue
1836
      if not (node.offline or node.drained) and node.master_capable:
1837
        mc_max += 1
1838
      if node.master_candidate:
1839
        mc_now += 1
1840
    mc_should = min(mc_max, self._config_data.cluster.candidate_pool_size)
1841
    return (mc_now, mc_should, mc_max)
1842

    
1843
  @locking.ssynchronized(_config_lock, shared=1)
1844
  def GetMasterCandidateStats(self, exceptions=None):
1845
    """Get the number of current and maximum possible candidates.
1846

1847
    This is just a wrapper over L{_UnlockedGetMasterCandidateStats}.
1848

1849
    @type exceptions: list
1850
    @param exceptions: if passed, list of nodes that should be ignored
1851
    @rtype: tuple
1852
    @return: tuple of (current, max)
1853

1854
    """
1855
    return self._UnlockedGetMasterCandidateStats(exceptions)
1856

    
1857
  @locking.ssynchronized(_config_lock)
1858
  def MaintainCandidatePool(self, exceptions):
1859
    """Try to grow the candidate pool to the desired size.
1860

1861
    @type exceptions: list
1862
    @param exceptions: if passed, list of nodes that should be ignored
1863
    @rtype: list
1864
    @return: list with the adjusted nodes (L{objects.Node} instances)
1865

1866
    """
1867
    mc_now, mc_max, _ = self._UnlockedGetMasterCandidateStats(exceptions)
1868
    mod_list = []
1869
    if mc_now < mc_max:
1870
      node_list = self._config_data.nodes.keys()
1871
      random.shuffle(node_list)
1872
      for name in node_list:
1873
        if mc_now >= mc_max:
1874
          break
1875
        node = self._config_data.nodes[name]
1876
        if (node.master_candidate or node.offline or node.drained or
1877
            node.name in exceptions or not node.master_capable):
1878
          continue
1879
        mod_list.append(node)
1880
        node.master_candidate = True
1881
        node.serial_no += 1
1882
        mc_now += 1
1883
      if mc_now != mc_max:
1884
        # this should not happen
1885
        logging.warning("Warning: MaintainCandidatePool didn't manage to"
1886
                        " fill the candidate pool (%d/%d)", mc_now, mc_max)
1887
      if mod_list:
1888
        self._config_data.cluster.serial_no += 1
1889
        self._WriteConfig()
1890

    
1891
    return mod_list
1892

    
1893
  def _UnlockedAddNodeToGroup(self, node_name, nodegroup_uuid):
1894
    """Add a given node to the specified group.
1895

1896
    """
1897
    if nodegroup_uuid not in self._config_data.nodegroups:
1898
      # This can happen if a node group gets deleted between its lookup and
1899
      # when we're adding the first node to it, since we don't keep a lock in
1900
      # the meantime. It's ok though, as we'll fail cleanly if the node group
1901
      # is not found anymore.
1902
      raise errors.OpExecError("Unknown node group: %s" % nodegroup_uuid)
1903
    if node_name not in self._config_data.nodegroups[nodegroup_uuid].members:
1904
      self._config_data.nodegroups[nodegroup_uuid].members.append(node_name)
1905

    
1906
  def _UnlockedRemoveNodeFromGroup(self, node):
1907
    """Remove a given node from its group.
1908

1909
    """
1910
    nodegroup = node.group
1911
    if nodegroup not in self._config_data.nodegroups:
1912
      logging.warning("Warning: node '%s' has unknown node group '%s'"
1913
                      " (while being removed from it)", node.name, nodegroup)
1914
    nodegroup_obj = self._config_data.nodegroups[nodegroup]
1915
    if node.name not in nodegroup_obj.members:
1916
      logging.warning("Warning: node '%s' not a member of its node group '%s'"
1917
                      " (while being removed from it)", node.name, nodegroup)
1918
    else:
1919
      nodegroup_obj.members.remove(node.name)
1920

    
1921
  @locking.ssynchronized(_config_lock)
1922
  def AssignGroupNodes(self, mods):
1923
    """Changes the group of a number of nodes.
1924

1925
    @type mods: list of tuples; (node name, new group UUID)
1926
    @param mods: Node membership modifications
1927

1928
    """
1929
    groups = self._config_data.nodegroups
1930
    nodes = self._config_data.nodes
1931

    
1932
    resmod = []
1933

    
1934
    # Try to resolve names/UUIDs first
1935
    for (node_name, new_group_uuid) in mods:
1936
      try:
1937
        node = nodes[node_name]
1938
      except KeyError:
1939
        raise errors.ConfigurationError("Unable to find node '%s'" % node_name)
1940

    
1941
      if node.group == new_group_uuid:
1942
        # Node is being assigned to its current group
1943
        logging.debug("Node '%s' was assigned to its current group (%s)",
1944
                      node_name, node.group)
1945
        continue
1946

    
1947
      # Try to find current group of node
1948
      try:
1949
        old_group = groups[node.group]
1950
      except KeyError:
1951
        raise errors.ConfigurationError("Unable to find old group '%s'" %
1952
                                        node.group)
1953

    
1954
      # Try to find new group for node
1955
      try:
1956
        new_group = groups[new_group_uuid]
1957
      except KeyError:
1958
        raise errors.ConfigurationError("Unable to find new group '%s'" %
1959
                                        new_group_uuid)
1960

    
1961
      assert node.name in old_group.members, \
1962
        ("Inconsistent configuration: node '%s' not listed in members for its"
1963
         " old group '%s'" % (node.name, old_group.uuid))
1964
      assert node.name not in new_group.members, \
1965
        ("Inconsistent configuration: node '%s' already listed in members for"
1966
         " its new group '%s'" % (node.name, new_group.uuid))
1967

    
1968
      resmod.append((node, old_group, new_group))
1969

    
1970
    # Apply changes
1971
    for (node, old_group, new_group) in resmod:
1972
      assert node.uuid != new_group.uuid and old_group.uuid != new_group.uuid, \
1973
        "Assigning to current group is not possible"
1974

    
1975
      node.group = new_group.uuid
1976

    
1977
      # Update members of involved groups
1978
      if node.name in old_group.members:
1979
        old_group.members.remove(node.name)
1980
      if node.name not in new_group.members:
1981
        new_group.members.append(node.name)
1982

    
1983
    # Update timestamps and serials (only once per node/group object)
1984
    now = time.time()
1985
    for obj in frozenset(itertools.chain(*resmod)): # pylint: disable=W0142
1986
      obj.serial_no += 1
1987
      obj.mtime = now
1988

    
1989
    # Force ssconf update
1990
    self._config_data.cluster.serial_no += 1
1991

    
1992
    self._WriteConfig()
1993

    
1994
  def _BumpSerialNo(self):
1995
    """Bump up the serial number of the config.
1996

1997
    """
1998
    self._config_data.serial_no += 1
1999
    self._config_data.mtime = time.time()
2000

    
2001
  def _AllUUIDObjects(self):
2002
    """Returns all objects with uuid attributes.
2003

2004
    """
2005
    return (self._config_data.instances.values() +
2006
            self._config_data.nodes.values() +
2007
            self._config_data.nodegroups.values() +
2008
            [self._config_data.cluster])
2009

    
2010
  def _OpenConfig(self, accept_foreign):
2011
    """Read the config data from disk.
2012

2013
    """
2014
    raw_data = utils.ReadFile(self._cfg_file)
2015

    
2016
    try:
2017
      data = objects.ConfigData.FromDict(serializer.Load(raw_data))
2018
    except Exception, err:
2019
      raise errors.ConfigurationError(err)
2020

    
2021
    # Make sure the configuration has the right version
2022
    _ValidateConfig(data)
2023

    
2024
    if (not hasattr(data, "cluster") or
2025
        not hasattr(data.cluster, "rsahostkeypub")):
2026
      raise errors.ConfigurationError("Incomplete configuration"
2027
                                      " (missing cluster.rsahostkeypub)")
2028

    
2029
    if data.cluster.master_node != self._my_hostname and not accept_foreign:
2030
      msg = ("The configuration denotes node %s as master, while my"
2031
             " hostname is %s; opening a foreign configuration is only"
2032
             " possible in accept_foreign mode" %
2033
             (data.cluster.master_node, self._my_hostname))
2034
      raise errors.ConfigurationError(msg)
2035

    
2036
    # Upgrade configuration if needed
2037
    data.UpgradeConfig()
2038

    
2039
    self._config_data = data
2040
    # reset the last serial as -1 so that the next write will cause
2041
    # ssconf update
2042
    self._last_cluster_serial = -1
2043

    
2044
    # And finally run our (custom) config upgrade sequence
2045
    self._UpgradeConfig()
2046

    
2047
    self._cfg_id = utils.GetFileID(path=self._cfg_file)
2048

    
2049
  def _UpgradeConfig(self):
2050
    """Run upgrade steps that cannot be done purely in the objects.
2051

2052
    This is because some data elements need uniqueness across the
2053
    whole configuration, etc.
2054

2055
    @warning: this function will call L{_WriteConfig()}, but also
2056
        L{DropECReservations} so it needs to be called only from a
2057
        "safe" place (the constructor). If one wanted to call it with
2058
        the lock held, a DropECReservationUnlocked would need to be
2059
        created first, to avoid causing deadlock.
2060

2061
    """
2062
    modified = False
2063
    for item in self._AllUUIDObjects():
2064
      if item.uuid is None:
2065
        item.uuid = self._GenerateUniqueID(_UPGRADE_CONFIG_JID)
2066
        modified = True
2067
    if not self._config_data.nodegroups:
2068
      default_nodegroup_name = constants.INITIAL_NODE_GROUP_NAME
2069
      default_nodegroup = objects.NodeGroup(name=default_nodegroup_name,
2070
                                            members=[])
2071
      self._UnlockedAddNodeGroup(default_nodegroup, _UPGRADE_CONFIG_JID, True)
2072
      modified = True
2073
    for node in self._config_data.nodes.values():
2074
      if not node.group:
2075
        node.group = self.LookupNodeGroup(None)
2076
        modified = True
2077
      # This is technically *not* an upgrade, but needs to be done both when
2078
      # nodegroups are being added, and upon normally loading the config,
2079
      # because the members list of a node group is discarded upon
2080
      # serializing/deserializing the object.
2081
      self._UnlockedAddNodeToGroup(node.name, node.group)
2082
    if modified:
2083
      self._WriteConfig()
2084
      # This is ok even if it acquires the internal lock, as _UpgradeConfig is
2085
      # only called at config init time, without the lock held
2086
      self.DropECReservations(_UPGRADE_CONFIG_JID)
2087

    
2088
  def _DistributeConfig(self, feedback_fn):
2089
    """Distribute the configuration to the other nodes.
2090

2091
    Currently, this only copies the configuration file. In the future,
2092
    it could be used to encapsulate the 2/3-phase update mechanism.
2093

2094
    """
2095
    if self._offline:
2096
      return True
2097

    
2098
    bad = False
2099

    
2100
    node_list = []
2101
    addr_list = []
2102
    myhostname = self._my_hostname
2103
    # we can skip checking whether _UnlockedGetNodeInfo returns None
2104
    # since the node list comes from _UnlocketGetNodeList, and we are
2105
    # called with the lock held, so no modifications should take place
2106
    # in between
2107
    for node_name in self._UnlockedGetNodeList():
2108
      if node_name == myhostname:
2109
        continue
2110
      node_info = self._UnlockedGetNodeInfo(node_name)
2111
      if not node_info.master_candidate:
2112
        continue
2113
      node_list.append(node_info.name)
2114
      addr_list.append(node_info.primary_ip)
2115

    
2116
    # TODO: Use dedicated resolver talking to config writer for name resolution
2117
    result = \
2118
      self._GetRpc(addr_list).call_upload_file(node_list, self._cfg_file)
2119
    for to_node, to_result in result.items():
2120
      msg = to_result.fail_msg
2121
      if msg:
2122
        msg = ("Copy of file %s to node %s failed: %s" %
2123
               (self._cfg_file, to_node, msg))
2124
        logging.error(msg)
2125

    
2126
        if feedback_fn:
2127
          feedback_fn(msg)
2128

    
2129
        bad = True
2130

    
2131
    return not bad
2132

    
2133
  def _WriteConfig(self, destination=None, feedback_fn=None):
2134
    """Write the configuration data to persistent storage.
2135

2136
    """
2137
    assert feedback_fn is None or callable(feedback_fn)
2138

    
2139
    # Warn on config errors, but don't abort the save - the
2140
    # configuration has already been modified, and we can't revert;
2141
    # the best we can do is to warn the user and save as is, leaving
2142
    # recovery to the user
2143
    config_errors = self._UnlockedVerifyConfig()
2144
    if config_errors:
2145
      errmsg = ("Configuration data is not consistent: %s" %
2146
                (utils.CommaJoin(config_errors)))
2147
      logging.critical(errmsg)
2148
      if feedback_fn:
2149
        feedback_fn(errmsg)
2150

    
2151
    if destination is None:
2152
      destination = self._cfg_file
2153
    self._BumpSerialNo()
2154
    txt = serializer.Dump(self._config_data.ToDict())
2155

    
2156
    getents = self._getents()
2157
    try:
2158
      fd = utils.SafeWriteFile(destination, self._cfg_id, data=txt,
2159
                               close=False, gid=getents.confd_gid, mode=0640)
2160
    except errors.LockError:
2161
      raise errors.ConfigurationError("The configuration file has been"
2162
                                      " modified since the last write, cannot"
2163
                                      " update")
2164
    try:
2165
      self._cfg_id = utils.GetFileID(fd=fd)
2166
    finally:
2167
      os.close(fd)
2168

    
2169
    self.write_count += 1
2170

    
2171
    # and redistribute the config file to master candidates
2172
    self._DistributeConfig(feedback_fn)
2173

    
2174
    # Write ssconf files on all nodes (including locally)
2175
    if self._last_cluster_serial < self._config_data.cluster.serial_no:
2176
      if not self._offline:
2177
        result = self._GetRpc(None).call_write_ssconf_files(
2178
          self._UnlockedGetOnlineNodeList(),
2179
          self._UnlockedGetSsconfValues())
2180

    
2181
        for nname, nresu in result.items():
2182
          msg = nresu.fail_msg
2183
          if msg:
2184
            errmsg = ("Error while uploading ssconf files to"
2185
                      " node %s: %s" % (nname, msg))
2186
            logging.warning(errmsg)
2187

    
2188
            if feedback_fn:
2189
              feedback_fn(errmsg)
2190

    
2191
      self._last_cluster_serial = self._config_data.cluster.serial_no
2192

    
2193
  def _UnlockedGetSsconfValues(self):
2194
    """Return the values needed by ssconf.
2195

2196
    @rtype: dict
2197
    @return: a dictionary with keys the ssconf names and values their
2198
        associated value
2199

2200
    """
2201
    fn = "\n".join
2202
    instance_names = utils.NiceSort(self._UnlockedGetInstanceList())
2203
    node_names = utils.NiceSort(self._UnlockedGetNodeList())
2204
    node_info = [self._UnlockedGetNodeInfo(name) for name in node_names]
2205
    node_pri_ips = ["%s %s" % (ninfo.name, ninfo.primary_ip)
2206
                    for ninfo in node_info]
2207
    node_snd_ips = ["%s %s" % (ninfo.name, ninfo.secondary_ip)
2208
                    for ninfo in node_info]
2209

    
2210
    instance_data = fn(instance_names)
2211
    off_data = fn(node.name for node in node_info if node.offline)
2212
    on_data = fn(node.name for node in node_info if not node.offline)
2213
    mc_data = fn(node.name for node in node_info if node.master_candidate)
2214
    mc_ips_data = fn(node.primary_ip for node in node_info
2215
                     if node.master_candidate)
2216
    node_data = fn(node_names)
2217
    node_pri_ips_data = fn(node_pri_ips)
2218
    node_snd_ips_data = fn(node_snd_ips)
2219

    
2220
    cluster = self._config_data.cluster
2221
    cluster_tags = fn(cluster.GetTags())
2222

    
2223
    hypervisor_list = fn(cluster.enabled_hypervisors)
2224

    
2225
    uid_pool = uidpool.FormatUidPool(cluster.uid_pool, separator="\n")
2226

    
2227
    nodegroups = ["%s %s" % (nodegroup.uuid, nodegroup.name) for nodegroup in
2228
                  self._config_data.nodegroups.values()]
2229
    nodegroups_data = fn(utils.NiceSort(nodegroups))
2230
    networks = ["%s %s" % (net.uuid, net.name) for net in
2231
                self._config_data.networks.values()]
2232
    networks_data = fn(utils.NiceSort(networks))
2233

    
2234
    ssconf_values = {
2235
      constants.SS_CLUSTER_NAME: cluster.cluster_name,
2236
      constants.SS_CLUSTER_TAGS: cluster_tags,
2237
      constants.SS_FILE_STORAGE_DIR: cluster.file_storage_dir,
2238
      constants.SS_SHARED_FILE_STORAGE_DIR: cluster.shared_file_storage_dir,
2239
      constants.SS_MASTER_CANDIDATES: mc_data,
2240
      constants.SS_MASTER_CANDIDATES_IPS: mc_ips_data,
2241
      constants.SS_MASTER_IP: cluster.master_ip,
2242
      constants.SS_MASTER_NETDEV: cluster.master_netdev,
2243
      constants.SS_MASTER_NETMASK: str(cluster.master_netmask),
2244
      constants.SS_MASTER_NODE: cluster.master_node,
2245
      constants.SS_NODE_LIST: node_data,
2246
      constants.SS_NODE_PRIMARY_IPS: node_pri_ips_data,
2247
      constants.SS_NODE_SECONDARY_IPS: node_snd_ips_data,
2248
      constants.SS_OFFLINE_NODES: off_data,
2249
      constants.SS_ONLINE_NODES: on_data,
2250
      constants.SS_PRIMARY_IP_FAMILY: str(cluster.primary_ip_family),
2251
      constants.SS_INSTANCE_LIST: instance_data,
2252
      constants.SS_RELEASE_VERSION: constants.RELEASE_VERSION,
2253
      constants.SS_HYPERVISOR_LIST: hypervisor_list,
2254
      constants.SS_MAINTAIN_NODE_HEALTH: str(cluster.maintain_node_health),
2255
      constants.SS_UID_POOL: uid_pool,
2256
      constants.SS_NODEGROUPS: nodegroups_data,
2257
      constants.SS_NETWORKS: networks_data,
2258
      }
2259
    bad_values = [(k, v) for k, v in ssconf_values.items()
2260
                  if not isinstance(v, (str, basestring))]
2261
    if bad_values:
2262
      err = utils.CommaJoin("%s=%s" % (k, v) for k, v in bad_values)
2263
      raise errors.ConfigurationError("Some ssconf key(s) have non-string"
2264
                                      " values: %s" % err)
2265
    return ssconf_values
2266

    
2267
  @locking.ssynchronized(_config_lock, shared=1)
2268
  def GetSsconfValues(self):
2269
    """Wrapper using lock around _UnlockedGetSsconf().
2270

2271
    """
2272
    return self._UnlockedGetSsconfValues()
2273

    
2274
  @locking.ssynchronized(_config_lock, shared=1)
2275
  def GetVGName(self):
2276
    """Return the volume group name.
2277

2278
    """
2279
    return self._config_data.cluster.volume_group_name
2280

    
2281
  @locking.ssynchronized(_config_lock)
2282
  def SetVGName(self, vg_name):
2283
    """Set the volume group name.
2284

2285
    """
2286
    self._config_data.cluster.volume_group_name = vg_name
2287
    self._config_data.cluster.serial_no += 1
2288
    self._WriteConfig()
2289

    
2290
  @locking.ssynchronized(_config_lock, shared=1)
2291
  def GetDRBDHelper(self):
2292
    """Return DRBD usermode helper.
2293

2294
    """
2295
    return self._config_data.cluster.drbd_usermode_helper
2296

    
2297
  @locking.ssynchronized(_config_lock)
2298
  def SetDRBDHelper(self, drbd_helper):
2299
    """Set DRBD usermode helper.
2300

2301
    """
2302
    self._config_data.cluster.drbd_usermode_helper = drbd_helper
2303
    self._config_data.cluster.serial_no += 1
2304
    self._WriteConfig()
2305

    
2306
  @locking.ssynchronized(_config_lock, shared=1)
2307
  def GetMACPrefix(self):
2308
    """Return the mac prefix.
2309

2310
    """
2311
    return self._config_data.cluster.mac_prefix
2312

    
2313
  @locking.ssynchronized(_config_lock, shared=1)
2314
  def GetClusterInfo(self):
2315
    """Returns information about the cluster
2316

2317
    @rtype: L{objects.Cluster}
2318
    @return: the cluster object
2319

2320
    """
2321
    return self._config_data.cluster
2322

    
2323
  @locking.ssynchronized(_config_lock, shared=1)
2324
  def HasAnyDiskOfType(self, dev_type):
2325
    """Check if in there is at disk of the given type in the configuration.
2326

2327
    """
2328
    return self._config_data.HasAnyDiskOfType(dev_type)
2329

    
2330
  @locking.ssynchronized(_config_lock)
2331
  def Update(self, target, feedback_fn, ec_id=None):
2332
    """Notify function to be called after updates.
2333

2334
    This function must be called when an object (as returned by
2335
    GetInstanceInfo, GetNodeInfo, GetCluster) has been updated and the
2336
    caller wants the modifications saved to the backing store. Note
2337
    that all modified objects will be saved, but the target argument
2338
    is the one the caller wants to ensure that it's saved.
2339

2340
    @param target: an instance of either L{objects.Cluster},
2341
        L{objects.Node} or L{objects.Instance} which is existing in
2342
        the cluster
2343
    @param feedback_fn: Callable feedback function
2344

2345
    """
2346
    if self._config_data is None:
2347
      raise errors.ProgrammerError("Configuration file not read,"
2348
                                   " cannot save.")
2349
    update_serial = False
2350
    if isinstance(target, objects.Cluster):
2351
      test = target == self._config_data.cluster
2352
    elif isinstance(target, objects.Node):
2353
      test = target in self._config_data.nodes.values()
2354
      update_serial = True
2355
    elif isinstance(target, objects.Instance):
2356
      test = target in self._config_data.instances.values()
2357
    elif isinstance(target, objects.NodeGroup):
2358
      test = target in self._config_data.nodegroups.values()
2359
    elif isinstance(target, objects.Network):
2360
      test = target in self._config_data.networks.values()
2361
    else:
2362
      raise errors.ProgrammerError("Invalid object type (%s) passed to"
2363
                                   " ConfigWriter.Update" % type(target))
2364
    if not test:
2365
      raise errors.ConfigurationError("Configuration updated since object"
2366
                                      " has been read or unknown object")
2367
    target.serial_no += 1
2368
    target.mtime = now = time.time()
2369

    
2370
    if update_serial:
2371
      # for node updates, we need to increase the cluster serial too
2372
      self._config_data.cluster.serial_no += 1
2373
      self._config_data.cluster.mtime = now
2374

    
2375
    if isinstance(target, objects.Instance):
2376
      self._UnlockedReleaseDRBDMinors(target.name)
2377

    
2378
    if ec_id is not None:
2379
      # Commit all ips reserved by OpInstanceSetParams and OpGroupSetParams
2380
      self._UnlockedCommitTemporaryIps(ec_id)
2381

    
2382
    self._WriteConfig(feedback_fn=feedback_fn)
2383

    
2384
  @locking.ssynchronized(_config_lock)
2385
  def DropECReservations(self, ec_id):
2386
    """Drop per-execution-context reservations
2387

2388
    """
2389
    for rm in self._all_rms:
2390
      rm.DropECReservations(ec_id)
2391

    
2392
  @locking.ssynchronized(_config_lock, shared=1)
2393
  def GetAllNetworksInfo(self):
2394
    """Get the configuration of all networks
2395

2396
    """
2397
    return dict(self._config_data.networks)
2398

    
2399
  def _UnlockedGetNetworkList(self):
2400
    """Get the list of networks.
2401

2402
    This function is for internal use, when the config lock is already held.
2403

2404
    """
2405
    return self._config_data.networks.keys()
2406

    
2407
  @locking.ssynchronized(_config_lock, shared=1)
2408
  def GetNetworkList(self):
2409
    """Get the list of networks.
2410

2411
    @return: array of networks, ex. ["main", "vlan100", "200]
2412

2413
    """
2414
    return self._UnlockedGetNetworkList()
2415

    
2416
  @locking.ssynchronized(_config_lock, shared=1)
2417
  def GetNetworkNames(self):
2418
    """Get a list of network names
2419

2420
    """
2421
    names = [net.name
2422
             for net in self._config_data.networks.values()]
2423
    return names
2424

    
2425
  def _UnlockedGetNetwork(self, uuid):
2426
    """Returns information about a network.
2427

2428
    This function is for internal use, when the config lock is already held.
2429

2430
    """
2431
    if uuid not in self._config_data.networks:
2432
      return None
2433

    
2434
    return self._config_data.networks[uuid]
2435

    
2436
  @locking.ssynchronized(_config_lock, shared=1)
2437
  def GetNetwork(self, uuid):
2438
    """Returns information about a network.
2439

2440
    It takes the information from the configuration file.
2441

2442
    @param uuid: UUID of the network
2443

2444
    @rtype: L{objects.Network}
2445
    @return: the network object
2446

2447
    """
2448
    return self._UnlockedGetNetwork(uuid)
2449

    
2450
  @locking.ssynchronized(_config_lock)
2451
  def AddNetwork(self, net, ec_id, check_uuid=True):
2452
    """Add a network to the configuration.
2453

2454
    @type net: L{objects.Network}
2455
    @param net: the Network object to add
2456
    @type ec_id: string
2457
    @param ec_id: unique id for the job to use when creating a missing UUID
2458

2459
    """
2460
    self._UnlockedAddNetwork(net, ec_id, check_uuid)
2461
    self._WriteConfig()
2462

    
2463
  def _UnlockedAddNetwork(self, net, ec_id, check_uuid):
2464
    """Add a network to the configuration.
2465

2466
    """
2467
    logging.info("Adding network %s to configuration", net.name)
2468

    
2469
    if check_uuid:
2470
      self._EnsureUUID(net, ec_id)
2471

    
2472
    existing_uuid = self._UnlockedLookupNetwork(net.name)
2473
    if existing_uuid:
2474
      raise errors.OpPrereqError("Desired network name '%s' already"
2475
                                 " exists as a network (UUID: %s)" %
2476
                                 (net.name, existing_uuid),
2477
                                 errors.ECODE_EXISTS)
2478
    net.serial_no = 1
2479
    self._config_data.networks[net.uuid] = net
2480
    self._config_data.cluster.serial_no += 1
2481

    
2482
  def _UnlockedLookupNetwork(self, target):
2483
    """Lookup a network's UUID.
2484

2485
    @type target: string
2486
    @param target: network name or UUID
2487
    @rtype: string
2488
    @return: network UUID
2489
    @raises errors.OpPrereqError: when the target network cannot be found
2490

2491
    """
2492
    if target in self._config_data.networks:
2493
      return target
2494
    for net in self._config_data.networks.values():
2495
      if net.name == target:
2496
        return net.uuid
2497
    return None
2498

    
2499
  @locking.ssynchronized(_config_lock, shared=1)
2500
  def LookupNetwork(self, target):
2501
    """Lookup a network's UUID.
2502

2503
    This function is just a wrapper over L{_UnlockedLookupNetwork}.
2504

2505
    @type target: string
2506
    @param target: network name or UUID
2507
    @rtype: string
2508
    @return: network UUID
2509

2510
    """
2511
    return self._UnlockedLookupNetwork(target)
2512

    
2513
  @locking.ssynchronized(_config_lock)
2514
  def RemoveNetwork(self, network_uuid):
2515
    """Remove a network from the configuration.
2516

2517
    @type network_uuid: string
2518
    @param network_uuid: the UUID of the network to remove
2519

2520
    """
2521
    logging.info("Removing network %s from configuration", network_uuid)
2522

    
2523
    if network_uuid not in self._config_data.networks:
2524
      raise errors.ConfigurationError("Unknown network '%s'" % network_uuid)
2525

    
2526
    del self._config_data.networks[network_uuid]
2527
    self._config_data.cluster.serial_no += 1
2528
    self._WriteConfig()
2529

    
2530
  def _UnlockedGetGroupNetParams(self, net, node):
2531
    """Get the netparams (mode, link) of a network.
2532

2533
    Get a network's netparams for a given node.
2534

2535
    @type net: string
2536
    @param net: network name
2537
    @type node: string
2538
    @param node: node name
2539
    @rtype: dict or None
2540
    @return: netparams
2541

2542
    """
2543
    net_uuid = self._UnlockedLookupNetwork(net)
2544
    if net_uuid is None:
2545
      return None
2546

    
2547
    node_info = self._UnlockedGetNodeInfo(node)
2548
    nodegroup_info = self._UnlockedGetNodeGroup(node_info.group)
2549
    netparams = nodegroup_info.networks.get(net_uuid, None)
2550

    
2551
    return netparams
2552

    
2553
  @locking.ssynchronized(_config_lock, shared=1)
2554
  def GetGroupNetParams(self, net, node):
2555
    """Locking wrapper of _UnlockedGetGroupNetParams()
2556

2557
    """
2558
    return self._UnlockedGetGroupNetParams(net, node)
2559

    
2560
  @locking.ssynchronized(_config_lock, shared=1)
2561
  def CheckIPInNodeGroup(self, ip, node):
2562
    """Check for conflictig IP.
2563

2564
    @type ip: string
2565
    @param ip: ip address
2566
    @type node: string
2567
    @param node: node name
2568
    @rtype: (string, dict) or (None, None)
2569
    @return: (network name, netparams)
2570

2571
    """
2572
    if ip is None:
2573
      return (None, None)
2574
    node_info = self._UnlockedGetNodeInfo(node)
2575
    nodegroup_info = self._UnlockedGetNodeGroup(node_info.group)
2576
    for net_uuid in nodegroup_info.networks.keys():
2577
      net_info = self._UnlockedGetNetwork(net_uuid)
2578
      pool = network.AddressPool(net_info)
2579
      if pool.Contains(ip):
2580
        return (net_info.name, nodegroup_info.networks[net_uuid])
2581

    
2582
    return (None, None)