Statistics
| Branch: | Tag: | Revision:

root / lib / backend.py @ 13a6c760

History | View | Annotate | Download (137 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
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

    
38
import os
39
import os.path
40
import shutil
41
import time
42
import stat
43
import errno
44
import re
45
import random
46
import logging
47
import tempfile
48
import zlib
49
import base64
50
import signal
51

    
52
from ganeti import errors
53
from ganeti import utils
54
from ganeti import ssh
55
from ganeti import hypervisor
56
from ganeti import constants
57
from ganeti.storage import bdev
58
from ganeti.storage import drbd
59
from ganeti.storage import filestorage
60
from ganeti import objects
61
from ganeti import ssconf
62
from ganeti import serializer
63
from ganeti import netutils
64
from ganeti import runtime
65
from ganeti import compat
66
from ganeti import pathutils
67
from ganeti import vcluster
68
from ganeti import ht
69
from ganeti.storage.base import BlockDev
70
from ganeti.storage.drbd import DRBD8
71
from ganeti import hooksmaster
72

    
73

    
74
_BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id"
75
_ALLOWED_CLEAN_DIRS = compat.UniqueFrozenset([
76
  pathutils.DATA_DIR,
77
  pathutils.JOB_QUEUE_ARCHIVE_DIR,
78
  pathutils.QUEUE_DIR,
79
  pathutils.CRYPTO_KEYS_DIR,
80
  ])
81
_MAX_SSL_CERT_VALIDITY = 7 * 24 * 60 * 60
82
_X509_KEY_FILE = "key"
83
_X509_CERT_FILE = "cert"
84
_IES_STATUS_FILE = "status"
85
_IES_PID_FILE = "pid"
86
_IES_CA_FILE = "ca"
87

    
88
#: Valid LVS output line regex
89
_LVSLINE_REGEX = re.compile("^ *([^|]+)\|([^|]+)\|([0-9.]+)\|([^|]{6,})\|?$")
90

    
91
# Actions for the master setup script
92
_MASTER_START = "start"
93
_MASTER_STOP = "stop"
94

    
95
#: Maximum file permissions for restricted command directory and executables
96
_RCMD_MAX_MODE = (stat.S_IRWXU |
97
                  stat.S_IRGRP | stat.S_IXGRP |
98
                  stat.S_IROTH | stat.S_IXOTH)
99

    
100
#: Delay before returning an error for restricted commands
101
_RCMD_INVALID_DELAY = 10
102

    
103
#: How long to wait to acquire lock for restricted commands (shorter than
104
#: L{_RCMD_INVALID_DELAY}) to reduce blockage of noded forks when many
105
#: command requests arrive
106
_RCMD_LOCK_TIMEOUT = _RCMD_INVALID_DELAY * 0.8
107

    
108

    
109
class RPCFail(Exception):
110
  """Class denoting RPC failure.
111

112
  Its argument is the error message.
113

114
  """
115

    
116

    
117
def _GetInstReasonFilename(instance_name):
118
  """Path of the file containing the reason of the instance status change.
119

120
  @type instance_name: string
121
  @param instance_name: The name of the instance
122
  @rtype: string
123
  @return: The path of the file
124

125
  """
126
  return utils.PathJoin(pathutils.INSTANCE_REASON_DIR, instance_name)
127

    
128

    
129
def _StoreInstReasonTrail(instance_name, trail):
130
  """Serialize a reason trail related to an instance change of state to file.
131

132
  The exact location of the file depends on the name of the instance and on
133
  the configuration of the Ganeti cluster defined at deploy time.
134

135
  @type instance_name: string
136
  @param instance_name: The name of the instance
137
  @rtype: None
138

139
  """
140
  json = serializer.DumpJson(trail)
141
  filename = _GetInstReasonFilename(instance_name)
142
  utils.WriteFile(filename, data=json)
143

    
144

    
145
def _Fail(msg, *args, **kwargs):
146
  """Log an error and the raise an RPCFail exception.
147

148
  This exception is then handled specially in the ganeti daemon and
149
  turned into a 'failed' return type. As such, this function is a
150
  useful shortcut for logging the error and returning it to the master
151
  daemon.
152

153
  @type msg: string
154
  @param msg: the text of the exception
155
  @raise RPCFail
156

157
  """
158
  if args:
159
    msg = msg % args
160
  if "log" not in kwargs or kwargs["log"]: # if we should log this error
161
    if "exc" in kwargs and kwargs["exc"]:
162
      logging.exception(msg)
163
    else:
164
      logging.error(msg)
165
  raise RPCFail(msg)
166

    
167

    
168
def _GetConfig():
169
  """Simple wrapper to return a SimpleStore.
170

171
  @rtype: L{ssconf.SimpleStore}
172
  @return: a SimpleStore instance
173

174
  """
175
  return ssconf.SimpleStore()
176

    
177

    
178
def _GetSshRunner(cluster_name):
179
  """Simple wrapper to return an SshRunner.
180

181
  @type cluster_name: str
182
  @param cluster_name: the cluster name, which is needed
183
      by the SshRunner constructor
184
  @rtype: L{ssh.SshRunner}
185
  @return: an SshRunner instance
186

187
  """
188
  return ssh.SshRunner(cluster_name)
189

    
190

    
191
def _Decompress(data):
192
  """Unpacks data compressed by the RPC client.
193

194
  @type data: list or tuple
195
  @param data: Data sent by RPC client
196
  @rtype: str
197
  @return: Decompressed data
198

199
  """
200
  assert isinstance(data, (list, tuple))
201
  assert len(data) == 2
202
  (encoding, content) = data
203
  if encoding == constants.RPC_ENCODING_NONE:
204
    return content
205
  elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
206
    return zlib.decompress(base64.b64decode(content))
207
  else:
208
    raise AssertionError("Unknown data encoding")
209

    
210

    
211
def _CleanDirectory(path, exclude=None):
212
  """Removes all regular files in a directory.
213

214
  @type path: str
215
  @param path: the directory to clean
216
  @type exclude: list
217
  @param exclude: list of files to be excluded, defaults
218
      to the empty list
219

220
  """
221
  if path not in _ALLOWED_CLEAN_DIRS:
222
    _Fail("Path passed to _CleanDirectory not in allowed clean targets: '%s'",
223
          path)
224

    
225
  if not os.path.isdir(path):
226
    return
227
  if exclude is None:
228
    exclude = []
229
  else:
230
    # Normalize excluded paths
231
    exclude = [os.path.normpath(i) for i in exclude]
232

    
233
  for rel_name in utils.ListVisibleFiles(path):
234
    full_name = utils.PathJoin(path, rel_name)
235
    if full_name in exclude:
236
      continue
237
    if os.path.isfile(full_name) and not os.path.islink(full_name):
238
      utils.RemoveFile(full_name)
239

    
240

    
241
def _BuildUploadFileList():
242
  """Build the list of allowed upload files.
243

244
  This is abstracted so that it's built only once at module import time.
245

246
  """
247
  allowed_files = set([
248
    pathutils.CLUSTER_CONF_FILE,
249
    pathutils.ETC_HOSTS,
250
    pathutils.SSH_KNOWN_HOSTS_FILE,
251
    pathutils.VNC_PASSWORD_FILE,
252
    pathutils.RAPI_CERT_FILE,
253
    pathutils.SPICE_CERT_FILE,
254
    pathutils.SPICE_CACERT_FILE,
255
    pathutils.RAPI_USERS_FILE,
256
    pathutils.CONFD_HMAC_KEY,
257
    pathutils.CLUSTER_DOMAIN_SECRET_FILE,
258
    ])
259

    
260
  for hv_name in constants.HYPER_TYPES:
261
    hv_class = hypervisor.GetHypervisorClass(hv_name)
262
    allowed_files.update(hv_class.GetAncillaryFiles()[0])
263

    
264
  assert pathutils.FILE_STORAGE_PATHS_FILE not in allowed_files, \
265
    "Allowed file storage paths should never be uploaded via RPC"
266

    
267
  return frozenset(allowed_files)
268

    
269

    
270
_ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
271

    
272

    
273
def JobQueuePurge():
274
  """Removes job queue files and archived jobs.
275

276
  @rtype: tuple
277
  @return: True, None
278

279
  """
280
  _CleanDirectory(pathutils.QUEUE_DIR, exclude=[pathutils.JOB_QUEUE_LOCK_FILE])
281
  _CleanDirectory(pathutils.JOB_QUEUE_ARCHIVE_DIR)
282

    
283

    
284
def GetMasterInfo():
285
  """Returns master information.
286

287
  This is an utility function to compute master information, either
288
  for consumption here or from the node daemon.
289

290
  @rtype: tuple
291
  @return: master_netdev, master_ip, master_name, primary_ip_family,
292
    master_netmask
293
  @raise RPCFail: in case of errors
294

295
  """
296
  try:
297
    cfg = _GetConfig()
298
    master_netdev = cfg.GetMasterNetdev()
299
    master_ip = cfg.GetMasterIP()
300
    master_netmask = cfg.GetMasterNetmask()
301
    master_node = cfg.GetMasterNode()
302
    primary_ip_family = cfg.GetPrimaryIPFamily()
303
  except errors.ConfigurationError, err:
304
    _Fail("Cluster configuration incomplete: %s", err, exc=True)
305
  return (master_netdev, master_ip, master_node, primary_ip_family,
306
          master_netmask)
307

    
308

    
309
def RunLocalHooks(hook_opcode, hooks_path, env_builder_fn):
310
  """Decorator that runs hooks before and after the decorated function.
311

312
  @type hook_opcode: string
313
  @param hook_opcode: opcode of the hook
314
  @type hooks_path: string
315
  @param hooks_path: path of the hooks
316
  @type env_builder_fn: function
317
  @param env_builder_fn: function that returns a dictionary containing the
318
    environment variables for the hooks. Will get all the parameters of the
319
    decorated function.
320
  @raise RPCFail: in case of pre-hook failure
321

322
  """
323
  def decorator(fn):
324
    def wrapper(*args, **kwargs):
325
      _, myself = ssconf.GetMasterAndMyself()
326
      nodes = ([myself], [myself])  # these hooks run locally
327

    
328
      env_fn = compat.partial(env_builder_fn, *args, **kwargs)
329

    
330
      cfg = _GetConfig()
331
      hr = HooksRunner()
332
      hm = hooksmaster.HooksMaster(hook_opcode, hooks_path, nodes,
333
                                   hr.RunLocalHooks, None, env_fn,
334
                                   logging.warning, cfg.GetClusterName(),
335
                                   cfg.GetMasterNode())
336
      hm.RunPhase(constants.HOOKS_PHASE_PRE)
337
      result = fn(*args, **kwargs)
338
      hm.RunPhase(constants.HOOKS_PHASE_POST)
339

    
340
      return result
341
    return wrapper
342
  return decorator
343

    
344

    
345
def _BuildMasterIpEnv(master_params, use_external_mip_script=None):
346
  """Builds environment variables for master IP hooks.
347

348
  @type master_params: L{objects.MasterNetworkParameters}
349
  @param master_params: network parameters of the master
350
  @type use_external_mip_script: boolean
351
  @param use_external_mip_script: whether to use an external master IP
352
    address setup script (unused, but necessary per the implementation of the
353
    _RunLocalHooks decorator)
354

355
  """
356
  # pylint: disable=W0613
357
  ver = netutils.IPAddress.GetVersionFromAddressFamily(master_params.ip_family)
358
  env = {
359
    "MASTER_NETDEV": master_params.netdev,
360
    "MASTER_IP": master_params.ip,
361
    "MASTER_NETMASK": str(master_params.netmask),
362
    "CLUSTER_IP_VERSION": str(ver),
363
  }
364

    
365
  return env
366

    
367

    
368
def _RunMasterSetupScript(master_params, action, use_external_mip_script):
369
  """Execute the master IP address setup script.
370

371
  @type master_params: L{objects.MasterNetworkParameters}
372
  @param master_params: network parameters of the master
373
  @type action: string
374
  @param action: action to pass to the script. Must be one of
375
    L{backend._MASTER_START} or L{backend._MASTER_STOP}
376
  @type use_external_mip_script: boolean
377
  @param use_external_mip_script: whether to use an external master IP
378
    address setup script
379
  @raise backend.RPCFail: if there are errors during the execution of the
380
    script
381

382
  """
383
  env = _BuildMasterIpEnv(master_params)
384

    
385
  if use_external_mip_script:
386
    setup_script = pathutils.EXTERNAL_MASTER_SETUP_SCRIPT
387
  else:
388
    setup_script = pathutils.DEFAULT_MASTER_SETUP_SCRIPT
389

    
390
  result = utils.RunCmd([setup_script, action], env=env, reset_env=True)
391

    
392
  if result.failed:
393
    _Fail("Failed to %s the master IP. Script return value: %s, output: '%s'" %
394
          (action, result.exit_code, result.output), log=True)
395

    
396

    
397
@RunLocalHooks(constants.FAKE_OP_MASTER_TURNUP, "master-ip-turnup",
398
               _BuildMasterIpEnv)
399
def ActivateMasterIp(master_params, use_external_mip_script):
400
  """Activate the IP address of the master daemon.
401

402
  @type master_params: L{objects.MasterNetworkParameters}
403
  @param master_params: network parameters of the master
404
  @type use_external_mip_script: boolean
405
  @param use_external_mip_script: whether to use an external master IP
406
    address setup script
407
  @raise RPCFail: in case of errors during the IP startup
408

409
  """
410
  _RunMasterSetupScript(master_params, _MASTER_START,
411
                        use_external_mip_script)
412

    
413

    
414
def StartMasterDaemons(no_voting):
415
  """Activate local node as master node.
416

417
  The function will start the master daemons (ganeti-masterd and ganeti-rapi).
418

419
  @type no_voting: boolean
420
  @param no_voting: whether to start ganeti-masterd without a node vote
421
      but still non-interactively
422
  @rtype: None
423

424
  """
425

    
426
  if no_voting:
427
    masterd_args = "--no-voting --yes-do-it"
428
  else:
429
    masterd_args = ""
430

    
431
  env = {
432
    "EXTRA_MASTERD_ARGS": masterd_args,
433
    }
434

    
435
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "start-master"], env=env)
436
  if result.failed:
437
    msg = "Can't start Ganeti master: %s" % result.output
438
    logging.error(msg)
439
    _Fail(msg)
440

    
441

    
442
@RunLocalHooks(constants.FAKE_OP_MASTER_TURNDOWN, "master-ip-turndown",
443
               _BuildMasterIpEnv)
444
def DeactivateMasterIp(master_params, use_external_mip_script):
445
  """Deactivate the master IP on this node.
446

447
  @type master_params: L{objects.MasterNetworkParameters}
448
  @param master_params: network parameters of the master
449
  @type use_external_mip_script: boolean
450
  @param use_external_mip_script: whether to use an external master IP
451
    address setup script
452
  @raise RPCFail: in case of errors during the IP turndown
453

454
  """
455
  _RunMasterSetupScript(master_params, _MASTER_STOP,
456
                        use_external_mip_script)
457

    
458

    
459
def StopMasterDaemons():
460
  """Stop the master daemons on this node.
461

462
  Stop the master daemons (ganeti-masterd and ganeti-rapi) on this node.
463

464
  @rtype: None
465

466
  """
467
  # TODO: log and report back to the caller the error failures; we
468
  # need to decide in which case we fail the RPC for this
469

    
470
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop-master"])
471
  if result.failed:
472
    logging.error("Could not stop Ganeti master, command %s had exitcode %s"
473
                  " and error %s",
474
                  result.cmd, result.exit_code, result.output)
475

    
476

    
477
def ChangeMasterNetmask(old_netmask, netmask, master_ip, master_netdev):
478
  """Change the netmask of the master IP.
479

480
  @param old_netmask: the old value of the netmask
481
  @param netmask: the new value of the netmask
482
  @param master_ip: the master IP
483
  @param master_netdev: the master network device
484

485
  """
486
  if old_netmask == netmask:
487
    return
488

    
489
  if not netutils.IPAddress.Own(master_ip):
490
    _Fail("The master IP address is not up, not attempting to change its"
491
          " netmask")
492

    
493
  result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "add",
494
                         "%s/%s" % (master_ip, netmask),
495
                         "dev", master_netdev, "label",
496
                         "%s:0" % master_netdev])
497
  if result.failed:
498
    _Fail("Could not set the new netmask on the master IP address")
499

    
500
  result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "del",
501
                         "%s/%s" % (master_ip, old_netmask),
502
                         "dev", master_netdev, "label",
503
                         "%s:0" % master_netdev])
504
  if result.failed:
505
    _Fail("Could not bring down the master IP address with the old netmask")
506

    
507

    
508
def EtcHostsModify(mode, host, ip):
509
  """Modify a host entry in /etc/hosts.
510

511
  @param mode: The mode to operate. Either add or remove entry
512
  @param host: The host to operate on
513
  @param ip: The ip associated with the entry
514

515
  """
516
  if mode == constants.ETC_HOSTS_ADD:
517
    if not ip:
518
      RPCFail("Mode 'add' needs 'ip' parameter, but parameter not"
519
              " present")
520
    utils.AddHostToEtcHosts(host, ip)
521
  elif mode == constants.ETC_HOSTS_REMOVE:
522
    if ip:
523
      RPCFail("Mode 'remove' does not allow 'ip' parameter, but"
524
              " parameter is present")
525
    utils.RemoveHostFromEtcHosts(host)
526
  else:
527
    RPCFail("Mode not supported")
528

    
529

    
530
def LeaveCluster(modify_ssh_setup):
531
  """Cleans up and remove the current node.
532

533
  This function cleans up and prepares the current node to be removed
534
  from the cluster.
535

536
  If processing is successful, then it raises an
537
  L{errors.QuitGanetiException} which is used as a special case to
538
  shutdown the node daemon.
539

540
  @param modify_ssh_setup: boolean
541

542
  """
543
  _CleanDirectory(pathutils.DATA_DIR)
544
  _CleanDirectory(pathutils.CRYPTO_KEYS_DIR)
545
  JobQueuePurge()
546

    
547
  if modify_ssh_setup:
548
    try:
549
      priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.SSH_LOGIN_USER)
550

    
551
      utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
552

    
553
      utils.RemoveFile(priv_key)
554
      utils.RemoveFile(pub_key)
555
    except errors.OpExecError:
556
      logging.exception("Error while processing ssh files")
557

    
558
  try:
559
    utils.RemoveFile(pathutils.CONFD_HMAC_KEY)
560
    utils.RemoveFile(pathutils.RAPI_CERT_FILE)
561
    utils.RemoveFile(pathutils.SPICE_CERT_FILE)
562
    utils.RemoveFile(pathutils.SPICE_CACERT_FILE)
563
    utils.RemoveFile(pathutils.NODED_CERT_FILE)
564
  except: # pylint: disable=W0702
565
    logging.exception("Error while removing cluster secrets")
566

    
567
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop", constants.CONFD])
568
  if result.failed:
569
    logging.error("Command %s failed with exitcode %s and error %s",
570
                  result.cmd, result.exit_code, result.output)
571

    
572
  # Raise a custom exception (handled in ganeti-noded)
573
  raise errors.QuitGanetiException(True, "Shutdown scheduled")
574

    
575

    
576
def _CheckStorageParams(params, num_params):
577
  """Performs sanity checks for storage parameters.
578

579
  @type params: list
580
  @param params: list of storage parameters
581
  @type num_params: int
582
  @param num_params: expected number of parameters
583

584
  """
585
  if params is None:
586
    raise errors.ProgrammerError("No storage parameters for storage"
587
                                 " reporting is provided.")
588
  if not isinstance(params, list):
589
    raise errors.ProgrammerError("The storage parameters are not of type"
590
                                 " list: '%s'" % params)
591
  if not len(params) == num_params:
592
    raise errors.ProgrammerError("Did not receive the expected number of"
593
                                 "storage parameters: expected %s,"
594
                                 " received '%s'" % (num_params, len(params)))
595

    
596

    
597
def _CheckLvmStorageParams(params):
598
  """Performs sanity check for the 'exclusive storage' flag.
599

600
  @see: C{_CheckStorageParams}
601

602
  """
603
  _CheckStorageParams(params, 1)
604
  excl_stor = params[0]
605
  if not isinstance(params[0], bool):
606
    raise errors.ProgrammerError("Exclusive storage parameter is not"
607
                                 " boolean: '%s'." % excl_stor)
608
  return excl_stor
609

    
610

    
611
def _GetLvmVgSpaceInfo(name, params):
612
  """Wrapper around C{_GetVgInfo} which checks the storage parameters.
613

614
  @type name: string
615
  @param name: name of the volume group
616
  @type params: list
617
  @param params: list of storage parameters, which in this case should be
618
    containing only one for exclusive storage
619

620
  """
621
  excl_stor = _CheckLvmStorageParams(params)
622
  return _GetVgInfo(name, excl_stor)
623

    
624

    
625
def _GetVgInfo(
626
    name, excl_stor, info_fn=bdev.LogicalVolume.GetVGInfo):
627
  """Retrieves information about a LVM volume group.
628

629
  """
630
  # TODO: GetVGInfo supports returning information for multiple VGs at once
631
  vginfo = info_fn([name], excl_stor)
632
  if vginfo:
633
    vg_free = int(round(vginfo[0][0], 0))
634
    vg_size = int(round(vginfo[0][1], 0))
635
  else:
636
    vg_free = None
637
    vg_size = None
638

    
639
  return {
640
    "type": constants.ST_LVM_VG,
641
    "name": name,
642
    "storage_free": vg_free,
643
    "storage_size": vg_size,
644
    }
645

    
646

    
647
def _GetLvmPvSpaceInfo(name, params):
648
  """Wrapper around C{_GetVgSpindlesInfo} with sanity checks.
649

650
  @see: C{_GetLvmVgSpaceInfo}
651

652
  """
653
  excl_stor = _CheckLvmStorageParams(params)
654
  return _GetVgSpindlesInfo(name, excl_stor)
655

    
656

    
657
def _GetVgSpindlesInfo(
658
    name, excl_stor, info_fn=bdev.LogicalVolume.GetVgSpindlesInfo):
659
  """Retrieves information about spindles in an LVM volume group.
660

661
  @type name: string
662
  @param name: VG name
663
  @type excl_stor: bool
664
  @param excl_stor: exclusive storage
665
  @rtype: dict
666
  @return: dictionary whose keys are "name", "vg_free", "vg_size" for VG name,
667
      free spindles, total spindles respectively
668

669
  """
670
  if excl_stor:
671
    (vg_free, vg_size) = info_fn(name)
672
  else:
673
    vg_free = 0
674
    vg_size = 0
675
  return {
676
    "type": constants.ST_LVM_PV,
677
    "name": name,
678
    "storage_free": vg_free,
679
    "storage_size": vg_size,
680
    }
681

    
682

    
683
def _GetHvInfo(name, hvparams, get_hv_fn=hypervisor.GetHypervisor):
684
  """Retrieves node information from a hypervisor.
685

686
  The information returned depends on the hypervisor. Common items:
687

688
    - vg_size is the size of the configured volume group in MiB
689
    - vg_free is the free size of the volume group in MiB
690
    - memory_dom0 is the memory allocated for domain0 in MiB
691
    - memory_free is the currently available (free) ram in MiB
692
    - memory_total is the total number of ram in MiB
693
    - hv_version: the hypervisor version, if available
694

695
  @type hvparams: dict of string
696
  @param hvparams: the hypervisor's hvparams
697

698
  """
699
  return get_hv_fn(name).GetNodeInfo(hvparams=hvparams)
700

    
701

    
702
def _GetHvInfoAll(hv_specs, get_hv_fn=hypervisor.GetHypervisor):
703
  """Retrieves node information for all hypervisors.
704

705
  See C{_GetHvInfo} for information on the output.
706

707
  @type hv_specs: list of pairs (string, dict of strings)
708
  @param hv_specs: list of pairs of a hypervisor's name and its hvparams
709

710
  """
711
  if hv_specs is None:
712
    return None
713

    
714
  result = []
715
  for hvname, hvparams in hv_specs:
716
    result.append(_GetHvInfo(hvname, hvparams, get_hv_fn))
717
  return result
718

    
719

    
720
def _GetNamedNodeInfo(names, fn):
721
  """Calls C{fn} for all names in C{names} and returns a dictionary.
722

723
  @rtype: None or dict
724

725
  """
726
  if names is None:
727
    return None
728
  else:
729
    return map(fn, names)
730

    
731

    
732
def GetNodeInfo(storage_units, hv_specs):
733
  """Gives back a hash with different information about the node.
734

735
  @type storage_units: list of tuples (string, string, list)
736
  @param storage_units: List of tuples (storage unit, identifier, parameters) to
737
    ask for disk space information. In case of lvm-vg, the identifier is
738
    the VG name. The parameters can contain additional, storage-type-specific
739
    parameters, for example exclusive storage for lvm storage.
740
  @type hv_specs: list of pairs (string, dict of strings)
741
  @param hv_specs: list of pairs of a hypervisor's name and its hvparams
742
  @rtype: tuple; (string, None/dict, None/dict)
743
  @return: Tuple containing boot ID, volume group information and hypervisor
744
    information
745

746
  """
747
  bootid = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
748
  storage_info = _GetNamedNodeInfo(
749
    storage_units,
750
    (lambda (storage_type, storage_key, storage_params):
751
        _ApplyStorageInfoFunction(storage_type, storage_key, storage_params)))
752
  hv_info = _GetHvInfoAll(hv_specs)
753
  return (bootid, storage_info, hv_info)
754

    
755

    
756
def _GetFileStorageSpaceInfo(path, params):
757
  """Wrapper around filestorage.GetSpaceInfo.
758

759
  The purpose of this wrapper is to call filestorage.GetFileStorageSpaceInfo
760
  and ignore the *args parameter to not leak it into the filestorage
761
  module's code.
762

763
  @see: C{filestorage.GetFileStorageSpaceInfo} for description of the
764
    parameters.
765

766
  """
767
  _CheckStorageParams(params, 0)
768
  return filestorage.GetFileStorageSpaceInfo(path)
769

    
770

    
771
# FIXME: implement storage reporting for all missing storage types.
772
_STORAGE_TYPE_INFO_FN = {
773
  constants.ST_BLOCK: None,
774
  constants.ST_DISKLESS: None,
775
  constants.ST_EXT: None,
776
  constants.ST_FILE: _GetFileStorageSpaceInfo,
777
  constants.ST_LVM_PV: _GetLvmPvSpaceInfo,
778
  constants.ST_LVM_VG: _GetLvmVgSpaceInfo,
779
  constants.ST_RADOS: None,
780
}
781

    
782

    
783
def _ApplyStorageInfoFunction(storage_type, storage_key, *args):
784
  """Looks up and applies the correct function to calculate free and total
785
  storage for the given storage type.
786

787
  @type storage_type: string
788
  @param storage_type: the storage type for which the storage shall be reported.
789
  @type storage_key: string
790
  @param storage_key: identifier of a storage unit, e.g. the volume group name
791
    of an LVM storage unit
792
  @type args: any
793
  @param args: various parameters that can be used for storage reporting. These
794
    parameters and their semantics vary from storage type to storage type and
795
    are just propagated in this function.
796
  @return: the results of the application of the storage space function (see
797
    _STORAGE_TYPE_INFO_FN) if storage space reporting is implemented for that
798
    storage type
799
  @raises NotImplementedError: for storage types who don't support space
800
    reporting yet
801
  """
802
  fn = _STORAGE_TYPE_INFO_FN[storage_type]
803
  if fn is not None:
804
    return fn(storage_key, *args)
805
  else:
806
    raise NotImplementedError
807

    
808

    
809
def _CheckExclusivePvs(pvi_list):
810
  """Check that PVs are not shared among LVs
811

812
  @type pvi_list: list of L{objects.LvmPvInfo} objects
813
  @param pvi_list: information about the PVs
814

815
  @rtype: list of tuples (string, list of strings)
816
  @return: offending volumes, as tuples: (pv_name, [lv1_name, lv2_name...])
817

818
  """
819
  res = []
820
  for pvi in pvi_list:
821
    if len(pvi.lv_list) > 1:
822
      res.append((pvi.name, pvi.lv_list))
823
  return res
824

    
825

    
826
def _VerifyHypervisors(what, vm_capable, result, all_hvparams,
827
                       get_hv_fn=hypervisor.GetHypervisor):
828
  """Verifies the hypervisor. Appends the results to the 'results' list.
829

830
  @type what: C{dict}
831
  @param what: a dictionary of things to check
832
  @type vm_capable: boolean
833
  @param vm_capable: whether or not this node is vm capable
834
  @type result: dict
835
  @param result: dictionary of verification results; results of the
836
    verifications in this function will be added here
837
  @type all_hvparams: dict of dict of string
838
  @param all_hvparams: dictionary mapping hypervisor names to hvparams
839
  @type get_hv_fn: function
840
  @param get_hv_fn: function to retrieve the hypervisor, to improve testability
841

842
  """
843
  if not vm_capable:
844
    return
845

    
846
  if constants.NV_HYPERVISOR in what:
847
    result[constants.NV_HYPERVISOR] = {}
848
    for hv_name in what[constants.NV_HYPERVISOR]:
849
      hvparams = all_hvparams[hv_name]
850
      try:
851
        val = get_hv_fn(hv_name).Verify(hvparams=hvparams)
852
      except errors.HypervisorError, err:
853
        val = "Error while checking hypervisor: %s" % str(err)
854
      result[constants.NV_HYPERVISOR][hv_name] = val
855

    
856

    
857
def _VerifyHvparams(what, vm_capable, result,
858
                    get_hv_fn=hypervisor.GetHypervisor):
859
  """Verifies the hvparams. Appends the results to the 'results' list.
860

861
  @type what: C{dict}
862
  @param what: a dictionary of things to check
863
  @type vm_capable: boolean
864
  @param vm_capable: whether or not this node is vm capable
865
  @type result: dict
866
  @param result: dictionary of verification results; results of the
867
    verifications in this function will be added here
868
  @type get_hv_fn: function
869
  @param get_hv_fn: function to retrieve the hypervisor, to improve testability
870

871
  """
872
  if not vm_capable:
873
    return
874

    
875
  if constants.NV_HVPARAMS in what:
876
    result[constants.NV_HVPARAMS] = []
877
    for source, hv_name, hvparms in what[constants.NV_HVPARAMS]:
878
      try:
879
        logging.info("Validating hv %s, %s", hv_name, hvparms)
880
        get_hv_fn(hv_name).ValidateParameters(hvparms)
881
      except errors.HypervisorError, err:
882
        result[constants.NV_HVPARAMS].append((source, hv_name, str(err)))
883

    
884

    
885
def _VerifyInstanceList(what, vm_capable, result, all_hvparams):
886
  """Verifies the instance list.
887

888
  @type what: C{dict}
889
  @param what: a dictionary of things to check
890
  @type vm_capable: boolean
891
  @param vm_capable: whether or not this node is vm capable
892
  @type result: dict
893
  @param result: dictionary of verification results; results of the
894
    verifications in this function will be added here
895
  @type all_hvparams: dict of dict of string
896
  @param all_hvparams: dictionary mapping hypervisor names to hvparams
897

898
  """
899
  if constants.NV_INSTANCELIST in what and vm_capable:
900
    # GetInstanceList can fail
901
    try:
902
      val = GetInstanceList(what[constants.NV_INSTANCELIST],
903
                            all_hvparams=all_hvparams)
904
    except RPCFail, err:
905
      val = str(err)
906
    result[constants.NV_INSTANCELIST] = val
907

    
908

    
909
def _VerifyNodeInfo(what, vm_capable, result, all_hvparams):
910
  """Verifies the node info.
911

912
  @type what: C{dict}
913
  @param what: a dictionary of things to check
914
  @type vm_capable: boolean
915
  @param vm_capable: whether or not this node is vm capable
916
  @type result: dict
917
  @param result: dictionary of verification results; results of the
918
    verifications in this function will be added here
919
  @type all_hvparams: dict of dict of string
920
  @param all_hvparams: dictionary mapping hypervisor names to hvparams
921

922
  """
923
  if constants.NV_HVINFO in what and vm_capable:
924
    hvname = what[constants.NV_HVINFO]
925
    hyper = hypervisor.GetHypervisor(hvname)
926
    hvparams = all_hvparams[hvname]
927
    result[constants.NV_HVINFO] = hyper.GetNodeInfo(hvparams=hvparams)
928

    
929

    
930
def VerifyNode(what, cluster_name, all_hvparams):
931
  """Verify the status of the local node.
932

933
  Based on the input L{what} parameter, various checks are done on the
934
  local node.
935

936
  If the I{filelist} key is present, this list of
937
  files is checksummed and the file/checksum pairs are returned.
938

939
  If the I{nodelist} key is present, we check that we have
940
  connectivity via ssh with the target nodes (and check the hostname
941
  report).
942

943
  If the I{node-net-test} key is present, we check that we have
944
  connectivity to the given nodes via both primary IP and, if
945
  applicable, secondary IPs.
946

947
  @type what: C{dict}
948
  @param what: a dictionary of things to check:
949
      - filelist: list of files for which to compute checksums
950
      - nodelist: list of nodes we should check ssh communication with
951
      - node-net-test: list of nodes we should check node daemon port
952
        connectivity with
953
      - hypervisor: list with hypervisors to run the verify for
954
  @type cluster_name: string
955
  @param cluster_name: the cluster's name
956
  @type all_hvparams: dict of dict of strings
957
  @param all_hvparams: a dictionary mapping hypervisor names to hvparams
958
  @rtype: dict
959
  @return: a dictionary with the same keys as the input dict, and
960
      values representing the result of the checks
961

962
  """
963
  result = {}
964
  my_name = netutils.Hostname.GetSysName()
965
  port = netutils.GetDaemonPort(constants.NODED)
966
  vm_capable = my_name not in what.get(constants.NV_VMNODES, [])
967

    
968
  _VerifyHypervisors(what, vm_capable, result, all_hvparams)
969
  _VerifyHvparams(what, vm_capable, result)
970

    
971
  if constants.NV_FILELIST in what:
972
    fingerprints = utils.FingerprintFiles(map(vcluster.LocalizeVirtualPath,
973
                                              what[constants.NV_FILELIST]))
974
    result[constants.NV_FILELIST] = \
975
      dict((vcluster.MakeVirtualPath(key), value)
976
           for (key, value) in fingerprints.items())
977

    
978
  if constants.NV_NODELIST in what:
979
    (nodes, bynode) = what[constants.NV_NODELIST]
980

    
981
    # Add nodes from other groups (different for each node)
982
    try:
983
      nodes.extend(bynode[my_name])
984
    except KeyError:
985
      pass
986

    
987
    # Use a random order
988
    random.shuffle(nodes)
989

    
990
    # Try to contact all nodes
991
    val = {}
992
    for node in nodes:
993
      success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
994
      if not success:
995
        val[node] = message
996

    
997
    result[constants.NV_NODELIST] = val
998

    
999
  if constants.NV_NODENETTEST in what:
1000
    result[constants.NV_NODENETTEST] = tmp = {}
1001
    my_pip = my_sip = None
1002
    for name, pip, sip in what[constants.NV_NODENETTEST]:
1003
      if name == my_name:
1004
        my_pip = pip
1005
        my_sip = sip
1006
        break
1007
    if not my_pip:
1008
      tmp[my_name] = ("Can't find my own primary/secondary IP"
1009
                      " in the node list")
1010
    else:
1011
      for name, pip, sip in what[constants.NV_NODENETTEST]:
1012
        fail = []
1013
        if not netutils.TcpPing(pip, port, source=my_pip):
1014
          fail.append("primary")
1015
        if sip != pip:
1016
          if not netutils.TcpPing(sip, port, source=my_sip):
1017
            fail.append("secondary")
1018
        if fail:
1019
          tmp[name] = ("failure using the %s interface(s)" %
1020
                       " and ".join(fail))
1021

    
1022
  if constants.NV_MASTERIP in what:
1023
    # FIXME: add checks on incoming data structures (here and in the
1024
    # rest of the function)
1025
    master_name, master_ip = what[constants.NV_MASTERIP]
1026
    if master_name == my_name:
1027
      source = constants.IP4_ADDRESS_LOCALHOST
1028
    else:
1029
      source = None
1030
    result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port,
1031
                                                     source=source)
1032

    
1033
  if constants.NV_USERSCRIPTS in what:
1034
    result[constants.NV_USERSCRIPTS] = \
1035
      [script for script in what[constants.NV_USERSCRIPTS]
1036
       if not utils.IsExecutable(script)]
1037

    
1038
  if constants.NV_OOB_PATHS in what:
1039
    result[constants.NV_OOB_PATHS] = tmp = []
1040
    for path in what[constants.NV_OOB_PATHS]:
1041
      try:
1042
        st = os.stat(path)
1043
      except OSError, err:
1044
        tmp.append("error stating out of band helper: %s" % err)
1045
      else:
1046
        if stat.S_ISREG(st.st_mode):
1047
          if stat.S_IMODE(st.st_mode) & stat.S_IXUSR:
1048
            tmp.append(None)
1049
          else:
1050
            tmp.append("out of band helper %s is not executable" % path)
1051
        else:
1052
          tmp.append("out of band helper %s is not a file" % path)
1053

    
1054
  if constants.NV_LVLIST in what and vm_capable:
1055
    try:
1056
      val = GetVolumeList(utils.ListVolumeGroups().keys())
1057
    except RPCFail, err:
1058
      val = str(err)
1059
    result[constants.NV_LVLIST] = val
1060

    
1061
  _VerifyInstanceList(what, vm_capable, result, all_hvparams)
1062

    
1063
  if constants.NV_VGLIST in what and vm_capable:
1064
    result[constants.NV_VGLIST] = utils.ListVolumeGroups()
1065

    
1066
  if constants.NV_PVLIST in what and vm_capable:
1067
    check_exclusive_pvs = constants.NV_EXCLUSIVEPVS in what
1068
    val = bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
1069
                                       filter_allocatable=False,
1070
                                       include_lvs=check_exclusive_pvs)
1071
    if check_exclusive_pvs:
1072
      result[constants.NV_EXCLUSIVEPVS] = _CheckExclusivePvs(val)
1073
      for pvi in val:
1074
        # Avoid sending useless data on the wire
1075
        pvi.lv_list = []
1076
    result[constants.NV_PVLIST] = map(objects.LvmPvInfo.ToDict, val)
1077

    
1078
  if constants.NV_VERSION in what:
1079
    result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
1080
                                    constants.RELEASE_VERSION)
1081

    
1082
  _VerifyNodeInfo(what, vm_capable, result, all_hvparams)
1083

    
1084
  if constants.NV_DRBDVERSION in what and vm_capable:
1085
    try:
1086
      drbd_version = DRBD8.GetProcInfo().GetVersionString()
1087
    except errors.BlockDeviceError, err:
1088
      logging.warning("Can't get DRBD version", exc_info=True)
1089
      drbd_version = str(err)
1090
    result[constants.NV_DRBDVERSION] = drbd_version
1091

    
1092
  if constants.NV_DRBDLIST in what and vm_capable:
1093
    try:
1094
      used_minors = drbd.DRBD8.GetUsedDevs()
1095
    except errors.BlockDeviceError, err:
1096
      logging.warning("Can't get used minors list", exc_info=True)
1097
      used_minors = str(err)
1098
    result[constants.NV_DRBDLIST] = used_minors
1099

    
1100
  if constants.NV_DRBDHELPER in what and vm_capable:
1101
    status = True
1102
    try:
1103
      payload = drbd.DRBD8.GetUsermodeHelper()
1104
    except errors.BlockDeviceError, err:
1105
      logging.error("Can't get DRBD usermode helper: %s", str(err))
1106
      status = False
1107
      payload = str(err)
1108
    result[constants.NV_DRBDHELPER] = (status, payload)
1109

    
1110
  if constants.NV_NODESETUP in what:
1111
    result[constants.NV_NODESETUP] = tmpr = []
1112
    if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
1113
      tmpr.append("The sysfs filesytem doesn't seem to be mounted"
1114
                  " under /sys, missing required directories /sys/block"
1115
                  " and /sys/class/net")
1116
    if (not os.path.isdir("/proc/sys") or
1117
        not os.path.isfile("/proc/sysrq-trigger")):
1118
      tmpr.append("The procfs filesystem doesn't seem to be mounted"
1119
                  " under /proc, missing required directory /proc/sys and"
1120
                  " the file /proc/sysrq-trigger")
1121

    
1122
  if constants.NV_TIME in what:
1123
    result[constants.NV_TIME] = utils.SplitTime(time.time())
1124

    
1125
  if constants.NV_OSLIST in what and vm_capable:
1126
    result[constants.NV_OSLIST] = DiagnoseOS()
1127

    
1128
  if constants.NV_BRIDGES in what and vm_capable:
1129
    result[constants.NV_BRIDGES] = [bridge
1130
                                    for bridge in what[constants.NV_BRIDGES]
1131
                                    if not utils.BridgeExists(bridge)]
1132

    
1133
  if what.get(constants.NV_ACCEPTED_STORAGE_PATHS) == my_name:
1134
    result[constants.NV_ACCEPTED_STORAGE_PATHS] = \
1135
        filestorage.ComputeWrongFileStoragePaths()
1136

    
1137
  return result
1138

    
1139

    
1140
def GetBlockDevSizes(devices):
1141
  """Return the size of the given block devices
1142

1143
  @type devices: list
1144
  @param devices: list of block device nodes to query
1145
  @rtype: dict
1146
  @return:
1147
    dictionary of all block devices under /dev (key). The value is their
1148
    size in MiB.
1149

1150
    {'/dev/disk/by-uuid/123456-12321231-312312-312': 124}
1151

1152
  """
1153
  DEV_PREFIX = "/dev/"
1154
  blockdevs = {}
1155

    
1156
  for devpath in devices:
1157
    if not utils.IsBelowDir(DEV_PREFIX, devpath):
1158
      continue
1159

    
1160
    try:
1161
      st = os.stat(devpath)
1162
    except EnvironmentError, err:
1163
      logging.warning("Error stat()'ing device %s: %s", devpath, str(err))
1164
      continue
1165

    
1166
    if stat.S_ISBLK(st.st_mode):
1167
      result = utils.RunCmd(["blockdev", "--getsize64", devpath])
1168
      if result.failed:
1169
        # We don't want to fail, just do not list this device as available
1170
        logging.warning("Cannot get size for block device %s", devpath)
1171
        continue
1172

    
1173
      size = int(result.stdout) / (1024 * 1024)
1174
      blockdevs[devpath] = size
1175
  return blockdevs
1176

    
1177

    
1178
def GetVolumeList(vg_names):
1179
  """Compute list of logical volumes and their size.
1180

1181
  @type vg_names: list
1182
  @param vg_names: the volume groups whose LVs we should list, or
1183
      empty for all volume groups
1184
  @rtype: dict
1185
  @return:
1186
      dictionary of all partions (key) with value being a tuple of
1187
      their size (in MiB), inactive and online status::
1188

1189
        {'xenvg/test1': ('20.06', True, True)}
1190

1191
      in case of errors, a string is returned with the error
1192
      details.
1193

1194
  """
1195
  lvs = {}
1196
  sep = "|"
1197
  if not vg_names:
1198
    vg_names = []
1199
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
1200
                         "--separator=%s" % sep,
1201
                         "-ovg_name,lv_name,lv_size,lv_attr"] + vg_names)
1202
  if result.failed:
1203
    _Fail("Failed to list logical volumes, lvs output: %s", result.output)
1204

    
1205
  for line in result.stdout.splitlines():
1206
    line = line.strip()
1207
    match = _LVSLINE_REGEX.match(line)
1208
    if not match:
1209
      logging.error("Invalid line returned from lvs output: '%s'", line)
1210
      continue
1211
    vg_name, name, size, attr = match.groups()
1212
    inactive = attr[4] == "-"
1213
    online = attr[5] == "o"
1214
    virtual = attr[0] == "v"
1215
    if virtual:
1216
      # we don't want to report such volumes as existing, since they
1217
      # don't really hold data
1218
      continue
1219
    lvs[vg_name + "/" + name] = (size, inactive, online)
1220

    
1221
  return lvs
1222

    
1223

    
1224
def ListVolumeGroups():
1225
  """List the volume groups and their size.
1226

1227
  @rtype: dict
1228
  @return: dictionary with keys volume name and values the
1229
      size of the volume
1230

1231
  """
1232
  return utils.ListVolumeGroups()
1233

    
1234

    
1235
def NodeVolumes():
1236
  """List all volumes on this node.
1237

1238
  @rtype: list
1239
  @return:
1240
    A list of dictionaries, each having four keys:
1241
      - name: the logical volume name,
1242
      - size: the size of the logical volume
1243
      - dev: the physical device on which the LV lives
1244
      - vg: the volume group to which it belongs
1245

1246
    In case of errors, we return an empty list and log the
1247
    error.
1248

1249
    Note that since a logical volume can live on multiple physical
1250
    volumes, the resulting list might include a logical volume
1251
    multiple times.
1252

1253
  """
1254
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
1255
                         "--separator=|",
1256
                         "--options=lv_name,lv_size,devices,vg_name"])
1257
  if result.failed:
1258
    _Fail("Failed to list logical volumes, lvs output: %s",
1259
          result.output)
1260

    
1261
  def parse_dev(dev):
1262
    return dev.split("(")[0]
1263

    
1264
  def handle_dev(dev):
1265
    return [parse_dev(x) for x in dev.split(",")]
1266

    
1267
  def map_line(line):
1268
    line = [v.strip() for v in line]
1269
    return [{"name": line[0], "size": line[1],
1270
             "dev": dev, "vg": line[3]} for dev in handle_dev(line[2])]
1271

    
1272
  all_devs = []
1273
  for line in result.stdout.splitlines():
1274
    if line.count("|") >= 3:
1275
      all_devs.extend(map_line(line.split("|")))
1276
    else:
1277
      logging.warning("Strange line in the output from lvs: '%s'", line)
1278
  return all_devs
1279

    
1280

    
1281
def BridgesExist(bridges_list):
1282
  """Check if a list of bridges exist on the current node.
1283

1284
  @rtype: boolean
1285
  @return: C{True} if all of them exist, C{False} otherwise
1286

1287
  """
1288
  missing = []
1289
  for bridge in bridges_list:
1290
    if not utils.BridgeExists(bridge):
1291
      missing.append(bridge)
1292

    
1293
  if missing:
1294
    _Fail("Missing bridges %s", utils.CommaJoin(missing))
1295

    
1296

    
1297
def GetInstanceListForHypervisor(hname, hvparams=None,
1298
                                 get_hv_fn=hypervisor.GetHypervisor):
1299
  """Provides a list of instances of the given hypervisor.
1300

1301
  @type hname: string
1302
  @param hname: name of the hypervisor
1303
  @type hvparams: dict of strings
1304
  @param hvparams: hypervisor parameters for the given hypervisor
1305
  @type get_hv_fn: function
1306
  @param get_hv_fn: function that returns a hypervisor for the given hypervisor
1307
    name; optional parameter to increase testability
1308

1309
  @rtype: list
1310
  @return: a list of all running instances on the current node
1311
    - instance1.example.com
1312
    - instance2.example.com
1313

1314
  """
1315
  results = []
1316
  try:
1317
    hv = get_hv_fn(hname)
1318
    names = hv.ListInstances(hvparams=hvparams)
1319
    results.extend(names)
1320
  except errors.HypervisorError, err:
1321
    _Fail("Error enumerating instances (hypervisor %s): %s",
1322
          hname, err, exc=True)
1323
  return results
1324

    
1325

    
1326
def GetInstanceList(hypervisor_list, all_hvparams=None,
1327
                    get_hv_fn=hypervisor.GetHypervisor):
1328
  """Provides a list of instances.
1329

1330
  @type hypervisor_list: list
1331
  @param hypervisor_list: the list of hypervisors to query information
1332
  @type all_hvparams: dict of dict of strings
1333
  @param all_hvparams: a dictionary mapping hypervisor types to respective
1334
    cluster-wide hypervisor parameters
1335
  @type get_hv_fn: function
1336
  @param get_hv_fn: function that returns a hypervisor for the given hypervisor
1337
    name; optional parameter to increase testability
1338

1339
  @rtype: list
1340
  @return: a list of all running instances on the current node
1341
    - instance1.example.com
1342
    - instance2.example.com
1343

1344
  """
1345
  results = []
1346
  for hname in hypervisor_list:
1347
    hvparams = all_hvparams[hname]
1348
    results.extend(GetInstanceListForHypervisor(hname, hvparams=hvparams,
1349
                                                get_hv_fn=get_hv_fn))
1350
  return results
1351

    
1352

    
1353
def GetInstanceInfo(instance, hname, hvparams=None):
1354
  """Gives back the information about an instance as a dictionary.
1355

1356
  @type instance: string
1357
  @param instance: the instance name
1358
  @type hname: string
1359
  @param hname: the hypervisor type of the instance
1360
  @type hvparams: dict of strings
1361
  @param hvparams: the instance's hvparams
1362

1363
  @rtype: dict
1364
  @return: dictionary with the following keys:
1365
      - memory: memory size of instance (int)
1366
      - state: xen state of instance (string)
1367
      - time: cpu time of instance (float)
1368
      - vcpus: the number of vcpus (int)
1369

1370
  """
1371
  output = {}
1372

    
1373
  iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance,
1374
                                                          hvparams=hvparams)
1375
  if iinfo is not None:
1376
    output["memory"] = iinfo[2]
1377
    output["vcpus"] = iinfo[3]
1378
    output["state"] = iinfo[4]
1379
    output["time"] = iinfo[5]
1380

    
1381
  return output
1382

    
1383

    
1384
def GetInstanceMigratable(instance):
1385
  """Computes whether an instance can be migrated.
1386

1387
  @type instance: L{objects.Instance}
1388
  @param instance: object representing the instance to be checked.
1389

1390
  @rtype: tuple
1391
  @return: tuple of (result, description) where:
1392
      - result: whether the instance can be migrated or not
1393
      - description: a description of the issue, if relevant
1394

1395
  """
1396
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1397
  iname = instance.name
1398
  if iname not in hyper.ListInstances(instance.hvparams):
1399
    _Fail("Instance %s is not running", iname)
1400

    
1401
  for idx in range(len(instance.disks)):
1402
    link_name = _GetBlockDevSymlinkPath(iname, idx)
1403
    if not os.path.islink(link_name):
1404
      logging.warning("Instance %s is missing symlink %s for disk %d",
1405
                      iname, link_name, idx)
1406

    
1407

    
1408
def GetAllInstancesInfo(hypervisor_list, all_hvparams):
1409
  """Gather data about all instances.
1410

1411
  This is the equivalent of L{GetInstanceInfo}, except that it
1412
  computes data for all instances at once, thus being faster if one
1413
  needs data about more than one instance.
1414

1415
  @type hypervisor_list: list
1416
  @param hypervisor_list: list of hypervisors to query for instance data
1417
  @type all_hvparams: dict of dict of strings
1418
  @param all_hvparams: mapping of hypervisor names to hvparams
1419

1420
  @rtype: dict
1421
  @return: dictionary of instance: data, with data having the following keys:
1422
      - memory: memory size of instance (int)
1423
      - state: xen state of instance (string)
1424
      - time: cpu time of instance (float)
1425
      - vcpus: the number of vcpus
1426

1427
  """
1428
  output = {}
1429

    
1430
  for hname in hypervisor_list:
1431
    hvparams = all_hvparams[hname]
1432
    iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo(hvparams)
1433
    if iinfo:
1434
      for name, _, memory, vcpus, state, times in iinfo:
1435
        value = {
1436
          "memory": memory,
1437
          "vcpus": vcpus,
1438
          "state": state,
1439
          "time": times,
1440
          }
1441
        if name in output:
1442
          # we only check static parameters, like memory and vcpus,
1443
          # and not state and time which can change between the
1444
          # invocations of the different hypervisors
1445
          for key in "memory", "vcpus":
1446
            if value[key] != output[name][key]:
1447
              _Fail("Instance %s is running twice"
1448
                    " with different parameters", name)
1449
        output[name] = value
1450

    
1451
  return output
1452

    
1453

    
1454
def _InstanceLogName(kind, os_name, instance, component):
1455
  """Compute the OS log filename for a given instance and operation.
1456

1457
  The instance name and os name are passed in as strings since not all
1458
  operations have these as part of an instance object.
1459

1460
  @type kind: string
1461
  @param kind: the operation type (e.g. add, import, etc.)
1462
  @type os_name: string
1463
  @param os_name: the os name
1464
  @type instance: string
1465
  @param instance: the name of the instance being imported/added/etc.
1466
  @type component: string or None
1467
  @param component: the name of the component of the instance being
1468
      transferred
1469

1470
  """
1471
  # TODO: Use tempfile.mkstemp to create unique filename
1472
  if component:
1473
    assert "/" not in component
1474
    c_msg = "-%s" % component
1475
  else:
1476
    c_msg = ""
1477
  base = ("%s-%s-%s%s-%s.log" %
1478
          (kind, os_name, instance, c_msg, utils.TimestampForFilename()))
1479
  return utils.PathJoin(pathutils.LOG_OS_DIR, base)
1480

    
1481

    
1482
def InstanceOsAdd(instance, reinstall, debug):
1483
  """Add an OS to an instance.
1484

1485
  @type instance: L{objects.Instance}
1486
  @param instance: Instance whose OS is to be installed
1487
  @type reinstall: boolean
1488
  @param reinstall: whether this is an instance reinstall
1489
  @type debug: integer
1490
  @param debug: debug level, passed to the OS scripts
1491
  @rtype: None
1492

1493
  """
1494
  inst_os = OSFromDisk(instance.os)
1495

    
1496
  create_env = OSEnvironment(instance, inst_os, debug)
1497
  if reinstall:
1498
    create_env["INSTANCE_REINSTALL"] = "1"
1499

    
1500
  logfile = _InstanceLogName("add", instance.os, instance.name, None)
1501

    
1502
  result = utils.RunCmd([inst_os.create_script], env=create_env,
1503
                        cwd=inst_os.path, output=logfile, reset_env=True)
1504
  if result.failed:
1505
    logging.error("os create command '%s' returned error: %s, logfile: %s,"
1506
                  " output: %s", result.cmd, result.fail_reason, logfile,
1507
                  result.output)
1508
    lines = [utils.SafeEncode(val)
1509
             for val in utils.TailFile(logfile, lines=20)]
1510
    _Fail("OS create script failed (%s), last lines in the"
1511
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1512

    
1513

    
1514
def RunRenameInstance(instance, old_name, debug):
1515
  """Run the OS rename script for an instance.
1516

1517
  @type instance: L{objects.Instance}
1518
  @param instance: Instance whose OS is to be installed
1519
  @type old_name: string
1520
  @param old_name: previous instance name
1521
  @type debug: integer
1522
  @param debug: debug level, passed to the OS scripts
1523
  @rtype: boolean
1524
  @return: the success of the operation
1525

1526
  """
1527
  inst_os = OSFromDisk(instance.os)
1528

    
1529
  rename_env = OSEnvironment(instance, inst_os, debug)
1530
  rename_env["OLD_INSTANCE_NAME"] = old_name
1531

    
1532
  logfile = _InstanceLogName("rename", instance.os,
1533
                             "%s-%s" % (old_name, instance.name), None)
1534

    
1535
  result = utils.RunCmd([inst_os.rename_script], env=rename_env,
1536
                        cwd=inst_os.path, output=logfile, reset_env=True)
1537

    
1538
  if result.failed:
1539
    logging.error("os create command '%s' returned error: %s output: %s",
1540
                  result.cmd, result.fail_reason, result.output)
1541
    lines = [utils.SafeEncode(val)
1542
             for val in utils.TailFile(logfile, lines=20)]
1543
    _Fail("OS rename script failed (%s), last lines in the"
1544
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1545

    
1546

    
1547
def _GetBlockDevSymlinkPath(instance_name, idx, _dir=None):
1548
  """Returns symlink path for block device.
1549

1550
  """
1551
  if _dir is None:
1552
    _dir = pathutils.DISK_LINKS_DIR
1553

    
1554
  return utils.PathJoin(_dir,
1555
                        ("%s%s%s" %
1556
                         (instance_name, constants.DISK_SEPARATOR, idx)))
1557

    
1558

    
1559
def _SymlinkBlockDev(instance_name, device_path, idx):
1560
  """Set up symlinks to a instance's block device.
1561

1562
  This is an auxiliary function run when an instance is start (on the primary
1563
  node) or when an instance is migrated (on the target node).
1564

1565

1566
  @param instance_name: the name of the target instance
1567
  @param device_path: path of the physical block device, on the node
1568
  @param idx: the disk index
1569
  @return: absolute path to the disk's symlink
1570

1571
  """
1572
  link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1573
  try:
1574
    os.symlink(device_path, link_name)
1575
  except OSError, err:
1576
    if err.errno == errno.EEXIST:
1577
      if (not os.path.islink(link_name) or
1578
          os.readlink(link_name) != device_path):
1579
        os.remove(link_name)
1580
        os.symlink(device_path, link_name)
1581
    else:
1582
      raise
1583

    
1584
  return link_name
1585

    
1586

    
1587
def _RemoveBlockDevLinks(instance_name, disks):
1588
  """Remove the block device symlinks belonging to the given instance.
1589

1590
  """
1591
  for idx, _ in enumerate(disks):
1592
    link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1593
    if os.path.islink(link_name):
1594
      try:
1595
        os.remove(link_name)
1596
      except OSError:
1597
        logging.exception("Can't remove symlink '%s'", link_name)
1598

    
1599

    
1600
def _GatherAndLinkBlockDevs(instance):
1601
  """Set up an instance's block device(s).
1602

1603
  This is run on the primary node at instance startup. The block
1604
  devices must be already assembled.
1605

1606
  @type instance: L{objects.Instance}
1607
  @param instance: the instance whose disks we shoul assemble
1608
  @rtype: list
1609
  @return: list of (disk_object, device_path)
1610

1611
  """
1612
  block_devices = []
1613
  for idx, disk in enumerate(instance.disks):
1614
    device = _RecursiveFindBD(disk)
1615
    if device is None:
1616
      raise errors.BlockDeviceError("Block device '%s' is not set up." %
1617
                                    str(disk))
1618
    device.Open()
1619
    try:
1620
      link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1621
    except OSError, e:
1622
      raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1623
                                    e.strerror)
1624

    
1625
    block_devices.append((disk, link_name))
1626

    
1627
  return block_devices
1628

    
1629

    
1630
def StartInstance(instance, startup_paused, reason, store_reason=True):
1631
  """Start an instance.
1632

1633
  @type instance: L{objects.Instance}
1634
  @param instance: the instance object
1635
  @type startup_paused: bool
1636
  @param instance: pause instance at startup?
1637
  @type reason: list of reasons
1638
  @param reason: the reason trail for this startup
1639
  @type store_reason: boolean
1640
  @param store_reason: whether to store the shutdown reason trail on file
1641
  @rtype: None
1642

1643
  """
1644
  running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1645
                                                   instance.hvparams)
1646

    
1647
  if instance.name in running_instances:
1648
    logging.info("Instance %s already running, not starting", instance.name)
1649
    return
1650

    
1651
  try:
1652
    block_devices = _GatherAndLinkBlockDevs(instance)
1653
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
1654
    hyper.StartInstance(instance, block_devices, startup_paused)
1655
    if store_reason:
1656
      _StoreInstReasonTrail(instance.name, reason)
1657
  except errors.BlockDeviceError, err:
1658
    _Fail("Block device error: %s", err, exc=True)
1659
  except errors.HypervisorError, err:
1660
    _RemoveBlockDevLinks(instance.name, instance.disks)
1661
    _Fail("Hypervisor error: %s", err, exc=True)
1662

    
1663

    
1664
def InstanceShutdown(instance, timeout, reason, store_reason=True):
1665
  """Shut an instance down.
1666

1667
  @note: this functions uses polling with a hardcoded timeout.
1668

1669
  @type instance: L{objects.Instance}
1670
  @param instance: the instance object
1671
  @type timeout: integer
1672
  @param timeout: maximum timeout for soft shutdown
1673
  @type reason: list of reasons
1674
  @param reason: the reason trail for this shutdown
1675
  @type store_reason: boolean
1676
  @param store_reason: whether to store the shutdown reason trail on file
1677
  @rtype: None
1678

1679
  """
1680
  hv_name = instance.hypervisor
1681
  hyper = hypervisor.GetHypervisor(hv_name)
1682
  iname = instance.name
1683

    
1684
  if instance.name not in hyper.ListInstances(instance.hvparams):
1685
    logging.info("Instance %s not running, doing nothing", iname)
1686
    return
1687

    
1688
  class _TryShutdown:
1689
    def __init__(self):
1690
      self.tried_once = False
1691

    
1692
    def __call__(self):
1693
      if iname not in hyper.ListInstances(instance.hvparams):
1694
        return
1695

    
1696
      try:
1697
        hyper.StopInstance(instance, retry=self.tried_once)
1698
        if store_reason:
1699
          _StoreInstReasonTrail(instance.name, reason)
1700
      except errors.HypervisorError, err:
1701
        if iname not in hyper.ListInstances(instance.hvparams):
1702
          # if the instance is no longer existing, consider this a
1703
          # success and go to cleanup
1704
          return
1705

    
1706
        _Fail("Failed to stop instance %s: %s", iname, err)
1707

    
1708
      self.tried_once = True
1709

    
1710
      raise utils.RetryAgain()
1711

    
1712
  try:
1713
    utils.Retry(_TryShutdown(), 5, timeout)
1714
  except utils.RetryTimeout:
1715
    # the shutdown did not succeed
1716
    logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1717

    
1718
    try:
1719
      hyper.StopInstance(instance, force=True)
1720
    except errors.HypervisorError, err:
1721
      if iname in hyper.ListInstances(instance.hvparams):
1722
        # only raise an error if the instance still exists, otherwise
1723
        # the error could simply be "instance ... unknown"!
1724
        _Fail("Failed to force stop instance %s: %s", iname, err)
1725

    
1726
    time.sleep(1)
1727

    
1728
    if iname in hyper.ListInstances(instance.hvparams):
1729
      _Fail("Could not shutdown instance %s even by destroy", iname)
1730

    
1731
  try:
1732
    hyper.CleanupInstance(instance.name)
1733
  except errors.HypervisorError, err:
1734
    logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1735

    
1736
  _RemoveBlockDevLinks(iname, instance.disks)
1737

    
1738

    
1739
def InstanceReboot(instance, reboot_type, shutdown_timeout, reason):
1740
  """Reboot an instance.
1741

1742
  @type instance: L{objects.Instance}
1743
  @param instance: the instance object to reboot
1744
  @type reboot_type: str
1745
  @param reboot_type: the type of reboot, one the following
1746
    constants:
1747
      - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1748
        instance OS, do not recreate the VM
1749
      - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1750
        restart the VM (at the hypervisor level)
1751
      - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1752
        not accepted here, since that mode is handled differently, in
1753
        cmdlib, and translates into full stop and start of the
1754
        instance (instead of a call_instance_reboot RPC)
1755
  @type shutdown_timeout: integer
1756
  @param shutdown_timeout: maximum timeout for soft shutdown
1757
  @type reason: list of reasons
1758
  @param reason: the reason trail for this reboot
1759
  @rtype: None
1760

1761
  """
1762
  running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1763
                                                   instance.hvparams)
1764

    
1765
  if instance.name not in running_instances:
1766
    _Fail("Cannot reboot instance %s that is not running", instance.name)
1767

    
1768
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1769
  if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1770
    try:
1771
      hyper.RebootInstance(instance)
1772
    except errors.HypervisorError, err:
1773
      _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1774
  elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1775
    try:
1776
      InstanceShutdown(instance, shutdown_timeout, reason, store_reason=False)
1777
      result = StartInstance(instance, False, reason, store_reason=False)
1778
      _StoreInstReasonTrail(instance.name, reason)
1779
      return result
1780
    except errors.HypervisorError, err:
1781
      _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1782
  else:
1783
    _Fail("Invalid reboot_type received: %s", reboot_type)
1784

    
1785

    
1786
def InstanceBalloonMemory(instance, memory):
1787
  """Resize an instance's memory.
1788

1789
  @type instance: L{objects.Instance}
1790
  @param instance: the instance object
1791
  @type memory: int
1792
  @param memory: new memory amount in MB
1793
  @rtype: None
1794

1795
  """
1796
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1797
  running = hyper.ListInstances(instance.hvparams)
1798
  if instance.name not in running:
1799
    logging.info("Instance %s is not running, cannot balloon", instance.name)
1800
    return
1801
  try:
1802
    hyper.BalloonInstanceMemory(instance, memory)
1803
  except errors.HypervisorError, err:
1804
    _Fail("Failed to balloon instance memory: %s", err, exc=True)
1805

    
1806

    
1807
def MigrationInfo(instance):
1808
  """Gather information about an instance to be migrated.
1809

1810
  @type instance: L{objects.Instance}
1811
  @param instance: the instance definition
1812

1813
  """
1814
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1815
  try:
1816
    info = hyper.MigrationInfo(instance)
1817
  except errors.HypervisorError, err:
1818
    _Fail("Failed to fetch migration information: %s", err, exc=True)
1819
  return info
1820

    
1821

    
1822
def AcceptInstance(instance, info, target):
1823
  """Prepare the node to accept an instance.
1824

1825
  @type instance: L{objects.Instance}
1826
  @param instance: the instance definition
1827
  @type info: string/data (opaque)
1828
  @param info: migration information, from the source node
1829
  @type target: string
1830
  @param target: target host (usually ip), on this node
1831

1832
  """
1833
  # TODO: why is this required only for DTS_EXT_MIRROR?
1834
  if instance.disk_template in constants.DTS_EXT_MIRROR:
1835
    # Create the symlinks, as the disks are not active
1836
    # in any way
1837
    try:
1838
      _GatherAndLinkBlockDevs(instance)
1839
    except errors.BlockDeviceError, err:
1840
      _Fail("Block device error: %s", err, exc=True)
1841

    
1842
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1843
  try:
1844
    hyper.AcceptInstance(instance, info, target)
1845
  except errors.HypervisorError, err:
1846
    if instance.disk_template in constants.DTS_EXT_MIRROR:
1847
      _RemoveBlockDevLinks(instance.name, instance.disks)
1848
    _Fail("Failed to accept instance: %s", err, exc=True)
1849

    
1850

    
1851
def FinalizeMigrationDst(instance, info, success):
1852
  """Finalize any preparation to accept an instance.
1853

1854
  @type instance: L{objects.Instance}
1855
  @param instance: the instance definition
1856
  @type info: string/data (opaque)
1857
  @param info: migration information, from the source node
1858
  @type success: boolean
1859
  @param success: whether the migration was a success or a failure
1860

1861
  """
1862
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1863
  try:
1864
    hyper.FinalizeMigrationDst(instance, info, success)
1865
  except errors.HypervisorError, err:
1866
    _Fail("Failed to finalize migration on the target node: %s", err, exc=True)
1867

    
1868

    
1869
def MigrateInstance(cluster_name, instance, target, live):
1870
  """Migrates an instance to another node.
1871

1872
  @type cluster_name: string
1873
  @param cluster_name: name of the cluster
1874
  @type instance: L{objects.Instance}
1875
  @param instance: the instance definition
1876
  @type target: string
1877
  @param target: the target node name
1878
  @type live: boolean
1879
  @param live: whether the migration should be done live or not (the
1880
      interpretation of this parameter is left to the hypervisor)
1881
  @raise RPCFail: if migration fails for some reason
1882

1883
  """
1884
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1885

    
1886
  try:
1887
    hyper.MigrateInstance(cluster_name, instance, target, live)
1888
  except errors.HypervisorError, err:
1889
    _Fail("Failed to migrate instance: %s", err, exc=True)
1890

    
1891

    
1892
def FinalizeMigrationSource(instance, success, live):
1893
  """Finalize the instance migration on the source node.
1894

1895
  @type instance: L{objects.Instance}
1896
  @param instance: the instance definition of the migrated instance
1897
  @type success: bool
1898
  @param success: whether the migration succeeded or not
1899
  @type live: bool
1900
  @param live: whether the user requested a live migration or not
1901
  @raise RPCFail: If the execution fails for some reason
1902

1903
  """
1904
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1905

    
1906
  try:
1907
    hyper.FinalizeMigrationSource(instance, success, live)
1908
  except Exception, err:  # pylint: disable=W0703
1909
    _Fail("Failed to finalize the migration on the source node: %s", err,
1910
          exc=True)
1911

    
1912

    
1913
def GetMigrationStatus(instance):
1914
  """Get the migration status
1915

1916
  @type instance: L{objects.Instance}
1917
  @param instance: the instance that is being migrated
1918
  @rtype: L{objects.MigrationStatus}
1919
  @return: the status of the current migration (one of
1920
           L{constants.HV_MIGRATION_VALID_STATUSES}), plus any additional
1921
           progress info that can be retrieved from the hypervisor
1922
  @raise RPCFail: If the migration status cannot be retrieved
1923

1924
  """
1925
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1926
  try:
1927
    return hyper.GetMigrationStatus(instance)
1928
  except Exception, err:  # pylint: disable=W0703
1929
    _Fail("Failed to get migration status: %s", err, exc=True)
1930

    
1931

    
1932
def BlockdevCreate(disk, size, owner, on_primary, info, excl_stor):
1933
  """Creates a block device for an instance.
1934

1935
  @type disk: L{objects.Disk}
1936
  @param disk: the object describing the disk we should create
1937
  @type size: int
1938
  @param size: the size of the physical underlying device, in MiB
1939
  @type owner: str
1940
  @param owner: the name of the instance for which disk is created,
1941
      used for device cache data
1942
  @type on_primary: boolean
1943
  @param on_primary:  indicates if it is the primary node or not
1944
  @type info: string
1945
  @param info: string that will be sent to the physical device
1946
      creation, used for example to set (LVM) tags on LVs
1947
  @type excl_stor: boolean
1948
  @param excl_stor: Whether exclusive_storage is active
1949

1950
  @return: the new unique_id of the device (this can sometime be
1951
      computed only after creation), or None. On secondary nodes,
1952
      it's not required to return anything.
1953

1954
  """
1955
  # TODO: remove the obsolete "size" argument
1956
  # pylint: disable=W0613
1957
  clist = []
1958
  if disk.children:
1959
    for child in disk.children:
1960
      try:
1961
        crdev = _RecursiveAssembleBD(child, owner, on_primary)
1962
      except errors.BlockDeviceError, err:
1963
        _Fail("Can't assemble device %s: %s", child, err)
1964
      if on_primary or disk.AssembleOnSecondary():
1965
        # we need the children open in case the device itself has to
1966
        # be assembled
1967
        try:
1968
          # pylint: disable=E1103
1969
          crdev.Open()
1970
        except errors.BlockDeviceError, err:
1971
          _Fail("Can't make child '%s' read-write: %s", child, err)
1972
      clist.append(crdev)
1973

    
1974
  try:
1975
    device = bdev.Create(disk, clist, excl_stor)
1976
  except errors.BlockDeviceError, err:
1977
    _Fail("Can't create block device: %s", err)
1978

    
1979
  if on_primary or disk.AssembleOnSecondary():
1980
    try:
1981
      device.Assemble()
1982
    except errors.BlockDeviceError, err:
1983
      _Fail("Can't assemble device after creation, unusual event: %s", err)
1984
    if on_primary or disk.OpenOnSecondary():
1985
      try:
1986
        device.Open(force=True)
1987
      except errors.BlockDeviceError, err:
1988
        _Fail("Can't make device r/w after creation, unusual event: %s", err)
1989
    DevCacheManager.UpdateCache(device.dev_path, owner,
1990
                                on_primary, disk.iv_name)
1991

    
1992
  device.SetInfo(info)
1993

    
1994
  return device.unique_id
1995

    
1996

    
1997
def _WipeDevice(path, offset, size):
1998
  """This function actually wipes the device.
1999

2000
  @param path: The path to the device to wipe
2001
  @param offset: The offset in MiB in the file
2002
  @param size: The size in MiB to write
2003

2004
  """
2005
  # Internal sizes are always in Mebibytes; if the following "dd" command
2006
  # should use a different block size the offset and size given to this
2007
  # function must be adjusted accordingly before being passed to "dd".
2008
  block_size = 1024 * 1024
2009

    
2010
  cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
2011
         "bs=%s" % block_size, "oflag=direct", "of=%s" % path,
2012
         "count=%d" % size]
2013
  result = utils.RunCmd(cmd)
2014

    
2015
  if result.failed:
2016
    _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
2017
          result.fail_reason, result.output)
2018

    
2019

    
2020
def BlockdevWipe(disk, offset, size):
2021
  """Wipes a block device.
2022

2023
  @type disk: L{objects.Disk}
2024
  @param disk: the disk object we want to wipe
2025
  @type offset: int
2026
  @param offset: The offset in MiB in the file
2027
  @type size: int
2028
  @param size: The size in MiB to write
2029

2030
  """
2031
  try:
2032
    rdev = _RecursiveFindBD(disk)
2033
  except errors.BlockDeviceError:
2034
    rdev = None
2035

    
2036
  if not rdev:
2037
    _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
2038

    
2039
  # Do cross verify some of the parameters
2040
  if offset < 0:
2041
    _Fail("Negative offset")
2042
  if size < 0:
2043
    _Fail("Negative size")
2044
  if offset > rdev.size:
2045
    _Fail("Offset is bigger than device size")
2046
  if (offset + size) > rdev.size:
2047
    _Fail("The provided offset and size to wipe is bigger than device size")
2048

    
2049
  _WipeDevice(rdev.dev_path, offset, size)
2050

    
2051

    
2052
def BlockdevPauseResumeSync(disks, pause):
2053
  """Pause or resume the sync of the block device.
2054

2055
  @type disks: list of L{objects.Disk}
2056
  @param disks: the disks object we want to pause/resume
2057
  @type pause: bool
2058
  @param pause: Wheater to pause or resume
2059

2060
  """
2061
  success = []
2062
  for disk in disks:
2063
    try:
2064
      rdev = _RecursiveFindBD(disk)
2065
    except errors.BlockDeviceError:
2066
      rdev = None
2067

    
2068
    if not rdev:
2069
      success.append((False, ("Cannot change sync for device %s:"
2070
                              " device not found" % disk.iv_name)))
2071
      continue
2072

    
2073
    result = rdev.PauseResumeSync(pause)
2074

    
2075
    if result:
2076
      success.append((result, None))
2077
    else:
2078
      if pause:
2079
        msg = "Pause"
2080
      else:
2081
        msg = "Resume"
2082
      success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
2083

    
2084
  return success
2085

    
2086

    
2087
def BlockdevRemove(disk):
2088
  """Remove a block device.
2089

2090
  @note: This is intended to be called recursively.
2091

2092
  @type disk: L{objects.Disk}
2093
  @param disk: the disk object we should remove
2094
  @rtype: boolean
2095
  @return: the success of the operation
2096

2097
  """
2098
  msgs = []
2099
  try:
2100
    rdev = _RecursiveFindBD(disk)
2101
  except errors.BlockDeviceError, err:
2102
    # probably can't attach
2103
    logging.info("Can't attach to device %s in remove", disk)
2104
    rdev = None
2105
  if rdev is not None:
2106
    r_path = rdev.dev_path
2107
    try:
2108
      rdev.Remove()
2109
    except errors.BlockDeviceError, err:
2110
      msgs.append(str(err))
2111
    if not msgs:
2112
      DevCacheManager.RemoveCache(r_path)
2113

    
2114
  if disk.children:
2115
    for child in disk.children:
2116
      try:
2117
        BlockdevRemove(child)
2118
      except RPCFail, err:
2119
        msgs.append(str(err))
2120

    
2121
  if msgs:
2122
    _Fail("; ".join(msgs))
2123

    
2124

    
2125
def _RecursiveAssembleBD(disk, owner, as_primary):
2126
  """Activate a block device for an instance.
2127

2128
  This is run on the primary and secondary nodes for an instance.
2129

2130
  @note: this function is called recursively.
2131

2132
  @type disk: L{objects.Disk}
2133
  @param disk: the disk we try to assemble
2134
  @type owner: str
2135
  @param owner: the name of the instance which owns the disk
2136
  @type as_primary: boolean
2137
  @param as_primary: if we should make the block device
2138
      read/write
2139

2140
  @return: the assembled device or None (in case no device
2141
      was assembled)
2142
  @raise errors.BlockDeviceError: in case there is an error
2143
      during the activation of the children or the device
2144
      itself
2145

2146
  """
2147
  children = []
2148
  if disk.children:
2149
    mcn = disk.ChildrenNeeded()
2150
    if mcn == -1:
2151
      mcn = 0 # max number of Nones allowed
2152
    else:
2153
      mcn = len(disk.children) - mcn # max number of Nones
2154
    for chld_disk in disk.children:
2155
      try:
2156
        cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
2157
      except errors.BlockDeviceError, err:
2158
        if children.count(None) >= mcn:
2159
          raise
2160
        cdev = None
2161
        logging.error("Error in child activation (but continuing): %s",
2162
                      str(err))
2163
      children.append(cdev)
2164

    
2165
  if as_primary or disk.AssembleOnSecondary():
2166
    r_dev = bdev.Assemble(disk, children)
2167
    result = r_dev
2168
    if as_primary or disk.OpenOnSecondary():
2169
      r_dev.Open()
2170
    DevCacheManager.UpdateCache(r_dev.dev_path, owner,
2171
                                as_primary, disk.iv_name)
2172

    
2173
  else:
2174
    result = True
2175
  return result
2176

    
2177

    
2178
def BlockdevAssemble(disk, owner, as_primary, idx):
2179
  """Activate a block device for an instance.
2180

2181
  This is a wrapper over _RecursiveAssembleBD.
2182

2183
  @rtype: str or boolean
2184
  @return: a C{/dev/...} path for primary nodes, and
2185
      C{True} for secondary nodes
2186

2187
  """
2188
  try:
2189
    result = _RecursiveAssembleBD(disk, owner, as_primary)
2190
    if isinstance(result, BlockDev):
2191
      # pylint: disable=E1103
2192
      result = result.dev_path
2193
      if as_primary:
2194
        _SymlinkBlockDev(owner, result, idx)
2195
  except errors.BlockDeviceError, err:
2196
    _Fail("Error while assembling disk: %s", err, exc=True)
2197
  except OSError, err:
2198
    _Fail("Error while symlinking disk: %s", err, exc=True)
2199

    
2200
  return result
2201

    
2202

    
2203
def BlockdevShutdown(disk):
2204
  """Shut down a block device.
2205

2206
  First, if the device is assembled (Attach() is successful), then
2207
  the device is shutdown. Then the children of the device are
2208
  shutdown.
2209

2210
  This function is called recursively. Note that we don't cache the
2211
  children or such, as oppossed to assemble, shutdown of different
2212
  devices doesn't require that the upper device was active.
2213

2214
  @type disk: L{objects.Disk}
2215
  @param disk: the description of the disk we should
2216
      shutdown
2217
  @rtype: None
2218

2219
  """
2220
  msgs = []
2221
  r_dev = _RecursiveFindBD(disk)
2222
  if r_dev is not None:
2223
    r_path = r_dev.dev_path
2224
    try:
2225
      r_dev.Shutdown()
2226
      DevCacheManager.RemoveCache(r_path)
2227
    except errors.BlockDeviceError, err:
2228
      msgs.append(str(err))
2229

    
2230
  if disk.children:
2231
    for child in disk.children:
2232
      try:
2233
        BlockdevShutdown(child)
2234
      except RPCFail, err:
2235
        msgs.append(str(err))
2236

    
2237
  if msgs:
2238
    _Fail("; ".join(msgs))
2239

    
2240

    
2241
def BlockdevAddchildren(parent_cdev, new_cdevs):
2242
  """Extend a mirrored block device.
2243

2244
  @type parent_cdev: L{objects.Disk}
2245
  @param parent_cdev: the disk to which we should add children
2246
  @type new_cdevs: list of L{objects.Disk}
2247
  @param new_cdevs: the list of children which we should add
2248
  @rtype: None
2249

2250
  """
2251
  parent_bdev = _RecursiveFindBD(parent_cdev)
2252
  if parent_bdev is None:
2253
    _Fail("Can't find parent device '%s' in add children", parent_cdev)
2254
  new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
2255
  if new_bdevs.count(None) > 0:
2256
    _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
2257
  parent_bdev.AddChildren(new_bdevs)
2258

    
2259

    
2260
def BlockdevRemovechildren(parent_cdev, new_cdevs):
2261
  """Shrink a mirrored block device.
2262

2263
  @type parent_cdev: L{objects.Disk}
2264
  @param parent_cdev: the disk from which we should remove children
2265
  @type new_cdevs: list of L{objects.Disk}
2266
  @param new_cdevs: the list of children which we should remove
2267
  @rtype: None
2268

2269
  """
2270
  parent_bdev = _RecursiveFindBD(parent_cdev)
2271
  if parent_bdev is None:
2272
    _Fail("Can't find parent device '%s' in remove children", parent_cdev)
2273
  devs = []
2274
  for disk in new_cdevs:
2275
    rpath = disk.StaticDevPath()
2276
    if rpath is None:
2277
      bd = _RecursiveFindBD(disk)
2278
      if bd is None:
2279
        _Fail("Can't find device %s while removing children", disk)
2280
      else:
2281
        devs.append(bd.dev_path)
2282
    else:
2283
      if not utils.IsNormAbsPath(rpath):
2284
        _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
2285
      devs.append(rpath)
2286
  parent_bdev.RemoveChildren(devs)
2287

    
2288

    
2289
def BlockdevGetmirrorstatus(disks):
2290
  """Get the mirroring status of a list of devices.
2291

2292
  @type disks: list of L{objects.Disk}
2293
  @param disks: the list of disks which we should query
2294
  @rtype: disk
2295
  @return: List of L{objects.BlockDevStatus}, one for each disk
2296
  @raise errors.BlockDeviceError: if any of the disks cannot be
2297
      found
2298

2299
  """
2300
  stats = []
2301
  for dsk in disks:
2302
    rbd = _RecursiveFindBD(dsk)
2303
    if rbd is None:
2304
      _Fail("Can't find device %s", dsk)
2305

    
2306
    stats.append(rbd.CombinedSyncStatus())
2307

    
2308
  return stats
2309

    
2310

    
2311
def BlockdevGetmirrorstatusMulti(disks):
2312
  """Get the mirroring status of a list of devices.
2313

2314
  @type disks: list of L{objects.Disk}
2315
  @param disks: the list of disks which we should query
2316
  @rtype: disk
2317
  @return: List of tuples, (bool, status), one for each disk; bool denotes
2318
    success/failure, status is L{objects.BlockDevStatus} on success, string
2319
    otherwise
2320

2321
  """
2322
  result = []
2323
  for disk in disks:
2324
    try:
2325
      rbd = _RecursiveFindBD(disk)
2326
      if rbd is None:
2327
        result.append((False, "Can't find device %s" % disk))
2328
        continue
2329

    
2330
      status = rbd.CombinedSyncStatus()
2331
    except errors.BlockDeviceError, err:
2332
      logging.exception("Error while getting disk status")
2333
      result.append((False, str(err)))
2334
    else:
2335
      result.append((True, status))
2336

    
2337
  assert len(disks) == len(result)
2338

    
2339
  return result
2340

    
2341

    
2342
def _RecursiveFindBD(disk):
2343
  """Check if a device is activated.
2344

2345
  If so, return information about the real device.
2346

2347
  @type disk: L{objects.Disk}
2348
  @param disk: the disk object we need to find
2349

2350
  @return: None if the device can't be found,
2351
      otherwise the device instance
2352

2353
  """
2354
  children = []
2355
  if disk.children:
2356
    for chdisk in disk.children:
2357
      children.append(_RecursiveFindBD(chdisk))
2358

    
2359
  return bdev.FindDevice(disk, children)
2360

    
2361

    
2362
def _OpenRealBD(disk):
2363
  """Opens the underlying block device of a disk.
2364

2365
  @type disk: L{objects.Disk}
2366
  @param disk: the disk object we want to open
2367

2368
  """
2369
  real_disk = _RecursiveFindBD(disk)
2370
  if real_disk is None:
2371
    _Fail("Block device '%s' is not set up", disk)
2372

    
2373
  real_disk.Open()
2374

    
2375
  return real_disk
2376

    
2377

    
2378
def BlockdevFind(disk):
2379
  """Check if a device is activated.
2380

2381
  If it is, return information about the real device.
2382

2383
  @type disk: L{objects.Disk}
2384
  @param disk: the disk to find
2385
  @rtype: None or objects.BlockDevStatus
2386
  @return: None if the disk cannot be found, otherwise a the current
2387
           information
2388

2389
  """
2390
  try:
2391
    rbd = _RecursiveFindBD(disk)
2392
  except errors.BlockDeviceError, err:
2393
    _Fail("Failed to find device: %s", err, exc=True)
2394

    
2395
  if rbd is None:
2396
    return None
2397

    
2398
  return rbd.GetSyncStatus()
2399

    
2400

    
2401
def BlockdevGetdimensions(disks):
2402
  """Computes the size of the given disks.
2403

2404
  If a disk is not found, returns None instead.
2405

2406
  @type disks: list of L{objects.Disk}
2407
  @param disks: the list of disk to compute the size for
2408
  @rtype: list
2409
  @return: list with elements None if the disk cannot be found,
2410
      otherwise the pair (size, spindles), where spindles is None if the
2411
      device doesn't support that
2412

2413
  """
2414
  result = []
2415
  for cf in disks:
2416
    try:
2417
      rbd = _RecursiveFindBD(cf)
2418
    except errors.BlockDeviceError:
2419
      result.append(None)
2420
      continue
2421
    if rbd is None:
2422
      result.append(None)
2423
    else:
2424
      result.append(rbd.GetActualDimensions())
2425
  return result
2426

    
2427

    
2428
def BlockdevExport(disk, dest_node, dest_path, cluster_name):
2429
  """Export a block device to a remote node.
2430

2431
  @type disk: L{objects.Disk}
2432
  @param disk: the description of the disk to export
2433
  @type dest_node: str
2434
  @param dest_node: the destination node to export to
2435
  @type dest_path: str
2436
  @param dest_path: the destination path on the target node
2437
  @type cluster_name: str
2438
  @param cluster_name: the cluster name, needed for SSH hostalias
2439
  @rtype: None
2440

2441
  """
2442
  real_disk = _OpenRealBD(disk)
2443

    
2444
  # the block size on the read dd is 1MiB to match our units
2445
  expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
2446
                               "dd if=%s bs=1048576 count=%s",
2447
                               real_disk.dev_path, str(disk.size))
2448

    
2449
  # we set here a smaller block size as, due to ssh buffering, more
2450
  # than 64-128k will mostly ignored; we use nocreat to fail if the
2451
  # device is not already there or we pass a wrong path; we use
2452
  # notrunc to no attempt truncate on an LV device; we use oflag=dsync
2453
  # to not buffer too much memory; this means that at best, we flush
2454
  # every 64k, which will not be very fast
2455
  destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
2456
                                " oflag=dsync", dest_path)
2457

    
2458
  remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
2459
                                                   constants.SSH_LOGIN_USER,
2460
                                                   destcmd)
2461

    
2462
  # all commands have been checked, so we're safe to combine them
2463
  command = "|".join([expcmd, utils.ShellQuoteArgs(remotecmd)])
2464

    
2465
  result = utils.RunCmd(["bash", "-c", command])
2466

    
2467
  if result.failed:
2468
    _Fail("Disk copy command '%s' returned error: %s"
2469
          " output: %s", command, result.fail_reason, result.output)
2470

    
2471

    
2472
def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
2473
  """Write a file to the filesystem.
2474

2475
  This allows the master to overwrite(!) a file. It will only perform
2476
  the operation if the file belongs to a list of configuration files.
2477

2478
  @type file_name: str
2479
  @param file_name: the target file name
2480
  @type data: str
2481
  @param data: the new contents of the file
2482
  @type mode: int
2483
  @param mode: the mode to give the file (can be None)
2484
  @type uid: string
2485
  @param uid: the owner of the file
2486
  @type gid: string
2487
  @param gid: the group of the file
2488
  @type atime: float
2489
  @param atime: the atime to set on the file (can be None)
2490
  @type mtime: float
2491
  @param mtime: the mtime to set on the file (can be None)
2492
  @rtype: None
2493

2494
  """
2495
  file_name = vcluster.LocalizeVirtualPath(file_name)
2496

    
2497
  if not os.path.isabs(file_name):
2498
    _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
2499

    
2500
  if file_name not in _ALLOWED_UPLOAD_FILES:
2501
    _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
2502
          file_name)
2503

    
2504
  raw_data = _Decompress(data)
2505

    
2506
  if not (isinstance(uid, basestring) and isinstance(gid, basestring)):
2507
    _Fail("Invalid username/groupname type")
2508

    
2509
  getents = runtime.GetEnts()
2510
  uid = getents.LookupUser(uid)
2511
  gid = getents.LookupGroup(gid)
2512

    
2513
  utils.SafeWriteFile(file_name, None,
2514
                      data=raw_data, mode=mode, uid=uid, gid=gid,
2515
                      atime=atime, mtime=mtime)
2516

    
2517

    
2518
def RunOob(oob_program, command, node, timeout):
2519
  """Executes oob_program with given command on given node.
2520

2521
  @param oob_program: The path to the executable oob_program
2522
  @param command: The command to invoke on oob_program
2523
  @param node: The node given as an argument to the program
2524
  @param timeout: Timeout after which we kill the oob program
2525

2526
  @return: stdout
2527
  @raise RPCFail: If execution fails for some reason
2528

2529
  """
2530
  result = utils.RunCmd([oob_program, command, node], timeout=timeout)
2531

    
2532
  if result.failed:
2533
    _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
2534
          result.fail_reason, result.output)
2535

    
2536
  return result.stdout
2537

    
2538

    
2539
def _OSOndiskAPIVersion(os_dir):
2540
  """Compute and return the API version of a given OS.
2541

2542
  This function will try to read the API version of the OS residing in
2543
  the 'os_dir' directory.
2544

2545
  @type os_dir: str
2546
  @param os_dir: the directory in which we should look for the OS
2547
  @rtype: tuple
2548
  @return: tuple (status, data) with status denoting the validity and
2549
      data holding either the vaid versions or an error message
2550

2551
  """
2552
  api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
2553

    
2554
  try:
2555
    st = os.stat(api_file)
2556
  except EnvironmentError, err:
2557
    return False, ("Required file '%s' not found under path %s: %s" %
2558
                   (constants.OS_API_FILE, os_dir, utils.ErrnoOrStr(err)))
2559

    
2560
  if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2561
    return False, ("File '%s' in %s is not a regular file" %
2562
                   (constants.OS_API_FILE, os_dir))
2563

    
2564
  try:
2565
    api_versions = utils.ReadFile(api_file).splitlines()
2566
  except EnvironmentError, err:
2567
    return False, ("Error while reading the API version file at %s: %s" %
2568
                   (api_file, utils.ErrnoOrStr(err)))
2569

    
2570
  try:
2571
    api_versions = [int(version.strip()) for version in api_versions]
2572
  except (TypeError, ValueError), err:
2573
    return False, ("API version(s) can't be converted to integer: %s" %
2574
                   str(err))
2575

    
2576
  return True, api_versions
2577

    
2578

    
2579
def DiagnoseOS(top_dirs=None):
2580
  """Compute the validity for all OSes.
2581

2582
  @type top_dirs: list
2583
  @param top_dirs: the list of directories in which to
2584
      search (if not given defaults to
2585
      L{pathutils.OS_SEARCH_PATH})
2586
  @rtype: list of L{objects.OS}
2587
  @return: a list of tuples (name, path, status, diagnose, variants,
2588
      parameters, api_version) for all (potential) OSes under all
2589
      search paths, where:
2590
          - name is the (potential) OS name
2591
          - path is the full path to the OS
2592
          - status True/False is the validity of the OS
2593
          - diagnose is the error message for an invalid OS, otherwise empty
2594
          - variants is a list of supported OS variants, if any
2595
          - parameters is a list of (name, help) parameters, if any
2596
          - api_version is a list of support OS API versions
2597

2598
  """
2599
  if top_dirs is None:
2600
    top_dirs = pathutils.OS_SEARCH_PATH
2601

    
2602
  result = []
2603
  for dir_name in top_dirs:
2604
    if os.path.isdir(dir_name):
2605
      try:
2606
        f_names = utils.ListVisibleFiles(dir_name)
2607
      except EnvironmentError, err:
2608
        logging.exception("Can't list the OS directory %s: %s", dir_name, err)
2609
        break
2610
      for name in f_names:
2611
        os_path = utils.PathJoin(dir_name, name)
2612
        status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
2613
        if status:
2614
          diagnose = ""
2615
          variants = os_inst.supported_variants
2616
          parameters = os_inst.supported_parameters
2617
          api_versions = os_inst.api_versions
2618
        else:
2619
          diagnose = os_inst
2620
          variants = parameters = api_versions = []
2621
        result.append((name, os_path, status, diagnose, variants,
2622
                       parameters, api_versions))
2623

    
2624
  return result
2625

    
2626

    
2627
def _TryOSFromDisk(name, base_dir=None):
2628
  """Create an OS instance from disk.
2629

2630
  This function will return an OS instance if the given name is a
2631
  valid OS name.
2632

2633
  @type base_dir: string
2634
  @keyword base_dir: Base directory containing OS installations.
2635
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2636
  @rtype: tuple
2637
  @return: success and either the OS instance if we find a valid one,
2638
      or error message
2639

2640
  """
2641
  if base_dir is None:
2642
    os_dir = utils.FindFile(name, pathutils.OS_SEARCH_PATH, os.path.isdir)
2643
  else:
2644
    os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2645

    
2646
  if os_dir is None:
2647
    return False, "Directory for OS %s not found in search path" % name
2648

    
2649
  status, api_versions = _OSOndiskAPIVersion(os_dir)
2650
  if not status:
2651
    # push the error up
2652
    return status, api_versions
2653

    
2654
  if not constants.OS_API_VERSIONS.intersection(api_versions):
2655
    return False, ("API version mismatch for path '%s': found %s, want %s." %
2656
                   (os_dir, api_versions, constants.OS_API_VERSIONS))
2657

    
2658
  # OS Files dictionary, we will populate it with the absolute path
2659
  # names; if the value is True, then it is a required file, otherwise
2660
  # an optional one
2661
  os_files = dict.fromkeys(constants.OS_SCRIPTS, True)
2662

    
2663
  if max(api_versions) >= constants.OS_API_V15:
2664
    os_files[constants.OS_VARIANTS_FILE] = False
2665

    
2666
  if max(api_versions) >= constants.OS_API_V20:
2667
    os_files[constants.OS_PARAMETERS_FILE] = True
2668
  else:
2669
    del os_files[constants.OS_SCRIPT_VERIFY]
2670

    
2671
  for (filename, required) in os_files.items():
2672
    os_files[filename] = utils.PathJoin(os_dir, filename)
2673

    
2674
    try:
2675
      st = os.stat(os_files[filename])
2676
    except EnvironmentError, err:
2677
      if err.errno == errno.ENOENT and not required:
2678
        del os_files[filename]
2679
        continue
2680
      return False, ("File '%s' under path '%s' is missing (%s)" %
2681
                     (filename, os_dir, utils.ErrnoOrStr(err)))
2682

    
2683
    if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2684
      return False, ("File '%s' under path '%s' is not a regular file" %
2685
                     (filename, os_dir))
2686

    
2687
    if filename in constants.OS_SCRIPTS:
2688
      if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2689
        return False, ("File '%s' under path '%s' is not executable" %
2690
                       (filename, os_dir))
2691

    
2692
  variants = []
2693
  if constants.OS_VARIANTS_FILE in os_files:
2694
    variants_file = os_files[constants.OS_VARIANTS_FILE]
2695
    try:
2696
      variants = \
2697
        utils.FilterEmptyLinesAndComments(utils.ReadFile(variants_file))
2698
    except EnvironmentError, err:
2699
      # we accept missing files, but not other errors
2700
      if err.errno != errno.ENOENT:
2701
        return False, ("Error while reading the OS variants file at %s: %s" %
2702
                       (variants_file, utils.ErrnoOrStr(err)))
2703

    
2704
  parameters = []
2705
  if constants.OS_PARAMETERS_FILE in os_files:
2706
    parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2707
    try:
2708
      parameters = utils.ReadFile(parameters_file).splitlines()
2709
    except EnvironmentError, err:
2710
      return False, ("Error while reading the OS parameters file at %s: %s" %
2711
                     (parameters_file, utils.ErrnoOrStr(err)))
2712
    parameters = [v.split(None, 1) for v in parameters]
2713

    
2714
  os_obj = objects.OS(name=name, path=os_dir,
2715
                      create_script=os_files[constants.OS_SCRIPT_CREATE],
2716
                      export_script=os_files[constants.OS_SCRIPT_EXPORT],
2717
                      import_script=os_files[constants.OS_SCRIPT_IMPORT],
2718
                      rename_script=os_files[constants.OS_SCRIPT_RENAME],
2719
                      verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2720
                                                 None),
2721
                      supported_variants=variants,
2722
                      supported_parameters=parameters,
2723
                      api_versions=api_versions)
2724
  return True, os_obj
2725

    
2726

    
2727
def OSFromDisk(name, base_dir=None):
2728
  """Create an OS instance from disk.
2729

2730
  This function will return an OS instance if the given name is a
2731
  valid OS name. Otherwise, it will raise an appropriate
2732
  L{RPCFail} exception, detailing why this is not a valid OS.
2733

2734
  This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2735
  an exception but returns true/false status data.
2736

2737
  @type base_dir: string
2738
  @keyword base_dir: Base directory containing OS installations.
2739
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2740
  @rtype: L{objects.OS}
2741
  @return: the OS instance if we find a valid one
2742
  @raise RPCFail: if we don't find a valid OS
2743

2744
  """
2745
  name_only = objects.OS.GetName(name)
2746
  status, payload = _TryOSFromDisk(name_only, base_dir)
2747

    
2748
  if not status:
2749
    _Fail(payload)
2750

    
2751
  return payload
2752

    
2753

    
2754
def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2755
  """Calculate the basic environment for an os script.
2756

2757
  @type os_name: str
2758
  @param os_name: full operating system name (including variant)
2759
  @type inst_os: L{objects.OS}
2760
  @param inst_os: operating system for which the environment is being built
2761
  @type os_params: dict
2762
  @param os_params: the OS parameters
2763
  @type debug: integer
2764
  @param debug: debug level (0 or 1, for OS Api 10)
2765
  @rtype: dict
2766
  @return: dict of environment variables
2767
  @raise errors.BlockDeviceError: if the block device
2768
      cannot be found
2769

2770
  """
2771
  result = {}
2772
  api_version = \
2773
    max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2774
  result["OS_API_VERSION"] = "%d" % api_version
2775
  result["OS_NAME"] = inst_os.name
2776
  result["DEBUG_LEVEL"] = "%d" % debug
2777

    
2778
  # OS variants
2779
  if api_version >= constants.OS_API_V15 and inst_os.supported_variants:
2780
    variant = objects.OS.GetVariant(os_name)
2781
    if not variant:
2782
      variant = inst_os.supported_variants[0]
2783
  else:
2784
    variant = ""
2785
  result["OS_VARIANT"] = variant
2786

    
2787
  # OS params
2788
  for pname, pvalue in os_params.items():
2789
    result["OSP_%s" % pname.upper()] = pvalue
2790

    
2791
  # Set a default path otherwise programs called by OS scripts (or
2792
  # even hooks called from OS scripts) might break, and we don't want
2793
  # to have each script require setting a PATH variable
2794
  result["PATH"] = constants.HOOKS_PATH
2795

    
2796
  return result
2797

    
2798

    
2799
def OSEnvironment(instance, inst_os, debug=0):
2800
  """Calculate the environment for an os script.
2801

2802
  @type instance: L{objects.Instance}
2803
  @param instance: target instance for the os script run
2804
  @type inst_os: L{objects.OS}
2805
  @param inst_os: operating system for which the environment is being built
2806
  @type debug: integer
2807
  @param debug: debug level (0 or 1, for OS Api 10)
2808
  @rtype: dict
2809
  @return: dict of environment variables
2810
  @raise errors.BlockDeviceError: if the block device
2811
      cannot be found
2812

2813
  """
2814
  result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2815

    
2816
  for attr in ["name", "os", "uuid", "ctime", "mtime", "primary_node"]:
2817
    result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2818

    
2819
  result["HYPERVISOR"] = instance.hypervisor
2820
  result["DISK_COUNT"] = "%d" % len(instance.disks)
2821
  result["NIC_COUNT"] = "%d" % len(instance.nics)
2822
  result["INSTANCE_SECONDARY_NODES"] = \
2823
      ("%s" % " ".join(instance.secondary_nodes))
2824

    
2825
  # Disks
2826
  for idx, disk in enumerate(instance.disks):
2827
    real_disk = _OpenRealBD(disk)
2828
    result["DISK_%d_PATH" % idx] = real_disk.dev_path
2829
    result["DISK_%d_ACCESS" % idx] = disk.mode
2830
    result["DISK_%d_UUID" % idx] = disk.uuid
2831
    if disk.name:
2832
      result["DISK_%d_NAME" % idx] = disk.name
2833
    if constants.HV_DISK_TYPE in instance.hvparams:
2834
      result["DISK_%d_FRONTEND_TYPE" % idx] = \
2835
        instance.hvparams[constants.HV_DISK_TYPE]
2836
    if disk.dev_type in constants.LDS_BLOCK:
2837
      result["DISK_%d_BACKEND_TYPE" % idx] = "block"
2838
    elif disk.dev_type == constants.LD_FILE:
2839
      result["DISK_%d_BACKEND_TYPE" % idx] = \
2840
        "file:%s" % disk.physical_id[0]
2841

    
2842
  # NICs
2843
  for idx, nic in enumerate(instance.nics):
2844
    result["NIC_%d_MAC" % idx] = nic.mac
2845
    result["NIC_%d_UUID" % idx] = nic.uuid
2846
    if nic.name:
2847
      result["NIC_%d_NAME" % idx] = nic.name
2848
    if nic.ip:
2849
      result["NIC_%d_IP" % idx] = nic.ip
2850
    result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE]
2851
    if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2852
      result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK]
2853
    if nic.nicparams[constants.NIC_LINK]:
2854
      result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK]
2855
    if nic.netinfo:
2856
      nobj = objects.Network.FromDict(nic.netinfo)
2857
      result.update(nobj.HooksDict("NIC_%d_" % idx))
2858
    if constants.HV_NIC_TYPE in instance.hvparams:
2859
      result["NIC_%d_FRONTEND_TYPE" % idx] = \
2860
        instance.hvparams[constants.HV_NIC_TYPE]
2861

    
2862
  # HV/BE params
2863
  for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2864
    for key, value in source.items():
2865
      result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2866

    
2867
  return result
2868

    
2869

    
2870
def DiagnoseExtStorage(top_dirs=None):
2871
  """Compute the validity for all ExtStorage Providers.
2872

2873
  @type top_dirs: list
2874
  @param top_dirs: the list of directories in which to
2875
      search (if not given defaults to
2876
      L{pathutils.ES_SEARCH_PATH})
2877
  @rtype: list of L{objects.ExtStorage}
2878
  @return: a list of tuples (name, path, status, diagnose, parameters)
2879
      for all (potential) ExtStorage Providers under all
2880
      search paths, where:
2881
          - name is the (potential) ExtStorage Provider
2882
          - path is the full path to the ExtStorage Provider
2883
          - status True/False is the validity of the ExtStorage Provider
2884
          - diagnose is the error message for an invalid ExtStorage Provider,
2885
            otherwise empty
2886
          - parameters is a list of (name, help) parameters, if any
2887

2888
  """
2889
  if top_dirs is None:
2890
    top_dirs = pathutils.ES_SEARCH_PATH
2891

    
2892
  result = []
2893
  for dir_name in top_dirs:
2894
    if os.path.isdir(dir_name):
2895
      try:
2896
        f_names = utils.ListVisibleFiles(dir_name)
2897
      except EnvironmentError, err:
2898
        logging.exception("Can't list the ExtStorage directory %s: %s",
2899
                          dir_name, err)
2900
        break
2901
      for name in f_names:
2902
        es_path = utils.PathJoin(dir_name, name)
2903
        status, es_inst = bdev.ExtStorageFromDisk(name, base_dir=dir_name)
2904
        if status:
2905
          diagnose = ""
2906
          parameters = es_inst.supported_parameters
2907
        else:
2908
          diagnose = es_inst
2909
          parameters = []
2910
        result.append((name, es_path, status, diagnose, parameters))
2911

    
2912
  return result
2913

    
2914

    
2915
def BlockdevGrow(disk, amount, dryrun, backingstore, excl_stor):
2916
  """Grow a stack of block devices.
2917

2918
  This function is called recursively, with the childrens being the
2919
  first ones to resize.
2920

2921
  @type disk: L{objects.Disk}
2922
  @param disk: the disk to be grown
2923
  @type amount: integer
2924
  @param amount: the amount (in mebibytes) to grow with
2925
  @type dryrun: boolean
2926
  @param dryrun: whether to execute the operation in simulation mode
2927
      only, without actually increasing the size
2928
  @param backingstore: whether to execute the operation on backing storage
2929
      only, or on "logical" storage only; e.g. DRBD is logical storage,
2930
      whereas LVM, file, RBD are backing storage
2931
  @rtype: (status, result)
2932
  @type excl_stor: boolean
2933
  @param excl_stor: Whether exclusive_storage is active
2934
  @return: a tuple with the status of the operation (True/False), and
2935
      the errors message if status is False
2936

2937
  """
2938
  r_dev = _RecursiveFindBD(disk)
2939
  if r_dev is None:
2940
    _Fail("Cannot find block device %s", disk)
2941

    
2942
  try:
2943
    r_dev.Grow(amount, dryrun, backingstore, excl_stor)
2944
  except errors.BlockDeviceError, err:
2945
    _Fail("Failed to grow block device: %s", err, exc=True)
2946

    
2947

    
2948
def BlockdevSnapshot(disk):
2949
  """Create a snapshot copy of a block device.
2950

2951
  This function is called recursively, and the snapshot is actually created
2952
  just for the leaf lvm backend device.
2953

2954
  @type disk: L{objects.Disk}
2955
  @param disk: the disk to be snapshotted
2956
  @rtype: string
2957
  @return: snapshot disk ID as (vg, lv)
2958

2959
  """
2960
  if disk.dev_type == constants.LD_DRBD8:
2961
    if not disk.children:
2962
      _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2963
            disk.unique_id)
2964
    return BlockdevSnapshot(disk.children[0])
2965
  elif disk.dev_type == constants.LD_LV:
2966
    r_dev = _RecursiveFindBD(disk)
2967
    if r_dev is not None:
2968
      # FIXME: choose a saner value for the snapshot size
2969
      # let's stay on the safe side and ask for the full size, for now
2970
      return r_dev.Snapshot(disk.size)
2971
    else:
2972
      _Fail("Cannot find block device %s", disk)
2973
  else:
2974
    _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2975
          disk.unique_id, disk.dev_type)
2976

    
2977

    
2978
def BlockdevSetInfo(disk, info):
2979
  """Sets 'metadata' information on block devices.
2980

2981
  This function sets 'info' metadata on block devices. Initial
2982
  information is set at device creation; this function should be used
2983
  for example after renames.
2984

2985
  @type disk: L{objects.Disk}
2986
  @param disk: the disk to be grown
2987
  @type info: string
2988
  @param info: new 'info' metadata
2989
  @rtype: (status, result)
2990
  @return: a tuple with the status of the operation (True/False), and
2991
      the errors message if status is False
2992

2993
  """
2994
  r_dev = _RecursiveFindBD(disk)
2995
  if r_dev is None:
2996
    _Fail("Cannot find block device %s", disk)
2997

    
2998
  try:
2999
    r_dev.SetInfo(info)
3000
  except errors.BlockDeviceError, err:
3001
    _Fail("Failed to set information on block device: %s", err, exc=True)
3002

    
3003

    
3004
def FinalizeExport(instance, snap_disks):
3005
  """Write out the export configuration information.
3006

3007
  @type instance: L{objects.Instance}
3008
  @param instance: the instance which we export, used for
3009
      saving configuration
3010
  @type snap_disks: list of L{objects.Disk}
3011
  @param snap_disks: list of snapshot block devices, which
3012
      will be used to get the actual name of the dump file
3013

3014
  @rtype: None
3015

3016
  """
3017
  destdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name + ".new")
3018
  finaldestdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name)
3019

    
3020
  config = objects.SerializableConfigParser()
3021

    
3022
  config.add_section(constants.INISECT_EXP)
3023
  config.set(constants.INISECT_EXP, "version", "0")
3024
  config.set(constants.INISECT_EXP, "timestamp", "%d" % int(time.time()))
3025
  config.set(constants.INISECT_EXP, "source", instance.primary_node)
3026
  config.set(constants.INISECT_EXP, "os", instance.os)
3027
  config.set(constants.INISECT_EXP, "compression", "none")
3028

    
3029
  config.add_section(constants.INISECT_INS)
3030
  config.set(constants.INISECT_INS, "name", instance.name)
3031
  config.set(constants.INISECT_INS, "maxmem", "%d" %
3032
             instance.beparams[constants.BE_MAXMEM])
3033
  config.set(constants.INISECT_INS, "minmem", "%d" %
3034
             instance.beparams[constants.BE_MINMEM])
3035
  # "memory" is deprecated, but useful for exporting to old ganeti versions
3036
  config.set(constants.INISECT_INS, "memory", "%d" %
3037
             instance.beparams[constants.BE_MAXMEM])
3038
  config.set(constants.INISECT_INS, "vcpus", "%d" %
3039
             instance.beparams[constants.BE_VCPUS])
3040
  config.set(constants.INISECT_INS, "disk_template", instance.disk_template)
3041
  config.set(constants.INISECT_INS, "hypervisor", instance.hypervisor)
3042
  config.set(constants.INISECT_INS, "tags", " ".join(instance.GetTags()))
3043

    
3044
  nic_total = 0
3045
  for nic_count, nic in enumerate(instance.nics):
3046
    nic_total += 1
3047
    config.set(constants.INISECT_INS, "nic%d_mac" %
3048
               nic_count, "%s" % nic.mac)
3049
    config.set(constants.INISECT_INS, "nic%d_ip" % nic_count, "%s" % nic.ip)
3050
    config.set(constants.INISECT_INS, "nic%d_network" % nic_count,
3051
               "%s" % nic.network)
3052
    for param in constants.NICS_PARAMETER_TYPES:
3053
      config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param),
3054
                 "%s" % nic.nicparams.get(param, None))
3055
  # TODO: redundant: on load can read nics until it doesn't exist
3056
  config.set(constants.INISECT_INS, "nic_count", "%d" % nic_total)
3057

    
3058
  disk_total = 0
3059
  for disk_count, disk in enumerate(snap_disks):
3060
    if disk:
3061
      disk_total += 1
3062
      config.set(constants.INISECT_INS, "disk%d_ivname" % disk_count,
3063
                 ("%s" % disk.iv_name))
3064
      config.set(constants.INISECT_INS, "disk%d_dump" % disk_count,
3065
                 ("%s" % disk.physical_id[1]))
3066
      config.set(constants.INISECT_INS, "disk%d_size" % disk_count,
3067
                 ("%d" % disk.size))
3068

    
3069
  config.set(constants.INISECT_INS, "disk_count", "%d" % disk_total)
3070

    
3071
  # New-style hypervisor/backend parameters
3072

    
3073
  config.add_section(constants.INISECT_HYP)
3074
  for name, value in instance.hvparams.items():
3075
    if name not in constants.HVC_GLOBALS:
3076
      config.set(constants.INISECT_HYP, name, str(value))
3077

    
3078
  config.add_section(constants.INISECT_BEP)
3079
  for name, value in instance.beparams.items():
3080
    config.set(constants.INISECT_BEP, name, str(value))
3081

    
3082
  config.add_section(constants.INISECT_OSP)
3083
  for name, value in instance.osparams.items():
3084
    config.set(constants.INISECT_OSP, name, str(value))
3085

    
3086
  utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
3087
                  data=config.Dumps())
3088
  shutil.rmtree(finaldestdir, ignore_errors=True)
3089
  shutil.move(destdir, finaldestdir)
3090

    
3091

    
3092
def ExportInfo(dest):
3093
  """Get export configuration information.
3094

3095
  @type dest: str
3096
  @param dest: directory containing the export
3097

3098
  @rtype: L{objects.SerializableConfigParser}
3099
  @return: a serializable config file containing the
3100
      export info
3101

3102
  """
3103
  cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
3104

    
3105
  config = objects.SerializableConfigParser()
3106
  config.read(cff)
3107

    
3108
  if (not config.has_section(constants.INISECT_EXP) or
3109
      not config.has_section(constants.INISECT_INS)):
3110
    _Fail("Export info file doesn't have the required fields")
3111

    
3112
  return config.Dumps()
3113

    
3114

    
3115
def ListExports():
3116
  """Return a list of exports currently available on this machine.
3117

3118
  @rtype: list
3119
  @return: list of the exports
3120

3121
  """
3122
  if os.path.isdir(pathutils.EXPORT_DIR):
3123
    return sorted(utils.ListVisibleFiles(pathutils.EXPORT_DIR))
3124
  else:
3125
    _Fail("No exports directory")
3126

    
3127

    
3128
def RemoveExport(export):
3129
  """Remove an existing export from the node.
3130

3131
  @type export: str
3132
  @param export: the name of the export to remove
3133
  @rtype: None
3134

3135
  """
3136
  target = utils.PathJoin(pathutils.EXPORT_DIR, export)
3137

    
3138
  try:
3139
    shutil.rmtree(target)
3140
  except EnvironmentError, err:
3141
    _Fail("Error while removing the export: %s", err, exc=True)
3142

    
3143

    
3144
def BlockdevRename(devlist):
3145
  """Rename a list of block devices.
3146

3147
  @type devlist: list of tuples
3148
  @param devlist: list of tuples of the form  (disk,
3149
      new_logical_id, new_physical_id); disk is an
3150
      L{objects.Disk} object describing the current disk,
3151
      and new logical_id/physical_id is the name we
3152
      rename it to
3153
  @rtype: boolean
3154
  @return: True if all renames succeeded, False otherwise
3155

3156
  """
3157
  msgs = []
3158
  result = True
3159
  for disk, unique_id in devlist:
3160
    dev = _RecursiveFindBD(disk)
3161
    if dev is None:
3162
      msgs.append("Can't find device %s in rename" % str(disk))
3163
      result = False
3164
      continue
3165
    try:
3166
      old_rpath = dev.dev_path
3167
      dev.Rename(unique_id)
3168
      new_rpath = dev.dev_path
3169
      if old_rpath != new_rpath:
3170
        DevCacheManager.RemoveCache(old_rpath)
3171
        # FIXME: we should add the new cache information here, like:
3172
        # DevCacheManager.UpdateCache(new_rpath, owner, ...)
3173
        # but we don't have the owner here - maybe parse from existing
3174
        # cache? for now, we only lose lvm data when we rename, which
3175
        # is less critical than DRBD or MD
3176
    except errors.BlockDeviceError, err:
3177
      msgs.append("Can't rename device '%s' to '%s': %s" %
3178
                  (dev, unique_id, err))
3179
      logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
3180
      result = False
3181
  if not result:
3182
    _Fail("; ".join(msgs))
3183

    
3184

    
3185
def _TransformFileStorageDir(fs_dir):
3186
  """Checks whether given file_storage_dir is valid.
3187

3188
  Checks wheter the given fs_dir is within the cluster-wide default
3189
  file_storage_dir or the shared_file_storage_dir, which are stored in
3190
  SimpleStore. Only paths under those directories are allowed.
3191

3192
  @type fs_dir: str
3193
  @param fs_dir: the path to check
3194

3195
  @return: the normalized path if valid, None otherwise
3196

3197
  """
3198
  filestorage.CheckFileStoragePath(fs_dir)
3199

    
3200
  return os.path.normpath(fs_dir)
3201

    
3202

    
3203
def CreateFileStorageDir(file_storage_dir):
3204
  """Create file storage directory.
3205

3206
  @type file_storage_dir: str
3207
  @param file_storage_dir: directory to create
3208

3209
  @rtype: tuple
3210
  @return: tuple with first element a boolean indicating wheter dir
3211
      creation was successful or not
3212

3213
  """
3214
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3215
  if os.path.exists(file_storage_dir):
3216
    if not os.path.isdir(file_storage_dir):
3217
      _Fail("Specified storage dir '%s' is not a directory",
3218
            file_storage_dir)
3219
  else:
3220
    try:
3221
      os.makedirs(file_storage_dir, 0750)
3222
    except OSError, err:
3223
      _Fail("Cannot create file storage directory '%s': %s",
3224
            file_storage_dir, err, exc=True)
3225

    
3226

    
3227
def RemoveFileStorageDir(file_storage_dir):
3228
  """Remove file storage directory.
3229

3230
  Remove it only if it's empty. If not log an error and return.
3231

3232
  @type file_storage_dir: str
3233
  @param file_storage_dir: the directory we should cleanup
3234
  @rtype: tuple (success,)
3235
  @return: tuple of one element, C{success}, denoting
3236
      whether the operation was successful
3237

3238
  """
3239
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3240
  if os.path.exists(file_storage_dir):
3241
    if not os.path.isdir(file_storage_dir):
3242
      _Fail("Specified Storage directory '%s' is not a directory",
3243
            file_storage_dir)
3244
    # deletes dir only if empty, otherwise we want to fail the rpc call
3245
    try:
3246
      os.rmdir(file_storage_dir)
3247
    except OSError, err:
3248
      _Fail("Cannot remove file storage directory '%s': %s",
3249
            file_storage_dir, err)
3250

    
3251

    
3252
def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
3253
  """Rename the file storage directory.
3254

3255
  @type old_file_storage_dir: str
3256
  @param old_file_storage_dir: the current path
3257
  @type new_file_storage_dir: str
3258
  @param new_file_storage_dir: the name we should rename to
3259
  @rtype: tuple (success,)
3260
  @return: tuple of one element, C{success}, denoting
3261
      whether the operation was successful
3262

3263
  """
3264
  old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
3265
  new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
3266
  if not os.path.exists(new_file_storage_dir):
3267
    if os.path.isdir(old_file_storage_dir):
3268
      try:
3269
        os.rename(old_file_storage_dir, new_file_storage_dir)
3270
      except OSError, err:
3271
        _Fail("Cannot rename '%s' to '%s': %s",
3272
              old_file_storage_dir, new_file_storage_dir, err)
3273
    else:
3274
      _Fail("Specified storage dir '%s' is not a directory",
3275
            old_file_storage_dir)
3276
  else:
3277
    if os.path.exists(old_file_storage_dir):
3278
      _Fail("Cannot rename '%s' to '%s': both locations exist",
3279
            old_file_storage_dir, new_file_storage_dir)
3280

    
3281

    
3282
def _EnsureJobQueueFile(file_name):
3283
  """Checks whether the given filename is in the queue directory.
3284

3285
  @type file_name: str
3286
  @param file_name: the file name we should check
3287
  @rtype: None
3288
  @raises RPCFail: if the file is not valid
3289

3290
  """
3291
  if not utils.IsBelowDir(pathutils.QUEUE_DIR, file_name):
3292
    _Fail("Passed job queue file '%s' does not belong to"
3293
          " the queue directory '%s'", file_name, pathutils.QUEUE_DIR)
3294

    
3295

    
3296
def JobQueueUpdate(file_name, content):
3297
  """Updates a file in the queue directory.
3298

3299
  This is just a wrapper over L{utils.io.WriteFile}, with proper
3300
  checking.
3301

3302
  @type file_name: str
3303
  @param file_name: the job file name
3304
  @type content: str
3305
  @param content: the new job contents
3306
  @rtype: boolean
3307
  @return: the success of the operation
3308

3309
  """
3310
  file_name = vcluster.LocalizeVirtualPath(file_name)
3311

    
3312
  _EnsureJobQueueFile(file_name)
3313
  getents = runtime.GetEnts()
3314

    
3315
  # Write and replace the file atomically
3316
  utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
3317
                  gid=getents.daemons_gid, mode=constants.JOB_QUEUE_FILES_PERMS)
3318

    
3319

    
3320
def JobQueueRename(old, new):
3321
  """Renames a job queue file.
3322

3323
  This is just a wrapper over os.rename with proper checking.
3324

3325
  @type old: str
3326
  @param old: the old (actual) file name
3327
  @type new: str
3328
  @param new: the desired file name
3329
  @rtype: tuple
3330
  @return: the success of the operation and payload
3331

3332
  """
3333
  old = vcluster.LocalizeVirtualPath(old)
3334
  new = vcluster.LocalizeVirtualPath(new)
3335

    
3336
  _EnsureJobQueueFile(old)
3337
  _EnsureJobQueueFile(new)
3338

    
3339
  getents = runtime.GetEnts()
3340

    
3341
  utils.RenameFile(old, new, mkdir=True, mkdir_mode=0750,
3342
                   dir_uid=getents.masterd_uid, dir_gid=getents.daemons_gid)
3343

    
3344

    
3345
def BlockdevClose(instance_name, disks):
3346
  """Closes the given block devices.
3347

3348
  This means they will be switched to secondary mode (in case of
3349
  DRBD).
3350

3351
  @param instance_name: if the argument is not empty, the symlinks
3352
      of this instance will be removed
3353
  @type disks: list of L{objects.Disk}
3354
  @param disks: the list of disks to be closed
3355
  @rtype: tuple (success, message)
3356
  @return: a tuple of success and message, where success
3357
      indicates the succes of the operation, and message
3358
      which will contain the error details in case we
3359
      failed
3360

3361
  """
3362
  bdevs = []
3363
  for cf in disks:
3364
    rd = _RecursiveFindBD(cf)
3365
    if rd is None:
3366
      _Fail("Can't find device %s", cf)
3367
    bdevs.append(rd)
3368

    
3369
  msg = []
3370
  for rd in bdevs:
3371
    try:
3372
      rd.Close()
3373
    except errors.BlockDeviceError, err:
3374
      msg.append(str(err))
3375
  if msg:
3376
    _Fail("Can't make devices secondary: %s", ",".join(msg))
3377
  else:
3378
    if instance_name:
3379
      _RemoveBlockDevLinks(instance_name, disks)
3380

    
3381

    
3382
def ValidateHVParams(hvname, hvparams):
3383
  """Validates the given hypervisor parameters.
3384

3385
  @type hvname: string
3386
  @param hvname: the hypervisor name
3387
  @type hvparams: dict
3388
  @param hvparams: the hypervisor parameters to be validated
3389
  @rtype: None
3390

3391
  """
3392
  try:
3393
    hv_type = hypervisor.GetHypervisor(hvname)
3394
    hv_type.ValidateParameters(hvparams)
3395
  except errors.HypervisorError, err:
3396
    _Fail(str(err), log=False)
3397

    
3398

    
3399
def _CheckOSPList(os_obj, parameters):
3400
  """Check whether a list of parameters is supported by the OS.
3401

3402
  @type os_obj: L{objects.OS}
3403
  @param os_obj: OS object to check
3404
  @type parameters: list
3405
  @param parameters: the list of parameters to check
3406

3407
  """
3408
  supported = [v[0] for v in os_obj.supported_parameters]
3409
  delta = frozenset(parameters).difference(supported)
3410
  if delta:
3411
    _Fail("The following parameters are not supported"
3412
          " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
3413

    
3414

    
3415
def ValidateOS(required, osname, checks, osparams):
3416
  """Validate the given OS' parameters.
3417

3418
  @type required: boolean
3419
  @param required: whether absence of the OS should translate into
3420
      failure or not
3421
  @type osname: string
3422
  @param osname: the OS to be validated
3423
  @type checks: list
3424
  @param checks: list of the checks to run (currently only 'parameters')
3425
  @type osparams: dict
3426
  @param osparams: dictionary with OS parameters
3427
  @rtype: boolean
3428
  @return: True if the validation passed, or False if the OS was not
3429
      found and L{required} was false
3430

3431
  """
3432
  if not constants.OS_VALIDATE_CALLS.issuperset(checks):
3433
    _Fail("Unknown checks required for OS %s: %s", osname,
3434
          set(checks).difference(constants.OS_VALIDATE_CALLS))
3435

    
3436
  name_only = objects.OS.GetName(osname)
3437
  status, tbv = _TryOSFromDisk(name_only, None)
3438

    
3439
  if not status:
3440
    if required:
3441
      _Fail(tbv)
3442
    else:
3443
      return False
3444

    
3445
  if max(tbv.api_versions) < constants.OS_API_V20:
3446
    return True
3447

    
3448
  if constants.OS_VALIDATE_PARAMETERS in checks:
3449
    _CheckOSPList(tbv, osparams.keys())
3450

    
3451
  validate_env = OSCoreEnv(osname, tbv, osparams)
3452
  result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
3453
                        cwd=tbv.path, reset_env=True)
3454
  if result.failed:
3455
    logging.error("os validate command '%s' returned error: %s output: %s",
3456
                  result.cmd, result.fail_reason, result.output)
3457
    _Fail("OS validation script failed (%s), output: %s",
3458
          result.fail_reason, result.output, log=False)
3459

    
3460
  return True
3461

    
3462

    
3463
def DemoteFromMC():
3464
  """Demotes the current node from master candidate role.
3465

3466
  """
3467
  # try to ensure we're not the master by mistake
3468
  master, myself = ssconf.GetMasterAndMyself()
3469
  if master == myself:
3470
    _Fail("ssconf status shows I'm the master node, will not demote")
3471

    
3472
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "check", constants.MASTERD])
3473
  if not result.failed:
3474
    _Fail("The master daemon is running, will not demote")
3475

    
3476
  try:
3477
    if os.path.isfile(pathutils.CLUSTER_CONF_FILE):
3478
      utils.CreateBackup(pathutils.CLUSTER_CONF_FILE)
3479
  except EnvironmentError, err:
3480
    if err.errno != errno.ENOENT:
3481
      _Fail("Error while backing up cluster file: %s", err, exc=True)
3482

    
3483
  utils.RemoveFile(pathutils.CLUSTER_CONF_FILE)
3484

    
3485

    
3486
def _GetX509Filenames(cryptodir, name):
3487
  """Returns the full paths for the private key and certificate.
3488

3489
  """
3490
  return (utils.PathJoin(cryptodir, name),
3491
          utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
3492
          utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
3493

    
3494

    
3495
def CreateX509Certificate(validity, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3496
  """Creates a new X509 certificate for SSL/TLS.
3497

3498
  @type validity: int
3499
  @param validity: Validity in seconds
3500
  @rtype: tuple; (string, string)
3501
  @return: Certificate name and public part
3502

3503
  """
3504
  (key_pem, cert_pem) = \
3505
    utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
3506
                                     min(validity, _MAX_SSL_CERT_VALIDITY))
3507

    
3508
  cert_dir = tempfile.mkdtemp(dir=cryptodir,
3509
                              prefix="x509-%s-" % utils.TimestampForFilename())
3510
  try:
3511
    name = os.path.basename(cert_dir)
3512
    assert len(name) > 5
3513

    
3514
    (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3515

    
3516
    utils.WriteFile(key_file, mode=0400, data=key_pem)
3517
    utils.WriteFile(cert_file, mode=0400, data=cert_pem)
3518

    
3519
    # Never return private key as it shouldn't leave the node
3520
    return (name, cert_pem)
3521
  except Exception:
3522
    shutil.rmtree(cert_dir, ignore_errors=True)
3523
    raise
3524

    
3525

    
3526
def RemoveX509Certificate(name, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3527
  """Removes a X509 certificate.
3528

3529
  @type name: string
3530
  @param name: Certificate name
3531

3532
  """
3533
  (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3534

    
3535
  utils.RemoveFile(key_file)
3536
  utils.RemoveFile(cert_file)
3537

    
3538
  try:
3539
    os.rmdir(cert_dir)
3540
  except EnvironmentError, err:
3541
    _Fail("Cannot remove certificate directory '%s': %s",
3542
          cert_dir, err)
3543

    
3544

    
3545
def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
3546
  """Returns the command for the requested input/output.
3547

3548
  @type instance: L{objects.Instance}
3549
  @param instance: The instance object
3550
  @param mode: Import/export mode
3551
  @param ieio: Input/output type
3552
  @param ieargs: Input/output arguments
3553

3554
  """
3555
  assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
3556

    
3557
  env = None
3558
  prefix = None
3559
  suffix = None
3560
  exp_size = None
3561

    
3562
  if ieio == constants.IEIO_FILE:
3563
    (filename, ) = ieargs
3564

    
3565
    if not utils.IsNormAbsPath(filename):
3566
      _Fail("Path '%s' is not normalized or absolute", filename)
3567

    
3568
    real_filename = os.path.realpath(filename)
3569
    directory = os.path.dirname(real_filename)
3570

    
3571
    if not utils.IsBelowDir(pathutils.EXPORT_DIR, real_filename):
3572
      _Fail("File '%s' is not under exports directory '%s': %s",
3573
            filename, pathutils.EXPORT_DIR, real_filename)
3574

    
3575
    # Create directory
3576
    utils.Makedirs(directory, mode=0750)
3577

    
3578
    quoted_filename = utils.ShellQuote(filename)
3579

    
3580
    if mode == constants.IEM_IMPORT:
3581
      suffix = "> %s" % quoted_filename
3582
    elif mode == constants.IEM_EXPORT:
3583
      suffix = "< %s" % quoted_filename
3584

    
3585
      # Retrieve file size
3586
      try:
3587
        st = os.stat(filename)
3588
      except EnvironmentError, err:
3589
        logging.error("Can't stat(2) %s: %s", filename, err)
3590
      else:
3591
        exp_size = utils.BytesToMebibyte(st.st_size)
3592

    
3593
  elif ieio == constants.IEIO_RAW_DISK:
3594
    (disk, ) = ieargs
3595

    
3596
    real_disk = _OpenRealBD(disk)
3597

    
3598
    if mode == constants.IEM_IMPORT:
3599
      # we set here a smaller block size as, due to transport buffering, more
3600
      # than 64-128k will mostly ignored; we use nocreat to fail if the device
3601
      # is not already there or we pass a wrong path; we use notrunc to no
3602
      # attempt truncate on an LV device; we use oflag=dsync to not buffer too
3603
      # much memory; this means that at best, we flush every 64k, which will
3604
      # not be very fast
3605
      suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
3606
                                    " bs=%s oflag=dsync"),
3607
                                    real_disk.dev_path,
3608
                                    str(64 * 1024))
3609

    
3610
    elif mode == constants.IEM_EXPORT:
3611
      # the block size on the read dd is 1MiB to match our units
3612
      prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
3613
                                   real_disk.dev_path,
3614
                                   str(1024 * 1024), # 1 MB
3615
                                   str(disk.size))
3616
      exp_size = disk.size
3617

    
3618
  elif ieio == constants.IEIO_SCRIPT:
3619
    (disk, disk_index, ) = ieargs
3620

    
3621
    assert isinstance(disk_index, (int, long))
3622

    
3623
    real_disk = _OpenRealBD(disk)
3624

    
3625
    inst_os = OSFromDisk(instance.os)
3626
    env = OSEnvironment(instance, inst_os)
3627

    
3628
    if mode == constants.IEM_IMPORT:
3629
      env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
3630
      env["IMPORT_INDEX"] = str(disk_index)
3631
      script = inst_os.import_script
3632

    
3633
    elif mode == constants.IEM_EXPORT:
3634
      env["EXPORT_DEVICE"] = real_disk.dev_path
3635
      env["EXPORT_INDEX"] = str(disk_index)
3636
      script = inst_os.export_script
3637

    
3638
    # TODO: Pass special environment only to script
3639
    script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
3640

    
3641
    if mode == constants.IEM_IMPORT:
3642
      suffix = "| %s" % script_cmd
3643

    
3644
    elif mode == constants.IEM_EXPORT:
3645
      prefix = "%s |" % script_cmd
3646

    
3647
    # Let script predict size
3648
    exp_size = constants.IE_CUSTOM_SIZE
3649

    
3650
  else:
3651
    _Fail("Invalid %s I/O mode %r", mode, ieio)
3652

    
3653
  return (env, prefix, suffix, exp_size)
3654

    
3655

    
3656
def _CreateImportExportStatusDir(prefix):
3657
  """Creates status directory for import/export.
3658

3659
  """
3660
  return tempfile.mkdtemp(dir=pathutils.IMPORT_EXPORT_DIR,
3661
                          prefix=("%s-%s-" %
3662
                                  (prefix, utils.TimestampForFilename())))
3663

    
3664

    
3665
def StartImportExportDaemon(mode, opts, host, port, instance, component,
3666
                            ieio, ieioargs):
3667
  """Starts an import or export daemon.
3668

3669
  @param mode: Import/output mode
3670
  @type opts: L{objects.ImportExportOptions}
3671
  @param opts: Daemon options
3672
  @type host: string
3673
  @param host: Remote host for export (None for import)
3674
  @type port: int
3675
  @param port: Remote port for export (None for import)
3676
  @type instance: L{objects.Instance}
3677
  @param instance: Instance object
3678
  @type component: string
3679
  @param component: which part of the instance is transferred now,
3680
      e.g. 'disk/0'
3681
  @param ieio: Input/output type
3682
  @param ieioargs: Input/output arguments
3683

3684
  """
3685
  if mode == constants.IEM_IMPORT:
3686
    prefix = "import"
3687

    
3688
    if not (host is None and port is None):
3689
      _Fail("Can not specify host or port on import")
3690

    
3691
  elif mode == constants.IEM_EXPORT:
3692
    prefix = "export"
3693

    
3694
    if host is None or port is None:
3695
      _Fail("Host and port must be specified for an export")
3696

    
3697
  else:
3698
    _Fail("Invalid mode %r", mode)
3699

    
3700
  if (opts.key_name is None) ^ (opts.ca_pem is None):
3701
    _Fail("Cluster certificate can only be used for both key and CA")
3702

    
3703
  (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
3704
    _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
3705

    
3706
  if opts.key_name is None:
3707
    # Use server.pem
3708
    key_path = pathutils.NODED_CERT_FILE
3709
    cert_path = pathutils.NODED_CERT_FILE
3710
    assert opts.ca_pem is None
3711
  else:
3712
    (_, key_path, cert_path) = _GetX509Filenames(pathutils.CRYPTO_KEYS_DIR,
3713
                                                 opts.key_name)
3714
    assert opts.ca_pem is not None
3715

    
3716
  for i in [key_path, cert_path]:
3717
    if not os.path.exists(i):
3718
      _Fail("File '%s' does not exist" % i)
3719

    
3720
  status_dir = _CreateImportExportStatusDir("%s-%s" % (prefix, component))
3721
  try:
3722
    status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
3723
    pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
3724
    ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
3725

    
3726
    if opts.ca_pem is None:
3727
      # Use server.pem
3728
      ca = utils.ReadFile(pathutils.NODED_CERT_FILE)
3729
    else:
3730
      ca = opts.ca_pem
3731

    
3732
    # Write CA file
3733
    utils.WriteFile(ca_file, data=ca, mode=0400)
3734

    
3735
    cmd = [
3736
      pathutils.IMPORT_EXPORT_DAEMON,
3737
      status_file, mode,
3738
      "--key=%s" % key_path,
3739
      "--cert=%s" % cert_path,
3740
      "--ca=%s" % ca_file,
3741
      ]
3742

    
3743
    if host:
3744
      cmd.append("--host=%s" % host)
3745

    
3746
    if port:
3747
      cmd.append("--port=%s" % port)
3748

    
3749
    if opts.ipv6:
3750
      cmd.append("--ipv6")
3751
    else:
3752
      cmd.append("--ipv4")
3753

    
3754
    if opts.compress:
3755
      cmd.append("--compress=%s" % opts.compress)
3756

    
3757
    if opts.magic:
3758
      cmd.append("--magic=%s" % opts.magic)
3759

    
3760
    if exp_size is not None:
3761
      cmd.append("--expected-size=%s" % exp_size)
3762

    
3763
    if cmd_prefix:
3764
      cmd.append("--cmd-prefix=%s" % cmd_prefix)
3765

    
3766
    if cmd_suffix:
3767
      cmd.append("--cmd-suffix=%s" % cmd_suffix)
3768

    
3769
    if mode == constants.IEM_EXPORT:
3770
      # Retry connection a few times when connecting to remote peer
3771
      cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3772
      cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3773
    elif opts.connect_timeout is not None:
3774
      assert mode == constants.IEM_IMPORT
3775
      # Overall timeout for establishing connection while listening
3776
      cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3777

    
3778
    logfile = _InstanceLogName(prefix, instance.os, instance.name, component)
3779

    
3780
    # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3781
    # support for receiving a file descriptor for output
3782
    utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3783
                      output=logfile)
3784

    
3785
    # The import/export name is simply the status directory name
3786
    return os.path.basename(status_dir)
3787

    
3788
  except Exception:
3789
    shutil.rmtree(status_dir, ignore_errors=True)
3790
    raise
3791

    
3792

    
3793
def GetImportExportStatus(names):
3794
  """Returns import/export daemon status.
3795

3796
  @type names: sequence
3797
  @param names: List of names
3798
  @rtype: List of dicts
3799
  @return: Returns a list of the state of each named import/export or None if a
3800
           status couldn't be read
3801

3802
  """
3803
  result = []
3804

    
3805
  for name in names:
3806
    status_file = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name,
3807
                                 _IES_STATUS_FILE)
3808

    
3809
    try:
3810
      data = utils.ReadFile(status_file)
3811
    except EnvironmentError, err:
3812
      if err.errno != errno.ENOENT:
3813
        raise
3814
      data = None
3815

    
3816
    if not data:
3817
      result.append(None)
3818
      continue
3819

    
3820
    result.append(serializer.LoadJson(data))
3821

    
3822
  return result
3823

    
3824

    
3825
def AbortImportExport(name):
3826
  """Sends SIGTERM to a running import/export daemon.
3827

3828
  """
3829
  logging.info("Abort import/export %s", name)
3830

    
3831
  status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3832
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3833

    
3834
  if pid:
3835
    logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3836
                 name, pid)
3837
    utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3838

    
3839

    
3840
def CleanupImportExport(name):
3841
  """Cleanup after an import or export.
3842

3843
  If the import/export daemon is still running it's killed. Afterwards the
3844
  whole status directory is removed.
3845

3846
  """
3847
  logging.info("Finalizing import/export %s", name)
3848

    
3849
  status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3850

    
3851
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3852

    
3853
  if pid:
3854
    logging.info("Import/export %s is still running with PID %s",
3855
                 name, pid)
3856
    utils.KillProcess(pid, waitpid=False)
3857

    
3858
  shutil.rmtree(status_dir, ignore_errors=True)
3859

    
3860

    
3861
def _SetPhysicalId(target_node_uuid, nodes_ip, disks):
3862
  """Sets the correct physical ID on all passed disks.
3863

3864
  """
3865
  for cf in disks:
3866
    cf.SetPhysicalID(target_node_uuid, nodes_ip)
3867

    
3868

    
3869
def _FindDisks(target_node_uuid, nodes_ip, disks):
3870
  """Sets the physical ID on disks and returns the block devices.
3871

3872
  """
3873
  _SetPhysicalId(target_node_uuid, nodes_ip, disks)
3874

    
3875
  bdevs = []
3876

    
3877
  for cf in disks:
3878
    rd = _RecursiveFindBD(cf)
3879
    if rd is None:
3880
      _Fail("Can't find device %s", cf)
3881
    bdevs.append(rd)
3882
  return bdevs
3883

    
3884

    
3885
def DrbdDisconnectNet(target_node_uuid, nodes_ip, disks):
3886
  """Disconnects the network on a list of drbd devices.
3887

3888
  """
3889
  bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
3890

    
3891
  # disconnect disks
3892
  for rd in bdevs:
3893
    try:
3894
      rd.DisconnectNet()
3895
    except errors.BlockDeviceError, err:
3896
      _Fail("Can't change network configuration to standalone mode: %s",
3897
            err, exc=True)
3898

    
3899

    
3900
def DrbdAttachNet(target_node_uuid, nodes_ip, disks, instance_name,
3901
                  multimaster):
3902
  """Attaches the network on a list of drbd devices.
3903

3904
  """
3905
  bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
3906

    
3907
  if multimaster:
3908
    for idx, rd in enumerate(bdevs):
3909
      try:
3910
        _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3911
      except EnvironmentError, err:
3912
        _Fail("Can't create symlink: %s", err)
3913
  # reconnect disks, switch to new master configuration and if
3914
  # needed primary mode
3915
  for rd in bdevs:
3916
    try:
3917
      rd.AttachNet(multimaster)
3918
    except errors.BlockDeviceError, err:
3919
      _Fail("Can't change network configuration: %s", err)
3920

    
3921
  # wait until the disks are connected; we need to retry the re-attach
3922
  # if the device becomes standalone, as this might happen if the one
3923
  # node disconnects and reconnects in a different mode before the
3924
  # other node reconnects; in this case, one or both of the nodes will
3925
  # decide it has wrong configuration and switch to standalone
3926

    
3927
  def _Attach():
3928
    all_connected = True
3929

    
3930
    for rd in bdevs:
3931
      stats = rd.GetProcStatus()
3932

    
3933
      all_connected = (all_connected and
3934
                       (stats.is_connected or stats.is_in_resync))
3935

    
3936
      if stats.is_standalone:
3937
        # peer had different config info and this node became
3938
        # standalone, even though this should not happen with the
3939
        # new staged way of changing disk configs
3940
        try:
3941
          rd.AttachNet(multimaster)
3942
        except errors.BlockDeviceError, err:
3943
          _Fail("Can't change network configuration: %s", err)
3944

    
3945
    if not all_connected:
3946
      raise utils.RetryAgain()
3947

    
3948
  try:
3949
    # Start with a delay of 100 miliseconds and go up to 5 seconds
3950
    utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
3951
  except utils.RetryTimeout:
3952
    _Fail("Timeout in disk reconnecting")
3953

    
3954
  if multimaster:
3955
    # change to primary mode
3956
    for rd in bdevs:
3957
      try:
3958
        rd.Open()
3959
      except errors.BlockDeviceError, err:
3960
        _Fail("Can't change to primary mode: %s", err)
3961

    
3962

    
3963
def DrbdWaitSync(target_node_uuid, nodes_ip, disks):
3964
  """Wait until DRBDs have synchronized.
3965

3966
  """
3967
  def _helper(rd):
3968
    stats = rd.GetProcStatus()
3969
    if not (stats.is_connected or stats.is_in_resync):
3970
      raise utils.RetryAgain()
3971
    return stats
3972

    
3973
  bdevs = _FindDisks(target_node_uuid, nodes_ip, disks)
3974

    
3975
  min_resync = 100
3976
  alldone = True
3977
  for rd in bdevs:
3978
    try:
3979
      # poll each second for 15 seconds
3980
      stats = utils.Retry(_helper, 1, 15, args=[rd])
3981
    except utils.RetryTimeout:
3982
      stats = rd.GetProcStatus()
3983
      # last check
3984
      if not (stats.is_connected or stats.is_in_resync):
3985
        _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
3986
    alldone = alldone and (not stats.is_in_resync)
3987
    if stats.sync_percent is not None:
3988
      min_resync = min(min_resync, stats.sync_percent)
3989

    
3990
  return (alldone, min_resync)
3991

    
3992

    
3993
def DrbdNeedsActivation(target_node_uuid, nodes_ip, disks):
3994
  """Checks which of the passed disks needs activation and returns their UUIDs.
3995

3996
  """
3997
  _SetPhysicalId(target_node_uuid, nodes_ip, disks)
3998
  faulty_disks = []
3999

    
4000
  for disk in disks:
4001
    rd = _RecursiveFindBD(disk)
4002
    if rd is None:
4003
      faulty_disks.append(disk)
4004
      continue
4005

    
4006
    stats = rd.GetProcStatus()
4007
    if stats.is_standalone or stats.is_diskless:
4008
      faulty_disks.append(disk)
4009

    
4010
  return [disk.uuid for disk in faulty_disks]
4011

    
4012

    
4013
def GetDrbdUsermodeHelper():
4014
  """Returns DRBD usermode helper currently configured.
4015

4016
  """
4017
  try:
4018
    return drbd.DRBD8.GetUsermodeHelper()
4019
  except errors.BlockDeviceError, err:
4020
    _Fail(str(err))
4021

    
4022

    
4023
def PowercycleNode(hypervisor_type, hvparams=None):
4024
  """Hard-powercycle the node.
4025

4026
  Because we need to return first, and schedule the powercycle in the
4027
  background, we won't be able to report failures nicely.
4028

4029
  """
4030
  hyper = hypervisor.GetHypervisor(hypervisor_type)
4031
  try:
4032
    pid = os.fork()
4033
  except OSError:
4034
    # if we can't fork, we'll pretend that we're in the child process
4035
    pid = 0
4036
  if pid > 0:
4037
    return "Reboot scheduled in 5 seconds"
4038
  # ensure the child is running on ram
4039
  try:
4040
    utils.Mlockall()
4041
  except Exception: # pylint: disable=W0703
4042
    pass
4043
  time.sleep(5)
4044
  hyper.PowercycleNode(hvparams=hvparams)
4045

    
4046

    
4047
def _VerifyRestrictedCmdName(cmd):
4048
  """Verifies a restricted command name.
4049

4050
  @type cmd: string
4051
  @param cmd: Command name
4052
  @rtype: tuple; (boolean, string or None)
4053
  @return: The tuple's first element is the status; if C{False}, the second
4054
    element is an error message string, otherwise it's C{None}
4055

4056
  """
4057
  if not cmd.strip():
4058
    return (False, "Missing command name")
4059

    
4060
  if os.path.basename(cmd) != cmd:
4061
    return (False, "Invalid command name")
4062

    
4063
  if not constants.EXT_PLUGIN_MASK.match(cmd):
4064
    return (False, "Command name contains forbidden characters")
4065

    
4066
  return (True, None)
4067

    
4068

    
4069
def _CommonRestrictedCmdCheck(path, owner):
4070
  """Common checks for restricted command file system directories and files.
4071

4072
  @type path: string
4073
  @param path: Path to check
4074
  @param owner: C{None} or tuple containing UID and GID
4075
  @rtype: tuple; (boolean, string or C{os.stat} result)
4076
  @return: The tuple's first element is the status; if C{False}, the second
4077
    element is an error message string, otherwise it's the result of C{os.stat}
4078

4079
  """
4080
  if owner is None:
4081
    # Default to root as owner
4082
    owner = (0, 0)
4083

    
4084
  try:
4085
    st = os.stat(path)
4086
  except EnvironmentError, err:
4087
    return (False, "Can't stat(2) '%s': %s" % (path, err))
4088

    
4089
  if stat.S_IMODE(st.st_mode) & (~_RCMD_MAX_MODE):
4090
    return (False, "Permissions on '%s' are too permissive" % path)
4091

    
4092
  if (st.st_uid, st.st_gid) != owner:
4093
    (owner_uid, owner_gid) = owner
4094
    return (False, "'%s' is not owned by %s:%s" % (path, owner_uid, owner_gid))
4095

    
4096
  return (True, st)
4097

    
4098

    
4099
def _VerifyRestrictedCmdDirectory(path, _owner=None):
4100
  """Verifies restricted command directory.
4101

4102
  @type path: string
4103
  @param path: Path to check
4104
  @rtype: tuple; (boolean, string or None)
4105
  @return: The tuple's first element is the status; if C{False}, the second
4106
    element is an error message string, otherwise it's C{None}
4107

4108
  """
4109
  (status, value) = _CommonRestrictedCmdCheck(path, _owner)
4110

    
4111
  if not status:
4112
    return (False, value)
4113

    
4114
  if not stat.S_ISDIR(value.st_mode):
4115
    return (False, "Path '%s' is not a directory" % path)
4116

    
4117
  return (True, None)
4118

    
4119

    
4120
def _VerifyRestrictedCmd(path, cmd, _owner=None):
4121
  """Verifies a whole restricted command and returns its executable filename.
4122

4123
  @type path: string
4124
  @param path: Directory containing restricted commands
4125
  @type cmd: string
4126
  @param cmd: Command name
4127
  @rtype: tuple; (boolean, string)
4128
  @return: The tuple's first element is the status; if C{False}, the second
4129
    element is an error message string, otherwise the second element is the
4130
    absolute path to the executable
4131

4132
  """
4133
  executable = utils.PathJoin(path, cmd)
4134

    
4135
  (status, msg) = _CommonRestrictedCmdCheck(executable, _owner)
4136

    
4137
  if not status:
4138
    return (False, msg)
4139

    
4140
  if not utils.IsExecutable(executable):
4141
    return (False, "access(2) thinks '%s' can't be executed" % executable)
4142

    
4143
  return (True, executable)
4144

    
4145

    
4146
def _PrepareRestrictedCmd(path, cmd,
4147
                          _verify_dir=_VerifyRestrictedCmdDirectory,
4148
                          _verify_name=_VerifyRestrictedCmdName,
4149
                          _verify_cmd=_VerifyRestrictedCmd):
4150
  """Performs a number of tests on a restricted command.
4151

4152
  @type path: string
4153
  @param path: Directory containing restricted commands
4154
  @type cmd: string
4155
  @param cmd: Command name
4156
  @return: Same as L{_VerifyRestrictedCmd}
4157

4158
  """
4159
  # Verify the directory first
4160
  (status, msg) = _verify_dir(path)
4161
  if status:
4162
    # Check command if everything was alright
4163
    (status, msg) = _verify_name(cmd)
4164

    
4165
  if not status:
4166
    return (False, msg)
4167

    
4168
  # Check actual executable
4169
  return _verify_cmd(path, cmd)
4170

    
4171

    
4172
def RunRestrictedCmd(cmd,
4173
                     _lock_timeout=_RCMD_LOCK_TIMEOUT,
4174
                     _lock_file=pathutils.RESTRICTED_COMMANDS_LOCK_FILE,
4175
                     _path=pathutils.RESTRICTED_COMMANDS_DIR,
4176
                     _sleep_fn=time.sleep,
4177
                     _prepare_fn=_PrepareRestrictedCmd,
4178
                     _runcmd_fn=utils.RunCmd,
4179
                     _enabled=constants.ENABLE_RESTRICTED_COMMANDS):
4180
  """Executes a restricted command after performing strict tests.
4181

4182
  @type cmd: string
4183
  @param cmd: Command name
4184
  @rtype: string
4185
  @return: Command output
4186
  @raise RPCFail: In case of an error
4187

4188
  """
4189
  logging.info("Preparing to run restricted command '%s'", cmd)
4190

    
4191
  if not _enabled:
4192
    _Fail("Restricted commands disabled at configure time")
4193

    
4194
  lock = None
4195
  try:
4196
    cmdresult = None
4197
    try:
4198
      lock = utils.FileLock.Open(_lock_file)
4199
      lock.Exclusive(blocking=True, timeout=_lock_timeout)
4200

    
4201
      (status, value) = _prepare_fn(_path, cmd)
4202

    
4203
      if status:
4204
        cmdresult = _runcmd_fn([value], env={}, reset_env=True,
4205
                               postfork_fn=lambda _: lock.Unlock())
4206
      else:
4207
        logging.error(value)
4208
    except Exception: # pylint: disable=W0703
4209
      # Keep original error in log
4210
      logging.exception("Caught exception")
4211

    
4212
    if cmdresult is None:
4213
      logging.info("Sleeping for %0.1f seconds before returning",
4214
                   _RCMD_INVALID_DELAY)
4215
      _sleep_fn(_RCMD_INVALID_DELAY)
4216

    
4217
      # Do not include original error message in returned error
4218
      _Fail("Executing command '%s' failed" % cmd)
4219
    elif cmdresult.failed or cmdresult.fail_reason:
4220
      _Fail("Restricted command '%s' failed: %s; output: %s",
4221
            cmd, cmdresult.fail_reason, cmdresult.output)
4222
    else:
4223
      return cmdresult.output
4224
  finally:
4225
    if lock is not None:
4226
      # Release lock at last
4227
      lock.Close()
4228
      lock = None
4229

    
4230

    
4231
def SetWatcherPause(until, _filename=pathutils.WATCHER_PAUSEFILE):
4232
  """Creates or removes the watcher pause file.
4233

4234
  @type until: None or number
4235
  @param until: Unix timestamp saying until when the watcher shouldn't run
4236

4237
  """
4238
  if until is None:
4239
    logging.info("Received request to no longer pause watcher")
4240
    utils.RemoveFile(_filename)
4241
  else:
4242
    logging.info("Received request to pause watcher until %s", until)
4243

    
4244
    if not ht.TNumber(until):
4245
      _Fail("Duration must be numeric")
4246

    
4247
    utils.WriteFile(_filename, data="%d\n" % (until, ), mode=0644)
4248

    
4249

    
4250
class HooksRunner(object):
4251
  """Hook runner.
4252

4253
  This class is instantiated on the node side (ganeti-noded) and not
4254
  on the master side.
4255

4256
  """
4257
  def __init__(self, hooks_base_dir=None):
4258
    """Constructor for hooks runner.
4259

4260
    @type hooks_base_dir: str or None
4261
    @param hooks_base_dir: if not None, this overrides the
4262
        L{pathutils.HOOKS_BASE_DIR} (useful for unittests)
4263

4264
    """
4265
    if hooks_base_dir is None:
4266
      hooks_base_dir = pathutils.HOOKS_BASE_DIR
4267
    # yeah, _BASE_DIR is not valid for attributes, we use it like a
4268
    # constant
4269
    self._BASE_DIR = hooks_base_dir # pylint: disable=C0103
4270

    
4271
  def RunLocalHooks(self, node_list, hpath, phase, env):
4272
    """Check that the hooks will be run only locally and then run them.
4273

4274
    """
4275
    assert len(node_list) == 1
4276
    node = node_list[0]
4277
    _, myself = ssconf.GetMasterAndMyself()
4278
    assert node == myself
4279

    
4280
    results = self.RunHooks(hpath, phase, env)
4281

    
4282
    # Return values in the form expected by HooksMaster
4283
    return {node: (None, False, results)}
4284

    
4285
  def RunHooks(self, hpath, phase, env):
4286
    """Run the scripts in the hooks directory.
4287

4288
    @type hpath: str
4289
    @param hpath: the path to the hooks directory which
4290
        holds the scripts
4291
    @type phase: str
4292
    @param phase: either L{constants.HOOKS_PHASE_PRE} or
4293
        L{constants.HOOKS_PHASE_POST}
4294
    @type env: dict
4295
    @param env: dictionary with the environment for the hook
4296
    @rtype: list
4297
    @return: list of 3-element tuples:
4298
      - script path
4299
      - script result, either L{constants.HKR_SUCCESS} or
4300
        L{constants.HKR_FAIL}
4301
      - output of the script
4302

4303
    @raise errors.ProgrammerError: for invalid input
4304
        parameters
4305

4306
    """
4307
    if phase == constants.HOOKS_PHASE_PRE:
4308
      suffix = "pre"
4309
    elif phase == constants.HOOKS_PHASE_POST:
4310
      suffix = "post"
4311
    else:
4312
      _Fail("Unknown hooks phase '%s'", phase)
4313

    
4314
    subdir = "%s-%s.d" % (hpath, suffix)
4315
    dir_name = utils.PathJoin(self._BASE_DIR, subdir)
4316

    
4317
    results = []
4318

    
4319
    if not os.path.isdir(dir_name):
4320
      # for non-existing/non-dirs, we simply exit instead of logging a
4321
      # warning at every operation
4322
      return results
4323

    
4324
    runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
4325

    
4326
    for (relname, relstatus, runresult) in runparts_results:
4327
      if relstatus == constants.RUNPARTS_SKIP:
4328
        rrval = constants.HKR_SKIP
4329
        output = ""
4330
      elif relstatus == constants.RUNPARTS_ERR:
4331
        rrval = constants.HKR_FAIL
4332
        output = "Hook script execution error: %s" % runresult
4333
      elif relstatus == constants.RUNPARTS_RUN:
4334
        if runresult.failed:
4335
          rrval = constants.HKR_FAIL
4336
        else:
4337
          rrval = constants.HKR_SUCCESS
4338
        output = utils.SafeEncode(runresult.output.strip())
4339
      results.append(("%s/%s" % (subdir, relname), rrval, output))
4340

    
4341
    return results
4342

    
4343

    
4344
class IAllocatorRunner(object):
4345
  """IAllocator runner.
4346

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

4350
  """
4351
  @staticmethod
4352
  def Run(name, idata):
4353
    """Run an iallocator script.
4354

4355
    @type name: str
4356
    @param name: the iallocator script name
4357
    @type idata: str
4358
    @param idata: the allocator input data
4359

4360
    @rtype: tuple
4361
    @return: two element tuple of:
4362
       - status
4363
       - either error message or stdout of allocator (for success)
4364

4365
    """
4366
    alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
4367
                                  os.path.isfile)
4368
    if alloc_script is None:
4369
      _Fail("iallocator module '%s' not found in the search path", name)
4370

    
4371
    fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
4372
    try:
4373
      os.write(fd, idata)
4374
      os.close(fd)
4375
      result = utils.RunCmd([alloc_script, fin_name])
4376
      if result.failed:
4377
        _Fail("iallocator module '%s' failed: %s, output '%s'",
4378
              name, result.fail_reason, result.output)
4379
    finally:
4380
      os.unlink(fin_name)
4381

    
4382
    return result.stdout
4383

    
4384

    
4385
class DevCacheManager(object):
4386
  """Simple class for managing a cache of block device information.
4387

4388
  """
4389
  _DEV_PREFIX = "/dev/"
4390
  _ROOT_DIR = pathutils.BDEV_CACHE_DIR
4391

    
4392
  @classmethod
4393
  def _ConvertPath(cls, dev_path):
4394
    """Converts a /dev/name path to the cache file name.
4395

4396
    This replaces slashes with underscores and strips the /dev
4397
    prefix. It then returns the full path to the cache file.
4398

4399
    @type dev_path: str
4400
    @param dev_path: the C{/dev/} path name
4401
    @rtype: str
4402
    @return: the converted path name
4403

4404
    """
4405
    if dev_path.startswith(cls._DEV_PREFIX):
4406
      dev_path = dev_path[len(cls._DEV_PREFIX):]
4407
    dev_path = dev_path.replace("/", "_")
4408
    fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
4409
    return fpath
4410

    
4411
  @classmethod
4412
  def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
4413
    """Updates the cache information for a given device.
4414

4415
    @type dev_path: str
4416
    @param dev_path: the pathname of the device
4417
    @type owner: str
4418
    @param owner: the owner (instance name) of the device
4419
    @type on_primary: bool
4420
    @param on_primary: whether this is the primary
4421
        node nor not
4422
    @type iv_name: str
4423
    @param iv_name: the instance-visible name of the
4424
        device, as in objects.Disk.iv_name
4425

4426
    @rtype: None
4427

4428
    """
4429
    if dev_path is None:
4430
      logging.error("DevCacheManager.UpdateCache got a None dev_path")
4431
      return
4432
    fpath = cls._ConvertPath(dev_path)
4433
    if on_primary:
4434
      state = "primary"
4435
    else:
4436
      state = "secondary"
4437
    if iv_name is None:
4438
      iv_name = "not_visible"
4439
    fdata = "%s %s %s\n" % (str(owner), state, iv_name)
4440
    try:
4441
      utils.WriteFile(fpath, data=fdata)
4442
    except EnvironmentError, err:
4443
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
4444

    
4445
  @classmethod
4446
  def RemoveCache(cls, dev_path):
4447
    """Remove data for a dev_path.
4448

4449
    This is just a wrapper over L{utils.io.RemoveFile} with a converted
4450
    path name and logging.
4451

4452
    @type dev_path: str
4453
    @param dev_path: the pathname of the device
4454

4455
    @rtype: None
4456

4457
    """
4458
    if dev_path is None:
4459
      logging.error("DevCacheManager.RemoveCache got a None dev_path")
4460
      return
4461
    fpath = cls._ConvertPath(dev_path)
4462
    try:
4463
      utils.RemoveFile(fpath)
4464
    except EnvironmentError, err:
4465
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)