Statistics
| Branch: | Tag: | Revision:

root / lib / config.py @ b691385f

History | View | Annotate | Download (92.8 kB)

1
#
2
#
3

    
4
# Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013 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 copy
38
import os
39
import random
40
import logging
41
import time
42
import itertools
43

    
44
from ganeti import errors
45
from ganeti import locking
46
from ganeti import utils
47
from ganeti import constants
48
from ganeti import rpc
49
from ganeti import objects
50
from ganeti import serializer
51
from ganeti import uidpool
52
from ganeti import netutils
53
from ganeti import runtime
54
from ganeti import pathutils
55
from ganeti import network
56

    
57

    
58
_config_lock = locking.SharedLock("ConfigWriter")
59

    
60
# job id used for resource management at config upgrade time
61
_UPGRADE_CONFIG_JID = "jid-cfg-upgrade"
62

    
63

    
64
def _ValidateConfig(data):
65
  """Verifies that a configuration objects looks valid.
66

67
  This only verifies the version of the configuration.
68

69
  @raise errors.ConfigurationError: if the version differs from what
70
      we expect
71

72
  """
73
  if data.version != constants.CONFIG_VERSION:
74
    raise errors.ConfigVersionMismatch(constants.CONFIG_VERSION, data.version)
75

    
76

    
77
class TemporaryReservationManager:
78
  """A temporary resource reservation manager.
79

80
  This is used to reserve resources in a job, before using them, making sure
81
  other jobs cannot get them in the meantime.
82

83
  """
84
  def __init__(self):
85
    self._ec_reserved = {}
86

    
87
  def Reserved(self, resource):
88
    for holder_reserved in self._ec_reserved.values():
89
      if resource in holder_reserved:
90
        return True
91
    return False
92

    
93
  def Reserve(self, ec_id, resource):
94
    if self.Reserved(resource):
95
      raise errors.ReservationError("Duplicate reservation for resource '%s'"
96
                                    % str(resource))
97
    if ec_id not in self._ec_reserved:
98
      self._ec_reserved[ec_id] = set([resource])
99
    else:
100
      self._ec_reserved[ec_id].add(resource)
101

    
102
  def DropECReservations(self, ec_id):
103
    if ec_id in self._ec_reserved:
104
      del self._ec_reserved[ec_id]
105

    
106
  def GetReserved(self):
107
    all_reserved = set()
108
    for holder_reserved in self._ec_reserved.values():
109
      all_reserved.update(holder_reserved)
110
    return all_reserved
111

    
112
  def GetECReserved(self, ec_id):
113
    """ Used when you want to retrieve all reservations for a specific
114
        execution context. E.g when commiting reserved IPs for a specific
115
        network.
116

117
    """
118
    ec_reserved = set()
119
    if ec_id in self._ec_reserved:
120
      ec_reserved.update(self._ec_reserved[ec_id])
121
    return ec_reserved
122

    
123
  def Generate(self, existing, generate_one_fn, ec_id):
124
    """Generate a new resource of this type
125

126
    """
127
    assert callable(generate_one_fn)
128

    
129
    all_elems = self.GetReserved()
130
    all_elems.update(existing)
131
    retries = 64
132
    while retries > 0:
133
      new_resource = generate_one_fn()
134
      if new_resource is not None and new_resource not in all_elems:
135
        break
136
    else:
137
      raise errors.ConfigurationError("Not able generate new resource"
138
                                      " (last tried: %s)" % new_resource)
139
    self.Reserve(ec_id, new_resource)
140
    return new_resource
141

    
142

    
143
def _MatchNameComponentIgnoreCase(short_name, names):
144
  """Wrapper around L{utils.text.MatchNameComponent}.
145

146
  """
147
  return utils.MatchNameComponent(short_name, names, case_sensitive=False)
148

    
149

    
150
def _CheckInstanceDiskIvNames(disks):
151
  """Checks if instance's disks' C{iv_name} attributes are in order.
152

153
  @type disks: list of L{objects.Disk}
154
  @param disks: List of disks
155
  @rtype: list of tuples; (int, string, string)
156
  @return: List of wrongly named disks, each tuple contains disk index,
157
    expected and actual name
158

159
  """
160
  result = []
161

    
162
  for (idx, disk) in enumerate(disks):
163
    exp_iv_name = "disk/%s" % idx
164
    if disk.iv_name != exp_iv_name:
165
      result.append((idx, exp_iv_name, disk.iv_name))
166

    
167
  return result
168

    
169

    
170
class ConfigWriter:
171
  """The interface to the cluster configuration.
172

173
  @ivar _temporary_lvs: reservation manager for temporary LVs
174
  @ivar _all_rms: a list of all temporary reservation managers
175

176
  """
177
  def __init__(self, cfg_file=None, offline=False, _getents=runtime.GetEnts,
178
               accept_foreign=False):
179
    self.write_count = 0
180
    self._lock = _config_lock
181
    self._config_data = None
182
    self._offline = offline
183
    if cfg_file is None:
184
      self._cfg_file = pathutils.CLUSTER_CONF_FILE
185
    else:
186
      self._cfg_file = cfg_file
187
    self._getents = _getents
188
    self._temporary_ids = TemporaryReservationManager()
189
    self._temporary_drbds = {}
190
    self._temporary_macs = TemporaryReservationManager()
191
    self._temporary_secrets = TemporaryReservationManager()
192
    self._temporary_lvs = TemporaryReservationManager()
193
    self._temporary_ips = TemporaryReservationManager()
194
    self._all_rms = [self._temporary_ids, self._temporary_macs,
195
                     self._temporary_secrets, self._temporary_lvs,
196
                     self._temporary_ips]
197
    # Note: in order to prevent errors when resolving our name in
198
    # _DistributeConfig, we compute it here once and reuse it; it's
199
    # better to raise an error before starting to modify the config
200
    # file than after it was modified
201
    self._my_hostname = netutils.Hostname.GetSysName()
202
    self._last_cluster_serial = -1
203
    self._cfg_id = None
204
    self._context = None
205
    self._OpenConfig(accept_foreign)
206

    
207
  def _GetRpc(self, address_list):
208
    """Returns RPC runner for configuration.
209

210
    """
211
    return rpc.ConfigRunner(self._context, address_list)
212

    
213
  def SetContext(self, context):
214
    """Sets Ganeti context.
215

216
    """
217
    self._context = context
218

    
219
  # this method needs to be static, so that we can call it on the class
220
  @staticmethod
221
  def IsCluster():
222
    """Check if the cluster is configured.
223

224
    """
225
    return os.path.exists(pathutils.CLUSTER_CONF_FILE)
226

    
227
  @locking.ssynchronized(_config_lock, shared=1)
228
  def GetNdParams(self, node):
229
    """Get the node params populated with cluster defaults.
230

231
    @type node: L{objects.Node}
232
    @param node: The node we want to know the params for
233
    @return: A dict with the filled in node params
234

235
    """
236
    nodegroup = self._UnlockedGetNodeGroup(node.group)
237
    return self._config_data.cluster.FillND(node, nodegroup)
238

    
239
  @locking.ssynchronized(_config_lock, shared=1)
240
  def GetInstanceDiskParams(self, instance):
241
    """Get the disk params populated with inherit chain.
242

243
    @type instance: L{objects.Instance}
244
    @param instance: The instance we want to know the params for
245
    @return: A dict with the filled in disk params
246

247
    """
248
    node = self._UnlockedGetNodeInfo(instance.primary_node)
249
    nodegroup = self._UnlockedGetNodeGroup(node.group)
250
    return self._UnlockedGetGroupDiskParams(nodegroup)
251

    
252
  @locking.ssynchronized(_config_lock, shared=1)
253
  def GetGroupDiskParams(self, group):
254
    """Get the disk params populated with inherit chain.
255

256
    @type group: L{objects.NodeGroup}
257
    @param group: The group we want to know the params for
258
    @return: A dict with the filled in disk params
259

260
    """
261
    return self._UnlockedGetGroupDiskParams(group)
262

    
263
  def _UnlockedGetGroupDiskParams(self, group):
264
    """Get the disk params populated with inherit chain down to node-group.
265

266
    @type group: L{objects.NodeGroup}
267
    @param group: The group we want to know the params for
268
    @return: A dict with the filled in disk params
269

270
    """
271
    return self._config_data.cluster.SimpleFillDP(group.diskparams)
272

    
273
  def _UnlockedGetNetworkMACPrefix(self, net_uuid):
274
    """Return the network mac prefix if it exists or the cluster level default.
275

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

    
283
    return prefix
284

    
285
  def _GenerateOneMAC(self, prefix=None):
286
    """Return a function that randomly generates a MAC suffic
287
       and appends it to the given prefix. If prefix is not given get
288
       the cluster level default.
289

290
    """
291
    if not prefix:
292
      prefix = self._config_data.cluster.mac_prefix
293

    
294
    def GenMac():
295
      byte1 = random.randrange(0, 256)
296
      byte2 = random.randrange(0, 256)
297
      byte3 = random.randrange(0, 256)
298
      mac = "%s:%02x:%02x:%02x" % (prefix, byte1, byte2, byte3)
299
      return mac
300

    
301
    return GenMac
302

    
303
  @locking.ssynchronized(_config_lock, shared=1)
304
  def GenerateMAC(self, net_uuid, ec_id):
305
    """Generate a MAC for an instance.
306

307
    This should check the current instances for duplicates.
308

309
    """
310
    existing = self._AllMACs()
311
    prefix = self._UnlockedGetNetworkMACPrefix(net_uuid)
312
    gen_mac = self._GenerateOneMAC(prefix)
313
    return self._temporary_ids.Generate(existing, gen_mac, ec_id)
314

    
315
  @locking.ssynchronized(_config_lock, shared=1)
316
  def ReserveMAC(self, mac, ec_id):
317
    """Reserve a MAC for an instance.
318

319
    This only checks instances managed by this cluster, it does not
320
    check for potential collisions elsewhere.
321

322
    """
323
    all_macs = self._AllMACs()
324
    if mac in all_macs:
325
      raise errors.ReservationError("mac already in use")
326
    else:
327
      self._temporary_macs.Reserve(ec_id, mac)
328

    
329
  def _UnlockedCommitTemporaryIps(self, ec_id):
330
    """Commit all reserved IP address to their respective pools
331

332
    """
333
    for action, address, net_uuid in self._temporary_ips.GetECReserved(ec_id):
334
      self._UnlockedCommitIp(action, net_uuid, address)
335

    
336
  def _UnlockedCommitIp(self, action, net_uuid, address):
337
    """Commit a reserved IP address to an IP pool.
338

339
    The IP address is taken from the network's IP pool and marked as reserved.
340

341
    """
342
    nobj = self._UnlockedGetNetwork(net_uuid)
343
    pool = network.AddressPool(nobj)
344
    if action == constants.RESERVE_ACTION:
345
      pool.Reserve(address)
346
    elif action == constants.RELEASE_ACTION:
347
      pool.Release(address)
348

    
349
  def _UnlockedReleaseIp(self, net_uuid, address, ec_id):
350
    """Give a specific IP address back to an IP pool.
351

352
    The IP address is returned to the IP pool designated by pool_id and marked
353
    as reserved.
354

355
    """
356
    self._temporary_ips.Reserve(ec_id,
357
                                (constants.RELEASE_ACTION, address, net_uuid))
358

    
359
  @locking.ssynchronized(_config_lock, shared=1)
360
  def ReleaseIp(self, net_uuid, address, ec_id):
361
    """Give a specified IP address back to an IP pool.
362

363
    This is just a wrapper around _UnlockedReleaseIp.
364

365
    """
366
    if net_uuid:
367
      self._UnlockedReleaseIp(net_uuid, address, ec_id)
368

    
369
  @locking.ssynchronized(_config_lock, shared=1)
370
  def GenerateIp(self, net_uuid, ec_id):
371
    """Find a free IPv4 address for an instance.
372

373
    """
374
    nobj = self._UnlockedGetNetwork(net_uuid)
375
    pool = network.AddressPool(nobj)
376

    
377
    def gen_one():
378
      try:
379
        ip = pool.GenerateFree()
380
      except errors.AddressPoolError:
381
        raise errors.ReservationError("Cannot generate IP. Network is full")
382
      return (constants.RESERVE_ACTION, ip, net_uuid)
383

    
384
    _, address, _ = self._temporary_ips.Generate([], gen_one, ec_id)
385
    return address
386

    
387
  def _UnlockedReserveIp(self, net_uuid, address, ec_id):
388
    """Reserve a given IPv4 address for use by an instance.
389

390
    """
391
    nobj = self._UnlockedGetNetwork(net_uuid)
392
    pool = network.AddressPool(nobj)
393
    try:
394
      isreserved = pool.IsReserved(address)
395
    except errors.AddressPoolError:
396
      raise errors.ReservationError("IP address not in network")
397
    if isreserved:
398
      raise errors.ReservationError("IP address already in use")
399

    
400
    return self._temporary_ips.Reserve(ec_id,
401
                                       (constants.RESERVE_ACTION,
402
                                        address, net_uuid))
403

    
404
  @locking.ssynchronized(_config_lock, shared=1)
405
  def ReserveIp(self, net_uuid, address, ec_id):
406
    """Reserve a given IPv4 address for use by an instance.
407

408
    """
409
    if net_uuid:
410
      return self._UnlockedReserveIp(net_uuid, address, ec_id)
411

    
412
  @locking.ssynchronized(_config_lock, shared=1)
413
  def ReserveLV(self, lv_name, ec_id):
414
    """Reserve an VG/LV pair for an instance.
415

416
    @type lv_name: string
417
    @param lv_name: the logical volume name to reserve
418

419
    """
420
    all_lvs = self._AllLVs()
421
    if lv_name in all_lvs:
422
      raise errors.ReservationError("LV already in use")
423
    else:
424
      self._temporary_lvs.Reserve(ec_id, lv_name)
425

    
426
  @locking.ssynchronized(_config_lock, shared=1)
427
  def GenerateDRBDSecret(self, ec_id):
428
    """Generate a DRBD secret.
429

430
    This checks the current disks for duplicates.
431

432
    """
433
    return self._temporary_secrets.Generate(self._AllDRBDSecrets(),
434
                                            utils.GenerateSecret,
435
                                            ec_id)
436

    
437
  def _AllLVs(self):
438
    """Compute the list of all LVs.
439

440
    """
441
    lvnames = set()
442
    for instance in self._config_data.instances.values():
443
      node_data = instance.MapLVsByNode()
444
      for lv_list in node_data.values():
445
        lvnames.update(lv_list)
446
    return lvnames
447

    
448
  def _AllDisks(self):
449
    """Compute the list of all Disks.
450

451
    """
452
    disks = []
453
    for instance in self._config_data.instances.values():
454
      disks.extend(instance.disks)
455
    return disks
456

    
457
  def _AllNICs(self):
458
    """Compute the list of all NICs.
459

460
    """
461
    nics = []
462
    for instance in self._config_data.instances.values():
463
      nics.extend(instance.nics)
464
    return nics
465

    
466
  def _AllIDs(self, include_temporary):
467
    """Compute the list of all UUIDs and names we have.
468

469
    @type include_temporary: boolean
470
    @param include_temporary: whether to include the _temporary_ids set
471
    @rtype: set
472
    @return: a set of IDs
473

474
    """
475
    existing = set()
476
    if include_temporary:
477
      existing.update(self._temporary_ids.GetReserved())
478
    existing.update(self._AllLVs())
479
    existing.update(self._config_data.instances.keys())
480
    existing.update(self._config_data.nodes.keys())
481
    existing.update([i.uuid for i in self._AllUUIDObjects() if i.uuid])
482
    return existing
483

    
484
  def _GenerateUniqueID(self, ec_id):
485
    """Generate an unique UUID.
486

487
    This checks the current node, instances and disk names for
488
    duplicates.
489

490
    @rtype: string
491
    @return: the unique id
492

493
    """
494
    existing = self._AllIDs(include_temporary=False)
495
    return self._temporary_ids.Generate(existing, utils.NewUUID, ec_id)
496

    
497
  @locking.ssynchronized(_config_lock, shared=1)
498
  def GenerateUniqueID(self, ec_id):
499
    """Generate an unique ID.
500

501
    This is just a wrapper over the unlocked version.
502

503
    @type ec_id: string
504
    @param ec_id: unique id for the job to reserve the id to
505

506
    """
507
    return self._GenerateUniqueID(ec_id)
508

    
509
  def _AllMACs(self):
510
    """Return all MACs present in the config.
511

512
    @rtype: list
513
    @return: the list of all MACs
514

515
    """
516
    result = []
517
    for instance in self._config_data.instances.values():
518
      for nic in instance.nics:
519
        result.append(nic.mac)
520

    
521
    return result
522

    
523
  def _AllDRBDSecrets(self):
524
    """Return all DRBD secrets present in the config.
525

526
    @rtype: list
527
    @return: the list of all DRBD secrets
528

529
    """
530
    def helper(disk, result):
531
      """Recursively gather secrets from this disk."""
532
      if disk.dev_type == constants.DT_DRBD8:
533
        result.append(disk.logical_id[5])
534
      if disk.children:
535
        for child in disk.children:
536
          helper(child, result)
537

    
538
    result = []
539
    for instance in self._config_data.instances.values():
540
      for disk in instance.disks:
541
        helper(disk, result)
542

    
543
    return result
544

    
545
  def _CheckDiskIDs(self, disk, l_ids, p_ids):
546
    """Compute duplicate disk IDs
547

548
    @type disk: L{objects.Disk}
549
    @param disk: the disk at which to start searching
550
    @type l_ids: list
551
    @param l_ids: list of current logical ids
552
    @type p_ids: list
553
    @param p_ids: list of current physical ids
554
    @rtype: list
555
    @return: a list of error messages
556

557
    """
558
    result = []
559
    if disk.logical_id is not None:
560
      if disk.logical_id in l_ids:
561
        result.append("duplicate logical id %s" % str(disk.logical_id))
562
      else:
563
        l_ids.append(disk.logical_id)
564
    if disk.physical_id is not None:
565
      if disk.physical_id in p_ids:
566
        result.append("duplicate physical id %s" % str(disk.physical_id))
567
      else:
568
        p_ids.append(disk.physical_id)
569

    
570
    if disk.children:
571
      for child in disk.children:
572
        result.extend(self._CheckDiskIDs(child, l_ids, p_ids))
573
    return result
574

    
575
  def _UnlockedVerifyConfig(self):
576
    """Verify function.
577

578
    @rtype: list
579
    @return: a list of error messages; a non-empty list signifies
580
        configuration errors
581

582
    """
583
    # pylint: disable=R0914
584
    result = []
585
    seen_macs = []
586
    ports = {}
587
    data = self._config_data
588
    cluster = data.cluster
589
    seen_lids = []
590
    seen_pids = []
591

    
592
    # global cluster checks
593
    if not cluster.enabled_hypervisors:
594
      result.append("enabled hypervisors list doesn't have any entries")
595
    invalid_hvs = set(cluster.enabled_hypervisors) - constants.HYPER_TYPES
596
    if invalid_hvs:
597
      result.append("enabled hypervisors contains invalid entries: %s" %
598
                    utils.CommaJoin(invalid_hvs))
599
    missing_hvp = (set(cluster.enabled_hypervisors) -
600
                   set(cluster.hvparams.keys()))
601
    if missing_hvp:
602
      result.append("hypervisor parameters missing for the enabled"
603
                    " hypervisor(s) %s" % utils.CommaJoin(missing_hvp))
604

    
605
    if not cluster.enabled_disk_templates:
606
      result.append("enabled disk templates list doesn't have any entries")
607
    invalid_disk_templates = set(cluster.enabled_disk_templates) \
608
                               - constants.DISK_TEMPLATES
609
    if invalid_disk_templates:
610
      result.append("enabled disk templates list contains invalid entries:"
611
                    " %s" % utils.CommaJoin(invalid_disk_templates))
612

    
613
    if cluster.master_node not in data.nodes:
614
      result.append("cluster has invalid primary node '%s'" %
615
                    cluster.master_node)
616

    
617
    def _helper(owner, attr, value, template):
618
      try:
619
        utils.ForceDictType(value, template)
620
      except errors.GenericError, err:
621
        result.append("%s has invalid %s: %s" % (owner, attr, err))
622

    
623
    def _helper_nic(owner, params):
624
      try:
625
        objects.NIC.CheckParameterSyntax(params)
626
      except errors.ConfigurationError, err:
627
        result.append("%s has invalid nicparams: %s" % (owner, err))
628

    
629
    def _helper_ipolicy(owner, ipolicy, iscluster):
630
      try:
631
        objects.InstancePolicy.CheckParameterSyntax(ipolicy, iscluster)
632
      except errors.ConfigurationError, err:
633
        result.append("%s has invalid instance policy: %s" % (owner, err))
634
      for key, value in ipolicy.items():
635
        if key == constants.ISPECS_MINMAX:
636
          for k in range(len(value)):
637
            _helper_ispecs(owner, "ipolicy/%s[%s]" % (key, k), value[k])
638
        elif key == constants.ISPECS_STD:
639
          _helper(owner, "ipolicy/" + key, value,
640
                  constants.ISPECS_PARAMETER_TYPES)
641
        else:
642
          # FIXME: assuming list type
643
          if key in constants.IPOLICY_PARAMETERS:
644
            exp_type = float
645
          else:
646
            exp_type = list
647
          if not isinstance(value, exp_type):
648
            result.append("%s has invalid instance policy: for %s,"
649
                          " expecting %s, got %s" %
650
                          (owner, key, exp_type.__name__, type(value)))
651

    
652
    def _helper_ispecs(owner, parentkey, params):
653
      for (key, value) in params.items():
654
        fullkey = "/".join([parentkey, key])
655
        _helper(owner, fullkey, value, constants.ISPECS_PARAMETER_TYPES)
656

    
657
    # check cluster parameters
658
    _helper("cluster", "beparams", cluster.SimpleFillBE({}),
659
            constants.BES_PARAMETER_TYPES)
660
    _helper("cluster", "nicparams", cluster.SimpleFillNIC({}),
661
            constants.NICS_PARAMETER_TYPES)
662
    _helper_nic("cluster", cluster.SimpleFillNIC({}))
663
    _helper("cluster", "ndparams", cluster.SimpleFillND({}),
664
            constants.NDS_PARAMETER_TYPES)
665
    _helper_ipolicy("cluster", cluster.ipolicy, True)
666

    
667
    # per-instance checks
668
    for instance_name in data.instances:
669
      instance = data.instances[instance_name]
670
      if instance.name != instance_name:
671
        result.append("instance '%s' is indexed by wrong name '%s'" %
672
                      (instance.name, instance_name))
673
      if instance.primary_node not in data.nodes:
674
        result.append("instance '%s' has invalid primary node '%s'" %
675
                      (instance_name, instance.primary_node))
676
      for snode in instance.secondary_nodes:
677
        if snode not in data.nodes:
678
          result.append("instance '%s' has invalid secondary node '%s'" %
679
                        (instance_name, snode))
680
      for idx, nic in enumerate(instance.nics):
681
        if nic.mac in seen_macs:
682
          result.append("instance '%s' has NIC %d mac %s duplicate" %
683
                        (instance_name, idx, nic.mac))
684
        else:
685
          seen_macs.append(nic.mac)
686
        if nic.nicparams:
687
          filled = cluster.SimpleFillNIC(nic.nicparams)
688
          owner = "instance %s nic %d" % (instance.name, idx)
689
          _helper(owner, "nicparams",
690
                  filled, constants.NICS_PARAMETER_TYPES)
691
          _helper_nic(owner, filled)
692

    
693
      # disk template checks
694
      if not instance.disk_template in data.cluster.enabled_disk_templates:
695
        result.append("instance '%s' uses the disabled disk template '%s'." %
696
                      (instance_name, instance.disk_template))
697

    
698
      # parameter checks
699
      if instance.beparams:
700
        _helper("instance %s" % instance.name, "beparams",
701
                cluster.FillBE(instance), constants.BES_PARAMETER_TYPES)
702

    
703
      # gather the drbd ports for duplicate checks
704
      for (idx, dsk) in enumerate(instance.disks):
705
        if dsk.dev_type in constants.LDS_DRBD:
706
          tcp_port = dsk.logical_id[2]
707
          if tcp_port not in ports:
708
            ports[tcp_port] = []
709
          ports[tcp_port].append((instance.name, "drbd disk %s" % idx))
710
      # gather network port reservation
711
      net_port = getattr(instance, "network_port", None)
712
      if net_port is not None:
713
        if net_port not in ports:
714
          ports[net_port] = []
715
        ports[net_port].append((instance.name, "network port"))
716

    
717
      # instance disk verify
718
      for idx, disk in enumerate(instance.disks):
719
        result.extend(["instance '%s' disk %d error: %s" %
720
                       (instance.name, idx, msg) for msg in disk.Verify()])
721
        result.extend(self._CheckDiskIDs(disk, seen_lids, seen_pids))
722

    
723
      wrong_names = _CheckInstanceDiskIvNames(instance.disks)
724
      if wrong_names:
725
        tmp = "; ".join(("name of disk %s should be '%s', but is '%s'" %
726
                         (idx, exp_name, actual_name))
727
                        for (idx, exp_name, actual_name) in wrong_names)
728

    
729
        result.append("Instance '%s' has wrongly named disks: %s" %
730
                      (instance.name, tmp))
731

    
732
    # cluster-wide pool of free ports
733
    for free_port in cluster.tcpudp_port_pool:
734
      if free_port not in ports:
735
        ports[free_port] = []
736
      ports[free_port].append(("cluster", "port marked as free"))
737

    
738
    # compute tcp/udp duplicate ports
739
    keys = ports.keys()
740
    keys.sort()
741
    for pnum in keys:
742
      pdata = ports[pnum]
743
      if len(pdata) > 1:
744
        txt = utils.CommaJoin(["%s/%s" % val for val in pdata])
745
        result.append("tcp/udp port %s has duplicates: %s" % (pnum, txt))
746

    
747
    # highest used tcp port check
748
    if keys:
749
      if keys[-1] > cluster.highest_used_port:
750
        result.append("Highest used port mismatch, saved %s, computed %s" %
751
                      (cluster.highest_used_port, keys[-1]))
752

    
753
    if not data.nodes[cluster.master_node].master_candidate:
754
      result.append("Master node is not a master candidate")
755

    
756
    # master candidate checks
757
    mc_now, mc_max, _ = self._UnlockedGetMasterCandidateStats()
758
    if mc_now < mc_max:
759
      result.append("Not enough master candidates: actual %d, target %d" %
760
                    (mc_now, mc_max))
761

    
762
    # node checks
763
    for node_uuid, node in data.nodes.items():
764
      if node.uuid != node_uuid:
765
        result.append("Node '%s' is indexed by wrong UUID '%s'" %
766
                      (node.name, node_uuid))
767
      if [node.master_candidate, node.drained, node.offline].count(True) > 1:
768
        result.append("Node %s state is invalid: master_candidate=%s,"
769
                      " drain=%s, offline=%s" %
770
                      (node.name, node.master_candidate, node.drained,
771
                       node.offline))
772
      if node.group not in data.nodegroups:
773
        result.append("Node '%s' has invalid group '%s'" %
774
                      (node.name, node.group))
775
      else:
776
        _helper("node %s" % node.name, "ndparams",
777
                cluster.FillND(node, data.nodegroups[node.group]),
778
                constants.NDS_PARAMETER_TYPES)
779
      used_globals = constants.NDC_GLOBALS.intersection(node.ndparams)
780
      if used_globals:
781
        result.append("Node '%s' has some global parameters set: %s" %
782
                      (node.name, utils.CommaJoin(used_globals)))
783

    
784
    # nodegroups checks
785
    nodegroups_names = set()
786
    for nodegroup_uuid in data.nodegroups:
787
      nodegroup = data.nodegroups[nodegroup_uuid]
788
      if nodegroup.uuid != nodegroup_uuid:
789
        result.append("node group '%s' (uuid: '%s') indexed by wrong uuid '%s'"
790
                      % (nodegroup.name, nodegroup.uuid, nodegroup_uuid))
791
      if utils.UUID_RE.match(nodegroup.name.lower()):
792
        result.append("node group '%s' (uuid: '%s') has uuid-like name" %
793
                      (nodegroup.name, nodegroup.uuid))
794
      if nodegroup.name in nodegroups_names:
795
        result.append("duplicate node group name '%s'" % nodegroup.name)
796
      else:
797
        nodegroups_names.add(nodegroup.name)
798
      group_name = "group %s" % nodegroup.name
799
      _helper_ipolicy(group_name, cluster.SimpleFillIPolicy(nodegroup.ipolicy),
800
                      False)
801
      if nodegroup.ndparams:
802
        _helper(group_name, "ndparams",
803
                cluster.SimpleFillND(nodegroup.ndparams),
804
                constants.NDS_PARAMETER_TYPES)
805

    
806
    # drbd minors check
807
    _, duplicates = self._UnlockedComputeDRBDMap()
808
    for node, minor, instance_a, instance_b in duplicates:
809
      result.append("DRBD minor %d on node %s is assigned twice to instances"
810
                    " %s and %s" % (minor, node, instance_a, instance_b))
811

    
812
    # IP checks
813
    default_nicparams = cluster.nicparams[constants.PP_DEFAULT]
814
    ips = {}
815

    
816
    def _AddIpAddress(ip, name):
817
      ips.setdefault(ip, []).append(name)
818

    
819
    _AddIpAddress(cluster.master_ip, "cluster_ip")
820

    
821
    for node in data.nodes.values():
822
      _AddIpAddress(node.primary_ip, "node:%s/primary" % node.name)
823
      if node.secondary_ip != node.primary_ip:
824
        _AddIpAddress(node.secondary_ip, "node:%s/secondary" % node.name)
825

    
826
    for instance in data.instances.values():
827
      for idx, nic in enumerate(instance.nics):
828
        if nic.ip is None:
829
          continue
830

    
831
        nicparams = objects.FillDict(default_nicparams, nic.nicparams)
832
        nic_mode = nicparams[constants.NIC_MODE]
833
        nic_link = nicparams[constants.NIC_LINK]
834

    
835
        if nic_mode == constants.NIC_MODE_BRIDGED:
836
          link = "bridge:%s" % nic_link
837
        elif nic_mode == constants.NIC_MODE_ROUTED:
838
          link = "route:%s" % nic_link
839
        else:
840
          raise errors.ProgrammerError("NIC mode '%s' not handled" % nic_mode)
841

    
842
        _AddIpAddress("%s/%s/%s" % (link, nic.ip, nic.network),
843
                      "instance:%s/nic:%d" % (instance.name, idx))
844

    
845
    for ip, owners in ips.items():
846
      if len(owners) > 1:
847
        result.append("IP address %s is used by multiple owners: %s" %
848
                      (ip, utils.CommaJoin(owners)))
849

    
850
    return result
851

    
852
  @locking.ssynchronized(_config_lock, shared=1)
853
  def VerifyConfig(self):
854
    """Verify function.
855

856
    This is just a wrapper over L{_UnlockedVerifyConfig}.
857

858
    @rtype: list
859
    @return: a list of error messages; a non-empty list signifies
860
        configuration errors
861

862
    """
863
    return self._UnlockedVerifyConfig()
864

    
865
  def _UnlockedSetDiskID(self, disk, node_uuid):
866
    """Convert the unique ID to the ID needed on the target nodes.
867

868
    This is used only for drbd, which needs ip/port configuration.
869

870
    The routine descends down and updates its children also, because
871
    this helps when the only the top device is passed to the remote
872
    node.
873

874
    This function is for internal use, when the config lock is already held.
875

876
    """
877
    if disk.children:
878
      for child in disk.children:
879
        self._UnlockedSetDiskID(child, node_uuid)
880

    
881
    if disk.logical_id is None and disk.physical_id is not None:
882
      return
883
    if disk.dev_type == constants.LD_DRBD8:
884
      pnode, snode, port, pminor, sminor, secret = disk.logical_id
885
      if node_uuid not in (pnode, snode):
886
        raise errors.ConfigurationError("DRBD device not knowing node %s" %
887
                                        node_uuid)
888
      pnode_info = self._UnlockedGetNodeInfo(pnode)
889
      snode_info = self._UnlockedGetNodeInfo(snode)
890
      if pnode_info is None or snode_info is None:
891
        raise errors.ConfigurationError("Can't find primary or secondary node"
892
                                        " for %s" % str(disk))
893
      p_data = (pnode_info.secondary_ip, port)
894
      s_data = (snode_info.secondary_ip, port)
895
      if pnode == node_uuid:
896
        disk.physical_id = p_data + s_data + (pminor, secret)
897
      else: # it must be secondary, we tested above
898
        disk.physical_id = s_data + p_data + (sminor, secret)
899
    else:
900
      disk.physical_id = disk.logical_id
901
    return
902

    
903
  @locking.ssynchronized(_config_lock)
904
  def SetDiskID(self, disk, node_uuid):
905
    """Convert the unique ID to the ID needed on the target nodes.
906

907
    This is used only for drbd, which needs ip/port configuration.
908

909
    The routine descends down and updates its children also, because
910
    this helps when the only the top device is passed to the remote
911
    node.
912

913
    """
914
    return self._UnlockedSetDiskID(disk, node_uuid)
915

    
916
  @locking.ssynchronized(_config_lock)
917
  def AddTcpUdpPort(self, port):
918
    """Adds a new port to the available port pool.
919

920
    @warning: this method does not "flush" the configuration (via
921
        L{_WriteConfig}); callers should do that themselves once the
922
        configuration is stable
923

924
    """
925
    if not isinstance(port, int):
926
      raise errors.ProgrammerError("Invalid type passed for port")
927

    
928
    self._config_data.cluster.tcpudp_port_pool.add(port)
929

    
930
  @locking.ssynchronized(_config_lock, shared=1)
931
  def GetPortList(self):
932
    """Returns a copy of the current port list.
933

934
    """
935
    return self._config_data.cluster.tcpudp_port_pool.copy()
936

    
937
  @locking.ssynchronized(_config_lock)
938
  def AllocatePort(self):
939
    """Allocate a port.
940

941
    The port will be taken from the available port pool or from the
942
    default port range (and in this case we increase
943
    highest_used_port).
944

945
    """
946
    # If there are TCP/IP ports configured, we use them first.
947
    if self._config_data.cluster.tcpudp_port_pool:
948
      port = self._config_data.cluster.tcpudp_port_pool.pop()
949
    else:
950
      port = self._config_data.cluster.highest_used_port + 1
951
      if port >= constants.LAST_DRBD_PORT:
952
        raise errors.ConfigurationError("The highest used port is greater"
953
                                        " than %s. Aborting." %
954
                                        constants.LAST_DRBD_PORT)
955
      self._config_data.cluster.highest_used_port = port
956

    
957
    self._WriteConfig()
958
    return port
959

    
960
  def _UnlockedComputeDRBDMap(self):
961
    """Compute the used DRBD minor/nodes.
962

963
    @rtype: (dict, list)
964
    @return: dictionary of node_uuid: dict of minor: instance_name;
965
        the returned dict will have all the nodes in it (even if with
966
        an empty list), and a list of duplicates; if the duplicates
967
        list is not empty, the configuration is corrupted and its caller
968
        should raise an exception
969

970
    """
971
    def _AppendUsedPorts(get_node_name_fn, instance_name, disk, used):
972
      duplicates = []
973
      if disk.dev_type == constants.LD_DRBD8 and len(disk.logical_id) >= 5:
974
        node_a, node_b, _, minor_a, minor_b = disk.logical_id[:5]
975
        for node_uuid, port in ((node_a, minor_a), (node_b, minor_b)):
976
          assert node_uuid in used, \
977
            ("Node '%s' of instance '%s' not found in node list" %
978
             (get_node_name_fn(node_uuid), instance_name))
979
          if port in used[node_uuid]:
980
            duplicates.append((node_uuid, port, instance_name,
981
                               used[node_uuid][port]))
982
          else:
983
            used[node_uuid][port] = instance_name
984
      if disk.children:
985
        for child in disk.children:
986
          duplicates.extend(_AppendUsedPorts(get_node_name_fn, instance_name,
987
                                             child, used))
988
      return duplicates
989

    
990
    duplicates = []
991
    my_dict = dict((node, {}) for node in self._config_data.nodes)
992
    for instance in self._config_data.instances.itervalues():
993
      for disk in instance.disks:
994
        duplicates.extend(_AppendUsedPorts(self._UnlockedGetNodeName,
995
                                           instance.name, disk, my_dict))
996
    for (node, minor), instance in self._temporary_drbds.iteritems():
997
      if minor in my_dict[node] and my_dict[node][minor] != instance:
998
        duplicates.append((node, minor, instance, my_dict[node][minor]))
999
      else:
1000
        my_dict[node][minor] = instance
1001
    return my_dict, duplicates
1002

    
1003
  @locking.ssynchronized(_config_lock)
1004
  def ComputeDRBDMap(self):
1005
    """Compute the used DRBD minor/nodes.
1006

1007
    This is just a wrapper over L{_UnlockedComputeDRBDMap}.
1008

1009
    @return: dictionary of node_uuid: dict of minor: instance_name;
1010
        the returned dict will have all the nodes in it (even if with
1011
        an empty list).
1012

1013
    """
1014
    d_map, duplicates = self._UnlockedComputeDRBDMap()
1015
    if duplicates:
1016
      raise errors.ConfigurationError("Duplicate DRBD ports detected: %s" %
1017
                                      str(duplicates))
1018
    return d_map
1019

    
1020
  @locking.ssynchronized(_config_lock)
1021
  def AllocateDRBDMinor(self, node_uuids, instance):
1022
    """Allocate a drbd minor.
1023

1024
    The free minor will be automatically computed from the existing
1025
    devices. A node can be given multiple times in order to allocate
1026
    multiple minors. The result is the list of minors, in the same
1027
    order as the passed nodes.
1028

1029
    @type instance: string
1030
    @param instance: the instance for which we allocate minors
1031

1032
    """
1033
    assert isinstance(instance, basestring), \
1034
           "Invalid argument '%s' passed to AllocateDRBDMinor" % instance
1035

    
1036
    d_map, duplicates = self._UnlockedComputeDRBDMap()
1037
    if duplicates:
1038
      raise errors.ConfigurationError("Duplicate DRBD ports detected: %s" %
1039
                                      str(duplicates))
1040
    result = []
1041
    for nuuid in node_uuids:
1042
      ndata = d_map[nuuid]
1043
      if not ndata:
1044
        # no minors used, we can start at 0
1045
        result.append(0)
1046
        ndata[0] = instance
1047
        self._temporary_drbds[(nuuid, 0)] = instance
1048
        continue
1049
      keys = ndata.keys()
1050
      keys.sort()
1051
      ffree = utils.FirstFree(keys)
1052
      if ffree is None:
1053
        # return the next minor
1054
        # TODO: implement high-limit check
1055
        minor = keys[-1] + 1
1056
      else:
1057
        minor = ffree
1058
      # double-check minor against current instances
1059
      assert minor not in d_map[nuuid], \
1060
             ("Attempt to reuse allocated DRBD minor %d on node %s,"
1061
              " already allocated to instance %s" %
1062
              (minor, nuuid, d_map[nuuid][minor]))
1063
      ndata[minor] = instance
1064
      # double-check minor against reservation
1065
      r_key = (nuuid, minor)
1066
      assert r_key not in self._temporary_drbds, \
1067
             ("Attempt to reuse reserved DRBD minor %d on node %s,"
1068
              " reserved for instance %s" %
1069
              (minor, nuuid, self._temporary_drbds[r_key]))
1070
      self._temporary_drbds[r_key] = instance
1071
      result.append(minor)
1072
    logging.debug("Request to allocate drbd minors, input: %s, returning %s",
1073
                  node_uuids, result)
1074
    return result
1075

    
1076
  def _UnlockedReleaseDRBDMinors(self, instance):
1077
    """Release temporary drbd minors allocated for a given instance.
1078

1079
    @type instance: string
1080
    @param instance: the instance for which temporary minors should be
1081
                     released
1082

1083
    """
1084
    assert isinstance(instance, basestring), \
1085
           "Invalid argument passed to ReleaseDRBDMinors"
1086
    for key, name in self._temporary_drbds.items():
1087
      if name == instance:
1088
        del self._temporary_drbds[key]
1089

    
1090
  @locking.ssynchronized(_config_lock)
1091
  def ReleaseDRBDMinors(self, instance):
1092
    """Release temporary drbd minors allocated for a given instance.
1093

1094
    This should be called on the error paths, on the success paths
1095
    it's automatically called by the ConfigWriter add and update
1096
    functions.
1097

1098
    This function is just a wrapper over L{_UnlockedReleaseDRBDMinors}.
1099

1100
    @type instance: string
1101
    @param instance: the instance for which temporary minors should be
1102
                     released
1103

1104
    """
1105
    self._UnlockedReleaseDRBDMinors(instance)
1106

    
1107
  @locking.ssynchronized(_config_lock, shared=1)
1108
  def GetConfigVersion(self):
1109
    """Get the configuration version.
1110

1111
    @return: Config version
1112

1113
    """
1114
    return self._config_data.version
1115

    
1116
  @locking.ssynchronized(_config_lock, shared=1)
1117
  def GetClusterName(self):
1118
    """Get cluster name.
1119

1120
    @return: Cluster name
1121

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

    
1125
  @locking.ssynchronized(_config_lock, shared=1)
1126
  def GetMasterNode(self):
1127
    """Get the UUID of the master node for this cluster.
1128

1129
    @return: Master node UUID
1130

1131
    """
1132
    return self._config_data.cluster.master_node
1133

    
1134
  @locking.ssynchronized(_config_lock, shared=1)
1135
  def GetMasterNodeName(self):
1136
    """Get the hostname of the master node for this cluster.
1137

1138
    @return: Master node hostname
1139

1140
    """
1141
    return self._UnlockedGetNodeName(self._config_data.cluster.master_node)
1142

    
1143
  @locking.ssynchronized(_config_lock, shared=1)
1144
  def GetMasterIP(self):
1145
    """Get the IP of the master node for this cluster.
1146

1147
    @return: Master IP
1148

1149
    """
1150
    return self._config_data.cluster.master_ip
1151

    
1152
  @locking.ssynchronized(_config_lock, shared=1)
1153
  def GetMasterNetdev(self):
1154
    """Get the master network device for this cluster.
1155

1156
    """
1157
    return self._config_data.cluster.master_netdev
1158

    
1159
  @locking.ssynchronized(_config_lock, shared=1)
1160
  def GetMasterNetmask(self):
1161
    """Get the netmask of the master node for this cluster.
1162

1163
    """
1164
    return self._config_data.cluster.master_netmask
1165

    
1166
  @locking.ssynchronized(_config_lock, shared=1)
1167
  def GetUseExternalMipScript(self):
1168
    """Get flag representing whether to use the external master IP setup script.
1169

1170
    """
1171
    return self._config_data.cluster.use_external_mip_script
1172

    
1173
  @locking.ssynchronized(_config_lock, shared=1)
1174
  def GetFileStorageDir(self):
1175
    """Get the file storage dir for this cluster.
1176

1177
    """
1178
    return self._config_data.cluster.file_storage_dir
1179

    
1180
  @locking.ssynchronized(_config_lock, shared=1)
1181
  def GetSharedFileStorageDir(self):
1182
    """Get the shared file storage dir for this cluster.
1183

1184
    """
1185
    return self._config_data.cluster.shared_file_storage_dir
1186

    
1187
  @locking.ssynchronized(_config_lock, shared=1)
1188
  def GetHypervisorType(self):
1189
    """Get the hypervisor type for this cluster.
1190

1191
    """
1192
    return self._config_data.cluster.enabled_hypervisors[0]
1193

    
1194
  @locking.ssynchronized(_config_lock, shared=1)
1195
  def GetHostKey(self):
1196
    """Return the rsa hostkey from the config.
1197

1198
    @rtype: string
1199
    @return: the rsa hostkey
1200

1201
    """
1202
    return self._config_data.cluster.rsahostkeypub
1203

    
1204
  @locking.ssynchronized(_config_lock, shared=1)
1205
  def GetDefaultIAllocator(self):
1206
    """Get the default instance allocator for this cluster.
1207

1208
    """
1209
    return self._config_data.cluster.default_iallocator
1210

    
1211
  @locking.ssynchronized(_config_lock, shared=1)
1212
  def GetPrimaryIPFamily(self):
1213
    """Get cluster primary ip family.
1214

1215
    @return: primary ip family
1216

1217
    """
1218
    return self._config_data.cluster.primary_ip_family
1219

    
1220
  @locking.ssynchronized(_config_lock, shared=1)
1221
  def GetMasterNetworkParameters(self):
1222
    """Get network parameters of the master node.
1223

1224
    @rtype: L{object.MasterNetworkParameters}
1225
    @return: network parameters of the master node
1226

1227
    """
1228
    cluster = self._config_data.cluster
1229
    result = objects.MasterNetworkParameters(
1230
      uuid=cluster.master_node, ip=cluster.master_ip,
1231
      netmask=cluster.master_netmask, netdev=cluster.master_netdev,
1232
      ip_family=cluster.primary_ip_family)
1233

    
1234
    return result
1235

    
1236
  @locking.ssynchronized(_config_lock)
1237
  def AddNodeGroup(self, group, ec_id, check_uuid=True):
1238
    """Add a node group to the configuration.
1239

1240
    This method calls group.UpgradeConfig() to fill any missing attributes
1241
    according to their default values.
1242

1243
    @type group: L{objects.NodeGroup}
1244
    @param group: the NodeGroup object to add
1245
    @type ec_id: string
1246
    @param ec_id: unique id for the job to use when creating a missing UUID
1247
    @type check_uuid: bool
1248
    @param check_uuid: add an UUID to the group if it doesn't have one or, if
1249
                       it does, ensure that it does not exist in the
1250
                       configuration already
1251

1252
    """
1253
    self._UnlockedAddNodeGroup(group, ec_id, check_uuid)
1254
    self._WriteConfig()
1255

    
1256
  def _UnlockedAddNodeGroup(self, group, ec_id, check_uuid):
1257
    """Add a node group to the configuration.
1258

1259
    """
1260
    logging.info("Adding node group %s to configuration", group.name)
1261

    
1262
    # Some code might need to add a node group with a pre-populated UUID
1263
    # generated with ConfigWriter.GenerateUniqueID(). We allow them to bypass
1264
    # the "does this UUID" exist already check.
1265
    if check_uuid:
1266
      self._EnsureUUID(group, ec_id)
1267

    
1268
    try:
1269
      existing_uuid = self._UnlockedLookupNodeGroup(group.name)
1270
    except errors.OpPrereqError:
1271
      pass
1272
    else:
1273
      raise errors.OpPrereqError("Desired group name '%s' already exists as a"
1274
                                 " node group (UUID: %s)" %
1275
                                 (group.name, existing_uuid),
1276
                                 errors.ECODE_EXISTS)
1277

    
1278
    group.serial_no = 1
1279
    group.ctime = group.mtime = time.time()
1280
    group.UpgradeConfig()
1281

    
1282
    self._config_data.nodegroups[group.uuid] = group
1283
    self._config_data.cluster.serial_no += 1
1284

    
1285
  @locking.ssynchronized(_config_lock)
1286
  def RemoveNodeGroup(self, group_uuid):
1287
    """Remove a node group from the configuration.
1288

1289
    @type group_uuid: string
1290
    @param group_uuid: the UUID of the node group to remove
1291

1292
    """
1293
    logging.info("Removing node group %s from configuration", group_uuid)
1294

    
1295
    if group_uuid not in self._config_data.nodegroups:
1296
      raise errors.ConfigurationError("Unknown node group '%s'" % group_uuid)
1297

    
1298
    assert len(self._config_data.nodegroups) != 1, \
1299
            "Group '%s' is the only group, cannot be removed" % group_uuid
1300

    
1301
    del self._config_data.nodegroups[group_uuid]
1302
    self._config_data.cluster.serial_no += 1
1303
    self._WriteConfig()
1304

    
1305
  def _UnlockedLookupNodeGroup(self, target):
1306
    """Lookup a node group's UUID.
1307

1308
    @type target: string or None
1309
    @param target: group name or UUID or None to look for the default
1310
    @rtype: string
1311
    @return: nodegroup UUID
1312
    @raises errors.OpPrereqError: when the target group cannot be found
1313

1314
    """
1315
    if target is None:
1316
      if len(self._config_data.nodegroups) != 1:
1317
        raise errors.OpPrereqError("More than one node group exists. Target"
1318
                                   " group must be specified explicitly.")
1319
      else:
1320
        return self._config_data.nodegroups.keys()[0]
1321
    if target in self._config_data.nodegroups:
1322
      return target
1323
    for nodegroup in self._config_data.nodegroups.values():
1324
      if nodegroup.name == target:
1325
        return nodegroup.uuid
1326
    raise errors.OpPrereqError("Node group '%s' not found" % target,
1327
                               errors.ECODE_NOENT)
1328

    
1329
  @locking.ssynchronized(_config_lock, shared=1)
1330
  def LookupNodeGroup(self, target):
1331
    """Lookup a node group's UUID.
1332

1333
    This function is just a wrapper over L{_UnlockedLookupNodeGroup}.
1334

1335
    @type target: string or None
1336
    @param target: group name or UUID or None to look for the default
1337
    @rtype: string
1338
    @return: nodegroup UUID
1339

1340
    """
1341
    return self._UnlockedLookupNodeGroup(target)
1342

    
1343
  def _UnlockedGetNodeGroup(self, uuid):
1344
    """Lookup a node group.
1345

1346
    @type uuid: string
1347
    @param uuid: group UUID
1348
    @rtype: L{objects.NodeGroup} or None
1349
    @return: nodegroup object, or None if not found
1350

1351
    """
1352
    if uuid not in self._config_data.nodegroups:
1353
      return None
1354

    
1355
    return self._config_data.nodegroups[uuid]
1356

    
1357
  @locking.ssynchronized(_config_lock, shared=1)
1358
  def GetNodeGroup(self, uuid):
1359
    """Lookup a node group.
1360

1361
    @type uuid: string
1362
    @param uuid: group UUID
1363
    @rtype: L{objects.NodeGroup} or None
1364
    @return: nodegroup object, or None if not found
1365

1366
    """
1367
    return self._UnlockedGetNodeGroup(uuid)
1368

    
1369
  @locking.ssynchronized(_config_lock, shared=1)
1370
  def GetAllNodeGroupsInfo(self):
1371
    """Get the configuration of all node groups.
1372

1373
    """
1374
    return dict(self._config_data.nodegroups)
1375

    
1376
  @locking.ssynchronized(_config_lock, shared=1)
1377
  def GetNodeGroupList(self):
1378
    """Get a list of node groups.
1379

1380
    """
1381
    return self._config_data.nodegroups.keys()
1382

    
1383
  @locking.ssynchronized(_config_lock, shared=1)
1384
  def GetNodeGroupMembersByNodes(self, nodes):
1385
    """Get nodes which are member in the same nodegroups as the given nodes.
1386

1387
    """
1388
    ngfn = lambda node_uuid: self._UnlockedGetNodeInfo(node_uuid).group
1389
    return frozenset(member_uuid
1390
                     for node_uuid in nodes
1391
                     for member_uuid in
1392
                       self._UnlockedGetNodeGroup(ngfn(node_uuid)).members)
1393

    
1394
  @locking.ssynchronized(_config_lock, shared=1)
1395
  def GetMultiNodeGroupInfo(self, group_uuids):
1396
    """Get the configuration of multiple node groups.
1397

1398
    @param group_uuids: List of node group UUIDs
1399
    @rtype: list
1400
    @return: List of tuples of (group_uuid, group_info)
1401

1402
    """
1403
    return [(uuid, self._UnlockedGetNodeGroup(uuid)) for uuid in group_uuids]
1404

    
1405
  @locking.ssynchronized(_config_lock)
1406
  def AddInstance(self, instance, ec_id):
1407
    """Add an instance to the config.
1408

1409
    This should be used after creating a new instance.
1410

1411
    @type instance: L{objects.Instance}
1412
    @param instance: the instance object
1413

1414
    """
1415
    if not isinstance(instance, objects.Instance):
1416
      raise errors.ProgrammerError("Invalid type passed to AddInstance")
1417

    
1418
    if instance.disk_template != constants.DT_DISKLESS:
1419
      all_lvs = instance.MapLVsByNode()
1420
      logging.info("Instance '%s' DISK_LAYOUT: %s", instance.name, all_lvs)
1421

    
1422
    all_macs = self._AllMACs()
1423
    for nic in instance.nics:
1424
      if nic.mac in all_macs:
1425
        raise errors.ConfigurationError("Cannot add instance %s:"
1426
                                        " MAC address '%s' already in use." %
1427
                                        (instance.name, nic.mac))
1428

    
1429
    self._EnsureUUID(instance, ec_id)
1430

    
1431
    instance.serial_no = 1
1432
    instance.ctime = instance.mtime = time.time()
1433
    self._config_data.instances[instance.name] = instance
1434
    self._config_data.cluster.serial_no += 1
1435
    self._UnlockedReleaseDRBDMinors(instance.name)
1436
    self._UnlockedCommitTemporaryIps(ec_id)
1437
    self._WriteConfig()
1438

    
1439
  def _EnsureUUID(self, item, ec_id):
1440
    """Ensures a given object has a valid UUID.
1441

1442
    @param item: the instance or node to be checked
1443
    @param ec_id: the execution context id for the uuid reservation
1444

1445
    """
1446
    if not item.uuid:
1447
      item.uuid = self._GenerateUniqueID(ec_id)
1448
    elif item.uuid in self._AllIDs(include_temporary=True):
1449
      raise errors.ConfigurationError("Cannot add '%s': UUID %s already"
1450
                                      " in use" % (item.name, item.uuid))
1451

    
1452
  def _SetInstanceStatus(self, instance_name, status, disks_active):
1453
    """Set the instance's status to a given value.
1454

1455
    """
1456
    if instance_name not in self._config_data.instances:
1457
      raise errors.ConfigurationError("Unknown instance '%s'" %
1458
                                      instance_name)
1459
    instance = self._config_data.instances[instance_name]
1460

    
1461
    if status is None:
1462
      status = instance.admin_state
1463
    if disks_active is None:
1464
      disks_active = instance.disks_active
1465

    
1466
    assert status in constants.ADMINST_ALL, \
1467
           "Invalid status '%s' passed to SetInstanceStatus" % (status,)
1468

    
1469
    if instance.admin_state != status or \
1470
       instance.disks_active != disks_active:
1471
      instance.admin_state = status
1472
      instance.disks_active = disks_active
1473
      instance.serial_no += 1
1474
      instance.mtime = time.time()
1475
      self._WriteConfig()
1476

    
1477
  @locking.ssynchronized(_config_lock)
1478
  def MarkInstanceUp(self, instance_name):
1479
    """Mark the instance status to up in the config.
1480

1481
    This also sets the instance disks active flag.
1482

1483
    """
1484
    self._SetInstanceStatus(instance_name, constants.ADMINST_UP, True)
1485

    
1486
  @locking.ssynchronized(_config_lock)
1487
  def MarkInstanceOffline(self, instance_name):
1488
    """Mark the instance status to down in the config.
1489

1490
    This also clears the instance disks active flag.
1491

1492
    """
1493
    self._SetInstanceStatus(instance_name, constants.ADMINST_OFFLINE, False)
1494

    
1495
  @locking.ssynchronized(_config_lock)
1496
  def RemoveInstance(self, instance_name):
1497
    """Remove the instance from the configuration.
1498

1499
    """
1500
    if instance_name not in self._config_data.instances:
1501
      raise errors.ConfigurationError("Unknown instance '%s'" % instance_name)
1502

    
1503
    # If a network port has been allocated to the instance,
1504
    # return it to the pool of free ports.
1505
    inst = self._config_data.instances[instance_name]
1506
    network_port = getattr(inst, "network_port", None)
1507
    if network_port is not None:
1508
      self._config_data.cluster.tcpudp_port_pool.add(network_port)
1509

    
1510
    instance = self._UnlockedGetInstanceInfo(instance_name)
1511

    
1512
    for nic in instance.nics:
1513
      if nic.network and nic.ip:
1514
        # Return all IP addresses to the respective address pools
1515
        self._UnlockedCommitIp(constants.RELEASE_ACTION, nic.network, nic.ip)
1516

    
1517
    del self._config_data.instances[instance_name]
1518
    self._config_data.cluster.serial_no += 1
1519
    self._WriteConfig()
1520

    
1521
  @locking.ssynchronized(_config_lock)
1522
  def RenameInstance(self, old_name, new_name):
1523
    """Rename an instance.
1524

1525
    This needs to be done in ConfigWriter and not by RemoveInstance
1526
    combined with AddInstance as only we can guarantee an atomic
1527
    rename.
1528

1529
    """
1530
    if old_name not in self._config_data.instances:
1531
      raise errors.ConfigurationError("Unknown instance '%s'" % old_name)
1532

    
1533
    # Operate on a copy to not loose instance object in case of a failure
1534
    inst = self._config_data.instances[old_name].Copy()
1535
    inst.name = new_name
1536

    
1537
    for (idx, disk) in enumerate(inst.disks):
1538
      if disk.dev_type == constants.LD_FILE:
1539
        # rename the file paths in logical and physical id
1540
        file_storage_dir = os.path.dirname(os.path.dirname(disk.logical_id[1]))
1541
        disk.logical_id = (disk.logical_id[0],
1542
                           utils.PathJoin(file_storage_dir, inst.name,
1543
                                          "disk%s" % idx))
1544
        disk.physical_id = disk.logical_id
1545

    
1546
    # Actually replace instance object
1547
    del self._config_data.instances[old_name]
1548
    self._config_data.instances[inst.name] = inst
1549

    
1550
    # Force update of ssconf files
1551
    self._config_data.cluster.serial_no += 1
1552

    
1553
    self._WriteConfig()
1554

    
1555
  @locking.ssynchronized(_config_lock)
1556
  def MarkInstanceDown(self, instance_name):
1557
    """Mark the status of an instance to down in the configuration.
1558

1559
    This does not touch the instance disks active flag, as shut down instances
1560
    can still have active disks.
1561

1562
    """
1563
    self._SetInstanceStatus(instance_name, constants.ADMINST_DOWN, None)
1564

    
1565
  @locking.ssynchronized(_config_lock)
1566
  def MarkInstanceDisksActive(self, instance_name):
1567
    """Mark the status of instance disks active.
1568

1569
    """
1570
    self._SetInstanceStatus(instance_name, None, True)
1571

    
1572
  @locking.ssynchronized(_config_lock)
1573
  def MarkInstanceDisksInactive(self, instance_name):
1574
    """Mark the status of instance disks inactive.
1575

1576
    """
1577
    self._SetInstanceStatus(instance_name, None, False)
1578

    
1579
  def _UnlockedGetInstanceList(self):
1580
    """Get the list of instances.
1581

1582
    This function is for internal use, when the config lock is already held.
1583

1584
    """
1585
    return self._config_data.instances.keys()
1586

    
1587
  @locking.ssynchronized(_config_lock, shared=1)
1588
  def GetInstanceList(self):
1589
    """Get the list of instances.
1590

1591
    @return: array of instances, ex. ['instance2.example.com',
1592
        'instance1.example.com']
1593

1594
    """
1595
    return self._UnlockedGetInstanceList()
1596

    
1597
  def ExpandInstanceName(self, short_name):
1598
    """Attempt to expand an incomplete instance name.
1599

1600
    """
1601
    # Locking is done in L{ConfigWriter.GetInstanceList}
1602
    return _MatchNameComponentIgnoreCase(short_name, self.GetInstanceList())
1603

    
1604
  def _UnlockedGetInstanceInfo(self, instance_name):
1605
    """Returns information about an instance.
1606

1607
    This function is for internal use, when the config lock is already held.
1608

1609
    """
1610
    if instance_name not in self._config_data.instances:
1611
      return None
1612

    
1613
    return self._config_data.instances[instance_name]
1614

    
1615
  @locking.ssynchronized(_config_lock, shared=1)
1616
  def GetInstanceInfo(self, instance_name):
1617
    """Returns information about an instance.
1618

1619
    It takes the information from the configuration file. Other information of
1620
    an instance are taken from the live systems.
1621

1622
    @param instance_name: name of the instance, e.g.
1623
        I{instance1.example.com}
1624

1625
    @rtype: L{objects.Instance}
1626
    @return: the instance object
1627

1628
    """
1629
    return self._UnlockedGetInstanceInfo(instance_name)
1630

    
1631
  @locking.ssynchronized(_config_lock, shared=1)
1632
  def GetInstanceNodeGroups(self, instance_name, primary_only=False):
1633
    """Returns set of node group UUIDs for instance's nodes.
1634

1635
    @rtype: frozenset
1636

1637
    """
1638
    instance = self._UnlockedGetInstanceInfo(instance_name)
1639
    if not instance:
1640
      raise errors.ConfigurationError("Unknown instance '%s'" % instance_name)
1641

    
1642
    if primary_only:
1643
      nodes = [instance.primary_node]
1644
    else:
1645
      nodes = instance.all_nodes
1646

    
1647
    return frozenset(self._UnlockedGetNodeInfo(node_uuid).group
1648
                     for node_uuid in nodes)
1649

    
1650
  @locking.ssynchronized(_config_lock, shared=1)
1651
  def GetInstanceNetworks(self, instance_name):
1652
    """Returns set of network UUIDs for instance's nics.
1653

1654
    @rtype: frozenset
1655

1656
    """
1657
    instance = self._UnlockedGetInstanceInfo(instance_name)
1658
    if not instance:
1659
      raise errors.ConfigurationError("Unknown instance '%s'" % instance_name)
1660

    
1661
    networks = set()
1662
    for nic in instance.nics:
1663
      if nic.network:
1664
        networks.add(nic.network)
1665

    
1666
    return frozenset(networks)
1667

    
1668
  @locking.ssynchronized(_config_lock, shared=1)
1669
  def GetMultiInstanceInfo(self, instances):
1670
    """Get the configuration of multiple instances.
1671

1672
    @param instances: list of instance names
1673
    @rtype: list
1674
    @return: list of tuples (instance, instance_info), where
1675
        instance_info is what would GetInstanceInfo return for the
1676
        node, while keeping the original order
1677

1678
    """
1679
    return [(name, self._UnlockedGetInstanceInfo(name)) for name in instances]
1680

    
1681
  @locking.ssynchronized(_config_lock, shared=1)
1682
  def GetAllInstancesInfo(self):
1683
    """Get the configuration of all instances.
1684

1685
    @rtype: dict
1686
    @return: dict of (instance, instance_info), where instance_info is what
1687
              would GetInstanceInfo return for the node
1688

1689
    """
1690
    my_dict = dict([(instance, self._UnlockedGetInstanceInfo(instance))
1691
                    for instance in self._UnlockedGetInstanceList()])
1692
    return my_dict
1693

    
1694
  @locking.ssynchronized(_config_lock, shared=1)
1695
  def GetInstancesInfoByFilter(self, filter_fn):
1696
    """Get instance configuration with a filter.
1697

1698
    @type filter_fn: callable
1699
    @param filter_fn: Filter function receiving instance object as parameter,
1700
      returning boolean. Important: this function is called while the
1701
      configuration locks is held. It must not do any complex work or call
1702
      functions potentially leading to a deadlock. Ideally it doesn't call any
1703
      other functions and just compares instance attributes.
1704

1705
    """
1706
    return dict((name, inst)
1707
                for (name, inst) in self._config_data.instances.items()
1708
                if filter_fn(inst))
1709

    
1710
  @locking.ssynchronized(_config_lock)
1711
  def AddNode(self, node, ec_id):
1712
    """Add a node to the configuration.
1713

1714
    @type node: L{objects.Node}
1715
    @param node: a Node instance
1716

1717
    """
1718
    logging.info("Adding node %s to configuration", node.name)
1719

    
1720
    self._EnsureUUID(node, ec_id)
1721

    
1722
    node.serial_no = 1
1723
    node.ctime = node.mtime = time.time()
1724
    self._UnlockedAddNodeToGroup(node.uuid, node.group)
1725
    self._config_data.nodes[node.uuid] = node
1726
    self._config_data.cluster.serial_no += 1
1727
    self._WriteConfig()
1728

    
1729
  @locking.ssynchronized(_config_lock)
1730
  def RemoveNode(self, node_uuid):
1731
    """Remove a node from the configuration.
1732

1733
    """
1734
    logging.info("Removing node %s from configuration", node_uuid)
1735

    
1736
    if node_uuid not in self._config_data.nodes:
1737
      raise errors.ConfigurationError("Unknown node '%s'" % node_uuid)
1738

    
1739
    self._UnlockedRemoveNodeFromGroup(self._config_data.nodes[node_uuid])
1740
    del self._config_data.nodes[node_uuid]
1741
    self._config_data.cluster.serial_no += 1
1742
    self._WriteConfig()
1743

    
1744
  def ExpandNodeName(self, short_name):
1745
    """Attempt to expand an incomplete node name into a node UUID.
1746

1747
    """
1748
    # Locking is done in L{ConfigWriter.GetAllNodesInfo}
1749
    all_nodes = self.GetAllNodesInfo().values()
1750
    expanded_name = _MatchNameComponentIgnoreCase(
1751
                      short_name, [node.name for node in all_nodes])
1752

    
1753
    if expanded_name is not None:
1754
      # there has to be exactly one node whith that name
1755
      node = (filter(lambda n: n.name == expanded_name, all_nodes)[0])
1756
      return (node.uuid, node.name)
1757
    else:
1758
      return None
1759

    
1760
  def _UnlockedGetNodeInfo(self, node_uuid):
1761
    """Get the configuration of a node, as stored in the config.
1762

1763
    This function is for internal use, when the config lock is already
1764
    held.
1765

1766
    @param node_uuid: the node UUID
1767

1768
    @rtype: L{objects.Node}
1769
    @return: the node object
1770

1771
    """
1772
    if node_uuid not in self._config_data.nodes:
1773
      return None
1774

    
1775
    return self._config_data.nodes[node_uuid]
1776

    
1777
  @locking.ssynchronized(_config_lock, shared=1)
1778
  def GetNodeInfo(self, node_uuid):
1779
    """Get the configuration of a node, as stored in the config.
1780

1781
    This is just a locked wrapper over L{_UnlockedGetNodeInfo}.
1782

1783
    @param node_uuid: the node UUID
1784

1785
    @rtype: L{objects.Node}
1786
    @return: the node object
1787

1788
    """
1789
    return self._UnlockedGetNodeInfo(node_uuid)
1790

    
1791
  @locking.ssynchronized(_config_lock, shared=1)
1792
  def GetNodeInstances(self, node_uuid):
1793
    """Get the instances of a node, as stored in the config.
1794

1795
    @param node_uuid: the node UUID
1796

1797
    @rtype: (list, list)
1798
    @return: a tuple with two lists: the primary and the secondary instances
1799

1800
    """
1801
    pri = []
1802
    sec = []
1803
    for inst in self._config_data.instances.values():
1804
      if inst.primary_node == node_uuid:
1805
        pri.append(inst.name)
1806
      if node_uuid in inst.secondary_nodes:
1807
        sec.append(inst.name)
1808
    return (pri, sec)
1809

    
1810
  @locking.ssynchronized(_config_lock, shared=1)
1811
  def GetNodeGroupInstances(self, uuid, primary_only=False):
1812
    """Get the instances of a node group.
1813

1814
    @param uuid: Node group UUID
1815
    @param primary_only: Whether to only consider primary nodes
1816
    @rtype: frozenset
1817
    @return: List of instance names in node group
1818

1819
    """
1820
    if primary_only:
1821
      nodes_fn = lambda inst: [inst.primary_node]
1822
    else:
1823
      nodes_fn = lambda inst: inst.all_nodes
1824

    
1825
    return frozenset(inst.name
1826
                     for inst in self._config_data.instances.values()
1827
                     for node_uuid in nodes_fn(inst)
1828
                     if self._UnlockedGetNodeInfo(node_uuid).group == uuid)
1829

    
1830
  def _UnlockedGetHvparamsString(self, hvname):
1831
    """Return the string representation of the list of hyervisor parameters of
1832
    the given hypervisor.
1833

1834
    @see: C{GetHvparams}
1835

1836
    """
1837
    result = ""
1838
    hvparams = self._config_data.cluster.hvparams[hvname]
1839
    for key in hvparams:
1840
      result += "%s=%s\n" % (key, hvparams[key])
1841
    return result
1842

    
1843
  @locking.ssynchronized(_config_lock, shared=1)
1844
  def GetHvparamsString(self, hvname):
1845
    """Return the hypervisor parameters of the given hypervisor.
1846

1847
    @type hvname: string
1848
    @param hvname: name of a hypervisor
1849
    @rtype: string
1850
    @return: string containing key-value-pairs, one pair on each line;
1851
      format: KEY=VALUE
1852

1853
    """
1854
    return self._UnlockedGetHvparamsString(hvname)
1855

    
1856
  def _UnlockedGetNodeList(self):
1857
    """Return the list of nodes which are in the configuration.
1858

1859
    This function is for internal use, when the config lock is already
1860
    held.
1861

1862
    @rtype: list
1863

1864
    """
1865
    return self._config_data.nodes.keys()
1866

    
1867
  @locking.ssynchronized(_config_lock, shared=1)
1868
  def GetNodeList(self):
1869
    """Return the list of nodes which are in the configuration.
1870

1871
    """
1872
    return self._UnlockedGetNodeList()
1873

    
1874
  def _UnlockedGetOnlineNodeList(self):
1875
    """Return the list of nodes which are online.
1876

1877
    """
1878
    all_nodes = [self._UnlockedGetNodeInfo(node)
1879
                 for node in self._UnlockedGetNodeList()]
1880
    return [node.uuid for node in all_nodes if not node.offline]
1881

    
1882
  @locking.ssynchronized(_config_lock, shared=1)
1883
  def GetOnlineNodeList(self):
1884
    """Return the list of nodes which are online.
1885

1886
    """
1887
    return self._UnlockedGetOnlineNodeList()
1888

    
1889
  @locking.ssynchronized(_config_lock, shared=1)
1890
  def GetVmCapableNodeList(self):
1891
    """Return the list of nodes which are not vm capable.
1892

1893
    """
1894
    all_nodes = [self._UnlockedGetNodeInfo(node)
1895
                 for node in self._UnlockedGetNodeList()]
1896
    return [node.uuid for node in all_nodes if node.vm_capable]
1897

    
1898
  @locking.ssynchronized(_config_lock, shared=1)
1899
  def GetNonVmCapableNodeList(self):
1900
    """Return the list of nodes which are not vm capable.
1901

1902
    """
1903
    all_nodes = [self._UnlockedGetNodeInfo(node)
1904
                 for node in self._UnlockedGetNodeList()]
1905
    return [node.uuid for node in all_nodes if not node.vm_capable]
1906

    
1907
  @locking.ssynchronized(_config_lock, shared=1)
1908
  def GetMultiNodeInfo(self, node_uuids):
1909
    """Get the configuration of multiple nodes.
1910

1911
    @param node_uuids: list of node UUIDs
1912
    @rtype: list
1913
    @return: list of tuples of (node, node_info), where node_info is
1914
        what would GetNodeInfo return for the node, in the original
1915
        order
1916

1917
    """
1918
    return [(uuid, self._UnlockedGetNodeInfo(uuid)) for uuid in node_uuids]
1919

    
1920
  def _UnlockedGetAllNodesInfo(self):
1921
    """Gets configuration of all nodes.
1922

1923
    @note: See L{GetAllNodesInfo}
1924

1925
    """
1926
    return dict([(node_uuid, self._UnlockedGetNodeInfo(node_uuid))
1927
                 for node_uuid in self._UnlockedGetNodeList()])
1928

    
1929
  @locking.ssynchronized(_config_lock, shared=1)
1930
  def GetAllNodesInfo(self):
1931
    """Get the configuration of all nodes.
1932

1933
    @rtype: dict
1934
    @return: dict of (node, node_info), where node_info is what
1935
              would GetNodeInfo return for the node
1936

1937
    """
1938
    return self._UnlockedGetAllNodesInfo()
1939

    
1940
  def _UnlockedGetNodeInfoByName(self, node_name):
1941
    for node in self._UnlockedGetAllNodesInfo().values():
1942
      if node.name == node_name:
1943
        return node
1944
    return None
1945

    
1946
  @locking.ssynchronized(_config_lock, shared=1)
1947
  def GetNodeInfoByName(self, node_name):
1948
    """Get the L{objects.Node} object for a named node.
1949

1950
    @param node_name: name of the node to get information for
1951
    @type node_name: string
1952
    @return: the corresponding L{objects.Node} instance or None if no
1953
          information is available
1954

1955
    """
1956
    return self._UnlockedGetNodeInfoByName(node_name)
1957

    
1958
  def _UnlockedGetNodeName(self, node_spec):
1959
    if isinstance(node_spec, objects.Node):
1960
      return node_spec.name
1961
    elif isinstance(node_spec, basestring):
1962
      node_info = self._UnlockedGetNodeInfo(node_spec)
1963
      if node_info is None:
1964
        raise errors.OpExecError("Unknown node: %s" % node_spec)
1965
      return node_info.name
1966
    else:
1967
      raise errors.ProgrammerError("Can't handle node spec '%s'" % node_spec)
1968

    
1969
  @locking.ssynchronized(_config_lock, shared=1)
1970
  def GetNodeName(self, node_spec):
1971
    """Gets the node name for the passed node.
1972

1973
    @param node_spec: node to get names for
1974
    @type node_spec: either node UUID or a L{objects.Node} object
1975
    @rtype: string
1976
    @return: node name
1977

1978
    """
1979
    return self._UnlockedGetNodeName(node_spec)
1980

    
1981
  def _UnlockedGetNodeNames(self, node_specs):
1982
    return [self._UnlockedGetNodeName(node_spec) for node_spec in node_specs]
1983

    
1984
  @locking.ssynchronized(_config_lock, shared=1)
1985
  def GetNodeNames(self, node_specs):
1986
    """Gets the node names for the passed list of nodes.
1987

1988
    @param node_specs: list of nodes to get names for
1989
    @type node_specs: list of either node UUIDs or L{objects.Node} objects
1990
    @rtype: list of strings
1991
    @return: list of node names
1992

1993
    """
1994
    return self._UnlockedGetNodeNames(node_specs)
1995

    
1996
  @locking.ssynchronized(_config_lock, shared=1)
1997
  def GetNodeGroupsFromNodes(self, node_uuids):
1998
    """Returns groups for a list of nodes.
1999

2000
    @type node_uuids: list of string
2001
    @param node_uuids: List of node UUIDs
2002
    @rtype: frozenset
2003

2004
    """
2005
    return frozenset(self._UnlockedGetNodeInfo(uuid).group
2006
                     for uuid in node_uuids)
2007

    
2008
  def _UnlockedGetMasterCandidateStats(self, exceptions=None):
2009
    """Get the number of current and maximum desired and possible candidates.
2010

2011
    @type exceptions: list
2012
    @param exceptions: if passed, list of nodes that should be ignored
2013
    @rtype: tuple
2014
    @return: tuple of (current, desired and possible, possible)
2015

2016
    """
2017
    mc_now = mc_should = mc_max = 0
2018
    for node in self._config_data.nodes.values():
2019
      if exceptions and node.uuid in exceptions:
2020
        continue
2021
      if not (node.offline or node.drained) and node.master_capable:
2022
        mc_max += 1
2023
      if node.master_candidate:
2024
        mc_now += 1
2025
    mc_should = min(mc_max, self._config_data.cluster.candidate_pool_size)
2026
    return (mc_now, mc_should, mc_max)
2027

    
2028
  @locking.ssynchronized(_config_lock, shared=1)
2029
  def GetMasterCandidateStats(self, exceptions=None):
2030
    """Get the number of current and maximum possible candidates.
2031

2032
    This is just a wrapper over L{_UnlockedGetMasterCandidateStats}.
2033

2034
    @type exceptions: list
2035
    @param exceptions: if passed, list of nodes that should be ignored
2036
    @rtype: tuple
2037
    @return: tuple of (current, max)
2038

2039
    """
2040
    return self._UnlockedGetMasterCandidateStats(exceptions)
2041

    
2042
  @locking.ssynchronized(_config_lock)
2043
  def MaintainCandidatePool(self, exception_node_uuids):
2044
    """Try to grow the candidate pool to the desired size.
2045

2046
    @type exception_node_uuids: list
2047
    @param exception_node_uuids: if passed, list of nodes that should be ignored
2048
    @rtype: list
2049
    @return: list with the adjusted nodes (L{objects.Node} instances)
2050

2051
    """
2052
    mc_now, mc_max, _ = self._UnlockedGetMasterCandidateStats(
2053
                          exception_node_uuids)
2054
    mod_list = []
2055
    if mc_now < mc_max:
2056
      node_list = self._config_data.nodes.keys()
2057
      random.shuffle(node_list)
2058
      for uuid in node_list:
2059
        if mc_now >= mc_max:
2060
          break
2061
        node = self._config_data.nodes[uuid]
2062
        if (node.master_candidate or node.offline or node.drained or
2063
            node.uuid in exception_node_uuids or not node.master_capable):
2064
          continue
2065
        mod_list.append(node)
2066
        node.master_candidate = True
2067
        node.serial_no += 1
2068
        mc_now += 1
2069
      if mc_now != mc_max:
2070
        # this should not happen
2071
        logging.warning("Warning: MaintainCandidatePool didn't manage to"
2072
                        " fill the candidate pool (%d/%d)", mc_now, mc_max)
2073
      if mod_list:
2074
        self._config_data.cluster.serial_no += 1
2075
        self._WriteConfig()
2076

    
2077
    return mod_list
2078

    
2079
  def _UnlockedAddNodeToGroup(self, node_uuid, nodegroup_uuid):
2080
    """Add a given node to the specified group.
2081

2082
    """
2083
    if nodegroup_uuid not in self._config_data.nodegroups:
2084
      # This can happen if a node group gets deleted between its lookup and
2085
      # when we're adding the first node to it, since we don't keep a lock in
2086
      # the meantime. It's ok though, as we'll fail cleanly if the node group
2087
      # is not found anymore.
2088
      raise errors.OpExecError("Unknown node group: %s" % nodegroup_uuid)
2089
    if node_uuid not in self._config_data.nodegroups[nodegroup_uuid].members:
2090
      self._config_data.nodegroups[nodegroup_uuid].members.append(node_uuid)
2091

    
2092
  def _UnlockedRemoveNodeFromGroup(self, node):
2093
    """Remove a given node from its group.
2094

2095
    """
2096
    nodegroup = node.group
2097
    if nodegroup not in self._config_data.nodegroups:
2098
      logging.warning("Warning: node '%s' has unknown node group '%s'"
2099
                      " (while being removed from it)", node.uuid, nodegroup)
2100
    nodegroup_obj = self._config_data.nodegroups[nodegroup]
2101
    if node.uuid not in nodegroup_obj.members:
2102
      logging.warning("Warning: node '%s' not a member of its node group '%s'"
2103
                      " (while being removed from it)", node.uuid, nodegroup)
2104
    else:
2105
      nodegroup_obj.members.remove(node.uuid)
2106

    
2107
  @locking.ssynchronized(_config_lock)
2108
  def AssignGroupNodes(self, mods):
2109
    """Changes the group of a number of nodes.
2110

2111
    @type mods: list of tuples; (node name, new group UUID)
2112
    @param mods: Node membership modifications
2113

2114
    """
2115
    groups = self._config_data.nodegroups
2116
    nodes = self._config_data.nodes
2117

    
2118
    resmod = []
2119

    
2120
    # Try to resolve UUIDs first
2121
    for (node_uuid, new_group_uuid) in mods:
2122
      try:
2123
        node = nodes[node_uuid]
2124
      except KeyError:
2125
        raise errors.ConfigurationError("Unable to find node '%s'" % node_uuid)
2126

    
2127
      if node.group == new_group_uuid:
2128
        # Node is being assigned to its current group
2129
        logging.debug("Node '%s' was assigned to its current group (%s)",
2130
                      node_uuid, node.group)
2131
        continue
2132

    
2133
      # Try to find current group of node
2134
      try:
2135
        old_group = groups[node.group]
2136
      except KeyError:
2137
        raise errors.ConfigurationError("Unable to find old group '%s'" %
2138
                                        node.group)
2139

    
2140
      # Try to find new group for node
2141
      try:
2142
        new_group = groups[new_group_uuid]
2143
      except KeyError:
2144
        raise errors.ConfigurationError("Unable to find new group '%s'" %
2145
                                        new_group_uuid)
2146

    
2147
      assert node.uuid in old_group.members, \
2148
        ("Inconsistent configuration: node '%s' not listed in members for its"
2149
         " old group '%s'" % (node.uuid, old_group.uuid))
2150
      assert node.uuid not in new_group.members, \
2151
        ("Inconsistent configuration: node '%s' already listed in members for"
2152
         " its new group '%s'" % (node.uuid, new_group.uuid))
2153

    
2154
      resmod.append((node, old_group, new_group))
2155

    
2156
    # Apply changes
2157
    for (node, old_group, new_group) in resmod:
2158
      assert node.uuid != new_group.uuid and old_group.uuid != new_group.uuid, \
2159
        "Assigning to current group is not possible"
2160

    
2161
      node.group = new_group.uuid
2162

    
2163
      # Update members of involved groups
2164
      if node.uuid in old_group.members:
2165
        old_group.members.remove(node.uuid)
2166
      if node.uuid not in new_group.members:
2167
        new_group.members.append(node.uuid)
2168

    
2169
    # Update timestamps and serials (only once per node/group object)
2170
    now = time.time()
2171
    for obj in frozenset(itertools.chain(*resmod)): # pylint: disable=W0142
2172
      obj.serial_no += 1
2173
      obj.mtime = now
2174

    
2175
    # Force ssconf update
2176
    self._config_data.cluster.serial_no += 1
2177

    
2178
    self._WriteConfig()
2179

    
2180
  def _BumpSerialNo(self):
2181
    """Bump up the serial number of the config.
2182

2183
    """
2184
    self._config_data.serial_no += 1
2185
    self._config_data.mtime = time.time()
2186

    
2187
  def _AllUUIDObjects(self):
2188
    """Returns all objects with uuid attributes.
2189

2190
    """
2191
    return (self._config_data.instances.values() +
2192
            self._config_data.nodes.values() +
2193
            self._config_data.nodegroups.values() +
2194
            self._config_data.networks.values() +
2195
            self._AllDisks() +
2196
            self._AllNICs() +
2197
            [self._config_data.cluster])
2198

    
2199
  def _OpenConfig(self, accept_foreign):
2200
    """Read the config data from disk.
2201

2202
    """
2203
    raw_data = utils.ReadFile(self._cfg_file)
2204

    
2205
    try:
2206
      data = objects.ConfigData.FromDict(serializer.Load(raw_data))
2207
    except Exception, err:
2208
      raise errors.ConfigurationError(err)
2209

    
2210
    # Make sure the configuration has the right version
2211
    _ValidateConfig(data)
2212

    
2213
    if (not hasattr(data, "cluster") or
2214
        not hasattr(data.cluster, "rsahostkeypub")):
2215
      raise errors.ConfigurationError("Incomplete configuration"
2216
                                      " (missing cluster.rsahostkeypub)")
2217

    
2218
    if not data.cluster.master_node in data.nodes:
2219
      msg = ("The configuration denotes node %s as master, but does not"
2220
             " contain information about this node" %
2221
             data.cluster.master_node)
2222
      raise errors.ConfigurationError(msg)
2223

    
2224
    master_info = data.nodes[data.cluster.master_node]
2225
    if master_info.name != self._my_hostname and not accept_foreign:
2226
      msg = ("The configuration denotes node %s as master, while my"
2227
             " hostname is %s; opening a foreign configuration is only"
2228
             " possible in accept_foreign mode" %
2229
             (master_info.name, self._my_hostname))
2230
      raise errors.ConfigurationError(msg)
2231

    
2232
    self._config_data = data
2233
    # reset the last serial as -1 so that the next write will cause
2234
    # ssconf update
2235
    self._last_cluster_serial = -1
2236

    
2237
    # Upgrade configuration if needed
2238
    self._UpgradeConfig()
2239

    
2240
    self._cfg_id = utils.GetFileID(path=self._cfg_file)
2241

    
2242
  def _UpgradeConfig(self):
2243
    """Run any upgrade steps.
2244

2245
    This method performs both in-object upgrades and also update some data
2246
    elements that need uniqueness across the whole configuration or interact
2247
    with other objects.
2248

2249
    @warning: this function will call L{_WriteConfig()}, but also
2250
        L{DropECReservations} so it needs to be called only from a
2251
        "safe" place (the constructor). If one wanted to call it with
2252
        the lock held, a DropECReservationUnlocked would need to be
2253
        created first, to avoid causing deadlock.
2254

2255
    """
2256
    # Keep a copy of the persistent part of _config_data to check for changes
2257
    # Serialization doesn't guarantee order in dictionaries
2258
    oldconf = copy.deepcopy(self._config_data.ToDict())
2259

    
2260
    # In-object upgrades
2261
    self._config_data.UpgradeConfig()
2262

    
2263
    for item in self._AllUUIDObjects():
2264
      if item.uuid is None:
2265
        item.uuid = self._GenerateUniqueID(_UPGRADE_CONFIG_JID)
2266
    if not self._config_data.nodegroups:
2267
      default_nodegroup_name = constants.INITIAL_NODE_GROUP_NAME
2268
      default_nodegroup = objects.NodeGroup(name=default_nodegroup_name,
2269
                                            members=[])
2270
      self._UnlockedAddNodeGroup(default_nodegroup, _UPGRADE_CONFIG_JID, True)
2271
    for node in self._config_data.nodes.values():
2272
      if not node.group:
2273
        node.group = self.LookupNodeGroup(None)
2274
      # This is technically *not* an upgrade, but needs to be done both when
2275
      # nodegroups are being added, and upon normally loading the config,
2276
      # because the members list of a node group is discarded upon
2277
      # serializing/deserializing the object.
2278
      self._UnlockedAddNodeToGroup(node.uuid, node.group)
2279

    
2280
    modified = (oldconf != self._config_data.ToDict())
2281
    if modified:
2282
      self._WriteConfig()
2283
      # This is ok even if it acquires the internal lock, as _UpgradeConfig is
2284
      # only called at config init time, without the lock held
2285
      self.DropECReservations(_UPGRADE_CONFIG_JID)
2286
    else:
2287
      config_errors = self._UnlockedVerifyConfig()
2288
      if config_errors:
2289
        errmsg = ("Loaded configuration data is not consistent: %s" %
2290
                  (utils.CommaJoin(config_errors)))
2291
        logging.critical(errmsg)
2292

    
2293
  def _DistributeConfig(self, feedback_fn):
2294
    """Distribute the configuration to the other nodes.
2295

2296
    Currently, this only copies the configuration file. In the future,
2297
    it could be used to encapsulate the 2/3-phase update mechanism.
2298

2299
    """
2300
    if self._offline:
2301
      return True
2302

    
2303
    bad = False
2304

    
2305
    node_list = []
2306
    addr_list = []
2307
    myhostname = self._my_hostname
2308
    # we can skip checking whether _UnlockedGetNodeInfo returns None
2309
    # since the node list comes from _UnlocketGetNodeList, and we are
2310
    # called with the lock held, so no modifications should take place
2311
    # in between
2312
    for node_uuid in self._UnlockedGetNodeList():
2313
      node_info = self._UnlockedGetNodeInfo(node_uuid)
2314
      if node_info.name == myhostname or not node_info.master_candidate:
2315
        continue
2316
      node_list.append(node_info.name)
2317
      addr_list.append(node_info.primary_ip)
2318

    
2319
    # TODO: Use dedicated resolver talking to config writer for name resolution
2320
    result = \
2321
      self._GetRpc(addr_list).call_upload_file(node_list, self._cfg_file)
2322
    for to_node, to_result in result.items():
2323
      msg = to_result.fail_msg
2324
      if msg:
2325
        msg = ("Copy of file %s to node %s failed: %s" %
2326
               (self._cfg_file, to_node, msg))
2327
        logging.error(msg)
2328

    
2329
        if feedback_fn:
2330
          feedback_fn(msg)
2331

    
2332
        bad = True
2333

    
2334
    return not bad
2335

    
2336
  def _WriteConfig(self, destination=None, feedback_fn=None):
2337
    """Write the configuration data to persistent storage.
2338

2339
    """
2340
    assert feedback_fn is None or callable(feedback_fn)
2341

    
2342
    # Warn on config errors, but don't abort the save - the
2343
    # configuration has already been modified, and we can't revert;
2344
    # the best we can do is to warn the user and save as is, leaving
2345
    # recovery to the user
2346
    config_errors = self._UnlockedVerifyConfig()
2347
    if config_errors:
2348
      errmsg = ("Configuration data is not consistent: %s" %
2349
                (utils.CommaJoin(config_errors)))
2350
      logging.critical(errmsg)
2351
      if feedback_fn:
2352
        feedback_fn(errmsg)
2353

    
2354
    if destination is None:
2355
      destination = self._cfg_file
2356
    self._BumpSerialNo()
2357
    txt = serializer.Dump(self._config_data.ToDict())
2358

    
2359
    getents = self._getents()
2360
    try:
2361
      fd = utils.SafeWriteFile(destination, self._cfg_id, data=txt,
2362
                               close=False, gid=getents.confd_gid, mode=0640)
2363
    except errors.LockError:
2364
      raise errors.ConfigurationError("The configuration file has been"
2365
                                      " modified since the last write, cannot"
2366
                                      " update")
2367
    try:
2368
      self._cfg_id = utils.GetFileID(fd=fd)
2369
    finally:
2370
      os.close(fd)
2371

    
2372
    self.write_count += 1
2373

    
2374
    # and redistribute the config file to master candidates
2375
    self._DistributeConfig(feedback_fn)
2376

    
2377
    # Write ssconf files on all nodes (including locally)
2378
    if self._last_cluster_serial < self._config_data.cluster.serial_no:
2379
      if not self._offline:
2380
        result = self._GetRpc(None).call_write_ssconf_files(
2381
          self._UnlockedGetNodeNames(self._UnlockedGetOnlineNodeList()),
2382
          self._UnlockedGetSsconfValues())
2383

    
2384
        for nname, nresu in result.items():
2385
          msg = nresu.fail_msg
2386
          if msg:
2387
            errmsg = ("Error while uploading ssconf files to"
2388
                      " node %s: %s" % (nname, msg))
2389
            logging.warning(errmsg)
2390

    
2391
            if feedback_fn:
2392
              feedback_fn(errmsg)
2393

    
2394
      self._last_cluster_serial = self._config_data.cluster.serial_no
2395

    
2396
  def _GetAllHvparamsStrings(self, hypervisors):
2397
    """Get the hvparams of all given hypervisors from the config.
2398

2399
    @type hypervisors: list of string
2400
    @param hypervisors: list of hypervisor names
2401
    @rtype: dict of strings
2402
    @returns: dictionary mapping the hypervisor name to a string representation
2403
      of the hypervisor's hvparams
2404

2405
    """
2406
    hvparams = {}
2407
    for hv in hypervisors:
2408
      hvparams[hv] = self._UnlockedGetHvparamsString(hv)
2409
    return hvparams
2410

    
2411
  @staticmethod
2412
  def _ExtendByAllHvparamsStrings(ssconf_values, all_hvparams):
2413
    """Extends the ssconf_values dictionary by hvparams.
2414

2415
    @type ssconf_values: dict of strings
2416
    @param ssconf_values: dictionary mapping ssconf_keys to strings
2417
      representing the content of ssconf files
2418
    @type all_hvparams: dict of strings
2419
    @param all_hvparams: dictionary mapping hypervisor names to a string
2420
      representation of their hvparams
2421
    @rtype: same as ssconf_values
2422
    @returns: the ssconf_values dictionary extended by hvparams
2423

2424
    """
2425
    for hv in all_hvparams:
2426
      ssconf_key = constants.SS_HVPARAMS_PREF + hv
2427
      ssconf_values[ssconf_key] = all_hvparams[hv]
2428
    return ssconf_values
2429

    
2430
  def _UnlockedGetSsconfValues(self):
2431
    """Return the values needed by ssconf.
2432

2433
    @rtype: dict
2434
    @return: a dictionary with keys the ssconf names and values their
2435
        associated value
2436

2437
    """
2438
    fn = "\n".join
2439
    instance_names = utils.NiceSort(self._UnlockedGetInstanceList())
2440
    node_infos = self._UnlockedGetAllNodesInfo().values()
2441
    node_names = [node.name for node in node_infos]
2442
    node_pri_ips = ["%s %s" % (ninfo.name, ninfo.primary_ip)
2443
                    for ninfo in node_infos]
2444
    node_snd_ips = ["%s %s" % (ninfo.name, ninfo.secondary_ip)
2445
                    for ninfo in node_infos]
2446

    
2447
    instance_data = fn(instance_names)
2448
    off_data = fn(node.name for node in node_infos if node.offline)
2449
    on_data = fn(node.name for node in node_infos if not node.offline)
2450
    mc_data = fn(node.name for node in node_infos if node.master_candidate)
2451
    mc_ips_data = fn(node.primary_ip for node in node_infos
2452
                     if node.master_candidate)
2453
    node_data = fn(node_names)
2454
    node_pri_ips_data = fn(node_pri_ips)
2455
    node_snd_ips_data = fn(node_snd_ips)
2456

    
2457
    cluster = self._config_data.cluster
2458
    cluster_tags = fn(cluster.GetTags())
2459

    
2460
    hypervisor_list = fn(cluster.enabled_hypervisors)
2461
    all_hvparams = self._GetAllHvparamsStrings(constants.HYPER_TYPES)
2462

    
2463
    uid_pool = uidpool.FormatUidPool(cluster.uid_pool, separator="\n")
2464

    
2465
    nodegroups = ["%s %s" % (nodegroup.uuid, nodegroup.name) for nodegroup in
2466
                  self._config_data.nodegroups.values()]
2467
    nodegroups_data = fn(utils.NiceSort(nodegroups))
2468
    networks = ["%s %s" % (net.uuid, net.name) for net in
2469
                self._config_data.networks.values()]
2470
    networks_data = fn(utils.NiceSort(networks))
2471

    
2472
    ssconf_values = {
2473
      constants.SS_CLUSTER_NAME: cluster.cluster_name,
2474
      constants.SS_CLUSTER_TAGS: cluster_tags,
2475
      constants.SS_FILE_STORAGE_DIR: cluster.file_storage_dir,
2476
      constants.SS_SHARED_FILE_STORAGE_DIR: cluster.shared_file_storage_dir,
2477
      constants.SS_MASTER_CANDIDATES: mc_data,
2478
      constants.SS_MASTER_CANDIDATES_IPS: mc_ips_data,
2479
      constants.SS_MASTER_IP: cluster.master_ip,
2480
      constants.SS_MASTER_NETDEV: cluster.master_netdev,
2481
      constants.SS_MASTER_NETMASK: str(cluster.master_netmask),
2482
      constants.SS_MASTER_NODE: self._UnlockedGetNodeName(cluster.master_node),
2483
      constants.SS_NODE_LIST: node_data,
2484
      constants.SS_NODE_PRIMARY_IPS: node_pri_ips_data,
2485
      constants.SS_NODE_SECONDARY_IPS: node_snd_ips_data,
2486
      constants.SS_OFFLINE_NODES: off_data,
2487
      constants.SS_ONLINE_NODES: on_data,
2488
      constants.SS_PRIMARY_IP_FAMILY: str(cluster.primary_ip_family),
2489
      constants.SS_INSTANCE_LIST: instance_data,
2490
      constants.SS_RELEASE_VERSION: constants.RELEASE_VERSION,
2491
      constants.SS_HYPERVISOR_LIST: hypervisor_list,
2492
      constants.SS_MAINTAIN_NODE_HEALTH: str(cluster.maintain_node_health),
2493
      constants.SS_UID_POOL: uid_pool,
2494
      constants.SS_NODEGROUPS: nodegroups_data,
2495
      constants.SS_NETWORKS: networks_data,
2496
      }
2497
    ssconf_values = self._ExtendByAllHvparamsStrings(ssconf_values,
2498
                                                     all_hvparams)
2499
    bad_values = [(k, v) for k, v in ssconf_values.items()
2500
                  if not isinstance(v, (str, basestring))]
2501
    if bad_values:
2502
      err = utils.CommaJoin("%s=%s" % (k, v) for k, v in bad_values)
2503
      raise errors.ConfigurationError("Some ssconf key(s) have non-string"
2504
                                      " values: %s" % err)
2505
    return ssconf_values
2506

    
2507
  @locking.ssynchronized(_config_lock, shared=1)
2508
  def GetSsconfValues(self):
2509
    """Wrapper using lock around _UnlockedGetSsconf().
2510

2511
    """
2512
    return self._UnlockedGetSsconfValues()
2513

    
2514
  @locking.ssynchronized(_config_lock, shared=1)
2515
  def GetVGName(self):
2516
    """Return the volume group name.
2517

2518
    """
2519
    return self._config_data.cluster.volume_group_name
2520

    
2521
  @locking.ssynchronized(_config_lock)
2522
  def SetVGName(self, vg_name):
2523
    """Set the volume group name.
2524

2525
    """
2526
    self._config_data.cluster.volume_group_name = vg_name
2527
    self._config_data.cluster.serial_no += 1
2528
    self._WriteConfig()
2529

    
2530
  @locking.ssynchronized(_config_lock, shared=1)
2531
  def GetDRBDHelper(self):
2532
    """Return DRBD usermode helper.
2533

2534
    """
2535
    return self._config_data.cluster.drbd_usermode_helper
2536

    
2537
  @locking.ssynchronized(_config_lock)
2538
  def SetDRBDHelper(self, drbd_helper):
2539
    """Set DRBD usermode helper.
2540

2541
    """
2542
    self._config_data.cluster.drbd_usermode_helper = drbd_helper
2543
    self._config_data.cluster.serial_no += 1
2544
    self._WriteConfig()
2545

    
2546
  @locking.ssynchronized(_config_lock, shared=1)
2547
  def GetMACPrefix(self):
2548
    """Return the mac prefix.
2549

2550
    """
2551
    return self._config_data.cluster.mac_prefix
2552

    
2553
  @locking.ssynchronized(_config_lock, shared=1)
2554
  def GetClusterInfo(self):
2555
    """Returns information about the cluster
2556

2557
    @rtype: L{objects.Cluster}
2558
    @return: the cluster object
2559

2560
    """
2561
    return self._config_data.cluster
2562

    
2563
  @locking.ssynchronized(_config_lock, shared=1)
2564
  def HasAnyDiskOfType(self, dev_type):
2565
    """Check if in there is at disk of the given type in the configuration.
2566

2567
    """
2568
    return self._config_data.HasAnyDiskOfType(dev_type)
2569

    
2570
  @locking.ssynchronized(_config_lock)
2571
  def Update(self, target, feedback_fn, ec_id=None):
2572
    """Notify function to be called after updates.
2573

2574
    This function must be called when an object (as returned by
2575
    GetInstanceInfo, GetNodeInfo, GetCluster) has been updated and the
2576
    caller wants the modifications saved to the backing store. Note
2577
    that all modified objects will be saved, but the target argument
2578
    is the one the caller wants to ensure that it's saved.
2579

2580
    @param target: an instance of either L{objects.Cluster},
2581
        L{objects.Node} or L{objects.Instance} which is existing in
2582
        the cluster
2583
    @param feedback_fn: Callable feedback function
2584

2585
    """
2586
    if self._config_data is None:
2587
      raise errors.ProgrammerError("Configuration file not read,"
2588
                                   " cannot save.")
2589
    update_serial = False
2590
    if isinstance(target, objects.Cluster):
2591
      test = target == self._config_data.cluster
2592
    elif isinstance(target, objects.Node):
2593
      test = target in self._config_data.nodes.values()
2594
      update_serial = True
2595
    elif isinstance(target, objects.Instance):
2596
      test = target in self._config_data.instances.values()
2597
    elif isinstance(target, objects.NodeGroup):
2598
      test = target in self._config_data.nodegroups.values()
2599
    elif isinstance(target, objects.Network):
2600
      test = target in self._config_data.networks.values()
2601
    else:
2602
      raise errors.ProgrammerError("Invalid object type (%s) passed to"
2603
                                   " ConfigWriter.Update" % type(target))
2604
    if not test:
2605
      raise errors.ConfigurationError("Configuration updated since object"
2606
                                      " has been read or unknown object")
2607
    target.serial_no += 1
2608
    target.mtime = now = time.time()
2609

    
2610
    if update_serial:
2611
      # for node updates, we need to increase the cluster serial too
2612
      self._config_data.cluster.serial_no += 1
2613
      self._config_data.cluster.mtime = now
2614

    
2615
    if isinstance(target, objects.Instance):
2616
      self._UnlockedReleaseDRBDMinors(target.name)
2617

    
2618
    if ec_id is not None:
2619
      # Commit all ips reserved by OpInstanceSetParams and OpGroupSetParams
2620
      self._UnlockedCommitTemporaryIps(ec_id)
2621

    
2622
    self._WriteConfig(feedback_fn=feedback_fn)
2623

    
2624
  @locking.ssynchronized(_config_lock)
2625
  def DropECReservations(self, ec_id):
2626
    """Drop per-execution-context reservations
2627

2628
    """
2629
    for rm in self._all_rms:
2630
      rm.DropECReservations(ec_id)
2631

    
2632
  @locking.ssynchronized(_config_lock, shared=1)
2633
  def GetAllNetworksInfo(self):
2634
    """Get configuration info of all the networks.
2635

2636
    """
2637
    return dict(self._config_data.networks)
2638

    
2639
  def _UnlockedGetNetworkList(self):
2640
    """Get the list of networks.
2641

2642
    This function is for internal use, when the config lock is already held.
2643

2644
    """
2645
    return self._config_data.networks.keys()
2646

    
2647
  @locking.ssynchronized(_config_lock, shared=1)
2648
  def GetNetworkList(self):
2649
    """Get the list of networks.
2650

2651
    @return: array of networks, ex. ["main", "vlan100", "200]
2652

2653
    """
2654
    return self._UnlockedGetNetworkList()
2655

    
2656
  @locking.ssynchronized(_config_lock, shared=1)
2657
  def GetNetworkNames(self):
2658
    """Get a list of network names
2659

2660
    """
2661
    names = [net.name
2662
             for net in self._config_data.networks.values()]
2663
    return names
2664

    
2665
  def _UnlockedGetNetwork(self, uuid):
2666
    """Returns information about a network.
2667

2668
    This function is for internal use, when the config lock is already held.
2669

2670
    """
2671
    if uuid not in self._config_data.networks:
2672
      return None
2673

    
2674
    return self._config_data.networks[uuid]
2675

    
2676
  @locking.ssynchronized(_config_lock, shared=1)
2677
  def GetNetwork(self, uuid):
2678
    """Returns information about a network.
2679

2680
    It takes the information from the configuration file.
2681

2682
    @param uuid: UUID of the network
2683

2684
    @rtype: L{objects.Network}
2685
    @return: the network object
2686

2687
    """
2688
    return self._UnlockedGetNetwork(uuid)
2689

    
2690
  @locking.ssynchronized(_config_lock)
2691
  def AddNetwork(self, net, ec_id, check_uuid=True):
2692
    """Add a network to the configuration.
2693

2694
    @type net: L{objects.Network}
2695
    @param net: the Network object to add
2696
    @type ec_id: string
2697
    @param ec_id: unique id for the job to use when creating a missing UUID
2698

2699
    """
2700
    self._UnlockedAddNetwork(net, ec_id, check_uuid)
2701
    self._WriteConfig()
2702

    
2703
  def _UnlockedAddNetwork(self, net, ec_id, check_uuid):
2704
    """Add a network to the configuration.
2705

2706
    """
2707
    logging.info("Adding network %s to configuration", net.name)
2708

    
2709
    if check_uuid:
2710
      self._EnsureUUID(net, ec_id)
2711

    
2712
    net.serial_no = 1
2713
    self._config_data.networks[net.uuid] = net
2714
    self._config_data.cluster.serial_no += 1
2715

    
2716
  def _UnlockedLookupNetwork(self, target):
2717
    """Lookup a network's UUID.
2718

2719
    @type target: string
2720
    @param target: network name or UUID
2721
    @rtype: string
2722
    @return: network UUID
2723
    @raises errors.OpPrereqError: when the target network cannot be found
2724

2725
    """
2726
    if target is None:
2727
      return None
2728
    if target in self._config_data.networks:
2729
      return target
2730
    for net in self._config_data.networks.values():
2731
      if net.name == target:
2732
        return net.uuid
2733
    raise errors.OpPrereqError("Network '%s' not found" % target,
2734
                               errors.ECODE_NOENT)
2735

    
2736
  @locking.ssynchronized(_config_lock, shared=1)
2737
  def LookupNetwork(self, target):
2738
    """Lookup a network's UUID.
2739

2740
    This function is just a wrapper over L{_UnlockedLookupNetwork}.
2741

2742
    @type target: string
2743
    @param target: network name or UUID
2744
    @rtype: string
2745
    @return: network UUID
2746

2747
    """
2748
    return self._UnlockedLookupNetwork(target)
2749

    
2750
  @locking.ssynchronized(_config_lock)
2751
  def RemoveNetwork(self, network_uuid):
2752
    """Remove a network from the configuration.
2753

2754
    @type network_uuid: string
2755
    @param network_uuid: the UUID of the network to remove
2756

2757
    """
2758
    logging.info("Removing network %s from configuration", network_uuid)
2759

    
2760
    if network_uuid not in self._config_data.networks:
2761
      raise errors.ConfigurationError("Unknown network '%s'" % network_uuid)
2762

    
2763
    del self._config_data.networks[network_uuid]
2764
    self._config_data.cluster.serial_no += 1
2765
    self._WriteConfig()
2766

    
2767
  def _UnlockedGetGroupNetParams(self, net_uuid, node_uuid):
2768
    """Get the netparams (mode, link) of a network.
2769

2770
    Get a network's netparams for a given node.
2771

2772
    @type net_uuid: string
2773
    @param net_uuid: network uuid
2774
    @type node_uuid: string
2775
    @param node_uuid: node UUID
2776
    @rtype: dict or None
2777
    @return: netparams
2778

2779
    """
2780
    node_info = self._UnlockedGetNodeInfo(node_uuid)
2781
    nodegroup_info = self._UnlockedGetNodeGroup(node_info.group)
2782
    netparams = nodegroup_info.networks.get(net_uuid, None)
2783

    
2784
    return netparams
2785

    
2786
  @locking.ssynchronized(_config_lock, shared=1)
2787
  def GetGroupNetParams(self, net_uuid, node_uuid):
2788
    """Locking wrapper of _UnlockedGetGroupNetParams()
2789

2790
    """
2791
    return self._UnlockedGetGroupNetParams(net_uuid, node_uuid)
2792

    
2793
  @locking.ssynchronized(_config_lock, shared=1)
2794
  def CheckIPInNodeGroup(self, ip, node_uuid):
2795
    """Check IP uniqueness in nodegroup.
2796

2797
    Check networks that are connected in the node's node group
2798
    if ip is contained in any of them. Used when creating/adding
2799
    a NIC to ensure uniqueness among nodegroups.
2800

2801
    @type ip: string
2802
    @param ip: ip address
2803
    @type node_uuid: string
2804
    @param node_uuid: node UUID
2805
    @rtype: (string, dict) or (None, None)
2806
    @return: (network name, netparams)
2807

2808
    """
2809
    if ip is None:
2810
      return (None, None)
2811
    node_info = self._UnlockedGetNodeInfo(node_uuid)
2812
    nodegroup_info = self._UnlockedGetNodeGroup(node_info.group)
2813
    for net_uuid in nodegroup_info.networks.keys():
2814
      net_info = self._UnlockedGetNetwork(net_uuid)
2815
      pool = network.AddressPool(net_info)
2816
      if pool.Contains(ip):
2817
        return (net_info.name, nodegroup_info.networks[net_uuid])
2818

    
2819
    return (None, None)