Statistics
| Branch: | Tag: | Revision:

root / lib / backend.py @ 9fe3c789

History | View | Annotate | Download (140.2 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
"""Functions used by the node daemon
23

24
@var _ALLOWED_UPLOAD_FILES: denotes which files are accepted in
25
     the L{UploadFile} function
26
@var _ALLOWED_CLEAN_DIRS: denotes which directories are accepted
27
     in the L{_CleanDirectory} function
28

29
"""
30

    
31
# pylint: disable=E1103,C0302
32

    
33
# E1103: %s %r has no %r member (but some types could not be
34
# inferred), because the _TryOSFromDisk returns either (True, os_obj)
35
# or (False, "string") which confuses pylint
36

    
37
# C0302: This module has become too big and should be split up
38

    
39

    
40
import os
41
import os.path
42
import shutil
43
import time
44
import stat
45
import errno
46
import re
47
import random
48
import logging
49
import tempfile
50
import zlib
51
import base64
52
import signal
53

    
54
from ganeti import errors
55
from ganeti import utils
56
from ganeti import ssh
57
from ganeti import hypervisor
58
from ganeti import constants
59
from ganeti.storage import bdev
60
from ganeti.storage import drbd
61
from ganeti.storage import filestorage
62
from ganeti import objects
63
from ganeti import ssconf
64
from ganeti import serializer
65
from ganeti import netutils
66
from ganeti import runtime
67
from ganeti import compat
68
from ganeti import pathutils
69
from ganeti import vcluster
70
from ganeti import ht
71
from ganeti.storage.base import BlockDev
72
from ganeti.storage.drbd import DRBD8
73
from ganeti import hooksmaster
74

    
75

    
76
_BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id"
77
_ALLOWED_CLEAN_DIRS = compat.UniqueFrozenset([
78
  pathutils.DATA_DIR,
79
  pathutils.JOB_QUEUE_ARCHIVE_DIR,
80
  pathutils.QUEUE_DIR,
81
  pathutils.CRYPTO_KEYS_DIR,
82
  ])
83
_MAX_SSL_CERT_VALIDITY = 7 * 24 * 60 * 60
84
_X509_KEY_FILE = "key"
85
_X509_CERT_FILE = "cert"
86
_IES_STATUS_FILE = "status"
87
_IES_PID_FILE = "pid"
88
_IES_CA_FILE = "ca"
89

    
90
#: Valid LVS output line regex
91
_LVSLINE_REGEX = re.compile(r"^ *([^|]+)\|([^|]+)\|([0-9.]+)\|([^|]{6,})\|?$")
92

    
93
# Actions for the master setup script
94
_MASTER_START = "start"
95
_MASTER_STOP = "stop"
96

    
97
#: Maximum file permissions for restricted command directory and executables
98
_RCMD_MAX_MODE = (stat.S_IRWXU |
99
                  stat.S_IRGRP | stat.S_IXGRP |
100
                  stat.S_IROTH | stat.S_IXOTH)
101

    
102
#: Delay before returning an error for restricted commands
103
_RCMD_INVALID_DELAY = 10
104

    
105
#: How long to wait to acquire lock for restricted commands (shorter than
106
#: L{_RCMD_INVALID_DELAY}) to reduce blockage of noded forks when many
107
#: command requests arrive
108
_RCMD_LOCK_TIMEOUT = _RCMD_INVALID_DELAY * 0.8
109

    
110

    
111
class RPCFail(Exception):
112
  """Class denoting RPC failure.
113

114
  Its argument is the error message.
115

116
  """
117

    
118

    
119
def _GetInstReasonFilename(instance_name):
120
  """Path of the file containing the reason of the instance status change.
121

122
  @type instance_name: string
123
  @param instance_name: The name of the instance
124
  @rtype: string
125
  @return: The path of the file
126

127
  """
128
  return utils.PathJoin(pathutils.INSTANCE_REASON_DIR, instance_name)
129

    
130

    
131
def _StoreInstReasonTrail(instance_name, trail):
132
  """Serialize a reason trail related to an instance change of state to file.
133

134
  The exact location of the file depends on the name of the instance and on
135
  the configuration of the Ganeti cluster defined at deploy time.
136

137
  @type instance_name: string
138
  @param instance_name: The name of the instance
139
  @rtype: None
140

141
  """
142
  json = serializer.DumpJson(trail)
143
  filename = _GetInstReasonFilename(instance_name)
144
  utils.WriteFile(filename, data=json)
145

    
146

    
147
def _Fail(msg, *args, **kwargs):
148
  """Log an error and the raise an RPCFail exception.
149

150
  This exception is then handled specially in the ganeti daemon and
151
  turned into a 'failed' return type. As such, this function is a
152
  useful shortcut for logging the error and returning it to the master
153
  daemon.
154

155
  @type msg: string
156
  @param msg: the text of the exception
157
  @raise RPCFail
158

159
  """
160
  if args:
161
    msg = msg % args
162
  if "log" not in kwargs or kwargs["log"]: # if we should log this error
163
    if "exc" in kwargs and kwargs["exc"]:
164
      logging.exception(msg)
165
    else:
166
      logging.error(msg)
167
  raise RPCFail(msg)
168

    
169

    
170
def _GetConfig():
171
  """Simple wrapper to return a SimpleStore.
172

173
  @rtype: L{ssconf.SimpleStore}
174
  @return: a SimpleStore instance
175

176
  """
177
  return ssconf.SimpleStore()
178

    
179

    
180
def _GetSshRunner(cluster_name):
181
  """Simple wrapper to return an SshRunner.
182

183
  @type cluster_name: str
184
  @param cluster_name: the cluster name, which is needed
185
      by the SshRunner constructor
186
  @rtype: L{ssh.SshRunner}
187
  @return: an SshRunner instance
188

189
  """
190
  return ssh.SshRunner(cluster_name)
191

    
192

    
193
def _Decompress(data):
194
  """Unpacks data compressed by the RPC client.
195

196
  @type data: list or tuple
197
  @param data: Data sent by RPC client
198
  @rtype: str
199
  @return: Decompressed data
200

201
  """
202
  assert isinstance(data, (list, tuple))
203
  assert len(data) == 2
204
  (encoding, content) = data
205
  if encoding == constants.RPC_ENCODING_NONE:
206
    return content
207
  elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
208
    return zlib.decompress(base64.b64decode(content))
209
  else:
210
    raise AssertionError("Unknown data encoding")
211

    
212

    
213
def _CleanDirectory(path, exclude=None):
214
  """Removes all regular files in a directory.
215

216
  @type path: str
217
  @param path: the directory to clean
218
  @type exclude: list
219
  @param exclude: list of files to be excluded, defaults
220
      to the empty list
221

222
  """
223
  if path not in _ALLOWED_CLEAN_DIRS:
224
    _Fail("Path passed to _CleanDirectory not in allowed clean targets: '%s'",
225
          path)
226

    
227
  if not os.path.isdir(path):
228
    return
229
  if exclude is None:
230
    exclude = []
231
  else:
232
    # Normalize excluded paths
233
    exclude = [os.path.normpath(i) for i in exclude]
234

    
235
  for rel_name in utils.ListVisibleFiles(path):
236
    full_name = utils.PathJoin(path, rel_name)
237
    if full_name in exclude:
238
      continue
239
    if os.path.isfile(full_name) and not os.path.islink(full_name):
240
      utils.RemoveFile(full_name)
241

    
242

    
243
def _BuildUploadFileList():
244
  """Build the list of allowed upload files.
245

246
  This is abstracted so that it's built only once at module import time.
247

248
  """
249
  allowed_files = set([
250
    pathutils.CLUSTER_CONF_FILE,
251
    pathutils.ETC_HOSTS,
252
    pathutils.SSH_KNOWN_HOSTS_FILE,
253
    pathutils.VNC_PASSWORD_FILE,
254
    pathutils.RAPI_CERT_FILE,
255
    pathutils.SPICE_CERT_FILE,
256
    pathutils.SPICE_CACERT_FILE,
257
    pathutils.RAPI_USERS_FILE,
258
    pathutils.CONFD_HMAC_KEY,
259
    pathutils.CLUSTER_DOMAIN_SECRET_FILE,
260
    ])
261

    
262
  for hv_name in constants.HYPER_TYPES:
263
    hv_class = hypervisor.GetHypervisorClass(hv_name)
264
    allowed_files.update(hv_class.GetAncillaryFiles()[0])
265

    
266
  assert pathutils.FILE_STORAGE_PATHS_FILE not in allowed_files, \
267
    "Allowed file storage paths should never be uploaded via RPC"
268

    
269
  return frozenset(allowed_files)
270

    
271

    
272
_ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
273

    
274

    
275
def JobQueuePurge():
276
  """Removes job queue files and archived jobs.
277

278
  @rtype: tuple
279
  @return: True, None
280

281
  """
282
  _CleanDirectory(pathutils.QUEUE_DIR, exclude=[pathutils.JOB_QUEUE_LOCK_FILE])
283
  _CleanDirectory(pathutils.JOB_QUEUE_ARCHIVE_DIR)
284

    
285

    
286
def GetMasterInfo():
287
  """Returns master information.
288

289
  This is an utility function to compute master information, either
290
  for consumption here or from the node daemon.
291

292
  @rtype: tuple
293
  @return: master_netdev, master_ip, master_name, primary_ip_family,
294
    master_netmask
295
  @raise RPCFail: in case of errors
296

297
  """
298
  try:
299
    cfg = _GetConfig()
300
    master_netdev = cfg.GetMasterNetdev()
301
    master_ip = cfg.GetMasterIP()
302
    master_netmask = cfg.GetMasterNetmask()
303
    master_node = cfg.GetMasterNode()
304
    primary_ip_family = cfg.GetPrimaryIPFamily()
305
  except errors.ConfigurationError, err:
306
    _Fail("Cluster configuration incomplete: %s", err, exc=True)
307
  return (master_netdev, master_ip, master_node, primary_ip_family,
308
          master_netmask)
309

    
310

    
311
def RunLocalHooks(hook_opcode, hooks_path, env_builder_fn):
312
  """Decorator that runs hooks before and after the decorated function.
313

314
  @type hook_opcode: string
315
  @param hook_opcode: opcode of the hook
316
  @type hooks_path: string
317
  @param hooks_path: path of the hooks
318
  @type env_builder_fn: function
319
  @param env_builder_fn: function that returns a dictionary containing the
320
    environment variables for the hooks. Will get all the parameters of the
321
    decorated function.
322
  @raise RPCFail: in case of pre-hook failure
323

324
  """
325
  def decorator(fn):
326
    def wrapper(*args, **kwargs):
327
      _, myself = ssconf.GetMasterAndMyself()
328
      nodes = ([myself], [myself])  # these hooks run locally
329

    
330
      env_fn = compat.partial(env_builder_fn, *args, **kwargs)
331

    
332
      cfg = _GetConfig()
333
      hr = HooksRunner()
334
      hm = hooksmaster.HooksMaster(hook_opcode, hooks_path, nodes,
335
                                   hr.RunLocalHooks, None, env_fn,
336
                                   logging.warning, cfg.GetClusterName(),
337
                                   cfg.GetMasterNode())
338
      hm.RunPhase(constants.HOOKS_PHASE_PRE)
339
      result = fn(*args, **kwargs)
340
      hm.RunPhase(constants.HOOKS_PHASE_POST)
341

    
342
      return result
343
    return wrapper
344
  return decorator
345

    
346

    
347
def _BuildMasterIpEnv(master_params, use_external_mip_script=None):
348
  """Builds environment variables for master IP hooks.
349

350
  @type master_params: L{objects.MasterNetworkParameters}
351
  @param master_params: network parameters of the master
352
  @type use_external_mip_script: boolean
353
  @param use_external_mip_script: whether to use an external master IP
354
    address setup script (unused, but necessary per the implementation of the
355
    _RunLocalHooks decorator)
356

357
  """
358
  # pylint: disable=W0613
359
  ver = netutils.IPAddress.GetVersionFromAddressFamily(master_params.ip_family)
360
  env = {
361
    "MASTER_NETDEV": master_params.netdev,
362
    "MASTER_IP": master_params.ip,
363
    "MASTER_NETMASK": str(master_params.netmask),
364
    "CLUSTER_IP_VERSION": str(ver),
365
  }
366

    
367
  return env
368

    
369

    
370
def _RunMasterSetupScript(master_params, action, use_external_mip_script):
371
  """Execute the master IP address setup script.
372

373
  @type master_params: L{objects.MasterNetworkParameters}
374
  @param master_params: network parameters of the master
375
  @type action: string
376
  @param action: action to pass to the script. Must be one of
377
    L{backend._MASTER_START} or L{backend._MASTER_STOP}
378
  @type use_external_mip_script: boolean
379
  @param use_external_mip_script: whether to use an external master IP
380
    address setup script
381
  @raise backend.RPCFail: if there are errors during the execution of the
382
    script
383

384
  """
385
  env = _BuildMasterIpEnv(master_params)
386

    
387
  if use_external_mip_script:
388
    setup_script = pathutils.EXTERNAL_MASTER_SETUP_SCRIPT
389
  else:
390
    setup_script = pathutils.DEFAULT_MASTER_SETUP_SCRIPT
391

    
392
  result = utils.RunCmd([setup_script, action], env=env, reset_env=True)
393

    
394
  if result.failed:
395
    _Fail("Failed to %s the master IP. Script return value: %s, output: '%s'" %
396
          (action, result.exit_code, result.output), log=True)
397

    
398

    
399
@RunLocalHooks(constants.FAKE_OP_MASTER_TURNUP, "master-ip-turnup",
400
               _BuildMasterIpEnv)
401
def ActivateMasterIp(master_params, use_external_mip_script):
402
  """Activate the IP address of the master daemon.
403

404
  @type master_params: L{objects.MasterNetworkParameters}
405
  @param master_params: network parameters of the master
406
  @type use_external_mip_script: boolean
407
  @param use_external_mip_script: whether to use an external master IP
408
    address setup script
409
  @raise RPCFail: in case of errors during the IP startup
410

411
  """
412
  _RunMasterSetupScript(master_params, _MASTER_START,
413
                        use_external_mip_script)
414

    
415

    
416
def StartMasterDaemons(no_voting):
417
  """Activate local node as master node.
418

419
  The function will start the master daemons (ganeti-masterd and ganeti-rapi).
420

421
  @type no_voting: boolean
422
  @param no_voting: whether to start ganeti-masterd without a node vote
423
      but still non-interactively
424
  @rtype: None
425

426
  """
427

    
428
  if no_voting:
429
    masterd_args = "--no-voting --yes-do-it"
430
  else:
431
    masterd_args = ""
432

    
433
  env = {
434
    "EXTRA_MASTERD_ARGS": masterd_args,
435
    }
436

    
437
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "start-master"], env=env)
438
  if result.failed:
439
    msg = "Can't start Ganeti master: %s" % result.output
440
    logging.error(msg)
441
    _Fail(msg)
442

    
443

    
444
@RunLocalHooks(constants.FAKE_OP_MASTER_TURNDOWN, "master-ip-turndown",
445
               _BuildMasterIpEnv)
446
def DeactivateMasterIp(master_params, use_external_mip_script):
447
  """Deactivate the master IP on this node.
448

449
  @type master_params: L{objects.MasterNetworkParameters}
450
  @param master_params: network parameters of the master
451
  @type use_external_mip_script: boolean
452
  @param use_external_mip_script: whether to use an external master IP
453
    address setup script
454
  @raise RPCFail: in case of errors during the IP turndown
455

456
  """
457
  _RunMasterSetupScript(master_params, _MASTER_STOP,
458
                        use_external_mip_script)
459

    
460

    
461
def StopMasterDaemons():
462
  """Stop the master daemons on this node.
463

464
  Stop the master daemons (ganeti-masterd and ganeti-rapi) on this node.
465

466
  @rtype: None
467

468
  """
469
  # TODO: log and report back to the caller the error failures; we
470
  # need to decide in which case we fail the RPC for this
471

    
472
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop-master"])
473
  if result.failed:
474
    logging.error("Could not stop Ganeti master, command %s had exitcode %s"
475
                  " and error %s",
476
                  result.cmd, result.exit_code, result.output)
477

    
478

    
479
def ChangeMasterNetmask(old_netmask, netmask, master_ip, master_netdev):
480
  """Change the netmask of the master IP.
481

482
  @param old_netmask: the old value of the netmask
483
  @param netmask: the new value of the netmask
484
  @param master_ip: the master IP
485
  @param master_netdev: the master network device
486

487
  """
488
  if old_netmask == netmask:
489
    return
490

    
491
  if not netutils.IPAddress.Own(master_ip):
492
    _Fail("The master IP address is not up, not attempting to change its"
493
          " netmask")
494

    
495
  result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "add",
496
                         "%s/%s" % (master_ip, netmask),
497
                         "dev", master_netdev, "label",
498
                         "%s:0" % master_netdev])
499
  if result.failed:
500
    _Fail("Could not set the new netmask on the master IP address")
501

    
502
  result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "del",
503
                         "%s/%s" % (master_ip, old_netmask),
504
                         "dev", master_netdev, "label",
505
                         "%s:0" % master_netdev])
506
  if result.failed:
507
    _Fail("Could not bring down the master IP address with the old netmask")
508

    
509

    
510
def EtcHostsModify(mode, host, ip):
511
  """Modify a host entry in /etc/hosts.
512

513
  @param mode: The mode to operate. Either add or remove entry
514
  @param host: The host to operate on
515
  @param ip: The ip associated with the entry
516

517
  """
518
  if mode == constants.ETC_HOSTS_ADD:
519
    if not ip:
520
      RPCFail("Mode 'add' needs 'ip' parameter, but parameter not"
521
              " present")
522
    utils.AddHostToEtcHosts(host, ip)
523
  elif mode == constants.ETC_HOSTS_REMOVE:
524
    if ip:
525
      RPCFail("Mode 'remove' does not allow 'ip' parameter, but"
526
              " parameter is present")
527
    utils.RemoveHostFromEtcHosts(host)
528
  else:
529
    RPCFail("Mode not supported")
530

    
531

    
532
def LeaveCluster(modify_ssh_setup):
533
  """Cleans up and remove the current node.
534

535
  This function cleans up and prepares the current node to be removed
536
  from the cluster.
537

538
  If processing is successful, then it raises an
539
  L{errors.QuitGanetiException} which is used as a special case to
540
  shutdown the node daemon.
541

542
  @param modify_ssh_setup: boolean
543

544
  """
545
  _CleanDirectory(pathutils.DATA_DIR)
546
  _CleanDirectory(pathutils.CRYPTO_KEYS_DIR)
547
  JobQueuePurge()
548

    
549
  if modify_ssh_setup:
550
    try:
551
      priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.SSH_LOGIN_USER)
552

    
553
      utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
554

    
555
      utils.RemoveFile(priv_key)
556
      utils.RemoveFile(pub_key)
557
    except errors.OpExecError:
558
      logging.exception("Error while processing ssh files")
559

    
560
  try:
561
    utils.RemoveFile(pathutils.CONFD_HMAC_KEY)
562
    utils.RemoveFile(pathutils.RAPI_CERT_FILE)
563
    utils.RemoveFile(pathutils.SPICE_CERT_FILE)
564
    utils.RemoveFile(pathutils.SPICE_CACERT_FILE)
565
    utils.RemoveFile(pathutils.NODED_CERT_FILE)
566
  except: # pylint: disable=W0702
567
    logging.exception("Error while removing cluster secrets")
568

    
569
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop", constants.CONFD])
570
  if result.failed:
571
    logging.error("Command %s failed with exitcode %s and error %s",
572
                  result.cmd, result.exit_code, result.output)
573

    
574
  # Raise a custom exception (handled in ganeti-noded)
575
  raise errors.QuitGanetiException(True, "Shutdown scheduled")
576

    
577

    
578
def _CheckStorageParams(params, num_params):
579
  """Performs sanity checks for storage parameters.
580

581
  @type params: list
582
  @param params: list of storage parameters
583
  @type num_params: int
584
  @param num_params: expected number of parameters
585

586
  """
587
  if params is None:
588
    raise errors.ProgrammerError("No storage parameters for storage"
589
                                 " reporting is provided.")
590
  if not isinstance(params, list):
591
    raise errors.ProgrammerError("The storage parameters are not of type"
592
                                 " list: '%s'" % params)
593
  if not len(params) == num_params:
594
    raise errors.ProgrammerError("Did not receive the expected number of"
595
                                 "storage parameters: expected %s,"
596
                                 " received '%s'" % (num_params, len(params)))
597

    
598

    
599
def _CheckLvmStorageParams(params):
600
  """Performs sanity check for the 'exclusive storage' flag.
601

602
  @see: C{_CheckStorageParams}
603

604
  """
605
  _CheckStorageParams(params, 1)
606
  excl_stor = params[0]
607
  if not isinstance(params[0], bool):
608
    raise errors.ProgrammerError("Exclusive storage parameter is not"
609
                                 " boolean: '%s'." % excl_stor)
610
  return excl_stor
611

    
612

    
613
def _GetLvmVgSpaceInfo(name, params):
614
  """Wrapper around C{_GetVgInfo} which checks the storage parameters.
615

616
  @type name: string
617
  @param name: name of the volume group
618
  @type params: list
619
  @param params: list of storage parameters, which in this case should be
620
    containing only one for exclusive storage
621

622
  """
623
  excl_stor = _CheckLvmStorageParams(params)
624
  return _GetVgInfo(name, excl_stor)
625

    
626

    
627
def _GetVgInfo(
628
    name, excl_stor, info_fn=bdev.LogicalVolume.GetVGInfo):
629
  """Retrieves information about a LVM volume group.
630

631
  """
632
  # TODO: GetVGInfo supports returning information for multiple VGs at once
633
  vginfo = info_fn([name], excl_stor)
634
  if vginfo:
635
    vg_free = int(round(vginfo[0][0], 0))
636
    vg_size = int(round(vginfo[0][1], 0))
637
  else:
638
    vg_free = None
639
    vg_size = None
640

    
641
  return {
642
    "type": constants.ST_LVM_VG,
643
    "name": name,
644
    "storage_free": vg_free,
645
    "storage_size": vg_size,
646
    }
647

    
648

    
649
def _GetLvmPvSpaceInfo(name, params):
650
  """Wrapper around C{_GetVgSpindlesInfo} with sanity checks.
651

652
  @see: C{_GetLvmVgSpaceInfo}
653

654
  """
655
  excl_stor = _CheckLvmStorageParams(params)
656
  return _GetVgSpindlesInfo(name, excl_stor)
657

    
658

    
659
def _GetVgSpindlesInfo(
660
    name, excl_stor, info_fn=bdev.LogicalVolume.GetVgSpindlesInfo):
661
  """Retrieves information about spindles in an LVM volume group.
662

663
  @type name: string
664
  @param name: VG name
665
  @type excl_stor: bool
666
  @param excl_stor: exclusive storage
667
  @rtype: dict
668
  @return: dictionary whose keys are "name", "vg_free", "vg_size" for VG name,
669
      free spindles, total spindles respectively
670

671
  """
672
  if excl_stor:
673
    (vg_free, vg_size) = info_fn(name)
674
  else:
675
    vg_free = 0
676
    vg_size = 0
677
  return {
678
    "type": constants.ST_LVM_PV,
679
    "name": name,
680
    "storage_free": vg_free,
681
    "storage_size": vg_size,
682
    }
683

    
684

    
685
def _GetHvInfo(name, hvparams, get_hv_fn=hypervisor.GetHypervisor):
686
  """Retrieves node information from a hypervisor.
687

688
  The information returned depends on the hypervisor. Common items:
689

690
    - vg_size is the size of the configured volume group in MiB
691
    - vg_free is the free size of the volume group in MiB
692
    - memory_dom0 is the memory allocated for domain0 in MiB
693
    - memory_free is the currently available (free) ram in MiB
694
    - memory_total is the total number of ram in MiB
695
    - hv_version: the hypervisor version, if available
696

697
  @type hvparams: dict of string
698
  @param hvparams: the hypervisor's hvparams
699

700
  """
701
  return get_hv_fn(name).GetNodeInfo(hvparams=hvparams)
702

    
703

    
704
def _GetHvInfoAll(hv_specs, get_hv_fn=hypervisor.GetHypervisor):
705
  """Retrieves node information for all hypervisors.
706

707
  See C{_GetHvInfo} for information on the output.
708

709
  @type hv_specs: list of pairs (string, dict of strings)
710
  @param hv_specs: list of pairs of a hypervisor's name and its hvparams
711

712
  """
713
  if hv_specs is None:
714
    return None
715

    
716
  result = []
717
  for hvname, hvparams in hv_specs:
718
    result.append(_GetHvInfo(hvname, hvparams, get_hv_fn))
719
  return result
720

    
721

    
722
def _GetNamedNodeInfo(names, fn):
723
  """Calls C{fn} for all names in C{names} and returns a dictionary.
724

725
  @rtype: None or dict
726

727
  """
728
  if names is None:
729
    return None
730
  else:
731
    return map(fn, names)
732

    
733

    
734
def GetNodeInfo(storage_units, hv_specs):
735
  """Gives back a hash with different information about the node.
736

737
  @type storage_units: list of tuples (string, string, list)
738
  @param storage_units: List of tuples (storage unit, identifier, parameters) to
739
    ask for disk space information. In case of lvm-vg, the identifier is
740
    the VG name. The parameters can contain additional, storage-type-specific
741
    parameters, for example exclusive storage for lvm storage.
742
  @type hv_specs: list of pairs (string, dict of strings)
743
  @param hv_specs: list of pairs of a hypervisor's name and its hvparams
744
  @rtype: tuple; (string, None/dict, None/dict)
745
  @return: Tuple containing boot ID, volume group information and hypervisor
746
    information
747

748
  """
749
  bootid = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
750
  storage_info = _GetNamedNodeInfo(
751
    storage_units,
752
    (lambda (storage_type, storage_key, storage_params):
753
        _ApplyStorageInfoFunction(storage_type, storage_key, storage_params)))
754
  hv_info = _GetHvInfoAll(hv_specs)
755
  return (bootid, storage_info, hv_info)
756

    
757

    
758
def _GetFileStorageSpaceInfo(path, params):
759
  """Wrapper around filestorage.GetSpaceInfo.
760

761
  The purpose of this wrapper is to call filestorage.GetFileStorageSpaceInfo
762
  and ignore the *args parameter to not leak it into the filestorage
763
  module's code.
764

765
  @see: C{filestorage.GetFileStorageSpaceInfo} for description of the
766
    parameters.
767

768
  """
769
  _CheckStorageParams(params, 0)
770
  return filestorage.GetFileStorageSpaceInfo(path)
771

    
772

    
773
# FIXME: implement storage reporting for all missing storage types.
774
_STORAGE_TYPE_INFO_FN = {
775
  constants.ST_BLOCK: None,
776
  constants.ST_DISKLESS: None,
777
  constants.ST_EXT: None,
778
  constants.ST_FILE: _GetFileStorageSpaceInfo,
779
  constants.ST_LVM_PV: _GetLvmPvSpaceInfo,
780
  constants.ST_LVM_VG: _GetLvmVgSpaceInfo,
781
  constants.ST_RADOS: None,
782
}
783

    
784

    
785
def _ApplyStorageInfoFunction(storage_type, storage_key, *args):
786
  """Looks up and applies the correct function to calculate free and total
787
  storage for the given storage type.
788

789
  @type storage_type: string
790
  @param storage_type: the storage type for which the storage shall be reported.
791
  @type storage_key: string
792
  @param storage_key: identifier of a storage unit, e.g. the volume group name
793
    of an LVM storage unit
794
  @type args: any
795
  @param args: various parameters that can be used for storage reporting. These
796
    parameters and their semantics vary from storage type to storage type and
797
    are just propagated in this function.
798
  @return: the results of the application of the storage space function (see
799
    _STORAGE_TYPE_INFO_FN) if storage space reporting is implemented for that
800
    storage type
801
  @raises NotImplementedError: for storage types who don't support space
802
    reporting yet
803
  """
804
  fn = _STORAGE_TYPE_INFO_FN[storage_type]
805
  if fn is not None:
806
    return fn(storage_key, *args)
807
  else:
808
    raise NotImplementedError
809

    
810

    
811
def _CheckExclusivePvs(pvi_list):
812
  """Check that PVs are not shared among LVs
813

814
  @type pvi_list: list of L{objects.LvmPvInfo} objects
815
  @param pvi_list: information about the PVs
816

817
  @rtype: list of tuples (string, list of strings)
818
  @return: offending volumes, as tuples: (pv_name, [lv1_name, lv2_name...])
819

820
  """
821
  res = []
822
  for pvi in pvi_list:
823
    if len(pvi.lv_list) > 1:
824
      res.append((pvi.name, pvi.lv_list))
825
  return res
826

    
827

    
828
def _VerifyHypervisors(what, vm_capable, result, all_hvparams,
829
                       get_hv_fn=hypervisor.GetHypervisor):
830
  """Verifies the hypervisor. Appends the results to the 'results' list.
831

832
  @type what: C{dict}
833
  @param what: a dictionary of things to check
834
  @type vm_capable: boolean
835
  @param vm_capable: whether or not this node is vm capable
836
  @type result: dict
837
  @param result: dictionary of verification results; results of the
838
    verifications in this function will be added here
839
  @type all_hvparams: dict of dict of string
840
  @param all_hvparams: dictionary mapping hypervisor names to hvparams
841
  @type get_hv_fn: function
842
  @param get_hv_fn: function to retrieve the hypervisor, to improve testability
843

844
  """
845
  if not vm_capable:
846
    return
847

    
848
  if constants.NV_HYPERVISOR in what:
849
    result[constants.NV_HYPERVISOR] = {}
850
    for hv_name in what[constants.NV_HYPERVISOR]:
851
      hvparams = all_hvparams[hv_name]
852
      try:
853
        val = get_hv_fn(hv_name).Verify(hvparams=hvparams)
854
      except errors.HypervisorError, err:
855
        val = "Error while checking hypervisor: %s" % str(err)
856
      result[constants.NV_HYPERVISOR][hv_name] = val
857

    
858

    
859
def _VerifyHvparams(what, vm_capable, result,
860
                    get_hv_fn=hypervisor.GetHypervisor):
861
  """Verifies the hvparams. Appends the results to the 'results' list.
862

863
  @type what: C{dict}
864
  @param what: a dictionary of things to check
865
  @type vm_capable: boolean
866
  @param vm_capable: whether or not this node is vm capable
867
  @type result: dict
868
  @param result: dictionary of verification results; results of the
869
    verifications in this function will be added here
870
  @type get_hv_fn: function
871
  @param get_hv_fn: function to retrieve the hypervisor, to improve testability
872

873
  """
874
  if not vm_capable:
875
    return
876

    
877
  if constants.NV_HVPARAMS in what:
878
    result[constants.NV_HVPARAMS] = []
879
    for source, hv_name, hvparms in what[constants.NV_HVPARAMS]:
880
      try:
881
        logging.info("Validating hv %s, %s", hv_name, hvparms)
882
        get_hv_fn(hv_name).ValidateParameters(hvparms)
883
      except errors.HypervisorError, err:
884
        result[constants.NV_HVPARAMS].append((source, hv_name, str(err)))
885

    
886

    
887
def _VerifyInstanceList(what, vm_capable, result, all_hvparams):
888
  """Verifies the instance list.
889

890
  @type what: C{dict}
891
  @param what: a dictionary of things to check
892
  @type vm_capable: boolean
893
  @param vm_capable: whether or not this node is vm capable
894
  @type result: dict
895
  @param result: dictionary of verification results; results of the
896
    verifications in this function will be added here
897
  @type all_hvparams: dict of dict of string
898
  @param all_hvparams: dictionary mapping hypervisor names to hvparams
899

900
  """
901
  if constants.NV_INSTANCELIST in what and vm_capable:
902
    # GetInstanceList can fail
903
    try:
904
      val = GetInstanceList(what[constants.NV_INSTANCELIST],
905
                            all_hvparams=all_hvparams)
906
    except RPCFail, err:
907
      val = str(err)
908
    result[constants.NV_INSTANCELIST] = val
909

    
910

    
911
def _VerifyNodeInfo(what, vm_capable, result, all_hvparams):
912
  """Verifies the node info.
913

914
  @type what: C{dict}
915
  @param what: a dictionary of things to check
916
  @type vm_capable: boolean
917
  @param vm_capable: whether or not this node is vm capable
918
  @type result: dict
919
  @param result: dictionary of verification results; results of the
920
    verifications in this function will be added here
921
  @type all_hvparams: dict of dict of string
922
  @param all_hvparams: dictionary mapping hypervisor names to hvparams
923

924
  """
925
  if constants.NV_HVINFO in what and vm_capable:
926
    hvname = what[constants.NV_HVINFO]
927
    hyper = hypervisor.GetHypervisor(hvname)
928
    hvparams = all_hvparams[hvname]
929
    result[constants.NV_HVINFO] = hyper.GetNodeInfo(hvparams=hvparams)
930

    
931

    
932
def VerifyNode(what, cluster_name, all_hvparams):
933
  """Verify the status of the local node.
934

935
  Based on the input L{what} parameter, various checks are done on the
936
  local node.
937

938
  If the I{filelist} key is present, this list of
939
  files is checksummed and the file/checksum pairs are returned.
940

941
  If the I{nodelist} key is present, we check that we have
942
  connectivity via ssh with the target nodes (and check the hostname
943
  report).
944

945
  If the I{node-net-test} key is present, we check that we have
946
  connectivity to the given nodes via both primary IP and, if
947
  applicable, secondary IPs.
948

949
  @type what: C{dict}
950
  @param what: a dictionary of things to check:
951
      - filelist: list of files for which to compute checksums
952
      - nodelist: list of nodes we should check ssh communication with
953
      - node-net-test: list of nodes we should check node daemon port
954
        connectivity with
955
      - hypervisor: list with hypervisors to run the verify for
956
  @type cluster_name: string
957
  @param cluster_name: the cluster's name
958
  @type all_hvparams: dict of dict of strings
959
  @param all_hvparams: a dictionary mapping hypervisor names to hvparams
960
  @rtype: dict
961
  @return: a dictionary with the same keys as the input dict, and
962
      values representing the result of the checks
963

964
  """
965
  result = {}
966
  my_name = netutils.Hostname.GetSysName()
967
  port = netutils.GetDaemonPort(constants.NODED)
968
  vm_capable = my_name not in what.get(constants.NV_VMNODES, [])
969

    
970
  _VerifyHypervisors(what, vm_capable, result, all_hvparams)
971
  _VerifyHvparams(what, vm_capable, result)
972

    
973
  if constants.NV_FILELIST in what:
974
    fingerprints = utils.FingerprintFiles(map(vcluster.LocalizeVirtualPath,
975
                                              what[constants.NV_FILELIST]))
976
    result[constants.NV_FILELIST] = \
977
      dict((vcluster.MakeVirtualPath(key), value)
978
           for (key, value) in fingerprints.items())
979

    
980
  if constants.NV_NODELIST in what:
981
    (nodes, bynode) = what[constants.NV_NODELIST]
982

    
983
    # Add nodes from other groups (different for each node)
984
    try:
985
      nodes.extend(bynode[my_name])
986
    except KeyError:
987
      pass
988

    
989
    # Use a random order
990
    random.shuffle(nodes)
991

    
992
    # Try to contact all nodes
993
    val = {}
994
    for node in nodes:
995
      success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
996
      if not success:
997
        val[node] = message
998

    
999
    result[constants.NV_NODELIST] = val
1000

    
1001
  if constants.NV_NODENETTEST in what:
1002
    result[constants.NV_NODENETTEST] = tmp = {}
1003
    my_pip = my_sip = None
1004
    for name, pip, sip in what[constants.NV_NODENETTEST]:
1005
      if name == my_name:
1006
        my_pip = pip
1007
        my_sip = sip
1008
        break
1009
    if not my_pip:
1010
      tmp[my_name] = ("Can't find my own primary/secondary IP"
1011
                      " in the node list")
1012
    else:
1013
      for name, pip, sip in what[constants.NV_NODENETTEST]:
1014
        fail = []
1015
        if not netutils.TcpPing(pip, port, source=my_pip):
1016
          fail.append("primary")
1017
        if sip != pip:
1018
          if not netutils.TcpPing(sip, port, source=my_sip):
1019
            fail.append("secondary")
1020
        if fail:
1021
          tmp[name] = ("failure using the %s interface(s)" %
1022
                       " and ".join(fail))
1023

    
1024
  if constants.NV_MASTERIP in what:
1025
    # FIXME: add checks on incoming data structures (here and in the
1026
    # rest of the function)
1027
    master_name, master_ip = what[constants.NV_MASTERIP]
1028
    if master_name == my_name:
1029
      source = constants.IP4_ADDRESS_LOCALHOST
1030
    else:
1031
      source = None
1032
    result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port,
1033
                                                     source=source)
1034

    
1035
  if constants.NV_USERSCRIPTS in what:
1036
    result[constants.NV_USERSCRIPTS] = \
1037
      [script for script in what[constants.NV_USERSCRIPTS]
1038
       if not utils.IsExecutable(script)]
1039

    
1040
  if constants.NV_OOB_PATHS in what:
1041
    result[constants.NV_OOB_PATHS] = tmp = []
1042
    for path in what[constants.NV_OOB_PATHS]:
1043
      try:
1044
        st = os.stat(path)
1045
      except OSError, err:
1046
        tmp.append("error stating out of band helper: %s" % err)
1047
      else:
1048
        if stat.S_ISREG(st.st_mode):
1049
          if stat.S_IMODE(st.st_mode) & stat.S_IXUSR:
1050
            tmp.append(None)
1051
          else:
1052
            tmp.append("out of band helper %s is not executable" % path)
1053
        else:
1054
          tmp.append("out of band helper %s is not a file" % path)
1055

    
1056
  if constants.NV_LVLIST in what and vm_capable:
1057
    try:
1058
      val = GetVolumeList(utils.ListVolumeGroups().keys())
1059
    except RPCFail, err:
1060
      val = str(err)
1061
    result[constants.NV_LVLIST] = val
1062

    
1063
  _VerifyInstanceList(what, vm_capable, result, all_hvparams)
1064

    
1065
  if constants.NV_VGLIST in what and vm_capable:
1066
    result[constants.NV_VGLIST] = utils.ListVolumeGroups()
1067

    
1068
  if constants.NV_PVLIST in what and vm_capable:
1069
    check_exclusive_pvs = constants.NV_EXCLUSIVEPVS in what
1070
    val = bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
1071
                                       filter_allocatable=False,
1072
                                       include_lvs=check_exclusive_pvs)
1073
    if check_exclusive_pvs:
1074
      result[constants.NV_EXCLUSIVEPVS] = _CheckExclusivePvs(val)
1075
      for pvi in val:
1076
        # Avoid sending useless data on the wire
1077
        pvi.lv_list = []
1078
    result[constants.NV_PVLIST] = map(objects.LvmPvInfo.ToDict, val)
1079

    
1080
  if constants.NV_VERSION in what:
1081
    result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
1082
                                    constants.RELEASE_VERSION)
1083

    
1084
  _VerifyNodeInfo(what, vm_capable, result, all_hvparams)
1085

    
1086
  if constants.NV_DRBDVERSION in what and vm_capable:
1087
    try:
1088
      drbd_version = DRBD8.GetProcInfo().GetVersionString()
1089
    except errors.BlockDeviceError, err:
1090
      logging.warning("Can't get DRBD version", exc_info=True)
1091
      drbd_version = str(err)
1092
    result[constants.NV_DRBDVERSION] = drbd_version
1093

    
1094
  if constants.NV_DRBDLIST in what and vm_capable:
1095
    try:
1096
      used_minors = drbd.DRBD8.GetUsedDevs()
1097
    except errors.BlockDeviceError, err:
1098
      logging.warning("Can't get used minors list", exc_info=True)
1099
      used_minors = str(err)
1100
    result[constants.NV_DRBDLIST] = used_minors
1101

    
1102
  if constants.NV_DRBDHELPER in what and vm_capable:
1103
    status = True
1104
    try:
1105
      payload = drbd.DRBD8.GetUsermodeHelper()
1106
    except errors.BlockDeviceError, err:
1107
      logging.error("Can't get DRBD usermode helper: %s", str(err))
1108
      status = False
1109
      payload = str(err)
1110
    result[constants.NV_DRBDHELPER] = (status, payload)
1111

    
1112
  if constants.NV_NODESETUP in what:
1113
    result[constants.NV_NODESETUP] = tmpr = []
1114
    if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
1115
      tmpr.append("The sysfs filesytem doesn't seem to be mounted"
1116
                  " under /sys, missing required directories /sys/block"
1117
                  " and /sys/class/net")
1118
    if (not os.path.isdir("/proc/sys") or
1119
        not os.path.isfile("/proc/sysrq-trigger")):
1120
      tmpr.append("The procfs filesystem doesn't seem to be mounted"
1121
                  " under /proc, missing required directory /proc/sys and"
1122
                  " the file /proc/sysrq-trigger")
1123

    
1124
  if constants.NV_TIME in what:
1125
    result[constants.NV_TIME] = utils.SplitTime(time.time())
1126

    
1127
  if constants.NV_OSLIST in what and vm_capable:
1128
    result[constants.NV_OSLIST] = DiagnoseOS()
1129

    
1130
  if constants.NV_BRIDGES in what and vm_capable:
1131
    result[constants.NV_BRIDGES] = [bridge
1132
                                    for bridge in what[constants.NV_BRIDGES]
1133
                                    if not utils.BridgeExists(bridge)]
1134

    
1135
  if what.get(constants.NV_ACCEPTED_STORAGE_PATHS) == my_name:
1136
    result[constants.NV_ACCEPTED_STORAGE_PATHS] = \
1137
        filestorage.ComputeWrongFileStoragePaths()
1138

    
1139
  if what.get(constants.NV_FILE_STORAGE_PATH):
1140
    pathresult = filestorage.CheckFileStoragePath(
1141
        what[constants.NV_FILE_STORAGE_PATH])
1142
    if pathresult:
1143
      result[constants.NV_FILE_STORAGE_PATH] = pathresult
1144

    
1145
  if what.get(constants.NV_SHARED_FILE_STORAGE_PATH):
1146
    pathresult = filestorage.CheckFileStoragePath(
1147
        what[constants.NV_SHARED_FILE_STORAGE_PATH])
1148
    if pathresult:
1149
      result[constants.NV_SHARED_FILE_STORAGE_PATH] = pathresult
1150

    
1151
  return result
1152

    
1153

    
1154
def GetBlockDevSizes(devices):
1155
  """Return the size of the given block devices
1156

1157
  @type devices: list
1158
  @param devices: list of block device nodes to query
1159
  @rtype: dict
1160
  @return:
1161
    dictionary of all block devices under /dev (key). The value is their
1162
    size in MiB.
1163

1164
    {'/dev/disk/by-uuid/123456-12321231-312312-312': 124}
1165

1166
  """
1167
  DEV_PREFIX = "/dev/"
1168
  blockdevs = {}
1169

    
1170
  for devpath in devices:
1171
    if not utils.IsBelowDir(DEV_PREFIX, devpath):
1172
      continue
1173

    
1174
    try:
1175
      st = os.stat(devpath)
1176
    except EnvironmentError, err:
1177
      logging.warning("Error stat()'ing device %s: %s", devpath, str(err))
1178
      continue
1179

    
1180
    if stat.S_ISBLK(st.st_mode):
1181
      result = utils.RunCmd(["blockdev", "--getsize64", devpath])
1182
      if result.failed:
1183
        # We don't want to fail, just do not list this device as available
1184
        logging.warning("Cannot get size for block device %s", devpath)
1185
        continue
1186

    
1187
      size = int(result.stdout) / (1024 * 1024)
1188
      blockdevs[devpath] = size
1189
  return blockdevs
1190

    
1191

    
1192
def GetVolumeList(vg_names):
1193
  """Compute list of logical volumes and their size.
1194

1195
  @type vg_names: list
1196
  @param vg_names: the volume groups whose LVs we should list, or
1197
      empty for all volume groups
1198
  @rtype: dict
1199
  @return:
1200
      dictionary of all partions (key) with value being a tuple of
1201
      their size (in MiB), inactive and online status::
1202

1203
        {'xenvg/test1': ('20.06', True, True)}
1204

1205
      in case of errors, a string is returned with the error
1206
      details.
1207

1208
  """
1209
  lvs = {}
1210
  sep = "|"
1211
  if not vg_names:
1212
    vg_names = []
1213
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
1214
                         "--separator=%s" % sep,
1215
                         "-ovg_name,lv_name,lv_size,lv_attr"] + vg_names)
1216
  if result.failed:
1217
    _Fail("Failed to list logical volumes, lvs output: %s", result.output)
1218

    
1219
  for line in result.stdout.splitlines():
1220
    line = line.strip()
1221
    match = _LVSLINE_REGEX.match(line)
1222
    if not match:
1223
      logging.error("Invalid line returned from lvs output: '%s'", line)
1224
      continue
1225
    vg_name, name, size, attr = match.groups()
1226
    inactive = attr[4] == "-"
1227
    online = attr[5] == "o"
1228
    virtual = attr[0] == "v"
1229
    if virtual:
1230
      # we don't want to report such volumes as existing, since they
1231
      # don't really hold data
1232
      continue
1233
    lvs[vg_name + "/" + name] = (size, inactive, online)
1234

    
1235
  return lvs
1236

    
1237

    
1238
def ListVolumeGroups():
1239
  """List the volume groups and their size.
1240

1241
  @rtype: dict
1242
  @return: dictionary with keys volume name and values the
1243
      size of the volume
1244

1245
  """
1246
  return utils.ListVolumeGroups()
1247

    
1248

    
1249
def NodeVolumes():
1250
  """List all volumes on this node.
1251

1252
  @rtype: list
1253
  @return:
1254
    A list of dictionaries, each having four keys:
1255
      - name: the logical volume name,
1256
      - size: the size of the logical volume
1257
      - dev: the physical device on which the LV lives
1258
      - vg: the volume group to which it belongs
1259

1260
    In case of errors, we return an empty list and log the
1261
    error.
1262

1263
    Note that since a logical volume can live on multiple physical
1264
    volumes, the resulting list might include a logical volume
1265
    multiple times.
1266

1267
  """
1268
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
1269
                         "--separator=|",
1270
                         "--options=lv_name,lv_size,devices,vg_name"])
1271
  if result.failed:
1272
    _Fail("Failed to list logical volumes, lvs output: %s",
1273
          result.output)
1274

    
1275
  def parse_dev(dev):
1276
    return dev.split("(")[0]
1277

    
1278
  def handle_dev(dev):
1279
    return [parse_dev(x) for x in dev.split(",")]
1280

    
1281
  def map_line(line):
1282
    line = [v.strip() for v in line]
1283
    return [{"name": line[0], "size": line[1],
1284
             "dev": dev, "vg": line[3]} for dev in handle_dev(line[2])]
1285

    
1286
  all_devs = []
1287
  for line in result.stdout.splitlines():
1288
    if line.count("|") >= 3:
1289
      all_devs.extend(map_line(line.split("|")))
1290
    else:
1291
      logging.warning("Strange line in the output from lvs: '%s'", line)
1292
  return all_devs
1293

    
1294

    
1295
def BridgesExist(bridges_list):
1296
  """Check if a list of bridges exist on the current node.
1297

1298
  @rtype: boolean
1299
  @return: C{True} if all of them exist, C{False} otherwise
1300

1301
  """
1302
  missing = []
1303
  for bridge in bridges_list:
1304
    if not utils.BridgeExists(bridge):
1305
      missing.append(bridge)
1306

    
1307
  if missing:
1308
    _Fail("Missing bridges %s", utils.CommaJoin(missing))
1309

    
1310

    
1311
def GetInstanceListForHypervisor(hname, hvparams=None,
1312
                                 get_hv_fn=hypervisor.GetHypervisor):
1313
  """Provides a list of instances of the given hypervisor.
1314

1315
  @type hname: string
1316
  @param hname: name of the hypervisor
1317
  @type hvparams: dict of strings
1318
  @param hvparams: hypervisor parameters for the given hypervisor
1319
  @type get_hv_fn: function
1320
  @param get_hv_fn: function that returns a hypervisor for the given hypervisor
1321
    name; optional parameter to increase testability
1322

1323
  @rtype: list
1324
  @return: a list of all running instances on the current node
1325
    - instance1.example.com
1326
    - instance2.example.com
1327

1328
  """
1329
  results = []
1330
  try:
1331
    hv = get_hv_fn(hname)
1332
    names = hv.ListInstances(hvparams=hvparams)
1333
    results.extend(names)
1334
  except errors.HypervisorError, err:
1335
    _Fail("Error enumerating instances (hypervisor %s): %s",
1336
          hname, err, exc=True)
1337
  return results
1338

    
1339

    
1340
def GetInstanceList(hypervisor_list, all_hvparams=None,
1341
                    get_hv_fn=hypervisor.GetHypervisor):
1342
  """Provides a list of instances.
1343

1344
  @type hypervisor_list: list
1345
  @param hypervisor_list: the list of hypervisors to query information
1346
  @type all_hvparams: dict of dict of strings
1347
  @param all_hvparams: a dictionary mapping hypervisor types to respective
1348
    cluster-wide hypervisor parameters
1349
  @type get_hv_fn: function
1350
  @param get_hv_fn: function that returns a hypervisor for the given hypervisor
1351
    name; optional parameter to increase testability
1352

1353
  @rtype: list
1354
  @return: a list of all running instances on the current node
1355
    - instance1.example.com
1356
    - instance2.example.com
1357

1358
  """
1359
  results = []
1360
  for hname in hypervisor_list:
1361
    hvparams = all_hvparams[hname]
1362
    results.extend(GetInstanceListForHypervisor(hname, hvparams=hvparams,
1363
                                                get_hv_fn=get_hv_fn))
1364
  return results
1365

    
1366

    
1367
def GetInstanceInfo(instance, hname, hvparams=None):
1368
  """Gives back the information about an instance as a dictionary.
1369

1370
  @type instance: string
1371
  @param instance: the instance name
1372
  @type hname: string
1373
  @param hname: the hypervisor type of the instance
1374
  @type hvparams: dict of strings
1375
  @param hvparams: the instance's hvparams
1376

1377
  @rtype: dict
1378
  @return: dictionary with the following keys:
1379
      - memory: memory size of instance (int)
1380
      - state: xen state of instance (string)
1381
      - time: cpu time of instance (float)
1382
      - vcpus: the number of vcpus (int)
1383

1384
  """
1385
  output = {}
1386

    
1387
  iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance,
1388
                                                          hvparams=hvparams)
1389
  if iinfo is not None:
1390
    output["memory"] = iinfo[2]
1391
    output["vcpus"] = iinfo[3]
1392
    output["state"] = iinfo[4]
1393
    output["time"] = iinfo[5]
1394

    
1395
  return output
1396

    
1397

    
1398
def GetInstanceMigratable(instance):
1399
  """Computes whether an instance can be migrated.
1400

1401
  @type instance: L{objects.Instance}
1402
  @param instance: object representing the instance to be checked.
1403

1404
  @rtype: tuple
1405
  @return: tuple of (result, description) where:
1406
      - result: whether the instance can be migrated or not
1407
      - description: a description of the issue, if relevant
1408

1409
  """
1410
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1411
  iname = instance.name
1412
  if iname not in hyper.ListInstances(instance.hvparams):
1413
    _Fail("Instance %s is not running", iname)
1414

    
1415
  for idx in range(len(instance.disks)):
1416
    link_name = _GetBlockDevSymlinkPath(iname, idx)
1417
    if not os.path.islink(link_name):
1418
      logging.warning("Instance %s is missing symlink %s for disk %d",
1419
                      iname, link_name, idx)
1420

    
1421

    
1422
def GetAllInstancesInfo(hypervisor_list, all_hvparams):
1423
  """Gather data about all instances.
1424

1425
  This is the equivalent of L{GetInstanceInfo}, except that it
1426
  computes data for all instances at once, thus being faster if one
1427
  needs data about more than one instance.
1428

1429
  @type hypervisor_list: list
1430
  @param hypervisor_list: list of hypervisors to query for instance data
1431
  @type all_hvparams: dict of dict of strings
1432
  @param all_hvparams: mapping of hypervisor names to hvparams
1433

1434
  @rtype: dict
1435
  @return: dictionary of instance: data, with data having the following keys:
1436
      - memory: memory size of instance (int)
1437
      - state: xen state of instance (string)
1438
      - time: cpu time of instance (float)
1439
      - vcpus: the number of vcpus
1440

1441
  """
1442
  output = {}
1443

    
1444
  for hname in hypervisor_list:
1445
    hvparams = all_hvparams[hname]
1446
    iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo(hvparams)
1447
    if iinfo:
1448
      for name, _, memory, vcpus, state, times in iinfo:
1449
        value = {
1450
          "memory": memory,
1451
          "vcpus": vcpus,
1452
          "state": state,
1453
          "time": times,
1454
          }
1455
        if name in output:
1456
          # we only check static parameters, like memory and vcpus,
1457
          # and not state and time which can change between the
1458
          # invocations of the different hypervisors
1459
          for key in "memory", "vcpus":
1460
            if value[key] != output[name][key]:
1461
              _Fail("Instance %s is running twice"
1462
                    " with different parameters", name)
1463
        output[name] = value
1464

    
1465
  return output
1466

    
1467

    
1468
def _InstanceLogName(kind, os_name, instance, component):
1469
  """Compute the OS log filename for a given instance and operation.
1470

1471
  The instance name and os name are passed in as strings since not all
1472
  operations have these as part of an instance object.
1473

1474
  @type kind: string
1475
  @param kind: the operation type (e.g. add, import, etc.)
1476
  @type os_name: string
1477
  @param os_name: the os name
1478
  @type instance: string
1479
  @param instance: the name of the instance being imported/added/etc.
1480
  @type component: string or None
1481
  @param component: the name of the component of the instance being
1482
      transferred
1483

1484
  """
1485
  # TODO: Use tempfile.mkstemp to create unique filename
1486
  if component:
1487
    assert "/" not in component
1488
    c_msg = "-%s" % component
1489
  else:
1490
    c_msg = ""
1491
  base = ("%s-%s-%s%s-%s.log" %
1492
          (kind, os_name, instance, c_msg, utils.TimestampForFilename()))
1493
  return utils.PathJoin(pathutils.LOG_OS_DIR, base)
1494

    
1495

    
1496
def InstanceOsAdd(instance, reinstall, debug):
1497
  """Add an OS to an instance.
1498

1499
  @type instance: L{objects.Instance}
1500
  @param instance: Instance whose OS is to be installed
1501
  @type reinstall: boolean
1502
  @param reinstall: whether this is an instance reinstall
1503
  @type debug: integer
1504
  @param debug: debug level, passed to the OS scripts
1505
  @rtype: None
1506

1507
  """
1508
  inst_os = OSFromDisk(instance.os)
1509

    
1510
  create_env = OSEnvironment(instance, inst_os, debug)
1511
  if reinstall:
1512
    create_env["INSTANCE_REINSTALL"] = "1"
1513

    
1514
  logfile = _InstanceLogName("add", instance.os, instance.name, None)
1515

    
1516
  result = utils.RunCmd([inst_os.create_script], env=create_env,
1517
                        cwd=inst_os.path, output=logfile, reset_env=True)
1518
  if result.failed:
1519
    logging.error("os create command '%s' returned error: %s, logfile: %s,"
1520
                  " output: %s", result.cmd, result.fail_reason, logfile,
1521
                  result.output)
1522
    lines = [utils.SafeEncode(val)
1523
             for val in utils.TailFile(logfile, lines=20)]
1524
    _Fail("OS create script failed (%s), last lines in the"
1525
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1526

    
1527

    
1528
def RunRenameInstance(instance, old_name, debug):
1529
  """Run the OS rename script for an instance.
1530

1531
  @type instance: L{objects.Instance}
1532
  @param instance: Instance whose OS is to be installed
1533
  @type old_name: string
1534
  @param old_name: previous instance name
1535
  @type debug: integer
1536
  @param debug: debug level, passed to the OS scripts
1537
  @rtype: boolean
1538
  @return: the success of the operation
1539

1540
  """
1541
  inst_os = OSFromDisk(instance.os)
1542

    
1543
  rename_env = OSEnvironment(instance, inst_os, debug)
1544
  rename_env["OLD_INSTANCE_NAME"] = old_name
1545

    
1546
  logfile = _InstanceLogName("rename", instance.os,
1547
                             "%s-%s" % (old_name, instance.name), None)
1548

    
1549
  result = utils.RunCmd([inst_os.rename_script], env=rename_env,
1550
                        cwd=inst_os.path, output=logfile, reset_env=True)
1551

    
1552
  if result.failed:
1553
    logging.error("os create command '%s' returned error: %s output: %s",
1554
                  result.cmd, result.fail_reason, result.output)
1555
    lines = [utils.SafeEncode(val)
1556
             for val in utils.TailFile(logfile, lines=20)]
1557
    _Fail("OS rename script failed (%s), last lines in the"
1558
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1559

    
1560

    
1561
def _GetBlockDevSymlinkPath(instance_name, idx, _dir=None):
1562
  """Returns symlink path for block device.
1563

1564
  """
1565
  if _dir is None:
1566
    _dir = pathutils.DISK_LINKS_DIR
1567

    
1568
  return utils.PathJoin(_dir,
1569
                        ("%s%s%s" %
1570
                         (instance_name, constants.DISK_SEPARATOR, idx)))
1571

    
1572

    
1573
def _SymlinkBlockDev(instance_name, device_path, idx):
1574
  """Set up symlinks to a instance's block device.
1575

1576
  This is an auxiliary function run when an instance is start (on the primary
1577
  node) or when an instance is migrated (on the target node).
1578

1579

1580
  @param instance_name: the name of the target instance
1581
  @param device_path: path of the physical block device, on the node
1582
  @param idx: the disk index
1583
  @return: absolute path to the disk's symlink
1584

1585
  """
1586
  link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1587
  try:
1588
    os.symlink(device_path, link_name)
1589
  except OSError, err:
1590
    if err.errno == errno.EEXIST:
1591
      if (not os.path.islink(link_name) or
1592
          os.readlink(link_name) != device_path):
1593
        os.remove(link_name)
1594
        os.symlink(device_path, link_name)
1595
    else:
1596
      raise
1597

    
1598
  return link_name
1599

    
1600

    
1601
def _RemoveBlockDevLinks(instance_name, disks):
1602
  """Remove the block device symlinks belonging to the given instance.
1603

1604
  """
1605
  for idx, _ in enumerate(disks):
1606
    link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1607
    if os.path.islink(link_name):
1608
      try:
1609
        os.remove(link_name)
1610
      except OSError:
1611
        logging.exception("Can't remove symlink '%s'", link_name)
1612

    
1613

    
1614
def _CalculateDeviceURI(instance, disk, device):
1615
  """Get the URI for the device.
1616

1617
  @type instance: L{objects.Instance}
1618
  @param instance: the instance which disk belongs to
1619
  @type disk: L{objects.Disk}
1620
  @param disk: the target disk object
1621
  @type device: L{bdev.BlockDev}
1622
  @param device: the corresponding BlockDevice
1623
  @rtype: string
1624
  @return: the device uri if any else None
1625

1626
  """
1627
  access_mode = disk.params.get(constants.LDP_ACCESS,
1628
                                constants.DISK_KERNELSPACE)
1629
  if access_mode == constants.DISK_USERSPACE:
1630
    try:
1631
      return device.GetUserspaceAccessUri(instance.hypervisor)
1632
    except errors.BlockDeviceError:
1633
      logging.exception("Access mode is userspace but cannot get URI")
1634

    
1635
  return None
1636

    
1637

    
1638
def _GatherAndLinkBlockDevs(instance):
1639
  """Set up an instance's block device(s).
1640

1641
  This is run on the primary node at instance startup. The block
1642
  devices must be already assembled.
1643

1644
  @type instance: L{objects.Instance}
1645
  @param instance: the instance whose disks we should assemble
1646
  @rtype: list
1647
  @return: list of (disk_object, link_name, drive_uri)
1648

1649
  """
1650
  block_devices = []
1651
  for idx, disk in enumerate(instance.disks):
1652
    device = _RecursiveFindBD(disk)
1653
    if device is None:
1654
      raise errors.BlockDeviceError("Block device '%s' is not set up." %
1655
                                    str(disk))
1656
    device.Open()
1657
    try:
1658
      link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1659
    except OSError, e:
1660
      raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1661
                                    e.strerror)
1662
    uri = _CalculateDeviceURI(instance, disk, device)
1663

    
1664
    block_devices.append((disk, link_name, uri))
1665

    
1666
  return block_devices
1667

    
1668

    
1669
def StartInstance(instance, startup_paused, reason, store_reason=True):
1670
  """Start an instance.
1671

1672
  @type instance: L{objects.Instance}
1673
  @param instance: the instance object
1674
  @type startup_paused: bool
1675
  @param instance: pause instance at startup?
1676
  @type reason: list of reasons
1677
  @param reason: the reason trail for this startup
1678
  @type store_reason: boolean
1679
  @param store_reason: whether to store the shutdown reason trail on file
1680
  @rtype: None
1681

1682
  """
1683
  running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1684
                                                   instance.hvparams)
1685

    
1686
  if instance.name in running_instances:
1687
    logging.info("Instance %s already running, not starting", instance.name)
1688
    return
1689

    
1690
  try:
1691
    block_devices = _GatherAndLinkBlockDevs(instance)
1692
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
1693
    hyper.StartInstance(instance, block_devices, startup_paused)
1694
    if store_reason:
1695
      _StoreInstReasonTrail(instance.name, reason)
1696
  except errors.BlockDeviceError, err:
1697
    _Fail("Block device error: %s", err, exc=True)
1698
  except errors.HypervisorError, err:
1699
    _RemoveBlockDevLinks(instance.name, instance.disks)
1700
    _Fail("Hypervisor error: %s", err, exc=True)
1701

    
1702

    
1703
def InstanceShutdown(instance, timeout, reason, store_reason=True):
1704
  """Shut an instance down.
1705

1706
  @note: this functions uses polling with a hardcoded timeout.
1707

1708
  @type instance: L{objects.Instance}
1709
  @param instance: the instance object
1710
  @type timeout: integer
1711
  @param timeout: maximum timeout for soft shutdown
1712
  @type reason: list of reasons
1713
  @param reason: the reason trail for this shutdown
1714
  @type store_reason: boolean
1715
  @param store_reason: whether to store the shutdown reason trail on file
1716
  @rtype: None
1717

1718
  """
1719
  hv_name = instance.hypervisor
1720
  hyper = hypervisor.GetHypervisor(hv_name)
1721
  iname = instance.name
1722

    
1723
  if instance.name not in hyper.ListInstances(instance.hvparams):
1724
    logging.info("Instance %s not running, doing nothing", iname)
1725
    return
1726

    
1727
  class _TryShutdown:
1728
    def __init__(self):
1729
      self.tried_once = False
1730

    
1731
    def __call__(self):
1732
      if iname not in hyper.ListInstances(instance.hvparams):
1733
        return
1734

    
1735
      try:
1736
        hyper.StopInstance(instance, retry=self.tried_once)
1737
        if store_reason:
1738
          _StoreInstReasonTrail(instance.name, reason)
1739
      except errors.HypervisorError, err:
1740
        if iname not in hyper.ListInstances(instance.hvparams):
1741
          # if the instance is no longer existing, consider this a
1742
          # success and go to cleanup
1743
          return
1744

    
1745
        _Fail("Failed to stop instance %s: %s", iname, err)
1746

    
1747
      self.tried_once = True
1748

    
1749
      raise utils.RetryAgain()
1750

    
1751
  try:
1752
    utils.Retry(_TryShutdown(), 5, timeout)
1753
  except utils.RetryTimeout:
1754
    # the shutdown did not succeed
1755
    logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1756

    
1757
    try:
1758
      hyper.StopInstance(instance, force=True)
1759
    except errors.HypervisorError, err:
1760
      if iname in hyper.ListInstances(instance.hvparams):
1761
        # only raise an error if the instance still exists, otherwise
1762
        # the error could simply be "instance ... unknown"!
1763
        _Fail("Failed to force stop instance %s: %s", iname, err)
1764

    
1765
    time.sleep(1)
1766

    
1767
    if iname in hyper.ListInstances(instance.hvparams):
1768
      _Fail("Could not shutdown instance %s even by destroy", iname)
1769

    
1770
  try:
1771
    hyper.CleanupInstance(instance.name)
1772
  except errors.HypervisorError, err:
1773
    logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1774

    
1775
  _RemoveBlockDevLinks(iname, instance.disks)
1776

    
1777

    
1778
def InstanceReboot(instance, reboot_type, shutdown_timeout, reason):
1779
  """Reboot an instance.
1780

1781
  @type instance: L{objects.Instance}
1782
  @param instance: the instance object to reboot
1783
  @type reboot_type: str
1784
  @param reboot_type: the type of reboot, one the following
1785
    constants:
1786
      - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1787
        instance OS, do not recreate the VM
1788
      - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1789
        restart the VM (at the hypervisor level)
1790
      - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1791
        not accepted here, since that mode is handled differently, in
1792
        cmdlib, and translates into full stop and start of the
1793
        instance (instead of a call_instance_reboot RPC)
1794
  @type shutdown_timeout: integer
1795
  @param shutdown_timeout: maximum timeout for soft shutdown
1796
  @type reason: list of reasons
1797
  @param reason: the reason trail for this reboot
1798
  @rtype: None
1799

1800
  """
1801
  running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1802
                                                   instance.hvparams)
1803

    
1804
  if instance.name not in running_instances:
1805
    _Fail("Cannot reboot instance %s that is not running", instance.name)
1806

    
1807
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1808
  if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1809
    try:
1810
      hyper.RebootInstance(instance)
1811
    except errors.HypervisorError, err:
1812
      _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1813
  elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1814
    try:
1815
      InstanceShutdown(instance, shutdown_timeout, reason, store_reason=False)
1816
      result = StartInstance(instance, False, reason, store_reason=False)
1817
      _StoreInstReasonTrail(instance.name, reason)
1818
      return result
1819
    except errors.HypervisorError, err:
1820
      _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1821
  else:
1822
    _Fail("Invalid reboot_type received: %s", reboot_type)
1823

    
1824

    
1825
def InstanceBalloonMemory(instance, memory):
1826
  """Resize an instance's memory.
1827

1828
  @type instance: L{objects.Instance}
1829
  @param instance: the instance object
1830
  @type memory: int
1831
  @param memory: new memory amount in MB
1832
  @rtype: None
1833

1834
  """
1835
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1836
  running = hyper.ListInstances(instance.hvparams)
1837
  if instance.name not in running:
1838
    logging.info("Instance %s is not running, cannot balloon", instance.name)
1839
    return
1840
  try:
1841
    hyper.BalloonInstanceMemory(instance, memory)
1842
  except errors.HypervisorError, err:
1843
    _Fail("Failed to balloon instance memory: %s", err, exc=True)
1844

    
1845

    
1846
def MigrationInfo(instance):
1847
  """Gather information about an instance to be migrated.
1848

1849
  @type instance: L{objects.Instance}
1850
  @param instance: the instance definition
1851

1852
  """
1853
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1854
  try:
1855
    info = hyper.MigrationInfo(instance)
1856
  except errors.HypervisorError, err:
1857
    _Fail("Failed to fetch migration information: %s", err, exc=True)
1858
  return info
1859

    
1860

    
1861
def AcceptInstance(instance, info, target):
1862
  """Prepare the node to accept an instance.
1863

1864
  @type instance: L{objects.Instance}
1865
  @param instance: the instance definition
1866
  @type info: string/data (opaque)
1867
  @param info: migration information, from the source node
1868
  @type target: string
1869
  @param target: target host (usually ip), on this node
1870

1871
  """
1872
  # TODO: why is this required only for DTS_EXT_MIRROR?
1873
  if instance.disk_template in constants.DTS_EXT_MIRROR:
1874
    # Create the symlinks, as the disks are not active
1875
    # in any way
1876
    try:
1877
      _GatherAndLinkBlockDevs(instance)
1878
    except errors.BlockDeviceError, err:
1879
      _Fail("Block device error: %s", err, exc=True)
1880

    
1881
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1882
  try:
1883
    hyper.AcceptInstance(instance, info, target)
1884
  except errors.HypervisorError, err:
1885
    if instance.disk_template in constants.DTS_EXT_MIRROR:
1886
      _RemoveBlockDevLinks(instance.name, instance.disks)
1887
    _Fail("Failed to accept instance: %s", err, exc=True)
1888

    
1889

    
1890
def FinalizeMigrationDst(instance, info, success):
1891
  """Finalize any preparation to accept an instance.
1892

1893
  @type instance: L{objects.Instance}
1894
  @param instance: the instance definition
1895
  @type info: string/data (opaque)
1896
  @param info: migration information, from the source node
1897
  @type success: boolean
1898
  @param success: whether the migration was a success or a failure
1899

1900
  """
1901
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1902
  try:
1903
    hyper.FinalizeMigrationDst(instance, info, success)
1904
  except errors.HypervisorError, err:
1905
    _Fail("Failed to finalize migration on the target node: %s", err, exc=True)
1906

    
1907

    
1908
def MigrateInstance(cluster_name, instance, target, live):
1909
  """Migrates an instance to another node.
1910

1911
  @type cluster_name: string
1912
  @param cluster_name: name of the cluster
1913
  @type instance: L{objects.Instance}
1914
  @param instance: the instance definition
1915
  @type target: string
1916
  @param target: the target node name
1917
  @type live: boolean
1918
  @param live: whether the migration should be done live or not (the
1919
      interpretation of this parameter is left to the hypervisor)
1920
  @raise RPCFail: if migration fails for some reason
1921

1922
  """
1923
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1924

    
1925
  try:
1926
    hyper.MigrateInstance(cluster_name, instance, target, live)
1927
  except errors.HypervisorError, err:
1928
    _Fail("Failed to migrate instance: %s", err, exc=True)
1929

    
1930

    
1931
def FinalizeMigrationSource(instance, success, live):
1932
  """Finalize the instance migration on the source node.
1933

1934
  @type instance: L{objects.Instance}
1935
  @param instance: the instance definition of the migrated instance
1936
  @type success: bool
1937
  @param success: whether the migration succeeded or not
1938
  @type live: bool
1939
  @param live: whether the user requested a live migration or not
1940
  @raise RPCFail: If the execution fails for some reason
1941

1942
  """
1943
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1944

    
1945
  try:
1946
    hyper.FinalizeMigrationSource(instance, success, live)
1947
  except Exception, err:  # pylint: disable=W0703
1948
    _Fail("Failed to finalize the migration on the source node: %s", err,
1949
          exc=True)
1950

    
1951

    
1952
def GetMigrationStatus(instance):
1953
  """Get the migration status
1954

1955
  @type instance: L{objects.Instance}
1956
  @param instance: the instance that is being migrated
1957
  @rtype: L{objects.MigrationStatus}
1958
  @return: the status of the current migration (one of
1959
           L{constants.HV_MIGRATION_VALID_STATUSES}), plus any additional
1960
           progress info that can be retrieved from the hypervisor
1961
  @raise RPCFail: If the migration status cannot be retrieved
1962

1963
  """
1964
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1965
  try:
1966
    return hyper.GetMigrationStatus(instance)
1967
  except Exception, err:  # pylint: disable=W0703
1968
    _Fail("Failed to get migration status: %s", err, exc=True)
1969

    
1970

    
1971
def HotplugDevice(instance, action, dev_type, device, extra, seq):
1972
  """Hotplug a device
1973

1974
  Hotplug is currently supported only for KVM Hypervisor.
1975
  @type instance: L{objects.Instance}
1976
  @param instance: the instance to which we hotplug a device
1977
  @type action: string
1978
  @param action: the hotplug action to perform
1979
  @type dev_type: string
1980
  @param dev_type: the device type to hotplug
1981
  @type device: either L{objects.NIC} or L{objects.Disk}
1982
  @param device: the device object to hotplug
1983
  @type extra: string
1984
  @param extra: extra info used by hotplug code (e.g. disk link)
1985
  @type seq: int
1986
  @param seq: the index of the device from master perspective
1987
  @raise RPCFail: in case instance does not have KVM hypervisor
1988

1989
  """
1990
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1991
  try:
1992
    hyper.HotplugSupported(instance, action, dev_type)
1993
  except (errors.HotplugError, errors.HypervisorError), err:
1994
    _Fail("Hotplug is not supported: %s", err)
1995

    
1996
  if action == constants.HOTPLUG_ACTION_ADD:
1997
    fn = hyper.HotAddDevice
1998
  elif action == constants.HOTPLUG_ACTION_REMOVE:
1999
    fn = hyper.HotDelDevice
2000
  elif action == constants.HOTPLUG_ACTION_MODIFY:
2001
    fn = hyper.HotModDevice
2002
  else:
2003
    assert action in constants.HOTPLUG_ALL_ACTIONS
2004

    
2005
  return fn(instance, dev_type, device, extra, seq)
2006

    
2007

    
2008
def BlockdevCreate(disk, size, owner, on_primary, info, excl_stor):
2009
  """Creates a block device for an instance.
2010

2011
  @type disk: L{objects.Disk}
2012
  @param disk: the object describing the disk we should create
2013
  @type size: int
2014
  @param size: the size of the physical underlying device, in MiB
2015
  @type owner: str
2016
  @param owner: the name of the instance for which disk is created,
2017
      used for device cache data
2018
  @type on_primary: boolean
2019
  @param on_primary:  indicates if it is the primary node or not
2020
  @type info: string
2021
  @param info: string that will be sent to the physical device
2022
      creation, used for example to set (LVM) tags on LVs
2023
  @type excl_stor: boolean
2024
  @param excl_stor: Whether exclusive_storage is active
2025

2026
  @return: the new unique_id of the device (this can sometime be
2027
      computed only after creation), or None. On secondary nodes,
2028
      it's not required to return anything.
2029

2030
  """
2031
  # TODO: remove the obsolete "size" argument
2032
  # pylint: disable=W0613
2033
  clist = []
2034
  if disk.children:
2035
    for child in disk.children:
2036
      try:
2037
        crdev = _RecursiveAssembleBD(child, owner, on_primary)
2038
      except errors.BlockDeviceError, err:
2039
        _Fail("Can't assemble device %s: %s", child, err)
2040
      if on_primary or disk.AssembleOnSecondary():
2041
        # we need the children open in case the device itself has to
2042
        # be assembled
2043
        try:
2044
          # pylint: disable=E1103
2045
          crdev.Open()
2046
        except errors.BlockDeviceError, err:
2047
          _Fail("Can't make child '%s' read-write: %s", child, err)
2048
      clist.append(crdev)
2049

    
2050
  try:
2051
    device = bdev.Create(disk, clist, excl_stor)
2052
  except errors.BlockDeviceError, err:
2053
    _Fail("Can't create block device: %s", err)
2054

    
2055
  if on_primary or disk.AssembleOnSecondary():
2056
    try:
2057
      device.Assemble()
2058
    except errors.BlockDeviceError, err:
2059
      _Fail("Can't assemble device after creation, unusual event: %s", err)
2060
    if on_primary or disk.OpenOnSecondary():
2061
      try:
2062
        device.Open(force=True)
2063
      except errors.BlockDeviceError, err:
2064
        _Fail("Can't make device r/w after creation, unusual event: %s", err)
2065
    DevCacheManager.UpdateCache(device.dev_path, owner,
2066
                                on_primary, disk.iv_name)
2067

    
2068
  device.SetInfo(info)
2069

    
2070
  return device.unique_id
2071

    
2072

    
2073
def _WipeDevice(path, offset, size):
2074
  """This function actually wipes the device.
2075

2076
  @param path: The path to the device to wipe
2077
  @param offset: The offset in MiB in the file
2078
  @param size: The size in MiB to write
2079

2080
  """
2081
  # Internal sizes are always in Mebibytes; if the following "dd" command
2082
  # should use a different block size the offset and size given to this
2083
  # function must be adjusted accordingly before being passed to "dd".
2084
  block_size = 1024 * 1024
2085

    
2086
  cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
2087
         "bs=%s" % block_size, "oflag=direct", "of=%s" % path,
2088
         "count=%d" % size]
2089
  result = utils.RunCmd(cmd)
2090

    
2091
  if result.failed:
2092
    _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
2093
          result.fail_reason, result.output)
2094

    
2095

    
2096
def BlockdevWipe(disk, offset, size):
2097
  """Wipes a block device.
2098

2099
  @type disk: L{objects.Disk}
2100
  @param disk: the disk object we want to wipe
2101
  @type offset: int
2102
  @param offset: The offset in MiB in the file
2103
  @type size: int
2104
  @param size: The size in MiB to write
2105

2106
  """
2107
  try:
2108
    rdev = _RecursiveFindBD(disk)
2109
  except errors.BlockDeviceError:
2110
    rdev = None
2111

    
2112
  if not rdev:
2113
    _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
2114

    
2115
  # Do cross verify some of the parameters
2116
  if offset < 0:
2117
    _Fail("Negative offset")
2118
  if size < 0:
2119
    _Fail("Negative size")
2120
  if offset > rdev.size:
2121
    _Fail("Offset is bigger than device size")
2122
  if (offset + size) > rdev.size:
2123
    _Fail("The provided offset and size to wipe is bigger than device size")
2124

    
2125
  _WipeDevice(rdev.dev_path, offset, size)
2126

    
2127

    
2128
def BlockdevPauseResumeSync(disks, pause):
2129
  """Pause or resume the sync of the block device.
2130

2131
  @type disks: list of L{objects.Disk}
2132
  @param disks: the disks object we want to pause/resume
2133
  @type pause: bool
2134
  @param pause: Wheater to pause or resume
2135

2136
  """
2137
  success = []
2138
  for disk in disks:
2139
    try:
2140
      rdev = _RecursiveFindBD(disk)
2141
    except errors.BlockDeviceError:
2142
      rdev = None
2143

    
2144
    if not rdev:
2145
      success.append((False, ("Cannot change sync for device %s:"
2146
                              " device not found" % disk.iv_name)))
2147
      continue
2148

    
2149
    result = rdev.PauseResumeSync(pause)
2150

    
2151
    if result:
2152
      success.append((result, None))
2153
    else:
2154
      if pause:
2155
        msg = "Pause"
2156
      else:
2157
        msg = "Resume"
2158
      success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
2159

    
2160
  return success
2161

    
2162

    
2163
def BlockdevRemove(disk):
2164
  """Remove a block device.
2165

2166
  @note: This is intended to be called recursively.
2167

2168
  @type disk: L{objects.Disk}
2169
  @param disk: the disk object we should remove
2170
  @rtype: boolean
2171
  @return: the success of the operation
2172

2173
  """
2174
  msgs = []
2175
  try:
2176
    rdev = _RecursiveFindBD(disk)
2177
  except errors.BlockDeviceError, err:
2178
    # probably can't attach
2179
    logging.info("Can't attach to device %s in remove", disk)
2180
    rdev = None
2181
  if rdev is not None:
2182
    r_path = rdev.dev_path
2183
    try:
2184
      rdev.Remove()
2185
    except errors.BlockDeviceError, err:
2186
      msgs.append(str(err))
2187
    if not msgs:
2188
      DevCacheManager.RemoveCache(r_path)
2189

    
2190
  if disk.children:
2191
    for child in disk.children:
2192
      try:
2193
        BlockdevRemove(child)
2194
      except RPCFail, err:
2195
        msgs.append(str(err))
2196

    
2197
  if msgs:
2198
    _Fail("; ".join(msgs))
2199

    
2200

    
2201
def _RecursiveAssembleBD(disk, owner, as_primary):
2202
  """Activate a block device for an instance.
2203

2204
  This is run on the primary and secondary nodes for an instance.
2205

2206
  @note: this function is called recursively.
2207

2208
  @type disk: L{objects.Disk}
2209
  @param disk: the disk we try to assemble
2210
  @type owner: str
2211
  @param owner: the name of the instance which owns the disk
2212
  @type as_primary: boolean
2213
  @param as_primary: if we should make the block device
2214
      read/write
2215

2216
  @return: the assembled device or None (in case no device
2217
      was assembled)
2218
  @raise errors.BlockDeviceError: in case there is an error
2219
      during the activation of the children or the device
2220
      itself
2221

2222
  """
2223
  children = []
2224
  if disk.children:
2225
    mcn = disk.ChildrenNeeded()
2226
    if mcn == -1:
2227
      mcn = 0 # max number of Nones allowed
2228
    else:
2229
      mcn = len(disk.children) - mcn # max number of Nones
2230
    for chld_disk in disk.children:
2231
      try:
2232
        cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
2233
      except errors.BlockDeviceError, err:
2234
        if children.count(None) >= mcn:
2235
          raise
2236
        cdev = None
2237
        logging.error("Error in child activation (but continuing): %s",
2238
                      str(err))
2239
      children.append(cdev)
2240

    
2241
  if as_primary or disk.AssembleOnSecondary():
2242
    r_dev = bdev.Assemble(disk, children)
2243
    result = r_dev
2244
    if as_primary or disk.OpenOnSecondary():
2245
      r_dev.Open()
2246
    DevCacheManager.UpdateCache(r_dev.dev_path, owner,
2247
                                as_primary, disk.iv_name)
2248

    
2249
  else:
2250
    result = True
2251
  return result
2252

    
2253

    
2254
def BlockdevAssemble(disk, owner, as_primary, idx):
2255
  """Activate a block device for an instance.
2256

2257
  This is a wrapper over _RecursiveAssembleBD.
2258

2259
  @rtype: str or boolean
2260
  @return: a C{/dev/...} path for primary nodes, and
2261
      C{True} for secondary nodes
2262

2263
  """
2264
  try:
2265
    result = _RecursiveAssembleBD(disk, owner, as_primary)
2266
    if isinstance(result, BlockDev):
2267
      # pylint: disable=E1103
2268
      result = result.dev_path
2269
      if as_primary:
2270
        _SymlinkBlockDev(owner, result, idx)
2271
  except errors.BlockDeviceError, err:
2272
    _Fail("Error while assembling disk: %s", err, exc=True)
2273
  except OSError, err:
2274
    _Fail("Error while symlinking disk: %s", err, exc=True)
2275

    
2276
  return result
2277

    
2278

    
2279
def BlockdevShutdown(disk):
2280
  """Shut down a block device.
2281

2282
  First, if the device is assembled (Attach() is successful), then
2283
  the device is shutdown. Then the children of the device are
2284
  shutdown.
2285

2286
  This function is called recursively. Note that we don't cache the
2287
  children or such, as oppossed to assemble, shutdown of different
2288
  devices doesn't require that the upper device was active.
2289

2290
  @type disk: L{objects.Disk}
2291
  @param disk: the description of the disk we should
2292
      shutdown
2293
  @rtype: None
2294

2295
  """
2296
  msgs = []
2297
  r_dev = _RecursiveFindBD(disk)
2298
  if r_dev is not None:
2299
    r_path = r_dev.dev_path
2300
    try:
2301
      r_dev.Shutdown()
2302
      DevCacheManager.RemoveCache(r_path)
2303
    except errors.BlockDeviceError, err:
2304
      msgs.append(str(err))
2305

    
2306
  if disk.children:
2307
    for child in disk.children:
2308
      try:
2309
        BlockdevShutdown(child)
2310
      except RPCFail, err:
2311
        msgs.append(str(err))
2312

    
2313
  if msgs:
2314
    _Fail("; ".join(msgs))
2315

    
2316

    
2317
def BlockdevAddchildren(parent_cdev, new_cdevs):
2318
  """Extend a mirrored block device.
2319

2320
  @type parent_cdev: L{objects.Disk}
2321
  @param parent_cdev: the disk to which we should add children
2322
  @type new_cdevs: list of L{objects.Disk}
2323
  @param new_cdevs: the list of children which we should add
2324
  @rtype: None
2325

2326
  """
2327
  parent_bdev = _RecursiveFindBD(parent_cdev)
2328
  if parent_bdev is None:
2329
    _Fail("Can't find parent device '%s' in add children", parent_cdev)
2330
  new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
2331
  if new_bdevs.count(None) > 0:
2332
    _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
2333
  parent_bdev.AddChildren(new_bdevs)
2334

    
2335

    
2336
def BlockdevRemovechildren(parent_cdev, new_cdevs):
2337
  """Shrink a mirrored block device.
2338

2339
  @type parent_cdev: L{objects.Disk}
2340
  @param parent_cdev: the disk from which we should remove children
2341
  @type new_cdevs: list of L{objects.Disk}
2342
  @param new_cdevs: the list of children which we should remove
2343
  @rtype: None
2344

2345
  """
2346
  parent_bdev = _RecursiveFindBD(parent_cdev)
2347
  if parent_bdev is None:
2348
    _Fail("Can't find parent device '%s' in remove children", parent_cdev)
2349
  devs = []
2350
  for disk in new_cdevs:
2351
    rpath = disk.StaticDevPath()
2352
    if rpath is None:
2353
      bd = _RecursiveFindBD(disk)
2354
      if bd is None:
2355
        _Fail("Can't find device %s while removing children", disk)
2356
      else:
2357
        devs.append(bd.dev_path)
2358
    else:
2359
      if not utils.IsNormAbsPath(rpath):
2360
        _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
2361
      devs.append(rpath)
2362
  parent_bdev.RemoveChildren(devs)
2363

    
2364

    
2365
def BlockdevGetmirrorstatus(disks):
2366
  """Get the mirroring status of a list of devices.
2367

2368
  @type disks: list of L{objects.Disk}
2369
  @param disks: the list of disks which we should query
2370
  @rtype: disk
2371
  @return: List of L{objects.BlockDevStatus}, one for each disk
2372
  @raise errors.BlockDeviceError: if any of the disks cannot be
2373
      found
2374

2375
  """
2376
  stats = []
2377
  for dsk in disks:
2378
    rbd = _RecursiveFindBD(dsk)
2379
    if rbd is None:
2380
      _Fail("Can't find device %s", dsk)
2381

    
2382
    stats.append(rbd.CombinedSyncStatus())
2383

    
2384
  return stats
2385

    
2386

    
2387
def BlockdevGetmirrorstatusMulti(disks):
2388
  """Get the mirroring status of a list of devices.
2389

2390
  @type disks: list of L{objects.Disk}
2391
  @param disks: the list of disks which we should query
2392
  @rtype: disk
2393
  @return: List of tuples, (bool, status), one for each disk; bool denotes
2394
    success/failure, status is L{objects.BlockDevStatus} on success, string
2395
    otherwise
2396

2397
  """
2398
  result = []
2399
  for disk in disks:
2400
    try:
2401
      rbd = _RecursiveFindBD(disk)
2402
      if rbd is None:
2403
        result.append((False, "Can't find device %s" % disk))
2404
        continue
2405

    
2406
      status = rbd.CombinedSyncStatus()
2407
    except errors.BlockDeviceError, err:
2408
      logging.exception("Error while getting disk status")
2409
      result.append((False, str(err)))
2410
    else:
2411
      result.append((True, status))
2412

    
2413
  assert len(disks) == len(result)
2414

    
2415
  return result
2416

    
2417

    
2418
def _RecursiveFindBD(disk):
2419
  """Check if a device is activated.
2420

2421
  If so, return information about the real device.
2422

2423
  @type disk: L{objects.Disk}
2424
  @param disk: the disk object we need to find
2425

2426
  @return: None if the device can't be found,
2427
      otherwise the device instance
2428

2429
  """
2430
  children = []
2431
  if disk.children:
2432
    for chdisk in disk.children:
2433
      children.append(_RecursiveFindBD(chdisk))
2434

    
2435
  return bdev.FindDevice(disk, children)
2436

    
2437

    
2438
def _OpenRealBD(disk):
2439
  """Opens the underlying block device of a disk.
2440

2441
  @type disk: L{objects.Disk}
2442
  @param disk: the disk object we want to open
2443

2444
  """
2445
  real_disk = _RecursiveFindBD(disk)
2446
  if real_disk is None:
2447
    _Fail("Block device '%s' is not set up", disk)
2448

    
2449
  real_disk.Open()
2450

    
2451
  return real_disk
2452

    
2453

    
2454
def BlockdevFind(disk):
2455
  """Check if a device is activated.
2456

2457
  If it is, return information about the real device.
2458

2459
  @type disk: L{objects.Disk}
2460
  @param disk: the disk to find
2461
  @rtype: None or objects.BlockDevStatus
2462
  @return: None if the disk cannot be found, otherwise a the current
2463
           information
2464

2465
  """
2466
  try:
2467
    rbd = _RecursiveFindBD(disk)
2468
  except errors.BlockDeviceError, err:
2469
    _Fail("Failed to find device: %s", err, exc=True)
2470

    
2471
  if rbd is None:
2472
    return None
2473

    
2474
  return rbd.GetSyncStatus()
2475

    
2476

    
2477
def BlockdevGetdimensions(disks):
2478
  """Computes the size of the given disks.
2479

2480
  If a disk is not found, returns None instead.
2481

2482
  @type disks: list of L{objects.Disk}
2483
  @param disks: the list of disk to compute the size for
2484
  @rtype: list
2485
  @return: list with elements None if the disk cannot be found,
2486
      otherwise the pair (size, spindles), where spindles is None if the
2487
      device doesn't support that
2488

2489
  """
2490
  result = []
2491
  for cf in disks:
2492
    try:
2493
      rbd = _RecursiveFindBD(cf)
2494
    except errors.BlockDeviceError:
2495
      result.append(None)
2496
      continue
2497
    if rbd is None:
2498
      result.append(None)
2499
    else:
2500
      result.append(rbd.GetActualDimensions())
2501
  return result
2502

    
2503

    
2504
def BlockdevExport(disk, dest_node_ip, dest_path, cluster_name):
2505
  """Export a block device to a remote node.
2506

2507
  @type disk: L{objects.Disk}
2508
  @param disk: the description of the disk to export
2509
  @type dest_node_ip: str
2510
  @param dest_node_ip: the destination node IP to export to
2511
  @type dest_path: str
2512
  @param dest_path: the destination path on the target node
2513
  @type cluster_name: str
2514
  @param cluster_name: the cluster name, needed for SSH hostalias
2515
  @rtype: None
2516

2517
  """
2518
  real_disk = _OpenRealBD(disk)
2519

    
2520
  # the block size on the read dd is 1MiB to match our units
2521
  expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
2522
                               "dd if=%s bs=1048576 count=%s",
2523
                               real_disk.dev_path, str(disk.size))
2524

    
2525
  # we set here a smaller block size as, due to ssh buffering, more
2526
  # than 64-128k will mostly ignored; we use nocreat to fail if the
2527
  # device is not already there or we pass a wrong path; we use
2528
  # notrunc to no attempt truncate on an LV device; we use oflag=dsync
2529
  # to not buffer too much memory; this means that at best, we flush
2530
  # every 64k, which will not be very fast
2531
  destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
2532
                                " oflag=dsync", dest_path)
2533

    
2534
  remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node_ip,
2535
                                                   constants.SSH_LOGIN_USER,
2536
                                                   destcmd)
2537

    
2538
  # all commands have been checked, so we're safe to combine them
2539
  command = "|".join([expcmd, utils.ShellQuoteArgs(remotecmd)])
2540

    
2541
  result = utils.RunCmd(["bash", "-c", command])
2542

    
2543
  if result.failed:
2544
    _Fail("Disk copy command '%s' returned error: %s"
2545
          " output: %s", command, result.fail_reason, result.output)
2546

    
2547

    
2548
def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
2549
  """Write a file to the filesystem.
2550

2551
  This allows the master to overwrite(!) a file. It will only perform
2552
  the operation if the file belongs to a list of configuration files.
2553

2554
  @type file_name: str
2555
  @param file_name: the target file name
2556
  @type data: str
2557
  @param data: the new contents of the file
2558
  @type mode: int
2559
  @param mode: the mode to give the file (can be None)
2560
  @type uid: string
2561
  @param uid: the owner of the file
2562
  @type gid: string
2563
  @param gid: the group of the file
2564
  @type atime: float
2565
  @param atime: the atime to set on the file (can be None)
2566
  @type mtime: float
2567
  @param mtime: the mtime to set on the file (can be None)
2568
  @rtype: None
2569

2570
  """
2571
  file_name = vcluster.LocalizeVirtualPath(file_name)
2572

    
2573
  if not os.path.isabs(file_name):
2574
    _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
2575

    
2576
  if file_name not in _ALLOWED_UPLOAD_FILES:
2577
    _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
2578
          file_name)
2579

    
2580
  raw_data = _Decompress(data)
2581

    
2582
  if not (isinstance(uid, basestring) and isinstance(gid, basestring)):
2583
    _Fail("Invalid username/groupname type")
2584

    
2585
  getents = runtime.GetEnts()
2586
  uid = getents.LookupUser(uid)
2587
  gid = getents.LookupGroup(gid)
2588

    
2589
  utils.SafeWriteFile(file_name, None,
2590
                      data=raw_data, mode=mode, uid=uid, gid=gid,
2591
                      atime=atime, mtime=mtime)
2592

    
2593

    
2594
def RunOob(oob_program, command, node, timeout):
2595
  """Executes oob_program with given command on given node.
2596

2597
  @param oob_program: The path to the executable oob_program
2598
  @param command: The command to invoke on oob_program
2599
  @param node: The node given as an argument to the program
2600
  @param timeout: Timeout after which we kill the oob program
2601

2602
  @return: stdout
2603
  @raise RPCFail: If execution fails for some reason
2604

2605
  """
2606
  result = utils.RunCmd([oob_program, command, node], timeout=timeout)
2607

    
2608
  if result.failed:
2609
    _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
2610
          result.fail_reason, result.output)
2611

    
2612
  return result.stdout
2613

    
2614

    
2615
def _OSOndiskAPIVersion(os_dir):
2616
  """Compute and return the API version of a given OS.
2617

2618
  This function will try to read the API version of the OS residing in
2619
  the 'os_dir' directory.
2620

2621
  @type os_dir: str
2622
  @param os_dir: the directory in which we should look for the OS
2623
  @rtype: tuple
2624
  @return: tuple (status, data) with status denoting the validity and
2625
      data holding either the vaid versions or an error message
2626

2627
  """
2628
  api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
2629

    
2630
  try:
2631
    st = os.stat(api_file)
2632
  except EnvironmentError, err:
2633
    return False, ("Required file '%s' not found under path %s: %s" %
2634
                   (constants.OS_API_FILE, os_dir, utils.ErrnoOrStr(err)))
2635

    
2636
  if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2637
    return False, ("File '%s' in %s is not a regular file" %
2638
                   (constants.OS_API_FILE, os_dir))
2639

    
2640
  try:
2641
    api_versions = utils.ReadFile(api_file).splitlines()
2642
  except EnvironmentError, err:
2643
    return False, ("Error while reading the API version file at %s: %s" %
2644
                   (api_file, utils.ErrnoOrStr(err)))
2645

    
2646
  try:
2647
    api_versions = [int(version.strip()) for version in api_versions]
2648
  except (TypeError, ValueError), err:
2649
    return False, ("API version(s) can't be converted to integer: %s" %
2650
                   str(err))
2651

    
2652
  return True, api_versions
2653

    
2654

    
2655
def DiagnoseOS(top_dirs=None):
2656
  """Compute the validity for all OSes.
2657

2658
  @type top_dirs: list
2659
  @param top_dirs: the list of directories in which to
2660
      search (if not given defaults to
2661
      L{pathutils.OS_SEARCH_PATH})
2662
  @rtype: list of L{objects.OS}
2663
  @return: a list of tuples (name, path, status, diagnose, variants,
2664
      parameters, api_version) for all (potential) OSes under all
2665
      search paths, where:
2666
          - name is the (potential) OS name
2667
          - path is the full path to the OS
2668
          - status True/False is the validity of the OS
2669
          - diagnose is the error message for an invalid OS, otherwise empty
2670
          - variants is a list of supported OS variants, if any
2671
          - parameters is a list of (name, help) parameters, if any
2672
          - api_version is a list of support OS API versions
2673

2674
  """
2675
  if top_dirs is None:
2676
    top_dirs = pathutils.OS_SEARCH_PATH
2677

    
2678
  result = []
2679
  for dir_name in top_dirs:
2680
    if os.path.isdir(dir_name):
2681
      try:
2682
        f_names = utils.ListVisibleFiles(dir_name)
2683
      except EnvironmentError, err:
2684
        logging.exception("Can't list the OS directory %s: %s", dir_name, err)
2685
        break
2686
      for name in f_names:
2687
        os_path = utils.PathJoin(dir_name, name)
2688
        status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
2689
        if status:
2690
          diagnose = ""
2691
          variants = os_inst.supported_variants
2692
          parameters = os_inst.supported_parameters
2693
          api_versions = os_inst.api_versions
2694
        else:
2695
          diagnose = os_inst
2696
          variants = parameters = api_versions = []
2697
        result.append((name, os_path, status, diagnose, variants,
2698
                       parameters, api_versions))
2699

    
2700
  return result
2701

    
2702

    
2703
def _TryOSFromDisk(name, base_dir=None):
2704
  """Create an OS instance from disk.
2705

2706
  This function will return an OS instance if the given name is a
2707
  valid OS name.
2708

2709
  @type base_dir: string
2710
  @keyword base_dir: Base directory containing OS installations.
2711
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2712
  @rtype: tuple
2713
  @return: success and either the OS instance if we find a valid one,
2714
      or error message
2715

2716
  """
2717
  if base_dir is None:
2718
    os_dir = utils.FindFile(name, pathutils.OS_SEARCH_PATH, os.path.isdir)
2719
  else:
2720
    os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2721

    
2722
  if os_dir is None:
2723
    return False, "Directory for OS %s not found in search path" % name
2724

    
2725
  status, api_versions = _OSOndiskAPIVersion(os_dir)
2726
  if not status:
2727
    # push the error up
2728
    return status, api_versions
2729

    
2730
  if not constants.OS_API_VERSIONS.intersection(api_versions):
2731
    return False, ("API version mismatch for path '%s': found %s, want %s." %
2732
                   (os_dir, api_versions, constants.OS_API_VERSIONS))
2733

    
2734
  # OS Files dictionary, we will populate it with the absolute path
2735
  # names; if the value is True, then it is a required file, otherwise
2736
  # an optional one
2737
  os_files = dict.fromkeys(constants.OS_SCRIPTS, True)
2738

    
2739
  if max(api_versions) >= constants.OS_API_V15:
2740
    os_files[constants.OS_VARIANTS_FILE] = False
2741

    
2742
  if max(api_versions) >= constants.OS_API_V20:
2743
    os_files[constants.OS_PARAMETERS_FILE] = True
2744
  else:
2745
    del os_files[constants.OS_SCRIPT_VERIFY]
2746

    
2747
  for (filename, required) in os_files.items():
2748
    os_files[filename] = utils.PathJoin(os_dir, filename)
2749

    
2750
    try:
2751
      st = os.stat(os_files[filename])
2752
    except EnvironmentError, err:
2753
      if err.errno == errno.ENOENT and not required:
2754
        del os_files[filename]
2755
        continue
2756
      return False, ("File '%s' under path '%s' is missing (%s)" %
2757
                     (filename, os_dir, utils.ErrnoOrStr(err)))
2758

    
2759
    if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2760
      return False, ("File '%s' under path '%s' is not a regular file" %
2761
                     (filename, os_dir))
2762

    
2763
    if filename in constants.OS_SCRIPTS:
2764
      if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2765
        return False, ("File '%s' under path '%s' is not executable" %
2766
                       (filename, os_dir))
2767

    
2768
  variants = []
2769
  if constants.OS_VARIANTS_FILE in os_files:
2770
    variants_file = os_files[constants.OS_VARIANTS_FILE]
2771
    try:
2772
      variants = \
2773
        utils.FilterEmptyLinesAndComments(utils.ReadFile(variants_file))
2774
    except EnvironmentError, err:
2775
      # we accept missing files, but not other errors
2776
      if err.errno != errno.ENOENT:
2777
        return False, ("Error while reading the OS variants file at %s: %s" %
2778
                       (variants_file, utils.ErrnoOrStr(err)))
2779

    
2780
  parameters = []
2781
  if constants.OS_PARAMETERS_FILE in os_files:
2782
    parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2783
    try:
2784
      parameters = utils.ReadFile(parameters_file).splitlines()
2785
    except EnvironmentError, err:
2786
      return False, ("Error while reading the OS parameters file at %s: %s" %
2787
                     (parameters_file, utils.ErrnoOrStr(err)))
2788
    parameters = [v.split(None, 1) for v in parameters]
2789

    
2790
  os_obj = objects.OS(name=name, path=os_dir,
2791
                      create_script=os_files[constants.OS_SCRIPT_CREATE],
2792
                      export_script=os_files[constants.OS_SCRIPT_EXPORT],
2793
                      import_script=os_files[constants.OS_SCRIPT_IMPORT],
2794
                      rename_script=os_files[constants.OS_SCRIPT_RENAME],
2795
                      verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2796
                                                 None),
2797
                      supported_variants=variants,
2798
                      supported_parameters=parameters,
2799
                      api_versions=api_versions)
2800
  return True, os_obj
2801

    
2802

    
2803
def OSFromDisk(name, base_dir=None):
2804
  """Create an OS instance from disk.
2805

2806
  This function will return an OS instance if the given name is a
2807
  valid OS name. Otherwise, it will raise an appropriate
2808
  L{RPCFail} exception, detailing why this is not a valid OS.
2809

2810
  This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2811
  an exception but returns true/false status data.
2812

2813
  @type base_dir: string
2814
  @keyword base_dir: Base directory containing OS installations.
2815
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2816
  @rtype: L{objects.OS}
2817
  @return: the OS instance if we find a valid one
2818
  @raise RPCFail: if we don't find a valid OS
2819

2820
  """
2821
  name_only = objects.OS.GetName(name)
2822
  status, payload = _TryOSFromDisk(name_only, base_dir)
2823

    
2824
  if not status:
2825
    _Fail(payload)
2826

    
2827
  return payload
2828

    
2829

    
2830
def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2831
  """Calculate the basic environment for an os script.
2832

2833
  @type os_name: str
2834
  @param os_name: full operating system name (including variant)
2835
  @type inst_os: L{objects.OS}
2836
  @param inst_os: operating system for which the environment is being built
2837
  @type os_params: dict
2838
  @param os_params: the OS parameters
2839
  @type debug: integer
2840
  @param debug: debug level (0 or 1, for OS Api 10)
2841
  @rtype: dict
2842
  @return: dict of environment variables
2843
  @raise errors.BlockDeviceError: if the block device
2844
      cannot be found
2845

2846
  """
2847
  result = {}
2848
  api_version = \
2849
    max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2850
  result["OS_API_VERSION"] = "%d" % api_version
2851
  result["OS_NAME"] = inst_os.name
2852
  result["DEBUG_LEVEL"] = "%d" % debug
2853

    
2854
  # OS variants
2855
  if api_version >= constants.OS_API_V15 and inst_os.supported_variants:
2856
    variant = objects.OS.GetVariant(os_name)
2857
    if not variant:
2858
      variant = inst_os.supported_variants[0]
2859
  else:
2860
    variant = ""
2861
  result["OS_VARIANT"] = variant
2862

    
2863
  # OS params
2864
  for pname, pvalue in os_params.items():
2865
    result["OSP_%s" % pname.upper()] = pvalue
2866

    
2867
  # Set a default path otherwise programs called by OS scripts (or
2868
  # even hooks called from OS scripts) might break, and we don't want
2869
  # to have each script require setting a PATH variable
2870
  result["PATH"] = constants.HOOKS_PATH
2871

    
2872
  return result
2873

    
2874

    
2875
def OSEnvironment(instance, inst_os, debug=0):
2876
  """Calculate the environment for an os script.
2877

2878
  @type instance: L{objects.Instance}
2879
  @param instance: target instance for the os script run
2880
  @type inst_os: L{objects.OS}
2881
  @param inst_os: operating system for which the environment is being built
2882
  @type debug: integer
2883
  @param debug: debug level (0 or 1, for OS Api 10)
2884
  @rtype: dict
2885
  @return: dict of environment variables
2886
  @raise errors.BlockDeviceError: if the block device
2887
      cannot be found
2888

2889
  """
2890
  result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2891

    
2892
  for attr in ["name", "os", "uuid", "ctime", "mtime", "primary_node"]:
2893
    result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2894

    
2895
  result["HYPERVISOR"] = instance.hypervisor
2896
  result["DISK_COUNT"] = "%d" % len(instance.disks)
2897
  result["NIC_COUNT"] = "%d" % len(instance.nics)
2898
  result["INSTANCE_SECONDARY_NODES"] = \
2899
      ("%s" % " ".join(instance.secondary_nodes))
2900

    
2901
  # Disks
2902
  for idx, disk in enumerate(instance.disks):
2903
    real_disk = _OpenRealBD(disk)
2904
    result["DISK_%d_PATH" % idx] = real_disk.dev_path
2905
    result["DISK_%d_ACCESS" % idx] = disk.mode
2906
    result["DISK_%d_UUID" % idx] = disk.uuid
2907
    if disk.name:
2908
      result["DISK_%d_NAME" % idx] = disk.name
2909
    if constants.HV_DISK_TYPE in instance.hvparams:
2910
      result["DISK_%d_FRONTEND_TYPE" % idx] = \
2911
        instance.hvparams[constants.HV_DISK_TYPE]
2912
    if disk.dev_type in constants.DTS_BLOCK:
2913
      result["DISK_%d_BACKEND_TYPE" % idx] = "block"
2914
    elif disk.dev_type in [constants.DT_FILE, constants.DT_SHARED_FILE]:
2915
      result["DISK_%d_BACKEND_TYPE" % idx] = \
2916
        "file:%s" % disk.logical_id[0]
2917

    
2918
  # NICs
2919
  for idx, nic in enumerate(instance.nics):
2920
    result["NIC_%d_MAC" % idx] = nic.mac
2921
    result["NIC_%d_UUID" % idx] = nic.uuid
2922
    if nic.name:
2923
      result["NIC_%d_NAME" % idx] = nic.name
2924
    if nic.ip:
2925
      result["NIC_%d_IP" % idx] = nic.ip
2926
    result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE]
2927
    if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2928
      result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK]
2929
    if nic.nicparams[constants.NIC_LINK]:
2930
      result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK]
2931
    if nic.netinfo:
2932
      nobj = objects.Network.FromDict(nic.netinfo)
2933
      result.update(nobj.HooksDict("NIC_%d_" % idx))
2934
    if constants.HV_NIC_TYPE in instance.hvparams:
2935
      result["NIC_%d_FRONTEND_TYPE" % idx] = \
2936
        instance.hvparams[constants.HV_NIC_TYPE]
2937

    
2938
  # HV/BE params
2939
  for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2940
    for key, value in source.items():
2941
      result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2942

    
2943
  return result
2944

    
2945

    
2946
def DiagnoseExtStorage(top_dirs=None):
2947
  """Compute the validity for all ExtStorage Providers.
2948

2949
  @type top_dirs: list
2950
  @param top_dirs: the list of directories in which to
2951
      search (if not given defaults to
2952
      L{pathutils.ES_SEARCH_PATH})
2953
  @rtype: list of L{objects.ExtStorage}
2954
  @return: a list of tuples (name, path, status, diagnose, parameters)
2955
      for all (potential) ExtStorage Providers under all
2956
      search paths, where:
2957
          - name is the (potential) ExtStorage Provider
2958
          - path is the full path to the ExtStorage Provider
2959
          - status True/False is the validity of the ExtStorage Provider
2960
          - diagnose is the error message for an invalid ExtStorage Provider,
2961
            otherwise empty
2962
          - parameters is a list of (name, help) parameters, if any
2963

2964
  """
2965
  if top_dirs is None:
2966
    top_dirs = pathutils.ES_SEARCH_PATH
2967

    
2968
  result = []
2969
  for dir_name in top_dirs:
2970
    if os.path.isdir(dir_name):
2971
      try:
2972
        f_names = utils.ListVisibleFiles(dir_name)
2973
      except EnvironmentError, err:
2974
        logging.exception("Can't list the ExtStorage directory %s: %s",
2975
                          dir_name, err)
2976
        break
2977
      for name in f_names:
2978
        es_path = utils.PathJoin(dir_name, name)
2979
        status, es_inst = bdev.ExtStorageFromDisk(name, base_dir=dir_name)
2980
        if status:
2981
          diagnose = ""
2982
          parameters = es_inst.supported_parameters
2983
        else:
2984
          diagnose = es_inst
2985
          parameters = []
2986
        result.append((name, es_path, status, diagnose, parameters))
2987

    
2988
  return result
2989

    
2990

    
2991
def BlockdevGrow(disk, amount, dryrun, backingstore, excl_stor):
2992
  """Grow a stack of block devices.
2993

2994
  This function is called recursively, with the childrens being the
2995
  first ones to resize.
2996

2997
  @type disk: L{objects.Disk}
2998
  @param disk: the disk to be grown
2999
  @type amount: integer
3000
  @param amount: the amount (in mebibytes) to grow with
3001
  @type dryrun: boolean
3002
  @param dryrun: whether to execute the operation in simulation mode
3003
      only, without actually increasing the size
3004
  @param backingstore: whether to execute the operation on backing storage
3005
      only, or on "logical" storage only; e.g. DRBD is logical storage,
3006
      whereas LVM, file, RBD are backing storage
3007
  @rtype: (status, result)
3008
  @type excl_stor: boolean
3009
  @param excl_stor: Whether exclusive_storage is active
3010
  @return: a tuple with the status of the operation (True/False), and
3011
      the errors message if status is False
3012

3013
  """
3014
  r_dev = _RecursiveFindBD(disk)
3015
  if r_dev is None:
3016
    _Fail("Cannot find block device %s", disk)
3017

    
3018
  try:
3019
    r_dev.Grow(amount, dryrun, backingstore, excl_stor)
3020
  except errors.BlockDeviceError, err:
3021
    _Fail("Failed to grow block device: %s", err, exc=True)
3022

    
3023

    
3024
def BlockdevSnapshot(disk):
3025
  """Create a snapshot copy of a block device.
3026

3027
  This function is called recursively, and the snapshot is actually created
3028
  just for the leaf lvm backend device.
3029

3030
  @type disk: L{objects.Disk}
3031
  @param disk: the disk to be snapshotted
3032
  @rtype: string
3033
  @return: snapshot disk ID as (vg, lv)
3034

3035
  """
3036
  if disk.dev_type == constants.DT_DRBD8:
3037
    if not disk.children:
3038
      _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
3039
            disk.unique_id)
3040
    return BlockdevSnapshot(disk.children[0])
3041
  elif disk.dev_type == constants.DT_PLAIN:
3042
    r_dev = _RecursiveFindBD(disk)
3043
    if r_dev is not None:
3044
      # FIXME: choose a saner value for the snapshot size
3045
      # let's stay on the safe side and ask for the full size, for now
3046
      return r_dev.Snapshot(disk.size)
3047
    else:
3048
      _Fail("Cannot find block device %s", disk)
3049
  else:
3050
    _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
3051
          disk.unique_id, disk.dev_type)
3052

    
3053

    
3054
def BlockdevSetInfo(disk, info):
3055
  """Sets 'metadata' information on block devices.
3056

3057
  This function sets 'info' metadata on block devices. Initial
3058
  information is set at device creation; this function should be used
3059
  for example after renames.
3060

3061
  @type disk: L{objects.Disk}
3062
  @param disk: the disk to be grown
3063
  @type info: string
3064
  @param info: new 'info' metadata
3065
  @rtype: (status, result)
3066
  @return: a tuple with the status of the operation (True/False), and
3067
      the errors message if status is False
3068

3069
  """
3070
  r_dev = _RecursiveFindBD(disk)
3071
  if r_dev is None:
3072
    _Fail("Cannot find block device %s", disk)
3073

    
3074
  try:
3075
    r_dev.SetInfo(info)
3076
  except errors.BlockDeviceError, err:
3077
    _Fail("Failed to set information on block device: %s", err, exc=True)
3078

    
3079

    
3080
def FinalizeExport(instance, snap_disks):
3081
  """Write out the export configuration information.
3082

3083
  @type instance: L{objects.Instance}
3084
  @param instance: the instance which we export, used for
3085
      saving configuration
3086
  @type snap_disks: list of L{objects.Disk}
3087
  @param snap_disks: list of snapshot block devices, which
3088
      will be used to get the actual name of the dump file
3089

3090
  @rtype: None
3091

3092
  """
3093
  destdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name + ".new")
3094
  finaldestdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name)
3095

    
3096
  config = objects.SerializableConfigParser()
3097

    
3098
  config.add_section(constants.INISECT_EXP)
3099
  config.set(constants.INISECT_EXP, "version", "0")
3100
  config.set(constants.INISECT_EXP, "timestamp", "%d" % int(time.time()))
3101
  config.set(constants.INISECT_EXP, "source", instance.primary_node)
3102
  config.set(constants.INISECT_EXP, "os", instance.os)
3103
  config.set(constants.INISECT_EXP, "compression", "none")
3104

    
3105
  config.add_section(constants.INISECT_INS)
3106
  config.set(constants.INISECT_INS, "name", instance.name)
3107
  config.set(constants.INISECT_INS, "maxmem", "%d" %
3108
             instance.beparams[constants.BE_MAXMEM])
3109
  config.set(constants.INISECT_INS, "minmem", "%d" %
3110
             instance.beparams[constants.BE_MINMEM])
3111
  # "memory" is deprecated, but useful for exporting to old ganeti versions
3112
  config.set(constants.INISECT_INS, "memory", "%d" %
3113
             instance.beparams[constants.BE_MAXMEM])
3114
  config.set(constants.INISECT_INS, "vcpus", "%d" %
3115
             instance.beparams[constants.BE_VCPUS])
3116
  config.set(constants.INISECT_INS, "disk_template", instance.disk_template)
3117
  config.set(constants.INISECT_INS, "hypervisor", instance.hypervisor)
3118
  config.set(constants.INISECT_INS, "tags", " ".join(instance.GetTags()))
3119

    
3120
  nic_total = 0
3121
  for nic_count, nic in enumerate(instance.nics):
3122
    nic_total += 1
3123
    config.set(constants.INISECT_INS, "nic%d_mac" %
3124
               nic_count, "%s" % nic.mac)
3125
    config.set(constants.INISECT_INS, "nic%d_ip" % nic_count, "%s" % nic.ip)
3126
    config.set(constants.INISECT_INS, "nic%d_network" % nic_count,
3127
               "%s" % nic.network)
3128
    for param in constants.NICS_PARAMETER_TYPES:
3129
      config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param),
3130
                 "%s" % nic.nicparams.get(param, None))
3131
  # TODO: redundant: on load can read nics until it doesn't exist
3132
  config.set(constants.INISECT_INS, "nic_count", "%d" % nic_total)
3133

    
3134
  disk_total = 0
3135
  for disk_count, disk in enumerate(snap_disks):
3136
    if disk:
3137
      disk_total += 1
3138
      config.set(constants.INISECT_INS, "disk%d_ivname" % disk_count,
3139
                 ("%s" % disk.iv_name))
3140
      config.set(constants.INISECT_INS, "disk%d_dump" % disk_count,
3141
                 ("%s" % disk.logical_id[1]))
3142
      config.set(constants.INISECT_INS, "disk%d_size" % disk_count,
3143
                 ("%d" % disk.size))
3144

    
3145
  config.set(constants.INISECT_INS, "disk_count", "%d" % disk_total)
3146

    
3147
  # New-style hypervisor/backend parameters
3148

    
3149
  config.add_section(constants.INISECT_HYP)
3150
  for name, value in instance.hvparams.items():
3151
    if name not in constants.HVC_GLOBALS:
3152
      config.set(constants.INISECT_HYP, name, str(value))
3153

    
3154
  config.add_section(constants.INISECT_BEP)
3155
  for name, value in instance.beparams.items():
3156
    config.set(constants.INISECT_BEP, name, str(value))
3157

    
3158
  config.add_section(constants.INISECT_OSP)
3159
  for name, value in instance.osparams.items():
3160
    config.set(constants.INISECT_OSP, name, str(value))
3161

    
3162
  utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
3163
                  data=config.Dumps())
3164
  shutil.rmtree(finaldestdir, ignore_errors=True)
3165
  shutil.move(destdir, finaldestdir)
3166

    
3167

    
3168
def ExportInfo(dest):
3169
  """Get export configuration information.
3170

3171
  @type dest: str
3172
  @param dest: directory containing the export
3173

3174
  @rtype: L{objects.SerializableConfigParser}
3175
  @return: a serializable config file containing the
3176
      export info
3177

3178
  """
3179
  cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
3180

    
3181
  config = objects.SerializableConfigParser()
3182
  config.read(cff)
3183

    
3184
  if (not config.has_section(constants.INISECT_EXP) or
3185
      not config.has_section(constants.INISECT_INS)):
3186
    _Fail("Export info file doesn't have the required fields")
3187

    
3188
  return config.Dumps()
3189

    
3190

    
3191
def ListExports():
3192
  """Return a list of exports currently available on this machine.
3193

3194
  @rtype: list
3195
  @return: list of the exports
3196

3197
  """
3198
  if os.path.isdir(pathutils.EXPORT_DIR):
3199
    return sorted(utils.ListVisibleFiles(pathutils.EXPORT_DIR))
3200
  else:
3201
    _Fail("No exports directory")
3202

    
3203

    
3204
def RemoveExport(export):
3205
  """Remove an existing export from the node.
3206

3207
  @type export: str
3208
  @param export: the name of the export to remove
3209
  @rtype: None
3210

3211
  """
3212
  target = utils.PathJoin(pathutils.EXPORT_DIR, export)
3213

    
3214
  try:
3215
    shutil.rmtree(target)
3216
  except EnvironmentError, err:
3217
    _Fail("Error while removing the export: %s", err, exc=True)
3218

    
3219

    
3220
def BlockdevRename(devlist):
3221
  """Rename a list of block devices.
3222

3223
  @type devlist: list of tuples
3224
  @param devlist: list of tuples of the form  (disk, new_unique_id); disk is
3225
      an L{objects.Disk} object describing the current disk, and new
3226
      unique_id is the name we rename it to
3227
  @rtype: boolean
3228
  @return: True if all renames succeeded, False otherwise
3229

3230
  """
3231
  msgs = []
3232
  result = True
3233
  for disk, unique_id in devlist:
3234
    dev = _RecursiveFindBD(disk)
3235
    if dev is None:
3236
      msgs.append("Can't find device %s in rename" % str(disk))
3237
      result = False
3238
      continue
3239
    try:
3240
      old_rpath = dev.dev_path
3241
      dev.Rename(unique_id)
3242
      new_rpath = dev.dev_path
3243
      if old_rpath != new_rpath:
3244
        DevCacheManager.RemoveCache(old_rpath)
3245
        # FIXME: we should add the new cache information here, like:
3246
        # DevCacheManager.UpdateCache(new_rpath, owner, ...)
3247
        # but we don't have the owner here - maybe parse from existing
3248
        # cache? for now, we only lose lvm data when we rename, which
3249
        # is less critical than DRBD or MD
3250
    except errors.BlockDeviceError, err:
3251
      msgs.append("Can't rename device '%s' to '%s': %s" %
3252
                  (dev, unique_id, err))
3253
      logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
3254
      result = False
3255
  if not result:
3256
    _Fail("; ".join(msgs))
3257

    
3258

    
3259
def _TransformFileStorageDir(fs_dir):
3260
  """Checks whether given file_storage_dir is valid.
3261

3262
  Checks wheter the given fs_dir is within the cluster-wide default
3263
  file_storage_dir or the shared_file_storage_dir, which are stored in
3264
  SimpleStore. Only paths under those directories are allowed.
3265

3266
  @type fs_dir: str
3267
  @param fs_dir: the path to check
3268

3269
  @return: the normalized path if valid, None otherwise
3270

3271
  """
3272
  filestorage.CheckFileStoragePath(fs_dir)
3273

    
3274
  return os.path.normpath(fs_dir)
3275

    
3276

    
3277
def CreateFileStorageDir(file_storage_dir):
3278
  """Create file storage directory.
3279

3280
  @type file_storage_dir: str
3281
  @param file_storage_dir: directory to create
3282

3283
  @rtype: tuple
3284
  @return: tuple with first element a boolean indicating wheter dir
3285
      creation was successful or not
3286

3287
  """
3288
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3289
  if os.path.exists(file_storage_dir):
3290
    if not os.path.isdir(file_storage_dir):
3291
      _Fail("Specified storage dir '%s' is not a directory",
3292
            file_storage_dir)
3293
  else:
3294
    try:
3295
      os.makedirs(file_storage_dir, 0750)
3296
    except OSError, err:
3297
      _Fail("Cannot create file storage directory '%s': %s",
3298
            file_storage_dir, err, exc=True)
3299

    
3300

    
3301
def RemoveFileStorageDir(file_storage_dir):
3302
  """Remove file storage directory.
3303

3304
  Remove it only if it's empty. If not log an error and return.
3305

3306
  @type file_storage_dir: str
3307
  @param file_storage_dir: the directory we should cleanup
3308
  @rtype: tuple (success,)
3309
  @return: tuple of one element, C{success}, denoting
3310
      whether the operation was successful
3311

3312
  """
3313
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3314
  if os.path.exists(file_storage_dir):
3315
    if not os.path.isdir(file_storage_dir):
3316
      _Fail("Specified Storage directory '%s' is not a directory",
3317
            file_storage_dir)
3318
    # deletes dir only if empty, otherwise we want to fail the rpc call
3319
    try:
3320
      os.rmdir(file_storage_dir)
3321
    except OSError, err:
3322
      _Fail("Cannot remove file storage directory '%s': %s",
3323
            file_storage_dir, err)
3324

    
3325

    
3326
def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
3327
  """Rename the file storage directory.
3328

3329
  @type old_file_storage_dir: str
3330
  @param old_file_storage_dir: the current path
3331
  @type new_file_storage_dir: str
3332
  @param new_file_storage_dir: the name we should rename to
3333
  @rtype: tuple (success,)
3334
  @return: tuple of one element, C{success}, denoting
3335
      whether the operation was successful
3336

3337
  """
3338
  old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
3339
  new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
3340
  if not os.path.exists(new_file_storage_dir):
3341
    if os.path.isdir(old_file_storage_dir):
3342
      try:
3343
        os.rename(old_file_storage_dir, new_file_storage_dir)
3344
      except OSError, err:
3345
        _Fail("Cannot rename '%s' to '%s': %s",
3346
              old_file_storage_dir, new_file_storage_dir, err)
3347
    else:
3348
      _Fail("Specified storage dir '%s' is not a directory",
3349
            old_file_storage_dir)
3350
  else:
3351
    if os.path.exists(old_file_storage_dir):
3352
      _Fail("Cannot rename '%s' to '%s': both locations exist",
3353
            old_file_storage_dir, new_file_storage_dir)
3354

    
3355

    
3356
def _EnsureJobQueueFile(file_name):
3357
  """Checks whether the given filename is in the queue directory.
3358

3359
  @type file_name: str
3360
  @param file_name: the file name we should check
3361
  @rtype: None
3362
  @raises RPCFail: if the file is not valid
3363

3364
  """
3365
  if not utils.IsBelowDir(pathutils.QUEUE_DIR, file_name):
3366
    _Fail("Passed job queue file '%s' does not belong to"
3367
          " the queue directory '%s'", file_name, pathutils.QUEUE_DIR)
3368

    
3369

    
3370
def JobQueueUpdate(file_name, content):
3371
  """Updates a file in the queue directory.
3372

3373
  This is just a wrapper over L{utils.io.WriteFile}, with proper
3374
  checking.
3375

3376
  @type file_name: str
3377
  @param file_name: the job file name
3378
  @type content: str
3379
  @param content: the new job contents
3380
  @rtype: boolean
3381
  @return: the success of the operation
3382

3383
  """
3384
  file_name = vcluster.LocalizeVirtualPath(file_name)
3385

    
3386
  _EnsureJobQueueFile(file_name)
3387
  getents = runtime.GetEnts()
3388

    
3389
  # Write and replace the file atomically
3390
  utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
3391
                  gid=getents.daemons_gid, mode=constants.JOB_QUEUE_FILES_PERMS)
3392

    
3393

    
3394
def JobQueueRename(old, new):
3395
  """Renames a job queue file.
3396

3397
  This is just a wrapper over os.rename with proper checking.
3398

3399
  @type old: str
3400
  @param old: the old (actual) file name
3401
  @type new: str
3402
  @param new: the desired file name
3403
  @rtype: tuple
3404
  @return: the success of the operation and payload
3405

3406
  """
3407
  old = vcluster.LocalizeVirtualPath(old)
3408
  new = vcluster.LocalizeVirtualPath(new)
3409

    
3410
  _EnsureJobQueueFile(old)
3411
  _EnsureJobQueueFile(new)
3412

    
3413
  getents = runtime.GetEnts()
3414

    
3415
  utils.RenameFile(old, new, mkdir=True, mkdir_mode=0750,
3416
                   dir_uid=getents.masterd_uid, dir_gid=getents.daemons_gid)
3417

    
3418

    
3419
def BlockdevClose(instance_name, disks):
3420
  """Closes the given block devices.
3421

3422
  This means they will be switched to secondary mode (in case of
3423
  DRBD).
3424

3425
  @param instance_name: if the argument is not empty, the symlinks
3426
      of this instance will be removed
3427
  @type disks: list of L{objects.Disk}
3428
  @param disks: the list of disks to be closed
3429
  @rtype: tuple (success, message)
3430
  @return: a tuple of success and message, where success
3431
      indicates the succes of the operation, and message
3432
      which will contain the error details in case we
3433
      failed
3434

3435
  """
3436
  bdevs = []
3437
  for cf in disks:
3438
    rd = _RecursiveFindBD(cf)
3439
    if rd is None:
3440
      _Fail("Can't find device %s", cf)
3441
    bdevs.append(rd)
3442

    
3443
  msg = []
3444
  for rd in bdevs:
3445
    try:
3446
      rd.Close()
3447
    except errors.BlockDeviceError, err:
3448
      msg.append(str(err))
3449
  if msg:
3450
    _Fail("Can't make devices secondary: %s", ",".join(msg))
3451
  else:
3452
    if instance_name:
3453
      _RemoveBlockDevLinks(instance_name, disks)
3454

    
3455

    
3456
def ValidateHVParams(hvname, hvparams):
3457
  """Validates the given hypervisor parameters.
3458

3459
  @type hvname: string
3460
  @param hvname: the hypervisor name
3461
  @type hvparams: dict
3462
  @param hvparams: the hypervisor parameters to be validated
3463
  @rtype: None
3464

3465
  """
3466
  try:
3467
    hv_type = hypervisor.GetHypervisor(hvname)
3468
    hv_type.ValidateParameters(hvparams)
3469
  except errors.HypervisorError, err:
3470
    _Fail(str(err), log=False)
3471

    
3472

    
3473
def _CheckOSPList(os_obj, parameters):
3474
  """Check whether a list of parameters is supported by the OS.
3475

3476
  @type os_obj: L{objects.OS}
3477
  @param os_obj: OS object to check
3478
  @type parameters: list
3479
  @param parameters: the list of parameters to check
3480

3481
  """
3482
  supported = [v[0] for v in os_obj.supported_parameters]
3483
  delta = frozenset(parameters).difference(supported)
3484
  if delta:
3485
    _Fail("The following parameters are not supported"
3486
          " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
3487

    
3488

    
3489
def ValidateOS(required, osname, checks, osparams):
3490
  """Validate the given OS' parameters.
3491

3492
  @type required: boolean
3493
  @param required: whether absence of the OS should translate into
3494
      failure or not
3495
  @type osname: string
3496
  @param osname: the OS to be validated
3497
  @type checks: list
3498
  @param checks: list of the checks to run (currently only 'parameters')
3499
  @type osparams: dict
3500
  @param osparams: dictionary with OS parameters
3501
  @rtype: boolean
3502
  @return: True if the validation passed, or False if the OS was not
3503
      found and L{required} was false
3504

3505
  """
3506
  if not constants.OS_VALIDATE_CALLS.issuperset(checks):
3507
    _Fail("Unknown checks required for OS %s: %s", osname,
3508
          set(checks).difference(constants.OS_VALIDATE_CALLS))
3509

    
3510
  name_only = objects.OS.GetName(osname)
3511
  status, tbv = _TryOSFromDisk(name_only, None)
3512

    
3513
  if not status:
3514
    if required:
3515
      _Fail(tbv)
3516
    else:
3517
      return False
3518

    
3519
  if max(tbv.api_versions) < constants.OS_API_V20:
3520
    return True
3521

    
3522
  if constants.OS_VALIDATE_PARAMETERS in checks:
3523
    _CheckOSPList(tbv, osparams.keys())
3524

    
3525
  validate_env = OSCoreEnv(osname, tbv, osparams)
3526
  result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
3527
                        cwd=tbv.path, reset_env=True)
3528
  if result.failed:
3529
    logging.error("os validate command '%s' returned error: %s output: %s",
3530
                  result.cmd, result.fail_reason, result.output)
3531
    _Fail("OS validation script failed (%s), output: %s",
3532
          result.fail_reason, result.output, log=False)
3533

    
3534
  return True
3535

    
3536

    
3537
def DemoteFromMC():
3538
  """Demotes the current node from master candidate role.
3539

3540
  """
3541
  # try to ensure we're not the master by mistake
3542
  master, myself = ssconf.GetMasterAndMyself()
3543
  if master == myself:
3544
    _Fail("ssconf status shows I'm the master node, will not demote")
3545

    
3546
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "check", constants.MASTERD])
3547
  if not result.failed:
3548
    _Fail("The master daemon is running, will not demote")
3549

    
3550
  try:
3551
    if os.path.isfile(pathutils.CLUSTER_CONF_FILE):
3552
      utils.CreateBackup(pathutils.CLUSTER_CONF_FILE)
3553
  except EnvironmentError, err:
3554
    if err.errno != errno.ENOENT:
3555
      _Fail("Error while backing up cluster file: %s", err, exc=True)
3556

    
3557
  utils.RemoveFile(pathutils.CLUSTER_CONF_FILE)
3558

    
3559

    
3560
def _GetX509Filenames(cryptodir, name):
3561
  """Returns the full paths for the private key and certificate.
3562

3563
  """
3564
  return (utils.PathJoin(cryptodir, name),
3565
          utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
3566
          utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
3567

    
3568

    
3569
def CreateX509Certificate(validity, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3570
  """Creates a new X509 certificate for SSL/TLS.
3571

3572
  @type validity: int
3573
  @param validity: Validity in seconds
3574
  @rtype: tuple; (string, string)
3575
  @return: Certificate name and public part
3576

3577
  """
3578
  (key_pem, cert_pem) = \
3579
    utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
3580
                                     min(validity, _MAX_SSL_CERT_VALIDITY))
3581

    
3582
  cert_dir = tempfile.mkdtemp(dir=cryptodir,
3583
                              prefix="x509-%s-" % utils.TimestampForFilename())
3584
  try:
3585
    name = os.path.basename(cert_dir)
3586
    assert len(name) > 5
3587

    
3588
    (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3589

    
3590
    utils.WriteFile(key_file, mode=0400, data=key_pem)
3591
    utils.WriteFile(cert_file, mode=0400, data=cert_pem)
3592

    
3593
    # Never return private key as it shouldn't leave the node
3594
    return (name, cert_pem)
3595
  except Exception:
3596
    shutil.rmtree(cert_dir, ignore_errors=True)
3597
    raise
3598

    
3599

    
3600
def RemoveX509Certificate(name, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3601
  """Removes a X509 certificate.
3602

3603
  @type name: string
3604
  @param name: Certificate name
3605

3606
  """
3607
  (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3608

    
3609
  utils.RemoveFile(key_file)
3610
  utils.RemoveFile(cert_file)
3611

    
3612
  try:
3613
    os.rmdir(cert_dir)
3614
  except EnvironmentError, err:
3615
    _Fail("Cannot remove certificate directory '%s': %s",
3616
          cert_dir, err)
3617

    
3618

    
3619
def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
3620
  """Returns the command for the requested input/output.
3621

3622
  @type instance: L{objects.Instance}
3623
  @param instance: The instance object
3624
  @param mode: Import/export mode
3625
  @param ieio: Input/output type
3626
  @param ieargs: Input/output arguments
3627

3628
  """
3629
  assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
3630

    
3631
  env = None
3632
  prefix = None
3633
  suffix = None
3634
  exp_size = None
3635

    
3636
  if ieio == constants.IEIO_FILE:
3637
    (filename, ) = ieargs
3638

    
3639
    if not utils.IsNormAbsPath(filename):
3640
      _Fail("Path '%s' is not normalized or absolute", filename)
3641

    
3642
    real_filename = os.path.realpath(filename)
3643
    directory = os.path.dirname(real_filename)
3644

    
3645
    if not utils.IsBelowDir(pathutils.EXPORT_DIR, real_filename):
3646
      _Fail("File '%s' is not under exports directory '%s': %s",
3647
            filename, pathutils.EXPORT_DIR, real_filename)
3648

    
3649
    # Create directory
3650
    utils.Makedirs(directory, mode=0750)
3651

    
3652
    quoted_filename = utils.ShellQuote(filename)
3653

    
3654
    if mode == constants.IEM_IMPORT:
3655
      suffix = "> %s" % quoted_filename
3656
    elif mode == constants.IEM_EXPORT:
3657
      suffix = "< %s" % quoted_filename
3658

    
3659
      # Retrieve file size
3660
      try:
3661
        st = os.stat(filename)
3662
      except EnvironmentError, err:
3663
        logging.error("Can't stat(2) %s: %s", filename, err)
3664
      else:
3665
        exp_size = utils.BytesToMebibyte(st.st_size)
3666

    
3667
  elif ieio == constants.IEIO_RAW_DISK:
3668
    (disk, ) = ieargs
3669

    
3670
    real_disk = _OpenRealBD(disk)
3671

    
3672
    if mode == constants.IEM_IMPORT:
3673
      # we set here a smaller block size as, due to transport buffering, more
3674
      # than 64-128k will mostly ignored; we use nocreat to fail if the device
3675
      # is not already there or we pass a wrong path; we use notrunc to no
3676
      # attempt truncate on an LV device; we use oflag=dsync to not buffer too
3677
      # much memory; this means that at best, we flush every 64k, which will
3678
      # not be very fast
3679
      suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
3680
                                    " bs=%s oflag=dsync"),
3681
                                    real_disk.dev_path,
3682
                                    str(64 * 1024))
3683

    
3684
    elif mode == constants.IEM_EXPORT:
3685
      # the block size on the read dd is 1MiB to match our units
3686
      prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
3687
                                   real_disk.dev_path,
3688
                                   str(1024 * 1024), # 1 MB
3689
                                   str(disk.size))
3690
      exp_size = disk.size
3691

    
3692
  elif ieio == constants.IEIO_SCRIPT:
3693
    (disk, disk_index, ) = ieargs
3694

    
3695
    assert isinstance(disk_index, (int, long))
3696

    
3697
    inst_os = OSFromDisk(instance.os)
3698
    env = OSEnvironment(instance, inst_os)
3699

    
3700
    if mode == constants.IEM_IMPORT:
3701
      env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
3702
      env["IMPORT_INDEX"] = str(disk_index)
3703
      script = inst_os.import_script
3704

    
3705
    elif mode == constants.IEM_EXPORT:
3706
      real_disk = _OpenRealBD(disk)
3707
      env["EXPORT_DEVICE"] = real_disk.dev_path
3708
      env["EXPORT_INDEX"] = str(disk_index)
3709
      script = inst_os.export_script
3710

    
3711
    # TODO: Pass special environment only to script
3712
    script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
3713

    
3714
    if mode == constants.IEM_IMPORT:
3715
      suffix = "| %s" % script_cmd
3716

    
3717
    elif mode == constants.IEM_EXPORT:
3718
      prefix = "%s |" % script_cmd
3719

    
3720
    # Let script predict size
3721
    exp_size = constants.IE_CUSTOM_SIZE
3722

    
3723
  else:
3724
    _Fail("Invalid %s I/O mode %r", mode, ieio)
3725

    
3726
  return (env, prefix, suffix, exp_size)
3727

    
3728

    
3729
def _CreateImportExportStatusDir(prefix):
3730
  """Creates status directory for import/export.
3731

3732
  """
3733
  return tempfile.mkdtemp(dir=pathutils.IMPORT_EXPORT_DIR,
3734
                          prefix=("%s-%s-" %
3735
                                  (prefix, utils.TimestampForFilename())))
3736

    
3737

    
3738
def StartImportExportDaemon(mode, opts, host, port, instance, component,
3739
                            ieio, ieioargs):
3740
  """Starts an import or export daemon.
3741

3742
  @param mode: Import/output mode
3743
  @type opts: L{objects.ImportExportOptions}
3744
  @param opts: Daemon options
3745
  @type host: string
3746
  @param host: Remote host for export (None for import)
3747
  @type port: int
3748
  @param port: Remote port for export (None for import)
3749
  @type instance: L{objects.Instance}
3750
  @param instance: Instance object
3751
  @type component: string
3752
  @param component: which part of the instance is transferred now,
3753
      e.g. 'disk/0'
3754
  @param ieio: Input/output type
3755
  @param ieioargs: Input/output arguments
3756

3757
  """
3758
  if mode == constants.IEM_IMPORT:
3759
    prefix = "import"
3760

    
3761
    if not (host is None and port is None):
3762
      _Fail("Can not specify host or port on import")
3763

    
3764
  elif mode == constants.IEM_EXPORT:
3765
    prefix = "export"
3766

    
3767
    if host is None or port is None:
3768
      _Fail("Host and port must be specified for an export")
3769

    
3770
  else:
3771
    _Fail("Invalid mode %r", mode)
3772

    
3773
  if (opts.key_name is None) ^ (opts.ca_pem is None):
3774
    _Fail("Cluster certificate can only be used for both key and CA")
3775

    
3776
  (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
3777
    _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
3778

    
3779
  if opts.key_name is None:
3780
    # Use server.pem
3781
    key_path = pathutils.NODED_CERT_FILE
3782
    cert_path = pathutils.NODED_CERT_FILE
3783
    assert opts.ca_pem is None
3784
  else:
3785
    (_, key_path, cert_path) = _GetX509Filenames(pathutils.CRYPTO_KEYS_DIR,
3786
                                                 opts.key_name)
3787
    assert opts.ca_pem is not None
3788

    
3789
  for i in [key_path, cert_path]:
3790
    if not os.path.exists(i):
3791
      _Fail("File '%s' does not exist" % i)
3792

    
3793
  status_dir = _CreateImportExportStatusDir("%s-%s" % (prefix, component))
3794
  try:
3795
    status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
3796
    pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
3797
    ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
3798

    
3799
    if opts.ca_pem is None:
3800
      # Use server.pem
3801
      ca = utils.ReadFile(pathutils.NODED_CERT_FILE)
3802
    else:
3803
      ca = opts.ca_pem
3804

    
3805
    # Write CA file
3806
    utils.WriteFile(ca_file, data=ca, mode=0400)
3807

    
3808
    cmd = [
3809
      pathutils.IMPORT_EXPORT_DAEMON,
3810
      status_file, mode,
3811
      "--key=%s" % key_path,
3812
      "--cert=%s" % cert_path,
3813
      "--ca=%s" % ca_file,
3814
      ]
3815

    
3816
    if host:
3817
      cmd.append("--host=%s" % host)
3818

    
3819
    if port:
3820
      cmd.append("--port=%s" % port)
3821

    
3822
    if opts.ipv6:
3823
      cmd.append("--ipv6")
3824
    else:
3825
      cmd.append("--ipv4")
3826

    
3827
    if opts.compress:
3828
      cmd.append("--compress=%s" % opts.compress)
3829

    
3830
    if opts.magic:
3831
      cmd.append("--magic=%s" % opts.magic)
3832

    
3833
    if exp_size is not None:
3834
      cmd.append("--expected-size=%s" % exp_size)
3835

    
3836
    if cmd_prefix:
3837
      cmd.append("--cmd-prefix=%s" % cmd_prefix)
3838

    
3839
    if cmd_suffix:
3840
      cmd.append("--cmd-suffix=%s" % cmd_suffix)
3841

    
3842
    if mode == constants.IEM_EXPORT:
3843
      # Retry connection a few times when connecting to remote peer
3844
      cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3845
      cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3846
    elif opts.connect_timeout is not None:
3847
      assert mode == constants.IEM_IMPORT
3848
      # Overall timeout for establishing connection while listening
3849
      cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3850

    
3851
    logfile = _InstanceLogName(prefix, instance.os, instance.name, component)
3852

    
3853
    # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3854
    # support for receiving a file descriptor for output
3855
    utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3856
                      output=logfile)
3857

    
3858
    # The import/export name is simply the status directory name
3859
    return os.path.basename(status_dir)
3860

    
3861
  except Exception:
3862
    shutil.rmtree(status_dir, ignore_errors=True)
3863
    raise
3864

    
3865

    
3866
def GetImportExportStatus(names):
3867
  """Returns import/export daemon status.
3868

3869
  @type names: sequence
3870
  @param names: List of names
3871
  @rtype: List of dicts
3872
  @return: Returns a list of the state of each named import/export or None if a
3873
           status couldn't be read
3874

3875
  """
3876
  result = []
3877

    
3878
  for name in names:
3879
    status_file = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name,
3880
                                 _IES_STATUS_FILE)
3881

    
3882
    try:
3883
      data = utils.ReadFile(status_file)
3884
    except EnvironmentError, err:
3885
      if err.errno != errno.ENOENT:
3886
        raise
3887
      data = None
3888

    
3889
    if not data:
3890
      result.append(None)
3891
      continue
3892

    
3893
    result.append(serializer.LoadJson(data))
3894

    
3895
  return result
3896

    
3897

    
3898
def AbortImportExport(name):
3899
  """Sends SIGTERM to a running import/export daemon.
3900

3901
  """
3902
  logging.info("Abort import/export %s", name)
3903

    
3904
  status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3905
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3906

    
3907
  if pid:
3908
    logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3909
                 name, pid)
3910
    utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3911

    
3912

    
3913
def CleanupImportExport(name):
3914
  """Cleanup after an import or export.
3915

3916
  If the import/export daemon is still running it's killed. Afterwards the
3917
  whole status directory is removed.
3918

3919
  """
3920
  logging.info("Finalizing import/export %s", name)
3921

    
3922
  status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3923

    
3924
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3925

    
3926
  if pid:
3927
    logging.info("Import/export %s is still running with PID %s",
3928
                 name, pid)
3929
    utils.KillProcess(pid, waitpid=False)
3930

    
3931
  shutil.rmtree(status_dir, ignore_errors=True)
3932

    
3933

    
3934
def _FindDisks(disks):
3935
  """Finds attached L{BlockDev}s for the given disks.
3936

3937
  @type disks: list of L{objects.Disk}
3938
  @param disks: the disk objects we need to find
3939

3940
  @return: list of L{BlockDev} objects or C{None} if a given disk
3941
           was not found or was no attached.
3942

3943
  """
3944
  bdevs = []
3945

    
3946
  for disk in disks:
3947
    rd = _RecursiveFindBD(disk)
3948
    if rd is None:
3949
      _Fail("Can't find device %s", disk)
3950
    bdevs.append(rd)
3951
  return bdevs
3952

    
3953

    
3954
def DrbdDisconnectNet(disks):
3955
  """Disconnects the network on a list of drbd devices.
3956

3957
  """
3958
  bdevs = _FindDisks(disks)
3959

    
3960
  # disconnect disks
3961
  for rd in bdevs:
3962
    try:
3963
      rd.DisconnectNet()
3964
    except errors.BlockDeviceError, err:
3965
      _Fail("Can't change network configuration to standalone mode: %s",
3966
            err, exc=True)
3967

    
3968

    
3969
def DrbdAttachNet(disks, instance_name, multimaster):
3970
  """Attaches the network on a list of drbd devices.
3971

3972
  """
3973
  bdevs = _FindDisks(disks)
3974

    
3975
  if multimaster:
3976
    for idx, rd in enumerate(bdevs):
3977
      try:
3978
        _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3979
      except EnvironmentError, err:
3980
        _Fail("Can't create symlink: %s", err)
3981
  # reconnect disks, switch to new master configuration and if
3982
  # needed primary mode
3983
  for rd in bdevs:
3984
    try:
3985
      rd.AttachNet(multimaster)
3986
    except errors.BlockDeviceError, err:
3987
      _Fail("Can't change network configuration: %s", err)
3988

    
3989
  # wait until the disks are connected; we need to retry the re-attach
3990
  # if the device becomes standalone, as this might happen if the one
3991
  # node disconnects and reconnects in a different mode before the
3992
  # other node reconnects; in this case, one or both of the nodes will
3993
  # decide it has wrong configuration and switch to standalone
3994

    
3995
  def _Attach():
3996
    all_connected = True
3997

    
3998
    for rd in bdevs:
3999
      stats = rd.GetProcStatus()
4000

    
4001
      all_connected = (all_connected and
4002
                       (stats.is_connected or stats.is_in_resync))
4003

    
4004
      if stats.is_standalone:
4005
        # peer had different config info and this node became
4006
        # standalone, even though this should not happen with the
4007
        # new staged way of changing disk configs
4008
        try:
4009
          rd.AttachNet(multimaster)
4010
        except errors.BlockDeviceError, err:
4011
          _Fail("Can't change network configuration: %s", err)
4012

    
4013
    if not all_connected:
4014
      raise utils.RetryAgain()
4015

    
4016
  try:
4017
    # Start with a delay of 100 miliseconds and go up to 5 seconds
4018
    utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
4019
  except utils.RetryTimeout:
4020
    _Fail("Timeout in disk reconnecting")
4021

    
4022
  if multimaster:
4023
    # change to primary mode
4024
    for rd in bdevs:
4025
      try:
4026
        rd.Open()
4027
      except errors.BlockDeviceError, err:
4028
        _Fail("Can't change to primary mode: %s", err)
4029

    
4030

    
4031
def DrbdWaitSync(disks):
4032
  """Wait until DRBDs have synchronized.
4033

4034
  """
4035
  def _helper(rd):
4036
    stats = rd.GetProcStatus()
4037
    if not (stats.is_connected or stats.is_in_resync):
4038
      raise utils.RetryAgain()
4039
    return stats
4040

    
4041
  bdevs = _FindDisks(disks)
4042

    
4043
  min_resync = 100
4044
  alldone = True
4045
  for rd in bdevs:
4046
    try:
4047
      # poll each second for 15 seconds
4048
      stats = utils.Retry(_helper, 1, 15, args=[rd])
4049
    except utils.RetryTimeout:
4050
      stats = rd.GetProcStatus()
4051
      # last check
4052
      if not (stats.is_connected or stats.is_in_resync):
4053
        _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
4054
    alldone = alldone and (not stats.is_in_resync)
4055
    if stats.sync_percent is not None:
4056
      min_resync = min(min_resync, stats.sync_percent)
4057

    
4058
  return (alldone, min_resync)
4059

    
4060

    
4061
def DrbdNeedsActivation(disks):
4062
  """Checks which of the passed disks needs activation and returns their UUIDs.
4063

4064
  """
4065
  faulty_disks = []
4066

    
4067
  for disk in disks:
4068
    rd = _RecursiveFindBD(disk)
4069
    if rd is None:
4070
      faulty_disks.append(disk)
4071
      continue
4072

    
4073
    stats = rd.GetProcStatus()
4074
    if stats.is_standalone or stats.is_diskless:
4075
      faulty_disks.append(disk)
4076

    
4077
  return [disk.uuid for disk in faulty_disks]
4078

    
4079

    
4080
def GetDrbdUsermodeHelper():
4081
  """Returns DRBD usermode helper currently configured.
4082

4083
  """
4084
  try:
4085
    return drbd.DRBD8.GetUsermodeHelper()
4086
  except errors.BlockDeviceError, err:
4087
    _Fail(str(err))
4088

    
4089

    
4090
def PowercycleNode(hypervisor_type, hvparams=None):
4091
  """Hard-powercycle the node.
4092

4093
  Because we need to return first, and schedule the powercycle in the
4094
  background, we won't be able to report failures nicely.
4095

4096
  """
4097
  hyper = hypervisor.GetHypervisor(hypervisor_type)
4098
  try:
4099
    pid = os.fork()
4100
  except OSError:
4101
    # if we can't fork, we'll pretend that we're in the child process
4102
    pid = 0
4103
  if pid > 0:
4104
    return "Reboot scheduled in 5 seconds"
4105
  # ensure the child is running on ram
4106
  try:
4107
    utils.Mlockall()
4108
  except Exception: # pylint: disable=W0703
4109
    pass
4110
  time.sleep(5)
4111
  hyper.PowercycleNode(hvparams=hvparams)
4112

    
4113

    
4114
def _VerifyRestrictedCmdName(cmd):
4115
  """Verifies a restricted command name.
4116

4117
  @type cmd: string
4118
  @param cmd: Command name
4119
  @rtype: tuple; (boolean, string or None)
4120
  @return: The tuple's first element is the status; if C{False}, the second
4121
    element is an error message string, otherwise it's C{None}
4122

4123
  """
4124
  if not cmd.strip():
4125
    return (False, "Missing command name")
4126

    
4127
  if os.path.basename(cmd) != cmd:
4128
    return (False, "Invalid command name")
4129

    
4130
  if not constants.EXT_PLUGIN_MASK.match(cmd):
4131
    return (False, "Command name contains forbidden characters")
4132

    
4133
  return (True, None)
4134

    
4135

    
4136
def _CommonRestrictedCmdCheck(path, owner):
4137
  """Common checks for restricted command file system directories and files.
4138

4139
  @type path: string
4140
  @param path: Path to check
4141
  @param owner: C{None} or tuple containing UID and GID
4142
  @rtype: tuple; (boolean, string or C{os.stat} result)
4143
  @return: The tuple's first element is the status; if C{False}, the second
4144
    element is an error message string, otherwise it's the result of C{os.stat}
4145

4146
  """
4147
  if owner is None:
4148
    # Default to root as owner
4149
    owner = (0, 0)
4150

    
4151
  try:
4152
    st = os.stat(path)
4153
  except EnvironmentError, err:
4154
    return (False, "Can't stat(2) '%s': %s" % (path, err))
4155

    
4156
  if stat.S_IMODE(st.st_mode) & (~_RCMD_MAX_MODE):
4157
    return (False, "Permissions on '%s' are too permissive" % path)
4158

    
4159
  if (st.st_uid, st.st_gid) != owner:
4160
    (owner_uid, owner_gid) = owner
4161
    return (False, "'%s' is not owned by %s:%s" % (path, owner_uid, owner_gid))
4162

    
4163
  return (True, st)
4164

    
4165

    
4166
def _VerifyRestrictedCmdDirectory(path, _owner=None):
4167
  """Verifies restricted command directory.
4168

4169
  @type path: string
4170
  @param path: Path to check
4171
  @rtype: tuple; (boolean, string or None)
4172
  @return: The tuple's first element is the status; if C{False}, the second
4173
    element is an error message string, otherwise it's C{None}
4174

4175
  """
4176
  (status, value) = _CommonRestrictedCmdCheck(path, _owner)
4177

    
4178
  if not status:
4179
    return (False, value)
4180

    
4181
  if not stat.S_ISDIR(value.st_mode):
4182
    return (False, "Path '%s' is not a directory" % path)
4183

    
4184
  return (True, None)
4185

    
4186

    
4187
def _VerifyRestrictedCmd(path, cmd, _owner=None):
4188
  """Verifies a whole restricted command and returns its executable filename.
4189

4190
  @type path: string
4191
  @param path: Directory containing restricted commands
4192
  @type cmd: string
4193
  @param cmd: Command name
4194
  @rtype: tuple; (boolean, string)
4195
  @return: The tuple's first element is the status; if C{False}, the second
4196
    element is an error message string, otherwise the second element is the
4197
    absolute path to the executable
4198

4199
  """
4200
  executable = utils.PathJoin(path, cmd)
4201

    
4202
  (status, msg) = _CommonRestrictedCmdCheck(executable, _owner)
4203

    
4204
  if not status:
4205
    return (False, msg)
4206

    
4207
  if not utils.IsExecutable(executable):
4208
    return (False, "access(2) thinks '%s' can't be executed" % executable)
4209

    
4210
  return (True, executable)
4211

    
4212

    
4213
def _PrepareRestrictedCmd(path, cmd,
4214
                          _verify_dir=_VerifyRestrictedCmdDirectory,
4215
                          _verify_name=_VerifyRestrictedCmdName,
4216
                          _verify_cmd=_VerifyRestrictedCmd):
4217
  """Performs a number of tests on a restricted command.
4218

4219
  @type path: string
4220
  @param path: Directory containing restricted commands
4221
  @type cmd: string
4222
  @param cmd: Command name
4223
  @return: Same as L{_VerifyRestrictedCmd}
4224

4225
  """
4226
  # Verify the directory first
4227
  (status, msg) = _verify_dir(path)
4228
  if status:
4229
    # Check command if everything was alright
4230
    (status, msg) = _verify_name(cmd)
4231

    
4232
  if not status:
4233
    return (False, msg)
4234

    
4235
  # Check actual executable
4236
  return _verify_cmd(path, cmd)
4237

    
4238

    
4239
def RunRestrictedCmd(cmd,
4240
                     _lock_timeout=_RCMD_LOCK_TIMEOUT,
4241
                     _lock_file=pathutils.RESTRICTED_COMMANDS_LOCK_FILE,
4242
                     _path=pathutils.RESTRICTED_COMMANDS_DIR,
4243
                     _sleep_fn=time.sleep,
4244
                     _prepare_fn=_PrepareRestrictedCmd,
4245
                     _runcmd_fn=utils.RunCmd,
4246
                     _enabled=constants.ENABLE_RESTRICTED_COMMANDS):
4247
  """Executes a restricted command after performing strict tests.
4248

4249
  @type cmd: string
4250
  @param cmd: Command name
4251
  @rtype: string
4252
  @return: Command output
4253
  @raise RPCFail: In case of an error
4254

4255
  """
4256
  logging.info("Preparing to run restricted command '%s'", cmd)
4257

    
4258
  if not _enabled:
4259
    _Fail("Restricted commands disabled at configure time")
4260

    
4261
  lock = None
4262
  try:
4263
    cmdresult = None
4264
    try:
4265
      lock = utils.FileLock.Open(_lock_file)
4266
      lock.Exclusive(blocking=True, timeout=_lock_timeout)
4267

    
4268
      (status, value) = _prepare_fn(_path, cmd)
4269

    
4270
      if status:
4271
        cmdresult = _runcmd_fn([value], env={}, reset_env=True,
4272
                               postfork_fn=lambda _: lock.Unlock())
4273
      else:
4274
        logging.error(value)
4275
    except Exception: # pylint: disable=W0703
4276
      # Keep original error in log
4277
      logging.exception("Caught exception")
4278

    
4279
    if cmdresult is None:
4280
      logging.info("Sleeping for %0.1f seconds before returning",
4281
                   _RCMD_INVALID_DELAY)
4282
      _sleep_fn(_RCMD_INVALID_DELAY)
4283

    
4284
      # Do not include original error message in returned error
4285
      _Fail("Executing command '%s' failed" % cmd)
4286
    elif cmdresult.failed or cmdresult.fail_reason:
4287
      _Fail("Restricted command '%s' failed: %s; output: %s",
4288
            cmd, cmdresult.fail_reason, cmdresult.output)
4289
    else:
4290
      return cmdresult.output
4291
  finally:
4292
    if lock is not None:
4293
      # Release lock at last
4294
      lock.Close()
4295
      lock = None
4296

    
4297

    
4298
def SetWatcherPause(until, _filename=pathutils.WATCHER_PAUSEFILE):
4299
  """Creates or removes the watcher pause file.
4300

4301
  @type until: None or number
4302
  @param until: Unix timestamp saying until when the watcher shouldn't run
4303

4304
  """
4305
  if until is None:
4306
    logging.info("Received request to no longer pause watcher")
4307
    utils.RemoveFile(_filename)
4308
  else:
4309
    logging.info("Received request to pause watcher until %s", until)
4310

    
4311
    if not ht.TNumber(until):
4312
      _Fail("Duration must be numeric")
4313

    
4314
    utils.WriteFile(_filename, data="%d\n" % (until, ), mode=0644)
4315

    
4316

    
4317
def ConfigureOVS(ovs_name, ovs_link):
4318
  """Creates a OpenvSwitch on the node.
4319

4320
  This function sets up a OpenvSwitch on the node with given name nad
4321
  connects it via a given eth device.
4322

4323
  @type ovs_name: string
4324
  @param ovs_name: Name of the OpenvSwitch to create.
4325
  @type ovs_link: None or string
4326
  @param ovs_link: Ethernet device for outside connection (can be missing)
4327

4328
  """
4329
  # Initialize the OpenvSwitch
4330
  result = utils.RunCmd(["ovs-vsctl", "add-br", ovs_name])
4331
  if result.failed:
4332
    _Fail("Failed to create openvswitch %s. Script return value: %s, output:"
4333
          " '%s'" % result.exit_code, result.output, log=True)
4334

    
4335
  # And connect it to a physical interface, if given
4336
  if ovs_link:
4337
    result = utils.RunCmd(["ovs-vsctl", "add-port", ovs_name, ovs_link])
4338
    if result.failed:
4339
      _Fail("Failed to connect openvswitch to  interface %s. Script return"
4340
            " value: %s, output: '%s'" % ovs_link, result.exit_code,
4341
            result.output, log=True)
4342

    
4343

    
4344
class HooksRunner(object):
4345
  """Hook runner.
4346

4347
  This class is instantiated on the node side (ganeti-noded) and not
4348
  on the master side.
4349

4350
  """
4351
  def __init__(self, hooks_base_dir=None):
4352
    """Constructor for hooks runner.
4353

4354
    @type hooks_base_dir: str or None
4355
    @param hooks_base_dir: if not None, this overrides the
4356
        L{pathutils.HOOKS_BASE_DIR} (useful for unittests)
4357

4358
    """
4359
    if hooks_base_dir is None:
4360
      hooks_base_dir = pathutils.HOOKS_BASE_DIR
4361
    # yeah, _BASE_DIR is not valid for attributes, we use it like a
4362
    # constant
4363
    self._BASE_DIR = hooks_base_dir # pylint: disable=C0103
4364

    
4365
  def RunLocalHooks(self, node_list, hpath, phase, env):
4366
    """Check that the hooks will be run only locally and then run them.
4367

4368
    """
4369
    assert len(node_list) == 1
4370
    node = node_list[0]
4371
    _, myself = ssconf.GetMasterAndMyself()
4372
    assert node == myself
4373

    
4374
    results = self.RunHooks(hpath, phase, env)
4375

    
4376
    # Return values in the form expected by HooksMaster
4377
    return {node: (None, False, results)}
4378

    
4379
  def RunHooks(self, hpath, phase, env):
4380
    """Run the scripts in the hooks directory.
4381

4382
    @type hpath: str
4383
    @param hpath: the path to the hooks directory which
4384
        holds the scripts
4385
    @type phase: str
4386
    @param phase: either L{constants.HOOKS_PHASE_PRE} or
4387
        L{constants.HOOKS_PHASE_POST}
4388
    @type env: dict
4389
    @param env: dictionary with the environment for the hook
4390
    @rtype: list
4391
    @return: list of 3-element tuples:
4392
      - script path
4393
      - script result, either L{constants.HKR_SUCCESS} or
4394
        L{constants.HKR_FAIL}
4395
      - output of the script
4396

4397
    @raise errors.ProgrammerError: for invalid input
4398
        parameters
4399

4400
    """
4401
    if phase == constants.HOOKS_PHASE_PRE:
4402
      suffix = "pre"
4403
    elif phase == constants.HOOKS_PHASE_POST:
4404
      suffix = "post"
4405
    else:
4406
      _Fail("Unknown hooks phase '%s'", phase)
4407

    
4408
    subdir = "%s-%s.d" % (hpath, suffix)
4409
    dir_name = utils.PathJoin(self._BASE_DIR, subdir)
4410

    
4411
    results = []
4412

    
4413
    if not os.path.isdir(dir_name):
4414
      # for non-existing/non-dirs, we simply exit instead of logging a
4415
      # warning at every operation
4416
      return results
4417

    
4418
    runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
4419

    
4420
    for (relname, relstatus, runresult) in runparts_results:
4421
      if relstatus == constants.RUNPARTS_SKIP:
4422
        rrval = constants.HKR_SKIP
4423
        output = ""
4424
      elif relstatus == constants.RUNPARTS_ERR:
4425
        rrval = constants.HKR_FAIL
4426
        output = "Hook script execution error: %s" % runresult
4427
      elif relstatus == constants.RUNPARTS_RUN:
4428
        if runresult.failed:
4429
          rrval = constants.HKR_FAIL
4430
        else:
4431
          rrval = constants.HKR_SUCCESS
4432
        output = utils.SafeEncode(runresult.output.strip())
4433
      results.append(("%s/%s" % (subdir, relname), rrval, output))
4434

    
4435
    return results
4436

    
4437

    
4438
class IAllocatorRunner(object):
4439
  """IAllocator runner.
4440

4441
  This class is instantiated on the node side (ganeti-noded) and not on
4442
  the master side.
4443

4444
  """
4445
  @staticmethod
4446
  def Run(name, idata):
4447
    """Run an iallocator script.
4448

4449
    @type name: str
4450
    @param name: the iallocator script name
4451
    @type idata: str
4452
    @param idata: the allocator input data
4453

4454
    @rtype: tuple
4455
    @return: two element tuple of:
4456
       - status
4457
       - either error message or stdout of allocator (for success)
4458

4459
    """
4460
    alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
4461
                                  os.path.isfile)
4462
    if alloc_script is None:
4463
      _Fail("iallocator module '%s' not found in the search path", name)
4464

    
4465
    fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
4466
    try:
4467
      os.write(fd, idata)
4468
      os.close(fd)
4469
      result = utils.RunCmd([alloc_script, fin_name])
4470
      if result.failed:
4471
        _Fail("iallocator module '%s' failed: %s, output '%s'",
4472
              name, result.fail_reason, result.output)
4473
    finally:
4474
      os.unlink(fin_name)
4475

    
4476
    return result.stdout
4477

    
4478

    
4479
class DevCacheManager(object):
4480
  """Simple class for managing a cache of block device information.
4481

4482
  """
4483
  _DEV_PREFIX = "/dev/"
4484
  _ROOT_DIR = pathutils.BDEV_CACHE_DIR
4485

    
4486
  @classmethod
4487
  def _ConvertPath(cls, dev_path):
4488
    """Converts a /dev/name path to the cache file name.
4489

4490
    This replaces slashes with underscores and strips the /dev
4491
    prefix. It then returns the full path to the cache file.
4492

4493
    @type dev_path: str
4494
    @param dev_path: the C{/dev/} path name
4495
    @rtype: str
4496
    @return: the converted path name
4497

4498
    """
4499
    if dev_path.startswith(cls._DEV_PREFIX):
4500
      dev_path = dev_path[len(cls._DEV_PREFIX):]
4501
    dev_path = dev_path.replace("/", "_")
4502
    fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
4503
    return fpath
4504

    
4505
  @classmethod
4506
  def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
4507
    """Updates the cache information for a given device.
4508

4509
    @type dev_path: str
4510
    @param dev_path: the pathname of the device
4511
    @type owner: str
4512
    @param owner: the owner (instance name) of the device
4513
    @type on_primary: bool
4514
    @param on_primary: whether this is the primary
4515
        node nor not
4516
    @type iv_name: str
4517
    @param iv_name: the instance-visible name of the
4518
        device, as in objects.Disk.iv_name
4519

4520
    @rtype: None
4521

4522
    """
4523
    if dev_path is None:
4524
      logging.error("DevCacheManager.UpdateCache got a None dev_path")
4525
      return
4526
    fpath = cls._ConvertPath(dev_path)
4527
    if on_primary:
4528
      state = "primary"
4529
    else:
4530
      state = "secondary"
4531
    if iv_name is None:
4532
      iv_name = "not_visible"
4533
    fdata = "%s %s %s\n" % (str(owner), state, iv_name)
4534
    try:
4535
      utils.WriteFile(fpath, data=fdata)
4536
    except EnvironmentError, err:
4537
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
4538

    
4539
  @classmethod
4540
  def RemoveCache(cls, dev_path):
4541
    """Remove data for a dev_path.
4542

4543
    This is just a wrapper over L{utils.io.RemoveFile} with a converted
4544
    path name and logging.
4545

4546
    @type dev_path: str
4547
    @param dev_path: the pathname of the device
4548

4549
    @rtype: None
4550

4551
    """
4552
    if dev_path is None:
4553
      logging.error("DevCacheManager.RemoveCache got a None dev_path")
4554
      return
4555
    fpath = cls._ConvertPath(dev_path)
4556
    try:
4557
      utils.RemoveFile(fpath)
4558
    except EnvironmentError, err:
4559
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)