Statistics
| Branch: | Tag: | Revision:

root / lib / config.py @ 3311e336

History | View | Annotate | Download (96.8 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(object):
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 (recursively, including children).
450

451
    """
452
    def DiskAndAllChildren(disk):
453
      """Returns a list containing the given disk and all of his children.
454

455
      """
456
      disks = [disk]
457
      if disk.children:
458
        for child_disk in disk.children:
459
          disks.extend(DiskAndAllChildren(child_disk))
460
      return disks
461

    
462
    disks = []
463
    for instance in self._config_data.instances.values():
464
      for disk in instance.disks:
465
        disks.extend(DiskAndAllChildren(disk))
466
    return disks
467

    
468
  def _AllNICs(self):
469
    """Compute the list of all NICs.
470

471
    """
472
    nics = []
473
    for instance in self._config_data.instances.values():
474
      nics.extend(instance.nics)
475
    return nics
476

    
477
  def _AllIDs(self, include_temporary):
478
    """Compute the list of all UUIDs and names we have.
479

480
    @type include_temporary: boolean
481
    @param include_temporary: whether to include the _temporary_ids set
482
    @rtype: set
483
    @return: a set of IDs
484

485
    """
486
    existing = set()
487
    if include_temporary:
488
      existing.update(self._temporary_ids.GetReserved())
489
    existing.update(self._AllLVs())
490
    existing.update(self._config_data.instances.keys())
491
    existing.update(self._config_data.nodes.keys())
492
    existing.update([i.uuid for i in self._AllUUIDObjects() if i.uuid])
493
    return existing
494

    
495
  def _GenerateUniqueID(self, ec_id):
496
    """Generate an unique UUID.
497

498
    This checks the current node, instances and disk names for
499
    duplicates.
500

501
    @rtype: string
502
    @return: the unique id
503

504
    """
505
    existing = self._AllIDs(include_temporary=False)
506
    return self._temporary_ids.Generate(existing, utils.NewUUID, ec_id)
507

    
508
  @locking.ssynchronized(_config_lock, shared=1)
509
  def GenerateUniqueID(self, ec_id):
510
    """Generate an unique ID.
511

512
    This is just a wrapper over the unlocked version.
513

514
    @type ec_id: string
515
    @param ec_id: unique id for the job to reserve the id to
516

517
    """
518
    return self._GenerateUniqueID(ec_id)
519

    
520
  def _AllMACs(self):
521
    """Return all MACs present in the config.
522

523
    @rtype: list
524
    @return: the list of all MACs
525

526
    """
527
    result = []
528
    for instance in self._config_data.instances.values():
529
      for nic in instance.nics:
530
        result.append(nic.mac)
531

    
532
    return result
533

    
534
  def _AllDRBDSecrets(self):
535
    """Return all DRBD secrets present in the config.
536

537
    @rtype: list
538
    @return: the list of all DRBD secrets
539

540
    """
541
    def helper(disk, result):
542
      """Recursively gather secrets from this disk."""
543
      if disk.dev_type == constants.DT_DRBD8:
544
        result.append(disk.logical_id[5])
545
      if disk.children:
546
        for child in disk.children:
547
          helper(child, result)
548

    
549
    result = []
550
    for instance in self._config_data.instances.values():
551
      for disk in instance.disks:
552
        helper(disk, result)
553

    
554
    return result
555

    
556
  def _CheckDiskIDs(self, disk, l_ids, p_ids):
557
    """Compute duplicate disk IDs
558

559
    @type disk: L{objects.Disk}
560
    @param disk: the disk at which to start searching
561
    @type l_ids: list
562
    @param l_ids: list of current logical ids
563
    @type p_ids: list
564
    @param p_ids: list of current physical ids
565
    @rtype: list
566
    @return: a list of error messages
567

568
    """
569
    result = []
570
    if disk.logical_id is not None:
571
      if disk.logical_id in l_ids:
572
        result.append("duplicate logical id %s" % str(disk.logical_id))
573
      else:
574
        l_ids.append(disk.logical_id)
575
    if disk.physical_id is not None:
576
      if disk.physical_id in p_ids:
577
        result.append("duplicate physical id %s" % str(disk.physical_id))
578
      else:
579
        p_ids.append(disk.physical_id)
580

    
581
    if disk.children:
582
      for child in disk.children:
583
        result.extend(self._CheckDiskIDs(child, l_ids, p_ids))
584
    return result
585

    
586
  def _UnlockedVerifyConfig(self):
587
    """Verify function.
588

589
    @rtype: list
590
    @return: a list of error messages; a non-empty list signifies
591
        configuration errors
592

593
    """
594
    # pylint: disable=R0914
595
    result = []
596
    seen_macs = []
597
    ports = {}
598
    data = self._config_data
599
    cluster = data.cluster
600
    seen_lids = []
601
    seen_pids = []
602

    
603
    # global cluster checks
604
    if not cluster.enabled_hypervisors:
605
      result.append("enabled hypervisors list doesn't have any entries")
606
    invalid_hvs = set(cluster.enabled_hypervisors) - constants.HYPER_TYPES
607
    if invalid_hvs:
608
      result.append("enabled hypervisors contains invalid entries: %s" %
609
                    utils.CommaJoin(invalid_hvs))
610
    missing_hvp = (set(cluster.enabled_hypervisors) -
611
                   set(cluster.hvparams.keys()))
612
    if missing_hvp:
613
      result.append("hypervisor parameters missing for the enabled"
614
                    " hypervisor(s) %s" % utils.CommaJoin(missing_hvp))
615

    
616
    if not cluster.enabled_disk_templates:
617
      result.append("enabled disk templates list doesn't have any entries")
618
    invalid_disk_templates = set(cluster.enabled_disk_templates) \
619
                               - constants.DISK_TEMPLATES
620
    if invalid_disk_templates:
621
      result.append("enabled disk templates list contains invalid entries:"
622
                    " %s" % utils.CommaJoin(invalid_disk_templates))
623

    
624
    if cluster.master_node not in data.nodes:
625
      result.append("cluster has invalid primary node '%s'" %
626
                    cluster.master_node)
627

    
628
    def _helper(owner, attr, value, template):
629
      try:
630
        utils.ForceDictType(value, template)
631
      except errors.GenericError, err:
632
        result.append("%s has invalid %s: %s" % (owner, attr, err))
633

    
634
    def _helper_nic(owner, params):
635
      try:
636
        objects.NIC.CheckParameterSyntax(params)
637
      except errors.ConfigurationError, err:
638
        result.append("%s has invalid nicparams: %s" % (owner, err))
639

    
640
    def _helper_ipolicy(owner, ipolicy, iscluster):
641
      try:
642
        objects.InstancePolicy.CheckParameterSyntax(ipolicy, iscluster)
643
      except errors.ConfigurationError, err:
644
        result.append("%s has invalid instance policy: %s" % (owner, err))
645
      for key, value in ipolicy.items():
646
        if key == constants.ISPECS_MINMAX:
647
          for k in range(len(value)):
648
            _helper_ispecs(owner, "ipolicy/%s[%s]" % (key, k), value[k])
649
        elif key == constants.ISPECS_STD:
650
          _helper(owner, "ipolicy/" + key, value,
651
                  constants.ISPECS_PARAMETER_TYPES)
652
        else:
653
          # FIXME: assuming list type
654
          if key in constants.IPOLICY_PARAMETERS:
655
            exp_type = float
656
          else:
657
            exp_type = list
658
          if not isinstance(value, exp_type):
659
            result.append("%s has invalid instance policy: for %s,"
660
                          " expecting %s, got %s" %
661
                          (owner, key, exp_type.__name__, type(value)))
662

    
663
    def _helper_ispecs(owner, parentkey, params):
664
      for (key, value) in params.items():
665
        fullkey = "/".join([parentkey, key])
666
        _helper(owner, fullkey, value, constants.ISPECS_PARAMETER_TYPES)
667

    
668
    # check cluster parameters
669
    _helper("cluster", "beparams", cluster.SimpleFillBE({}),
670
            constants.BES_PARAMETER_TYPES)
671
    _helper("cluster", "nicparams", cluster.SimpleFillNIC({}),
672
            constants.NICS_PARAMETER_TYPES)
673
    _helper_nic("cluster", cluster.SimpleFillNIC({}))
674
    _helper("cluster", "ndparams", cluster.SimpleFillND({}),
675
            constants.NDS_PARAMETER_TYPES)
676
    _helper_ipolicy("cluster", cluster.ipolicy, True)
677

    
678
    # per-instance checks
679
    for instance_uuid in data.instances:
680
      instance = data.instances[instance_uuid]
681
      if instance.uuid != instance_uuid:
682
        result.append("instance '%s' is indexed by wrong UUID '%s'" %
683
                      (instance.name, instance_uuid))
684
      if instance.primary_node not in data.nodes:
685
        result.append("instance '%s' has invalid primary node '%s'" %
686
                      (instance.name, instance.primary_node))
687
      for snode in instance.secondary_nodes:
688
        if snode not in data.nodes:
689
          result.append("instance '%s' has invalid secondary node '%s'" %
690
                        (instance.name, snode))
691
      for idx, nic in enumerate(instance.nics):
692
        if nic.mac in seen_macs:
693
          result.append("instance '%s' has NIC %d mac %s duplicate" %
694
                        (instance.name, idx, nic.mac))
695
        else:
696
          seen_macs.append(nic.mac)
697
        if nic.nicparams:
698
          filled = cluster.SimpleFillNIC(nic.nicparams)
699
          owner = "instance %s nic %d" % (instance.name, idx)
700
          _helper(owner, "nicparams",
701
                  filled, constants.NICS_PARAMETER_TYPES)
702
          _helper_nic(owner, filled)
703

    
704
      # disk template checks
705
      if not instance.disk_template in data.cluster.enabled_disk_templates:
706
        result.append("instance '%s' uses the disabled disk template '%s'." %
707
                      (instance.name, instance.disk_template))
708

    
709
      # parameter checks
710
      if instance.beparams:
711
        _helper("instance %s" % instance.name, "beparams",
712
                cluster.FillBE(instance), constants.BES_PARAMETER_TYPES)
713

    
714
      # gather the drbd ports for duplicate checks
715
      for (idx, dsk) in enumerate(instance.disks):
716
        if dsk.dev_type in constants.LDS_DRBD:
717
          tcp_port = dsk.logical_id[2]
718
          if tcp_port not in ports:
719
            ports[tcp_port] = []
720
          ports[tcp_port].append((instance.name, "drbd disk %s" % idx))
721
      # gather network port reservation
722
      net_port = getattr(instance, "network_port", None)
723
      if net_port is not None:
724
        if net_port not in ports:
725
          ports[net_port] = []
726
        ports[net_port].append((instance.name, "network port"))
727

    
728
      # instance disk verify
729
      for idx, disk in enumerate(instance.disks):
730
        result.extend(["instance '%s' disk %d error: %s" %
731
                       (instance.name, idx, msg) for msg in disk.Verify()])
732
        result.extend(self._CheckDiskIDs(disk, seen_lids, seen_pids))
733

    
734
      wrong_names = _CheckInstanceDiskIvNames(instance.disks)
735
      if wrong_names:
736
        tmp = "; ".join(("name of disk %s should be '%s', but is '%s'" %
737
                         (idx, exp_name, actual_name))
738
                        for (idx, exp_name, actual_name) in wrong_names)
739

    
740
        result.append("Instance '%s' has wrongly named disks: %s" %
741
                      (instance.name, tmp))
742

    
743
    # cluster-wide pool of free ports
744
    for free_port in cluster.tcpudp_port_pool:
745
      if free_port not in ports:
746
        ports[free_port] = []
747
      ports[free_port].append(("cluster", "port marked as free"))
748

    
749
    # compute tcp/udp duplicate ports
750
    keys = ports.keys()
751
    keys.sort()
752
    for pnum in keys:
753
      pdata = ports[pnum]
754
      if len(pdata) > 1:
755
        txt = utils.CommaJoin(["%s/%s" % val for val in pdata])
756
        result.append("tcp/udp port %s has duplicates: %s" % (pnum, txt))
757

    
758
    # highest used tcp port check
759
    if keys:
760
      if keys[-1] > cluster.highest_used_port:
761
        result.append("Highest used port mismatch, saved %s, computed %s" %
762
                      (cluster.highest_used_port, keys[-1]))
763

    
764
    if not data.nodes[cluster.master_node].master_candidate:
765
      result.append("Master node is not a master candidate")
766

    
767
    # master candidate checks
768
    mc_now, mc_max, _ = self._UnlockedGetMasterCandidateStats()
769
    if mc_now < mc_max:
770
      result.append("Not enough master candidates: actual %d, target %d" %
771
                    (mc_now, mc_max))
772

    
773
    # node checks
774
    for node_uuid, node in data.nodes.items():
775
      if node.uuid != node_uuid:
776
        result.append("Node '%s' is indexed by wrong UUID '%s'" %
777
                      (node.name, node_uuid))
778
      if [node.master_candidate, node.drained, node.offline].count(True) > 1:
779
        result.append("Node %s state is invalid: master_candidate=%s,"
780
                      " drain=%s, offline=%s" %
781
                      (node.name, node.master_candidate, node.drained,
782
                       node.offline))
783
      if node.group not in data.nodegroups:
784
        result.append("Node '%s' has invalid group '%s'" %
785
                      (node.name, node.group))
786
      else:
787
        _helper("node %s" % node.name, "ndparams",
788
                cluster.FillND(node, data.nodegroups[node.group]),
789
                constants.NDS_PARAMETER_TYPES)
790
      used_globals = constants.NDC_GLOBALS.intersection(node.ndparams)
791
      if used_globals:
792
        result.append("Node '%s' has some global parameters set: %s" %
793
                      (node.name, utils.CommaJoin(used_globals)))
794

    
795
    # nodegroups checks
796
    nodegroups_names = set()
797
    for nodegroup_uuid in data.nodegroups:
798
      nodegroup = data.nodegroups[nodegroup_uuid]
799
      if nodegroup.uuid != nodegroup_uuid:
800
        result.append("node group '%s' (uuid: '%s') indexed by wrong uuid '%s'"
801
                      % (nodegroup.name, nodegroup.uuid, nodegroup_uuid))
802
      if utils.UUID_RE.match(nodegroup.name.lower()):
803
        result.append("node group '%s' (uuid: '%s') has uuid-like name" %
804
                      (nodegroup.name, nodegroup.uuid))
805
      if nodegroup.name in nodegroups_names:
806
        result.append("duplicate node group name '%s'" % nodegroup.name)
807
      else:
808
        nodegroups_names.add(nodegroup.name)
809
      group_name = "group %s" % nodegroup.name
810
      _helper_ipolicy(group_name, cluster.SimpleFillIPolicy(nodegroup.ipolicy),
811
                      False)
812
      if nodegroup.ndparams:
813
        _helper(group_name, "ndparams",
814
                cluster.SimpleFillND(nodegroup.ndparams),
815
                constants.NDS_PARAMETER_TYPES)
816

    
817
    # drbd minors check
818
    _, duplicates = self._UnlockedComputeDRBDMap()
819
    for node, minor, instance_a, instance_b in duplicates:
820
      result.append("DRBD minor %d on node %s is assigned twice to instances"
821
                    " %s and %s" % (minor, node, instance_a, instance_b))
822

    
823
    # IP checks
824
    default_nicparams = cluster.nicparams[constants.PP_DEFAULT]
825
    ips = {}
826

    
827
    def _AddIpAddress(ip, name):
828
      ips.setdefault(ip, []).append(name)
829

    
830
    _AddIpAddress(cluster.master_ip, "cluster_ip")
831

    
832
    for node in data.nodes.values():
833
      _AddIpAddress(node.primary_ip, "node:%s/primary" % node.name)
834
      if node.secondary_ip != node.primary_ip:
835
        _AddIpAddress(node.secondary_ip, "node:%s/secondary" % node.name)
836

    
837
    for instance in data.instances.values():
838
      for idx, nic in enumerate(instance.nics):
839
        if nic.ip is None:
840
          continue
841

    
842
        nicparams = objects.FillDict(default_nicparams, nic.nicparams)
843
        nic_mode = nicparams[constants.NIC_MODE]
844
        nic_link = nicparams[constants.NIC_LINK]
845

    
846
        if nic_mode == constants.NIC_MODE_BRIDGED:
847
          link = "bridge:%s" % nic_link
848
        elif nic_mode == constants.NIC_MODE_ROUTED:
849
          link = "route:%s" % nic_link
850
        else:
851
          raise errors.ProgrammerError("NIC mode '%s' not handled" % nic_mode)
852

    
853
        _AddIpAddress("%s/%s/%s" % (link, nic.ip, nic.network),
854
                      "instance:%s/nic:%d" % (instance.name, idx))
855

    
856
    for ip, owners in ips.items():
857
      if len(owners) > 1:
858
        result.append("IP address %s is used by multiple owners: %s" %
859
                      (ip, utils.CommaJoin(owners)))
860

    
861
    return result
862

    
863
  @locking.ssynchronized(_config_lock, shared=1)
864
  def VerifyConfig(self):
865
    """Verify function.
866

867
    This is just a wrapper over L{_UnlockedVerifyConfig}.
868

869
    @rtype: list
870
    @return: a list of error messages; a non-empty list signifies
871
        configuration errors
872

873
    """
874
    return self._UnlockedVerifyConfig()
875

    
876
  def _UnlockedSetDiskID(self, disk, node_uuid):
877
    """Convert the unique ID to the ID needed on the target nodes.
878

879
    This is used only for drbd, which needs ip/port configuration.
880

881
    The routine descends down and updates its children also, because
882
    this helps when the only the top device is passed to the remote
883
    node.
884

885
    This function is for internal use, when the config lock is already held.
886

887
    """
888
    if disk.children:
889
      for child in disk.children:
890
        self._UnlockedSetDiskID(child, node_uuid)
891

    
892
    if disk.logical_id is None and disk.physical_id is not None:
893
      return
894
    if disk.dev_type == constants.DT_DRBD8:
895
      pnode, snode, port, pminor, sminor, secret = disk.logical_id
896
      if node_uuid not in (pnode, snode):
897
        raise errors.ConfigurationError("DRBD device not knowing node %s" %
898
                                        node_uuid)
899
      pnode_info = self._UnlockedGetNodeInfo(pnode)
900
      snode_info = self._UnlockedGetNodeInfo(snode)
901
      if pnode_info is None or snode_info is None:
902
        raise errors.ConfigurationError("Can't find primary or secondary node"
903
                                        " for %s" % str(disk))
904
      p_data = (pnode_info.secondary_ip, port)
905
      s_data = (snode_info.secondary_ip, port)
906
      if pnode == node_uuid:
907
        disk.physical_id = p_data + s_data + (pminor, secret)
908
      else: # it must be secondary, we tested above
909
        disk.physical_id = s_data + p_data + (sminor, secret)
910
    else:
911
      disk.physical_id = disk.logical_id
912
    return
913

    
914
  @locking.ssynchronized(_config_lock)
915
  def SetDiskID(self, disk, node_uuid):
916
    """Convert the unique ID to the ID needed on the target nodes.
917

918
    This is used only for drbd, which needs ip/port configuration.
919

920
    The routine descends down and updates its children also, because
921
    this helps when the only the top device is passed to the remote
922
    node.
923

924
    """
925
    return self._UnlockedSetDiskID(disk, node_uuid)
926

    
927
  @locking.ssynchronized(_config_lock)
928
  def AddTcpUdpPort(self, port):
929
    """Adds a new port to the available port pool.
930

931
    @warning: this method does not "flush" the configuration (via
932
        L{_WriteConfig}); callers should do that themselves once the
933
        configuration is stable
934

935
    """
936
    if not isinstance(port, int):
937
      raise errors.ProgrammerError("Invalid type passed for port")
938

    
939
    self._config_data.cluster.tcpudp_port_pool.add(port)
940

    
941
  @locking.ssynchronized(_config_lock, shared=1)
942
  def GetPortList(self):
943
    """Returns a copy of the current port list.
944

945
    """
946
    return self._config_data.cluster.tcpudp_port_pool.copy()
947

    
948
  @locking.ssynchronized(_config_lock)
949
  def AllocatePort(self):
950
    """Allocate a port.
951

952
    The port will be taken from the available port pool or from the
953
    default port range (and in this case we increase
954
    highest_used_port).
955

956
    """
957
    # If there are TCP/IP ports configured, we use them first.
958
    if self._config_data.cluster.tcpudp_port_pool:
959
      port = self._config_data.cluster.tcpudp_port_pool.pop()
960
    else:
961
      port = self._config_data.cluster.highest_used_port + 1
962
      if port >= constants.LAST_DRBD_PORT:
963
        raise errors.ConfigurationError("The highest used port is greater"
964
                                        " than %s. Aborting." %
965
                                        constants.LAST_DRBD_PORT)
966
      self._config_data.cluster.highest_used_port = port
967

    
968
    self._WriteConfig()
969
    return port
970

    
971
  def _UnlockedComputeDRBDMap(self):
972
    """Compute the used DRBD minor/nodes.
973

974
    @rtype: (dict, list)
975
    @return: dictionary of node_uuid: dict of minor: instance_uuid;
976
        the returned dict will have all the nodes in it (even if with
977
        an empty list), and a list of duplicates; if the duplicates
978
        list is not empty, the configuration is corrupted and its caller
979
        should raise an exception
980

981
    """
982
    def _AppendUsedMinors(get_node_name_fn, instance, disk, used):
983
      duplicates = []
984
      if disk.dev_type == constants.DT_DRBD8 and len(disk.logical_id) >= 5:
985
        node_a, node_b, _, minor_a, minor_b = disk.logical_id[:5]
986
        for node_uuid, minor in ((node_a, minor_a), (node_b, minor_b)):
987
          assert node_uuid in used, \
988
            ("Node '%s' of instance '%s' not found in node list" %
989
             (get_node_name_fn(node_uuid), instance.name))
990
          if minor in used[node_uuid]:
991
            duplicates.append((node_uuid, minor, instance.uuid,
992
                               used[node_uuid][minor]))
993
          else:
994
            used[node_uuid][minor] = instance.uuid
995
      if disk.children:
996
        for child in disk.children:
997
          duplicates.extend(_AppendUsedMinors(get_node_name_fn, instance, child,
998
                                              used))
999
      return duplicates
1000

    
1001
    duplicates = []
1002
    my_dict = dict((node_uuid, {}) for node_uuid in self._config_data.nodes)
1003
    for instance in self._config_data.instances.itervalues():
1004
      for disk in instance.disks:
1005
        duplicates.extend(_AppendUsedMinors(self._UnlockedGetNodeName,
1006
                                            instance, disk, my_dict))
1007
    for (node_uuid, minor), inst_uuid in self._temporary_drbds.iteritems():
1008
      if minor in my_dict[node_uuid] and my_dict[node_uuid][minor] != inst_uuid:
1009
        duplicates.append((node_uuid, minor, inst_uuid,
1010
                           my_dict[node_uuid][minor]))
1011
      else:
1012
        my_dict[node_uuid][minor] = inst_uuid
1013
    return my_dict, duplicates
1014

    
1015
  @locking.ssynchronized(_config_lock)
1016
  def ComputeDRBDMap(self):
1017
    """Compute the used DRBD minor/nodes.
1018

1019
    This is just a wrapper over L{_UnlockedComputeDRBDMap}.
1020

1021
    @return: dictionary of node_uuid: dict of minor: instance_uuid;
1022
        the returned dict will have all the nodes in it (even if with
1023
        an empty list).
1024

1025
    """
1026
    d_map, duplicates = self._UnlockedComputeDRBDMap()
1027
    if duplicates:
1028
      raise errors.ConfigurationError("Duplicate DRBD ports detected: %s" %
1029
                                      str(duplicates))
1030
    return d_map
1031

    
1032
  @locking.ssynchronized(_config_lock)
1033
  def AllocateDRBDMinor(self, node_uuids, inst_uuid):
1034
    """Allocate a drbd minor.
1035

1036
    The free minor will be automatically computed from the existing
1037
    devices. A node can be given multiple times in order to allocate
1038
    multiple minors. The result is the list of minors, in the same
1039
    order as the passed nodes.
1040

1041
    @type inst_uuid: string
1042
    @param inst_uuid: the instance for which we allocate minors
1043

1044
    """
1045
    assert isinstance(inst_uuid, basestring), \
1046
           "Invalid argument '%s' passed to AllocateDRBDMinor" % inst_uuid
1047

    
1048
    d_map, duplicates = self._UnlockedComputeDRBDMap()
1049
    if duplicates:
1050
      raise errors.ConfigurationError("Duplicate DRBD ports detected: %s" %
1051
                                      str(duplicates))
1052
    result = []
1053
    for nuuid in node_uuids:
1054
      ndata = d_map[nuuid]
1055
      if not ndata:
1056
        # no minors used, we can start at 0
1057
        result.append(0)
1058
        ndata[0] = inst_uuid
1059
        self._temporary_drbds[(nuuid, 0)] = inst_uuid
1060
        continue
1061
      keys = ndata.keys()
1062
      keys.sort()
1063
      ffree = utils.FirstFree(keys)
1064
      if ffree is None:
1065
        # return the next minor
1066
        # TODO: implement high-limit check
1067
        minor = keys[-1] + 1
1068
      else:
1069
        minor = ffree
1070
      # double-check minor against current instances
1071
      assert minor not in d_map[nuuid], \
1072
             ("Attempt to reuse allocated DRBD minor %d on node %s,"
1073
              " already allocated to instance %s" %
1074
              (minor, nuuid, d_map[nuuid][minor]))
1075
      ndata[minor] = inst_uuid
1076
      # double-check minor against reservation
1077
      r_key = (nuuid, minor)
1078
      assert r_key not in self._temporary_drbds, \
1079
             ("Attempt to reuse reserved DRBD minor %d on node %s,"
1080
              " reserved for instance %s" %
1081
              (minor, nuuid, self._temporary_drbds[r_key]))
1082
      self._temporary_drbds[r_key] = inst_uuid
1083
      result.append(minor)
1084
    logging.debug("Request to allocate drbd minors, input: %s, returning %s",
1085
                  node_uuids, result)
1086
    return result
1087

    
1088
  def _UnlockedReleaseDRBDMinors(self, inst_uuid):
1089
    """Release temporary drbd minors allocated for a given instance.
1090

1091
    @type inst_uuid: string
1092
    @param inst_uuid: the instance for which temporary minors should be
1093
                      released
1094

1095
    """
1096
    assert isinstance(inst_uuid, basestring), \
1097
           "Invalid argument passed to ReleaseDRBDMinors"
1098
    for key, uuid in self._temporary_drbds.items():
1099
      if uuid == inst_uuid:
1100
        del self._temporary_drbds[key]
1101

    
1102
  @locking.ssynchronized(_config_lock)
1103
  def ReleaseDRBDMinors(self, inst_uuid):
1104
    """Release temporary drbd minors allocated for a given instance.
1105

1106
    This should be called on the error paths, on the success paths
1107
    it's automatically called by the ConfigWriter add and update
1108
    functions.
1109

1110
    This function is just a wrapper over L{_UnlockedReleaseDRBDMinors}.
1111

1112
    @type inst_uuid: string
1113
    @param inst_uuid: the instance for which temporary minors should be
1114
                      released
1115

1116
    """
1117
    self._UnlockedReleaseDRBDMinors(inst_uuid)
1118

    
1119
  @locking.ssynchronized(_config_lock, shared=1)
1120
  def GetConfigVersion(self):
1121
    """Get the configuration version.
1122

1123
    @return: Config version
1124

1125
    """
1126
    return self._config_data.version
1127

    
1128
  @locking.ssynchronized(_config_lock, shared=1)
1129
  def GetClusterName(self):
1130
    """Get cluster name.
1131

1132
    @return: Cluster name
1133

1134
    """
1135
    return self._config_data.cluster.cluster_name
1136

    
1137
  @locking.ssynchronized(_config_lock, shared=1)
1138
  def GetMasterNode(self):
1139
    """Get the UUID of the master node for this cluster.
1140

1141
    @return: Master node UUID
1142

1143
    """
1144
    return self._config_data.cluster.master_node
1145

    
1146
  @locking.ssynchronized(_config_lock, shared=1)
1147
  def GetMasterNodeName(self):
1148
    """Get the hostname of the master node for this cluster.
1149

1150
    @return: Master node hostname
1151

1152
    """
1153
    return self._UnlockedGetNodeName(self._config_data.cluster.master_node)
1154

    
1155
  @locking.ssynchronized(_config_lock, shared=1)
1156
  def GetMasterNodeInfo(self):
1157
    """Get the master node information for this cluster.
1158

1159
    @rtype: objects.Node
1160
    @return: Master node L{objects.Node} object
1161

1162
    """
1163
    return self._UnlockedGetNodeInfo(self._config_data.cluster.master_node)
1164

    
1165
  @locking.ssynchronized(_config_lock, shared=1)
1166
  def GetMasterIP(self):
1167
    """Get the IP of the master node for this cluster.
1168

1169
    @return: Master IP
1170

1171
    """
1172
    return self._config_data.cluster.master_ip
1173

    
1174
  @locking.ssynchronized(_config_lock, shared=1)
1175
  def GetMasterNetdev(self):
1176
    """Get the master network device for this cluster.
1177

1178
    """
1179
    return self._config_data.cluster.master_netdev
1180

    
1181
  @locking.ssynchronized(_config_lock, shared=1)
1182
  def GetMasterNetmask(self):
1183
    """Get the netmask of the master node for this cluster.
1184

1185
    """
1186
    return self._config_data.cluster.master_netmask
1187

    
1188
  @locking.ssynchronized(_config_lock, shared=1)
1189
  def GetUseExternalMipScript(self):
1190
    """Get flag representing whether to use the external master IP setup script.
1191

1192
    """
1193
    return self._config_data.cluster.use_external_mip_script
1194

    
1195
  @locking.ssynchronized(_config_lock, shared=1)
1196
  def GetFileStorageDir(self):
1197
    """Get the file storage dir for this cluster.
1198

1199
    """
1200
    return self._config_data.cluster.file_storage_dir
1201

    
1202
  @locking.ssynchronized(_config_lock, shared=1)
1203
  def GetSharedFileStorageDir(self):
1204
    """Get the shared file storage dir for this cluster.
1205

1206
    """
1207
    return self._config_data.cluster.shared_file_storage_dir
1208

    
1209
  @locking.ssynchronized(_config_lock, shared=1)
1210
  def GetHypervisorType(self):
1211
    """Get the hypervisor type for this cluster.
1212

1213
    """
1214
    return self._config_data.cluster.enabled_hypervisors[0]
1215

    
1216
  @locking.ssynchronized(_config_lock, shared=1)
1217
  def GetRsaHostKey(self):
1218
    """Return the rsa hostkey from the config.
1219

1220
    @rtype: string
1221
    @return: the rsa hostkey
1222

1223
    """
1224
    return self._config_data.cluster.rsahostkeypub
1225

    
1226
  @locking.ssynchronized(_config_lock, shared=1)
1227
  def GetDsaHostKey(self):
1228
    """Return the dsa hostkey from the config.
1229

1230
    @rtype: string
1231
    @return: the dsa hostkey
1232

1233
    """
1234
    return self._config_data.cluster.dsahostkeypub
1235

    
1236
  @locking.ssynchronized(_config_lock, shared=1)
1237
  def GetDefaultIAllocator(self):
1238
    """Get the default instance allocator for this cluster.
1239

1240
    """
1241
    return self._config_data.cluster.default_iallocator
1242

    
1243
  @locking.ssynchronized(_config_lock, shared=1)
1244
  def GetPrimaryIPFamily(self):
1245
    """Get cluster primary ip family.
1246

1247
    @return: primary ip family
1248

1249
    """
1250
    return self._config_data.cluster.primary_ip_family
1251

    
1252
  @locking.ssynchronized(_config_lock, shared=1)
1253
  def GetMasterNetworkParameters(self):
1254
    """Get network parameters of the master node.
1255

1256
    @rtype: L{object.MasterNetworkParameters}
1257
    @return: network parameters of the master node
1258

1259
    """
1260
    cluster = self._config_data.cluster
1261
    result = objects.MasterNetworkParameters(
1262
      uuid=cluster.master_node, ip=cluster.master_ip,
1263
      netmask=cluster.master_netmask, netdev=cluster.master_netdev,
1264
      ip_family=cluster.primary_ip_family)
1265

    
1266
    return result
1267

    
1268
  @locking.ssynchronized(_config_lock)
1269
  def AddNodeGroup(self, group, ec_id, check_uuid=True):
1270
    """Add a node group to the configuration.
1271

1272
    This method calls group.UpgradeConfig() to fill any missing attributes
1273
    according to their default values.
1274

1275
    @type group: L{objects.NodeGroup}
1276
    @param group: the NodeGroup object to add
1277
    @type ec_id: string
1278
    @param ec_id: unique id for the job to use when creating a missing UUID
1279
    @type check_uuid: bool
1280
    @param check_uuid: add an UUID to the group if it doesn't have one or, if
1281
                       it does, ensure that it does not exist in the
1282
                       configuration already
1283

1284
    """
1285
    self._UnlockedAddNodeGroup(group, ec_id, check_uuid)
1286
    self._WriteConfig()
1287

    
1288
  def _UnlockedAddNodeGroup(self, group, ec_id, check_uuid):
1289
    """Add a node group to the configuration.
1290

1291
    """
1292
    logging.info("Adding node group %s to configuration", group.name)
1293

    
1294
    # Some code might need to add a node group with a pre-populated UUID
1295
    # generated with ConfigWriter.GenerateUniqueID(). We allow them to bypass
1296
    # the "does this UUID" exist already check.
1297
    if check_uuid:
1298
      self._EnsureUUID(group, ec_id)
1299

    
1300
    try:
1301
      existing_uuid = self._UnlockedLookupNodeGroup(group.name)
1302
    except errors.OpPrereqError:
1303
      pass
1304
    else:
1305
      raise errors.OpPrereqError("Desired group name '%s' already exists as a"
1306
                                 " node group (UUID: %s)" %
1307
                                 (group.name, existing_uuid),
1308
                                 errors.ECODE_EXISTS)
1309

    
1310
    group.serial_no = 1
1311
    group.ctime = group.mtime = time.time()
1312
    group.UpgradeConfig()
1313

    
1314
    self._config_data.nodegroups[group.uuid] = group
1315
    self._config_data.cluster.serial_no += 1
1316

    
1317
  @locking.ssynchronized(_config_lock)
1318
  def RemoveNodeGroup(self, group_uuid):
1319
    """Remove a node group from the configuration.
1320

1321
    @type group_uuid: string
1322
    @param group_uuid: the UUID of the node group to remove
1323

1324
    """
1325
    logging.info("Removing node group %s from configuration", group_uuid)
1326

    
1327
    if group_uuid not in self._config_data.nodegroups:
1328
      raise errors.ConfigurationError("Unknown node group '%s'" % group_uuid)
1329

    
1330
    assert len(self._config_data.nodegroups) != 1, \
1331
            "Group '%s' is the only group, cannot be removed" % group_uuid
1332

    
1333
    del self._config_data.nodegroups[group_uuid]
1334
    self._config_data.cluster.serial_no += 1
1335
    self._WriteConfig()
1336

    
1337
  def _UnlockedLookupNodeGroup(self, target):
1338
    """Lookup a node group's UUID.
1339

1340
    @type target: string or None
1341
    @param target: group name or UUID or None to look for the default
1342
    @rtype: string
1343
    @return: nodegroup UUID
1344
    @raises errors.OpPrereqError: when the target group cannot be found
1345

1346
    """
1347
    if target is None:
1348
      if len(self._config_data.nodegroups) != 1:
1349
        raise errors.OpPrereqError("More than one node group exists. Target"
1350
                                   " group must be specified explicitly.")
1351
      else:
1352
        return self._config_data.nodegroups.keys()[0]
1353
    if target in self._config_data.nodegroups:
1354
      return target
1355
    for nodegroup in self._config_data.nodegroups.values():
1356
      if nodegroup.name == target:
1357
        return nodegroup.uuid
1358
    raise errors.OpPrereqError("Node group '%s' not found" % target,
1359
                               errors.ECODE_NOENT)
1360

    
1361
  @locking.ssynchronized(_config_lock, shared=1)
1362
  def LookupNodeGroup(self, target):
1363
    """Lookup a node group's UUID.
1364

1365
    This function is just a wrapper over L{_UnlockedLookupNodeGroup}.
1366

1367
    @type target: string or None
1368
    @param target: group name or UUID or None to look for the default
1369
    @rtype: string
1370
    @return: nodegroup UUID
1371

1372
    """
1373
    return self._UnlockedLookupNodeGroup(target)
1374

    
1375
  def _UnlockedGetNodeGroup(self, uuid):
1376
    """Lookup a node group.
1377

1378
    @type uuid: string
1379
    @param uuid: group UUID
1380
    @rtype: L{objects.NodeGroup} or None
1381
    @return: nodegroup object, or None if not found
1382

1383
    """
1384
    if uuid not in self._config_data.nodegroups:
1385
      return None
1386

    
1387
    return self._config_data.nodegroups[uuid]
1388

    
1389
  @locking.ssynchronized(_config_lock, shared=1)
1390
  def GetNodeGroup(self, uuid):
1391
    """Lookup a node group.
1392

1393
    @type uuid: string
1394
    @param uuid: group UUID
1395
    @rtype: L{objects.NodeGroup} or None
1396
    @return: nodegroup object, or None if not found
1397

1398
    """
1399
    return self._UnlockedGetNodeGroup(uuid)
1400

    
1401
  @locking.ssynchronized(_config_lock, shared=1)
1402
  def GetAllNodeGroupsInfo(self):
1403
    """Get the configuration of all node groups.
1404

1405
    """
1406
    return dict(self._config_data.nodegroups)
1407

    
1408
  @locking.ssynchronized(_config_lock, shared=1)
1409
  def GetNodeGroupList(self):
1410
    """Get a list of node groups.
1411

1412
    """
1413
    return self._config_data.nodegroups.keys()
1414

    
1415
  @locking.ssynchronized(_config_lock, shared=1)
1416
  def GetNodeGroupMembersByNodes(self, nodes):
1417
    """Get nodes which are member in the same nodegroups as the given nodes.
1418

1419
    """
1420
    ngfn = lambda node_uuid: self._UnlockedGetNodeInfo(node_uuid).group
1421
    return frozenset(member_uuid
1422
                     for node_uuid in nodes
1423
                     for member_uuid in
1424
                       self._UnlockedGetNodeGroup(ngfn(node_uuid)).members)
1425

    
1426
  @locking.ssynchronized(_config_lock, shared=1)
1427
  def GetMultiNodeGroupInfo(self, group_uuids):
1428
    """Get the configuration of multiple node groups.
1429

1430
    @param group_uuids: List of node group UUIDs
1431
    @rtype: list
1432
    @return: List of tuples of (group_uuid, group_info)
1433

1434
    """
1435
    return [(uuid, self._UnlockedGetNodeGroup(uuid)) for uuid in group_uuids]
1436

    
1437
  @locking.ssynchronized(_config_lock)
1438
  def AddInstance(self, instance, ec_id):
1439
    """Add an instance to the config.
1440

1441
    This should be used after creating a new instance.
1442

1443
    @type instance: L{objects.Instance}
1444
    @param instance: the instance object
1445

1446
    """
1447
    if not isinstance(instance, objects.Instance):
1448
      raise errors.ProgrammerError("Invalid type passed to AddInstance")
1449

    
1450
    if instance.disk_template != constants.DT_DISKLESS:
1451
      all_lvs = instance.MapLVsByNode()
1452
      logging.info("Instance '%s' DISK_LAYOUT: %s", instance.name, all_lvs)
1453

    
1454
    all_macs = self._AllMACs()
1455
    for nic in instance.nics:
1456
      if nic.mac in all_macs:
1457
        raise errors.ConfigurationError("Cannot add instance %s:"
1458
                                        " MAC address '%s' already in use." %
1459
                                        (instance.name, nic.mac))
1460

    
1461
    self._CheckUniqueUUID(instance, include_temporary=False)
1462

    
1463
    instance.serial_no = 1
1464
    instance.ctime = instance.mtime = time.time()
1465
    self._config_data.instances[instance.uuid] = instance
1466
    self._config_data.cluster.serial_no += 1
1467
    self._UnlockedReleaseDRBDMinors(instance.uuid)
1468
    self._UnlockedCommitTemporaryIps(ec_id)
1469
    self._WriteConfig()
1470

    
1471
  def _EnsureUUID(self, item, ec_id):
1472
    """Ensures a given object has a valid UUID.
1473

1474
    @param item: the instance or node to be checked
1475
    @param ec_id: the execution context id for the uuid reservation
1476

1477
    """
1478
    if not item.uuid:
1479
      item.uuid = self._GenerateUniqueID(ec_id)
1480
    else:
1481
      self._CheckUniqueUUID(item, include_temporary=True)
1482

    
1483
  def _CheckUniqueUUID(self, item, include_temporary):
1484
    """Checks that the UUID of the given object is unique.
1485

1486
    @param item: the instance or node to be checked
1487
    @param include_temporary: whether temporarily generated UUID's should be
1488
              included in the check. If the UUID of the item to be checked is
1489
              a temporarily generated one, this has to be C{False}.
1490

1491
    """
1492
    if not item.uuid:
1493
      raise errors.ConfigurationError("'%s' must have an UUID" % (item.name,))
1494
    if item.uuid in self._AllIDs(include_temporary=include_temporary):
1495
      raise errors.ConfigurationError("Cannot add '%s': UUID %s already"
1496
                                      " in use" % (item.name, item.uuid))
1497

    
1498
  def _SetInstanceStatus(self, inst_uuid, status, disks_active):
1499
    """Set the instance's status to a given value.
1500

1501
    """
1502
    if inst_uuid not in self._config_data.instances:
1503
      raise errors.ConfigurationError("Unknown instance '%s'" %
1504
                                      inst_uuid)
1505
    instance = self._config_data.instances[inst_uuid]
1506

    
1507
    if status is None:
1508
      status = instance.admin_state
1509
    if disks_active is None:
1510
      disks_active = instance.disks_active
1511

    
1512
    assert status in constants.ADMINST_ALL, \
1513
           "Invalid status '%s' passed to SetInstanceStatus" % (status,)
1514

    
1515
    if instance.admin_state != status or \
1516
       instance.disks_active != disks_active:
1517
      instance.admin_state = status
1518
      instance.disks_active = disks_active
1519
      instance.serial_no += 1
1520
      instance.mtime = time.time()
1521
      self._WriteConfig()
1522

    
1523
  @locking.ssynchronized(_config_lock)
1524
  def MarkInstanceUp(self, inst_uuid):
1525
    """Mark the instance status to up in the config.
1526

1527
    This also sets the instance disks active flag.
1528

1529
    """
1530
    self._SetInstanceStatus(inst_uuid, constants.ADMINST_UP, True)
1531

    
1532
  @locking.ssynchronized(_config_lock)
1533
  def MarkInstanceOffline(self, inst_uuid):
1534
    """Mark the instance status to down in the config.
1535

1536
    This also clears the instance disks active flag.
1537

1538
    """
1539
    self._SetInstanceStatus(inst_uuid, constants.ADMINST_OFFLINE, False)
1540

    
1541
  @locking.ssynchronized(_config_lock)
1542
  def RemoveInstance(self, inst_uuid):
1543
    """Remove the instance from the configuration.
1544

1545
    """
1546
    if inst_uuid not in self._config_data.instances:
1547
      raise errors.ConfigurationError("Unknown instance '%s'" % inst_uuid)
1548

    
1549
    # If a network port has been allocated to the instance,
1550
    # return it to the pool of free ports.
1551
    inst = self._config_data.instances[inst_uuid]
1552
    network_port = getattr(inst, "network_port", None)
1553
    if network_port is not None:
1554
      self._config_data.cluster.tcpudp_port_pool.add(network_port)
1555

    
1556
    instance = self._UnlockedGetInstanceInfo(inst_uuid)
1557

    
1558
    for nic in instance.nics:
1559
      if nic.network and nic.ip:
1560
        # Return all IP addresses to the respective address pools
1561
        self._UnlockedCommitIp(constants.RELEASE_ACTION, nic.network, nic.ip)
1562

    
1563
    del self._config_data.instances[inst_uuid]
1564
    self._config_data.cluster.serial_no += 1
1565
    self._WriteConfig()
1566

    
1567
  @locking.ssynchronized(_config_lock)
1568
  def RenameInstance(self, inst_uuid, new_name):
1569
    """Rename an instance.
1570

1571
    This needs to be done in ConfigWriter and not by RemoveInstance
1572
    combined with AddInstance as only we can guarantee an atomic
1573
    rename.
1574

1575
    """
1576
    if inst_uuid not in self._config_data.instances:
1577
      raise errors.ConfigurationError("Unknown instance '%s'" % inst_uuid)
1578

    
1579
    inst = self._config_data.instances[inst_uuid]
1580
    inst.name = new_name
1581

    
1582
    for (idx, disk) in enumerate(inst.disks):
1583
      if disk.dev_type in [constants.DT_FILE, constants.DT_SHARED_FILE]:
1584
        # rename the file paths in logical and physical id
1585
        file_storage_dir = os.path.dirname(os.path.dirname(disk.logical_id[1]))
1586
        disk.logical_id = (disk.logical_id[0],
1587
                           utils.PathJoin(file_storage_dir, inst.name,
1588
                                          "disk%s" % idx))
1589
        disk.physical_id = disk.logical_id
1590

    
1591
    # Force update of ssconf files
1592
    self._config_data.cluster.serial_no += 1
1593

    
1594
    self._WriteConfig()
1595

    
1596
  @locking.ssynchronized(_config_lock)
1597
  def MarkInstanceDown(self, inst_uuid):
1598
    """Mark the status of an instance to down in the configuration.
1599

1600
    This does not touch the instance disks active flag, as shut down instances
1601
    can still have active disks.
1602

1603
    """
1604
    self._SetInstanceStatus(inst_uuid, constants.ADMINST_DOWN, None)
1605

    
1606
  @locking.ssynchronized(_config_lock)
1607
  def MarkInstanceDisksActive(self, inst_uuid):
1608
    """Mark the status of instance disks active.
1609

1610
    """
1611
    self._SetInstanceStatus(inst_uuid, None, True)
1612

    
1613
  @locking.ssynchronized(_config_lock)
1614
  def MarkInstanceDisksInactive(self, inst_uuid):
1615
    """Mark the status of instance disks inactive.
1616

1617
    """
1618
    self._SetInstanceStatus(inst_uuid, None, False)
1619

    
1620
  def _UnlockedGetInstanceList(self):
1621
    """Get the list of instances.
1622

1623
    This function is for internal use, when the config lock is already held.
1624

1625
    """
1626
    return self._config_data.instances.keys()
1627

    
1628
  @locking.ssynchronized(_config_lock, shared=1)
1629
  def GetInstanceList(self):
1630
    """Get the list of instances.
1631

1632
    @return: array of instances, ex. ['instance2-uuid', 'instance1-uuid']
1633

1634
    """
1635
    return self._UnlockedGetInstanceList()
1636

    
1637
  def ExpandInstanceName(self, short_name):
1638
    """Attempt to expand an incomplete instance name.
1639

1640
    """
1641
    # Locking is done in L{ConfigWriter.GetAllInstancesInfo}
1642
    all_insts = self.GetAllInstancesInfo().values()
1643
    expanded_name = _MatchNameComponentIgnoreCase(
1644
                      short_name, [inst.name for inst in all_insts])
1645

    
1646
    if expanded_name is not None:
1647
      # there has to be exactly one instance with that name
1648
      inst = (filter(lambda n: n.name == expanded_name, all_insts)[0])
1649
      return (inst.uuid, inst.name)
1650
    else:
1651
      return (None, None)
1652

    
1653
  def _UnlockedGetInstanceInfo(self, inst_uuid):
1654
    """Returns information about an instance.
1655

1656
    This function is for internal use, when the config lock is already held.
1657

1658
    """
1659
    if inst_uuid not in self._config_data.instances:
1660
      return None
1661

    
1662
    return self._config_data.instances[inst_uuid]
1663

    
1664
  @locking.ssynchronized(_config_lock, shared=1)
1665
  def GetInstanceInfo(self, inst_uuid):
1666
    """Returns information about an instance.
1667

1668
    It takes the information from the configuration file. Other information of
1669
    an instance are taken from the live systems.
1670

1671
    @param inst_uuid: UUID of the instance
1672

1673
    @rtype: L{objects.Instance}
1674
    @return: the instance object
1675

1676
    """
1677
    return self._UnlockedGetInstanceInfo(inst_uuid)
1678

    
1679
  @locking.ssynchronized(_config_lock, shared=1)
1680
  def GetInstanceNodeGroups(self, inst_uuid, primary_only=False):
1681
    """Returns set of node group UUIDs for instance's nodes.
1682

1683
    @rtype: frozenset
1684

1685
    """
1686
    instance = self._UnlockedGetInstanceInfo(inst_uuid)
1687
    if not instance:
1688
      raise errors.ConfigurationError("Unknown instance '%s'" % inst_uuid)
1689

    
1690
    if primary_only:
1691
      nodes = [instance.primary_node]
1692
    else:
1693
      nodes = instance.all_nodes
1694

    
1695
    return frozenset(self._UnlockedGetNodeInfo(node_uuid).group
1696
                     for node_uuid in nodes)
1697

    
1698
  @locking.ssynchronized(_config_lock, shared=1)
1699
  def GetInstanceNetworks(self, inst_uuid):
1700
    """Returns set of network UUIDs for instance's nics.
1701

1702
    @rtype: frozenset
1703

1704
    """
1705
    instance = self._UnlockedGetInstanceInfo(inst_uuid)
1706
    if not instance:
1707
      raise errors.ConfigurationError("Unknown instance '%s'" % inst_uuid)
1708

    
1709
    networks = set()
1710
    for nic in instance.nics:
1711
      if nic.network:
1712
        networks.add(nic.network)
1713

    
1714
    return frozenset(networks)
1715

    
1716
  @locking.ssynchronized(_config_lock, shared=1)
1717
  def GetMultiInstanceInfo(self, inst_uuids):
1718
    """Get the configuration of multiple instances.
1719

1720
    @param inst_uuids: list of instance UUIDs
1721
    @rtype: list
1722
    @return: list of tuples (instance UUID, instance_info), where
1723
        instance_info is what would GetInstanceInfo return for the
1724
        node, while keeping the original order
1725

1726
    """
1727
    return [(uuid, self._UnlockedGetInstanceInfo(uuid)) for uuid in inst_uuids]
1728

    
1729
  @locking.ssynchronized(_config_lock, shared=1)
1730
  def GetMultiInstanceInfoByName(self, inst_names):
1731
    """Get the configuration of multiple instances.
1732

1733
    @param inst_names: list of instance names
1734
    @rtype: list
1735
    @return: list of tuples (instance, instance_info), where
1736
        instance_info is what would GetInstanceInfo return for the
1737
        node, while keeping the original order
1738

1739
    """
1740
    result = []
1741
    for name in inst_names:
1742
      instance = self._UnlockedGetInstanceInfoByName(name)
1743
      result.append((instance.uuid, instance))
1744
    return result
1745

    
1746
  @locking.ssynchronized(_config_lock, shared=1)
1747
  def GetAllInstancesInfo(self):
1748
    """Get the configuration of all instances.
1749

1750
    @rtype: dict
1751
    @return: dict of (instance, instance_info), where instance_info is what
1752
              would GetInstanceInfo return for the node
1753

1754
    """
1755
    return self._UnlockedGetAllInstancesInfo()
1756

    
1757
  def _UnlockedGetAllInstancesInfo(self):
1758
    my_dict = dict([(inst_uuid, self._UnlockedGetInstanceInfo(inst_uuid))
1759
                    for inst_uuid in self._UnlockedGetInstanceList()])
1760
    return my_dict
1761

    
1762
  @locking.ssynchronized(_config_lock, shared=1)
1763
  def GetInstancesInfoByFilter(self, filter_fn):
1764
    """Get instance configuration with a filter.
1765

1766
    @type filter_fn: callable
1767
    @param filter_fn: Filter function receiving instance object as parameter,
1768
      returning boolean. Important: this function is called while the
1769
      configuration locks is held. It must not do any complex work or call
1770
      functions potentially leading to a deadlock. Ideally it doesn't call any
1771
      other functions and just compares instance attributes.
1772

1773
    """
1774
    return dict((uuid, inst)
1775
                for (uuid, inst) in self._config_data.instances.items()
1776
                if filter_fn(inst))
1777

    
1778
  @locking.ssynchronized(_config_lock, shared=1)
1779
  def GetInstanceInfoByName(self, inst_name):
1780
    """Get the L{objects.Instance} object for a named instance.
1781

1782
    @param inst_name: name of the instance to get information for
1783
    @type inst_name: string
1784
    @return: the corresponding L{objects.Instance} instance or None if no
1785
          information is available
1786

1787
    """
1788
    return self._UnlockedGetInstanceInfoByName(inst_name)
1789

    
1790
  def _UnlockedGetInstanceInfoByName(self, inst_name):
1791
    for inst in self._UnlockedGetAllInstancesInfo().values():
1792
      if inst.name == inst_name:
1793
        return inst
1794
    return None
1795

    
1796
  def _UnlockedGetInstanceName(self, inst_uuid):
1797
    inst_info = self._UnlockedGetInstanceInfo(inst_uuid)
1798
    if inst_info is None:
1799
      raise errors.OpExecError("Unknown instance: %s" % inst_uuid)
1800
    return inst_info.name
1801

    
1802
  @locking.ssynchronized(_config_lock, shared=1)
1803
  def GetInstanceName(self, inst_uuid):
1804
    """Gets the instance name for the passed instance.
1805

1806
    @param inst_uuid: instance UUID to get name for
1807
    @type inst_uuid: string
1808
    @rtype: string
1809
    @return: instance name
1810

1811
    """
1812
    return self._UnlockedGetInstanceName(inst_uuid)
1813

    
1814
  @locking.ssynchronized(_config_lock, shared=1)
1815
  def GetInstanceNames(self, inst_uuids):
1816
    """Gets the instance names for the passed list of nodes.
1817

1818
    @param inst_uuids: list of instance UUIDs to get names for
1819
    @type inst_uuids: list of strings
1820
    @rtype: list of strings
1821
    @return: list of instance names
1822

1823
    """
1824
    return self._UnlockedGetInstanceNames(inst_uuids)
1825

    
1826
  def _UnlockedGetInstanceNames(self, inst_uuids):
1827
    return [self._UnlockedGetInstanceName(uuid) for uuid in inst_uuids]
1828

    
1829
  @locking.ssynchronized(_config_lock)
1830
  def AddNode(self, node, ec_id):
1831
    """Add a node to the configuration.
1832

1833
    @type node: L{objects.Node}
1834
    @param node: a Node instance
1835

1836
    """
1837
    logging.info("Adding node %s to configuration", node.name)
1838

    
1839
    self._EnsureUUID(node, ec_id)
1840

    
1841
    node.serial_no = 1
1842
    node.ctime = node.mtime = time.time()
1843
    self._UnlockedAddNodeToGroup(node.uuid, node.group)
1844
    self._config_data.nodes[node.uuid] = node
1845
    self._config_data.cluster.serial_no += 1
1846
    self._WriteConfig()
1847

    
1848
  @locking.ssynchronized(_config_lock)
1849
  def RemoveNode(self, node_uuid):
1850
    """Remove a node from the configuration.
1851

1852
    """
1853
    logging.info("Removing node %s from configuration", node_uuid)
1854

    
1855
    if node_uuid not in self._config_data.nodes:
1856
      raise errors.ConfigurationError("Unknown node '%s'" % node_uuid)
1857

    
1858
    self._UnlockedRemoveNodeFromGroup(self._config_data.nodes[node_uuid])
1859
    del self._config_data.nodes[node_uuid]
1860
    self._config_data.cluster.serial_no += 1
1861
    self._WriteConfig()
1862

    
1863
  def ExpandNodeName(self, short_name):
1864
    """Attempt to expand an incomplete node name into a node UUID.
1865

1866
    """
1867
    # Locking is done in L{ConfigWriter.GetAllNodesInfo}
1868
    all_nodes = self.GetAllNodesInfo().values()
1869
    expanded_name = _MatchNameComponentIgnoreCase(
1870
                      short_name, [node.name for node in all_nodes])
1871

    
1872
    if expanded_name is not None:
1873
      # there has to be exactly one node with that name
1874
      node = (filter(lambda n: n.name == expanded_name, all_nodes)[0])
1875
      return (node.uuid, node.name)
1876
    else:
1877
      return (None, None)
1878

    
1879
  def _UnlockedGetNodeInfo(self, node_uuid):
1880
    """Get the configuration of a node, as stored in the config.
1881

1882
    This function is for internal use, when the config lock is already
1883
    held.
1884

1885
    @param node_uuid: the node UUID
1886

1887
    @rtype: L{objects.Node}
1888
    @return: the node object
1889

1890
    """
1891
    if node_uuid not in self._config_data.nodes:
1892
      return None
1893

    
1894
    return self._config_data.nodes[node_uuid]
1895

    
1896
  @locking.ssynchronized(_config_lock, shared=1)
1897
  def GetNodeInfo(self, node_uuid):
1898
    """Get the configuration of a node, as stored in the config.
1899

1900
    This is just a locked wrapper over L{_UnlockedGetNodeInfo}.
1901

1902
    @param node_uuid: the node UUID
1903

1904
    @rtype: L{objects.Node}
1905
    @return: the node object
1906

1907
    """
1908
    return self._UnlockedGetNodeInfo(node_uuid)
1909

    
1910
  @locking.ssynchronized(_config_lock, shared=1)
1911
  def GetNodeInstances(self, node_uuid):
1912
    """Get the instances of a node, as stored in the config.
1913

1914
    @param node_uuid: the node UUID
1915

1916
    @rtype: (list, list)
1917
    @return: a tuple with two lists: the primary and the secondary instances
1918

1919
    """
1920
    pri = []
1921
    sec = []
1922
    for inst in self._config_data.instances.values():
1923
      if inst.primary_node == node_uuid:
1924
        pri.append(inst.uuid)
1925
      if node_uuid in inst.secondary_nodes:
1926
        sec.append(inst.uuid)
1927
    return (pri, sec)
1928

    
1929
  @locking.ssynchronized(_config_lock, shared=1)
1930
  def GetNodeGroupInstances(self, uuid, primary_only=False):
1931
    """Get the instances of a node group.
1932

1933
    @param uuid: Node group UUID
1934
    @param primary_only: Whether to only consider primary nodes
1935
    @rtype: frozenset
1936
    @return: List of instance UUIDs in node group
1937

1938
    """
1939
    if primary_only:
1940
      nodes_fn = lambda inst: [inst.primary_node]
1941
    else:
1942
      nodes_fn = lambda inst: inst.all_nodes
1943

    
1944
    return frozenset(inst.uuid
1945
                     for inst in self._config_data.instances.values()
1946
                     for node_uuid in nodes_fn(inst)
1947
                     if self._UnlockedGetNodeInfo(node_uuid).group == uuid)
1948

    
1949
  def _UnlockedGetHvparamsString(self, hvname):
1950
    """Return the string representation of the list of hyervisor parameters of
1951
    the given hypervisor.
1952

1953
    @see: C{GetHvparams}
1954

1955
    """
1956
    result = ""
1957
    hvparams = self._config_data.cluster.hvparams[hvname]
1958
    for key in hvparams:
1959
      result += "%s=%s\n" % (key, hvparams[key])
1960
    return result
1961

    
1962
  @locking.ssynchronized(_config_lock, shared=1)
1963
  def GetHvparamsString(self, hvname):
1964
    """Return the hypervisor parameters of the given hypervisor.
1965

1966
    @type hvname: string
1967
    @param hvname: name of a hypervisor
1968
    @rtype: string
1969
    @return: string containing key-value-pairs, one pair on each line;
1970
      format: KEY=VALUE
1971

1972
    """
1973
    return self._UnlockedGetHvparamsString(hvname)
1974

    
1975
  def _UnlockedGetNodeList(self):
1976
    """Return the list of nodes which are in the configuration.
1977

1978
    This function is for internal use, when the config lock is already
1979
    held.
1980

1981
    @rtype: list
1982

1983
    """
1984
    return self._config_data.nodes.keys()
1985

    
1986
  @locking.ssynchronized(_config_lock, shared=1)
1987
  def GetNodeList(self):
1988
    """Return the list of nodes which are in the configuration.
1989

1990
    """
1991
    return self._UnlockedGetNodeList()
1992

    
1993
  def _UnlockedGetOnlineNodeList(self):
1994
    """Return the list of nodes which are online.
1995

1996
    """
1997
    all_nodes = [self._UnlockedGetNodeInfo(node)
1998
                 for node in self._UnlockedGetNodeList()]
1999
    return [node.uuid for node in all_nodes if not node.offline]
2000

    
2001
  @locking.ssynchronized(_config_lock, shared=1)
2002
  def GetOnlineNodeList(self):
2003
    """Return the list of nodes which are online.
2004

2005
    """
2006
    return self._UnlockedGetOnlineNodeList()
2007

    
2008
  @locking.ssynchronized(_config_lock, shared=1)
2009
  def GetVmCapableNodeList(self):
2010
    """Return the list of nodes which are not vm capable.
2011

2012
    """
2013
    all_nodes = [self._UnlockedGetNodeInfo(node)
2014
                 for node in self._UnlockedGetNodeList()]
2015
    return [node.uuid for node in all_nodes if node.vm_capable]
2016

    
2017
  @locking.ssynchronized(_config_lock, shared=1)
2018
  def GetNonVmCapableNodeList(self):
2019
    """Return the list of nodes which are not vm capable.
2020

2021
    """
2022
    all_nodes = [self._UnlockedGetNodeInfo(node)
2023
                 for node in self._UnlockedGetNodeList()]
2024
    return [node.uuid for node in all_nodes if not node.vm_capable]
2025

    
2026
  @locking.ssynchronized(_config_lock, shared=1)
2027
  def GetMultiNodeInfo(self, node_uuids):
2028
    """Get the configuration of multiple nodes.
2029

2030
    @param node_uuids: list of node UUIDs
2031
    @rtype: list
2032
    @return: list of tuples of (node, node_info), where node_info is
2033
        what would GetNodeInfo return for the node, in the original
2034
        order
2035

2036
    """
2037
    return [(uuid, self._UnlockedGetNodeInfo(uuid)) for uuid in node_uuids]
2038

    
2039
  def _UnlockedGetAllNodesInfo(self):
2040
    """Gets configuration of all nodes.
2041

2042
    @note: See L{GetAllNodesInfo}
2043

2044
    """
2045
    return dict([(node_uuid, self._UnlockedGetNodeInfo(node_uuid))
2046
                 for node_uuid in self._UnlockedGetNodeList()])
2047

    
2048
  @locking.ssynchronized(_config_lock, shared=1)
2049
  def GetAllNodesInfo(self):
2050
    """Get the configuration of all nodes.
2051

2052
    @rtype: dict
2053
    @return: dict of (node, node_info), where node_info is what
2054
              would GetNodeInfo return for the node
2055

2056
    """
2057
    return self._UnlockedGetAllNodesInfo()
2058

    
2059
  def _UnlockedGetNodeInfoByName(self, node_name):
2060
    for node in self._UnlockedGetAllNodesInfo().values():
2061
      if node.name == node_name:
2062
        return node
2063
    return None
2064

    
2065
  @locking.ssynchronized(_config_lock, shared=1)
2066
  def GetNodeInfoByName(self, node_name):
2067
    """Get the L{objects.Node} object for a named node.
2068

2069
    @param node_name: name of the node to get information for
2070
    @type node_name: string
2071
    @return: the corresponding L{objects.Node} instance or None if no
2072
          information is available
2073

2074
    """
2075
    return self._UnlockedGetNodeInfoByName(node_name)
2076

    
2077
  def _UnlockedGetNodeName(self, node_spec):
2078
    if isinstance(node_spec, objects.Node):
2079
      return node_spec.name
2080
    elif isinstance(node_spec, basestring):
2081
      node_info = self._UnlockedGetNodeInfo(node_spec)
2082
      if node_info is None:
2083
        raise errors.OpExecError("Unknown node: %s" % node_spec)
2084
      return node_info.name
2085
    else:
2086
      raise errors.ProgrammerError("Can't handle node spec '%s'" % node_spec)
2087

    
2088
  @locking.ssynchronized(_config_lock, shared=1)
2089
  def GetNodeName(self, node_spec):
2090
    """Gets the node name for the passed node.
2091

2092
    @param node_spec: node to get names for
2093
    @type node_spec: either node UUID or a L{objects.Node} object
2094
    @rtype: string
2095
    @return: node name
2096

2097
    """
2098
    return self._UnlockedGetNodeName(node_spec)
2099

    
2100
  def _UnlockedGetNodeNames(self, node_specs):
2101
    return [self._UnlockedGetNodeName(node_spec) for node_spec in node_specs]
2102

    
2103
  @locking.ssynchronized(_config_lock, shared=1)
2104
  def GetNodeNames(self, node_specs):
2105
    """Gets the node names for the passed list of nodes.
2106

2107
    @param node_specs: list of nodes to get names for
2108
    @type node_specs: list of either node UUIDs or L{objects.Node} objects
2109
    @rtype: list of strings
2110
    @return: list of node names
2111

2112
    """
2113
    return self._UnlockedGetNodeNames(node_specs)
2114

    
2115
  @locking.ssynchronized(_config_lock, shared=1)
2116
  def GetNodeGroupsFromNodes(self, node_uuids):
2117
    """Returns groups for a list of nodes.
2118

2119
    @type node_uuids: list of string
2120
    @param node_uuids: List of node UUIDs
2121
    @rtype: frozenset
2122

2123
    """
2124
    return frozenset(self._UnlockedGetNodeInfo(uuid).group
2125
                     for uuid in node_uuids)
2126

    
2127
  def _UnlockedGetMasterCandidateStats(self, exceptions=None):
2128
    """Get the number of current and maximum desired and possible candidates.
2129

2130
    @type exceptions: list
2131
    @param exceptions: if passed, list of nodes that should be ignored
2132
    @rtype: tuple
2133
    @return: tuple of (current, desired and possible, possible)
2134

2135
    """
2136
    mc_now = mc_should = mc_max = 0
2137
    for node in self._config_data.nodes.values():
2138
      if exceptions and node.uuid in exceptions:
2139
        continue
2140
      if not (node.offline or node.drained) and node.master_capable:
2141
        mc_max += 1
2142
      if node.master_candidate:
2143
        mc_now += 1
2144
    mc_should = min(mc_max, self._config_data.cluster.candidate_pool_size)
2145
    return (mc_now, mc_should, mc_max)
2146

    
2147
  @locking.ssynchronized(_config_lock, shared=1)
2148
  def GetMasterCandidateStats(self, exceptions=None):
2149
    """Get the number of current and maximum possible candidates.
2150

2151
    This is just a wrapper over L{_UnlockedGetMasterCandidateStats}.
2152

2153
    @type exceptions: list
2154
    @param exceptions: if passed, list of nodes that should be ignored
2155
    @rtype: tuple
2156
    @return: tuple of (current, max)
2157

2158
    """
2159
    return self._UnlockedGetMasterCandidateStats(exceptions)
2160

    
2161
  @locking.ssynchronized(_config_lock)
2162
  def MaintainCandidatePool(self, exception_node_uuids):
2163
    """Try to grow the candidate pool to the desired size.
2164

2165
    @type exception_node_uuids: list
2166
    @param exception_node_uuids: if passed, list of nodes that should be ignored
2167
    @rtype: list
2168
    @return: list with the adjusted nodes (L{objects.Node} instances)
2169

2170
    """
2171
    mc_now, mc_max, _ = self._UnlockedGetMasterCandidateStats(
2172
                          exception_node_uuids)
2173
    mod_list = []
2174
    if mc_now < mc_max:
2175
      node_list = self._config_data.nodes.keys()
2176
      random.shuffle(node_list)
2177
      for uuid in node_list:
2178
        if mc_now >= mc_max:
2179
          break
2180
        node = self._config_data.nodes[uuid]
2181
        if (node.master_candidate or node.offline or node.drained or
2182
            node.uuid in exception_node_uuids or not node.master_capable):
2183
          continue
2184
        mod_list.append(node)
2185
        node.master_candidate = True
2186
        node.serial_no += 1
2187
        mc_now += 1
2188
      if mc_now != mc_max:
2189
        # this should not happen
2190
        logging.warning("Warning: MaintainCandidatePool didn't manage to"
2191
                        " fill the candidate pool (%d/%d)", mc_now, mc_max)
2192
      if mod_list:
2193
        self._config_data.cluster.serial_no += 1
2194
        self._WriteConfig()
2195

    
2196
    return mod_list
2197

    
2198
  def _UnlockedAddNodeToGroup(self, node_uuid, nodegroup_uuid):
2199
    """Add a given node to the specified group.
2200

2201
    """
2202
    if nodegroup_uuid not in self._config_data.nodegroups:
2203
      # This can happen if a node group gets deleted between its lookup and
2204
      # when we're adding the first node to it, since we don't keep a lock in
2205
      # the meantime. It's ok though, as we'll fail cleanly if the node group
2206
      # is not found anymore.
2207
      raise errors.OpExecError("Unknown node group: %s" % nodegroup_uuid)
2208
    if node_uuid not in self._config_data.nodegroups[nodegroup_uuid].members:
2209
      self._config_data.nodegroups[nodegroup_uuid].members.append(node_uuid)
2210

    
2211
  def _UnlockedRemoveNodeFromGroup(self, node):
2212
    """Remove a given node from its group.
2213

2214
    """
2215
    nodegroup = node.group
2216
    if nodegroup not in self._config_data.nodegroups:
2217
      logging.warning("Warning: node '%s' has unknown node group '%s'"
2218
                      " (while being removed from it)", node.uuid, nodegroup)
2219
    nodegroup_obj = self._config_data.nodegroups[nodegroup]
2220
    if node.uuid not in nodegroup_obj.members:
2221
      logging.warning("Warning: node '%s' not a member of its node group '%s'"
2222
                      " (while being removed from it)", node.uuid, nodegroup)
2223
    else:
2224
      nodegroup_obj.members.remove(node.uuid)
2225

    
2226
  @locking.ssynchronized(_config_lock)
2227
  def AssignGroupNodes(self, mods):
2228
    """Changes the group of a number of nodes.
2229

2230
    @type mods: list of tuples; (node name, new group UUID)
2231
    @param mods: Node membership modifications
2232

2233
    """
2234
    groups = self._config_data.nodegroups
2235
    nodes = self._config_data.nodes
2236

    
2237
    resmod = []
2238

    
2239
    # Try to resolve UUIDs first
2240
    for (node_uuid, new_group_uuid) in mods:
2241
      try:
2242
        node = nodes[node_uuid]
2243
      except KeyError:
2244
        raise errors.ConfigurationError("Unable to find node '%s'" % node_uuid)
2245

    
2246
      if node.group == new_group_uuid:
2247
        # Node is being assigned to its current group
2248
        logging.debug("Node '%s' was assigned to its current group (%s)",
2249
                      node_uuid, node.group)
2250
        continue
2251

    
2252
      # Try to find current group of node
2253
      try:
2254
        old_group = groups[node.group]
2255
      except KeyError:
2256
        raise errors.ConfigurationError("Unable to find old group '%s'" %
2257
                                        node.group)
2258

    
2259
      # Try to find new group for node
2260
      try:
2261
        new_group = groups[new_group_uuid]
2262
      except KeyError:
2263
        raise errors.ConfigurationError("Unable to find new group '%s'" %
2264
                                        new_group_uuid)
2265

    
2266
      assert node.uuid in old_group.members, \
2267
        ("Inconsistent configuration: node '%s' not listed in members for its"
2268
         " old group '%s'" % (node.uuid, old_group.uuid))
2269
      assert node.uuid not in new_group.members, \
2270
        ("Inconsistent configuration: node '%s' already listed in members for"
2271
         " its new group '%s'" % (node.uuid, new_group.uuid))
2272

    
2273
      resmod.append((node, old_group, new_group))
2274

    
2275
    # Apply changes
2276
    for (node, old_group, new_group) in resmod:
2277
      assert node.uuid != new_group.uuid and old_group.uuid != new_group.uuid, \
2278
        "Assigning to current group is not possible"
2279

    
2280
      node.group = new_group.uuid
2281

    
2282
      # Update members of involved groups
2283
      if node.uuid in old_group.members:
2284
        old_group.members.remove(node.uuid)
2285
      if node.uuid not in new_group.members:
2286
        new_group.members.append(node.uuid)
2287

    
2288
    # Update timestamps and serials (only once per node/group object)
2289
    now = time.time()
2290
    for obj in frozenset(itertools.chain(*resmod)): # pylint: disable=W0142
2291
      obj.serial_no += 1
2292
      obj.mtime = now
2293

    
2294
    # Force ssconf update
2295
    self._config_data.cluster.serial_no += 1
2296

    
2297
    self._WriteConfig()
2298

    
2299
  def _BumpSerialNo(self):
2300
    """Bump up the serial number of the config.
2301

2302
    """
2303
    self._config_data.serial_no += 1
2304
    self._config_data.mtime = time.time()
2305

    
2306
  def _AllUUIDObjects(self):
2307
    """Returns all objects with uuid attributes.
2308

2309
    """
2310
    return (self._config_data.instances.values() +
2311
            self._config_data.nodes.values() +
2312
            self._config_data.nodegroups.values() +
2313
            self._config_data.networks.values() +
2314
            self._AllDisks() +
2315
            self._AllNICs() +
2316
            [self._config_data.cluster])
2317

    
2318
  def _OpenConfig(self, accept_foreign):
2319
    """Read the config data from disk.
2320

2321
    """
2322
    raw_data = utils.ReadFile(self._cfg_file)
2323

    
2324
    try:
2325
      data = objects.ConfigData.FromDict(serializer.Load(raw_data))
2326
    except Exception, err:
2327
      raise errors.ConfigurationError(err)
2328

    
2329
    # Make sure the configuration has the right version
2330
    _ValidateConfig(data)
2331

    
2332
    if (not hasattr(data, "cluster") or
2333
        not hasattr(data.cluster, "rsahostkeypub")):
2334
      raise errors.ConfigurationError("Incomplete configuration"
2335
                                      " (missing cluster.rsahostkeypub)")
2336

    
2337
    if not data.cluster.master_node in data.nodes:
2338
      msg = ("The configuration denotes node %s as master, but does not"
2339
             " contain information about this node" %
2340
             data.cluster.master_node)
2341
      raise errors.ConfigurationError(msg)
2342

    
2343
    master_info = data.nodes[data.cluster.master_node]
2344
    if master_info.name != self._my_hostname and not accept_foreign:
2345
      msg = ("The configuration denotes node %s as master, while my"
2346
             " hostname is %s; opening a foreign configuration is only"
2347
             " possible in accept_foreign mode" %
2348
             (master_info.name, self._my_hostname))
2349
      raise errors.ConfigurationError(msg)
2350

    
2351
    self._config_data = data
2352
    # reset the last serial as -1 so that the next write will cause
2353
    # ssconf update
2354
    self._last_cluster_serial = -1
2355

    
2356
    # Upgrade configuration if needed
2357
    self._UpgradeConfig()
2358

    
2359
    self._cfg_id = utils.GetFileID(path=self._cfg_file)
2360

    
2361
  def _UpgradeConfig(self):
2362
    """Run any upgrade steps.
2363

2364
    This method performs both in-object upgrades and also update some data
2365
    elements that need uniqueness across the whole configuration or interact
2366
    with other objects.
2367

2368
    @warning: this function will call L{_WriteConfig()}, but also
2369
        L{DropECReservations} so it needs to be called only from a
2370
        "safe" place (the constructor). If one wanted to call it with
2371
        the lock held, a DropECReservationUnlocked would need to be
2372
        created first, to avoid causing deadlock.
2373

2374
    """
2375
    # Keep a copy of the persistent part of _config_data to check for changes
2376
    # Serialization doesn't guarantee order in dictionaries
2377
    oldconf = copy.deepcopy(self._config_data.ToDict())
2378

    
2379
    # In-object upgrades
2380
    self._config_data.UpgradeConfig()
2381

    
2382
    for item in self._AllUUIDObjects():
2383
      if item.uuid is None:
2384
        item.uuid = self._GenerateUniqueID(_UPGRADE_CONFIG_JID)
2385
    if not self._config_data.nodegroups:
2386
      default_nodegroup_name = constants.INITIAL_NODE_GROUP_NAME
2387
      default_nodegroup = objects.NodeGroup(name=default_nodegroup_name,
2388
                                            members=[])
2389
      self._UnlockedAddNodeGroup(default_nodegroup, _UPGRADE_CONFIG_JID, True)
2390
    for node in self._config_data.nodes.values():
2391
      if not node.group:
2392
        node.group = self.LookupNodeGroup(None)
2393
      # This is technically *not* an upgrade, but needs to be done both when
2394
      # nodegroups are being added, and upon normally loading the config,
2395
      # because the members list of a node group is discarded upon
2396
      # serializing/deserializing the object.
2397
      self._UnlockedAddNodeToGroup(node.uuid, node.group)
2398

    
2399
    modified = (oldconf != self._config_data.ToDict())
2400
    if modified:
2401
      self._WriteConfig()
2402
      # This is ok even if it acquires the internal lock, as _UpgradeConfig is
2403
      # only called at config init time, without the lock held
2404
      self.DropECReservations(_UPGRADE_CONFIG_JID)
2405
    else:
2406
      config_errors = self._UnlockedVerifyConfig()
2407
      if config_errors:
2408
        errmsg = ("Loaded configuration data is not consistent: %s" %
2409
                  (utils.CommaJoin(config_errors)))
2410
        logging.critical(errmsg)
2411

    
2412
  def _DistributeConfig(self, feedback_fn):
2413
    """Distribute the configuration to the other nodes.
2414

2415
    Currently, this only copies the configuration file. In the future,
2416
    it could be used to encapsulate the 2/3-phase update mechanism.
2417

2418
    """
2419
    if self._offline:
2420
      return True
2421

    
2422
    bad = False
2423

    
2424
    node_list = []
2425
    addr_list = []
2426
    myhostname = self._my_hostname
2427
    # we can skip checking whether _UnlockedGetNodeInfo returns None
2428
    # since the node list comes from _UnlocketGetNodeList, and we are
2429
    # called with the lock held, so no modifications should take place
2430
    # in between
2431
    for node_uuid in self._UnlockedGetNodeList():
2432
      node_info = self._UnlockedGetNodeInfo(node_uuid)
2433
      if node_info.name == myhostname or not node_info.master_candidate:
2434
        continue
2435
      node_list.append(node_info.name)
2436
      addr_list.append(node_info.primary_ip)
2437

    
2438
    # TODO: Use dedicated resolver talking to config writer for name resolution
2439
    result = \
2440
      self._GetRpc(addr_list).call_upload_file(node_list, self._cfg_file)
2441
    for to_node, to_result in result.items():
2442
      msg = to_result.fail_msg
2443
      if msg:
2444
        msg = ("Copy of file %s to node %s failed: %s" %
2445
               (self._cfg_file, to_node, msg))
2446
        logging.error(msg)
2447

    
2448
        if feedback_fn:
2449
          feedback_fn(msg)
2450

    
2451
        bad = True
2452

    
2453
    return not bad
2454

    
2455
  def _WriteConfig(self, destination=None, feedback_fn=None):
2456
    """Write the configuration data to persistent storage.
2457

2458
    """
2459
    assert feedback_fn is None or callable(feedback_fn)
2460

    
2461
    # Warn on config errors, but don't abort the save - the
2462
    # configuration has already been modified, and we can't revert;
2463
    # the best we can do is to warn the user and save as is, leaving
2464
    # recovery to the user
2465
    config_errors = self._UnlockedVerifyConfig()
2466
    if config_errors:
2467
      errmsg = ("Configuration data is not consistent: %s" %
2468
                (utils.CommaJoin(config_errors)))
2469
      logging.critical(errmsg)
2470
      if feedback_fn:
2471
        feedback_fn(errmsg)
2472

    
2473
    if destination is None:
2474
      destination = self._cfg_file
2475
    self._BumpSerialNo()
2476
    txt = serializer.Dump(self._config_data.ToDict())
2477

    
2478
    getents = self._getents()
2479
    try:
2480
      fd = utils.SafeWriteFile(destination, self._cfg_id, data=txt,
2481
                               close=False, gid=getents.confd_gid, mode=0640)
2482
    except errors.LockError:
2483
      raise errors.ConfigurationError("The configuration file has been"
2484
                                      " modified since the last write, cannot"
2485
                                      " update")
2486
    try:
2487
      self._cfg_id = utils.GetFileID(fd=fd)
2488
    finally:
2489
      os.close(fd)
2490

    
2491
    self.write_count += 1
2492

    
2493
    # and redistribute the config file to master candidates
2494
    self._DistributeConfig(feedback_fn)
2495

    
2496
    # Write ssconf files on all nodes (including locally)
2497
    if self._last_cluster_serial < self._config_data.cluster.serial_no:
2498
      if not self._offline:
2499
        result = self._GetRpc(None).call_write_ssconf_files(
2500
          self._UnlockedGetNodeNames(self._UnlockedGetOnlineNodeList()),
2501
          self._UnlockedGetSsconfValues())
2502

    
2503
        for nname, nresu in result.items():
2504
          msg = nresu.fail_msg
2505
          if msg:
2506
            errmsg = ("Error while uploading ssconf files to"
2507
                      " node %s: %s" % (nname, msg))
2508
            logging.warning(errmsg)
2509

    
2510
            if feedback_fn:
2511
              feedback_fn(errmsg)
2512

    
2513
      self._last_cluster_serial = self._config_data.cluster.serial_no
2514

    
2515
  def _GetAllHvparamsStrings(self, hypervisors):
2516
    """Get the hvparams of all given hypervisors from the config.
2517

2518
    @type hypervisors: list of string
2519
    @param hypervisors: list of hypervisor names
2520
    @rtype: dict of strings
2521
    @returns: dictionary mapping the hypervisor name to a string representation
2522
      of the hypervisor's hvparams
2523

2524
    """
2525
    hvparams = {}
2526
    for hv in hypervisors:
2527
      hvparams[hv] = self._UnlockedGetHvparamsString(hv)
2528
    return hvparams
2529

    
2530
  @staticmethod
2531
  def _ExtendByAllHvparamsStrings(ssconf_values, all_hvparams):
2532
    """Extends the ssconf_values dictionary by hvparams.
2533

2534
    @type ssconf_values: dict of strings
2535
    @param ssconf_values: dictionary mapping ssconf_keys to strings
2536
      representing the content of ssconf files
2537
    @type all_hvparams: dict of strings
2538
    @param all_hvparams: dictionary mapping hypervisor names to a string
2539
      representation of their hvparams
2540
    @rtype: same as ssconf_values
2541
    @returns: the ssconf_values dictionary extended by hvparams
2542

2543
    """
2544
    for hv in all_hvparams:
2545
      ssconf_key = constants.SS_HVPARAMS_PREF + hv
2546
      ssconf_values[ssconf_key] = all_hvparams[hv]
2547
    return ssconf_values
2548

    
2549
  def _UnlockedGetSsconfValues(self):
2550
    """Return the values needed by ssconf.
2551

2552
    @rtype: dict
2553
    @return: a dictionary with keys the ssconf names and values their
2554
        associated value
2555

2556
    """
2557
    fn = "\n".join
2558
    instance_names = utils.NiceSort(
2559
                       [inst.name for inst in
2560
                        self._UnlockedGetAllInstancesInfo().values()])
2561
    node_infos = self._UnlockedGetAllNodesInfo().values()
2562
    node_names = [node.name for node in node_infos]
2563
    node_pri_ips = ["%s %s" % (ninfo.name, ninfo.primary_ip)
2564
                    for ninfo in node_infos]
2565
    node_snd_ips = ["%s %s" % (ninfo.name, ninfo.secondary_ip)
2566
                    for ninfo in node_infos]
2567

    
2568
    instance_data = fn(instance_names)
2569
    off_data = fn(node.name for node in node_infos if node.offline)
2570
    on_data = fn(node.name for node in node_infos if not node.offline)
2571
    mc_data = fn(node.name for node in node_infos if node.master_candidate)
2572
    mc_ips_data = fn(node.primary_ip for node in node_infos
2573
                     if node.master_candidate)
2574
    node_data = fn(node_names)
2575
    node_pri_ips_data = fn(node_pri_ips)
2576
    node_snd_ips_data = fn(node_snd_ips)
2577

    
2578
    cluster = self._config_data.cluster
2579
    cluster_tags = fn(cluster.GetTags())
2580

    
2581
    hypervisor_list = fn(cluster.enabled_hypervisors)
2582
    all_hvparams = self._GetAllHvparamsStrings(constants.HYPER_TYPES)
2583

    
2584
    uid_pool = uidpool.FormatUidPool(cluster.uid_pool, separator="\n")
2585

    
2586
    nodegroups = ["%s %s" % (nodegroup.uuid, nodegroup.name) for nodegroup in
2587
                  self._config_data.nodegroups.values()]
2588
    nodegroups_data = fn(utils.NiceSort(nodegroups))
2589
    networks = ["%s %s" % (net.uuid, net.name) for net in
2590
                self._config_data.networks.values()]
2591
    networks_data = fn(utils.NiceSort(networks))
2592

    
2593
    ssconf_values = {
2594
      constants.SS_CLUSTER_NAME: cluster.cluster_name,
2595
      constants.SS_CLUSTER_TAGS: cluster_tags,
2596
      constants.SS_FILE_STORAGE_DIR: cluster.file_storage_dir,
2597
      constants.SS_SHARED_FILE_STORAGE_DIR: cluster.shared_file_storage_dir,
2598
      constants.SS_MASTER_CANDIDATES: mc_data,
2599
      constants.SS_MASTER_CANDIDATES_IPS: mc_ips_data,
2600
      constants.SS_MASTER_IP: cluster.master_ip,
2601
      constants.SS_MASTER_NETDEV: cluster.master_netdev,
2602
      constants.SS_MASTER_NETMASK: str(cluster.master_netmask),
2603
      constants.SS_MASTER_NODE: self._UnlockedGetNodeName(cluster.master_node),
2604
      constants.SS_NODE_LIST: node_data,
2605
      constants.SS_NODE_PRIMARY_IPS: node_pri_ips_data,
2606
      constants.SS_NODE_SECONDARY_IPS: node_snd_ips_data,
2607
      constants.SS_OFFLINE_NODES: off_data,
2608
      constants.SS_ONLINE_NODES: on_data,
2609
      constants.SS_PRIMARY_IP_FAMILY: str(cluster.primary_ip_family),
2610
      constants.SS_INSTANCE_LIST: instance_data,
2611
      constants.SS_RELEASE_VERSION: constants.RELEASE_VERSION,
2612
      constants.SS_HYPERVISOR_LIST: hypervisor_list,
2613
      constants.SS_MAINTAIN_NODE_HEALTH: str(cluster.maintain_node_health),
2614
      constants.SS_UID_POOL: uid_pool,
2615
      constants.SS_NODEGROUPS: nodegroups_data,
2616
      constants.SS_NETWORKS: networks_data,
2617
      }
2618
    ssconf_values = self._ExtendByAllHvparamsStrings(ssconf_values,
2619
                                                     all_hvparams)
2620
    bad_values = [(k, v) for k, v in ssconf_values.items()
2621
                  if not isinstance(v, (str, basestring))]
2622
    if bad_values:
2623
      err = utils.CommaJoin("%s=%s" % (k, v) for k, v in bad_values)
2624
      raise errors.ConfigurationError("Some ssconf key(s) have non-string"
2625
                                      " values: %s" % err)
2626
    return ssconf_values
2627

    
2628
  @locking.ssynchronized(_config_lock, shared=1)
2629
  def GetSsconfValues(self):
2630
    """Wrapper using lock around _UnlockedGetSsconf().
2631

2632
    """
2633
    return self._UnlockedGetSsconfValues()
2634

    
2635
  @locking.ssynchronized(_config_lock, shared=1)
2636
  def GetVGName(self):
2637
    """Return the volume group name.
2638

2639
    """
2640
    return self._config_data.cluster.volume_group_name
2641

    
2642
  @locking.ssynchronized(_config_lock)
2643
  def SetVGName(self, vg_name):
2644
    """Set the volume group name.
2645

2646
    """
2647
    self._config_data.cluster.volume_group_name = vg_name
2648
    self._config_data.cluster.serial_no += 1
2649
    self._WriteConfig()
2650

    
2651
  @locking.ssynchronized(_config_lock, shared=1)
2652
  def GetDRBDHelper(self):
2653
    """Return DRBD usermode helper.
2654

2655
    """
2656
    return self._config_data.cluster.drbd_usermode_helper
2657

    
2658
  @locking.ssynchronized(_config_lock)
2659
  def SetDRBDHelper(self, drbd_helper):
2660
    """Set DRBD usermode helper.
2661

2662
    """
2663
    self._config_data.cluster.drbd_usermode_helper = drbd_helper
2664
    self._config_data.cluster.serial_no += 1
2665
    self._WriteConfig()
2666

    
2667
  @locking.ssynchronized(_config_lock, shared=1)
2668
  def GetMACPrefix(self):
2669
    """Return the mac prefix.
2670

2671
    """
2672
    return self._config_data.cluster.mac_prefix
2673

    
2674
  @locking.ssynchronized(_config_lock, shared=1)
2675
  def GetClusterInfo(self):
2676
    """Returns information about the cluster
2677

2678
    @rtype: L{objects.Cluster}
2679
    @return: the cluster object
2680

2681
    """
2682
    return self._config_data.cluster
2683

    
2684
  @locking.ssynchronized(_config_lock, shared=1)
2685
  def HasAnyDiskOfType(self, dev_type):
2686
    """Check if in there is at disk of the given type in the configuration.
2687

2688
    """
2689
    return self._config_data.HasAnyDiskOfType(dev_type)
2690

    
2691
  @locking.ssynchronized(_config_lock)
2692
  def Update(self, target, feedback_fn, ec_id=None):
2693
    """Notify function to be called after updates.
2694

2695
    This function must be called when an object (as returned by
2696
    GetInstanceInfo, GetNodeInfo, GetCluster) has been updated and the
2697
    caller wants the modifications saved to the backing store. Note
2698
    that all modified objects will be saved, but the target argument
2699
    is the one the caller wants to ensure that it's saved.
2700

2701
    @param target: an instance of either L{objects.Cluster},
2702
        L{objects.Node} or L{objects.Instance} which is existing in
2703
        the cluster
2704
    @param feedback_fn: Callable feedback function
2705

2706
    """
2707
    if self._config_data is None:
2708
      raise errors.ProgrammerError("Configuration file not read,"
2709
                                   " cannot save.")
2710
    update_serial = False
2711
    if isinstance(target, objects.Cluster):
2712
      test = target == self._config_data.cluster
2713
    elif isinstance(target, objects.Node):
2714
      test = target in self._config_data.nodes.values()
2715
      update_serial = True
2716
    elif isinstance(target, objects.Instance):
2717
      test = target in self._config_data.instances.values()
2718
    elif isinstance(target, objects.NodeGroup):
2719
      test = target in self._config_data.nodegroups.values()
2720
    elif isinstance(target, objects.Network):
2721
      test = target in self._config_data.networks.values()
2722
    else:
2723
      raise errors.ProgrammerError("Invalid object type (%s) passed to"
2724
                                   " ConfigWriter.Update" % type(target))
2725
    if not test:
2726
      raise errors.ConfigurationError("Configuration updated since object"
2727
                                      " has been read or unknown object")
2728
    target.serial_no += 1
2729
    target.mtime = now = time.time()
2730

    
2731
    if update_serial:
2732
      # for node updates, we need to increase the cluster serial too
2733
      self._config_data.cluster.serial_no += 1
2734
      self._config_data.cluster.mtime = now
2735

    
2736
    if isinstance(target, objects.Instance):
2737
      self._UnlockedReleaseDRBDMinors(target.uuid)
2738

    
2739
    if ec_id is not None:
2740
      # Commit all ips reserved by OpInstanceSetParams and OpGroupSetParams
2741
      self._UnlockedCommitTemporaryIps(ec_id)
2742

    
2743
    self._WriteConfig(feedback_fn=feedback_fn)
2744

    
2745
  @locking.ssynchronized(_config_lock)
2746
  def DropECReservations(self, ec_id):
2747
    """Drop per-execution-context reservations
2748

2749
    """
2750
    for rm in self._all_rms:
2751
      rm.DropECReservations(ec_id)
2752

    
2753
  @locking.ssynchronized(_config_lock, shared=1)
2754
  def GetAllNetworksInfo(self):
2755
    """Get configuration info of all the networks.
2756

2757
    """
2758
    return dict(self._config_data.networks)
2759

    
2760
  def _UnlockedGetNetworkList(self):
2761
    """Get the list of networks.
2762

2763
    This function is for internal use, when the config lock is already held.
2764

2765
    """
2766
    return self._config_data.networks.keys()
2767

    
2768
  @locking.ssynchronized(_config_lock, shared=1)
2769
  def GetNetworkList(self):
2770
    """Get the list of networks.
2771

2772
    @return: array of networks, ex. ["main", "vlan100", "200]
2773

2774
    """
2775
    return self._UnlockedGetNetworkList()
2776

    
2777
  @locking.ssynchronized(_config_lock, shared=1)
2778
  def GetNetworkNames(self):
2779
    """Get a list of network names
2780

2781
    """
2782
    names = [net.name
2783
             for net in self._config_data.networks.values()]
2784
    return names
2785

    
2786
  def _UnlockedGetNetwork(self, uuid):
2787
    """Returns information about a network.
2788

2789
    This function is for internal use, when the config lock is already held.
2790

2791
    """
2792
    if uuid not in self._config_data.networks:
2793
      return None
2794

    
2795
    return self._config_data.networks[uuid]
2796

    
2797
  @locking.ssynchronized(_config_lock, shared=1)
2798
  def GetNetwork(self, uuid):
2799
    """Returns information about a network.
2800

2801
    It takes the information from the configuration file.
2802

2803
    @param uuid: UUID of the network
2804

2805
    @rtype: L{objects.Network}
2806
    @return: the network object
2807

2808
    """
2809
    return self._UnlockedGetNetwork(uuid)
2810

    
2811
  @locking.ssynchronized(_config_lock)
2812
  def AddNetwork(self, net, ec_id, check_uuid=True):
2813
    """Add a network to the configuration.
2814

2815
    @type net: L{objects.Network}
2816
    @param net: the Network object to add
2817
    @type ec_id: string
2818
    @param ec_id: unique id for the job to use when creating a missing UUID
2819

2820
    """
2821
    self._UnlockedAddNetwork(net, ec_id, check_uuid)
2822
    self._WriteConfig()
2823

    
2824
  def _UnlockedAddNetwork(self, net, ec_id, check_uuid):
2825
    """Add a network to the configuration.
2826

2827
    """
2828
    logging.info("Adding network %s to configuration", net.name)
2829

    
2830
    if check_uuid:
2831
      self._EnsureUUID(net, ec_id)
2832

    
2833
    net.serial_no = 1
2834
    net.ctime = net.mtime = time.time()
2835
    self._config_data.networks[net.uuid] = net
2836
    self._config_data.cluster.serial_no += 1
2837

    
2838
  def _UnlockedLookupNetwork(self, target):
2839
    """Lookup a network's UUID.
2840

2841
    @type target: string
2842
    @param target: network name or UUID
2843
    @rtype: string
2844
    @return: network UUID
2845
    @raises errors.OpPrereqError: when the target network cannot be found
2846

2847
    """
2848
    if target is None:
2849
      return None
2850
    if target in self._config_data.networks:
2851
      return target
2852
    for net in self._config_data.networks.values():
2853
      if net.name == target:
2854
        return net.uuid
2855
    raise errors.OpPrereqError("Network '%s' not found" % target,
2856
                               errors.ECODE_NOENT)
2857

    
2858
  @locking.ssynchronized(_config_lock, shared=1)
2859
  def LookupNetwork(self, target):
2860
    """Lookup a network's UUID.
2861

2862
    This function is just a wrapper over L{_UnlockedLookupNetwork}.
2863

2864
    @type target: string
2865
    @param target: network name or UUID
2866
    @rtype: string
2867
    @return: network UUID
2868

2869
    """
2870
    return self._UnlockedLookupNetwork(target)
2871

    
2872
  @locking.ssynchronized(_config_lock)
2873
  def RemoveNetwork(self, network_uuid):
2874
    """Remove a network from the configuration.
2875

2876
    @type network_uuid: string
2877
    @param network_uuid: the UUID of the network to remove
2878

2879
    """
2880
    logging.info("Removing network %s from configuration", network_uuid)
2881

    
2882
    if network_uuid not in self._config_data.networks:
2883
      raise errors.ConfigurationError("Unknown network '%s'" % network_uuid)
2884

    
2885
    del self._config_data.networks[network_uuid]
2886
    self._config_data.cluster.serial_no += 1
2887
    self._WriteConfig()
2888

    
2889
  def _UnlockedGetGroupNetParams(self, net_uuid, node_uuid):
2890
    """Get the netparams (mode, link) of a network.
2891

2892
    Get a network's netparams for a given node.
2893

2894
    @type net_uuid: string
2895
    @param net_uuid: network uuid
2896
    @type node_uuid: string
2897
    @param node_uuid: node UUID
2898
    @rtype: dict or None
2899
    @return: netparams
2900

2901
    """
2902
    node_info = self._UnlockedGetNodeInfo(node_uuid)
2903
    nodegroup_info = self._UnlockedGetNodeGroup(node_info.group)
2904
    netparams = nodegroup_info.networks.get(net_uuid, None)
2905

    
2906
    return netparams
2907

    
2908
  @locking.ssynchronized(_config_lock, shared=1)
2909
  def GetGroupNetParams(self, net_uuid, node_uuid):
2910
    """Locking wrapper of _UnlockedGetGroupNetParams()
2911

2912
    """
2913
    return self._UnlockedGetGroupNetParams(net_uuid, node_uuid)
2914

    
2915
  @locking.ssynchronized(_config_lock, shared=1)
2916
  def CheckIPInNodeGroup(self, ip, node_uuid):
2917
    """Check IP uniqueness in nodegroup.
2918

2919
    Check networks that are connected in the node's node group
2920
    if ip is contained in any of them. Used when creating/adding
2921
    a NIC to ensure uniqueness among nodegroups.
2922

2923
    @type ip: string
2924
    @param ip: ip address
2925
    @type node_uuid: string
2926
    @param node_uuid: node UUID
2927
    @rtype: (string, dict) or (None, None)
2928
    @return: (network name, netparams)
2929

2930
    """
2931
    if ip is None:
2932
      return (None, None)
2933
    node_info = self._UnlockedGetNodeInfo(node_uuid)
2934
    nodegroup_info = self._UnlockedGetNodeGroup(node_info.group)
2935
    for net_uuid in nodegroup_info.networks.keys():
2936
      net_info = self._UnlockedGetNetwork(net_uuid)
2937
      pool = network.AddressPool(net_info)
2938
      if pool.Contains(ip):
2939
        return (net_info.name, nodegroup_info.networks[net_uuid])
2940

    
2941
    return (None, None)