Statistics
| Branch: | Tag: | Revision:

root / lib / config.py @ 738436bf

History | View | Annotate | Download (95.9 kB)

1
#
2
#
3

    
4
# Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013 Google Inc.
5
#
6
# This program is free software; you can redistribute it and/or modify
7
# it under the terms of the GNU General Public License as published by
8
# the Free Software Foundation; either version 2 of the License, or
9
# (at your option) any later version.
10
#
11
# This program is distributed in the hope that it will be useful, but
12
# WITHOUT ANY WARRANTY; without even the implied warranty of
13
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14
# General Public License for more details.
15
#
16
# You should have received a copy of the GNU General Public License
17
# along with this program; if not, write to the Free Software
18
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
19
# 02110-1301, USA.
20

    
21

    
22
"""Configuration management for Ganeti
23

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

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

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

32
"""
33

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

    
37
import copy
38
import os
39
import random
40
import logging
41
import time
42
import itertools
43

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

    
57

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

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

    
63

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

67
  This only verifies the version of the configuration.
68

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

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

    
76

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

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

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

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

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

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

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

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

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

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

126
    """
127
    assert callable(generate_one_fn)
128

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

    
142

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

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

    
149

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

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

159
  """
160
  result = []
161

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

    
167
  return result
168

    
169

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

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

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

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

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

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

216
    """
217
    self._context = context
218

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
283
    return prefix
284

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

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

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

    
301
    return GenMac
302

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

307
    This should check the current instances for duplicates.
308

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

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

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

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

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

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

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

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

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

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

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

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

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

363
    This is just a wrapper around _UnlockedReleaseIp.
364

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

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

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

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

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

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

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

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

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

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

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

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

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

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

430
    This checks the current disks for duplicates.
431

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
521
    return result
522

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

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

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

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

    
543
    return result
544

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
850
    return result
851

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

    
957
    self._WriteConfig()
958
    return port
959

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

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

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

    
990
    duplicates = []
991
    my_dict = dict((node_uuid, {}) for node_uuid in self._config_data.nodes)
992
    for instance in self._config_data.instances.itervalues():
993
      for disk in instance.disks:
994
        duplicates.extend(_AppendUsedMinors(self._UnlockedGetNodeName,
995
                                            instance, disk, my_dict))
996
    for (node_uuid, minor), inst_uuid in self._temporary_drbds.iteritems():
997
      if minor in my_dict[node_uuid] and my_dict[node_uuid][minor] != inst_uuid:
998
        duplicates.append((node_uuid, minor, inst_uuid,
999
                           my_dict[node_uuid][minor]))
1000
      else:
1001
        my_dict[node_uuid][minor] = inst_uuid
1002
    return my_dict, duplicates
1003

    
1004
  @locking.ssynchronized(_config_lock)
1005
  def ComputeDRBDMap(self):
1006
    """Compute the used DRBD minor/nodes.
1007

1008
    This is just a wrapper over L{_UnlockedComputeDRBDMap}.
1009

1010
    @return: dictionary of node_uuid: dict of minor: instance_uuid;
1011
        the returned dict will have all the nodes in it (even if with
1012
        an empty list).
1013

1014
    """
1015
    d_map, duplicates = self._UnlockedComputeDRBDMap()
1016
    if duplicates:
1017
      raise errors.ConfigurationError("Duplicate DRBD ports detected: %s" %
1018
                                      str(duplicates))
1019
    return d_map
1020

    
1021
  @locking.ssynchronized(_config_lock)
1022
  def AllocateDRBDMinor(self, node_uuids, inst_uuid):
1023
    """Allocate a drbd minor.
1024

1025
    The free minor will be automatically computed from the existing
1026
    devices. A node can be given multiple times in order to allocate
1027
    multiple minors. The result is the list of minors, in the same
1028
    order as the passed nodes.
1029

1030
    @type inst_uuid: string
1031
    @param inst_uuid: the instance for which we allocate minors
1032

1033
    """
1034
    assert isinstance(inst_uuid, basestring), \
1035
           "Invalid argument '%s' passed to AllocateDRBDMinor" % inst_uuid
1036

    
1037
    d_map, duplicates = self._UnlockedComputeDRBDMap()
1038
    if duplicates:
1039
      raise errors.ConfigurationError("Duplicate DRBD ports detected: %s" %
1040
                                      str(duplicates))
1041
    result = []
1042
    for nuuid in node_uuids:
1043
      ndata = d_map[nuuid]
1044
      if not ndata:
1045
        # no minors used, we can start at 0
1046
        result.append(0)
1047
        ndata[0] = inst_uuid
1048
        self._temporary_drbds[(nuuid, 0)] = inst_uuid
1049
        continue
1050
      keys = ndata.keys()
1051
      keys.sort()
1052
      ffree = utils.FirstFree(keys)
1053
      if ffree is None:
1054
        # return the next minor
1055
        # TODO: implement high-limit check
1056
        minor = keys[-1] + 1
1057
      else:
1058
        minor = ffree
1059
      # double-check minor against current instances
1060
      assert minor not in d_map[nuuid], \
1061
             ("Attempt to reuse allocated DRBD minor %d on node %s,"
1062
              " already allocated to instance %s" %
1063
              (minor, nuuid, d_map[nuuid][minor]))
1064
      ndata[minor] = inst_uuid
1065
      # double-check minor against reservation
1066
      r_key = (nuuid, minor)
1067
      assert r_key not in self._temporary_drbds, \
1068
             ("Attempt to reuse reserved DRBD minor %d on node %s,"
1069
              " reserved for instance %s" %
1070
              (minor, nuuid, self._temporary_drbds[r_key]))
1071
      self._temporary_drbds[r_key] = inst_uuid
1072
      result.append(minor)
1073
    logging.debug("Request to allocate drbd minors, input: %s, returning %s",
1074
                  node_uuids, result)
1075
    return result
1076

    
1077
  def _UnlockedReleaseDRBDMinors(self, inst_uuid):
1078
    """Release temporary drbd minors allocated for a given instance.
1079

1080
    @type inst_uuid: string
1081
    @param inst_uuid: the instance for which temporary minors should be
1082
                      released
1083

1084
    """
1085
    assert isinstance(inst_uuid, basestring), \
1086
           "Invalid argument passed to ReleaseDRBDMinors"
1087
    for key, uuid in self._temporary_drbds.items():
1088
      if uuid == inst_uuid:
1089
        del self._temporary_drbds[key]
1090

    
1091
  @locking.ssynchronized(_config_lock)
1092
  def ReleaseDRBDMinors(self, inst_uuid):
1093
    """Release temporary drbd minors allocated for a given instance.
1094

1095
    This should be called on the error paths, on the success paths
1096
    it's automatically called by the ConfigWriter add and update
1097
    functions.
1098

1099
    This function is just a wrapper over L{_UnlockedReleaseDRBDMinors}.
1100

1101
    @type inst_uuid: string
1102
    @param inst_uuid: the instance for which temporary minors should be
1103
                      released
1104

1105
    """
1106
    self._UnlockedReleaseDRBDMinors(inst_uuid)
1107

    
1108
  @locking.ssynchronized(_config_lock, shared=1)
1109
  def GetConfigVersion(self):
1110
    """Get the configuration version.
1111

1112
    @return: Config version
1113

1114
    """
1115
    return self._config_data.version
1116

    
1117
  @locking.ssynchronized(_config_lock, shared=1)
1118
  def GetClusterName(self):
1119
    """Get cluster name.
1120

1121
    @return: Cluster name
1122

1123
    """
1124
    return self._config_data.cluster.cluster_name
1125

    
1126
  @locking.ssynchronized(_config_lock, shared=1)
1127
  def GetMasterNode(self):
1128
    """Get the UUID of the master node for this cluster.
1129

1130
    @return: Master node UUID
1131

1132
    """
1133
    return self._config_data.cluster.master_node
1134

    
1135
  @locking.ssynchronized(_config_lock, shared=1)
1136
  def GetMasterNodeName(self):
1137
    """Get the hostname of the master node for this cluster.
1138

1139
    @return: Master node hostname
1140

1141
    """
1142
    return self._UnlockedGetNodeName(self._config_data.cluster.master_node)
1143

    
1144
  @locking.ssynchronized(_config_lock, shared=1)
1145
  def GetMasterIP(self):
1146
    """Get the IP of the master node for this cluster.
1147

1148
    @return: Master IP
1149

1150
    """
1151
    return self._config_data.cluster.master_ip
1152

    
1153
  @locking.ssynchronized(_config_lock, shared=1)
1154
  def GetMasterNetdev(self):
1155
    """Get the master network device for this cluster.
1156

1157
    """
1158
    return self._config_data.cluster.master_netdev
1159

    
1160
  @locking.ssynchronized(_config_lock, shared=1)
1161
  def GetMasterNetmask(self):
1162
    """Get the netmask of the master node for this cluster.
1163

1164
    """
1165
    return self._config_data.cluster.master_netmask
1166

    
1167
  @locking.ssynchronized(_config_lock, shared=1)
1168
  def GetUseExternalMipScript(self):
1169
    """Get flag representing whether to use the external master IP setup script.
1170

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

    
1174
  @locking.ssynchronized(_config_lock, shared=1)
1175
  def GetFileStorageDir(self):
1176
    """Get the file storage dir for this cluster.
1177

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

    
1181
  @locking.ssynchronized(_config_lock, shared=1)
1182
  def GetSharedFileStorageDir(self):
1183
    """Get the shared file storage dir for this cluster.
1184

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

    
1188
  @locking.ssynchronized(_config_lock, shared=1)
1189
  def GetHypervisorType(self):
1190
    """Get the hypervisor type for this cluster.
1191

1192
    """
1193
    return self._config_data.cluster.enabled_hypervisors[0]
1194

    
1195
  @locking.ssynchronized(_config_lock, shared=1)
1196
  def GetHostKey(self):
1197
    """Return the rsa hostkey from the config.
1198

1199
    @rtype: string
1200
    @return: the rsa hostkey
1201

1202
    """
1203
    return self._config_data.cluster.rsahostkeypub
1204

    
1205
  @locking.ssynchronized(_config_lock, shared=1)
1206
  def GetDefaultIAllocator(self):
1207
    """Get the default instance allocator for this cluster.
1208

1209
    """
1210
    return self._config_data.cluster.default_iallocator
1211

    
1212
  @locking.ssynchronized(_config_lock, shared=1)
1213
  def GetPrimaryIPFamily(self):
1214
    """Get cluster primary ip family.
1215

1216
    @return: primary ip family
1217

1218
    """
1219
    return self._config_data.cluster.primary_ip_family
1220

    
1221
  @locking.ssynchronized(_config_lock, shared=1)
1222
  def GetMasterNetworkParameters(self):
1223
    """Get network parameters of the master node.
1224

1225
    @rtype: L{object.MasterNetworkParameters}
1226
    @return: network parameters of the master node
1227

1228
    """
1229
    cluster = self._config_data.cluster
1230
    result = objects.MasterNetworkParameters(
1231
      uuid=cluster.master_node, ip=cluster.master_ip,
1232
      netmask=cluster.master_netmask, netdev=cluster.master_netdev,
1233
      ip_family=cluster.primary_ip_family)
1234

    
1235
    return result
1236

    
1237
  @locking.ssynchronized(_config_lock)
1238
  def AddNodeGroup(self, group, ec_id, check_uuid=True):
1239
    """Add a node group to the configuration.
1240

1241
    This method calls group.UpgradeConfig() to fill any missing attributes
1242
    according to their default values.
1243

1244
    @type group: L{objects.NodeGroup}
1245
    @param group: the NodeGroup object to add
1246
    @type ec_id: string
1247
    @param ec_id: unique id for the job to use when creating a missing UUID
1248
    @type check_uuid: bool
1249
    @param check_uuid: add an UUID to the group if it doesn't have one or, if
1250
                       it does, ensure that it does not exist in the
1251
                       configuration already
1252

1253
    """
1254
    self._UnlockedAddNodeGroup(group, ec_id, check_uuid)
1255
    self._WriteConfig()
1256

    
1257
  def _UnlockedAddNodeGroup(self, group, ec_id, check_uuid):
1258
    """Add a node group to the configuration.
1259

1260
    """
1261
    logging.info("Adding node group %s to configuration", group.name)
1262

    
1263
    # Some code might need to add a node group with a pre-populated UUID
1264
    # generated with ConfigWriter.GenerateUniqueID(). We allow them to bypass
1265
    # the "does this UUID" exist already check.
1266
    if check_uuid:
1267
      self._EnsureUUID(group, ec_id)
1268

    
1269
    try:
1270
      existing_uuid = self._UnlockedLookupNodeGroup(group.name)
1271
    except errors.OpPrereqError:
1272
      pass
1273
    else:
1274
      raise errors.OpPrereqError("Desired group name '%s' already exists as a"
1275
                                 " node group (UUID: %s)" %
1276
                                 (group.name, existing_uuid),
1277
                                 errors.ECODE_EXISTS)
1278

    
1279
    group.serial_no = 1
1280
    group.ctime = group.mtime = time.time()
1281
    group.UpgradeConfig()
1282

    
1283
    self._config_data.nodegroups[group.uuid] = group
1284
    self._config_data.cluster.serial_no += 1
1285

    
1286
  @locking.ssynchronized(_config_lock)
1287
  def RemoveNodeGroup(self, group_uuid):
1288
    """Remove a node group from the configuration.
1289

1290
    @type group_uuid: string
1291
    @param group_uuid: the UUID of the node group to remove
1292

1293
    """
1294
    logging.info("Removing node group %s from configuration", group_uuid)
1295

    
1296
    if group_uuid not in self._config_data.nodegroups:
1297
      raise errors.ConfigurationError("Unknown node group '%s'" % group_uuid)
1298

    
1299
    assert len(self._config_data.nodegroups) != 1, \
1300
            "Group '%s' is the only group, cannot be removed" % group_uuid
1301

    
1302
    del self._config_data.nodegroups[group_uuid]
1303
    self._config_data.cluster.serial_no += 1
1304
    self._WriteConfig()
1305

    
1306
  def _UnlockedLookupNodeGroup(self, target):
1307
    """Lookup a node group's UUID.
1308

1309
    @type target: string or None
1310
    @param target: group name or UUID or None to look for the default
1311
    @rtype: string
1312
    @return: nodegroup UUID
1313
    @raises errors.OpPrereqError: when the target group cannot be found
1314

1315
    """
1316
    if target is None:
1317
      if len(self._config_data.nodegroups) != 1:
1318
        raise errors.OpPrereqError("More than one node group exists. Target"
1319
                                   " group must be specified explicitly.")
1320
      else:
1321
        return self._config_data.nodegroups.keys()[0]
1322
    if target in self._config_data.nodegroups:
1323
      return target
1324
    for nodegroup in self._config_data.nodegroups.values():
1325
      if nodegroup.name == target:
1326
        return nodegroup.uuid
1327
    raise errors.OpPrereqError("Node group '%s' not found" % target,
1328
                               errors.ECODE_NOENT)
1329

    
1330
  @locking.ssynchronized(_config_lock, shared=1)
1331
  def LookupNodeGroup(self, target):
1332
    """Lookup a node group's UUID.
1333

1334
    This function is just a wrapper over L{_UnlockedLookupNodeGroup}.
1335

1336
    @type target: string or None
1337
    @param target: group name or UUID or None to look for the default
1338
    @rtype: string
1339
    @return: nodegroup UUID
1340

1341
    """
1342
    return self._UnlockedLookupNodeGroup(target)
1343

    
1344
  def _UnlockedGetNodeGroup(self, uuid):
1345
    """Lookup a node group.
1346

1347
    @type uuid: string
1348
    @param uuid: group UUID
1349
    @rtype: L{objects.NodeGroup} or None
1350
    @return: nodegroup object, or None if not found
1351

1352
    """
1353
    if uuid not in self._config_data.nodegroups:
1354
      return None
1355

    
1356
    return self._config_data.nodegroups[uuid]
1357

    
1358
  @locking.ssynchronized(_config_lock, shared=1)
1359
  def GetNodeGroup(self, uuid):
1360
    """Lookup a node group.
1361

1362
    @type uuid: string
1363
    @param uuid: group UUID
1364
    @rtype: L{objects.NodeGroup} or None
1365
    @return: nodegroup object, or None if not found
1366

1367
    """
1368
    return self._UnlockedGetNodeGroup(uuid)
1369

    
1370
  @locking.ssynchronized(_config_lock, shared=1)
1371
  def GetAllNodeGroupsInfo(self):
1372
    """Get the configuration of all node groups.
1373

1374
    """
1375
    return dict(self._config_data.nodegroups)
1376

    
1377
  @locking.ssynchronized(_config_lock, shared=1)
1378
  def GetNodeGroupList(self):
1379
    """Get a list of node groups.
1380

1381
    """
1382
    return self._config_data.nodegroups.keys()
1383

    
1384
  @locking.ssynchronized(_config_lock, shared=1)
1385
  def GetNodeGroupMembersByNodes(self, nodes):
1386
    """Get nodes which are member in the same nodegroups as the given nodes.
1387

1388
    """
1389
    ngfn = lambda node_uuid: self._UnlockedGetNodeInfo(node_uuid).group
1390
    return frozenset(member_uuid
1391
                     for node_uuid in nodes
1392
                     for member_uuid in
1393
                       self._UnlockedGetNodeGroup(ngfn(node_uuid)).members)
1394

    
1395
  @locking.ssynchronized(_config_lock, shared=1)
1396
  def GetMultiNodeGroupInfo(self, group_uuids):
1397
    """Get the configuration of multiple node groups.
1398

1399
    @param group_uuids: List of node group UUIDs
1400
    @rtype: list
1401
    @return: List of tuples of (group_uuid, group_info)
1402

1403
    """
1404
    return [(uuid, self._UnlockedGetNodeGroup(uuid)) for uuid in group_uuids]
1405

    
1406
  @locking.ssynchronized(_config_lock)
1407
  def AddInstance(self, instance, ec_id):
1408
    """Add an instance to the config.
1409

1410
    This should be used after creating a new instance.
1411

1412
    @type instance: L{objects.Instance}
1413
    @param instance: the instance object
1414

1415
    """
1416
    if not isinstance(instance, objects.Instance):
1417
      raise errors.ProgrammerError("Invalid type passed to AddInstance")
1418

    
1419
    if instance.disk_template != constants.DT_DISKLESS:
1420
      all_lvs = instance.MapLVsByNode()
1421
      logging.info("Instance '%s' DISK_LAYOUT: %s", instance.name, all_lvs)
1422

    
1423
    all_macs = self._AllMACs()
1424
    for nic in instance.nics:
1425
      if nic.mac in all_macs:
1426
        raise errors.ConfigurationError("Cannot add instance %s:"
1427
                                        " MAC address '%s' already in use." %
1428
                                        (instance.name, nic.mac))
1429

    
1430
    self._CheckUniqueUUID(instance, include_temporary=False)
1431

    
1432
    instance.serial_no = 1
1433
    instance.ctime = instance.mtime = time.time()
1434
    self._config_data.instances[instance.uuid] = instance
1435
    self._config_data.cluster.serial_no += 1
1436
    self._UnlockedReleaseDRBDMinors(instance.uuid)
1437
    self._UnlockedCommitTemporaryIps(ec_id)
1438
    self._WriteConfig()
1439

    
1440
  def _EnsureUUID(self, item, ec_id):
1441
    """Ensures a given object has a valid UUID.
1442

1443
    @param item: the instance or node to be checked
1444
    @param ec_id: the execution context id for the uuid reservation
1445

1446
    """
1447
    if not item.uuid:
1448
      item.uuid = self._GenerateUniqueID(ec_id)
1449
    else:
1450
      self._CheckUniqueUUID(item, include_temporary=True)
1451

    
1452
  def _CheckUniqueUUID(self, item, include_temporary):
1453
    """Checks that the UUID of the given object is unique.
1454

1455
    @param item: the instance or node to be checked
1456
    @param include_temporary: whether temporarily generated UUID's should be
1457
              included in the check. If the UUID of the item to be checked is
1458
              a temporarily generated one, this has to be C{False}.
1459

1460
    """
1461
    if not item.uuid:
1462
      raise errors.ConfigurationError("'%s' must have an UUID" % (item.name,))
1463
    if item.uuid in self._AllIDs(include_temporary=include_temporary):
1464
      raise errors.ConfigurationError("Cannot add '%s': UUID %s already"
1465
                                      " in use" % (item.name, item.uuid))
1466

    
1467
  def _SetInstanceStatus(self, inst_uuid, status, disks_active):
1468
    """Set the instance's status to a given value.
1469

1470
    """
1471
    if inst_uuid not in self._config_data.instances:
1472
      raise errors.ConfigurationError("Unknown instance '%s'" %
1473
                                      inst_uuid)
1474
    instance = self._config_data.instances[inst_uuid]
1475

    
1476
    if status is None:
1477
      status = instance.admin_state
1478
    if disks_active is None:
1479
      disks_active = instance.disks_active
1480

    
1481
    assert status in constants.ADMINST_ALL, \
1482
           "Invalid status '%s' passed to SetInstanceStatus" % (status,)
1483

    
1484
    if instance.admin_state != status or \
1485
       instance.disks_active != disks_active:
1486
      instance.admin_state = status
1487
      instance.disks_active = disks_active
1488
      instance.serial_no += 1
1489
      instance.mtime = time.time()
1490
      self._WriteConfig()
1491

    
1492
  @locking.ssynchronized(_config_lock)
1493
  def MarkInstanceUp(self, inst_uuid):
1494
    """Mark the instance status to up in the config.
1495

1496
    This also sets the instance disks active flag.
1497

1498
    """
1499
    self._SetInstanceStatus(inst_uuid, constants.ADMINST_UP, True)
1500

    
1501
  @locking.ssynchronized(_config_lock)
1502
  def MarkInstanceOffline(self, inst_uuid):
1503
    """Mark the instance status to down in the config.
1504

1505
    This also clears the instance disks active flag.
1506

1507
    """
1508
    self._SetInstanceStatus(inst_uuid, constants.ADMINST_OFFLINE, False)
1509

    
1510
  @locking.ssynchronized(_config_lock)
1511
  def RemoveInstance(self, inst_uuid):
1512
    """Remove the instance from the configuration.
1513

1514
    """
1515
    if inst_uuid not in self._config_data.instances:
1516
      raise errors.ConfigurationError("Unknown instance '%s'" % inst_uuid)
1517

    
1518
    # If a network port has been allocated to the instance,
1519
    # return it to the pool of free ports.
1520
    inst = self._config_data.instances[inst_uuid]
1521
    network_port = getattr(inst, "network_port", None)
1522
    if network_port is not None:
1523
      self._config_data.cluster.tcpudp_port_pool.add(network_port)
1524

    
1525
    instance = self._UnlockedGetInstanceInfo(inst_uuid)
1526

    
1527
    for nic in instance.nics:
1528
      if nic.network and nic.ip:
1529
        # Return all IP addresses to the respective address pools
1530
        self._UnlockedCommitIp(constants.RELEASE_ACTION, nic.network, nic.ip)
1531

    
1532
    del self._config_data.instances[inst_uuid]
1533
    self._config_data.cluster.serial_no += 1
1534
    self._WriteConfig()
1535

    
1536
  @locking.ssynchronized(_config_lock)
1537
  def RenameInstance(self, inst_uuid, new_name):
1538
    """Rename an instance.
1539

1540
    This needs to be done in ConfigWriter and not by RemoveInstance
1541
    combined with AddInstance as only we can guarantee an atomic
1542
    rename.
1543

1544
    """
1545
    if inst_uuid not in self._config_data.instances:
1546
      raise errors.ConfigurationError("Unknown instance '%s'" % inst_uuid)
1547

    
1548
    inst = self._config_data.instances[inst_uuid]
1549
    inst.name = new_name
1550

    
1551
    for (idx, disk) in enumerate(inst.disks):
1552
      if disk.dev_type == constants.LD_FILE:
1553
        # rename the file paths in logical and physical id
1554
        file_storage_dir = os.path.dirname(os.path.dirname(disk.logical_id[1]))
1555
        disk.logical_id = (disk.logical_id[0],
1556
                           utils.PathJoin(file_storage_dir, inst.name,
1557
                                          "disk%s" % idx))
1558
        disk.physical_id = disk.logical_id
1559

    
1560
    # Force update of ssconf files
1561
    self._config_data.cluster.serial_no += 1
1562

    
1563
    self._WriteConfig()
1564

    
1565
  @locking.ssynchronized(_config_lock)
1566
  def MarkInstanceDown(self, inst_uuid):
1567
    """Mark the status of an instance to down in the configuration.
1568

1569
    This does not touch the instance disks active flag, as shut down instances
1570
    can still have active disks.
1571

1572
    """
1573
    self._SetInstanceStatus(inst_uuid, constants.ADMINST_DOWN, None)
1574

    
1575
  @locking.ssynchronized(_config_lock)
1576
  def MarkInstanceDisksActive(self, inst_uuid):
1577
    """Mark the status of instance disks active.
1578

1579
    """
1580
    self._SetInstanceStatus(inst_uuid, None, True)
1581

    
1582
  @locking.ssynchronized(_config_lock)
1583
  def MarkInstanceDisksInactive(self, inst_uuid):
1584
    """Mark the status of instance disks inactive.
1585

1586
    """
1587
    self._SetInstanceStatus(inst_uuid, None, False)
1588

    
1589
  def _UnlockedGetInstanceList(self):
1590
    """Get the list of instances.
1591

1592
    This function is for internal use, when the config lock is already held.
1593

1594
    """
1595
    return self._config_data.instances.keys()
1596

    
1597
  @locking.ssynchronized(_config_lock, shared=1)
1598
  def GetInstanceList(self):
1599
    """Get the list of instances.
1600

1601
    @return: array of instances, ex. ['instance2-uuid', 'instance1-uuid']
1602

1603
    """
1604
    return self._UnlockedGetInstanceList()
1605

    
1606
  def ExpandInstanceName(self, short_name):
1607
    """Attempt to expand an incomplete instance name.
1608

1609
    """
1610
    # Locking is done in L{ConfigWriter.GetAllInstancesInfo}
1611
    all_insts = self.GetAllInstancesInfo().values()
1612
    expanded_name = _MatchNameComponentIgnoreCase(
1613
                      short_name, [inst.name for inst in all_insts])
1614

    
1615
    if expanded_name is not None:
1616
      # there has to be exactly one instance with that name
1617
      inst = (filter(lambda n: n.name == expanded_name, all_insts)[0])
1618
      return (inst.uuid, inst.name)
1619
    else:
1620
      return (None, None)
1621

    
1622
  def _UnlockedGetInstanceInfo(self, inst_uuid):
1623
    """Returns information about an instance.
1624

1625
    This function is for internal use, when the config lock is already held.
1626

1627
    """
1628
    if inst_uuid not in self._config_data.instances:
1629
      return None
1630

    
1631
    return self._config_data.instances[inst_uuid]
1632

    
1633
  @locking.ssynchronized(_config_lock, shared=1)
1634
  def GetInstanceInfo(self, inst_uuid):
1635
    """Returns information about an instance.
1636

1637
    It takes the information from the configuration file. Other information of
1638
    an instance are taken from the live systems.
1639

1640
    @param inst_uuid: UUID of the instance
1641

1642
    @rtype: L{objects.Instance}
1643
    @return: the instance object
1644

1645
    """
1646
    return self._UnlockedGetInstanceInfo(inst_uuid)
1647

    
1648
  @locking.ssynchronized(_config_lock, shared=1)
1649
  def GetInstanceNodeGroups(self, inst_uuid, primary_only=False):
1650
    """Returns set of node group UUIDs for instance's nodes.
1651

1652
    @rtype: frozenset
1653

1654
    """
1655
    instance = self._UnlockedGetInstanceInfo(inst_uuid)
1656
    if not instance:
1657
      raise errors.ConfigurationError("Unknown instance '%s'" % inst_uuid)
1658

    
1659
    if primary_only:
1660
      nodes = [instance.primary_node]
1661
    else:
1662
      nodes = instance.all_nodes
1663

    
1664
    return frozenset(self._UnlockedGetNodeInfo(node_uuid).group
1665
                     for node_uuid in nodes)
1666

    
1667
  @locking.ssynchronized(_config_lock, shared=1)
1668
  def GetInstanceNetworks(self, inst_uuid):
1669
    """Returns set of network UUIDs for instance's nics.
1670

1671
    @rtype: frozenset
1672

1673
    """
1674
    instance = self._UnlockedGetInstanceInfo(inst_uuid)
1675
    if not instance:
1676
      raise errors.ConfigurationError("Unknown instance '%s'" % inst_uuid)
1677

    
1678
    networks = set()
1679
    for nic in instance.nics:
1680
      if nic.network:
1681
        networks.add(nic.network)
1682

    
1683
    return frozenset(networks)
1684

    
1685
  @locking.ssynchronized(_config_lock, shared=1)
1686
  def GetMultiInstanceInfo(self, inst_uuids):
1687
    """Get the configuration of multiple instances.
1688

1689
    @param inst_uuids: list of instance UUIDs
1690
    @rtype: list
1691
    @return: list of tuples (instance UUID, instance_info), where
1692
        instance_info is what would GetInstanceInfo return for the
1693
        node, while keeping the original order
1694

1695
    """
1696
    return [(uuid, self._UnlockedGetInstanceInfo(uuid)) for uuid in inst_uuids]
1697

    
1698
  @locking.ssynchronized(_config_lock, shared=1)
1699
  def GetMultiInstanceInfoByName(self, inst_names):
1700
    """Get the configuration of multiple instances.
1701

1702
    @param inst_names: list of instance names
1703
    @rtype: list
1704
    @return: list of tuples (instance, instance_info), where
1705
        instance_info is what would GetInstanceInfo return for the
1706
        node, while keeping the original order
1707

1708
    """
1709
    result = []
1710
    for name in inst_names:
1711
      instance = self._UnlockedGetInstanceInfoByName(name)
1712
      result.append((instance.uuid, instance))
1713
    return result
1714

    
1715
  @locking.ssynchronized(_config_lock, shared=1)
1716
  def GetAllInstancesInfo(self):
1717
    """Get the configuration of all instances.
1718

1719
    @rtype: dict
1720
    @return: dict of (instance, instance_info), where instance_info is what
1721
              would GetInstanceInfo return for the node
1722

1723
    """
1724
    return self._UnlockedGetAllInstancesInfo()
1725

    
1726
  def _UnlockedGetAllInstancesInfo(self):
1727
    my_dict = dict([(inst_uuid, self._UnlockedGetInstanceInfo(inst_uuid))
1728
                    for inst_uuid in self._UnlockedGetInstanceList()])
1729
    return my_dict
1730

    
1731
  @locking.ssynchronized(_config_lock, shared=1)
1732
  def GetInstancesInfoByFilter(self, filter_fn):
1733
    """Get instance configuration with a filter.
1734

1735
    @type filter_fn: callable
1736
    @param filter_fn: Filter function receiving instance object as parameter,
1737
      returning boolean. Important: this function is called while the
1738
      configuration locks is held. It must not do any complex work or call
1739
      functions potentially leading to a deadlock. Ideally it doesn't call any
1740
      other functions and just compares instance attributes.
1741

1742
    """
1743
    return dict((uuid, inst)
1744
                for (uuid, inst) in self._config_data.instances.items()
1745
                if filter_fn(inst))
1746

    
1747
  @locking.ssynchronized(_config_lock, shared=1)
1748
  def GetInstanceInfoByName(self, inst_name):
1749
    """Get the L{objects.Instance} object for a named instance.
1750

1751
    @param inst_name: name of the instance to get information for
1752
    @type inst_name: string
1753
    @return: the corresponding L{objects.Instance} instance or None if no
1754
          information is available
1755

1756
    """
1757
    return self._UnlockedGetInstanceInfoByName(inst_name)
1758

    
1759
  def _UnlockedGetInstanceInfoByName(self, inst_name):
1760
    for inst in self._UnlockedGetAllInstancesInfo().values():
1761
      if inst.name == inst_name:
1762
        return inst
1763
    return None
1764

    
1765
  def _UnlockedGetInstanceName(self, inst_uuid):
1766
    inst_info = self._UnlockedGetInstanceInfo(inst_uuid)
1767
    if inst_info is None:
1768
      raise errors.OpExecError("Unknown instance: %s" % inst_uuid)
1769
    return inst_info.name
1770

    
1771
  @locking.ssynchronized(_config_lock, shared=1)
1772
  def GetInstanceName(self, inst_uuid):
1773
    """Gets the instance name for the passed instance.
1774

1775
    @param inst_uuid: instance UUID to get name for
1776
    @type inst_uuid: string
1777
    @rtype: string
1778
    @return: instance name
1779

1780
    """
1781
    return self._UnlockedGetInstanceName(inst_uuid)
1782

    
1783
  @locking.ssynchronized(_config_lock, shared=1)
1784
  def GetInstanceNames(self, inst_uuids):
1785
    """Gets the instance names for the passed list of nodes.
1786

1787
    @param inst_uuids: list of instance UUIDs to get names for
1788
    @type inst_uuids: list of strings
1789
    @rtype: list of strings
1790
    @return: list of instance names
1791

1792
    """
1793
    return self._UnlockedGetInstanceNames(inst_uuids)
1794

    
1795
  def _UnlockedGetInstanceNames(self, inst_uuids):
1796
    return [self._UnlockedGetInstanceName(uuid) for uuid in inst_uuids]
1797

    
1798
  @locking.ssynchronized(_config_lock)
1799
  def AddNode(self, node, ec_id):
1800
    """Add a node to the configuration.
1801

1802
    @type node: L{objects.Node}
1803
    @param node: a Node instance
1804

1805
    """
1806
    logging.info("Adding node %s to configuration", node.name)
1807

    
1808
    self._EnsureUUID(node, ec_id)
1809

    
1810
    node.serial_no = 1
1811
    node.ctime = node.mtime = time.time()
1812
    self._UnlockedAddNodeToGroup(node.uuid, node.group)
1813
    self._config_data.nodes[node.uuid] = node
1814
    self._config_data.cluster.serial_no += 1
1815
    self._WriteConfig()
1816

    
1817
  @locking.ssynchronized(_config_lock)
1818
  def RemoveNode(self, node_uuid):
1819
    """Remove a node from the configuration.
1820

1821
    """
1822
    logging.info("Removing node %s from configuration", node_uuid)
1823

    
1824
    if node_uuid not in self._config_data.nodes:
1825
      raise errors.ConfigurationError("Unknown node '%s'" % node_uuid)
1826

    
1827
    self._UnlockedRemoveNodeFromGroup(self._config_data.nodes[node_uuid])
1828
    del self._config_data.nodes[node_uuid]
1829
    self._config_data.cluster.serial_no += 1
1830
    self._WriteConfig()
1831

    
1832
  def ExpandNodeName(self, short_name):
1833
    """Attempt to expand an incomplete node name into a node UUID.
1834

1835
    """
1836
    # Locking is done in L{ConfigWriter.GetAllNodesInfo}
1837
    all_nodes = self.GetAllNodesInfo().values()
1838
    expanded_name = _MatchNameComponentIgnoreCase(
1839
                      short_name, [node.name for node in all_nodes])
1840

    
1841
    if expanded_name is not None:
1842
      # there has to be exactly one node with that name
1843
      node = (filter(lambda n: n.name == expanded_name, all_nodes)[0])
1844
      return (node.uuid, node.name)
1845
    else:
1846
      return (None, None)
1847

    
1848
  def _UnlockedGetNodeInfo(self, node_uuid):
1849
    """Get the configuration of a node, as stored in the config.
1850

1851
    This function is for internal use, when the config lock is already
1852
    held.
1853

1854
    @param node_uuid: the node UUID
1855

1856
    @rtype: L{objects.Node}
1857
    @return: the node object
1858

1859
    """
1860
    if node_uuid not in self._config_data.nodes:
1861
      return None
1862

    
1863
    return self._config_data.nodes[node_uuid]
1864

    
1865
  @locking.ssynchronized(_config_lock, shared=1)
1866
  def GetNodeInfo(self, node_uuid):
1867
    """Get the configuration of a node, as stored in the config.
1868

1869
    This is just a locked wrapper over L{_UnlockedGetNodeInfo}.
1870

1871
    @param node_uuid: the node UUID
1872

1873
    @rtype: L{objects.Node}
1874
    @return: the node object
1875

1876
    """
1877
    return self._UnlockedGetNodeInfo(node_uuid)
1878

    
1879
  @locking.ssynchronized(_config_lock, shared=1)
1880
  def GetNodeInstances(self, node_uuid):
1881
    """Get the instances of a node, as stored in the config.
1882

1883
    @param node_uuid: the node UUID
1884

1885
    @rtype: (list, list)
1886
    @return: a tuple with two lists: the primary and the secondary instances
1887

1888
    """
1889
    pri = []
1890
    sec = []
1891
    for inst in self._config_data.instances.values():
1892
      if inst.primary_node == node_uuid:
1893
        pri.append(inst.uuid)
1894
      if node_uuid in inst.secondary_nodes:
1895
        sec.append(inst.uuid)
1896
    return (pri, sec)
1897

    
1898
  @locking.ssynchronized(_config_lock, shared=1)
1899
  def GetNodeGroupInstances(self, uuid, primary_only=False):
1900
    """Get the instances of a node group.
1901

1902
    @param uuid: Node group UUID
1903
    @param primary_only: Whether to only consider primary nodes
1904
    @rtype: frozenset
1905
    @return: List of instance UUIDs in node group
1906

1907
    """
1908
    if primary_only:
1909
      nodes_fn = lambda inst: [inst.primary_node]
1910
    else:
1911
      nodes_fn = lambda inst: inst.all_nodes
1912

    
1913
    return frozenset(inst.uuid
1914
                     for inst in self._config_data.instances.values()
1915
                     for node_uuid in nodes_fn(inst)
1916
                     if self._UnlockedGetNodeInfo(node_uuid).group == uuid)
1917

    
1918
  def _UnlockedGetHvparamsString(self, hvname):
1919
    """Return the string representation of the list of hyervisor parameters of
1920
    the given hypervisor.
1921

1922
    @see: C{GetHvparams}
1923

1924
    """
1925
    result = ""
1926
    hvparams = self._config_data.cluster.hvparams[hvname]
1927
    for key in hvparams:
1928
      result += "%s=%s\n" % (key, hvparams[key])
1929
    return result
1930

    
1931
  @locking.ssynchronized(_config_lock, shared=1)
1932
  def GetHvparamsString(self, hvname):
1933
    """Return the hypervisor parameters of the given hypervisor.
1934

1935
    @type hvname: string
1936
    @param hvname: name of a hypervisor
1937
    @rtype: string
1938
    @return: string containing key-value-pairs, one pair on each line;
1939
      format: KEY=VALUE
1940

1941
    """
1942
    return self._UnlockedGetHvparamsString(hvname)
1943

    
1944
  def _UnlockedGetNodeList(self):
1945
    """Return the list of nodes which are in the configuration.
1946

1947
    This function is for internal use, when the config lock is already
1948
    held.
1949

1950
    @rtype: list
1951

1952
    """
1953
    return self._config_data.nodes.keys()
1954

    
1955
  @locking.ssynchronized(_config_lock, shared=1)
1956
  def GetNodeList(self):
1957
    """Return the list of nodes which are in the configuration.
1958

1959
    """
1960
    return self._UnlockedGetNodeList()
1961

    
1962
  def _UnlockedGetOnlineNodeList(self):
1963
    """Return the list of nodes which are online.
1964

1965
    """
1966
    all_nodes = [self._UnlockedGetNodeInfo(node)
1967
                 for node in self._UnlockedGetNodeList()]
1968
    return [node.uuid for node in all_nodes if not node.offline]
1969

    
1970
  @locking.ssynchronized(_config_lock, shared=1)
1971
  def GetOnlineNodeList(self):
1972
    """Return the list of nodes which are online.
1973

1974
    """
1975
    return self._UnlockedGetOnlineNodeList()
1976

    
1977
  @locking.ssynchronized(_config_lock, shared=1)
1978
  def GetVmCapableNodeList(self):
1979
    """Return the list of nodes which are not vm capable.
1980

1981
    """
1982
    all_nodes = [self._UnlockedGetNodeInfo(node)
1983
                 for node in self._UnlockedGetNodeList()]
1984
    return [node.uuid for node in all_nodes if node.vm_capable]
1985

    
1986
  @locking.ssynchronized(_config_lock, shared=1)
1987
  def GetNonVmCapableNodeList(self):
1988
    """Return the list of nodes which are not vm capable.
1989

1990
    """
1991
    all_nodes = [self._UnlockedGetNodeInfo(node)
1992
                 for node in self._UnlockedGetNodeList()]
1993
    return [node.uuid for node in all_nodes if not node.vm_capable]
1994

    
1995
  @locking.ssynchronized(_config_lock, shared=1)
1996
  def GetMultiNodeInfo(self, node_uuids):
1997
    """Get the configuration of multiple nodes.
1998

1999
    @param node_uuids: list of node UUIDs
2000
    @rtype: list
2001
    @return: list of tuples of (node, node_info), where node_info is
2002
        what would GetNodeInfo return for the node, in the original
2003
        order
2004

2005
    """
2006
    return [(uuid, self._UnlockedGetNodeInfo(uuid)) for uuid in node_uuids]
2007

    
2008
  def _UnlockedGetAllNodesInfo(self):
2009
    """Gets configuration of all nodes.
2010

2011
    @note: See L{GetAllNodesInfo}
2012

2013
    """
2014
    return dict([(node_uuid, self._UnlockedGetNodeInfo(node_uuid))
2015
                 for node_uuid in self._UnlockedGetNodeList()])
2016

    
2017
  @locking.ssynchronized(_config_lock, shared=1)
2018
  def GetAllNodesInfo(self):
2019
    """Get the configuration of all nodes.
2020

2021
    @rtype: dict
2022
    @return: dict of (node, node_info), where node_info is what
2023
              would GetNodeInfo return for the node
2024

2025
    """
2026
    return self._UnlockedGetAllNodesInfo()
2027

    
2028
  def _UnlockedGetNodeInfoByName(self, node_name):
2029
    for node in self._UnlockedGetAllNodesInfo().values():
2030
      if node.name == node_name:
2031
        return node
2032
    return None
2033

    
2034
  @locking.ssynchronized(_config_lock, shared=1)
2035
  def GetNodeInfoByName(self, node_name):
2036
    """Get the L{objects.Node} object for a named node.
2037

2038
    @param node_name: name of the node to get information for
2039
    @type node_name: string
2040
    @return: the corresponding L{objects.Node} instance or None if no
2041
          information is available
2042

2043
    """
2044
    return self._UnlockedGetNodeInfoByName(node_name)
2045

    
2046
  def _UnlockedGetNodeName(self, node_spec):
2047
    if isinstance(node_spec, objects.Node):
2048
      return node_spec.name
2049
    elif isinstance(node_spec, basestring):
2050
      node_info = self._UnlockedGetNodeInfo(node_spec)
2051
      if node_info is None:
2052
        raise errors.OpExecError("Unknown node: %s" % node_spec)
2053
      return node_info.name
2054
    else:
2055
      raise errors.ProgrammerError("Can't handle node spec '%s'" % node_spec)
2056

    
2057
  @locking.ssynchronized(_config_lock, shared=1)
2058
  def GetNodeName(self, node_spec):
2059
    """Gets the node name for the passed node.
2060

2061
    @param node_spec: node to get names for
2062
    @type node_spec: either node UUID or a L{objects.Node} object
2063
    @rtype: string
2064
    @return: node name
2065

2066
    """
2067
    return self._UnlockedGetNodeName(node_spec)
2068

    
2069
  def _UnlockedGetNodeNames(self, node_specs):
2070
    return [self._UnlockedGetNodeName(node_spec) for node_spec in node_specs]
2071

    
2072
  @locking.ssynchronized(_config_lock, shared=1)
2073
  def GetNodeNames(self, node_specs):
2074
    """Gets the node names for the passed list of nodes.
2075

2076
    @param node_specs: list of nodes to get names for
2077
    @type node_specs: list of either node UUIDs or L{objects.Node} objects
2078
    @rtype: list of strings
2079
    @return: list of node names
2080

2081
    """
2082
    return self._UnlockedGetNodeNames(node_specs)
2083

    
2084
  @locking.ssynchronized(_config_lock, shared=1)
2085
  def GetNodeGroupsFromNodes(self, node_uuids):
2086
    """Returns groups for a list of nodes.
2087

2088
    @type node_uuids: list of string
2089
    @param node_uuids: List of node UUIDs
2090
    @rtype: frozenset
2091

2092
    """
2093
    return frozenset(self._UnlockedGetNodeInfo(uuid).group
2094
                     for uuid in node_uuids)
2095

    
2096
  def _UnlockedGetMasterCandidateStats(self, exceptions=None):
2097
    """Get the number of current and maximum desired and possible candidates.
2098

2099
    @type exceptions: list
2100
    @param exceptions: if passed, list of nodes that should be ignored
2101
    @rtype: tuple
2102
    @return: tuple of (current, desired and possible, possible)
2103

2104
    """
2105
    mc_now = mc_should = mc_max = 0
2106
    for node in self._config_data.nodes.values():
2107
      if exceptions and node.uuid in exceptions:
2108
        continue
2109
      if not (node.offline or node.drained) and node.master_capable:
2110
        mc_max += 1
2111
      if node.master_candidate:
2112
        mc_now += 1
2113
    mc_should = min(mc_max, self._config_data.cluster.candidate_pool_size)
2114
    return (mc_now, mc_should, mc_max)
2115

    
2116
  @locking.ssynchronized(_config_lock, shared=1)
2117
  def GetMasterCandidateStats(self, exceptions=None):
2118
    """Get the number of current and maximum possible candidates.
2119

2120
    This is just a wrapper over L{_UnlockedGetMasterCandidateStats}.
2121

2122
    @type exceptions: list
2123
    @param exceptions: if passed, list of nodes that should be ignored
2124
    @rtype: tuple
2125
    @return: tuple of (current, max)
2126

2127
    """
2128
    return self._UnlockedGetMasterCandidateStats(exceptions)
2129

    
2130
  @locking.ssynchronized(_config_lock)
2131
  def MaintainCandidatePool(self, exception_node_uuids):
2132
    """Try to grow the candidate pool to the desired size.
2133

2134
    @type exception_node_uuids: list
2135
    @param exception_node_uuids: if passed, list of nodes that should be ignored
2136
    @rtype: list
2137
    @return: list with the adjusted nodes (L{objects.Node} instances)
2138

2139
    """
2140
    mc_now, mc_max, _ = self._UnlockedGetMasterCandidateStats(
2141
                          exception_node_uuids)
2142
    mod_list = []
2143
    if mc_now < mc_max:
2144
      node_list = self._config_data.nodes.keys()
2145
      random.shuffle(node_list)
2146
      for uuid in node_list:
2147
        if mc_now >= mc_max:
2148
          break
2149
        node = self._config_data.nodes[uuid]
2150
        if (node.master_candidate or node.offline or node.drained or
2151
            node.uuid in exception_node_uuids or not node.master_capable):
2152
          continue
2153
        mod_list.append(node)
2154
        node.master_candidate = True
2155
        node.serial_no += 1
2156
        mc_now += 1
2157
      if mc_now != mc_max:
2158
        # this should not happen
2159
        logging.warning("Warning: MaintainCandidatePool didn't manage to"
2160
                        " fill the candidate pool (%d/%d)", mc_now, mc_max)
2161
      if mod_list:
2162
        self._config_data.cluster.serial_no += 1
2163
        self._WriteConfig()
2164

    
2165
    return mod_list
2166

    
2167
  def _UnlockedAddNodeToGroup(self, node_uuid, nodegroup_uuid):
2168
    """Add a given node to the specified group.
2169

2170
    """
2171
    if nodegroup_uuid not in self._config_data.nodegroups:
2172
      # This can happen if a node group gets deleted between its lookup and
2173
      # when we're adding the first node to it, since we don't keep a lock in
2174
      # the meantime. It's ok though, as we'll fail cleanly if the node group
2175
      # is not found anymore.
2176
      raise errors.OpExecError("Unknown node group: %s" % nodegroup_uuid)
2177
    if node_uuid not in self._config_data.nodegroups[nodegroup_uuid].members:
2178
      self._config_data.nodegroups[nodegroup_uuid].members.append(node_uuid)
2179

    
2180
  def _UnlockedRemoveNodeFromGroup(self, node):
2181
    """Remove a given node from its group.
2182

2183
    """
2184
    nodegroup = node.group
2185
    if nodegroup not in self._config_data.nodegroups:
2186
      logging.warning("Warning: node '%s' has unknown node group '%s'"
2187
                      " (while being removed from it)", node.uuid, nodegroup)
2188
    nodegroup_obj = self._config_data.nodegroups[nodegroup]
2189
    if node.uuid not in nodegroup_obj.members:
2190
      logging.warning("Warning: node '%s' not a member of its node group '%s'"
2191
                      " (while being removed from it)", node.uuid, nodegroup)
2192
    else:
2193
      nodegroup_obj.members.remove(node.uuid)
2194

    
2195
  @locking.ssynchronized(_config_lock)
2196
  def AssignGroupNodes(self, mods):
2197
    """Changes the group of a number of nodes.
2198

2199
    @type mods: list of tuples; (node name, new group UUID)
2200
    @param mods: Node membership modifications
2201

2202
    """
2203
    groups = self._config_data.nodegroups
2204
    nodes = self._config_data.nodes
2205

    
2206
    resmod = []
2207

    
2208
    # Try to resolve UUIDs first
2209
    for (node_uuid, new_group_uuid) in mods:
2210
      try:
2211
        node = nodes[node_uuid]
2212
      except KeyError:
2213
        raise errors.ConfigurationError("Unable to find node '%s'" % node_uuid)
2214

    
2215
      if node.group == new_group_uuid:
2216
        # Node is being assigned to its current group
2217
        logging.debug("Node '%s' was assigned to its current group (%s)",
2218
                      node_uuid, node.group)
2219
        continue
2220

    
2221
      # Try to find current group of node
2222
      try:
2223
        old_group = groups[node.group]
2224
      except KeyError:
2225
        raise errors.ConfigurationError("Unable to find old group '%s'" %
2226
                                        node.group)
2227

    
2228
      # Try to find new group for node
2229
      try:
2230
        new_group = groups[new_group_uuid]
2231
      except KeyError:
2232
        raise errors.ConfigurationError("Unable to find new group '%s'" %
2233
                                        new_group_uuid)
2234

    
2235
      assert node.uuid in old_group.members, \
2236
        ("Inconsistent configuration: node '%s' not listed in members for its"
2237
         " old group '%s'" % (node.uuid, old_group.uuid))
2238
      assert node.uuid not in new_group.members, \
2239
        ("Inconsistent configuration: node '%s' already listed in members for"
2240
         " its new group '%s'" % (node.uuid, new_group.uuid))
2241

    
2242
      resmod.append((node, old_group, new_group))
2243

    
2244
    # Apply changes
2245
    for (node, old_group, new_group) in resmod:
2246
      assert node.uuid != new_group.uuid and old_group.uuid != new_group.uuid, \
2247
        "Assigning to current group is not possible"
2248

    
2249
      node.group = new_group.uuid
2250

    
2251
      # Update members of involved groups
2252
      if node.uuid in old_group.members:
2253
        old_group.members.remove(node.uuid)
2254
      if node.uuid not in new_group.members:
2255
        new_group.members.append(node.uuid)
2256

    
2257
    # Update timestamps and serials (only once per node/group object)
2258
    now = time.time()
2259
    for obj in frozenset(itertools.chain(*resmod)): # pylint: disable=W0142
2260
      obj.serial_no += 1
2261
      obj.mtime = now
2262

    
2263
    # Force ssconf update
2264
    self._config_data.cluster.serial_no += 1
2265

    
2266
    self._WriteConfig()
2267

    
2268
  def _BumpSerialNo(self):
2269
    """Bump up the serial number of the config.
2270

2271
    """
2272
    self._config_data.serial_no += 1
2273
    self._config_data.mtime = time.time()
2274

    
2275
  def _AllUUIDObjects(self):
2276
    """Returns all objects with uuid attributes.
2277

2278
    """
2279
    return (self._config_data.instances.values() +
2280
            self._config_data.nodes.values() +
2281
            self._config_data.nodegroups.values() +
2282
            self._config_data.networks.values() +
2283
            self._AllDisks() +
2284
            self._AllNICs() +
2285
            [self._config_data.cluster])
2286

    
2287
  def _OpenConfig(self, accept_foreign):
2288
    """Read the config data from disk.
2289

2290
    """
2291
    raw_data = utils.ReadFile(self._cfg_file)
2292

    
2293
    try:
2294
      data = objects.ConfigData.FromDict(serializer.Load(raw_data))
2295
    except Exception, err:
2296
      raise errors.ConfigurationError(err)
2297

    
2298
    # Make sure the configuration has the right version
2299
    _ValidateConfig(data)
2300

    
2301
    if (not hasattr(data, "cluster") or
2302
        not hasattr(data.cluster, "rsahostkeypub")):
2303
      raise errors.ConfigurationError("Incomplete configuration"
2304
                                      " (missing cluster.rsahostkeypub)")
2305

    
2306
    if not data.cluster.master_node in data.nodes:
2307
      msg = ("The configuration denotes node %s as master, but does not"
2308
             " contain information about this node" %
2309
             data.cluster.master_node)
2310
      raise errors.ConfigurationError(msg)
2311

    
2312
    master_info = data.nodes[data.cluster.master_node]
2313
    if master_info.name != self._my_hostname and not accept_foreign:
2314
      msg = ("The configuration denotes node %s as master, while my"
2315
             " hostname is %s; opening a foreign configuration is only"
2316
             " possible in accept_foreign mode" %
2317
             (master_info.name, self._my_hostname))
2318
      raise errors.ConfigurationError(msg)
2319

    
2320
    self._config_data = data
2321
    # reset the last serial as -1 so that the next write will cause
2322
    # ssconf update
2323
    self._last_cluster_serial = -1
2324

    
2325
    # Upgrade configuration if needed
2326
    self._UpgradeConfig()
2327

    
2328
    self._cfg_id = utils.GetFileID(path=self._cfg_file)
2329

    
2330
  def _UpgradeConfig(self):
2331
    """Run any upgrade steps.
2332

2333
    This method performs both in-object upgrades and also update some data
2334
    elements that need uniqueness across the whole configuration or interact
2335
    with other objects.
2336

2337
    @warning: this function will call L{_WriteConfig()}, but also
2338
        L{DropECReservations} so it needs to be called only from a
2339
        "safe" place (the constructor). If one wanted to call it with
2340
        the lock held, a DropECReservationUnlocked would need to be
2341
        created first, to avoid causing deadlock.
2342

2343
    """
2344
    # Keep a copy of the persistent part of _config_data to check for changes
2345
    # Serialization doesn't guarantee order in dictionaries
2346
    oldconf = copy.deepcopy(self._config_data.ToDict())
2347

    
2348
    # In-object upgrades
2349
    self._config_data.UpgradeConfig()
2350

    
2351
    for item in self._AllUUIDObjects():
2352
      if item.uuid is None:
2353
        item.uuid = self._GenerateUniqueID(_UPGRADE_CONFIG_JID)
2354
    if not self._config_data.nodegroups:
2355
      default_nodegroup_name = constants.INITIAL_NODE_GROUP_NAME
2356
      default_nodegroup = objects.NodeGroup(name=default_nodegroup_name,
2357
                                            members=[])
2358
      self._UnlockedAddNodeGroup(default_nodegroup, _UPGRADE_CONFIG_JID, True)
2359
    for node in self._config_data.nodes.values():
2360
      if not node.group:
2361
        node.group = self.LookupNodeGroup(None)
2362
      # This is technically *not* an upgrade, but needs to be done both when
2363
      # nodegroups are being added, and upon normally loading the config,
2364
      # because the members list of a node group is discarded upon
2365
      # serializing/deserializing the object.
2366
      self._UnlockedAddNodeToGroup(node.uuid, node.group)
2367

    
2368
    modified = (oldconf != self._config_data.ToDict())
2369
    if modified:
2370
      self._WriteConfig()
2371
      # This is ok even if it acquires the internal lock, as _UpgradeConfig is
2372
      # only called at config init time, without the lock held
2373
      self.DropECReservations(_UPGRADE_CONFIG_JID)
2374
    else:
2375
      config_errors = self._UnlockedVerifyConfig()
2376
      if config_errors:
2377
        errmsg = ("Loaded configuration data is not consistent: %s" %
2378
                  (utils.CommaJoin(config_errors)))
2379
        logging.critical(errmsg)
2380

    
2381
  def _DistributeConfig(self, feedback_fn):
2382
    """Distribute the configuration to the other nodes.
2383

2384
    Currently, this only copies the configuration file. In the future,
2385
    it could be used to encapsulate the 2/3-phase update mechanism.
2386

2387
    """
2388
    if self._offline:
2389
      return True
2390

    
2391
    bad = False
2392

    
2393
    node_list = []
2394
    addr_list = []
2395
    myhostname = self._my_hostname
2396
    # we can skip checking whether _UnlockedGetNodeInfo returns None
2397
    # since the node list comes from _UnlocketGetNodeList, and we are
2398
    # called with the lock held, so no modifications should take place
2399
    # in between
2400
    for node_uuid in self._UnlockedGetNodeList():
2401
      node_info = self._UnlockedGetNodeInfo(node_uuid)
2402
      if node_info.name == myhostname or not node_info.master_candidate:
2403
        continue
2404
      node_list.append(node_info.name)
2405
      addr_list.append(node_info.primary_ip)
2406

    
2407
    # TODO: Use dedicated resolver talking to config writer for name resolution
2408
    result = \
2409
      self._GetRpc(addr_list).call_upload_file(node_list, self._cfg_file)
2410
    for to_node, to_result in result.items():
2411
      msg = to_result.fail_msg
2412
      if msg:
2413
        msg = ("Copy of file %s to node %s failed: %s" %
2414
               (self._cfg_file, to_node, msg))
2415
        logging.error(msg)
2416

    
2417
        if feedback_fn:
2418
          feedback_fn(msg)
2419

    
2420
        bad = True
2421

    
2422
    return not bad
2423

    
2424
  def _WriteConfig(self, destination=None, feedback_fn=None):
2425
    """Write the configuration data to persistent storage.
2426

2427
    """
2428
    assert feedback_fn is None or callable(feedback_fn)
2429

    
2430
    # Warn on config errors, but don't abort the save - the
2431
    # configuration has already been modified, and we can't revert;
2432
    # the best we can do is to warn the user and save as is, leaving
2433
    # recovery to the user
2434
    config_errors = self._UnlockedVerifyConfig()
2435
    if config_errors:
2436
      errmsg = ("Configuration data is not consistent: %s" %
2437
                (utils.CommaJoin(config_errors)))
2438
      logging.critical(errmsg)
2439
      if feedback_fn:
2440
        feedback_fn(errmsg)
2441

    
2442
    if destination is None:
2443
      destination = self._cfg_file
2444
    self._BumpSerialNo()
2445
    txt = serializer.Dump(self._config_data.ToDict())
2446

    
2447
    getents = self._getents()
2448
    try:
2449
      fd = utils.SafeWriteFile(destination, self._cfg_id, data=txt,
2450
                               close=False, gid=getents.confd_gid, mode=0640)
2451
    except errors.LockError:
2452
      raise errors.ConfigurationError("The configuration file has been"
2453
                                      " modified since the last write, cannot"
2454
                                      " update")
2455
    try:
2456
      self._cfg_id = utils.GetFileID(fd=fd)
2457
    finally:
2458
      os.close(fd)
2459

    
2460
    self.write_count += 1
2461

    
2462
    # and redistribute the config file to master candidates
2463
    self._DistributeConfig(feedback_fn)
2464

    
2465
    # Write ssconf files on all nodes (including locally)
2466
    if self._last_cluster_serial < self._config_data.cluster.serial_no:
2467
      if not self._offline:
2468
        result = self._GetRpc(None).call_write_ssconf_files(
2469
          self._UnlockedGetNodeNames(self._UnlockedGetOnlineNodeList()),
2470
          self._UnlockedGetSsconfValues())
2471

    
2472
        for nname, nresu in result.items():
2473
          msg = nresu.fail_msg
2474
          if msg:
2475
            errmsg = ("Error while uploading ssconf files to"
2476
                      " node %s: %s" % (nname, msg))
2477
            logging.warning(errmsg)
2478

    
2479
            if feedback_fn:
2480
              feedback_fn(errmsg)
2481

    
2482
      self._last_cluster_serial = self._config_data.cluster.serial_no
2483

    
2484
  def _GetAllHvparamsStrings(self, hypervisors):
2485
    """Get the hvparams of all given hypervisors from the config.
2486

2487
    @type hypervisors: list of string
2488
    @param hypervisors: list of hypervisor names
2489
    @rtype: dict of strings
2490
    @returns: dictionary mapping the hypervisor name to a string representation
2491
      of the hypervisor's hvparams
2492

2493
    """
2494
    hvparams = {}
2495
    for hv in hypervisors:
2496
      hvparams[hv] = self._UnlockedGetHvparamsString(hv)
2497
    return hvparams
2498

    
2499
  @staticmethod
2500
  def _ExtendByAllHvparamsStrings(ssconf_values, all_hvparams):
2501
    """Extends the ssconf_values dictionary by hvparams.
2502

2503
    @type ssconf_values: dict of strings
2504
    @param ssconf_values: dictionary mapping ssconf_keys to strings
2505
      representing the content of ssconf files
2506
    @type all_hvparams: dict of strings
2507
    @param all_hvparams: dictionary mapping hypervisor names to a string
2508
      representation of their hvparams
2509
    @rtype: same as ssconf_values
2510
    @returns: the ssconf_values dictionary extended by hvparams
2511

2512
    """
2513
    for hv in all_hvparams:
2514
      ssconf_key = constants.SS_HVPARAMS_PREF + hv
2515
      ssconf_values[ssconf_key] = all_hvparams[hv]
2516
    return ssconf_values
2517

    
2518
  def _UnlockedGetSsconfValues(self):
2519
    """Return the values needed by ssconf.
2520

2521
    @rtype: dict
2522
    @return: a dictionary with keys the ssconf names and values their
2523
        associated value
2524

2525
    """
2526
    fn = "\n".join
2527
    instance_names = utils.NiceSort(
2528
                       [inst.name for inst in
2529
                        self._UnlockedGetAllInstancesInfo().values()])
2530
    node_infos = self._UnlockedGetAllNodesInfo().values()
2531
    node_names = [node.name for node in node_infos]
2532
    node_pri_ips = ["%s %s" % (ninfo.name, ninfo.primary_ip)
2533
                    for ninfo in node_infos]
2534
    node_snd_ips = ["%s %s" % (ninfo.name, ninfo.secondary_ip)
2535
                    for ninfo in node_infos]
2536

    
2537
    instance_data = fn(instance_names)
2538
    off_data = fn(node.name for node in node_infos if node.offline)
2539
    on_data = fn(node.name for node in node_infos if not node.offline)
2540
    mc_data = fn(node.name for node in node_infos if node.master_candidate)
2541
    mc_ips_data = fn(node.primary_ip for node in node_infos
2542
                     if node.master_candidate)
2543
    node_data = fn(node_names)
2544
    node_pri_ips_data = fn(node_pri_ips)
2545
    node_snd_ips_data = fn(node_snd_ips)
2546

    
2547
    cluster = self._config_data.cluster
2548
    cluster_tags = fn(cluster.GetTags())
2549

    
2550
    hypervisor_list = fn(cluster.enabled_hypervisors)
2551
    all_hvparams = self._GetAllHvparamsStrings(constants.HYPER_TYPES)
2552

    
2553
    uid_pool = uidpool.FormatUidPool(cluster.uid_pool, separator="\n")
2554

    
2555
    nodegroups = ["%s %s" % (nodegroup.uuid, nodegroup.name) for nodegroup in
2556
                  self._config_data.nodegroups.values()]
2557
    nodegroups_data = fn(utils.NiceSort(nodegroups))
2558
    networks = ["%s %s" % (net.uuid, net.name) for net in
2559
                self._config_data.networks.values()]
2560
    networks_data = fn(utils.NiceSort(networks))
2561

    
2562
    ssconf_values = {
2563
      constants.SS_CLUSTER_NAME: cluster.cluster_name,
2564
      constants.SS_CLUSTER_TAGS: cluster_tags,
2565
      constants.SS_FILE_STORAGE_DIR: cluster.file_storage_dir,
2566
      constants.SS_SHARED_FILE_STORAGE_DIR: cluster.shared_file_storage_dir,
2567
      constants.SS_MASTER_CANDIDATES: mc_data,
2568
      constants.SS_MASTER_CANDIDATES_IPS: mc_ips_data,
2569
      constants.SS_MASTER_IP: cluster.master_ip,
2570
      constants.SS_MASTER_NETDEV: cluster.master_netdev,
2571
      constants.SS_MASTER_NETMASK: str(cluster.master_netmask),
2572
      constants.SS_MASTER_NODE: self._UnlockedGetNodeName(cluster.master_node),
2573
      constants.SS_NODE_LIST: node_data,
2574
      constants.SS_NODE_PRIMARY_IPS: node_pri_ips_data,
2575
      constants.SS_NODE_SECONDARY_IPS: node_snd_ips_data,
2576
      constants.SS_OFFLINE_NODES: off_data,
2577
      constants.SS_ONLINE_NODES: on_data,
2578
      constants.SS_PRIMARY_IP_FAMILY: str(cluster.primary_ip_family),
2579
      constants.SS_INSTANCE_LIST: instance_data,
2580
      constants.SS_RELEASE_VERSION: constants.RELEASE_VERSION,
2581
      constants.SS_HYPERVISOR_LIST: hypervisor_list,
2582
      constants.SS_MAINTAIN_NODE_HEALTH: str(cluster.maintain_node_health),
2583
      constants.SS_UID_POOL: uid_pool,
2584
      constants.SS_NODEGROUPS: nodegroups_data,
2585
      constants.SS_NETWORKS: networks_data,
2586
      }
2587
    ssconf_values = self._ExtendByAllHvparamsStrings(ssconf_values,
2588
                                                     all_hvparams)
2589
    bad_values = [(k, v) for k, v in ssconf_values.items()
2590
                  if not isinstance(v, (str, basestring))]
2591
    if bad_values:
2592
      err = utils.CommaJoin("%s=%s" % (k, v) for k, v in bad_values)
2593
      raise errors.ConfigurationError("Some ssconf key(s) have non-string"
2594
                                      " values: %s" % err)
2595
    return ssconf_values
2596

    
2597
  @locking.ssynchronized(_config_lock, shared=1)
2598
  def GetSsconfValues(self):
2599
    """Wrapper using lock around _UnlockedGetSsconf().
2600

2601
    """
2602
    return self._UnlockedGetSsconfValues()
2603

    
2604
  @locking.ssynchronized(_config_lock, shared=1)
2605
  def GetVGName(self):
2606
    """Return the volume group name.
2607

2608
    """
2609
    return self._config_data.cluster.volume_group_name
2610

    
2611
  @locking.ssynchronized(_config_lock)
2612
  def SetVGName(self, vg_name):
2613
    """Set the volume group name.
2614

2615
    """
2616
    self._config_data.cluster.volume_group_name = vg_name
2617
    self._config_data.cluster.serial_no += 1
2618
    self._WriteConfig()
2619

    
2620
  @locking.ssynchronized(_config_lock, shared=1)
2621
  def GetDRBDHelper(self):
2622
    """Return DRBD usermode helper.
2623

2624
    """
2625
    return self._config_data.cluster.drbd_usermode_helper
2626

    
2627
  @locking.ssynchronized(_config_lock)
2628
  def SetDRBDHelper(self, drbd_helper):
2629
    """Set DRBD usermode helper.
2630

2631
    """
2632
    self._config_data.cluster.drbd_usermode_helper = drbd_helper
2633
    self._config_data.cluster.serial_no += 1
2634
    self._WriteConfig()
2635

    
2636
  @locking.ssynchronized(_config_lock, shared=1)
2637
  def GetMACPrefix(self):
2638
    """Return the mac prefix.
2639

2640
    """
2641
    return self._config_data.cluster.mac_prefix
2642

    
2643
  @locking.ssynchronized(_config_lock, shared=1)
2644
  def GetClusterInfo(self):
2645
    """Returns information about the cluster
2646

2647
    @rtype: L{objects.Cluster}
2648
    @return: the cluster object
2649

2650
    """
2651
    return self._config_data.cluster
2652

    
2653
  @locking.ssynchronized(_config_lock, shared=1)
2654
  def HasAnyDiskOfType(self, dev_type):
2655
    """Check if in there is at disk of the given type in the configuration.
2656

2657
    """
2658
    return self._config_data.HasAnyDiskOfType(dev_type)
2659

    
2660
  @locking.ssynchronized(_config_lock)
2661
  def Update(self, target, feedback_fn, ec_id=None):
2662
    """Notify function to be called after updates.
2663

2664
    This function must be called when an object (as returned by
2665
    GetInstanceInfo, GetNodeInfo, GetCluster) has been updated and the
2666
    caller wants the modifications saved to the backing store. Note
2667
    that all modified objects will be saved, but the target argument
2668
    is the one the caller wants to ensure that it's saved.
2669

2670
    @param target: an instance of either L{objects.Cluster},
2671
        L{objects.Node} or L{objects.Instance} which is existing in
2672
        the cluster
2673
    @param feedback_fn: Callable feedback function
2674

2675
    """
2676
    if self._config_data is None:
2677
      raise errors.ProgrammerError("Configuration file not read,"
2678
                                   " cannot save.")
2679
    update_serial = False
2680
    if isinstance(target, objects.Cluster):
2681
      test = target == self._config_data.cluster
2682
    elif isinstance(target, objects.Node):
2683
      test = target in self._config_data.nodes.values()
2684
      update_serial = True
2685
    elif isinstance(target, objects.Instance):
2686
      test = target in self._config_data.instances.values()
2687
    elif isinstance(target, objects.NodeGroup):
2688
      test = target in self._config_data.nodegroups.values()
2689
    elif isinstance(target, objects.Network):
2690
      test = target in self._config_data.networks.values()
2691
    else:
2692
      raise errors.ProgrammerError("Invalid object type (%s) passed to"
2693
                                   " ConfigWriter.Update" % type(target))
2694
    if not test:
2695
      raise errors.ConfigurationError("Configuration updated since object"
2696
                                      " has been read or unknown object")
2697
    target.serial_no += 1
2698
    target.mtime = now = time.time()
2699

    
2700
    if update_serial:
2701
      # for node updates, we need to increase the cluster serial too
2702
      self._config_data.cluster.serial_no += 1
2703
      self._config_data.cluster.mtime = now
2704

    
2705
    if isinstance(target, objects.Instance):
2706
      self._UnlockedReleaseDRBDMinors(target.uuid)
2707

    
2708
    if ec_id is not None:
2709
      # Commit all ips reserved by OpInstanceSetParams and OpGroupSetParams
2710
      self._UnlockedCommitTemporaryIps(ec_id)
2711

    
2712
    self._WriteConfig(feedback_fn=feedback_fn)
2713

    
2714
  @locking.ssynchronized(_config_lock)
2715
  def DropECReservations(self, ec_id):
2716
    """Drop per-execution-context reservations
2717

2718
    """
2719
    for rm in self._all_rms:
2720
      rm.DropECReservations(ec_id)
2721

    
2722
  @locking.ssynchronized(_config_lock, shared=1)
2723
  def GetAllNetworksInfo(self):
2724
    """Get configuration info of all the networks.
2725

2726
    """
2727
    return dict(self._config_data.networks)
2728

    
2729
  def _UnlockedGetNetworkList(self):
2730
    """Get the list of networks.
2731

2732
    This function is for internal use, when the config lock is already held.
2733

2734
    """
2735
    return self._config_data.networks.keys()
2736

    
2737
  @locking.ssynchronized(_config_lock, shared=1)
2738
  def GetNetworkList(self):
2739
    """Get the list of networks.
2740

2741
    @return: array of networks, ex. ["main", "vlan100", "200]
2742

2743
    """
2744
    return self._UnlockedGetNetworkList()
2745

    
2746
  @locking.ssynchronized(_config_lock, shared=1)
2747
  def GetNetworkNames(self):
2748
    """Get a list of network names
2749

2750
    """
2751
    names = [net.name
2752
             for net in self._config_data.networks.values()]
2753
    return names
2754

    
2755
  def _UnlockedGetNetwork(self, uuid):
2756
    """Returns information about a network.
2757

2758
    This function is for internal use, when the config lock is already held.
2759

2760
    """
2761
    if uuid not in self._config_data.networks:
2762
      return None
2763

    
2764
    return self._config_data.networks[uuid]
2765

    
2766
  @locking.ssynchronized(_config_lock, shared=1)
2767
  def GetNetwork(self, uuid):
2768
    """Returns information about a network.
2769

2770
    It takes the information from the configuration file.
2771

2772
    @param uuid: UUID of the network
2773

2774
    @rtype: L{objects.Network}
2775
    @return: the network object
2776

2777
    """
2778
    return self._UnlockedGetNetwork(uuid)
2779

    
2780
  @locking.ssynchronized(_config_lock)
2781
  def AddNetwork(self, net, ec_id, check_uuid=True):
2782
    """Add a network to the configuration.
2783

2784
    @type net: L{objects.Network}
2785
    @param net: the Network object to add
2786
    @type ec_id: string
2787
    @param ec_id: unique id for the job to use when creating a missing UUID
2788

2789
    """
2790
    self._UnlockedAddNetwork(net, ec_id, check_uuid)
2791
    self._WriteConfig()
2792

    
2793
  def _UnlockedAddNetwork(self, net, ec_id, check_uuid):
2794
    """Add a network to the configuration.
2795

2796
    """
2797
    logging.info("Adding network %s to configuration", net.name)
2798

    
2799
    if check_uuid:
2800
      self._EnsureUUID(net, ec_id)
2801

    
2802
    net.serial_no = 1
2803
    net.ctime = net.mtime = time.time()
2804
    self._config_data.networks[net.uuid] = net
2805
    self._config_data.cluster.serial_no += 1
2806

    
2807
  def _UnlockedLookupNetwork(self, target):
2808
    """Lookup a network's UUID.
2809

2810
    @type target: string
2811
    @param target: network name or UUID
2812
    @rtype: string
2813
    @return: network UUID
2814
    @raises errors.OpPrereqError: when the target network cannot be found
2815

2816
    """
2817
    if target is None:
2818
      return None
2819
    if target in self._config_data.networks:
2820
      return target
2821
    for net in self._config_data.networks.values():
2822
      if net.name == target:
2823
        return net.uuid
2824
    raise errors.OpPrereqError("Network '%s' not found" % target,
2825
                               errors.ECODE_NOENT)
2826

    
2827
  @locking.ssynchronized(_config_lock, shared=1)
2828
  def LookupNetwork(self, target):
2829
    """Lookup a network's UUID.
2830

2831
    This function is just a wrapper over L{_UnlockedLookupNetwork}.
2832

2833
    @type target: string
2834
    @param target: network name or UUID
2835
    @rtype: string
2836
    @return: network UUID
2837

2838
    """
2839
    return self._UnlockedLookupNetwork(target)
2840

    
2841
  @locking.ssynchronized(_config_lock)
2842
  def RemoveNetwork(self, network_uuid):
2843
    """Remove a network from the configuration.
2844

2845
    @type network_uuid: string
2846
    @param network_uuid: the UUID of the network to remove
2847

2848
    """
2849
    logging.info("Removing network %s from configuration", network_uuid)
2850

    
2851
    if network_uuid not in self._config_data.networks:
2852
      raise errors.ConfigurationError("Unknown network '%s'" % network_uuid)
2853

    
2854
    del self._config_data.networks[network_uuid]
2855
    self._config_data.cluster.serial_no += 1
2856
    self._WriteConfig()
2857

    
2858
  def _UnlockedGetGroupNetParams(self, net_uuid, node_uuid):
2859
    """Get the netparams (mode, link) of a network.
2860

2861
    Get a network's netparams for a given node.
2862

2863
    @type net_uuid: string
2864
    @param net_uuid: network uuid
2865
    @type node_uuid: string
2866
    @param node_uuid: node UUID
2867
    @rtype: dict or None
2868
    @return: netparams
2869

2870
    """
2871
    node_info = self._UnlockedGetNodeInfo(node_uuid)
2872
    nodegroup_info = self._UnlockedGetNodeGroup(node_info.group)
2873
    netparams = nodegroup_info.networks.get(net_uuid, None)
2874

    
2875
    return netparams
2876

    
2877
  @locking.ssynchronized(_config_lock, shared=1)
2878
  def GetGroupNetParams(self, net_uuid, node_uuid):
2879
    """Locking wrapper of _UnlockedGetGroupNetParams()
2880

2881
    """
2882
    return self._UnlockedGetGroupNetParams(net_uuid, node_uuid)
2883

    
2884
  @locking.ssynchronized(_config_lock, shared=1)
2885
  def CheckIPInNodeGroup(self, ip, node_uuid):
2886
    """Check IP uniqueness in nodegroup.
2887

2888
    Check networks that are connected in the node's node group
2889
    if ip is contained in any of them. Used when creating/adding
2890
    a NIC to ensure uniqueness among nodegroups.
2891

2892
    @type ip: string
2893
    @param ip: ip address
2894
    @type node_uuid: string
2895
    @param node_uuid: node UUID
2896
    @rtype: (string, dict) or (None, None)
2897
    @return: (network name, netparams)
2898

2899
    """
2900
    if ip is None:
2901
      return (None, None)
2902
    node_info = self._UnlockedGetNodeInfo(node_uuid)
2903
    nodegroup_info = self._UnlockedGetNodeGroup(node_info.group)
2904
    for net_uuid in nodegroup_info.networks.keys():
2905
      net_info = self._UnlockedGetNetwork(net_uuid)
2906
      pool = network.AddressPool(net_info)
2907
      if pool.Contains(ip):
2908
        return (net_info.name, nodegroup_info.networks[net_uuid])
2909

    
2910
    return (None, None)