Statistics
| Branch: | Tag: | Revision:

root / lib / backend.py @ ff5def9b

History | View | Annotate | Download (140.3 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
    # This can raise errors.BlockDeviceError
1631
    return device.GetUserspaceAccessUri(instance.hypervisor)
1632
  else:
1633
    return None
1634

    
1635

    
1636
def _GatherAndLinkBlockDevs(instance):
1637
  """Set up an instance's block device(s).
1638

1639
  This is run on the primary node at instance startup. The block
1640
  devices must be already assembled.
1641

1642
  @type instance: L{objects.Instance}
1643
  @param instance: the instance whose disks we should assemble
1644
  @rtype: list
1645
  @return: list of (disk_object, link_name, drive_uri)
1646

1647
  """
1648
  block_devices = []
1649
  for idx, disk in enumerate(instance.disks):
1650
    device = _RecursiveFindBD(disk)
1651
    if device is None:
1652
      raise errors.BlockDeviceError("Block device '%s' is not set up." %
1653
                                    str(disk))
1654
    device.Open()
1655
    try:
1656
      link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1657
    except OSError, e:
1658
      raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1659
                                    e.strerror)
1660
    uri = _CalculateDeviceURI(instance, disk, device)
1661

    
1662
    block_devices.append((disk, link_name, uri))
1663

    
1664
  return block_devices
1665

    
1666

    
1667
def StartInstance(instance, startup_paused, reason, store_reason=True):
1668
  """Start an instance.
1669

1670
  @type instance: L{objects.Instance}
1671
  @param instance: the instance object
1672
  @type startup_paused: bool
1673
  @param instance: pause instance at startup?
1674
  @type reason: list of reasons
1675
  @param reason: the reason trail for this startup
1676
  @type store_reason: boolean
1677
  @param store_reason: whether to store the shutdown reason trail on file
1678
  @rtype: None
1679

1680
  """
1681
  running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1682
                                                   instance.hvparams)
1683

    
1684
  if instance.name in running_instances:
1685
    logging.info("Instance %s already running, not starting", instance.name)
1686
    return
1687

    
1688
  try:
1689
    block_devices = _GatherAndLinkBlockDevs(instance)
1690
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
1691
    hyper.StartInstance(instance, block_devices, startup_paused)
1692
    if store_reason:
1693
      _StoreInstReasonTrail(instance.name, reason)
1694
  except errors.BlockDeviceError, err:
1695
    _Fail("Block device error: %s", err, exc=True)
1696
  except errors.HypervisorError, err:
1697
    _RemoveBlockDevLinks(instance.name, instance.disks)
1698
    _Fail("Hypervisor error: %s", err, exc=True)
1699

    
1700

    
1701
def InstanceShutdown(instance, timeout, reason, store_reason=True):
1702
  """Shut an instance down.
1703

1704
  @note: this functions uses polling with a hardcoded timeout.
1705

1706
  @type instance: L{objects.Instance}
1707
  @param instance: the instance object
1708
  @type timeout: integer
1709
  @param timeout: maximum timeout for soft shutdown
1710
  @type reason: list of reasons
1711
  @param reason: the reason trail for this shutdown
1712
  @type store_reason: boolean
1713
  @param store_reason: whether to store the shutdown reason trail on file
1714
  @rtype: None
1715

1716
  """
1717
  hv_name = instance.hypervisor
1718
  hyper = hypervisor.GetHypervisor(hv_name)
1719
  iname = instance.name
1720

    
1721
  if instance.name not in hyper.ListInstances(instance.hvparams):
1722
    logging.info("Instance %s not running, doing nothing", iname)
1723
    return
1724

    
1725
  class _TryShutdown:
1726
    def __init__(self):
1727
      self.tried_once = False
1728

    
1729
    def __call__(self):
1730
      if iname not in hyper.ListInstances(instance.hvparams):
1731
        return
1732

    
1733
      try:
1734
        hyper.StopInstance(instance, retry=self.tried_once)
1735
        if store_reason:
1736
          _StoreInstReasonTrail(instance.name, reason)
1737
      except errors.HypervisorError, err:
1738
        if iname not in hyper.ListInstances(instance.hvparams):
1739
          # if the instance is no longer existing, consider this a
1740
          # success and go to cleanup
1741
          return
1742

    
1743
        _Fail("Failed to stop instance %s: %s", iname, err)
1744

    
1745
      self.tried_once = True
1746

    
1747
      raise utils.RetryAgain()
1748

    
1749
  try:
1750
    utils.Retry(_TryShutdown(), 5, timeout)
1751
  except utils.RetryTimeout:
1752
    # the shutdown did not succeed
1753
    logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1754

    
1755
    try:
1756
      hyper.StopInstance(instance, force=True)
1757
    except errors.HypervisorError, err:
1758
      if iname in hyper.ListInstances(instance.hvparams):
1759
        # only raise an error if the instance still exists, otherwise
1760
        # the error could simply be "instance ... unknown"!
1761
        _Fail("Failed to force stop instance %s: %s", iname, err)
1762

    
1763
    time.sleep(1)
1764

    
1765
    if iname in hyper.ListInstances(instance.hvparams):
1766
      _Fail("Could not shutdown instance %s even by destroy", iname)
1767

    
1768
  try:
1769
    hyper.CleanupInstance(instance.name)
1770
  except errors.HypervisorError, err:
1771
    logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1772

    
1773
  _RemoveBlockDevLinks(iname, instance.disks)
1774

    
1775

    
1776
def InstanceReboot(instance, reboot_type, shutdown_timeout, reason):
1777
  """Reboot an instance.
1778

1779
  @type instance: L{objects.Instance}
1780
  @param instance: the instance object to reboot
1781
  @type reboot_type: str
1782
  @param reboot_type: the type of reboot, one the following
1783
    constants:
1784
      - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1785
        instance OS, do not recreate the VM
1786
      - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1787
        restart the VM (at the hypervisor level)
1788
      - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1789
        not accepted here, since that mode is handled differently, in
1790
        cmdlib, and translates into full stop and start of the
1791
        instance (instead of a call_instance_reboot RPC)
1792
  @type shutdown_timeout: integer
1793
  @param shutdown_timeout: maximum timeout for soft shutdown
1794
  @type reason: list of reasons
1795
  @param reason: the reason trail for this reboot
1796
  @rtype: None
1797

1798
  """
1799
  running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1800
                                                   instance.hvparams)
1801

    
1802
  if instance.name not in running_instances:
1803
    _Fail("Cannot reboot instance %s that is not running", instance.name)
1804

    
1805
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1806
  if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1807
    try:
1808
      hyper.RebootInstance(instance)
1809
    except errors.HypervisorError, err:
1810
      _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1811
  elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1812
    try:
1813
      InstanceShutdown(instance, shutdown_timeout, reason, store_reason=False)
1814
      result = StartInstance(instance, False, reason, store_reason=False)
1815
      _StoreInstReasonTrail(instance.name, reason)
1816
      return result
1817
    except errors.HypervisorError, err:
1818
      _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1819
  else:
1820
    _Fail("Invalid reboot_type received: %s", reboot_type)
1821

    
1822

    
1823
def InstanceBalloonMemory(instance, memory):
1824
  """Resize an instance's memory.
1825

1826
  @type instance: L{objects.Instance}
1827
  @param instance: the instance object
1828
  @type memory: int
1829
  @param memory: new memory amount in MB
1830
  @rtype: None
1831

1832
  """
1833
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1834
  running = hyper.ListInstances(instance.hvparams)
1835
  if instance.name not in running:
1836
    logging.info("Instance %s is not running, cannot balloon", instance.name)
1837
    return
1838
  try:
1839
    hyper.BalloonInstanceMemory(instance, memory)
1840
  except errors.HypervisorError, err:
1841
    _Fail("Failed to balloon instance memory: %s", err, exc=True)
1842

    
1843

    
1844
def MigrationInfo(instance):
1845
  """Gather information about an instance to be migrated.
1846

1847
  @type instance: L{objects.Instance}
1848
  @param instance: the instance definition
1849

1850
  """
1851
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1852
  try:
1853
    info = hyper.MigrationInfo(instance)
1854
  except errors.HypervisorError, err:
1855
    _Fail("Failed to fetch migration information: %s", err, exc=True)
1856
  return info
1857

    
1858

    
1859
def AcceptInstance(instance, info, target):
1860
  """Prepare the node to accept an instance.
1861

1862
  @type instance: L{objects.Instance}
1863
  @param instance: the instance definition
1864
  @type info: string/data (opaque)
1865
  @param info: migration information, from the source node
1866
  @type target: string
1867
  @param target: target host (usually ip), on this node
1868

1869
  """
1870
  # TODO: why is this required only for DTS_EXT_MIRROR?
1871
  if instance.disk_template in constants.DTS_EXT_MIRROR:
1872
    # Create the symlinks, as the disks are not active
1873
    # in any way
1874
    try:
1875
      _GatherAndLinkBlockDevs(instance)
1876
    except errors.BlockDeviceError, err:
1877
      _Fail("Block device error: %s", err, exc=True)
1878

    
1879
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1880
  try:
1881
    hyper.AcceptInstance(instance, info, target)
1882
  except errors.HypervisorError, err:
1883
    if instance.disk_template in constants.DTS_EXT_MIRROR:
1884
      _RemoveBlockDevLinks(instance.name, instance.disks)
1885
    _Fail("Failed to accept instance: %s", err, exc=True)
1886

    
1887

    
1888
def FinalizeMigrationDst(instance, info, success):
1889
  """Finalize any preparation to accept an instance.
1890

1891
  @type instance: L{objects.Instance}
1892
  @param instance: the instance definition
1893
  @type info: string/data (opaque)
1894
  @param info: migration information, from the source node
1895
  @type success: boolean
1896
  @param success: whether the migration was a success or a failure
1897

1898
  """
1899
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1900
  try:
1901
    hyper.FinalizeMigrationDst(instance, info, success)
1902
  except errors.HypervisorError, err:
1903
    _Fail("Failed to finalize migration on the target node: %s", err, exc=True)
1904

    
1905

    
1906
def MigrateInstance(cluster_name, instance, target, live):
1907
  """Migrates an instance to another node.
1908

1909
  @type cluster_name: string
1910
  @param cluster_name: name of the cluster
1911
  @type instance: L{objects.Instance}
1912
  @param instance: the instance definition
1913
  @type target: string
1914
  @param target: the target node name
1915
  @type live: boolean
1916
  @param live: whether the migration should be done live or not (the
1917
      interpretation of this parameter is left to the hypervisor)
1918
  @raise RPCFail: if migration fails for some reason
1919

1920
  """
1921
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1922

    
1923
  try:
1924
    hyper.MigrateInstance(cluster_name, instance, target, live)
1925
  except errors.HypervisorError, err:
1926
    _Fail("Failed to migrate instance: %s", err, exc=True)
1927

    
1928

    
1929
def FinalizeMigrationSource(instance, success, live):
1930
  """Finalize the instance migration on the source node.
1931

1932
  @type instance: L{objects.Instance}
1933
  @param instance: the instance definition of the migrated instance
1934
  @type success: bool
1935
  @param success: whether the migration succeeded or not
1936
  @type live: bool
1937
  @param live: whether the user requested a live migration or not
1938
  @raise RPCFail: If the execution fails for some reason
1939

1940
  """
1941
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1942

    
1943
  try:
1944
    hyper.FinalizeMigrationSource(instance, success, live)
1945
  except Exception, err:  # pylint: disable=W0703
1946
    _Fail("Failed to finalize the migration on the source node: %s", err,
1947
          exc=True)
1948

    
1949

    
1950
def GetMigrationStatus(instance):
1951
  """Get the migration status
1952

1953
  @type instance: L{objects.Instance}
1954
  @param instance: the instance that is being migrated
1955
  @rtype: L{objects.MigrationStatus}
1956
  @return: the status of the current migration (one of
1957
           L{constants.HV_MIGRATION_VALID_STATUSES}), plus any additional
1958
           progress info that can be retrieved from the hypervisor
1959
  @raise RPCFail: If the migration status cannot be retrieved
1960

1961
  """
1962
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1963
  try:
1964
    return hyper.GetMigrationStatus(instance)
1965
  except Exception, err:  # pylint: disable=W0703
1966
    _Fail("Failed to get migration status: %s", err, exc=True)
1967

    
1968

    
1969
def HotplugDevice(instance, action, dev_type, device, extra, seq):
1970
  """Hotplug a device
1971

1972
  Hotplug is currently supported only for KVM Hypervisor.
1973
  @type instance: L{objects.Instance}
1974
  @param instance: the instance to which we hotplug a device
1975
  @type action: string
1976
  @param action: the hotplug action to perform
1977
  @type dev_type: string
1978
  @param dev_type: the device type to hotplug
1979
  @type device: either L{objects.NIC} or L{objects.Disk}
1980
  @param device: the device object to hotplug
1981
  @type extra: string
1982
  @param extra: extra info used by hotplug code (e.g. disk link)
1983
  @type seq: int
1984
  @param seq: the index of the device from master perspective
1985
  @raise RPCFail: in case instance does not have KVM hypervisor
1986

1987
  """
1988
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1989
  try:
1990
    hyper.HotplugSupported(instance, action, dev_type)
1991
  except (errors.HotplugError, errors.HypervisorError), err:
1992
    _Fail("Hotplug is not supported: %s", err)
1993

    
1994
  if action == constants.HOTPLUG_ACTION_ADD:
1995
    fn = hyper.HotAddDevice
1996
  elif action == constants.HOTPLUG_ACTION_REMOVE:
1997
    fn = hyper.HotDelDevice
1998
  elif action == constants.HOTPLUG_ACTION_MODIFY:
1999
    fn = hyper.HotModDevice
2000
  else:
2001
    assert action in constants.HOTPLUG_ALL_ACTIONS
2002

    
2003
  return fn(instance, dev_type, device, extra, seq)
2004

    
2005

    
2006
def BlockdevCreate(disk, size, owner, on_primary, info, excl_stor):
2007
  """Creates a block device for an instance.
2008

2009
  @type disk: L{objects.Disk}
2010
  @param disk: the object describing the disk we should create
2011
  @type size: int
2012
  @param size: the size of the physical underlying device, in MiB
2013
  @type owner: str
2014
  @param owner: the name of the instance for which disk is created,
2015
      used for device cache data
2016
  @type on_primary: boolean
2017
  @param on_primary:  indicates if it is the primary node or not
2018
  @type info: string
2019
  @param info: string that will be sent to the physical device
2020
      creation, used for example to set (LVM) tags on LVs
2021
  @type excl_stor: boolean
2022
  @param excl_stor: Whether exclusive_storage is active
2023

2024
  @return: the new unique_id of the device (this can sometime be
2025
      computed only after creation), or None. On secondary nodes,
2026
      it's not required to return anything.
2027

2028
  """
2029
  # TODO: remove the obsolete "size" argument
2030
  # pylint: disable=W0613
2031
  clist = []
2032
  if disk.children:
2033
    for child in disk.children:
2034
      try:
2035
        crdev = _RecursiveAssembleBD(child, owner, on_primary)
2036
      except errors.BlockDeviceError, err:
2037
        _Fail("Can't assemble device %s: %s", child, err)
2038
      if on_primary or disk.AssembleOnSecondary():
2039
        # we need the children open in case the device itself has to
2040
        # be assembled
2041
        try:
2042
          # pylint: disable=E1103
2043
          crdev.Open()
2044
        except errors.BlockDeviceError, err:
2045
          _Fail("Can't make child '%s' read-write: %s", child, err)
2046
      clist.append(crdev)
2047

    
2048
  try:
2049
    device = bdev.Create(disk, clist, excl_stor)
2050
  except errors.BlockDeviceError, err:
2051
    _Fail("Can't create block device: %s", err)
2052

    
2053
  if on_primary or disk.AssembleOnSecondary():
2054
    try:
2055
      device.Assemble()
2056
    except errors.BlockDeviceError, err:
2057
      _Fail("Can't assemble device after creation, unusual event: %s", err)
2058
    if on_primary or disk.OpenOnSecondary():
2059
      try:
2060
        device.Open(force=True)
2061
      except errors.BlockDeviceError, err:
2062
        _Fail("Can't make device r/w after creation, unusual event: %s", err)
2063
    DevCacheManager.UpdateCache(device.dev_path, owner,
2064
                                on_primary, disk.iv_name)
2065

    
2066
  device.SetInfo(info)
2067

    
2068
  return device.unique_id
2069

    
2070

    
2071
def _WipeDevice(path, offset, size):
2072
  """This function actually wipes the device.
2073

2074
  @param path: The path to the device to wipe
2075
  @param offset: The offset in MiB in the file
2076
  @param size: The size in MiB to write
2077

2078
  """
2079
  # Internal sizes are always in Mebibytes; if the following "dd" command
2080
  # should use a different block size the offset and size given to this
2081
  # function must be adjusted accordingly before being passed to "dd".
2082
  block_size = 1024 * 1024
2083

    
2084
  cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
2085
         "bs=%s" % block_size, "oflag=direct", "of=%s" % path,
2086
         "count=%d" % size]
2087
  result = utils.RunCmd(cmd)
2088

    
2089
  if result.failed:
2090
    _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
2091
          result.fail_reason, result.output)
2092

    
2093

    
2094
def BlockdevWipe(disk, offset, size):
2095
  """Wipes a block device.
2096

2097
  @type disk: L{objects.Disk}
2098
  @param disk: the disk object we want to wipe
2099
  @type offset: int
2100
  @param offset: The offset in MiB in the file
2101
  @type size: int
2102
  @param size: The size in MiB to write
2103

2104
  """
2105
  try:
2106
    rdev = _RecursiveFindBD(disk)
2107
  except errors.BlockDeviceError:
2108
    rdev = None
2109

    
2110
  if not rdev:
2111
    _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
2112

    
2113
  # Do cross verify some of the parameters
2114
  if offset < 0:
2115
    _Fail("Negative offset")
2116
  if size < 0:
2117
    _Fail("Negative size")
2118
  if offset > rdev.size:
2119
    _Fail("Offset is bigger than device size")
2120
  if (offset + size) > rdev.size:
2121
    _Fail("The provided offset and size to wipe is bigger than device size")
2122

    
2123
  _WipeDevice(rdev.dev_path, offset, size)
2124

    
2125

    
2126
def BlockdevPauseResumeSync(disks, pause):
2127
  """Pause or resume the sync of the block device.
2128

2129
  @type disks: list of L{objects.Disk}
2130
  @param disks: the disks object we want to pause/resume
2131
  @type pause: bool
2132
  @param pause: Wheater to pause or resume
2133

2134
  """
2135
  success = []
2136
  for disk in disks:
2137
    try:
2138
      rdev = _RecursiveFindBD(disk)
2139
    except errors.BlockDeviceError:
2140
      rdev = None
2141

    
2142
    if not rdev:
2143
      success.append((False, ("Cannot change sync for device %s:"
2144
                              " device not found" % disk.iv_name)))
2145
      continue
2146

    
2147
    result = rdev.PauseResumeSync(pause)
2148

    
2149
    if result:
2150
      success.append((result, None))
2151
    else:
2152
      if pause:
2153
        msg = "Pause"
2154
      else:
2155
        msg = "Resume"
2156
      success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
2157

    
2158
  return success
2159

    
2160

    
2161
def BlockdevRemove(disk):
2162
  """Remove a block device.
2163

2164
  @note: This is intended to be called recursively.
2165

2166
  @type disk: L{objects.Disk}
2167
  @param disk: the disk object we should remove
2168
  @rtype: boolean
2169
  @return: the success of the operation
2170

2171
  """
2172
  msgs = []
2173
  try:
2174
    rdev = _RecursiveFindBD(disk)
2175
  except errors.BlockDeviceError, err:
2176
    # probably can't attach
2177
    logging.info("Can't attach to device %s in remove", disk)
2178
    rdev = None
2179
  if rdev is not None:
2180
    r_path = rdev.dev_path
2181
    try:
2182
      rdev.Remove()
2183
    except errors.BlockDeviceError, err:
2184
      msgs.append(str(err))
2185
    if not msgs:
2186
      DevCacheManager.RemoveCache(r_path)
2187

    
2188
  if disk.children:
2189
    for child in disk.children:
2190
      try:
2191
        BlockdevRemove(child)
2192
      except RPCFail, err:
2193
        msgs.append(str(err))
2194

    
2195
  if msgs:
2196
    _Fail("; ".join(msgs))
2197

    
2198

    
2199
def _RecursiveAssembleBD(disk, owner, as_primary):
2200
  """Activate a block device for an instance.
2201

2202
  This is run on the primary and secondary nodes for an instance.
2203

2204
  @note: this function is called recursively.
2205

2206
  @type disk: L{objects.Disk}
2207
  @param disk: the disk we try to assemble
2208
  @type owner: str
2209
  @param owner: the name of the instance which owns the disk
2210
  @type as_primary: boolean
2211
  @param as_primary: if we should make the block device
2212
      read/write
2213

2214
  @return: the assembled device or None (in case no device
2215
      was assembled)
2216
  @raise errors.BlockDeviceError: in case there is an error
2217
      during the activation of the children or the device
2218
      itself
2219

2220
  """
2221
  children = []
2222
  if disk.children:
2223
    mcn = disk.ChildrenNeeded()
2224
    if mcn == -1:
2225
      mcn = 0 # max number of Nones allowed
2226
    else:
2227
      mcn = len(disk.children) - mcn # max number of Nones
2228
    for chld_disk in disk.children:
2229
      try:
2230
        cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
2231
      except errors.BlockDeviceError, err:
2232
        if children.count(None) >= mcn:
2233
          raise
2234
        cdev = None
2235
        logging.error("Error in child activation (but continuing): %s",
2236
                      str(err))
2237
      children.append(cdev)
2238

    
2239
  if as_primary or disk.AssembleOnSecondary():
2240
    r_dev = bdev.Assemble(disk, children)
2241
    result = r_dev
2242
    if as_primary or disk.OpenOnSecondary():
2243
      r_dev.Open()
2244
    DevCacheManager.UpdateCache(r_dev.dev_path, owner,
2245
                                as_primary, disk.iv_name)
2246

    
2247
  else:
2248
    result = True
2249
  return result
2250

    
2251

    
2252
def BlockdevAssemble(disk, owner, as_primary, idx):
2253
  """Activate a block device for an instance.
2254

2255
  This is a wrapper over _RecursiveAssembleBD.
2256

2257
  @rtype: str or boolean
2258
  @return: a tuple with the C{/dev/...} path and the created symlink
2259
      for primary nodes, and (C{True}, C{True}) for secondary nodes
2260

2261
  """
2262
  try:
2263
    result = _RecursiveAssembleBD(disk, owner, as_primary)
2264
    if isinstance(result, BlockDev):
2265
      # pylint: disable=E1103
2266
      dev_path = result.dev_path
2267
      link_name = None
2268
      if as_primary:
2269
        link_name = _SymlinkBlockDev(owner, dev_path, idx)
2270
    elif result:
2271
      return result, result
2272
    else:
2273
      _Fail("Unexpected result from _RecursiveAssembleBD")
2274
  except errors.BlockDeviceError, err:
2275
    _Fail("Error while assembling disk: %s", err, exc=True)
2276
  except OSError, err:
2277
    _Fail("Error while symlinking disk: %s", err, exc=True)
2278

    
2279
  return dev_path, link_name
2280

    
2281

    
2282
def BlockdevShutdown(disk):
2283
  """Shut down a block device.
2284

2285
  First, if the device is assembled (Attach() is successful), then
2286
  the device is shutdown. Then the children of the device are
2287
  shutdown.
2288

2289
  This function is called recursively. Note that we don't cache the
2290
  children or such, as oppossed to assemble, shutdown of different
2291
  devices doesn't require that the upper device was active.
2292

2293
  @type disk: L{objects.Disk}
2294
  @param disk: the description of the disk we should
2295
      shutdown
2296
  @rtype: None
2297

2298
  """
2299
  msgs = []
2300
  r_dev = _RecursiveFindBD(disk)
2301
  if r_dev is not None:
2302
    r_path = r_dev.dev_path
2303
    try:
2304
      r_dev.Shutdown()
2305
      DevCacheManager.RemoveCache(r_path)
2306
    except errors.BlockDeviceError, err:
2307
      msgs.append(str(err))
2308

    
2309
  if disk.children:
2310
    for child in disk.children:
2311
      try:
2312
        BlockdevShutdown(child)
2313
      except RPCFail, err:
2314
        msgs.append(str(err))
2315

    
2316
  if msgs:
2317
    _Fail("; ".join(msgs))
2318

    
2319

    
2320
def BlockdevAddchildren(parent_cdev, new_cdevs):
2321
  """Extend a mirrored block device.
2322

2323
  @type parent_cdev: L{objects.Disk}
2324
  @param parent_cdev: the disk to which we should add children
2325
  @type new_cdevs: list of L{objects.Disk}
2326
  @param new_cdevs: the list of children which we should add
2327
  @rtype: None
2328

2329
  """
2330
  parent_bdev = _RecursiveFindBD(parent_cdev)
2331
  if parent_bdev is None:
2332
    _Fail("Can't find parent device '%s' in add children", parent_cdev)
2333
  new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
2334
  if new_bdevs.count(None) > 0:
2335
    _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
2336
  parent_bdev.AddChildren(new_bdevs)
2337

    
2338

    
2339
def BlockdevRemovechildren(parent_cdev, new_cdevs):
2340
  """Shrink a mirrored block device.
2341

2342
  @type parent_cdev: L{objects.Disk}
2343
  @param parent_cdev: the disk from which we should remove children
2344
  @type new_cdevs: list of L{objects.Disk}
2345
  @param new_cdevs: the list of children which we should remove
2346
  @rtype: None
2347

2348
  """
2349
  parent_bdev = _RecursiveFindBD(parent_cdev)
2350
  if parent_bdev is None:
2351
    _Fail("Can't find parent device '%s' in remove children", parent_cdev)
2352
  devs = []
2353
  for disk in new_cdevs:
2354
    rpath = disk.StaticDevPath()
2355
    if rpath is None:
2356
      bd = _RecursiveFindBD(disk)
2357
      if bd is None:
2358
        _Fail("Can't find device %s while removing children", disk)
2359
      else:
2360
        devs.append(bd.dev_path)
2361
    else:
2362
      if not utils.IsNormAbsPath(rpath):
2363
        _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
2364
      devs.append(rpath)
2365
  parent_bdev.RemoveChildren(devs)
2366

    
2367

    
2368
def BlockdevGetmirrorstatus(disks):
2369
  """Get the mirroring status of a list of devices.
2370

2371
  @type disks: list of L{objects.Disk}
2372
  @param disks: the list of disks which we should query
2373
  @rtype: disk
2374
  @return: List of L{objects.BlockDevStatus}, one for each disk
2375
  @raise errors.BlockDeviceError: if any of the disks cannot be
2376
      found
2377

2378
  """
2379
  stats = []
2380
  for dsk in disks:
2381
    rbd = _RecursiveFindBD(dsk)
2382
    if rbd is None:
2383
      _Fail("Can't find device %s", dsk)
2384

    
2385
    stats.append(rbd.CombinedSyncStatus())
2386

    
2387
  return stats
2388

    
2389

    
2390
def BlockdevGetmirrorstatusMulti(disks):
2391
  """Get the mirroring status of a list of devices.
2392

2393
  @type disks: list of L{objects.Disk}
2394
  @param disks: the list of disks which we should query
2395
  @rtype: disk
2396
  @return: List of tuples, (bool, status), one for each disk; bool denotes
2397
    success/failure, status is L{objects.BlockDevStatus} on success, string
2398
    otherwise
2399

2400
  """
2401
  result = []
2402
  for disk in disks:
2403
    try:
2404
      rbd = _RecursiveFindBD(disk)
2405
      if rbd is None:
2406
        result.append((False, "Can't find device %s" % disk))
2407
        continue
2408

    
2409
      status = rbd.CombinedSyncStatus()
2410
    except errors.BlockDeviceError, err:
2411
      logging.exception("Error while getting disk status")
2412
      result.append((False, str(err)))
2413
    else:
2414
      result.append((True, status))
2415

    
2416
  assert len(disks) == len(result)
2417

    
2418
  return result
2419

    
2420

    
2421
def _RecursiveFindBD(disk):
2422
  """Check if a device is activated.
2423

2424
  If so, return information about the real device.
2425

2426
  @type disk: L{objects.Disk}
2427
  @param disk: the disk object we need to find
2428

2429
  @return: None if the device can't be found,
2430
      otherwise the device instance
2431

2432
  """
2433
  children = []
2434
  if disk.children:
2435
    for chdisk in disk.children:
2436
      children.append(_RecursiveFindBD(chdisk))
2437

    
2438
  return bdev.FindDevice(disk, children)
2439

    
2440

    
2441
def _OpenRealBD(disk):
2442
  """Opens the underlying block device of a disk.
2443

2444
  @type disk: L{objects.Disk}
2445
  @param disk: the disk object we want to open
2446

2447
  """
2448
  real_disk = _RecursiveFindBD(disk)
2449
  if real_disk is None:
2450
    _Fail("Block device '%s' is not set up", disk)
2451

    
2452
  real_disk.Open()
2453

    
2454
  return real_disk
2455

    
2456

    
2457
def BlockdevFind(disk):
2458
  """Check if a device is activated.
2459

2460
  If it is, return information about the real device.
2461

2462
  @type disk: L{objects.Disk}
2463
  @param disk: the disk to find
2464
  @rtype: None or objects.BlockDevStatus
2465
  @return: None if the disk cannot be found, otherwise a the current
2466
           information
2467

2468
  """
2469
  try:
2470
    rbd = _RecursiveFindBD(disk)
2471
  except errors.BlockDeviceError, err:
2472
    _Fail("Failed to find device: %s", err, exc=True)
2473

    
2474
  if rbd is None:
2475
    return None
2476

    
2477
  return rbd.GetSyncStatus()
2478

    
2479

    
2480
def BlockdevGetdimensions(disks):
2481
  """Computes the size of the given disks.
2482

2483
  If a disk is not found, returns None instead.
2484

2485
  @type disks: list of L{objects.Disk}
2486
  @param disks: the list of disk to compute the size for
2487
  @rtype: list
2488
  @return: list with elements None if the disk cannot be found,
2489
      otherwise the pair (size, spindles), where spindles is None if the
2490
      device doesn't support that
2491

2492
  """
2493
  result = []
2494
  for cf in disks:
2495
    try:
2496
      rbd = _RecursiveFindBD(cf)
2497
    except errors.BlockDeviceError:
2498
      result.append(None)
2499
      continue
2500
    if rbd is None:
2501
      result.append(None)
2502
    else:
2503
      result.append(rbd.GetActualDimensions())
2504
  return result
2505

    
2506

    
2507
def BlockdevExport(disk, dest_node_ip, dest_path, cluster_name):
2508
  """Export a block device to a remote node.
2509

2510
  @type disk: L{objects.Disk}
2511
  @param disk: the description of the disk to export
2512
  @type dest_node_ip: str
2513
  @param dest_node_ip: the destination node IP to export to
2514
  @type dest_path: str
2515
  @param dest_path: the destination path on the target node
2516
  @type cluster_name: str
2517
  @param cluster_name: the cluster name, needed for SSH hostalias
2518
  @rtype: None
2519

2520
  """
2521
  real_disk = _OpenRealBD(disk)
2522

    
2523
  # the block size on the read dd is 1MiB to match our units
2524
  expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
2525
                               "dd if=%s bs=1048576 count=%s",
2526
                               real_disk.dev_path, str(disk.size))
2527

    
2528
  # we set here a smaller block size as, due to ssh buffering, more
2529
  # than 64-128k will mostly ignored; we use nocreat to fail if the
2530
  # device is not already there or we pass a wrong path; we use
2531
  # notrunc to no attempt truncate on an LV device; we use oflag=dsync
2532
  # to not buffer too much memory; this means that at best, we flush
2533
  # every 64k, which will not be very fast
2534
  destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
2535
                                " oflag=dsync", dest_path)
2536

    
2537
  remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node_ip,
2538
                                                   constants.SSH_LOGIN_USER,
2539
                                                   destcmd)
2540

    
2541
  # all commands have been checked, so we're safe to combine them
2542
  command = "|".join([expcmd, utils.ShellQuoteArgs(remotecmd)])
2543

    
2544
  result = utils.RunCmd(["bash", "-c", command])
2545

    
2546
  if result.failed:
2547
    _Fail("Disk copy command '%s' returned error: %s"
2548
          " output: %s", command, result.fail_reason, result.output)
2549

    
2550

    
2551
def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
2552
  """Write a file to the filesystem.
2553

2554
  This allows the master to overwrite(!) a file. It will only perform
2555
  the operation if the file belongs to a list of configuration files.
2556

2557
  @type file_name: str
2558
  @param file_name: the target file name
2559
  @type data: str
2560
  @param data: the new contents of the file
2561
  @type mode: int
2562
  @param mode: the mode to give the file (can be None)
2563
  @type uid: string
2564
  @param uid: the owner of the file
2565
  @type gid: string
2566
  @param gid: the group of the file
2567
  @type atime: float
2568
  @param atime: the atime to set on the file (can be None)
2569
  @type mtime: float
2570
  @param mtime: the mtime to set on the file (can be None)
2571
  @rtype: None
2572

2573
  """
2574
  file_name = vcluster.LocalizeVirtualPath(file_name)
2575

    
2576
  if not os.path.isabs(file_name):
2577
    _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
2578

    
2579
  if file_name not in _ALLOWED_UPLOAD_FILES:
2580
    _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
2581
          file_name)
2582

    
2583
  raw_data = _Decompress(data)
2584

    
2585
  if not (isinstance(uid, basestring) and isinstance(gid, basestring)):
2586
    _Fail("Invalid username/groupname type")
2587

    
2588
  getents = runtime.GetEnts()
2589
  uid = getents.LookupUser(uid)
2590
  gid = getents.LookupGroup(gid)
2591

    
2592
  utils.SafeWriteFile(file_name, None,
2593
                      data=raw_data, mode=mode, uid=uid, gid=gid,
2594
                      atime=atime, mtime=mtime)
2595

    
2596

    
2597
def RunOob(oob_program, command, node, timeout):
2598
  """Executes oob_program with given command on given node.
2599

2600
  @param oob_program: The path to the executable oob_program
2601
  @param command: The command to invoke on oob_program
2602
  @param node: The node given as an argument to the program
2603
  @param timeout: Timeout after which we kill the oob program
2604

2605
  @return: stdout
2606
  @raise RPCFail: If execution fails for some reason
2607

2608
  """
2609
  result = utils.RunCmd([oob_program, command, node], timeout=timeout)
2610

    
2611
  if result.failed:
2612
    _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
2613
          result.fail_reason, result.output)
2614

    
2615
  return result.stdout
2616

    
2617

    
2618
def _OSOndiskAPIVersion(os_dir):
2619
  """Compute and return the API version of a given OS.
2620

2621
  This function will try to read the API version of the OS residing in
2622
  the 'os_dir' directory.
2623

2624
  @type os_dir: str
2625
  @param os_dir: the directory in which we should look for the OS
2626
  @rtype: tuple
2627
  @return: tuple (status, data) with status denoting the validity and
2628
      data holding either the vaid versions or an error message
2629

2630
  """
2631
  api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
2632

    
2633
  try:
2634
    st = os.stat(api_file)
2635
  except EnvironmentError, err:
2636
    return False, ("Required file '%s' not found under path %s: %s" %
2637
                   (constants.OS_API_FILE, os_dir, utils.ErrnoOrStr(err)))
2638

    
2639
  if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2640
    return False, ("File '%s' in %s is not a regular file" %
2641
                   (constants.OS_API_FILE, os_dir))
2642

    
2643
  try:
2644
    api_versions = utils.ReadFile(api_file).splitlines()
2645
  except EnvironmentError, err:
2646
    return False, ("Error while reading the API version file at %s: %s" %
2647
                   (api_file, utils.ErrnoOrStr(err)))
2648

    
2649
  try:
2650
    api_versions = [int(version.strip()) for version in api_versions]
2651
  except (TypeError, ValueError), err:
2652
    return False, ("API version(s) can't be converted to integer: %s" %
2653
                   str(err))
2654

    
2655
  return True, api_versions
2656

    
2657

    
2658
def DiagnoseOS(top_dirs=None):
2659
  """Compute the validity for all OSes.
2660

2661
  @type top_dirs: list
2662
  @param top_dirs: the list of directories in which to
2663
      search (if not given defaults to
2664
      L{pathutils.OS_SEARCH_PATH})
2665
  @rtype: list of L{objects.OS}
2666
  @return: a list of tuples (name, path, status, diagnose, variants,
2667
      parameters, api_version) for all (potential) OSes under all
2668
      search paths, where:
2669
          - name is the (potential) OS name
2670
          - path is the full path to the OS
2671
          - status True/False is the validity of the OS
2672
          - diagnose is the error message for an invalid OS, otherwise empty
2673
          - variants is a list of supported OS variants, if any
2674
          - parameters is a list of (name, help) parameters, if any
2675
          - api_version is a list of support OS API versions
2676

2677
  """
2678
  if top_dirs is None:
2679
    top_dirs = pathutils.OS_SEARCH_PATH
2680

    
2681
  result = []
2682
  for dir_name in top_dirs:
2683
    if os.path.isdir(dir_name):
2684
      try:
2685
        f_names = utils.ListVisibleFiles(dir_name)
2686
      except EnvironmentError, err:
2687
        logging.exception("Can't list the OS directory %s: %s", dir_name, err)
2688
        break
2689
      for name in f_names:
2690
        os_path = utils.PathJoin(dir_name, name)
2691
        status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
2692
        if status:
2693
          diagnose = ""
2694
          variants = os_inst.supported_variants
2695
          parameters = os_inst.supported_parameters
2696
          api_versions = os_inst.api_versions
2697
        else:
2698
          diagnose = os_inst
2699
          variants = parameters = api_versions = []
2700
        result.append((name, os_path, status, diagnose, variants,
2701
                       parameters, api_versions))
2702

    
2703
  return result
2704

    
2705

    
2706
def _TryOSFromDisk(name, base_dir=None):
2707
  """Create an OS instance from disk.
2708

2709
  This function will return an OS instance if the given name is a
2710
  valid OS name.
2711

2712
  @type base_dir: string
2713
  @keyword base_dir: Base directory containing OS installations.
2714
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2715
  @rtype: tuple
2716
  @return: success and either the OS instance if we find a valid one,
2717
      or error message
2718

2719
  """
2720
  if base_dir is None:
2721
    os_dir = utils.FindFile(name, pathutils.OS_SEARCH_PATH, os.path.isdir)
2722
  else:
2723
    os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2724

    
2725
  if os_dir is None:
2726
    return False, "Directory for OS %s not found in search path" % name
2727

    
2728
  status, api_versions = _OSOndiskAPIVersion(os_dir)
2729
  if not status:
2730
    # push the error up
2731
    return status, api_versions
2732

    
2733
  if not constants.OS_API_VERSIONS.intersection(api_versions):
2734
    return False, ("API version mismatch for path '%s': found %s, want %s." %
2735
                   (os_dir, api_versions, constants.OS_API_VERSIONS))
2736

    
2737
  # OS Files dictionary, we will populate it with the absolute path
2738
  # names; if the value is True, then it is a required file, otherwise
2739
  # an optional one
2740
  os_files = dict.fromkeys(constants.OS_SCRIPTS, True)
2741

    
2742
  if max(api_versions) >= constants.OS_API_V15:
2743
    os_files[constants.OS_VARIANTS_FILE] = False
2744

    
2745
  if max(api_versions) >= constants.OS_API_V20:
2746
    os_files[constants.OS_PARAMETERS_FILE] = True
2747
  else:
2748
    del os_files[constants.OS_SCRIPT_VERIFY]
2749

    
2750
  for (filename, required) in os_files.items():
2751
    os_files[filename] = utils.PathJoin(os_dir, filename)
2752

    
2753
    try:
2754
      st = os.stat(os_files[filename])
2755
    except EnvironmentError, err:
2756
      if err.errno == errno.ENOENT and not required:
2757
        del os_files[filename]
2758
        continue
2759
      return False, ("File '%s' under path '%s' is missing (%s)" %
2760
                     (filename, os_dir, utils.ErrnoOrStr(err)))
2761

    
2762
    if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2763
      return False, ("File '%s' under path '%s' is not a regular file" %
2764
                     (filename, os_dir))
2765

    
2766
    if filename in constants.OS_SCRIPTS:
2767
      if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2768
        return False, ("File '%s' under path '%s' is not executable" %
2769
                       (filename, os_dir))
2770

    
2771
  variants = []
2772
  if constants.OS_VARIANTS_FILE in os_files:
2773
    variants_file = os_files[constants.OS_VARIANTS_FILE]
2774
    try:
2775
      variants = \
2776
        utils.FilterEmptyLinesAndComments(utils.ReadFile(variants_file))
2777
    except EnvironmentError, err:
2778
      # we accept missing files, but not other errors
2779
      if err.errno != errno.ENOENT:
2780
        return False, ("Error while reading the OS variants file at %s: %s" %
2781
                       (variants_file, utils.ErrnoOrStr(err)))
2782

    
2783
  parameters = []
2784
  if constants.OS_PARAMETERS_FILE in os_files:
2785
    parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2786
    try:
2787
      parameters = utils.ReadFile(parameters_file).splitlines()
2788
    except EnvironmentError, err:
2789
      return False, ("Error while reading the OS parameters file at %s: %s" %
2790
                     (parameters_file, utils.ErrnoOrStr(err)))
2791
    parameters = [v.split(None, 1) for v in parameters]
2792

    
2793
  os_obj = objects.OS(name=name, path=os_dir,
2794
                      create_script=os_files[constants.OS_SCRIPT_CREATE],
2795
                      export_script=os_files[constants.OS_SCRIPT_EXPORT],
2796
                      import_script=os_files[constants.OS_SCRIPT_IMPORT],
2797
                      rename_script=os_files[constants.OS_SCRIPT_RENAME],
2798
                      verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2799
                                                 None),
2800
                      supported_variants=variants,
2801
                      supported_parameters=parameters,
2802
                      api_versions=api_versions)
2803
  return True, os_obj
2804

    
2805

    
2806
def OSFromDisk(name, base_dir=None):
2807
  """Create an OS instance from disk.
2808

2809
  This function will return an OS instance if the given name is a
2810
  valid OS name. Otherwise, it will raise an appropriate
2811
  L{RPCFail} exception, detailing why this is not a valid OS.
2812

2813
  This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2814
  an exception but returns true/false status data.
2815

2816
  @type base_dir: string
2817
  @keyword base_dir: Base directory containing OS installations.
2818
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2819
  @rtype: L{objects.OS}
2820
  @return: the OS instance if we find a valid one
2821
  @raise RPCFail: if we don't find a valid OS
2822

2823
  """
2824
  name_only = objects.OS.GetName(name)
2825
  status, payload = _TryOSFromDisk(name_only, base_dir)
2826

    
2827
  if not status:
2828
    _Fail(payload)
2829

    
2830
  return payload
2831

    
2832

    
2833
def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2834
  """Calculate the basic environment for an os script.
2835

2836
  @type os_name: str
2837
  @param os_name: full operating system name (including variant)
2838
  @type inst_os: L{objects.OS}
2839
  @param inst_os: operating system for which the environment is being built
2840
  @type os_params: dict
2841
  @param os_params: the OS parameters
2842
  @type debug: integer
2843
  @param debug: debug level (0 or 1, for OS Api 10)
2844
  @rtype: dict
2845
  @return: dict of environment variables
2846
  @raise errors.BlockDeviceError: if the block device
2847
      cannot be found
2848

2849
  """
2850
  result = {}
2851
  api_version = \
2852
    max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2853
  result["OS_API_VERSION"] = "%d" % api_version
2854
  result["OS_NAME"] = inst_os.name
2855
  result["DEBUG_LEVEL"] = "%d" % debug
2856

    
2857
  # OS variants
2858
  if api_version >= constants.OS_API_V15 and inst_os.supported_variants:
2859
    variant = objects.OS.GetVariant(os_name)
2860
    if not variant:
2861
      variant = inst_os.supported_variants[0]
2862
  else:
2863
    variant = ""
2864
  result["OS_VARIANT"] = variant
2865

    
2866
  # OS params
2867
  for pname, pvalue in os_params.items():
2868
    result["OSP_%s" % pname.upper()] = pvalue
2869

    
2870
  # Set a default path otherwise programs called by OS scripts (or
2871
  # even hooks called from OS scripts) might break, and we don't want
2872
  # to have each script require setting a PATH variable
2873
  result["PATH"] = constants.HOOKS_PATH
2874

    
2875
  return result
2876

    
2877

    
2878
def OSEnvironment(instance, inst_os, debug=0):
2879
  """Calculate the environment for an os script.
2880

2881
  @type instance: L{objects.Instance}
2882
  @param instance: target instance for the os script run
2883
  @type inst_os: L{objects.OS}
2884
  @param inst_os: operating system for which the environment is being built
2885
  @type debug: integer
2886
  @param debug: debug level (0 or 1, for OS Api 10)
2887
  @rtype: dict
2888
  @return: dict of environment variables
2889
  @raise errors.BlockDeviceError: if the block device
2890
      cannot be found
2891

2892
  """
2893
  result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2894

    
2895
  for attr in ["name", "os", "uuid", "ctime", "mtime", "primary_node"]:
2896
    result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2897

    
2898
  result["HYPERVISOR"] = instance.hypervisor
2899
  result["DISK_COUNT"] = "%d" % len(instance.disks)
2900
  result["NIC_COUNT"] = "%d" % len(instance.nics)
2901
  result["INSTANCE_SECONDARY_NODES"] = \
2902
      ("%s" % " ".join(instance.secondary_nodes))
2903

    
2904
  # Disks
2905
  for idx, disk in enumerate(instance.disks):
2906
    real_disk = _OpenRealBD(disk)
2907
    result["DISK_%d_PATH" % idx] = real_disk.dev_path
2908
    result["DISK_%d_ACCESS" % idx] = disk.mode
2909
    result["DISK_%d_UUID" % idx] = disk.uuid
2910
    if disk.name:
2911
      result["DISK_%d_NAME" % idx] = disk.name
2912
    if constants.HV_DISK_TYPE in instance.hvparams:
2913
      result["DISK_%d_FRONTEND_TYPE" % idx] = \
2914
        instance.hvparams[constants.HV_DISK_TYPE]
2915
    if disk.dev_type in constants.DTS_BLOCK:
2916
      result["DISK_%d_BACKEND_TYPE" % idx] = "block"
2917
    elif disk.dev_type in [constants.DT_FILE, constants.DT_SHARED_FILE]:
2918
      result["DISK_%d_BACKEND_TYPE" % idx] = \
2919
        "file:%s" % disk.logical_id[0]
2920

    
2921
  # NICs
2922
  for idx, nic in enumerate(instance.nics):
2923
    result["NIC_%d_MAC" % idx] = nic.mac
2924
    result["NIC_%d_UUID" % idx] = nic.uuid
2925
    if nic.name:
2926
      result["NIC_%d_NAME" % idx] = nic.name
2927
    if nic.ip:
2928
      result["NIC_%d_IP" % idx] = nic.ip
2929
    result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE]
2930
    if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2931
      result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK]
2932
    if nic.nicparams[constants.NIC_LINK]:
2933
      result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK]
2934
    if nic.netinfo:
2935
      nobj = objects.Network.FromDict(nic.netinfo)
2936
      result.update(nobj.HooksDict("NIC_%d_" % idx))
2937
    if constants.HV_NIC_TYPE in instance.hvparams:
2938
      result["NIC_%d_FRONTEND_TYPE" % idx] = \
2939
        instance.hvparams[constants.HV_NIC_TYPE]
2940

    
2941
  # HV/BE params
2942
  for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2943
    for key, value in source.items():
2944
      result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2945

    
2946
  return result
2947

    
2948

    
2949
def DiagnoseExtStorage(top_dirs=None):
2950
  """Compute the validity for all ExtStorage Providers.
2951

2952
  @type top_dirs: list
2953
  @param top_dirs: the list of directories in which to
2954
      search (if not given defaults to
2955
      L{pathutils.ES_SEARCH_PATH})
2956
  @rtype: list of L{objects.ExtStorage}
2957
  @return: a list of tuples (name, path, status, diagnose, parameters)
2958
      for all (potential) ExtStorage Providers under all
2959
      search paths, where:
2960
          - name is the (potential) ExtStorage Provider
2961
          - path is the full path to the ExtStorage Provider
2962
          - status True/False is the validity of the ExtStorage Provider
2963
          - diagnose is the error message for an invalid ExtStorage Provider,
2964
            otherwise empty
2965
          - parameters is a list of (name, help) parameters, if any
2966

2967
  """
2968
  if top_dirs is None:
2969
    top_dirs = pathutils.ES_SEARCH_PATH
2970

    
2971
  result = []
2972
  for dir_name in top_dirs:
2973
    if os.path.isdir(dir_name):
2974
      try:
2975
        f_names = utils.ListVisibleFiles(dir_name)
2976
      except EnvironmentError, err:
2977
        logging.exception("Can't list the ExtStorage directory %s: %s",
2978
                          dir_name, err)
2979
        break
2980
      for name in f_names:
2981
        es_path = utils.PathJoin(dir_name, name)
2982
        status, es_inst = bdev.ExtStorageFromDisk(name, base_dir=dir_name)
2983
        if status:
2984
          diagnose = ""
2985
          parameters = es_inst.supported_parameters
2986
        else:
2987
          diagnose = es_inst
2988
          parameters = []
2989
        result.append((name, es_path, status, diagnose, parameters))
2990

    
2991
  return result
2992

    
2993

    
2994
def BlockdevGrow(disk, amount, dryrun, backingstore, excl_stor):
2995
  """Grow a stack of block devices.
2996

2997
  This function is called recursively, with the childrens being the
2998
  first ones to resize.
2999

3000
  @type disk: L{objects.Disk}
3001
  @param disk: the disk to be grown
3002
  @type amount: integer
3003
  @param amount: the amount (in mebibytes) to grow with
3004
  @type dryrun: boolean
3005
  @param dryrun: whether to execute the operation in simulation mode
3006
      only, without actually increasing the size
3007
  @param backingstore: whether to execute the operation on backing storage
3008
      only, or on "logical" storage only; e.g. DRBD is logical storage,
3009
      whereas LVM, file, RBD are backing storage
3010
  @rtype: (status, result)
3011
  @type excl_stor: boolean
3012
  @param excl_stor: Whether exclusive_storage is active
3013
  @return: a tuple with the status of the operation (True/False), and
3014
      the errors message if status is False
3015

3016
  """
3017
  r_dev = _RecursiveFindBD(disk)
3018
  if r_dev is None:
3019
    _Fail("Cannot find block device %s", disk)
3020

    
3021
  try:
3022
    r_dev.Grow(amount, dryrun, backingstore, excl_stor)
3023
  except errors.BlockDeviceError, err:
3024
    _Fail("Failed to grow block device: %s", err, exc=True)
3025

    
3026

    
3027
def BlockdevSnapshot(disk):
3028
  """Create a snapshot copy of a block device.
3029

3030
  This function is called recursively, and the snapshot is actually created
3031
  just for the leaf lvm backend device.
3032

3033
  @type disk: L{objects.Disk}
3034
  @param disk: the disk to be snapshotted
3035
  @rtype: string
3036
  @return: snapshot disk ID as (vg, lv)
3037

3038
  """
3039
  if disk.dev_type == constants.DT_DRBD8:
3040
    if not disk.children:
3041
      _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
3042
            disk.unique_id)
3043
    return BlockdevSnapshot(disk.children[0])
3044
  elif disk.dev_type == constants.DT_PLAIN:
3045
    r_dev = _RecursiveFindBD(disk)
3046
    if r_dev is not None:
3047
      # FIXME: choose a saner value for the snapshot size
3048
      # let's stay on the safe side and ask for the full size, for now
3049
      return r_dev.Snapshot(disk.size)
3050
    else:
3051
      _Fail("Cannot find block device %s", disk)
3052
  else:
3053
    _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
3054
          disk.unique_id, disk.dev_type)
3055

    
3056

    
3057
def BlockdevSetInfo(disk, info):
3058
  """Sets 'metadata' information on block devices.
3059

3060
  This function sets 'info' metadata on block devices. Initial
3061
  information is set at device creation; this function should be used
3062
  for example after renames.
3063

3064
  @type disk: L{objects.Disk}
3065
  @param disk: the disk to be grown
3066
  @type info: string
3067
  @param info: new 'info' metadata
3068
  @rtype: (status, result)
3069
  @return: a tuple with the status of the operation (True/False), and
3070
      the errors message if status is False
3071

3072
  """
3073
  r_dev = _RecursiveFindBD(disk)
3074
  if r_dev is None:
3075
    _Fail("Cannot find block device %s", disk)
3076

    
3077
  try:
3078
    r_dev.SetInfo(info)
3079
  except errors.BlockDeviceError, err:
3080
    _Fail("Failed to set information on block device: %s", err, exc=True)
3081

    
3082

    
3083
def FinalizeExport(instance, snap_disks):
3084
  """Write out the export configuration information.
3085

3086
  @type instance: L{objects.Instance}
3087
  @param instance: the instance which we export, used for
3088
      saving configuration
3089
  @type snap_disks: list of L{objects.Disk}
3090
  @param snap_disks: list of snapshot block devices, which
3091
      will be used to get the actual name of the dump file
3092

3093
  @rtype: None
3094

3095
  """
3096
  destdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name + ".new")
3097
  finaldestdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name)
3098

    
3099
  config = objects.SerializableConfigParser()
3100

    
3101
  config.add_section(constants.INISECT_EXP)
3102
  config.set(constants.INISECT_EXP, "version", "0")
3103
  config.set(constants.INISECT_EXP, "timestamp", "%d" % int(time.time()))
3104
  config.set(constants.INISECT_EXP, "source", instance.primary_node)
3105
  config.set(constants.INISECT_EXP, "os", instance.os)
3106
  config.set(constants.INISECT_EXP, "compression", "none")
3107

    
3108
  config.add_section(constants.INISECT_INS)
3109
  config.set(constants.INISECT_INS, "name", instance.name)
3110
  config.set(constants.INISECT_INS, "maxmem", "%d" %
3111
             instance.beparams[constants.BE_MAXMEM])
3112
  config.set(constants.INISECT_INS, "minmem", "%d" %
3113
             instance.beparams[constants.BE_MINMEM])
3114
  # "memory" is deprecated, but useful for exporting to old ganeti versions
3115
  config.set(constants.INISECT_INS, "memory", "%d" %
3116
             instance.beparams[constants.BE_MAXMEM])
3117
  config.set(constants.INISECT_INS, "vcpus", "%d" %
3118
             instance.beparams[constants.BE_VCPUS])
3119
  config.set(constants.INISECT_INS, "disk_template", instance.disk_template)
3120
  config.set(constants.INISECT_INS, "hypervisor", instance.hypervisor)
3121
  config.set(constants.INISECT_INS, "tags", " ".join(instance.GetTags()))
3122

    
3123
  nic_total = 0
3124
  for nic_count, nic in enumerate(instance.nics):
3125
    nic_total += 1
3126
    config.set(constants.INISECT_INS, "nic%d_mac" %
3127
               nic_count, "%s" % nic.mac)
3128
    config.set(constants.INISECT_INS, "nic%d_ip" % nic_count, "%s" % nic.ip)
3129
    config.set(constants.INISECT_INS, "nic%d_network" % nic_count,
3130
               "%s" % nic.network)
3131
    for param in constants.NICS_PARAMETER_TYPES:
3132
      config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param),
3133
                 "%s" % nic.nicparams.get(param, None))
3134
  # TODO: redundant: on load can read nics until it doesn't exist
3135
  config.set(constants.INISECT_INS, "nic_count", "%d" % nic_total)
3136

    
3137
  disk_total = 0
3138
  for disk_count, disk in enumerate(snap_disks):
3139
    if disk:
3140
      disk_total += 1
3141
      config.set(constants.INISECT_INS, "disk%d_ivname" % disk_count,
3142
                 ("%s" % disk.iv_name))
3143
      config.set(constants.INISECT_INS, "disk%d_dump" % disk_count,
3144
                 ("%s" % disk.logical_id[1]))
3145
      config.set(constants.INISECT_INS, "disk%d_size" % disk_count,
3146
                 ("%d" % disk.size))
3147

    
3148
  config.set(constants.INISECT_INS, "disk_count", "%d" % disk_total)
3149

    
3150
  # New-style hypervisor/backend parameters
3151

    
3152
  config.add_section(constants.INISECT_HYP)
3153
  for name, value in instance.hvparams.items():
3154
    if name not in constants.HVC_GLOBALS:
3155
      config.set(constants.INISECT_HYP, name, str(value))
3156

    
3157
  config.add_section(constants.INISECT_BEP)
3158
  for name, value in instance.beparams.items():
3159
    config.set(constants.INISECT_BEP, name, str(value))
3160

    
3161
  config.add_section(constants.INISECT_OSP)
3162
  for name, value in instance.osparams.items():
3163
    config.set(constants.INISECT_OSP, name, str(value))
3164

    
3165
  utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
3166
                  data=config.Dumps())
3167
  shutil.rmtree(finaldestdir, ignore_errors=True)
3168
  shutil.move(destdir, finaldestdir)
3169

    
3170

    
3171
def ExportInfo(dest):
3172
  """Get export configuration information.
3173

3174
  @type dest: str
3175
  @param dest: directory containing the export
3176

3177
  @rtype: L{objects.SerializableConfigParser}
3178
  @return: a serializable config file containing the
3179
      export info
3180

3181
  """
3182
  cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
3183

    
3184
  config = objects.SerializableConfigParser()
3185
  config.read(cff)
3186

    
3187
  if (not config.has_section(constants.INISECT_EXP) or
3188
      not config.has_section(constants.INISECT_INS)):
3189
    _Fail("Export info file doesn't have the required fields")
3190

    
3191
  return config.Dumps()
3192

    
3193

    
3194
def ListExports():
3195
  """Return a list of exports currently available on this machine.
3196

3197
  @rtype: list
3198
  @return: list of the exports
3199

3200
  """
3201
  if os.path.isdir(pathutils.EXPORT_DIR):
3202
    return sorted(utils.ListVisibleFiles(pathutils.EXPORT_DIR))
3203
  else:
3204
    _Fail("No exports directory")
3205

    
3206

    
3207
def RemoveExport(export):
3208
  """Remove an existing export from the node.
3209

3210
  @type export: str
3211
  @param export: the name of the export to remove
3212
  @rtype: None
3213

3214
  """
3215
  target = utils.PathJoin(pathutils.EXPORT_DIR, export)
3216

    
3217
  try:
3218
    shutil.rmtree(target)
3219
  except EnvironmentError, err:
3220
    _Fail("Error while removing the export: %s", err, exc=True)
3221

    
3222

    
3223
def BlockdevRename(devlist):
3224
  """Rename a list of block devices.
3225

3226
  @type devlist: list of tuples
3227
  @param devlist: list of tuples of the form  (disk, new_unique_id); disk is
3228
      an L{objects.Disk} object describing the current disk, and new
3229
      unique_id is the name we rename it to
3230
  @rtype: boolean
3231
  @return: True if all renames succeeded, False otherwise
3232

3233
  """
3234
  msgs = []
3235
  result = True
3236
  for disk, unique_id in devlist:
3237
    dev = _RecursiveFindBD(disk)
3238
    if dev is None:
3239
      msgs.append("Can't find device %s in rename" % str(disk))
3240
      result = False
3241
      continue
3242
    try:
3243
      old_rpath = dev.dev_path
3244
      dev.Rename(unique_id)
3245
      new_rpath = dev.dev_path
3246
      if old_rpath != new_rpath:
3247
        DevCacheManager.RemoveCache(old_rpath)
3248
        # FIXME: we should add the new cache information here, like:
3249
        # DevCacheManager.UpdateCache(new_rpath, owner, ...)
3250
        # but we don't have the owner here - maybe parse from existing
3251
        # cache? for now, we only lose lvm data when we rename, which
3252
        # is less critical than DRBD or MD
3253
    except errors.BlockDeviceError, err:
3254
      msgs.append("Can't rename device '%s' to '%s': %s" %
3255
                  (dev, unique_id, err))
3256
      logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
3257
      result = False
3258
  if not result:
3259
    _Fail("; ".join(msgs))
3260

    
3261

    
3262
def _TransformFileStorageDir(fs_dir):
3263
  """Checks whether given file_storage_dir is valid.
3264

3265
  Checks wheter the given fs_dir is within the cluster-wide default
3266
  file_storage_dir or the shared_file_storage_dir, which are stored in
3267
  SimpleStore. Only paths under those directories are allowed.
3268

3269
  @type fs_dir: str
3270
  @param fs_dir: the path to check
3271

3272
  @return: the normalized path if valid, None otherwise
3273

3274
  """
3275
  filestorage.CheckFileStoragePath(fs_dir)
3276

    
3277
  return os.path.normpath(fs_dir)
3278

    
3279

    
3280
def CreateFileStorageDir(file_storage_dir):
3281
  """Create file storage directory.
3282

3283
  @type file_storage_dir: str
3284
  @param file_storage_dir: directory to create
3285

3286
  @rtype: tuple
3287
  @return: tuple with first element a boolean indicating wheter dir
3288
      creation was successful or not
3289

3290
  """
3291
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3292
  if os.path.exists(file_storage_dir):
3293
    if not os.path.isdir(file_storage_dir):
3294
      _Fail("Specified storage dir '%s' is not a directory",
3295
            file_storage_dir)
3296
  else:
3297
    try:
3298
      os.makedirs(file_storage_dir, 0750)
3299
    except OSError, err:
3300
      _Fail("Cannot create file storage directory '%s': %s",
3301
            file_storage_dir, err, exc=True)
3302

    
3303

    
3304
def RemoveFileStorageDir(file_storage_dir):
3305
  """Remove file storage directory.
3306

3307
  Remove it only if it's empty. If not log an error and return.
3308

3309
  @type file_storage_dir: str
3310
  @param file_storage_dir: the directory we should cleanup
3311
  @rtype: tuple (success,)
3312
  @return: tuple of one element, C{success}, denoting
3313
      whether the operation was successful
3314

3315
  """
3316
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3317
  if os.path.exists(file_storage_dir):
3318
    if not os.path.isdir(file_storage_dir):
3319
      _Fail("Specified Storage directory '%s' is not a directory",
3320
            file_storage_dir)
3321
    # deletes dir only if empty, otherwise we want to fail the rpc call
3322
    try:
3323
      os.rmdir(file_storage_dir)
3324
    except OSError, err:
3325
      _Fail("Cannot remove file storage directory '%s': %s",
3326
            file_storage_dir, err)
3327

    
3328

    
3329
def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
3330
  """Rename the file storage directory.
3331

3332
  @type old_file_storage_dir: str
3333
  @param old_file_storage_dir: the current path
3334
  @type new_file_storage_dir: str
3335
  @param new_file_storage_dir: the name we should rename to
3336
  @rtype: tuple (success,)
3337
  @return: tuple of one element, C{success}, denoting
3338
      whether the operation was successful
3339

3340
  """
3341
  old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
3342
  new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
3343
  if not os.path.exists(new_file_storage_dir):
3344
    if os.path.isdir(old_file_storage_dir):
3345
      try:
3346
        os.rename(old_file_storage_dir, new_file_storage_dir)
3347
      except OSError, err:
3348
        _Fail("Cannot rename '%s' to '%s': %s",
3349
              old_file_storage_dir, new_file_storage_dir, err)
3350
    else:
3351
      _Fail("Specified storage dir '%s' is not a directory",
3352
            old_file_storage_dir)
3353
  else:
3354
    if os.path.exists(old_file_storage_dir):
3355
      _Fail("Cannot rename '%s' to '%s': both locations exist",
3356
            old_file_storage_dir, new_file_storage_dir)
3357

    
3358

    
3359
def _EnsureJobQueueFile(file_name):
3360
  """Checks whether the given filename is in the queue directory.
3361

3362
  @type file_name: str
3363
  @param file_name: the file name we should check
3364
  @rtype: None
3365
  @raises RPCFail: if the file is not valid
3366

3367
  """
3368
  if not utils.IsBelowDir(pathutils.QUEUE_DIR, file_name):
3369
    _Fail("Passed job queue file '%s' does not belong to"
3370
          " the queue directory '%s'", file_name, pathutils.QUEUE_DIR)
3371

    
3372

    
3373
def JobQueueUpdate(file_name, content):
3374
  """Updates a file in the queue directory.
3375

3376
  This is just a wrapper over L{utils.io.WriteFile}, with proper
3377
  checking.
3378

3379
  @type file_name: str
3380
  @param file_name: the job file name
3381
  @type content: str
3382
  @param content: the new job contents
3383
  @rtype: boolean
3384
  @return: the success of the operation
3385

3386
  """
3387
  file_name = vcluster.LocalizeVirtualPath(file_name)
3388

    
3389
  _EnsureJobQueueFile(file_name)
3390
  getents = runtime.GetEnts()
3391

    
3392
  # Write and replace the file atomically
3393
  utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
3394
                  gid=getents.daemons_gid, mode=constants.JOB_QUEUE_FILES_PERMS)
3395

    
3396

    
3397
def JobQueueRename(old, new):
3398
  """Renames a job queue file.
3399

3400
  This is just a wrapper over os.rename with proper checking.
3401

3402
  @type old: str
3403
  @param old: the old (actual) file name
3404
  @type new: str
3405
  @param new: the desired file name
3406
  @rtype: tuple
3407
  @return: the success of the operation and payload
3408

3409
  """
3410
  old = vcluster.LocalizeVirtualPath(old)
3411
  new = vcluster.LocalizeVirtualPath(new)
3412

    
3413
  _EnsureJobQueueFile(old)
3414
  _EnsureJobQueueFile(new)
3415

    
3416
  getents = runtime.GetEnts()
3417

    
3418
  utils.RenameFile(old, new, mkdir=True, mkdir_mode=0750,
3419
                   dir_uid=getents.masterd_uid, dir_gid=getents.daemons_gid)
3420

    
3421

    
3422
def BlockdevClose(instance_name, disks):
3423
  """Closes the given block devices.
3424

3425
  This means they will be switched to secondary mode (in case of
3426
  DRBD).
3427

3428
  @param instance_name: if the argument is not empty, the symlinks
3429
      of this instance will be removed
3430
  @type disks: list of L{objects.Disk}
3431
  @param disks: the list of disks to be closed
3432
  @rtype: tuple (success, message)
3433
  @return: a tuple of success and message, where success
3434
      indicates the succes of the operation, and message
3435
      which will contain the error details in case we
3436
      failed
3437

3438
  """
3439
  bdevs = []
3440
  for cf in disks:
3441
    rd = _RecursiveFindBD(cf)
3442
    if rd is None:
3443
      _Fail("Can't find device %s", cf)
3444
    bdevs.append(rd)
3445

    
3446
  msg = []
3447
  for rd in bdevs:
3448
    try:
3449
      rd.Close()
3450
    except errors.BlockDeviceError, err:
3451
      msg.append(str(err))
3452
  if msg:
3453
    _Fail("Can't make devices secondary: %s", ",".join(msg))
3454
  else:
3455
    if instance_name:
3456
      _RemoveBlockDevLinks(instance_name, disks)
3457

    
3458

    
3459
def ValidateHVParams(hvname, hvparams):
3460
  """Validates the given hypervisor parameters.
3461

3462
  @type hvname: string
3463
  @param hvname: the hypervisor name
3464
  @type hvparams: dict
3465
  @param hvparams: the hypervisor parameters to be validated
3466
  @rtype: None
3467

3468
  """
3469
  try:
3470
    hv_type = hypervisor.GetHypervisor(hvname)
3471
    hv_type.ValidateParameters(hvparams)
3472
  except errors.HypervisorError, err:
3473
    _Fail(str(err), log=False)
3474

    
3475

    
3476
def _CheckOSPList(os_obj, parameters):
3477
  """Check whether a list of parameters is supported by the OS.
3478

3479
  @type os_obj: L{objects.OS}
3480
  @param os_obj: OS object to check
3481
  @type parameters: list
3482
  @param parameters: the list of parameters to check
3483

3484
  """
3485
  supported = [v[0] for v in os_obj.supported_parameters]
3486
  delta = frozenset(parameters).difference(supported)
3487
  if delta:
3488
    _Fail("The following parameters are not supported"
3489
          " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
3490

    
3491

    
3492
def ValidateOS(required, osname, checks, osparams):
3493
  """Validate the given OS' parameters.
3494

3495
  @type required: boolean
3496
  @param required: whether absence of the OS should translate into
3497
      failure or not
3498
  @type osname: string
3499
  @param osname: the OS to be validated
3500
  @type checks: list
3501
  @param checks: list of the checks to run (currently only 'parameters')
3502
  @type osparams: dict
3503
  @param osparams: dictionary with OS parameters
3504
  @rtype: boolean
3505
  @return: True if the validation passed, or False if the OS was not
3506
      found and L{required} was false
3507

3508
  """
3509
  if not constants.OS_VALIDATE_CALLS.issuperset(checks):
3510
    _Fail("Unknown checks required for OS %s: %s", osname,
3511
          set(checks).difference(constants.OS_VALIDATE_CALLS))
3512

    
3513
  name_only = objects.OS.GetName(osname)
3514
  status, tbv = _TryOSFromDisk(name_only, None)
3515

    
3516
  if not status:
3517
    if required:
3518
      _Fail(tbv)
3519
    else:
3520
      return False
3521

    
3522
  if max(tbv.api_versions) < constants.OS_API_V20:
3523
    return True
3524

    
3525
  if constants.OS_VALIDATE_PARAMETERS in checks:
3526
    _CheckOSPList(tbv, osparams.keys())
3527

    
3528
  validate_env = OSCoreEnv(osname, tbv, osparams)
3529
  result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
3530
                        cwd=tbv.path, reset_env=True)
3531
  if result.failed:
3532
    logging.error("os validate command '%s' returned error: %s output: %s",
3533
                  result.cmd, result.fail_reason, result.output)
3534
    _Fail("OS validation script failed (%s), output: %s",
3535
          result.fail_reason, result.output, log=False)
3536

    
3537
  return True
3538

    
3539

    
3540
def DemoteFromMC():
3541
  """Demotes the current node from master candidate role.
3542

3543
  """
3544
  # try to ensure we're not the master by mistake
3545
  master, myself = ssconf.GetMasterAndMyself()
3546
  if master == myself:
3547
    _Fail("ssconf status shows I'm the master node, will not demote")
3548

    
3549
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "check", constants.MASTERD])
3550
  if not result.failed:
3551
    _Fail("The master daemon is running, will not demote")
3552

    
3553
  try:
3554
    if os.path.isfile(pathutils.CLUSTER_CONF_FILE):
3555
      utils.CreateBackup(pathutils.CLUSTER_CONF_FILE)
3556
  except EnvironmentError, err:
3557
    if err.errno != errno.ENOENT:
3558
      _Fail("Error while backing up cluster file: %s", err, exc=True)
3559

    
3560
  utils.RemoveFile(pathutils.CLUSTER_CONF_FILE)
3561

    
3562

    
3563
def _GetX509Filenames(cryptodir, name):
3564
  """Returns the full paths for the private key and certificate.
3565

3566
  """
3567
  return (utils.PathJoin(cryptodir, name),
3568
          utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
3569
          utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
3570

    
3571

    
3572
def CreateX509Certificate(validity, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3573
  """Creates a new X509 certificate for SSL/TLS.
3574

3575
  @type validity: int
3576
  @param validity: Validity in seconds
3577
  @rtype: tuple; (string, string)
3578
  @return: Certificate name and public part
3579

3580
  """
3581
  (key_pem, cert_pem) = \
3582
    utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
3583
                                     min(validity, _MAX_SSL_CERT_VALIDITY))
3584

    
3585
  cert_dir = tempfile.mkdtemp(dir=cryptodir,
3586
                              prefix="x509-%s-" % utils.TimestampForFilename())
3587
  try:
3588
    name = os.path.basename(cert_dir)
3589
    assert len(name) > 5
3590

    
3591
    (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3592

    
3593
    utils.WriteFile(key_file, mode=0400, data=key_pem)
3594
    utils.WriteFile(cert_file, mode=0400, data=cert_pem)
3595

    
3596
    # Never return private key as it shouldn't leave the node
3597
    return (name, cert_pem)
3598
  except Exception:
3599
    shutil.rmtree(cert_dir, ignore_errors=True)
3600
    raise
3601

    
3602

    
3603
def RemoveX509Certificate(name, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3604
  """Removes a X509 certificate.
3605

3606
  @type name: string
3607
  @param name: Certificate name
3608

3609
  """
3610
  (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3611

    
3612
  utils.RemoveFile(key_file)
3613
  utils.RemoveFile(cert_file)
3614

    
3615
  try:
3616
    os.rmdir(cert_dir)
3617
  except EnvironmentError, err:
3618
    _Fail("Cannot remove certificate directory '%s': %s",
3619
          cert_dir, err)
3620

    
3621

    
3622
def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
3623
  """Returns the command for the requested input/output.
3624

3625
  @type instance: L{objects.Instance}
3626
  @param instance: The instance object
3627
  @param mode: Import/export mode
3628
  @param ieio: Input/output type
3629
  @param ieargs: Input/output arguments
3630

3631
  """
3632
  assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
3633

    
3634
  env = None
3635
  prefix = None
3636
  suffix = None
3637
  exp_size = None
3638

    
3639
  if ieio == constants.IEIO_FILE:
3640
    (filename, ) = ieargs
3641

    
3642
    if not utils.IsNormAbsPath(filename):
3643
      _Fail("Path '%s' is not normalized or absolute", filename)
3644

    
3645
    real_filename = os.path.realpath(filename)
3646
    directory = os.path.dirname(real_filename)
3647

    
3648
    if not utils.IsBelowDir(pathutils.EXPORT_DIR, real_filename):
3649
      _Fail("File '%s' is not under exports directory '%s': %s",
3650
            filename, pathutils.EXPORT_DIR, real_filename)
3651

    
3652
    # Create directory
3653
    utils.Makedirs(directory, mode=0750)
3654

    
3655
    quoted_filename = utils.ShellQuote(filename)
3656

    
3657
    if mode == constants.IEM_IMPORT:
3658
      suffix = "> %s" % quoted_filename
3659
    elif mode == constants.IEM_EXPORT:
3660
      suffix = "< %s" % quoted_filename
3661

    
3662
      # Retrieve file size
3663
      try:
3664
        st = os.stat(filename)
3665
      except EnvironmentError, err:
3666
        logging.error("Can't stat(2) %s: %s", filename, err)
3667
      else:
3668
        exp_size = utils.BytesToMebibyte(st.st_size)
3669

    
3670
  elif ieio == constants.IEIO_RAW_DISK:
3671
    (disk, ) = ieargs
3672

    
3673
    real_disk = _OpenRealBD(disk)
3674

    
3675
    if mode == constants.IEM_IMPORT:
3676
      # we set here a smaller block size as, due to transport buffering, more
3677
      # than 64-128k will mostly ignored; we use nocreat to fail if the device
3678
      # is not already there or we pass a wrong path; we use notrunc to no
3679
      # attempt truncate on an LV device; we use oflag=dsync to not buffer too
3680
      # much memory; this means that at best, we flush every 64k, which will
3681
      # not be very fast
3682
      suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
3683
                                    " bs=%s oflag=dsync"),
3684
                                    real_disk.dev_path,
3685
                                    str(64 * 1024))
3686

    
3687
    elif mode == constants.IEM_EXPORT:
3688
      # the block size on the read dd is 1MiB to match our units
3689
      prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
3690
                                   real_disk.dev_path,
3691
                                   str(1024 * 1024), # 1 MB
3692
                                   str(disk.size))
3693
      exp_size = disk.size
3694

    
3695
  elif ieio == constants.IEIO_SCRIPT:
3696
    (disk, disk_index, ) = ieargs
3697

    
3698
    assert isinstance(disk_index, (int, long))
3699

    
3700
    inst_os = OSFromDisk(instance.os)
3701
    env = OSEnvironment(instance, inst_os)
3702

    
3703
    if mode == constants.IEM_IMPORT:
3704
      env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
3705
      env["IMPORT_INDEX"] = str(disk_index)
3706
      script = inst_os.import_script
3707

    
3708
    elif mode == constants.IEM_EXPORT:
3709
      real_disk = _OpenRealBD(disk)
3710
      env["EXPORT_DEVICE"] = real_disk.dev_path
3711
      env["EXPORT_INDEX"] = str(disk_index)
3712
      script = inst_os.export_script
3713

    
3714
    # TODO: Pass special environment only to script
3715
    script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
3716

    
3717
    if mode == constants.IEM_IMPORT:
3718
      suffix = "| %s" % script_cmd
3719

    
3720
    elif mode == constants.IEM_EXPORT:
3721
      prefix = "%s |" % script_cmd
3722

    
3723
    # Let script predict size
3724
    exp_size = constants.IE_CUSTOM_SIZE
3725

    
3726
  else:
3727
    _Fail("Invalid %s I/O mode %r", mode, ieio)
3728

    
3729
  return (env, prefix, suffix, exp_size)
3730

    
3731

    
3732
def _CreateImportExportStatusDir(prefix):
3733
  """Creates status directory for import/export.
3734

3735
  """
3736
  return tempfile.mkdtemp(dir=pathutils.IMPORT_EXPORT_DIR,
3737
                          prefix=("%s-%s-" %
3738
                                  (prefix, utils.TimestampForFilename())))
3739

    
3740

    
3741
def StartImportExportDaemon(mode, opts, host, port, instance, component,
3742
                            ieio, ieioargs):
3743
  """Starts an import or export daemon.
3744

3745
  @param mode: Import/output mode
3746
  @type opts: L{objects.ImportExportOptions}
3747
  @param opts: Daemon options
3748
  @type host: string
3749
  @param host: Remote host for export (None for import)
3750
  @type port: int
3751
  @param port: Remote port for export (None for import)
3752
  @type instance: L{objects.Instance}
3753
  @param instance: Instance object
3754
  @type component: string
3755
  @param component: which part of the instance is transferred now,
3756
      e.g. 'disk/0'
3757
  @param ieio: Input/output type
3758
  @param ieioargs: Input/output arguments
3759

3760
  """
3761
  if mode == constants.IEM_IMPORT:
3762
    prefix = "import"
3763

    
3764
    if not (host is None and port is None):
3765
      _Fail("Can not specify host or port on import")
3766

    
3767
  elif mode == constants.IEM_EXPORT:
3768
    prefix = "export"
3769

    
3770
    if host is None or port is None:
3771
      _Fail("Host and port must be specified for an export")
3772

    
3773
  else:
3774
    _Fail("Invalid mode %r", mode)
3775

    
3776
  if (opts.key_name is None) ^ (opts.ca_pem is None):
3777
    _Fail("Cluster certificate can only be used for both key and CA")
3778

    
3779
  (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
3780
    _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
3781

    
3782
  if opts.key_name is None:
3783
    # Use server.pem
3784
    key_path = pathutils.NODED_CERT_FILE
3785
    cert_path = pathutils.NODED_CERT_FILE
3786
    assert opts.ca_pem is None
3787
  else:
3788
    (_, key_path, cert_path) = _GetX509Filenames(pathutils.CRYPTO_KEYS_DIR,
3789
                                                 opts.key_name)
3790
    assert opts.ca_pem is not None
3791

    
3792
  for i in [key_path, cert_path]:
3793
    if not os.path.exists(i):
3794
      _Fail("File '%s' does not exist" % i)
3795

    
3796
  status_dir = _CreateImportExportStatusDir("%s-%s" % (prefix, component))
3797
  try:
3798
    status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
3799
    pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
3800
    ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
3801

    
3802
    if opts.ca_pem is None:
3803
      # Use server.pem
3804
      ca = utils.ReadFile(pathutils.NODED_CERT_FILE)
3805
    else:
3806
      ca = opts.ca_pem
3807

    
3808
    # Write CA file
3809
    utils.WriteFile(ca_file, data=ca, mode=0400)
3810

    
3811
    cmd = [
3812
      pathutils.IMPORT_EXPORT_DAEMON,
3813
      status_file, mode,
3814
      "--key=%s" % key_path,
3815
      "--cert=%s" % cert_path,
3816
      "--ca=%s" % ca_file,
3817
      ]
3818

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

    
3822
    if port:
3823
      cmd.append("--port=%s" % port)
3824

    
3825
    if opts.ipv6:
3826
      cmd.append("--ipv6")
3827
    else:
3828
      cmd.append("--ipv4")
3829

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

    
3833
    if opts.magic:
3834
      cmd.append("--magic=%s" % opts.magic)
3835

    
3836
    if exp_size is not None:
3837
      cmd.append("--expected-size=%s" % exp_size)
3838

    
3839
    if cmd_prefix:
3840
      cmd.append("--cmd-prefix=%s" % cmd_prefix)
3841

    
3842
    if cmd_suffix:
3843
      cmd.append("--cmd-suffix=%s" % cmd_suffix)
3844

    
3845
    if mode == constants.IEM_EXPORT:
3846
      # Retry connection a few times when connecting to remote peer
3847
      cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3848
      cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3849
    elif opts.connect_timeout is not None:
3850
      assert mode == constants.IEM_IMPORT
3851
      # Overall timeout for establishing connection while listening
3852
      cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3853

    
3854
    logfile = _InstanceLogName(prefix, instance.os, instance.name, component)
3855

    
3856
    # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3857
    # support for receiving a file descriptor for output
3858
    utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3859
                      output=logfile)
3860

    
3861
    # The import/export name is simply the status directory name
3862
    return os.path.basename(status_dir)
3863

    
3864
  except Exception:
3865
    shutil.rmtree(status_dir, ignore_errors=True)
3866
    raise
3867

    
3868

    
3869
def GetImportExportStatus(names):
3870
  """Returns import/export daemon status.
3871

3872
  @type names: sequence
3873
  @param names: List of names
3874
  @rtype: List of dicts
3875
  @return: Returns a list of the state of each named import/export or None if a
3876
           status couldn't be read
3877

3878
  """
3879
  result = []
3880

    
3881
  for name in names:
3882
    status_file = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name,
3883
                                 _IES_STATUS_FILE)
3884

    
3885
    try:
3886
      data = utils.ReadFile(status_file)
3887
    except EnvironmentError, err:
3888
      if err.errno != errno.ENOENT:
3889
        raise
3890
      data = None
3891

    
3892
    if not data:
3893
      result.append(None)
3894
      continue
3895

    
3896
    result.append(serializer.LoadJson(data))
3897

    
3898
  return result
3899

    
3900

    
3901
def AbortImportExport(name):
3902
  """Sends SIGTERM to a running import/export daemon.
3903

3904
  """
3905
  logging.info("Abort import/export %s", name)
3906

    
3907
  status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3908
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3909

    
3910
  if pid:
3911
    logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3912
                 name, pid)
3913
    utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3914

    
3915

    
3916
def CleanupImportExport(name):
3917
  """Cleanup after an import or export.
3918

3919
  If the import/export daemon is still running it's killed. Afterwards the
3920
  whole status directory is removed.
3921

3922
  """
3923
  logging.info("Finalizing import/export %s", name)
3924

    
3925
  status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3926

    
3927
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3928

    
3929
  if pid:
3930
    logging.info("Import/export %s is still running with PID %s",
3931
                 name, pid)
3932
    utils.KillProcess(pid, waitpid=False)
3933

    
3934
  shutil.rmtree(status_dir, ignore_errors=True)
3935

    
3936

    
3937
def _FindDisks(disks):
3938
  """Finds attached L{BlockDev}s for the given disks.
3939

3940
  @type disks: list of L{objects.Disk}
3941
  @param disks: the disk objects we need to find
3942

3943
  @return: list of L{BlockDev} objects or C{None} if a given disk
3944
           was not found or was no attached.
3945

3946
  """
3947
  bdevs = []
3948

    
3949
  for disk in disks:
3950
    rd = _RecursiveFindBD(disk)
3951
    if rd is None:
3952
      _Fail("Can't find device %s", disk)
3953
    bdevs.append(rd)
3954
  return bdevs
3955

    
3956

    
3957
def DrbdDisconnectNet(disks):
3958
  """Disconnects the network on a list of drbd devices.
3959

3960
  """
3961
  bdevs = _FindDisks(disks)
3962

    
3963
  # disconnect disks
3964
  for rd in bdevs:
3965
    try:
3966
      rd.DisconnectNet()
3967
    except errors.BlockDeviceError, err:
3968
      _Fail("Can't change network configuration to standalone mode: %s",
3969
            err, exc=True)
3970

    
3971

    
3972
def DrbdAttachNet(disks, instance_name, multimaster):
3973
  """Attaches the network on a list of drbd devices.
3974

3975
  """
3976
  bdevs = _FindDisks(disks)
3977

    
3978
  if multimaster:
3979
    for idx, rd in enumerate(bdevs):
3980
      try:
3981
        _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3982
      except EnvironmentError, err:
3983
        _Fail("Can't create symlink: %s", err)
3984
  # reconnect disks, switch to new master configuration and if
3985
  # needed primary mode
3986
  for rd in bdevs:
3987
    try:
3988
      rd.AttachNet(multimaster)
3989
    except errors.BlockDeviceError, err:
3990
      _Fail("Can't change network configuration: %s", err)
3991

    
3992
  # wait until the disks are connected; we need to retry the re-attach
3993
  # if the device becomes standalone, as this might happen if the one
3994
  # node disconnects and reconnects in a different mode before the
3995
  # other node reconnects; in this case, one or both of the nodes will
3996
  # decide it has wrong configuration and switch to standalone
3997

    
3998
  def _Attach():
3999
    all_connected = True
4000

    
4001
    for rd in bdevs:
4002
      stats = rd.GetProcStatus()
4003

    
4004
      all_connected = (all_connected and
4005
                       (stats.is_connected or stats.is_in_resync))
4006

    
4007
      if stats.is_standalone:
4008
        # peer had different config info and this node became
4009
        # standalone, even though this should not happen with the
4010
        # new staged way of changing disk configs
4011
        try:
4012
          rd.AttachNet(multimaster)
4013
        except errors.BlockDeviceError, err:
4014
          _Fail("Can't change network configuration: %s", err)
4015

    
4016
    if not all_connected:
4017
      raise utils.RetryAgain()
4018

    
4019
  try:
4020
    # Start with a delay of 100 miliseconds and go up to 5 seconds
4021
    utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
4022
  except utils.RetryTimeout:
4023
    _Fail("Timeout in disk reconnecting")
4024

    
4025
  if multimaster:
4026
    # change to primary mode
4027
    for rd in bdevs:
4028
      try:
4029
        rd.Open()
4030
      except errors.BlockDeviceError, err:
4031
        _Fail("Can't change to primary mode: %s", err)
4032

    
4033

    
4034
def DrbdWaitSync(disks):
4035
  """Wait until DRBDs have synchronized.
4036

4037
  """
4038
  def _helper(rd):
4039
    stats = rd.GetProcStatus()
4040
    if not (stats.is_connected or stats.is_in_resync):
4041
      raise utils.RetryAgain()
4042
    return stats
4043

    
4044
  bdevs = _FindDisks(disks)
4045

    
4046
  min_resync = 100
4047
  alldone = True
4048
  for rd in bdevs:
4049
    try:
4050
      # poll each second for 15 seconds
4051
      stats = utils.Retry(_helper, 1, 15, args=[rd])
4052
    except utils.RetryTimeout:
4053
      stats = rd.GetProcStatus()
4054
      # last check
4055
      if not (stats.is_connected or stats.is_in_resync):
4056
        _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
4057
    alldone = alldone and (not stats.is_in_resync)
4058
    if stats.sync_percent is not None:
4059
      min_resync = min(min_resync, stats.sync_percent)
4060

    
4061
  return (alldone, min_resync)
4062

    
4063

    
4064
def DrbdNeedsActivation(disks):
4065
  """Checks which of the passed disks needs activation and returns their UUIDs.
4066

4067
  """
4068
  faulty_disks = []
4069

    
4070
  for disk in disks:
4071
    rd = _RecursiveFindBD(disk)
4072
    if rd is None:
4073
      faulty_disks.append(disk)
4074
      continue
4075

    
4076
    stats = rd.GetProcStatus()
4077
    if stats.is_standalone or stats.is_diskless:
4078
      faulty_disks.append(disk)
4079

    
4080
  return [disk.uuid for disk in faulty_disks]
4081

    
4082

    
4083
def GetDrbdUsermodeHelper():
4084
  """Returns DRBD usermode helper currently configured.
4085

4086
  """
4087
  try:
4088
    return drbd.DRBD8.GetUsermodeHelper()
4089
  except errors.BlockDeviceError, err:
4090
    _Fail(str(err))
4091

    
4092

    
4093
def PowercycleNode(hypervisor_type, hvparams=None):
4094
  """Hard-powercycle the node.
4095

4096
  Because we need to return first, and schedule the powercycle in the
4097
  background, we won't be able to report failures nicely.
4098

4099
  """
4100
  hyper = hypervisor.GetHypervisor(hypervisor_type)
4101
  try:
4102
    pid = os.fork()
4103
  except OSError:
4104
    # if we can't fork, we'll pretend that we're in the child process
4105
    pid = 0
4106
  if pid > 0:
4107
    return "Reboot scheduled in 5 seconds"
4108
  # ensure the child is running on ram
4109
  try:
4110
    utils.Mlockall()
4111
  except Exception: # pylint: disable=W0703
4112
    pass
4113
  time.sleep(5)
4114
  hyper.PowercycleNode(hvparams=hvparams)
4115

    
4116

    
4117
def _VerifyRestrictedCmdName(cmd):
4118
  """Verifies a restricted command name.
4119

4120
  @type cmd: string
4121
  @param cmd: Command name
4122
  @rtype: tuple; (boolean, string or None)
4123
  @return: The tuple's first element is the status; if C{False}, the second
4124
    element is an error message string, otherwise it's C{None}
4125

4126
  """
4127
  if not cmd.strip():
4128
    return (False, "Missing command name")
4129

    
4130
  if os.path.basename(cmd) != cmd:
4131
    return (False, "Invalid command name")
4132

    
4133
  if not constants.EXT_PLUGIN_MASK.match(cmd):
4134
    return (False, "Command name contains forbidden characters")
4135

    
4136
  return (True, None)
4137

    
4138

    
4139
def _CommonRestrictedCmdCheck(path, owner):
4140
  """Common checks for restricted command file system directories and files.
4141

4142
  @type path: string
4143
  @param path: Path to check
4144
  @param owner: C{None} or tuple containing UID and GID
4145
  @rtype: tuple; (boolean, string or C{os.stat} result)
4146
  @return: The tuple's first element is the status; if C{False}, the second
4147
    element is an error message string, otherwise it's the result of C{os.stat}
4148

4149
  """
4150
  if owner is None:
4151
    # Default to root as owner
4152
    owner = (0, 0)
4153

    
4154
  try:
4155
    st = os.stat(path)
4156
  except EnvironmentError, err:
4157
    return (False, "Can't stat(2) '%s': %s" % (path, err))
4158

    
4159
  if stat.S_IMODE(st.st_mode) & (~_RCMD_MAX_MODE):
4160
    return (False, "Permissions on '%s' are too permissive" % path)
4161

    
4162
  if (st.st_uid, st.st_gid) != owner:
4163
    (owner_uid, owner_gid) = owner
4164
    return (False, "'%s' is not owned by %s:%s" % (path, owner_uid, owner_gid))
4165

    
4166
  return (True, st)
4167

    
4168

    
4169
def _VerifyRestrictedCmdDirectory(path, _owner=None):
4170
  """Verifies restricted command directory.
4171

4172
  @type path: string
4173
  @param path: Path to check
4174
  @rtype: tuple; (boolean, string or None)
4175
  @return: The tuple's first element is the status; if C{False}, the second
4176
    element is an error message string, otherwise it's C{None}
4177

4178
  """
4179
  (status, value) = _CommonRestrictedCmdCheck(path, _owner)
4180

    
4181
  if not status:
4182
    return (False, value)
4183

    
4184
  if not stat.S_ISDIR(value.st_mode):
4185
    return (False, "Path '%s' is not a directory" % path)
4186

    
4187
  return (True, None)
4188

    
4189

    
4190
def _VerifyRestrictedCmd(path, cmd, _owner=None):
4191
  """Verifies a whole restricted command and returns its executable filename.
4192

4193
  @type path: string
4194
  @param path: Directory containing restricted commands
4195
  @type cmd: string
4196
  @param cmd: Command name
4197
  @rtype: tuple; (boolean, string)
4198
  @return: The tuple's first element is the status; if C{False}, the second
4199
    element is an error message string, otherwise the second element is the
4200
    absolute path to the executable
4201

4202
  """
4203
  executable = utils.PathJoin(path, cmd)
4204

    
4205
  (status, msg) = _CommonRestrictedCmdCheck(executable, _owner)
4206

    
4207
  if not status:
4208
    return (False, msg)
4209

    
4210
  if not utils.IsExecutable(executable):
4211
    return (False, "access(2) thinks '%s' can't be executed" % executable)
4212

    
4213
  return (True, executable)
4214

    
4215

    
4216
def _PrepareRestrictedCmd(path, cmd,
4217
                          _verify_dir=_VerifyRestrictedCmdDirectory,
4218
                          _verify_name=_VerifyRestrictedCmdName,
4219
                          _verify_cmd=_VerifyRestrictedCmd):
4220
  """Performs a number of tests on a restricted command.
4221

4222
  @type path: string
4223
  @param path: Directory containing restricted commands
4224
  @type cmd: string
4225
  @param cmd: Command name
4226
  @return: Same as L{_VerifyRestrictedCmd}
4227

4228
  """
4229
  # Verify the directory first
4230
  (status, msg) = _verify_dir(path)
4231
  if status:
4232
    # Check command if everything was alright
4233
    (status, msg) = _verify_name(cmd)
4234

    
4235
  if not status:
4236
    return (False, msg)
4237

    
4238
  # Check actual executable
4239
  return _verify_cmd(path, cmd)
4240

    
4241

    
4242
def RunRestrictedCmd(cmd,
4243
                     _lock_timeout=_RCMD_LOCK_TIMEOUT,
4244
                     _lock_file=pathutils.RESTRICTED_COMMANDS_LOCK_FILE,
4245
                     _path=pathutils.RESTRICTED_COMMANDS_DIR,
4246
                     _sleep_fn=time.sleep,
4247
                     _prepare_fn=_PrepareRestrictedCmd,
4248
                     _runcmd_fn=utils.RunCmd,
4249
                     _enabled=constants.ENABLE_RESTRICTED_COMMANDS):
4250
  """Executes a restricted command after performing strict tests.
4251

4252
  @type cmd: string
4253
  @param cmd: Command name
4254
  @rtype: string
4255
  @return: Command output
4256
  @raise RPCFail: In case of an error
4257

4258
  """
4259
  logging.info("Preparing to run restricted command '%s'", cmd)
4260

    
4261
  if not _enabled:
4262
    _Fail("Restricted commands disabled at configure time")
4263

    
4264
  lock = None
4265
  try:
4266
    cmdresult = None
4267
    try:
4268
      lock = utils.FileLock.Open(_lock_file)
4269
      lock.Exclusive(blocking=True, timeout=_lock_timeout)
4270

    
4271
      (status, value) = _prepare_fn(_path, cmd)
4272

    
4273
      if status:
4274
        cmdresult = _runcmd_fn([value], env={}, reset_env=True,
4275
                               postfork_fn=lambda _: lock.Unlock())
4276
      else:
4277
        logging.error(value)
4278
    except Exception: # pylint: disable=W0703
4279
      # Keep original error in log
4280
      logging.exception("Caught exception")
4281

    
4282
    if cmdresult is None:
4283
      logging.info("Sleeping for %0.1f seconds before returning",
4284
                   _RCMD_INVALID_DELAY)
4285
      _sleep_fn(_RCMD_INVALID_DELAY)
4286

    
4287
      # Do not include original error message in returned error
4288
      _Fail("Executing command '%s' failed" % cmd)
4289
    elif cmdresult.failed or cmdresult.fail_reason:
4290
      _Fail("Restricted command '%s' failed: %s; output: %s",
4291
            cmd, cmdresult.fail_reason, cmdresult.output)
4292
    else:
4293
      return cmdresult.output
4294
  finally:
4295
    if lock is not None:
4296
      # Release lock at last
4297
      lock.Close()
4298
      lock = None
4299

    
4300

    
4301
def SetWatcherPause(until, _filename=pathutils.WATCHER_PAUSEFILE):
4302
  """Creates or removes the watcher pause file.
4303

4304
  @type until: None or number
4305
  @param until: Unix timestamp saying until when the watcher shouldn't run
4306

4307
  """
4308
  if until is None:
4309
    logging.info("Received request to no longer pause watcher")
4310
    utils.RemoveFile(_filename)
4311
  else:
4312
    logging.info("Received request to pause watcher until %s", until)
4313

    
4314
    if not ht.TNumber(until):
4315
      _Fail("Duration must be numeric")
4316

    
4317
    utils.WriteFile(_filename, data="%d\n" % (until, ), mode=0644)
4318

    
4319

    
4320
def ConfigureOVS(ovs_name, ovs_link):
4321
  """Creates a OpenvSwitch on the node.
4322

4323
  This function sets up a OpenvSwitch on the node with given name nad
4324
  connects it via a given eth device.
4325

4326
  @type ovs_name: string
4327
  @param ovs_name: Name of the OpenvSwitch to create.
4328
  @type ovs_link: None or string
4329
  @param ovs_link: Ethernet device for outside connection (can be missing)
4330

4331
  """
4332
  # Initialize the OpenvSwitch
4333
  result = utils.RunCmd(["ovs-vsctl", "add-br", ovs_name])
4334
  if result.failed:
4335
    _Fail("Failed to create openvswitch %s. Script return value: %s, output:"
4336
          " '%s'" % result.exit_code, result.output, log=True)
4337

    
4338
  # And connect it to a physical interface, if given
4339
  if ovs_link:
4340
    result = utils.RunCmd(["ovs-vsctl", "add-port", ovs_name, ovs_link])
4341
    if result.failed:
4342
      _Fail("Failed to connect openvswitch to  interface %s. Script return"
4343
            " value: %s, output: '%s'" % ovs_link, result.exit_code,
4344
            result.output, log=True)
4345

    
4346

    
4347
class HooksRunner(object):
4348
  """Hook runner.
4349

4350
  This class is instantiated on the node side (ganeti-noded) and not
4351
  on the master side.
4352

4353
  """
4354
  def __init__(self, hooks_base_dir=None):
4355
    """Constructor for hooks runner.
4356

4357
    @type hooks_base_dir: str or None
4358
    @param hooks_base_dir: if not None, this overrides the
4359
        L{pathutils.HOOKS_BASE_DIR} (useful for unittests)
4360

4361
    """
4362
    if hooks_base_dir is None:
4363
      hooks_base_dir = pathutils.HOOKS_BASE_DIR
4364
    # yeah, _BASE_DIR is not valid for attributes, we use it like a
4365
    # constant
4366
    self._BASE_DIR = hooks_base_dir # pylint: disable=C0103
4367

    
4368
  def RunLocalHooks(self, node_list, hpath, phase, env):
4369
    """Check that the hooks will be run only locally and then run them.
4370

4371
    """
4372
    assert len(node_list) == 1
4373
    node = node_list[0]
4374
    _, myself = ssconf.GetMasterAndMyself()
4375
    assert node == myself
4376

    
4377
    results = self.RunHooks(hpath, phase, env)
4378

    
4379
    # Return values in the form expected by HooksMaster
4380
    return {node: (None, False, results)}
4381

    
4382
  def RunHooks(self, hpath, phase, env):
4383
    """Run the scripts in the hooks directory.
4384

4385
    @type hpath: str
4386
    @param hpath: the path to the hooks directory which
4387
        holds the scripts
4388
    @type phase: str
4389
    @param phase: either L{constants.HOOKS_PHASE_PRE} or
4390
        L{constants.HOOKS_PHASE_POST}
4391
    @type env: dict
4392
    @param env: dictionary with the environment for the hook
4393
    @rtype: list
4394
    @return: list of 3-element tuples:
4395
      - script path
4396
      - script result, either L{constants.HKR_SUCCESS} or
4397
        L{constants.HKR_FAIL}
4398
      - output of the script
4399

4400
    @raise errors.ProgrammerError: for invalid input
4401
        parameters
4402

4403
    """
4404
    if phase == constants.HOOKS_PHASE_PRE:
4405
      suffix = "pre"
4406
    elif phase == constants.HOOKS_PHASE_POST:
4407
      suffix = "post"
4408
    else:
4409
      _Fail("Unknown hooks phase '%s'", phase)
4410

    
4411
    subdir = "%s-%s.d" % (hpath, suffix)
4412
    dir_name = utils.PathJoin(self._BASE_DIR, subdir)
4413

    
4414
    results = []
4415

    
4416
    if not os.path.isdir(dir_name):
4417
      # for non-existing/non-dirs, we simply exit instead of logging a
4418
      # warning at every operation
4419
      return results
4420

    
4421
    runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
4422

    
4423
    for (relname, relstatus, runresult) in runparts_results:
4424
      if relstatus == constants.RUNPARTS_SKIP:
4425
        rrval = constants.HKR_SKIP
4426
        output = ""
4427
      elif relstatus == constants.RUNPARTS_ERR:
4428
        rrval = constants.HKR_FAIL
4429
        output = "Hook script execution error: %s" % runresult
4430
      elif relstatus == constants.RUNPARTS_RUN:
4431
        if runresult.failed:
4432
          rrval = constants.HKR_FAIL
4433
        else:
4434
          rrval = constants.HKR_SUCCESS
4435
        output = utils.SafeEncode(runresult.output.strip())
4436
      results.append(("%s/%s" % (subdir, relname), rrval, output))
4437

    
4438
    return results
4439

    
4440

    
4441
class IAllocatorRunner(object):
4442
  """IAllocator runner.
4443

4444
  This class is instantiated on the node side (ganeti-noded) and not on
4445
  the master side.
4446

4447
  """
4448
  @staticmethod
4449
  def Run(name, idata):
4450
    """Run an iallocator script.
4451

4452
    @type name: str
4453
    @param name: the iallocator script name
4454
    @type idata: str
4455
    @param idata: the allocator input data
4456

4457
    @rtype: tuple
4458
    @return: two element tuple of:
4459
       - status
4460
       - either error message or stdout of allocator (for success)
4461

4462
    """
4463
    alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
4464
                                  os.path.isfile)
4465
    if alloc_script is None:
4466
      _Fail("iallocator module '%s' not found in the search path", name)
4467

    
4468
    fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
4469
    try:
4470
      os.write(fd, idata)
4471
      os.close(fd)
4472
      result = utils.RunCmd([alloc_script, fin_name])
4473
      if result.failed:
4474
        _Fail("iallocator module '%s' failed: %s, output '%s'",
4475
              name, result.fail_reason, result.output)
4476
    finally:
4477
      os.unlink(fin_name)
4478

    
4479
    return result.stdout
4480

    
4481

    
4482
class DevCacheManager(object):
4483
  """Simple class for managing a cache of block device information.
4484

4485
  """
4486
  _DEV_PREFIX = "/dev/"
4487
  _ROOT_DIR = pathutils.BDEV_CACHE_DIR
4488

    
4489
  @classmethod
4490
  def _ConvertPath(cls, dev_path):
4491
    """Converts a /dev/name path to the cache file name.
4492

4493
    This replaces slashes with underscores and strips the /dev
4494
    prefix. It then returns the full path to the cache file.
4495

4496
    @type dev_path: str
4497
    @param dev_path: the C{/dev/} path name
4498
    @rtype: str
4499
    @return: the converted path name
4500

4501
    """
4502
    if dev_path.startswith(cls._DEV_PREFIX):
4503
      dev_path = dev_path[len(cls._DEV_PREFIX):]
4504
    dev_path = dev_path.replace("/", "_")
4505
    fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
4506
    return fpath
4507

    
4508
  @classmethod
4509
  def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
4510
    """Updates the cache information for a given device.
4511

4512
    @type dev_path: str
4513
    @param dev_path: the pathname of the device
4514
    @type owner: str
4515
    @param owner: the owner (instance name) of the device
4516
    @type on_primary: bool
4517
    @param on_primary: whether this is the primary
4518
        node nor not
4519
    @type iv_name: str
4520
    @param iv_name: the instance-visible name of the
4521
        device, as in objects.Disk.iv_name
4522

4523
    @rtype: None
4524

4525
    """
4526
    if dev_path is None:
4527
      logging.error("DevCacheManager.UpdateCache got a None dev_path")
4528
      return
4529
    fpath = cls._ConvertPath(dev_path)
4530
    if on_primary:
4531
      state = "primary"
4532
    else:
4533
      state = "secondary"
4534
    if iv_name is None:
4535
      iv_name = "not_visible"
4536
    fdata = "%s %s %s\n" % (str(owner), state, iv_name)
4537
    try:
4538
      utils.WriteFile(fpath, data=fdata)
4539
    except EnvironmentError, err:
4540
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
4541

    
4542
  @classmethod
4543
  def RemoveCache(cls, dev_path):
4544
    """Remove data for a dev_path.
4545

4546
    This is just a wrapper over L{utils.io.RemoveFile} with a converted
4547
    path name and logging.
4548

4549
    @type dev_path: str
4550
    @param dev_path: the pathname of the device
4551

4552
    @rtype: None
4553

4554
    """
4555
    if dev_path is None:
4556
      logging.error("DevCacheManager.RemoveCache got a None dev_path")
4557
      return
4558
    fpath = cls._ConvertPath(dev_path)
4559
    try:
4560
      utils.RemoveFile(fpath)
4561
    except EnvironmentError, err:
4562
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)