Statistics
| Branch: | Tag: | Revision:

root / lib / cmdlib / instance_storage.py @ da4a52a3

History | View | Annotate | Download (97.9 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
"""Logical units dealing with storage of instances."""
23

    
24
import itertools
25
import logging
26
import os
27
import time
28

    
29
from ganeti import compat
30
from ganeti import constants
31
from ganeti import errors
32
from ganeti import ht
33
from ganeti import locking
34
from ganeti.masterd import iallocator
35
from ganeti import objects
36
from ganeti import utils
37
from ganeti import opcodes
38
from ganeti import rpc
39
from ganeti.cmdlib.base import LogicalUnit, NoHooksLU, Tasklet
40
from ganeti.cmdlib.common import INSTANCE_DOWN, INSTANCE_NOT_RUNNING, \
41
  AnnotateDiskParams, CheckIAllocatorOrNode, ExpandNodeUuidAndName, \
42
  CheckNodeOnline, CheckInstanceNodeGroups, CheckInstanceState, \
43
  IsExclusiveStorageEnabledNode, FindFaultyInstanceDisks, GetWantedNodes
44
from ganeti.cmdlib.instance_utils import GetInstanceInfoText, \
45
  CopyLockList, ReleaseLocks, CheckNodeVmCapable, \
46
  BuildInstanceHookEnvByObject, CheckNodeNotDrained, CheckTargetNodeIPolicy
47

    
48
import ganeti.masterd.instance
49

    
50

    
51
_DISK_TEMPLATE_NAME_PREFIX = {
52
  constants.DT_PLAIN: "",
53
  constants.DT_RBD: ".rbd",
54
  constants.DT_EXT: ".ext",
55
  }
56

    
57

    
58
_DISK_TEMPLATE_DEVICE_TYPE = {
59
  constants.DT_PLAIN: constants.LD_LV,
60
  constants.DT_FILE: constants.LD_FILE,
61
  constants.DT_SHARED_FILE: constants.LD_FILE,
62
  constants.DT_BLOCK: constants.LD_BLOCKDEV,
63
  constants.DT_RBD: constants.LD_RBD,
64
  constants.DT_EXT: constants.LD_EXT,
65
  }
66

    
67

    
68
def CreateSingleBlockDev(lu, node_uuid, instance, device, info, force_open,
69
                         excl_stor):
70
  """Create a single block device on a given node.
71

72
  This will not recurse over children of the device, so they must be
73
  created in advance.
74

75
  @param lu: the lu on whose behalf we execute
76
  @param node_uuid: the node on which to create the device
77
  @type instance: L{objects.Instance}
78
  @param instance: the instance which owns the device
79
  @type device: L{objects.Disk}
80
  @param device: the device to create
81
  @param info: the extra 'metadata' we should attach to the device
82
      (this will be represented as a LVM tag)
83
  @type force_open: boolean
84
  @param force_open: this parameter will be passes to the
85
      L{backend.BlockdevCreate} function where it specifies
86
      whether we run on primary or not, and it affects both
87
      the child assembly and the device own Open() execution
88
  @type excl_stor: boolean
89
  @param excl_stor: Whether exclusive_storage is active for the node
90

91
  """
92
  lu.cfg.SetDiskID(device, node_uuid)
93
  result = lu.rpc.call_blockdev_create(node_uuid, device, device.size,
94
                                       instance.name, force_open, info,
95
                                       excl_stor)
96
  result.Raise("Can't create block device %s on"
97
               " node %s for instance %s" % (device,
98
                                             lu.cfg.GetNodeName(node_uuid),
99
                                             instance.name))
100
  if device.physical_id is None:
101
    device.physical_id = result.payload
102

    
103

    
104
def _CreateBlockDevInner(lu, node_uuid, instance, device, force_create,
105
                         info, force_open, excl_stor):
106
  """Create a tree of block devices on a given node.
107

108
  If this device type has to be created on secondaries, create it and
109
  all its children.
110

111
  If not, just recurse to children keeping the same 'force' value.
112

113
  @attention: The device has to be annotated already.
114

115
  @param lu: the lu on whose behalf we execute
116
  @param node_uuid: the node on which to create the device
117
  @type instance: L{objects.Instance}
118
  @param instance: the instance which owns the device
119
  @type device: L{objects.Disk}
120
  @param device: the device to create
121
  @type force_create: boolean
122
  @param force_create: whether to force creation of this device; this
123
      will be change to True whenever we find a device which has
124
      CreateOnSecondary() attribute
125
  @param info: the extra 'metadata' we should attach to the device
126
      (this will be represented as a LVM tag)
127
  @type force_open: boolean
128
  @param force_open: this parameter will be passes to the
129
      L{backend.BlockdevCreate} function where it specifies
130
      whether we run on primary or not, and it affects both
131
      the child assembly and the device own Open() execution
132
  @type excl_stor: boolean
133
  @param excl_stor: Whether exclusive_storage is active for the node
134

135
  @return: list of created devices
136
  """
137
  created_devices = []
138
  try:
139
    if device.CreateOnSecondary():
140
      force_create = True
141

    
142
    if device.children:
143
      for child in device.children:
144
        devs = _CreateBlockDevInner(lu, node_uuid, instance, child,
145
                                    force_create, info, force_open, excl_stor)
146
        created_devices.extend(devs)
147

    
148
    if not force_create:
149
      return created_devices
150

    
151
    CreateSingleBlockDev(lu, node_uuid, instance, device, info, force_open,
152
                         excl_stor)
153
    # The device has been completely created, so there is no point in keeping
154
    # its subdevices in the list. We just add the device itself instead.
155
    created_devices = [(node_uuid, device)]
156
    return created_devices
157

    
158
  except errors.DeviceCreationError, e:
159
    e.created_devices.extend(created_devices)
160
    raise e
161
  except errors.OpExecError, e:
162
    raise errors.DeviceCreationError(str(e), created_devices)
163

    
164

    
165
def IsExclusiveStorageEnabledNodeUuid(cfg, node_uuid):
166
  """Whether exclusive_storage is in effect for the given node.
167

168
  @type cfg: L{config.ConfigWriter}
169
  @param cfg: The cluster configuration
170
  @type node_uuid: string
171
  @param node_uuid: The node UUID
172
  @rtype: bool
173
  @return: The effective value of exclusive_storage
174
  @raise errors.OpPrereqError: if no node exists with the given name
175

176
  """
177
  ni = cfg.GetNodeInfo(node_uuid)
178
  if ni is None:
179
    raise errors.OpPrereqError("Invalid node UUID %s" % node_uuid,
180
                               errors.ECODE_NOENT)
181
  return IsExclusiveStorageEnabledNode(cfg, ni)
182

    
183

    
184
def _CreateBlockDev(lu, node_uuid, instance, device, force_create, info,
185
                    force_open):
186
  """Wrapper around L{_CreateBlockDevInner}.
187

188
  This method annotates the root device first.
189

190
  """
191
  (disk,) = AnnotateDiskParams(instance, [device], lu.cfg)
192
  excl_stor = IsExclusiveStorageEnabledNodeUuid(lu.cfg, node_uuid)
193
  return _CreateBlockDevInner(lu, node_uuid, instance, disk, force_create, info,
194
                              force_open, excl_stor)
195

    
196

    
197
def _UndoCreateDisks(lu, disks_created):
198
  """Undo the work performed by L{CreateDisks}.
199

200
  This function is called in case of an error to undo the work of
201
  L{CreateDisks}.
202

203
  @type lu: L{LogicalUnit}
204
  @param lu: the logical unit on whose behalf we execute
205
  @param disks_created: the result returned by L{CreateDisks}
206

207
  """
208
  for (node_uuid, disk) in disks_created:
209
    lu.cfg.SetDiskID(disk, node_uuid)
210
    result = lu.rpc.call_blockdev_remove(node_uuid, disk)
211
    result.Warn("Failed to remove newly-created disk %s on node %s" %
212
                (disk, lu.cfg.GetNodeName(node_uuid)), logging.warning)
213

    
214

    
215
def CreateDisks(lu, instance, to_skip=None, target_node_uuid=None, disks=None):
216
  """Create all disks for an instance.
217

218
  This abstracts away some work from AddInstance.
219

220
  @type lu: L{LogicalUnit}
221
  @param lu: the logical unit on whose behalf we execute
222
  @type instance: L{objects.Instance}
223
  @param instance: the instance whose disks we should create
224
  @type to_skip: list
225
  @param to_skip: list of indices to skip
226
  @type target_node_uuid: string
227
  @param target_node_uuid: if passed, overrides the target node for creation
228
  @type disks: list of {objects.Disk}
229
  @param disks: the disks to create; if not specified, all the disks of the
230
      instance are created
231
  @return: information about the created disks, to be used to call
232
      L{_UndoCreateDisks}
233
  @raise errors.OpPrereqError: in case of error
234

235
  """
236
  info = GetInstanceInfoText(instance)
237
  if target_node_uuid is None:
238
    pnode_uuid = instance.primary_node
239
    all_node_uuids = instance.all_nodes
240
  else:
241
    pnode_uuid = target_node_uuid
242
    all_node_uuids = [pnode_uuid]
243

    
244
  if disks is None:
245
    disks = instance.disks
246

    
247
  if instance.disk_template in constants.DTS_FILEBASED:
248
    file_storage_dir = os.path.dirname(instance.disks[0].logical_id[1])
249
    result = lu.rpc.call_file_storage_dir_create(pnode_uuid, file_storage_dir)
250

    
251
    result.Raise("Failed to create directory '%s' on"
252
                 " node %s" % (file_storage_dir,
253
                               lu.cfg.GetNodeName(pnode_uuid)))
254

    
255
  disks_created = []
256
  for idx, device in enumerate(disks):
257
    if to_skip and idx in to_skip:
258
      continue
259
    logging.info("Creating disk %s for instance '%s'", idx, instance.name)
260
    for node_uuid in all_node_uuids:
261
      f_create = node_uuid == pnode_uuid
262
      try:
263
        _CreateBlockDev(lu, node_uuid, instance, device, f_create, info,
264
                        f_create)
265
        disks_created.append((node_uuid, device))
266
      except errors.DeviceCreationError, e:
267
        logging.warning("Creating disk %s for instance '%s' failed",
268
                        idx, instance.name)
269
        disks_created.extend(e.created_devices)
270
        _UndoCreateDisks(lu, disks_created)
271
        raise errors.OpExecError(e.message)
272
  return disks_created
273

    
274

    
275
def ComputeDiskSizePerVG(disk_template, disks):
276
  """Compute disk size requirements in the volume group
277

278
  """
279
  def _compute(disks, payload):
280
    """Universal algorithm.
281

282
    """
283
    vgs = {}
284
    for disk in disks:
285
      vgs[disk[constants.IDISK_VG]] = \
286
        vgs.get(constants.IDISK_VG, 0) + disk[constants.IDISK_SIZE] + payload
287

    
288
    return vgs
289

    
290
  # Required free disk space as a function of disk and swap space
291
  req_size_dict = {
292
    constants.DT_DISKLESS: {},
293
    constants.DT_PLAIN: _compute(disks, 0),
294
    # 128 MB are added for drbd metadata for each disk
295
    constants.DT_DRBD8: _compute(disks, constants.DRBD_META_SIZE),
296
    constants.DT_FILE: {},
297
    constants.DT_SHARED_FILE: {},
298
    }
299

    
300
  if disk_template not in req_size_dict:
301
    raise errors.ProgrammerError("Disk template '%s' size requirement"
302
                                 " is unknown" % disk_template)
303

    
304
  return req_size_dict[disk_template]
305

    
306

    
307
def ComputeDisks(op, default_vg):
308
  """Computes the instance disks.
309

310
  @param op: The instance opcode
311
  @param default_vg: The default_vg to assume
312

313
  @return: The computed disks
314

315
  """
316
  disks = []
317
  for disk in op.disks:
318
    mode = disk.get(constants.IDISK_MODE, constants.DISK_RDWR)
319
    if mode not in constants.DISK_ACCESS_SET:
320
      raise errors.OpPrereqError("Invalid disk access mode '%s'" %
321
                                 mode, errors.ECODE_INVAL)
322
    size = disk.get(constants.IDISK_SIZE, None)
323
    if size is None:
324
      raise errors.OpPrereqError("Missing disk size", errors.ECODE_INVAL)
325
    try:
326
      size = int(size)
327
    except (TypeError, ValueError):
328
      raise errors.OpPrereqError("Invalid disk size '%s'" % size,
329
                                 errors.ECODE_INVAL)
330

    
331
    ext_provider = disk.get(constants.IDISK_PROVIDER, None)
332
    if ext_provider and op.disk_template != constants.DT_EXT:
333
      raise errors.OpPrereqError("The '%s' option is only valid for the %s"
334
                                 " disk template, not %s" %
335
                                 (constants.IDISK_PROVIDER, constants.DT_EXT,
336
                                  op.disk_template), errors.ECODE_INVAL)
337

    
338
    data_vg = disk.get(constants.IDISK_VG, default_vg)
339
    name = disk.get(constants.IDISK_NAME, None)
340
    if name is not None and name.lower() == constants.VALUE_NONE:
341
      name = None
342
    new_disk = {
343
      constants.IDISK_SIZE: size,
344
      constants.IDISK_MODE: mode,
345
      constants.IDISK_VG: data_vg,
346
      constants.IDISK_NAME: name,
347
      }
348

    
349
    for key in [
350
      constants.IDISK_METAVG,
351
      constants.IDISK_ADOPT,
352
      constants.IDISK_SPINDLES,
353
      ]:
354
      if key in disk:
355
        new_disk[key] = disk[key]
356

    
357
    # For extstorage, demand the `provider' option and add any
358
    # additional parameters (ext-params) to the dict
359
    if op.disk_template == constants.DT_EXT:
360
      if ext_provider:
361
        new_disk[constants.IDISK_PROVIDER] = ext_provider
362
        for key in disk:
363
          if key not in constants.IDISK_PARAMS:
364
            new_disk[key] = disk[key]
365
      else:
366
        raise errors.OpPrereqError("Missing provider for template '%s'" %
367
                                   constants.DT_EXT, errors.ECODE_INVAL)
368

    
369
    disks.append(new_disk)
370

    
371
  return disks
372

    
373

    
374
def CheckRADOSFreeSpace():
375
  """Compute disk size requirements inside the RADOS cluster.
376

377
  """
378
  # For the RADOS cluster we assume there is always enough space.
379
  pass
380

    
381

    
382
def _GenerateDRBD8Branch(lu, primary_uuid, secondary_uuid, size, vgnames, names,
383
                         iv_name, p_minor, s_minor):
384
  """Generate a drbd8 device complete with its children.
385

386
  """
387
  assert len(vgnames) == len(names) == 2
388
  port = lu.cfg.AllocatePort()
389
  shared_secret = lu.cfg.GenerateDRBDSecret(lu.proc.GetECId())
390

    
391
  dev_data = objects.Disk(dev_type=constants.LD_LV, size=size,
392
                          logical_id=(vgnames[0], names[0]),
393
                          params={})
394
  dev_data.uuid = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
395
  dev_meta = objects.Disk(dev_type=constants.LD_LV,
396
                          size=constants.DRBD_META_SIZE,
397
                          logical_id=(vgnames[1], names[1]),
398
                          params={})
399
  dev_meta.uuid = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
400
  drbd_dev = objects.Disk(dev_type=constants.LD_DRBD8, size=size,
401
                          logical_id=(primary_uuid, secondary_uuid, port,
402
                                      p_minor, s_minor,
403
                                      shared_secret),
404
                          children=[dev_data, dev_meta],
405
                          iv_name=iv_name, params={})
406
  drbd_dev.uuid = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
407
  return drbd_dev
408

    
409

    
410
def GenerateDiskTemplate(
411
  lu, template_name, instance_uuid, primary_node_uuid, secondary_node_uuids,
412
  disk_info, file_storage_dir, file_driver, base_index,
413
  feedback_fn, full_disk_params, _req_file_storage=opcodes.RequireFileStorage,
414
  _req_shr_file_storage=opcodes.RequireSharedFileStorage):
415
  """Generate the entire disk layout for a given template type.
416

417
  """
418
  vgname = lu.cfg.GetVGName()
419
  disk_count = len(disk_info)
420
  disks = []
421

    
422
  if template_name == constants.DT_DISKLESS:
423
    pass
424
  elif template_name == constants.DT_DRBD8:
425
    if len(secondary_node_uuids) != 1:
426
      raise errors.ProgrammerError("Wrong template configuration")
427
    remote_node_uuid = secondary_node_uuids[0]
428
    minors = lu.cfg.AllocateDRBDMinor(
429
      [primary_node_uuid, remote_node_uuid] * len(disk_info), instance_uuid)
430

    
431
    (drbd_params, _, _) = objects.Disk.ComputeLDParams(template_name,
432
                                                       full_disk_params)
433
    drbd_default_metavg = drbd_params[constants.LDP_DEFAULT_METAVG]
434

    
435
    names = []
436
    for lv_prefix in _GenerateUniqueNames(lu, [".disk%d" % (base_index + i)
437
                                               for i in range(disk_count)]):
438
      names.append(lv_prefix + "_data")
439
      names.append(lv_prefix + "_meta")
440
    for idx, disk in enumerate(disk_info):
441
      disk_index = idx + base_index
442
      data_vg = disk.get(constants.IDISK_VG, vgname)
443
      meta_vg = disk.get(constants.IDISK_METAVG, drbd_default_metavg)
444
      disk_dev = _GenerateDRBD8Branch(lu, primary_node_uuid, remote_node_uuid,
445
                                      disk[constants.IDISK_SIZE],
446
                                      [data_vg, meta_vg],
447
                                      names[idx * 2:idx * 2 + 2],
448
                                      "disk/%d" % disk_index,
449
                                      minors[idx * 2], minors[idx * 2 + 1])
450
      disk_dev.mode = disk[constants.IDISK_MODE]
451
      disk_dev.name = disk.get(constants.IDISK_NAME, None)
452
      disks.append(disk_dev)
453
  else:
454
    if secondary_node_uuids:
455
      raise errors.ProgrammerError("Wrong template configuration")
456

    
457
    if template_name == constants.DT_FILE:
458
      _req_file_storage()
459
    elif template_name == constants.DT_SHARED_FILE:
460
      _req_shr_file_storage()
461

    
462
    name_prefix = _DISK_TEMPLATE_NAME_PREFIX.get(template_name, None)
463
    if name_prefix is None:
464
      names = None
465
    else:
466
      names = _GenerateUniqueNames(lu, ["%s.disk%s" %
467
                                        (name_prefix, base_index + i)
468
                                        for i in range(disk_count)])
469

    
470
    if template_name == constants.DT_PLAIN:
471

    
472
      def logical_id_fn(idx, _, disk):
473
        vg = disk.get(constants.IDISK_VG, vgname)
474
        return (vg, names[idx])
475

    
476
    elif template_name in (constants.DT_FILE, constants.DT_SHARED_FILE):
477
      logical_id_fn = \
478
        lambda _, disk_index, disk: (file_driver,
479
                                     "%s/disk%d" % (file_storage_dir,
480
                                                    disk_index))
481
    elif template_name == constants.DT_BLOCK:
482
      logical_id_fn = \
483
        lambda idx, disk_index, disk: (constants.BLOCKDEV_DRIVER_MANUAL,
484
                                       disk[constants.IDISK_ADOPT])
485
    elif template_name == constants.DT_RBD:
486
      logical_id_fn = lambda idx, _, disk: ("rbd", names[idx])
487
    elif template_name == constants.DT_EXT:
488
      def logical_id_fn(idx, _, disk):
489
        provider = disk.get(constants.IDISK_PROVIDER, None)
490
        if provider is None:
491
          raise errors.ProgrammerError("Disk template is %s, but '%s' is"
492
                                       " not found", constants.DT_EXT,
493
                                       constants.IDISK_PROVIDER)
494
        return (provider, names[idx])
495
    else:
496
      raise errors.ProgrammerError("Unknown disk template '%s'" % template_name)
497

    
498
    dev_type = _DISK_TEMPLATE_DEVICE_TYPE[template_name]
499

    
500
    for idx, disk in enumerate(disk_info):
501
      params = {}
502
      # Only for the Ext template add disk_info to params
503
      if template_name == constants.DT_EXT:
504
        params[constants.IDISK_PROVIDER] = disk[constants.IDISK_PROVIDER]
505
        for key in disk:
506
          if key not in constants.IDISK_PARAMS:
507
            params[key] = disk[key]
508
      disk_index = idx + base_index
509
      size = disk[constants.IDISK_SIZE]
510
      feedback_fn("* disk %s, size %s" %
511
                  (disk_index, utils.FormatUnit(size, "h")))
512
      disk_dev = objects.Disk(dev_type=dev_type, size=size,
513
                              logical_id=logical_id_fn(idx, disk_index, disk),
514
                              iv_name="disk/%d" % disk_index,
515
                              mode=disk[constants.IDISK_MODE],
516
                              params=params,
517
                              spindles=disk.get(constants.IDISK_SPINDLES))
518
      disk_dev.name = disk.get(constants.IDISK_NAME, None)
519
      disk_dev.uuid = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
520
      disks.append(disk_dev)
521

    
522
  return disks
523

    
524

    
525
def CheckSpindlesExclusiveStorage(diskdict, es_flag, required):
526
  """Check the presence of the spindle options with exclusive_storage.
527

528
  @type diskdict: dict
529
  @param diskdict: disk parameters
530
  @type es_flag: bool
531
  @param es_flag: the effective value of the exlusive_storage flag
532
  @type required: bool
533
  @param required: whether spindles are required or just optional
534
  @raise errors.OpPrereqError when spindles are given and they should not
535

536
  """
537
  if (not es_flag and constants.IDISK_SPINDLES in diskdict and
538
      diskdict[constants.IDISK_SPINDLES] is not None):
539
    raise errors.OpPrereqError("Spindles in instance disks cannot be specified"
540
                               " when exclusive storage is not active",
541
                               errors.ECODE_INVAL)
542
  if (es_flag and required and (constants.IDISK_SPINDLES not in diskdict or
543
                                diskdict[constants.IDISK_SPINDLES] is None)):
544
    raise errors.OpPrereqError("You must specify spindles in instance disks"
545
                               " when exclusive storage is active",
546
                               errors.ECODE_INVAL)
547

    
548

    
549
class LUInstanceRecreateDisks(LogicalUnit):
550
  """Recreate an instance's missing disks.
551

552
  """
553
  HPATH = "instance-recreate-disks"
554
  HTYPE = constants.HTYPE_INSTANCE
555
  REQ_BGL = False
556

    
557
  _MODIFYABLE = compat.UniqueFrozenset([
558
    constants.IDISK_SIZE,
559
    constants.IDISK_MODE,
560
    constants.IDISK_SPINDLES,
561
    ])
562

    
563
  # New or changed disk parameters may have different semantics
564
  assert constants.IDISK_PARAMS == (_MODIFYABLE | frozenset([
565
    constants.IDISK_ADOPT,
566

    
567
    # TODO: Implement support changing VG while recreating
568
    constants.IDISK_VG,
569
    constants.IDISK_METAVG,
570
    constants.IDISK_PROVIDER,
571
    constants.IDISK_NAME,
572
    ]))
573

    
574
  def _RunAllocator(self):
575
    """Run the allocator based on input opcode.
576

577
    """
578
    be_full = self.cfg.GetClusterInfo().FillBE(self.instance)
579

    
580
    # FIXME
581
    # The allocator should actually run in "relocate" mode, but current
582
    # allocators don't support relocating all the nodes of an instance at
583
    # the same time. As a workaround we use "allocate" mode, but this is
584
    # suboptimal for two reasons:
585
    # - The instance name passed to the allocator is present in the list of
586
    #   existing instances, so there could be a conflict within the
587
    #   internal structures of the allocator. This doesn't happen with the
588
    #   current allocators, but it's a liability.
589
    # - The allocator counts the resources used by the instance twice: once
590
    #   because the instance exists already, and once because it tries to
591
    #   allocate a new instance.
592
    # The allocator could choose some of the nodes on which the instance is
593
    # running, but that's not a problem. If the instance nodes are broken,
594
    # they should be already be marked as drained or offline, and hence
595
    # skipped by the allocator. If instance disks have been lost for other
596
    # reasons, then recreating the disks on the same nodes should be fine.
597
    disk_template = self.instance.disk_template
598
    spindle_use = be_full[constants.BE_SPINDLE_USE]
599
    disks = [{
600
      constants.IDISK_SIZE: d.size,
601
      constants.IDISK_MODE: d.mode,
602
      constants.IDISK_SPINDLES: d.spindles,
603
      } for d in self.instance.disks]
604
    req = iallocator.IAReqInstanceAlloc(name=self.op.instance_name,
605
                                        disk_template=disk_template,
606
                                        tags=list(self.instance.GetTags()),
607
                                        os=self.instance.os,
608
                                        nics=[{}],
609
                                        vcpus=be_full[constants.BE_VCPUS],
610
                                        memory=be_full[constants.BE_MAXMEM],
611
                                        spindle_use=spindle_use,
612
                                        disks=disks,
613
                                        hypervisor=self.instance.hypervisor,
614
                                        node_whitelist=None)
615
    ial = iallocator.IAllocator(self.cfg, self.rpc, req)
616

    
617
    ial.Run(self.op.iallocator)
618

    
619
    assert req.RequiredNodes() == len(self.instance.all_nodes)
620

    
621
    if not ial.success:
622
      raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
623
                                 " %s" % (self.op.iallocator, ial.info),
624
                                 errors.ECODE_NORES)
625

    
626
    (self.op.node_uuids, self.op.nodes) = GetWantedNodes(self, ial.result)
627
    self.LogInfo("Selected nodes for instance %s via iallocator %s: %s",
628
                 self.op.instance_name, self.op.iallocator,
629
                 utils.CommaJoin(self.op.nodes))
630

    
631
  def CheckArguments(self):
632
    if self.op.disks and ht.TNonNegativeInt(self.op.disks[0]):
633
      # Normalize and convert deprecated list of disk indices
634
      self.op.disks = [(idx, {}) for idx in sorted(frozenset(self.op.disks))]
635

    
636
    duplicates = utils.FindDuplicates(map(compat.fst, self.op.disks))
637
    if duplicates:
638
      raise errors.OpPrereqError("Some disks have been specified more than"
639
                                 " once: %s" % utils.CommaJoin(duplicates),
640
                                 errors.ECODE_INVAL)
641

    
642
    # We don't want _CheckIAllocatorOrNode selecting the default iallocator
643
    # when neither iallocator nor nodes are specified
644
    if self.op.iallocator or self.op.nodes:
645
      CheckIAllocatorOrNode(self, "iallocator", "nodes")
646

    
647
    for (idx, params) in self.op.disks:
648
      utils.ForceDictType(params, constants.IDISK_PARAMS_TYPES)
649
      unsupported = frozenset(params.keys()) - self._MODIFYABLE
650
      if unsupported:
651
        raise errors.OpPrereqError("Parameters for disk %s try to change"
652
                                   " unmodifyable parameter(s): %s" %
653
                                   (idx, utils.CommaJoin(unsupported)),
654
                                   errors.ECODE_INVAL)
655

    
656
  def ExpandNames(self):
657
    self._ExpandAndLockInstance()
658
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
659

    
660
    if self.op.nodes:
661
      (self.op.node_uuids, self.op.nodes) = GetWantedNodes(self, self.op.nodes)
662
      self.needed_locks[locking.LEVEL_NODE] = list(self.op.node_uuids)
663
    else:
664
      self.needed_locks[locking.LEVEL_NODE] = []
665
      if self.op.iallocator:
666
        # iallocator will select a new node in the same group
667
        self.needed_locks[locking.LEVEL_NODEGROUP] = []
668
        self.needed_locks[locking.LEVEL_NODE_ALLOC] = locking.ALL_SET
669

    
670
    self.needed_locks[locking.LEVEL_NODE_RES] = []
671

    
672
  def DeclareLocks(self, level):
673
    if level == locking.LEVEL_NODEGROUP:
674
      assert self.op.iallocator is not None
675
      assert not self.op.nodes
676
      assert not self.needed_locks[locking.LEVEL_NODEGROUP]
677
      self.share_locks[locking.LEVEL_NODEGROUP] = 1
678
      # Lock the primary group used by the instance optimistically; this
679
      # requires going via the node before it's locked, requiring
680
      # verification later on
681
      self.needed_locks[locking.LEVEL_NODEGROUP] = \
682
        self.cfg.GetInstanceNodeGroups(self.op.instance_uuid, primary_only=True)
683

    
684
    elif level == locking.LEVEL_NODE:
685
      # If an allocator is used, then we lock all the nodes in the current
686
      # instance group, as we don't know yet which ones will be selected;
687
      # if we replace the nodes without using an allocator, locks are
688
      # already declared in ExpandNames; otherwise, we need to lock all the
689
      # instance nodes for disk re-creation
690
      if self.op.iallocator:
691
        assert not self.op.nodes
692
        assert not self.needed_locks[locking.LEVEL_NODE]
693
        assert len(self.owned_locks(locking.LEVEL_NODEGROUP)) == 1
694

    
695
        # Lock member nodes of the group of the primary node
696
        for group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP):
697
          self.needed_locks[locking.LEVEL_NODE].extend(
698
            self.cfg.GetNodeGroup(group_uuid).members)
699

    
700
        assert locking.NAL in self.owned_locks(locking.LEVEL_NODE_ALLOC)
701
      elif not self.op.nodes:
702
        self._LockInstancesNodes(primary_only=False)
703
    elif level == locking.LEVEL_NODE_RES:
704
      # Copy node locks
705
      self.needed_locks[locking.LEVEL_NODE_RES] = \
706
        CopyLockList(self.needed_locks[locking.LEVEL_NODE])
707

    
708
  def BuildHooksEnv(self):
709
    """Build hooks env.
710

711
    This runs on master, primary and secondary nodes of the instance.
712

713
    """
714
    return BuildInstanceHookEnvByObject(self, self.instance)
715

    
716
  def BuildHooksNodes(self):
717
    """Build hooks nodes.
718

719
    """
720
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
721
    return (nl, nl)
722

    
723
  def CheckPrereq(self):
724
    """Check prerequisites.
725

726
    This checks that the instance is in the cluster and is not running.
727

728
    """
729
    instance = self.cfg.GetInstanceInfo(self.op.instance_uuid)
730
    assert instance is not None, \
731
      "Cannot retrieve locked instance %s" % self.op.instance_name
732
    if self.op.node_uuids:
733
      if len(self.op.node_uuids) != len(instance.all_nodes):
734
        raise errors.OpPrereqError("Instance %s currently has %d nodes, but"
735
                                   " %d replacement nodes were specified" %
736
                                   (instance.name, len(instance.all_nodes),
737
                                    len(self.op.node_uuids)),
738
                                   errors.ECODE_INVAL)
739
      assert instance.disk_template != constants.DT_DRBD8 or \
740
             len(self.op.node_uuids) == 2
741
      assert instance.disk_template != constants.DT_PLAIN or \
742
             len(self.op.node_uuids) == 1
743
      primary_node = self.op.node_uuids[0]
744
    else:
745
      primary_node = instance.primary_node
746
    if not self.op.iallocator:
747
      CheckNodeOnline(self, primary_node)
748

    
749
    if instance.disk_template == constants.DT_DISKLESS:
750
      raise errors.OpPrereqError("Instance '%s' has no disks" %
751
                                 self.op.instance_name, errors.ECODE_INVAL)
752

    
753
    # Verify if node group locks are still correct
754
    owned_groups = self.owned_locks(locking.LEVEL_NODEGROUP)
755
    if owned_groups:
756
      # Node group locks are acquired only for the primary node (and only
757
      # when the allocator is used)
758
      CheckInstanceNodeGroups(self.cfg, instance.uuid, owned_groups,
759
                              primary_only=True)
760

    
761
    # if we replace nodes *and* the old primary is offline, we don't
762
    # check the instance state
763
    old_pnode = self.cfg.GetNodeInfo(instance.primary_node)
764
    if not ((self.op.iallocator or self.op.node_uuids) and old_pnode.offline):
765
      CheckInstanceState(self, instance, INSTANCE_NOT_RUNNING,
766
                         msg="cannot recreate disks")
767

    
768
    if self.op.disks:
769
      self.disks = dict(self.op.disks)
770
    else:
771
      self.disks = dict((idx, {}) for idx in range(len(instance.disks)))
772

    
773
    maxidx = max(self.disks.keys())
774
    if maxidx >= len(instance.disks):
775
      raise errors.OpPrereqError("Invalid disk index '%s'" % maxidx,
776
                                 errors.ECODE_INVAL)
777

    
778
    if ((self.op.node_uuids or self.op.iallocator) and
779
         sorted(self.disks.keys()) != range(len(instance.disks))):
780
      raise errors.OpPrereqError("Can't recreate disks partially and"
781
                                 " change the nodes at the same time",
782
                                 errors.ECODE_INVAL)
783

    
784
    self.instance = instance
785

    
786
    if self.op.iallocator:
787
      self._RunAllocator()
788
      # Release unneeded node and node resource locks
789
      ReleaseLocks(self, locking.LEVEL_NODE, keep=self.op.node_uuids)
790
      ReleaseLocks(self, locking.LEVEL_NODE_RES, keep=self.op.node_uuids)
791
      ReleaseLocks(self, locking.LEVEL_NODE_ALLOC)
792

    
793
    assert not self.glm.is_owned(locking.LEVEL_NODE_ALLOC)
794

    
795
    if self.op.node_uuids:
796
      node_uuids = self.op.node_uuids
797
    else:
798
      node_uuids = instance.all_nodes
799
    excl_stor = compat.any(
800
      rpc.GetExclusiveStorageForNodes(self.cfg, node_uuids).values()
801
      )
802
    for new_params in self.disks.values():
803
      CheckSpindlesExclusiveStorage(new_params, excl_stor, False)
804

    
805
  def Exec(self, feedback_fn):
806
    """Recreate the disks.
807

808
    """
809
    assert (self.owned_locks(locking.LEVEL_NODE) ==
810
            self.owned_locks(locking.LEVEL_NODE_RES))
811

    
812
    to_skip = []
813
    mods = [] # keeps track of needed changes
814

    
815
    for idx, disk in enumerate(self.instance.disks):
816
      try:
817
        changes = self.disks[idx]
818
      except KeyError:
819
        # Disk should not be recreated
820
        to_skip.append(idx)
821
        continue
822

    
823
      # update secondaries for disks, if needed
824
      if self.op.node_uuids and disk.dev_type == constants.LD_DRBD8:
825
        # need to update the nodes and minors
826
        assert len(self.op.node_uuids) == 2
827
        assert len(disk.logical_id) == 6 # otherwise disk internals
828
                                         # have changed
829
        (_, _, old_port, _, _, old_secret) = disk.logical_id
830
        new_minors = self.cfg.AllocateDRBDMinor(self.op.node_uuids,
831
                                                self.instance.uuid)
832
        new_id = (self.op.node_uuids[0], self.op.node_uuids[1], old_port,
833
                  new_minors[0], new_minors[1], old_secret)
834
        assert len(disk.logical_id) == len(new_id)
835
      else:
836
        new_id = None
837

    
838
      mods.append((idx, new_id, changes))
839

    
840
    # now that we have passed all asserts above, we can apply the mods
841
    # in a single run (to avoid partial changes)
842
    for idx, new_id, changes in mods:
843
      disk = self.instance.disks[idx]
844
      if new_id is not None:
845
        assert disk.dev_type == constants.LD_DRBD8
846
        disk.logical_id = new_id
847
      if changes:
848
        disk.Update(size=changes.get(constants.IDISK_SIZE, None),
849
                    mode=changes.get(constants.IDISK_MODE, None),
850
                    spindles=changes.get(constants.IDISK_SPINDLES, None))
851

    
852
    # change primary node, if needed
853
    if self.op.node_uuids:
854
      self.instance.primary_node = self.op.node_uuids[0]
855
      self.LogWarning("Changing the instance's nodes, you will have to"
856
                      " remove any disks left on the older nodes manually")
857

    
858
    if self.op.node_uuids:
859
      self.cfg.Update(self.instance, feedback_fn)
860

    
861
    # All touched nodes must be locked
862
    mylocks = self.owned_locks(locking.LEVEL_NODE)
863
    assert mylocks.issuperset(frozenset(self.instance.all_nodes))
864
    new_disks = CreateDisks(self, self.instance, to_skip=to_skip)
865

    
866
    # TODO: Release node locks before wiping, or explain why it's not possible
867
    if self.cfg.GetClusterInfo().prealloc_wipe_disks:
868
      wipedisks = [(idx, disk, 0)
869
                   for (idx, disk) in enumerate(self.instance.disks)
870
                   if idx not in to_skip]
871
      WipeOrCleanupDisks(self, self.instance, disks=wipedisks,
872
                         cleanup=new_disks)
873

    
874

    
875
def _CheckNodesFreeDiskOnVG(lu, node_uuids, vg, requested):
876
  """Checks if nodes have enough free disk space in the specified VG.
877

878
  This function checks if all given nodes have the needed amount of
879
  free disk. In case any node has less disk or we cannot get the
880
  information from the node, this function raises an OpPrereqError
881
  exception.
882

883
  @type lu: C{LogicalUnit}
884
  @param lu: a logical unit from which we get configuration data
885
  @type node_uuids: C{list}
886
  @param node_uuids: the list of node UUIDs to check
887
  @type vg: C{str}
888
  @param vg: the volume group to check
889
  @type requested: C{int}
890
  @param requested: the amount of disk in MiB to check for
891
  @raise errors.OpPrereqError: if the node doesn't have enough disk,
892
      or we cannot check the node
893

894
  """
895
  es_flags = rpc.GetExclusiveStorageForNodes(lu.cfg, node_uuids)
896
  hvname = lu.cfg.GetHypervisorType()
897
  hvparams = lu.cfg.GetClusterInfo().hvparams
898
  nodeinfo = lu.rpc.call_node_info(node_uuids, [(constants.ST_LVM_VG, vg)],
899
                                   [(hvname, hvparams[hvname])], es_flags)
900
  for node in node_uuids:
901
    node_name = lu.cfg.GetNodeName(node)
902

    
903
    info = nodeinfo[node]
904
    info.Raise("Cannot get current information from node %s" % node_name,
905
               prereq=True, ecode=errors.ECODE_ENVIRON)
906
    (_, (vg_info, ), _) = info.payload
907
    vg_free = vg_info.get("storage_free", None)
908
    if not isinstance(vg_free, int):
909
      raise errors.OpPrereqError("Can't compute free disk space on node"
910
                                 " %s for vg %s, result was '%s'" %
911
                                 (node_name, vg, vg_free), errors.ECODE_ENVIRON)
912
    if requested > vg_free:
913
      raise errors.OpPrereqError("Not enough disk space on target node %s"
914
                                 " vg %s: required %d MiB, available %d MiB" %
915
                                 (node_name, vg, requested, vg_free),
916
                                 errors.ECODE_NORES)
917

    
918

    
919
def CheckNodesFreeDiskPerVG(lu, node_uuids, req_sizes):
920
  """Checks if nodes have enough free disk space in all the VGs.
921

922
  This function checks if all given nodes have the needed amount of
923
  free disk. In case any node has less disk or we cannot get the
924
  information from the node, this function raises an OpPrereqError
925
  exception.
926

927
  @type lu: C{LogicalUnit}
928
  @param lu: a logical unit from which we get configuration data
929
  @type node_uuids: C{list}
930
  @param node_uuids: the list of node UUIDs to check
931
  @type req_sizes: C{dict}
932
  @param req_sizes: the hash of vg and corresponding amount of disk in
933
      MiB to check for
934
  @raise errors.OpPrereqError: if the node doesn't have enough disk,
935
      or we cannot check the node
936

937
  """
938
  for vg, req_size in req_sizes.items():
939
    _CheckNodesFreeDiskOnVG(lu, node_uuids, vg, req_size)
940

    
941

    
942
def _DiskSizeInBytesToMebibytes(lu, size):
943
  """Converts a disk size in bytes to mebibytes.
944

945
  Warns and rounds up if the size isn't an even multiple of 1 MiB.
946

947
  """
948
  (mib, remainder) = divmod(size, 1024 * 1024)
949

    
950
  if remainder != 0:
951
    lu.LogWarning("Disk size is not an even multiple of 1 MiB; rounding up"
952
                  " to not overwrite existing data (%s bytes will not be"
953
                  " wiped)", (1024 * 1024) - remainder)
954
    mib += 1
955

    
956
  return mib
957

    
958

    
959
def _CalcEta(time_taken, written, total_size):
960
  """Calculates the ETA based on size written and total size.
961

962
  @param time_taken: The time taken so far
963
  @param written: amount written so far
964
  @param total_size: The total size of data to be written
965
  @return: The remaining time in seconds
966

967
  """
968
  avg_time = time_taken / float(written)
969
  return (total_size - written) * avg_time
970

    
971

    
972
def WipeDisks(lu, instance, disks=None):
973
  """Wipes instance disks.
974

975
  @type lu: L{LogicalUnit}
976
  @param lu: the logical unit on whose behalf we execute
977
  @type instance: L{objects.Instance}
978
  @param instance: the instance whose disks we should create
979
  @type disks: None or list of tuple of (number, L{objects.Disk}, number)
980
  @param disks: Disk details; tuple contains disk index, disk object and the
981
    start offset
982

983
  """
984
  node_uuid = instance.primary_node
985
  node_name = lu.cfg.GetNodeName(node_uuid)
986

    
987
  if disks is None:
988
    disks = [(idx, disk, 0)
989
             for (idx, disk) in enumerate(instance.disks)]
990

    
991
  for (_, device, _) in disks:
992
    lu.cfg.SetDiskID(device, node_uuid)
993

    
994
  logging.info("Pausing synchronization of disks of instance '%s'",
995
               instance.name)
996
  result = lu.rpc.call_blockdev_pause_resume_sync(node_uuid,
997
                                                  (map(compat.snd, disks),
998
                                                   instance),
999
                                                  True)
1000
  result.Raise("Failed to pause disk synchronization on node '%s'" % node_name)
1001

    
1002
  for idx, success in enumerate(result.payload):
1003
    if not success:
1004
      logging.warn("Pausing synchronization of disk %s of instance '%s'"
1005
                   " failed", idx, instance.name)
1006

    
1007
  try:
1008
    for (idx, device, offset) in disks:
1009
      # The wipe size is MIN_WIPE_CHUNK_PERCENT % of the instance disk but
1010
      # MAX_WIPE_CHUNK at max. Truncating to integer to avoid rounding errors.
1011
      wipe_chunk_size = \
1012
        int(min(constants.MAX_WIPE_CHUNK,
1013
                device.size / 100.0 * constants.MIN_WIPE_CHUNK_PERCENT))
1014

    
1015
      size = device.size
1016
      last_output = 0
1017
      start_time = time.time()
1018

    
1019
      if offset == 0:
1020
        info_text = ""
1021
      else:
1022
        info_text = (" (from %s to %s)" %
1023
                     (utils.FormatUnit(offset, "h"),
1024
                      utils.FormatUnit(size, "h")))
1025

    
1026
      lu.LogInfo("* Wiping disk %s%s", idx, info_text)
1027

    
1028
      logging.info("Wiping disk %d for instance %s on node %s using"
1029
                   " chunk size %s", idx, instance.name, node_name,
1030
                   wipe_chunk_size)
1031

    
1032
      while offset < size:
1033
        wipe_size = min(wipe_chunk_size, size - offset)
1034

    
1035
        logging.debug("Wiping disk %d, offset %s, chunk %s",
1036
                      idx, offset, wipe_size)
1037

    
1038
        result = lu.rpc.call_blockdev_wipe(node_uuid, (device, instance),
1039
                                           offset, wipe_size)
1040
        result.Raise("Could not wipe disk %d at offset %d for size %d" %
1041
                     (idx, offset, wipe_size))
1042

    
1043
        now = time.time()
1044
        offset += wipe_size
1045
        if now - last_output >= 60:
1046
          eta = _CalcEta(now - start_time, offset, size)
1047
          lu.LogInfo(" - done: %.1f%% ETA: %s",
1048
                     offset / float(size) * 100, utils.FormatSeconds(eta))
1049
          last_output = now
1050
  finally:
1051
    logging.info("Resuming synchronization of disks for instance '%s'",
1052
                 instance.name)
1053

    
1054
    result = lu.rpc.call_blockdev_pause_resume_sync(node_uuid,
1055
                                                    (map(compat.snd, disks),
1056
                                                     instance),
1057
                                                    False)
1058

    
1059
    if result.fail_msg:
1060
      lu.LogWarning("Failed to resume disk synchronization on node '%s': %s",
1061
                    node_name, result.fail_msg)
1062
    else:
1063
      for idx, success in enumerate(result.payload):
1064
        if not success:
1065
          lu.LogWarning("Resuming synchronization of disk %s of instance '%s'"
1066
                        " failed", idx, instance.name)
1067

    
1068

    
1069
def WipeOrCleanupDisks(lu, instance, disks=None, cleanup=None):
1070
  """Wrapper for L{WipeDisks} that handles errors.
1071

1072
  @type lu: L{LogicalUnit}
1073
  @param lu: the logical unit on whose behalf we execute
1074
  @type instance: L{objects.Instance}
1075
  @param instance: the instance whose disks we should wipe
1076
  @param disks: see L{WipeDisks}
1077
  @param cleanup: the result returned by L{CreateDisks}, used for cleanup in
1078
      case of error
1079
  @raise errors.OpPrereqError: in case of failure
1080

1081
  """
1082
  try:
1083
    WipeDisks(lu, instance, disks=disks)
1084
  except errors.OpExecError:
1085
    logging.warning("Wiping disks for instance '%s' failed",
1086
                    instance.name)
1087
    _UndoCreateDisks(lu, cleanup)
1088
    raise
1089

    
1090

    
1091
def ExpandCheckDisks(instance, disks):
1092
  """Return the instance disks selected by the disks list
1093

1094
  @type disks: list of L{objects.Disk} or None
1095
  @param disks: selected disks
1096
  @rtype: list of L{objects.Disk}
1097
  @return: selected instance disks to act on
1098

1099
  """
1100
  if disks is None:
1101
    return instance.disks
1102
  else:
1103
    if not set(disks).issubset(instance.disks):
1104
      raise errors.ProgrammerError("Can only act on disks belonging to the"
1105
                                   " target instance: expected a subset of %r,"
1106
                                   " got %r" % (instance.disks, disks))
1107
    return disks
1108

    
1109

    
1110
def WaitForSync(lu, instance, disks=None, oneshot=False):
1111
  """Sleep and poll for an instance's disk to sync.
1112

1113
  """
1114
  if not instance.disks or disks is not None and not disks:
1115
    return True
1116

    
1117
  disks = ExpandCheckDisks(instance, disks)
1118

    
1119
  if not oneshot:
1120
    lu.LogInfo("Waiting for instance %s to sync disks", instance.name)
1121

    
1122
  node_uuid = instance.primary_node
1123
  node_name = lu.cfg.GetNodeName(node_uuid)
1124

    
1125
  for dev in disks:
1126
    lu.cfg.SetDiskID(dev, node_uuid)
1127

    
1128
  # TODO: Convert to utils.Retry
1129

    
1130
  retries = 0
1131
  degr_retries = 10 # in seconds, as we sleep 1 second each time
1132
  while True:
1133
    max_time = 0
1134
    done = True
1135
    cumul_degraded = False
1136
    rstats = lu.rpc.call_blockdev_getmirrorstatus(node_uuid, (disks, instance))
1137
    msg = rstats.fail_msg
1138
    if msg:
1139
      lu.LogWarning("Can't get any data from node %s: %s", node_name, msg)
1140
      retries += 1
1141
      if retries >= 10:
1142
        raise errors.RemoteError("Can't contact node %s for mirror data,"
1143
                                 " aborting." % node_name)
1144
      time.sleep(6)
1145
      continue
1146
    rstats = rstats.payload
1147
    retries = 0
1148
    for i, mstat in enumerate(rstats):
1149
      if mstat is None:
1150
        lu.LogWarning("Can't compute data for node %s/%s",
1151
                      node_name, disks[i].iv_name)
1152
        continue
1153

    
1154
      cumul_degraded = (cumul_degraded or
1155
                        (mstat.is_degraded and mstat.sync_percent is None))
1156
      if mstat.sync_percent is not None:
1157
        done = False
1158
        if mstat.estimated_time is not None:
1159
          rem_time = ("%s remaining (estimated)" %
1160
                      utils.FormatSeconds(mstat.estimated_time))
1161
          max_time = mstat.estimated_time
1162
        else:
1163
          rem_time = "no time estimate"
1164
        lu.LogInfo("- device %s: %5.2f%% done, %s",
1165
                   disks[i].iv_name, mstat.sync_percent, rem_time)
1166

    
1167
    # if we're done but degraded, let's do a few small retries, to
1168
    # make sure we see a stable and not transient situation; therefore
1169
    # we force restart of the loop
1170
    if (done or oneshot) and cumul_degraded and degr_retries > 0:
1171
      logging.info("Degraded disks found, %d retries left", degr_retries)
1172
      degr_retries -= 1
1173
      time.sleep(1)
1174
      continue
1175

    
1176
    if done or oneshot:
1177
      break
1178

    
1179
    time.sleep(min(60, max_time))
1180

    
1181
  if done:
1182
    lu.LogInfo("Instance %s's disks are in sync", instance.name)
1183

    
1184
  return not cumul_degraded
1185

    
1186

    
1187
def ShutdownInstanceDisks(lu, instance, disks=None, ignore_primary=False):
1188
  """Shutdown block devices of an instance.
1189

1190
  This does the shutdown on all nodes of the instance.
1191

1192
  If the ignore_primary is false, errors on the primary node are
1193
  ignored.
1194

1195
  """
1196
  lu.cfg.MarkInstanceDisksInactive(instance.uuid)
1197
  all_result = True
1198
  disks = ExpandCheckDisks(instance, disks)
1199

    
1200
  for disk in disks:
1201
    for node_uuid, top_disk in disk.ComputeNodeTree(instance.primary_node):
1202
      lu.cfg.SetDiskID(top_disk, node_uuid)
1203
      result = lu.rpc.call_blockdev_shutdown(node_uuid, (top_disk, instance))
1204
      msg = result.fail_msg
1205
      if msg:
1206
        lu.LogWarning("Could not shutdown block device %s on node %s: %s",
1207
                      disk.iv_name, lu.cfg.GetNodeName(node_uuid), msg)
1208
        if ((node_uuid == instance.primary_node and not ignore_primary) or
1209
            (node_uuid != instance.primary_node and not result.offline)):
1210
          all_result = False
1211
  return all_result
1212

    
1213

    
1214
def _SafeShutdownInstanceDisks(lu, instance, disks=None):
1215
  """Shutdown block devices of an instance.
1216

1217
  This function checks if an instance is running, before calling
1218
  _ShutdownInstanceDisks.
1219

1220
  """
1221
  CheckInstanceState(lu, instance, INSTANCE_DOWN, msg="cannot shutdown disks")
1222
  ShutdownInstanceDisks(lu, instance, disks=disks)
1223

    
1224

    
1225
def AssembleInstanceDisks(lu, instance, disks=None, ignore_secondaries=False,
1226
                           ignore_size=False):
1227
  """Prepare the block devices for an instance.
1228

1229
  This sets up the block devices on all nodes.
1230

1231
  @type lu: L{LogicalUnit}
1232
  @param lu: the logical unit on whose behalf we execute
1233
  @type instance: L{objects.Instance}
1234
  @param instance: the instance for whose disks we assemble
1235
  @type disks: list of L{objects.Disk} or None
1236
  @param disks: which disks to assemble (or all, if None)
1237
  @type ignore_secondaries: boolean
1238
  @param ignore_secondaries: if true, errors on secondary nodes
1239
      won't result in an error return from the function
1240
  @type ignore_size: boolean
1241
  @param ignore_size: if true, the current known size of the disk
1242
      will not be used during the disk activation, useful for cases
1243
      when the size is wrong
1244
  @return: False if the operation failed, otherwise a list of
1245
      (host, instance_visible_name, node_visible_name)
1246
      with the mapping from node devices to instance devices
1247

1248
  """
1249
  device_info = []
1250
  disks_ok = True
1251
  disks = ExpandCheckDisks(instance, disks)
1252

    
1253
  # With the two passes mechanism we try to reduce the window of
1254
  # opportunity for the race condition of switching DRBD to primary
1255
  # before handshaking occured, but we do not eliminate it
1256

    
1257
  # The proper fix would be to wait (with some limits) until the
1258
  # connection has been made and drbd transitions from WFConnection
1259
  # into any other network-connected state (Connected, SyncTarget,
1260
  # SyncSource, etc.)
1261

    
1262
  # mark instance disks as active before doing actual work, so watcher does
1263
  # not try to shut them down erroneously
1264
  lu.cfg.MarkInstanceDisksActive(instance.uuid)
1265

    
1266
  # 1st pass, assemble on all nodes in secondary mode
1267
  for idx, inst_disk in enumerate(disks):
1268
    for node_uuid, node_disk in inst_disk.ComputeNodeTree(
1269
                                  instance.primary_node):
1270
      if ignore_size:
1271
        node_disk = node_disk.Copy()
1272
        node_disk.UnsetSize()
1273
      lu.cfg.SetDiskID(node_disk, node_uuid)
1274
      result = lu.rpc.call_blockdev_assemble(node_uuid, (node_disk, instance),
1275
                                             instance.name, False, idx)
1276
      msg = result.fail_msg
1277
      if msg:
1278
        is_offline_secondary = (node_uuid in instance.secondary_nodes and
1279
                                result.offline)
1280
        lu.LogWarning("Could not prepare block device %s on node %s"
1281
                      " (is_primary=False, pass=1): %s",
1282
                      inst_disk.iv_name, lu.cfg.GetNodeName(node_uuid), msg)
1283
        if not (ignore_secondaries or is_offline_secondary):
1284
          disks_ok = False
1285

    
1286
  # FIXME: race condition on drbd migration to primary
1287

    
1288
  # 2nd pass, do only the primary node
1289
  for idx, inst_disk in enumerate(disks):
1290
    dev_path = None
1291

    
1292
    for node_uuid, node_disk in inst_disk.ComputeNodeTree(
1293
                                  instance.primary_node):
1294
      if node_uuid != instance.primary_node:
1295
        continue
1296
      if ignore_size:
1297
        node_disk = node_disk.Copy()
1298
        node_disk.UnsetSize()
1299
      lu.cfg.SetDiskID(node_disk, node_uuid)
1300
      result = lu.rpc.call_blockdev_assemble(node_uuid, (node_disk, instance),
1301
                                             instance.name, True, idx)
1302
      msg = result.fail_msg
1303
      if msg:
1304
        lu.LogWarning("Could not prepare block device %s on node %s"
1305
                      " (is_primary=True, pass=2): %s",
1306
                      inst_disk.iv_name, lu.cfg.GetNodeName(node_uuid), msg)
1307
        disks_ok = False
1308
      else:
1309
        dev_path = result.payload
1310

    
1311
    device_info.append((lu.cfg.GetNodeName(instance.primary_node),
1312
                        inst_disk.iv_name, dev_path))
1313

    
1314
  # leave the disks configured for the primary node
1315
  # this is a workaround that would be fixed better by
1316
  # improving the logical/physical id handling
1317
  for disk in disks:
1318
    lu.cfg.SetDiskID(disk, instance.primary_node)
1319

    
1320
  if not disks_ok:
1321
    lu.cfg.MarkInstanceDisksInactive(instance.uuid)
1322

    
1323
  return disks_ok, device_info
1324

    
1325

    
1326
def StartInstanceDisks(lu, instance, force):
1327
  """Start the disks of an instance.
1328

1329
  """
1330
  disks_ok, _ = AssembleInstanceDisks(lu, instance,
1331
                                      ignore_secondaries=force)
1332
  if not disks_ok:
1333
    ShutdownInstanceDisks(lu, instance)
1334
    if force is not None and not force:
1335
      lu.LogWarning("",
1336
                    hint=("If the message above refers to a secondary node,"
1337
                          " you can retry the operation using '--force'"))
1338
    raise errors.OpExecError("Disk consistency error")
1339

    
1340

    
1341
class LUInstanceGrowDisk(LogicalUnit):
1342
  """Grow a disk of an instance.
1343

1344
  """
1345
  HPATH = "disk-grow"
1346
  HTYPE = constants.HTYPE_INSTANCE
1347
  REQ_BGL = False
1348

    
1349
  def ExpandNames(self):
1350
    self._ExpandAndLockInstance()
1351
    self.needed_locks[locking.LEVEL_NODE] = []
1352
    self.needed_locks[locking.LEVEL_NODE_RES] = []
1353
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
1354
    self.recalculate_locks[locking.LEVEL_NODE_RES] = constants.LOCKS_REPLACE
1355

    
1356
  def DeclareLocks(self, level):
1357
    if level == locking.LEVEL_NODE:
1358
      self._LockInstancesNodes()
1359
    elif level == locking.LEVEL_NODE_RES:
1360
      # Copy node locks
1361
      self.needed_locks[locking.LEVEL_NODE_RES] = \
1362
        CopyLockList(self.needed_locks[locking.LEVEL_NODE])
1363

    
1364
  def BuildHooksEnv(self):
1365
    """Build hooks env.
1366

1367
    This runs on the master, the primary and all the secondaries.
1368

1369
    """
1370
    env = {
1371
      "DISK": self.op.disk,
1372
      "AMOUNT": self.op.amount,
1373
      "ABSOLUTE": self.op.absolute,
1374
      }
1375
    env.update(BuildInstanceHookEnvByObject(self, self.instance))
1376
    return env
1377

    
1378
  def BuildHooksNodes(self):
1379
    """Build hooks nodes.
1380

1381
    """
1382
    nl = [self.cfg.GetMasterNode()] + list(self.instance.all_nodes)
1383
    return (nl, nl)
1384

    
1385
  def CheckPrereq(self):
1386
    """Check prerequisites.
1387

1388
    This checks that the instance is in the cluster.
1389

1390
    """
1391
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_uuid)
1392
    assert self.instance is not None, \
1393
      "Cannot retrieve locked instance %s" % self.op.instance_name
1394
    node_uuids = list(self.instance.all_nodes)
1395
    for node_uuid in node_uuids:
1396
      CheckNodeOnline(self, node_uuid)
1397

    
1398
    if self.instance.disk_template not in constants.DTS_GROWABLE:
1399
      raise errors.OpPrereqError("Instance's disk layout does not support"
1400
                                 " growing", errors.ECODE_INVAL)
1401

    
1402
    self.disk = self.instance.FindDisk(self.op.disk)
1403

    
1404
    if self.op.absolute:
1405
      self.target = self.op.amount
1406
      self.delta = self.target - self.disk.size
1407
      if self.delta < 0:
1408
        raise errors.OpPrereqError("Requested size (%s) is smaller than "
1409
                                   "current disk size (%s)" %
1410
                                   (utils.FormatUnit(self.target, "h"),
1411
                                    utils.FormatUnit(self.disk.size, "h")),
1412
                                   errors.ECODE_STATE)
1413
    else:
1414
      self.delta = self.op.amount
1415
      self.target = self.disk.size + self.delta
1416
      if self.delta < 0:
1417
        raise errors.OpPrereqError("Requested increment (%s) is negative" %
1418
                                   utils.FormatUnit(self.delta, "h"),
1419
                                   errors.ECODE_INVAL)
1420

    
1421
    self._CheckDiskSpace(node_uuids, self.disk.ComputeGrowth(self.delta))
1422

    
1423
  def _CheckDiskSpace(self, node_uuids, req_vgspace):
1424
    template = self.instance.disk_template
1425
    if template not in (constants.DTS_NO_FREE_SPACE_CHECK):
1426
      # TODO: check the free disk space for file, when that feature will be
1427
      # supported
1428
      nodes = map(self.cfg.GetNodeInfo, node_uuids)
1429
      es_nodes = filter(lambda n: IsExclusiveStorageEnabledNode(self.cfg, n),
1430
                        nodes)
1431
      if es_nodes:
1432
        # With exclusive storage we need to something smarter than just looking
1433
        # at free space; for now, let's simply abort the operation.
1434
        raise errors.OpPrereqError("Cannot grow disks when exclusive_storage"
1435
                                   " is enabled", errors.ECODE_STATE)
1436
      CheckNodesFreeDiskPerVG(self, node_uuids, req_vgspace)
1437

    
1438
  def Exec(self, feedback_fn):
1439
    """Execute disk grow.
1440

1441
    """
1442
    assert set([self.instance.name]) == self.owned_locks(locking.LEVEL_INSTANCE)
1443
    assert (self.owned_locks(locking.LEVEL_NODE) ==
1444
            self.owned_locks(locking.LEVEL_NODE_RES))
1445

    
1446
    wipe_disks = self.cfg.GetClusterInfo().prealloc_wipe_disks
1447

    
1448
    disks_ok, _ = AssembleInstanceDisks(self, self.instance, disks=[self.disk])
1449
    if not disks_ok:
1450
      raise errors.OpExecError("Cannot activate block device to grow")
1451

    
1452
    feedback_fn("Growing disk %s of instance '%s' by %s to %s" %
1453
                (self.op.disk, self.instance.name,
1454
                 utils.FormatUnit(self.delta, "h"),
1455
                 utils.FormatUnit(self.target, "h")))
1456

    
1457
    # First run all grow ops in dry-run mode
1458
    for node_uuid in self.instance.all_nodes:
1459
      self.cfg.SetDiskID(self.disk, node_uuid)
1460
      result = self.rpc.call_blockdev_grow(node_uuid,
1461
                                           (self.disk, self.instance),
1462
                                           self.delta, True, True)
1463
      result.Raise("Dry-run grow request failed to node %s" %
1464
                   self.cfg.GetNodeName(node_uuid))
1465

    
1466
    if wipe_disks:
1467
      # Get disk size from primary node for wiping
1468
      result = self.rpc.call_blockdev_getdimensions(self.instance.primary_node,
1469
                                                    [self.disk])
1470
      result.Raise("Failed to retrieve disk size from node '%s'" %
1471
                   self.instance.primary_node)
1472

    
1473
      (disk_dimensions, ) = result.payload
1474

    
1475
      if disk_dimensions is None:
1476
        raise errors.OpExecError("Failed to retrieve disk size from primary"
1477
                                 " node '%s'" % self.instance.primary_node)
1478
      (disk_size_in_bytes, _) = disk_dimensions
1479

    
1480
      old_disk_size = _DiskSizeInBytesToMebibytes(self, disk_size_in_bytes)
1481

    
1482
      assert old_disk_size >= self.disk.size, \
1483
        ("Retrieved disk size too small (got %s, should be at least %s)" %
1484
         (old_disk_size, self.disk.size))
1485
    else:
1486
      old_disk_size = None
1487

    
1488
    # We know that (as far as we can test) operations across different
1489
    # nodes will succeed, time to run it for real on the backing storage
1490
    for node_uuid in self.instance.all_nodes:
1491
      self.cfg.SetDiskID(self.disk, node_uuid)
1492
      result = self.rpc.call_blockdev_grow(node_uuid,
1493
                                           (self.disk, self.instance),
1494
                                           self.delta, False, True)
1495
      result.Raise("Grow request failed to node %s" %
1496
                   self.cfg.GetNodeName(node_uuid))
1497

    
1498
    # And now execute it for logical storage, on the primary node
1499
    node_uuid = self.instance.primary_node
1500
    self.cfg.SetDiskID(self.disk, node_uuid)
1501
    result = self.rpc.call_blockdev_grow(node_uuid, (self.disk, self.instance),
1502
                                         self.delta, False, False)
1503
    result.Raise("Grow request failed to node %s" %
1504
                 self.cfg.GetNodeName(node_uuid))
1505

    
1506
    self.disk.RecordGrow(self.delta)
1507
    self.cfg.Update(self.instance, feedback_fn)
1508

    
1509
    # Changes have been recorded, release node lock
1510
    ReleaseLocks(self, locking.LEVEL_NODE)
1511

    
1512
    # Downgrade lock while waiting for sync
1513
    self.glm.downgrade(locking.LEVEL_INSTANCE)
1514

    
1515
    assert wipe_disks ^ (old_disk_size is None)
1516

    
1517
    if wipe_disks:
1518
      assert self.instance.disks[self.op.disk] == self.disk
1519

    
1520
      # Wipe newly added disk space
1521
      WipeDisks(self, self.instance,
1522
                disks=[(self.op.disk, self.disk, old_disk_size)])
1523

    
1524
    if self.op.wait_for_sync:
1525
      disk_abort = not WaitForSync(self, self.instance, disks=[self.disk])
1526
      if disk_abort:
1527
        self.LogWarning("Disk syncing has not returned a good status; check"
1528
                        " the instance")
1529
      if not self.instance.disks_active:
1530
        _SafeShutdownInstanceDisks(self, self.instance, disks=[self.disk])
1531
    elif not self.instance.disks_active:
1532
      self.LogWarning("Not shutting down the disk even if the instance is"
1533
                      " not supposed to be running because no wait for"
1534
                      " sync mode was requested")
1535

    
1536
    assert self.owned_locks(locking.LEVEL_NODE_RES)
1537
    assert set([self.instance.name]) == self.owned_locks(locking.LEVEL_INSTANCE)
1538

    
1539

    
1540
class LUInstanceReplaceDisks(LogicalUnit):
1541
  """Replace the disks of an instance.
1542

1543
  """
1544
  HPATH = "mirrors-replace"
1545
  HTYPE = constants.HTYPE_INSTANCE
1546
  REQ_BGL = False
1547

    
1548
  def CheckArguments(self):
1549
    """Check arguments.
1550

1551
    """
1552
    if self.op.mode == constants.REPLACE_DISK_CHG:
1553
      if self.op.remote_node is None and self.op.iallocator is None:
1554
        raise errors.OpPrereqError("When changing the secondary either an"
1555
                                   " iallocator script must be used or the"
1556
                                   " new node given", errors.ECODE_INVAL)
1557
      else:
1558
        CheckIAllocatorOrNode(self, "iallocator", "remote_node")
1559

    
1560
    elif self.op.remote_node is not None or self.op.iallocator is not None:
1561
      # Not replacing the secondary
1562
      raise errors.OpPrereqError("The iallocator and new node options can"
1563
                                 " only be used when changing the"
1564
                                 " secondary node", errors.ECODE_INVAL)
1565

    
1566
  def ExpandNames(self):
1567
    self._ExpandAndLockInstance()
1568

    
1569
    assert locking.LEVEL_NODE not in self.needed_locks
1570
    assert locking.LEVEL_NODE_RES not in self.needed_locks
1571
    assert locking.LEVEL_NODEGROUP not in self.needed_locks
1572

    
1573
    assert self.op.iallocator is None or self.op.remote_node is None, \
1574
      "Conflicting options"
1575

    
1576
    if self.op.remote_node is not None:
1577
      (self.op.remote_node_uuid, self.op.remote_node) = \
1578
        ExpandNodeUuidAndName(self.cfg, self.op.remote_node_uuid,
1579
                              self.op.remote_node)
1580

    
1581
      # Warning: do not remove the locking of the new secondary here
1582
      # unless DRBD8Dev.AddChildren is changed to work in parallel;
1583
      # currently it doesn't since parallel invocations of
1584
      # FindUnusedMinor will conflict
1585
      self.needed_locks[locking.LEVEL_NODE] = [self.op.remote_node_uuid]
1586
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_APPEND
1587
    else:
1588
      self.needed_locks[locking.LEVEL_NODE] = []
1589
      self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
1590

    
1591
      if self.op.iallocator is not None:
1592
        # iallocator will select a new node in the same group
1593
        self.needed_locks[locking.LEVEL_NODEGROUP] = []
1594
        self.needed_locks[locking.LEVEL_NODE_ALLOC] = locking.ALL_SET
1595

    
1596
    self.needed_locks[locking.LEVEL_NODE_RES] = []
1597

    
1598
    self.replacer = TLReplaceDisks(self, self.op.instance_uuid,
1599
                                   self.op.instance_name, self.op.mode,
1600
                                   self.op.iallocator, self.op.remote_node_uuid,
1601
                                   self.op.disks, self.op.early_release,
1602
                                   self.op.ignore_ipolicy)
1603

    
1604
    self.tasklets = [self.replacer]
1605

    
1606
  def DeclareLocks(self, level):
1607
    if level == locking.LEVEL_NODEGROUP:
1608
      assert self.op.remote_node_uuid is None
1609
      assert self.op.iallocator is not None
1610
      assert not self.needed_locks[locking.LEVEL_NODEGROUP]
1611

    
1612
      self.share_locks[locking.LEVEL_NODEGROUP] = 1
1613
      # Lock all groups used by instance optimistically; this requires going
1614
      # via the node before it's locked, requiring verification later on
1615
      self.needed_locks[locking.LEVEL_NODEGROUP] = \
1616
        self.cfg.GetInstanceNodeGroups(self.op.instance_uuid)
1617

    
1618
    elif level == locking.LEVEL_NODE:
1619
      if self.op.iallocator is not None:
1620
        assert self.op.remote_node_uuid is None
1621
        assert not self.needed_locks[locking.LEVEL_NODE]
1622
        assert locking.NAL in self.owned_locks(locking.LEVEL_NODE_ALLOC)
1623

    
1624
        # Lock member nodes of all locked groups
1625
        self.needed_locks[locking.LEVEL_NODE] = \
1626
          [node_uuid
1627
           for group_uuid in self.owned_locks(locking.LEVEL_NODEGROUP)
1628
           for node_uuid in self.cfg.GetNodeGroup(group_uuid).members]
1629
      else:
1630
        assert not self.glm.is_owned(locking.LEVEL_NODE_ALLOC)
1631

    
1632
        self._LockInstancesNodes()
1633

    
1634
    elif level == locking.LEVEL_NODE_RES:
1635
      # Reuse node locks
1636
      self.needed_locks[locking.LEVEL_NODE_RES] = \
1637
        self.needed_locks[locking.LEVEL_NODE]
1638

    
1639
  def BuildHooksEnv(self):
1640
    """Build hooks env.
1641

1642
    This runs on the master, the primary and all the secondaries.
1643

1644
    """
1645
    instance = self.replacer.instance
1646
    env = {
1647
      "MODE": self.op.mode,
1648
      "NEW_SECONDARY": self.op.remote_node,
1649
      "OLD_SECONDARY": self.cfg.GetNodeName(instance.secondary_nodes[0]),
1650
      }
1651
    env.update(BuildInstanceHookEnvByObject(self, instance))
1652
    return env
1653

    
1654
  def BuildHooksNodes(self):
1655
    """Build hooks nodes.
1656

1657
    """
1658
    instance = self.replacer.instance
1659
    nl = [
1660
      self.cfg.GetMasterNode(),
1661
      instance.primary_node,
1662
      ]
1663
    if self.op.remote_node_uuid is not None:
1664
      nl.append(self.op.remote_node_uuid)
1665
    return nl, nl
1666

    
1667
  def CheckPrereq(self):
1668
    """Check prerequisites.
1669

1670
    """
1671
    assert (self.glm.is_owned(locking.LEVEL_NODEGROUP) or
1672
            self.op.iallocator is None)
1673

    
1674
    # Verify if node group locks are still correct
1675
    owned_groups = self.owned_locks(locking.LEVEL_NODEGROUP)
1676
    if owned_groups:
1677
      CheckInstanceNodeGroups(self.cfg, self.op.instance_uuid, owned_groups)
1678

    
1679
    return LogicalUnit.CheckPrereq(self)
1680

    
1681

    
1682
class LUInstanceActivateDisks(NoHooksLU):
1683
  """Bring up an instance's disks.
1684

1685
  """
1686
  REQ_BGL = False
1687

    
1688
  def ExpandNames(self):
1689
    self._ExpandAndLockInstance()
1690
    self.needed_locks[locking.LEVEL_NODE] = []
1691
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
1692

    
1693
  def DeclareLocks(self, level):
1694
    if level == locking.LEVEL_NODE:
1695
      self._LockInstancesNodes()
1696

    
1697
  def CheckPrereq(self):
1698
    """Check prerequisites.
1699

1700
    This checks that the instance is in the cluster.
1701

1702
    """
1703
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_uuid)
1704
    assert self.instance is not None, \
1705
      "Cannot retrieve locked instance %s" % self.op.instance_name
1706
    CheckNodeOnline(self, self.instance.primary_node)
1707

    
1708
  def Exec(self, feedback_fn):
1709
    """Activate the disks.
1710

1711
    """
1712
    disks_ok, disks_info = \
1713
              AssembleInstanceDisks(self, self.instance,
1714
                                    ignore_size=self.op.ignore_size)
1715
    if not disks_ok:
1716
      raise errors.OpExecError("Cannot activate block devices")
1717

    
1718
    if self.op.wait_for_sync:
1719
      if not WaitForSync(self, self.instance):
1720
        self.cfg.MarkInstanceDisksInactive(self.instance.uuid)
1721
        raise errors.OpExecError("Some disks of the instance are degraded!")
1722

    
1723
    return disks_info
1724

    
1725

    
1726
class LUInstanceDeactivateDisks(NoHooksLU):
1727
  """Shutdown an instance's disks.
1728

1729
  """
1730
  REQ_BGL = False
1731

    
1732
  def ExpandNames(self):
1733
    self._ExpandAndLockInstance()
1734
    self.needed_locks[locking.LEVEL_NODE] = []
1735
    self.recalculate_locks[locking.LEVEL_NODE] = constants.LOCKS_REPLACE
1736

    
1737
  def DeclareLocks(self, level):
1738
    if level == locking.LEVEL_NODE:
1739
      self._LockInstancesNodes()
1740

    
1741
  def CheckPrereq(self):
1742
    """Check prerequisites.
1743

1744
    This checks that the instance is in the cluster.
1745

1746
    """
1747
    self.instance = self.cfg.GetInstanceInfo(self.op.instance_uuid)
1748
    assert self.instance is not None, \
1749
      "Cannot retrieve locked instance %s" % self.op.instance_name
1750

    
1751
  def Exec(self, feedback_fn):
1752
    """Deactivate the disks
1753

1754
    """
1755
    if self.op.force:
1756
      ShutdownInstanceDisks(self, self.instance)
1757
    else:
1758
      _SafeShutdownInstanceDisks(self, self.instance)
1759

    
1760

    
1761
def _CheckDiskConsistencyInner(lu, instance, dev, node_uuid, on_primary,
1762
                               ldisk=False):
1763
  """Check that mirrors are not degraded.
1764

1765
  @attention: The device has to be annotated already.
1766

1767
  The ldisk parameter, if True, will change the test from the
1768
  is_degraded attribute (which represents overall non-ok status for
1769
  the device(s)) to the ldisk (representing the local storage status).
1770

1771
  """
1772
  lu.cfg.SetDiskID(dev, node_uuid)
1773

    
1774
  result = True
1775

    
1776
  if on_primary or dev.AssembleOnSecondary():
1777
    rstats = lu.rpc.call_blockdev_find(node_uuid, dev)
1778
    msg = rstats.fail_msg
1779
    if msg:
1780
      lu.LogWarning("Can't find disk on node %s: %s",
1781
                    lu.cfg.GetNodeName(node_uuid), msg)
1782
      result = False
1783
    elif not rstats.payload:
1784
      lu.LogWarning("Can't find disk on node %s", lu.cfg.GetNodeName(node_uuid))
1785
      result = False
1786
    else:
1787
      if ldisk:
1788
        result = result and rstats.payload.ldisk_status == constants.LDS_OKAY
1789
      else:
1790
        result = result and not rstats.payload.is_degraded
1791

    
1792
  if dev.children:
1793
    for child in dev.children:
1794
      result = result and _CheckDiskConsistencyInner(lu, instance, child,
1795
                                                     node_uuid, on_primary)
1796

    
1797
  return result
1798

    
1799

    
1800
def CheckDiskConsistency(lu, instance, dev, node_uuid, on_primary, ldisk=False):
1801
  """Wrapper around L{_CheckDiskConsistencyInner}.
1802

1803
  """
1804
  (disk,) = AnnotateDiskParams(instance, [dev], lu.cfg)
1805
  return _CheckDiskConsistencyInner(lu, instance, disk, node_uuid, on_primary,
1806
                                    ldisk=ldisk)
1807

    
1808

    
1809
def _BlockdevFind(lu, node_uuid, dev, instance):
1810
  """Wrapper around call_blockdev_find to annotate diskparams.
1811

1812
  @param lu: A reference to the lu object
1813
  @param node_uuid: The node to call out
1814
  @param dev: The device to find
1815
  @param instance: The instance object the device belongs to
1816
  @returns The result of the rpc call
1817

1818
  """
1819
  (disk,) = AnnotateDiskParams(instance, [dev], lu.cfg)
1820
  return lu.rpc.call_blockdev_find(node_uuid, disk)
1821

    
1822

    
1823
def _GenerateUniqueNames(lu, exts):
1824
  """Generate a suitable LV name.
1825

1826
  This will generate a logical volume name for the given instance.
1827

1828
  """
1829
  results = []
1830
  for val in exts:
1831
    new_id = lu.cfg.GenerateUniqueID(lu.proc.GetECId())
1832
    results.append("%s%s" % (new_id, val))
1833
  return results
1834

    
1835

    
1836
class TLReplaceDisks(Tasklet):
1837
  """Replaces disks for an instance.
1838

1839
  Note: Locking is not within the scope of this class.
1840

1841
  """
1842
  def __init__(self, lu, instance_uuid, instance_name, mode, iallocator_name,
1843
               remote_node_uuid, disks, early_release, ignore_ipolicy):
1844
    """Initializes this class.
1845

1846
    """
1847
    Tasklet.__init__(self, lu)
1848

    
1849
    # Parameters
1850
    self.instance_uuid = instance_uuid
1851
    self.instance_name = instance_name
1852
    self.mode = mode
1853
    self.iallocator_name = iallocator_name
1854
    self.remote_node_uuid = remote_node_uuid
1855
    self.disks = disks
1856
    self.early_release = early_release
1857
    self.ignore_ipolicy = ignore_ipolicy
1858

    
1859
    # Runtime data
1860
    self.instance = None
1861
    self.new_node_uuid = None
1862
    self.target_node_uuid = None
1863
    self.other_node_uuid = None
1864
    self.remote_node_info = None
1865
    self.node_secondary_ip = None
1866

    
1867
  @staticmethod
1868
  def _RunAllocator(lu, iallocator_name, instance_uuid,
1869
                    relocate_from_node_uuids):
1870
    """Compute a new secondary node using an IAllocator.
1871

1872
    """
1873
    req = iallocator.IAReqRelocate(
1874
          inst_uuid=instance_uuid,
1875
          relocate_from_node_uuids=list(relocate_from_node_uuids))
1876
    ial = iallocator.IAllocator(lu.cfg, lu.rpc, req)
1877

    
1878
    ial.Run(iallocator_name)
1879

    
1880
    if not ial.success:
1881
      raise errors.OpPrereqError("Can't compute nodes using iallocator '%s':"
1882
                                 " %s" % (iallocator_name, ial.info),
1883
                                 errors.ECODE_NORES)
1884

    
1885
    remote_node_name = ial.result[0]
1886
    remote_node = lu.cfg.GetNodeInfoByName(remote_node_name)
1887

    
1888
    if remote_node is None:
1889
      raise errors.OpPrereqError("Node %s not found in configuration" %
1890
                                 remote_node_name, errors.ECODE_NOENT)
1891

    
1892
    lu.LogInfo("Selected new secondary for instance '%s': %s",
1893
               instance_uuid, remote_node_name)
1894

    
1895
    return remote_node.uuid
1896

    
1897
  def _FindFaultyDisks(self, node_uuid):
1898
    """Wrapper for L{FindFaultyInstanceDisks}.
1899

1900
    """
1901
    return FindFaultyInstanceDisks(self.cfg, self.rpc, self.instance,
1902
                                   node_uuid, True)
1903

    
1904
  def _CheckDisksActivated(self, instance):
1905
    """Checks if the instance disks are activated.
1906

1907
    @param instance: The instance to check disks
1908
    @return: True if they are activated, False otherwise
1909

1910
    """
1911
    node_uuids = instance.all_nodes
1912

    
1913
    for idx, dev in enumerate(instance.disks):
1914
      for node_uuid in node_uuids:
1915
        self.lu.LogInfo("Checking disk/%d on %s", idx,
1916
                        self.cfg.GetNodeName(node_uuid))
1917
        self.cfg.SetDiskID(dev, node_uuid)
1918

    
1919
        result = _BlockdevFind(self, node_uuid, dev, instance)
1920

    
1921
        if result.offline:
1922
          continue
1923
        elif result.fail_msg or not result.payload:
1924
          return False
1925

    
1926
    return True
1927

    
1928
  def CheckPrereq(self):
1929
    """Check prerequisites.
1930

1931
    This checks that the instance is in the cluster.
1932

1933
    """
1934
    self.instance = self.cfg.GetInstanceInfo(self.instance_uuid)
1935
    assert self.instance is not None, \
1936
      "Cannot retrieve locked instance %s" % self.instance_name
1937

    
1938
    if self.instance.disk_template != constants.DT_DRBD8:
1939
      raise errors.OpPrereqError("Can only run replace disks for DRBD8-based"
1940
                                 " instances", errors.ECODE_INVAL)
1941

    
1942
    if len(self.instance.secondary_nodes) != 1:
1943
      raise errors.OpPrereqError("The instance has a strange layout,"
1944
                                 " expected one secondary but found %d" %
1945
                                 len(self.instance.secondary_nodes),
1946
                                 errors.ECODE_FAULT)
1947

    
1948
    secondary_node_uuid = self.instance.secondary_nodes[0]
1949

    
1950
    if self.iallocator_name is None:
1951
      remote_node_uuid = self.remote_node_uuid
1952
    else:
1953
      remote_node_uuid = self._RunAllocator(self.lu, self.iallocator_name,
1954
                                            self.instance.uuid,
1955
                                            self.instance.secondary_nodes)
1956

    
1957
    if remote_node_uuid is None:
1958
      self.remote_node_info = None
1959
    else:
1960
      assert remote_node_uuid in self.lu.owned_locks(locking.LEVEL_NODE), \
1961
             "Remote node '%s' is not locked" % remote_node_uuid
1962

    
1963
      self.remote_node_info = self.cfg.GetNodeInfo(remote_node_uuid)
1964
      assert self.remote_node_info is not None, \
1965
        "Cannot retrieve locked node %s" % remote_node_uuid
1966

    
1967
    if remote_node_uuid == self.instance.primary_node:
1968
      raise errors.OpPrereqError("The specified node is the primary node of"
1969
                                 " the instance", errors.ECODE_INVAL)
1970

    
1971
    if remote_node_uuid == secondary_node_uuid:
1972
      raise errors.OpPrereqError("The specified node is already the"
1973
                                 " secondary node of the instance",
1974
                                 errors.ECODE_INVAL)
1975

    
1976
    if self.disks and self.mode in (constants.REPLACE_DISK_AUTO,
1977
                                    constants.REPLACE_DISK_CHG):
1978
      raise errors.OpPrereqError("Cannot specify disks to be replaced",
1979
                                 errors.ECODE_INVAL)
1980

    
1981
    if self.mode == constants.REPLACE_DISK_AUTO:
1982
      if not self._CheckDisksActivated(self.instance):
1983
        raise errors.OpPrereqError("Please run activate-disks on instance %s"
1984
                                   " first" % self.instance_name,
1985
                                   errors.ECODE_STATE)
1986
      faulty_primary = self._FindFaultyDisks(self.instance.primary_node)
1987
      faulty_secondary = self._FindFaultyDisks(secondary_node_uuid)
1988

    
1989
      if faulty_primary and faulty_secondary:
1990
        raise errors.OpPrereqError("Instance %s has faulty disks on more than"
1991
                                   " one node and can not be repaired"
1992
                                   " automatically" % self.instance_name,
1993
                                   errors.ECODE_STATE)
1994

    
1995
      if faulty_primary:
1996
        self.disks = faulty_primary
1997
        self.target_node_uuid = self.instance.primary_node
1998
        self.other_node_uuid = secondary_node_uuid
1999
        check_nodes = [self.target_node_uuid, self.other_node_uuid]
2000
      elif faulty_secondary:
2001
        self.disks = faulty_secondary
2002
        self.target_node_uuid = secondary_node_uuid
2003
        self.other_node_uuid = self.instance.primary_node
2004
        check_nodes = [self.target_node_uuid, self.other_node_uuid]
2005
      else:
2006
        self.disks = []
2007
        check_nodes = []
2008

    
2009
    else:
2010
      # Non-automatic modes
2011
      if self.mode == constants.REPLACE_DISK_PRI:
2012
        self.target_node_uuid = self.instance.primary_node
2013
        self.other_node_uuid = secondary_node_uuid
2014
        check_nodes = [self.target_node_uuid, self.other_node_uuid]
2015

    
2016
      elif self.mode == constants.REPLACE_DISK_SEC:
2017
        self.target_node_uuid = secondary_node_uuid
2018
        self.other_node_uuid = self.instance.primary_node
2019
        check_nodes = [self.target_node_uuid, self.other_node_uuid]
2020

    
2021
      elif self.mode == constants.REPLACE_DISK_CHG:
2022
        self.new_node_uuid = remote_node_uuid
2023
        self.other_node_uuid = self.instance.primary_node
2024
        self.target_node_uuid = secondary_node_uuid
2025
        check_nodes = [self.new_node_uuid, self.other_node_uuid]
2026

    
2027
        CheckNodeNotDrained(self.lu, remote_node_uuid)
2028
        CheckNodeVmCapable(self.lu, remote_node_uuid)
2029

    
2030
        old_node_info = self.cfg.GetNodeInfo(secondary_node_uuid)
2031
        assert old_node_info is not None
2032
        if old_node_info.offline and not self.early_release:
2033
          # doesn't make sense to delay the release
2034
          self.early_release = True
2035
          self.lu.LogInfo("Old secondary %s is offline, automatically enabling"
2036
                          " early-release mode", secondary_node_uuid)
2037

    
2038
      else:
2039
        raise errors.ProgrammerError("Unhandled disk replace mode (%s)" %
2040
                                     self.mode)
2041

    
2042
      # If not specified all disks should be replaced
2043
      if not self.disks:
2044
        self.disks = range(len(self.instance.disks))
2045

    
2046
    # TODO: This is ugly, but right now we can't distinguish between internal
2047
    # submitted opcode and external one. We should fix that.
2048
    if self.remote_node_info:
2049
      # We change the node, lets verify it still meets instance policy
2050
      new_group_info = self.cfg.GetNodeGroup(self.remote_node_info.group)
2051
      cluster = self.cfg.GetClusterInfo()
2052
      ipolicy = ganeti.masterd.instance.CalculateGroupIPolicy(cluster,
2053
                                                              new_group_info)
2054
      CheckTargetNodeIPolicy(self, ipolicy, self.instance,
2055
                             self.remote_node_info, self.cfg,
2056
                             ignore=self.ignore_ipolicy)
2057

    
2058
    for node_uuid in check_nodes:
2059
      CheckNodeOnline(self.lu, node_uuid)
2060

    
2061
    touched_nodes = frozenset(node_uuid for node_uuid in [self.new_node_uuid,
2062
                                                          self.other_node_uuid,
2063
                                                          self.target_node_uuid]
2064
                              if node_uuid is not None)
2065

    
2066
    # Release unneeded node and node resource locks
2067
    ReleaseLocks(self.lu, locking.LEVEL_NODE, keep=touched_nodes)
2068
    ReleaseLocks(self.lu, locking.LEVEL_NODE_RES, keep=touched_nodes)
2069
    ReleaseLocks(self.lu, locking.LEVEL_NODE_ALLOC)
2070

    
2071
    # Release any owned node group
2072
    ReleaseLocks(self.lu, locking.LEVEL_NODEGROUP)
2073

    
2074
    # Check whether disks are valid
2075
    for disk_idx in self.disks:
2076
      self.instance.FindDisk(disk_idx)
2077

    
2078
    # Get secondary node IP addresses
2079
    self.node_secondary_ip = dict((uuid, node.secondary_ip) for (uuid, node)
2080
                                  in self.cfg.GetMultiNodeInfo(touched_nodes))
2081

    
2082
  def Exec(self, feedback_fn):
2083
    """Execute disk replacement.
2084

2085
    This dispatches the disk replacement to the appropriate handler.
2086

2087
    """
2088
    if __debug__:
2089
      # Verify owned locks before starting operation
2090
      owned_nodes = self.lu.owned_locks(locking.LEVEL_NODE)
2091
      assert set(owned_nodes) == set(self.node_secondary_ip), \
2092
          ("Incorrect node locks, owning %s, expected %s" %
2093
           (owned_nodes, self.node_secondary_ip.keys()))
2094
      assert (self.lu.owned_locks(locking.LEVEL_NODE) ==
2095
              self.lu.owned_locks(locking.LEVEL_NODE_RES))
2096
      assert not self.lu.glm.is_owned(locking.LEVEL_NODE_ALLOC)
2097

    
2098
      owned_instances = self.lu.owned_locks(locking.LEVEL_INSTANCE)
2099
      assert list(owned_instances) == [self.instance_name], \
2100
          "Instance '%s' not locked" % self.instance_name
2101

    
2102
      assert not self.lu.glm.is_owned(locking.LEVEL_NODEGROUP), \
2103
          "Should not own any node group lock at this point"
2104

    
2105
    if not self.disks:
2106
      feedback_fn("No disks need replacement for instance '%s'" %
2107
                  self.instance.name)
2108
      return
2109

    
2110
    feedback_fn("Replacing disk(s) %s for instance '%s'" %
2111
                (utils.CommaJoin(self.disks), self.instance.name))
2112
    feedback_fn("Current primary node: %s" %
2113
                self.cfg.GetNodeName(self.instance.primary_node))
2114
    feedback_fn("Current seconary node: %s" %
2115
                utils.CommaJoin(self.cfg.GetNodeNames(
2116
                                  self.instance.secondary_nodes)))
2117

    
2118
    activate_disks = not self.instance.disks_active
2119

    
2120
    # Activate the instance disks if we're replacing them on a down instance
2121
    if activate_disks:
2122
      StartInstanceDisks(self.lu, self.instance, True)
2123

    
2124
    try:
2125
      # Should we replace the secondary node?
2126
      if self.new_node_uuid is not None:
2127
        fn = self._ExecDrbd8Secondary
2128
      else:
2129
        fn = self._ExecDrbd8DiskOnly
2130

    
2131
      result = fn(feedback_fn)
2132
    finally:
2133
      # Deactivate the instance disks if we're replacing them on a
2134
      # down instance
2135
      if activate_disks:
2136
        _SafeShutdownInstanceDisks(self.lu, self.instance)
2137

    
2138
    assert not self.lu.owned_locks(locking.LEVEL_NODE)
2139

    
2140
    if __debug__:
2141
      # Verify owned locks
2142
      owned_nodes = self.lu.owned_locks(locking.LEVEL_NODE_RES)
2143
      nodes = frozenset(self.node_secondary_ip)
2144
      assert ((self.early_release and not owned_nodes) or
2145
              (not self.early_release and not (set(owned_nodes) - nodes))), \
2146
        ("Not owning the correct locks, early_release=%s, owned=%r,"
2147
         " nodes=%r" % (self.early_release, owned_nodes, nodes))
2148

    
2149
    return result
2150

    
2151
  def _CheckVolumeGroup(self, node_uuids):
2152
    self.lu.LogInfo("Checking volume groups")
2153

    
2154
    vgname = self.cfg.GetVGName()
2155

    
2156
    # Make sure volume group exists on all involved nodes
2157
    results = self.rpc.call_vg_list(node_uuids)
2158
    if not results:
2159
      raise errors.OpExecError("Can't list volume groups on the nodes")
2160

    
2161
    for node_uuid in node_uuids:
2162
      res = results[node_uuid]
2163
      res.Raise("Error checking node %s" % self.cfg.GetNodeName(node_uuid))
2164
      if vgname not in res.payload:
2165
        raise errors.OpExecError("Volume group '%s' not found on node %s" %
2166
                                 (vgname, self.cfg.GetNodeName(node_uuid)))
2167

    
2168
  def _CheckDisksExistence(self, node_uuids):
2169
    # Check disk existence
2170
    for idx, dev in enumerate(self.instance.disks):
2171
      if idx not in self.disks:
2172
        continue
2173

    
2174
      for node_uuid in node_uuids:
2175
        self.lu.LogInfo("Checking disk/%d on %s", idx,
2176
                        self.cfg.GetNodeName(node_uuid))
2177
        self.cfg.SetDiskID(dev, node_uuid)
2178

    
2179
        result = _BlockdevFind(self, node_uuid, dev, self.instance)
2180

    
2181
        msg = result.fail_msg
2182
        if msg or not result.payload:
2183
          if not msg:
2184
            msg = "disk not found"
2185
          raise errors.OpExecError("Can't find disk/%d on node %s: %s" %
2186
                                   (idx, self.cfg.GetNodeName(node_uuid), msg))
2187

    
2188
  def _CheckDisksConsistency(self, node_uuid, on_primary, ldisk):
2189
    for idx, dev in enumerate(self.instance.disks):
2190
      if idx not in self.disks:
2191
        continue
2192

    
2193
      self.lu.LogInfo("Checking disk/%d consistency on node %s" %
2194
                      (idx, self.cfg.GetNodeName(node_uuid)))
2195

    
2196
      if not CheckDiskConsistency(self.lu, self.instance, dev, node_uuid,
2197
                                  on_primary, ldisk=ldisk):
2198
        raise errors.OpExecError("Node %s has degraded storage, unsafe to"
2199
                                 " replace disks for instance %s" %
2200
                                 (self.cfg.GetNodeName(node_uuid),
2201
                                  self.instance.name))
2202

    
2203
  def _CreateNewStorage(self, node_uuid):
2204
    """Create new storage on the primary or secondary node.
2205

2206
    This is only used for same-node replaces, not for changing the
2207
    secondary node, hence we don't want to modify the existing disk.
2208

2209
    """
2210
    iv_names = {}
2211

    
2212
    disks = AnnotateDiskParams(self.instance, self.instance.disks, self.cfg)
2213
    for idx, dev in enumerate(disks):
2214
      if idx not in self.disks:
2215
        continue
2216

    
2217
      self.lu.LogInfo("Adding storage on %s for disk/%d",
2218
                      self.cfg.GetNodeName(node_uuid), idx)
2219

    
2220
      self.cfg.SetDiskID(dev, node_uuid)
2221

    
2222
      lv_names = [".disk%d_%s" % (idx, suffix) for suffix in ["data", "meta"]]
2223
      names = _GenerateUniqueNames(self.lu, lv_names)
2224

    
2225
      (data_disk, meta_disk) = dev.children
2226
      vg_data = data_disk.logical_id[0]
2227
      lv_data = objects.Disk(dev_type=constants.LD_LV, size=dev.size,
2228
                             logical_id=(vg_data, names[0]),
2229
                             params=data_disk.params)
2230
      vg_meta = meta_disk.logical_id[0]
2231
      lv_meta = objects.Disk(dev_type=constants.LD_LV,
2232
                             size=constants.DRBD_META_SIZE,
2233
                             logical_id=(vg_meta, names[1]),
2234
                             params=meta_disk.params)
2235

    
2236
      new_lvs = [lv_data, lv_meta]
2237
      old_lvs = [child.Copy() for child in dev.children]
2238
      iv_names[dev.iv_name] = (dev, old_lvs, new_lvs)
2239
      excl_stor = IsExclusiveStorageEnabledNodeUuid(self.lu.cfg, node_uuid)
2240

    
2241
      # we pass force_create=True to force the LVM creation
2242
      for new_lv in new_lvs:
2243
        _CreateBlockDevInner(self.lu, node_uuid, self.instance, new_lv, True,
2244
                             GetInstanceInfoText(self.instance), False,
2245
                             excl_stor)
2246

    
2247
    return iv_names
2248

    
2249
  def _CheckDevices(self, node_uuid, iv_names):
2250
    for name, (dev, _, _) in iv_names.iteritems():
2251
      self.cfg.SetDiskID(dev, node_uuid)
2252

    
2253
      result = _BlockdevFind(self, node_uuid, dev, self.instance)
2254

    
2255
      msg = result.fail_msg
2256
      if msg or not result.payload:
2257
        if not msg:
2258
          msg = "disk not found"
2259
        raise errors.OpExecError("Can't find DRBD device %s: %s" %
2260
                                 (name, msg))
2261

    
2262
      if result.payload.is_degraded:
2263
        raise errors.OpExecError("DRBD device %s is degraded!" % name)
2264

    
2265
  def _RemoveOldStorage(self, node_uuid, iv_names):
2266
    for name, (_, old_lvs, _) in iv_names.iteritems():
2267
      self.lu.LogInfo("Remove logical volumes for %s", name)
2268

    
2269
      for lv in old_lvs:
2270
        self.cfg.SetDiskID(lv, node_uuid)
2271

    
2272
        msg = self.rpc.call_blockdev_remove(node_uuid, lv).fail_msg
2273
        if msg:
2274
          self.lu.LogWarning("Can't remove old LV: %s", msg,
2275
                             hint="remove unused LVs manually")
2276

    
2277
  def _ExecDrbd8DiskOnly(self, feedback_fn): # pylint: disable=W0613
2278
    """Replace a disk on the primary or secondary for DRBD 8.
2279

2280
    The algorithm for replace is quite complicated:
2281

2282
      1. for each disk to be replaced:
2283

2284
        1. create new LVs on the target node with unique names
2285
        1. detach old LVs from the drbd device
2286
        1. rename old LVs to name_replaced.<time_t>
2287
        1. rename new LVs to old LVs
2288
        1. attach the new LVs (with the old names now) to the drbd device
2289

2290
      1. wait for sync across all devices
2291

2292
      1. for each modified disk:
2293

2294
        1. remove old LVs (which have the name name_replaces.<time_t>)
2295

2296
    Failures are not very well handled.
2297

2298
    """
2299
    steps_total = 6
2300

    
2301
    # Step: check device activation
2302
    self.lu.LogStep(1, steps_total, "Check device existence")
2303
    self._CheckDisksExistence([self.other_node_uuid, self.target_node_uuid])
2304
    self._CheckVolumeGroup([self.target_node_uuid, self.other_node_uuid])
2305

    
2306
    # Step: check other node consistency
2307
    self.lu.LogStep(2, steps_total, "Check peer consistency")
2308
    self._CheckDisksConsistency(
2309
      self.other_node_uuid, self.other_node_uuid == self.instance.primary_node,
2310
      False)
2311

    
2312
    # Step: create new storage
2313
    self.lu.LogStep(3, steps_total, "Allocate new storage")
2314
    iv_names = self._CreateNewStorage(self.target_node_uuid)
2315

    
2316
    # Step: for each lv, detach+rename*2+attach
2317
    self.lu.LogStep(4, steps_total, "Changing drbd configuration")
2318
    for dev, old_lvs, new_lvs in iv_names.itervalues():
2319
      self.lu.LogInfo("Detaching %s drbd from local storage", dev.iv_name)
2320

    
2321
      result = self.rpc.call_blockdev_removechildren(self.target_node_uuid, dev,
2322
                                                     old_lvs)
2323
      result.Raise("Can't detach drbd from local storage on node"
2324
                   " %s for device %s" %
2325
                   (self.cfg.GetNodeName(self.target_node_uuid), dev.iv_name))
2326
      #dev.children = []
2327
      #cfg.Update(instance)
2328

    
2329
      # ok, we created the new LVs, so now we know we have the needed
2330
      # storage; as such, we proceed on the target node to rename
2331
      # old_lv to _old, and new_lv to old_lv; note that we rename LVs
2332
      # using the assumption that logical_id == physical_id (which in
2333
      # turn is the unique_id on that node)
2334

    
2335
      # FIXME(iustin): use a better name for the replaced LVs
2336
      temp_suffix = int(time.time())
2337
      ren_fn = lambda d, suff: (d.physical_id[0],
2338
                                d.physical_id[1] + "_replaced-%s" % suff)
2339

    
2340
      # Build the rename list based on what LVs exist on the node
2341
      rename_old_to_new = []
2342
      for to_ren in old_lvs:
2343
        result = self.rpc.call_blockdev_find(self.target_node_uuid, to_ren)
2344
        if not result.fail_msg and result.payload:
2345
          # device exists
2346
          rename_old_to_new.append((to_ren, ren_fn(to_ren, temp_suffix)))
2347

    
2348
      self.lu.LogInfo("Renaming the old LVs on the target node")
2349
      result = self.rpc.call_blockdev_rename(self.target_node_uuid,
2350
                                             rename_old_to_new)
2351
      result.Raise("Can't rename old LVs on node %s" %
2352
                   self.cfg.GetNodeName(self.target_node_uuid))
2353

    
2354
      # Now we rename the new LVs to the old LVs
2355
      self.lu.LogInfo("Renaming the new LVs on the target node")
2356
      rename_new_to_old = [(new, old.physical_id)
2357
                           for old, new in zip(old_lvs, new_lvs)]
2358
      result = self.rpc.call_blockdev_rename(self.target_node_uuid,
2359
                                             rename_new_to_old)
2360
      result.Raise("Can't rename new LVs on node %s" %
2361
                   self.cfg.GetNodeName(self.target_node_uuid))
2362

    
2363
      # Intermediate steps of in memory modifications
2364
      for old, new in zip(old_lvs, new_lvs):
2365
        new.logical_id = old.logical_id
2366
        self.cfg.SetDiskID(new, self.target_node_uuid)
2367

    
2368
      # We need to modify old_lvs so that removal later removes the
2369
      # right LVs, not the newly added ones; note that old_lvs is a
2370
      # copy here
2371
      for disk in old_lvs:
2372
        disk.logical_id = ren_fn(disk, temp_suffix)
2373
        self.cfg.SetDiskID(disk, self.target_node_uuid)
2374

    
2375
      # Now that the new lvs have the old name, we can add them to the device
2376
      self.lu.LogInfo("Adding new mirror component on %s",
2377
                      self.cfg.GetNodeName(self.target_node_uuid))
2378
      result = self.rpc.call_blockdev_addchildren(self.target_node_uuid,
2379
                                                  (dev, self.instance), new_lvs)
2380
      msg = result.fail_msg
2381
      if msg:
2382
        for new_lv in new_lvs:
2383
          msg2 = self.rpc.call_blockdev_remove(self.target_node_uuid,
2384
                                               new_lv).fail_msg
2385
          if msg2:
2386
            self.lu.LogWarning("Can't rollback device %s: %s", dev, msg2,
2387
                               hint=("cleanup manually the unused logical"
2388
                                     "volumes"))
2389
        raise errors.OpExecError("Can't add local storage to drbd: %s" % msg)
2390

    
2391
    cstep = itertools.count(5)
2392

    
2393
    if self.early_release:
2394
      self.lu.LogStep(cstep.next(), steps_total, "Removing old storage")
2395
      self._RemoveOldStorage(self.target_node_uuid, iv_names)
2396
      # TODO: Check if releasing locks early still makes sense
2397
      ReleaseLocks(self.lu, locking.LEVEL_NODE_RES)
2398
    else:
2399
      # Release all resource locks except those used by the instance
2400
      ReleaseLocks(self.lu, locking.LEVEL_NODE_RES,
2401
                   keep=self.node_secondary_ip.keys())
2402

    
2403
    # Release all node locks while waiting for sync
2404
    ReleaseLocks(self.lu, locking.LEVEL_NODE)
2405

    
2406
    # TODO: Can the instance lock be downgraded here? Take the optional disk
2407
    # shutdown in the caller into consideration.
2408

    
2409
    # Wait for sync
2410
    # This can fail as the old devices are degraded and _WaitForSync
2411
    # does a combined result over all disks, so we don't check its return value
2412
    self.lu.LogStep(cstep.next(), steps_total, "Sync devices")
2413
    WaitForSync(self.lu, self.instance)
2414

    
2415
    # Check all devices manually
2416
    self._CheckDevices(self.instance.primary_node, iv_names)
2417

    
2418
    # Step: remove old storage
2419
    if not self.early_release:
2420
      self.lu.LogStep(cstep.next(), steps_total, "Removing old storage")
2421
      self._RemoveOldStorage(self.target_node_uuid, iv_names)
2422

    
2423
  def _ExecDrbd8Secondary(self, feedback_fn):
2424
    """Replace the secondary node for DRBD 8.
2425

2426
    The algorithm for replace is quite complicated:
2427
      - for all disks of the instance:
2428
        - create new LVs on the new node with same names
2429
        - shutdown the drbd device on the old secondary
2430
        - disconnect the drbd network on the primary
2431
        - create the drbd device on the new secondary
2432
        - network attach the drbd on the primary, using an artifice:
2433
          the drbd code for Attach() will connect to the network if it
2434
          finds a device which is connected to the good local disks but
2435
          not network enabled
2436
      - wait for sync across all devices
2437
      - remove all disks from the old secondary
2438

2439
    Failures are not very well handled.
2440

2441
    """
2442
    steps_total = 6
2443

    
2444
    pnode = self.instance.primary_node
2445

    
2446
    # Step: check device activation
2447
    self.lu.LogStep(1, steps_total, "Check device existence")
2448
    self._CheckDisksExistence([self.instance.primary_node])
2449
    self._CheckVolumeGroup([self.instance.primary_node])
2450

    
2451
    # Step: check other node consistency
2452
    self.lu.LogStep(2, steps_total, "Check peer consistency")
2453
    self._CheckDisksConsistency(self.instance.primary_node, True, True)
2454

    
2455
    # Step: create new storage
2456
    self.lu.LogStep(3, steps_total, "Allocate new storage")
2457
    disks = AnnotateDiskParams(self.instance, self.instance.disks, self.cfg)
2458
    excl_stor = IsExclusiveStorageEnabledNodeUuid(self.lu.cfg,
2459
                                                  self.new_node_uuid)
2460
    for idx, dev in enumerate(disks):
2461
      self.lu.LogInfo("Adding new local storage on %s for disk/%d" %
2462
                      (self.cfg.GetNodeName(self.new_node_uuid), idx))
2463
      # we pass force_create=True to force LVM creation
2464
      for new_lv in dev.children:
2465
        _CreateBlockDevInner(self.lu, self.new_node_uuid, self.instance, new_lv,
2466
                             True, GetInstanceInfoText(self.instance), False,
2467
                             excl_stor)
2468

    
2469
    # Step 4: dbrd minors and drbd setups changes
2470
    # after this, we must manually remove the drbd minors on both the
2471
    # error and the success paths
2472
    self.lu.LogStep(4, steps_total, "Changing drbd configuration")
2473
    minors = self.cfg.AllocateDRBDMinor([self.new_node_uuid
2474
                                         for _ in self.instance.disks],
2475
                                        self.instance.uuid)
2476
    logging.debug("Allocated minors %r", minors)
2477

    
2478
    iv_names = {}
2479
    for idx, (dev, new_minor) in enumerate(zip(self.instance.disks, minors)):
2480
      self.lu.LogInfo("activating a new drbd on %s for disk/%d" %
2481
                      (self.cfg.GetNodeName(self.new_node_uuid), idx))
2482
      # create new devices on new_node; note that we create two IDs:
2483
      # one without port, so the drbd will be activated without
2484
      # networking information on the new node at this stage, and one
2485
      # with network, for the latter activation in step 4
2486
      (o_node1, o_node2, o_port, o_minor1, o_minor2, o_secret) = dev.logical_id
2487
      if self.instance.primary_node == o_node1:
2488
        p_minor = o_minor1
2489
      else:
2490
        assert self.instance.primary_node == o_node2, "Three-node instance?"
2491
        p_minor = o_minor2
2492

    
2493
      new_alone_id = (self.instance.primary_node, self.new_node_uuid, None,
2494
                      p_minor, new_minor, o_secret)
2495
      new_net_id = (self.instance.primary_node, self.new_node_uuid, o_port,
2496
                    p_minor, new_minor, o_secret)
2497

    
2498
      iv_names[idx] = (dev, dev.children, new_net_id)
2499
      logging.debug("Allocated new_minor: %s, new_logical_id: %s", new_minor,
2500
                    new_net_id)
2501
      new_drbd = objects.Disk(dev_type=constants.LD_DRBD8,
2502
                              logical_id=new_alone_id,
2503
                              children=dev.children,
2504
                              size=dev.size,
2505
                              params={})
2506
      (anno_new_drbd,) = AnnotateDiskParams(self.instance, [new_drbd],
2507
                                            self.cfg)
2508
      try:
2509
        CreateSingleBlockDev(self.lu, self.new_node_uuid, self.instance,
2510
                             anno_new_drbd,
2511
                             GetInstanceInfoText(self.instance), False,
2512
                             excl_stor)
2513
      except errors.GenericError:
2514
        self.cfg.ReleaseDRBDMinors(self.instance.uuid)
2515
        raise
2516

    
2517
    # We have new devices, shutdown the drbd on the old secondary
2518
    for idx, dev in enumerate(self.instance.disks):
2519
      self.lu.LogInfo("Shutting down drbd for disk/%d on old node", idx)
2520
      self.cfg.SetDiskID(dev, self.target_node_uuid)
2521
      msg = self.rpc.call_blockdev_shutdown(self.target_node_uuid,
2522
                                            (dev, self.instance)).fail_msg
2523
      if msg:
2524
        self.lu.LogWarning("Failed to shutdown drbd for disk/%d on old"
2525
                           "node: %s" % (idx, msg),
2526
                           hint=("Please cleanup this device manually as"
2527
                                 " soon as possible"))
2528

    
2529
    self.lu.LogInfo("Detaching primary drbds from the network (=> standalone)")
2530
    result = self.rpc.call_drbd_disconnect_net([pnode], self.node_secondary_ip,
2531
                                               self.instance.disks)[pnode]
2532

    
2533
    msg = result.fail_msg
2534
    if msg:
2535
      # detaches didn't succeed (unlikely)
2536
      self.cfg.ReleaseDRBDMinors(self.instance.uuid)
2537
      raise errors.OpExecError("Can't detach the disks from the network on"
2538
                               " old node: %s" % (msg,))
2539

    
2540
    # if we managed to detach at least one, we update all the disks of
2541
    # the instance to point to the new secondary
2542
    self.lu.LogInfo("Updating instance configuration")
2543
    for dev, _, new_logical_id in iv_names.itervalues():
2544
      dev.logical_id = new_logical_id
2545
      self.cfg.SetDiskID(dev, self.instance.primary_node)
2546

    
2547
    self.cfg.Update(self.instance, feedback_fn)
2548

    
2549
    # Release all node locks (the configuration has been updated)
2550
    ReleaseLocks(self.lu, locking.LEVEL_NODE)
2551

    
2552
    # and now perform the drbd attach
2553
    self.lu.LogInfo("Attaching primary drbds to new secondary"
2554
                    " (standalone => connected)")
2555
    result = self.rpc.call_drbd_attach_net([self.instance.primary_node,
2556
                                            self.new_node_uuid],
2557
                                           self.node_secondary_ip,
2558
                                           (self.instance.disks, self.instance),
2559
                                           self.instance.name,
2560
                                           False)
2561
    for to_node, to_result in result.items():
2562
      msg = to_result.fail_msg
2563
      if msg:
2564
        self.lu.LogWarning("Can't attach drbd disks on node %s: %s",
2565
                           self.cfg.GetNodeName(to_node), msg,
2566
                           hint=("please do a gnt-instance info to see the"
2567
                                 " status of disks"))
2568

    
2569
    cstep = itertools.count(5)
2570

    
2571
    if self.early_release:
2572
      self.lu.LogStep(cstep.next(), steps_total, "Removing old storage")
2573
      self._RemoveOldStorage(self.target_node_uuid, iv_names)
2574
      # TODO: Check if releasing locks early still makes sense
2575
      ReleaseLocks(self.lu, locking.LEVEL_NODE_RES)
2576
    else:
2577
      # Release all resource locks except those used by the instance
2578
      ReleaseLocks(self.lu, locking.LEVEL_NODE_RES,
2579
                   keep=self.node_secondary_ip.keys())
2580

    
2581
    # TODO: Can the instance lock be downgraded here? Take the optional disk
2582
    # shutdown in the caller into consideration.
2583

    
2584
    # Wait for sync
2585
    # This can fail as the old devices are degraded and _WaitForSync
2586
    # does a combined result over all disks, so we don't check its return value
2587
    self.lu.LogStep(cstep.next(), steps_total, "Sync devices")
2588
    WaitForSync(self.lu, self.instance)
2589

    
2590
    # Check all devices manually
2591
    self._CheckDevices(self.instance.primary_node, iv_names)
2592

    
2593
    # Step: remove old storage
2594
    if not self.early_release:
2595
      self.lu.LogStep(cstep.next(), steps_total, "Removing old storage")
2596
      self._RemoveOldStorage(self.target_node_uuid, iv_names)