Statistics
| Branch: | Tag: | Revision:

root / lib / backend.py @ 3361ab37

History | View | Annotate | Download (129.7 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 import objects
60
from ganeti import ssconf
61
from ganeti import serializer
62
from ganeti import netutils
63
from ganeti import runtime
64
from ganeti import compat
65
from ganeti import pathutils
66
from ganeti import vcluster
67
from ganeti import ht
68
from ganeti.storage.base import BlockDev
69
from ganeti.storage.drbd import DRBD8
70
from ganeti import hooksmaster
71

    
72

    
73
_BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id"
74
_ALLOWED_CLEAN_DIRS = compat.UniqueFrozenset([
75
  pathutils.DATA_DIR,
76
  pathutils.JOB_QUEUE_ARCHIVE_DIR,
77
  pathutils.QUEUE_DIR,
78
  pathutils.CRYPTO_KEYS_DIR,
79
  ])
80
_MAX_SSL_CERT_VALIDITY = 7 * 24 * 60 * 60
81
_X509_KEY_FILE = "key"
82
_X509_CERT_FILE = "cert"
83
_IES_STATUS_FILE = "status"
84
_IES_PID_FILE = "pid"
85
_IES_CA_FILE = "ca"
86

    
87
#: Valid LVS output line regex
88
_LVSLINE_REGEX = re.compile("^ *([^|]+)\|([^|]+)\|([0-9.]+)\|([^|]{6,})\|?$")
89

    
90
# Actions for the master setup script
91
_MASTER_START = "start"
92
_MASTER_STOP = "stop"
93

    
94
#: Maximum file permissions for restricted command directory and executables
95
_RCMD_MAX_MODE = (stat.S_IRWXU |
96
                  stat.S_IRGRP | stat.S_IXGRP |
97
                  stat.S_IROTH | stat.S_IXOTH)
98

    
99
#: Delay before returning an error for restricted commands
100
_RCMD_INVALID_DELAY = 10
101

    
102
#: How long to wait to acquire lock for restricted commands (shorter than
103
#: L{_RCMD_INVALID_DELAY}) to reduce blockage of noded forks when many
104
#: command requests arrive
105
_RCMD_LOCK_TIMEOUT = _RCMD_INVALID_DELAY * 0.8
106

    
107

    
108
class RPCFail(Exception):
109
  """Class denoting RPC failure.
110

111
  Its argument is the error message.
112

113
  """
114

    
115

    
116
def _GetInstReasonFilename(instance_name):
117
  """Path of the file containing the reason of the instance status change.
118

119
  @type instance_name: string
120
  @param instance_name: The name of the instance
121
  @rtype: string
122
  @return: The path of the file
123

124
  """
125
  return utils.PathJoin(pathutils.INSTANCE_REASON_DIR, instance_name)
126

    
127

    
128
def _StoreInstReasonTrail(instance_name, trail):
129
  """Serialize a reason trail related to an instance change of state to file.
130

131
  The exact location of the file depends on the name of the instance and on
132
  the configuration of the Ganeti cluster defined at deploy time.
133

134
  @type instance_name: string
135
  @param instance_name: The name of the instance
136
  @rtype: None
137

138
  """
139
  json = serializer.DumpJson(trail)
140
  filename = _GetInstReasonFilename(instance_name)
141
  utils.WriteFile(filename, data=json)
142

    
143

    
144
def _Fail(msg, *args, **kwargs):
145
  """Log an error and the raise an RPCFail exception.
146

147
  This exception is then handled specially in the ganeti daemon and
148
  turned into a 'failed' return type. As such, this function is a
149
  useful shortcut for logging the error and returning it to the master
150
  daemon.
151

152
  @type msg: string
153
  @param msg: the text of the exception
154
  @raise RPCFail
155

156
  """
157
  if args:
158
    msg = msg % args
159
  if "log" not in kwargs or kwargs["log"]: # if we should log this error
160
    if "exc" in kwargs and kwargs["exc"]:
161
      logging.exception(msg)
162
    else:
163
      logging.error(msg)
164
  raise RPCFail(msg)
165

    
166

    
167
def _GetConfig():
168
  """Simple wrapper to return a SimpleStore.
169

170
  @rtype: L{ssconf.SimpleStore}
171
  @return: a SimpleStore instance
172

173
  """
174
  return ssconf.SimpleStore()
175

    
176

    
177
def _GetSshRunner(cluster_name):
178
  """Simple wrapper to return an SshRunner.
179

180
  @type cluster_name: str
181
  @param cluster_name: the cluster name, which is needed
182
      by the SshRunner constructor
183
  @rtype: L{ssh.SshRunner}
184
  @return: an SshRunner instance
185

186
  """
187
  return ssh.SshRunner(cluster_name)
188

    
189

    
190
def _Decompress(data):
191
  """Unpacks data compressed by the RPC client.
192

193
  @type data: list or tuple
194
  @param data: Data sent by RPC client
195
  @rtype: str
196
  @return: Decompressed data
197

198
  """
199
  assert isinstance(data, (list, tuple))
200
  assert len(data) == 2
201
  (encoding, content) = data
202
  if encoding == constants.RPC_ENCODING_NONE:
203
    return content
204
  elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
205
    return zlib.decompress(base64.b64decode(content))
206
  else:
207
    raise AssertionError("Unknown data encoding")
208

    
209

    
210
def _CleanDirectory(path, exclude=None):
211
  """Removes all regular files in a directory.
212

213
  @type path: str
214
  @param path: the directory to clean
215
  @type exclude: list
216
  @param exclude: list of files to be excluded, defaults
217
      to the empty list
218

219
  """
220
  if path not in _ALLOWED_CLEAN_DIRS:
221
    _Fail("Path passed to _CleanDirectory not in allowed clean targets: '%s'",
222
          path)
223

    
224
  if not os.path.isdir(path):
225
    return
226
  if exclude is None:
227
    exclude = []
228
  else:
229
    # Normalize excluded paths
230
    exclude = [os.path.normpath(i) for i in exclude]
231

    
232
  for rel_name in utils.ListVisibleFiles(path):
233
    full_name = utils.PathJoin(path, rel_name)
234
    if full_name in exclude:
235
      continue
236
    if os.path.isfile(full_name) and not os.path.islink(full_name):
237
      utils.RemoveFile(full_name)
238

    
239

    
240
def _BuildUploadFileList():
241
  """Build the list of allowed upload files.
242

243
  This is abstracted so that it's built only once at module import time.
244

245
  """
246
  allowed_files = set([
247
    pathutils.CLUSTER_CONF_FILE,
248
    pathutils.ETC_HOSTS,
249
    pathutils.SSH_KNOWN_HOSTS_FILE,
250
    pathutils.VNC_PASSWORD_FILE,
251
    pathutils.RAPI_CERT_FILE,
252
    pathutils.SPICE_CERT_FILE,
253
    pathutils.SPICE_CACERT_FILE,
254
    pathutils.RAPI_USERS_FILE,
255
    pathutils.CONFD_HMAC_KEY,
256
    pathutils.CLUSTER_DOMAIN_SECRET_FILE,
257
    ])
258

    
259
  for hv_name in constants.HYPER_TYPES:
260
    hv_class = hypervisor.GetHypervisorClass(hv_name)
261
    allowed_files.update(hv_class.GetAncillaryFiles()[0])
262

    
263
  assert pathutils.FILE_STORAGE_PATHS_FILE not in allowed_files, \
264
    "Allowed file storage paths should never be uploaded via RPC"
265

    
266
  return frozenset(allowed_files)
267

    
268

    
269
_ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
270

    
271

    
272
def JobQueuePurge():
273
  """Removes job queue files and archived jobs.
274

275
  @rtype: tuple
276
  @return: True, None
277

278
  """
279
  _CleanDirectory(pathutils.QUEUE_DIR, exclude=[pathutils.JOB_QUEUE_LOCK_FILE])
280
  _CleanDirectory(pathutils.JOB_QUEUE_ARCHIVE_DIR)
281

    
282

    
283
def GetMasterInfo():
284
  """Returns master information.
285

286
  This is an utility function to compute master information, either
287
  for consumption here or from the node daemon.
288

289
  @rtype: tuple
290
  @return: master_netdev, master_ip, master_name, primary_ip_family,
291
    master_netmask
292
  @raise RPCFail: in case of errors
293

294
  """
295
  try:
296
    cfg = _GetConfig()
297
    master_netdev = cfg.GetMasterNetdev()
298
    master_ip = cfg.GetMasterIP()
299
    master_netmask = cfg.GetMasterNetmask()
300
    master_node = cfg.GetMasterNode()
301
    primary_ip_family = cfg.GetPrimaryIPFamily()
302
  except errors.ConfigurationError, err:
303
    _Fail("Cluster configuration incomplete: %s", err, exc=True)
304
  return (master_netdev, master_ip, master_node, primary_ip_family,
305
          master_netmask)
306

    
307

    
308
def RunLocalHooks(hook_opcode, hooks_path, env_builder_fn):
309
  """Decorator that runs hooks before and after the decorated function.
310

311
  @type hook_opcode: string
312
  @param hook_opcode: opcode of the hook
313
  @type hooks_path: string
314
  @param hooks_path: path of the hooks
315
  @type env_builder_fn: function
316
  @param env_builder_fn: function that returns a dictionary containing the
317
    environment variables for the hooks. Will get all the parameters of the
318
    decorated function.
319
  @raise RPCFail: in case of pre-hook failure
320

321
  """
322
  def decorator(fn):
323
    def wrapper(*args, **kwargs):
324
      _, myself = ssconf.GetMasterAndMyself()
325
      nodes = ([myself], [myself])  # these hooks run locally
326

    
327
      env_fn = compat.partial(env_builder_fn, *args, **kwargs)
328

    
329
      cfg = _GetConfig()
330
      hr = HooksRunner()
331
      hm = hooksmaster.HooksMaster(hook_opcode, hooks_path, nodes,
332
                                   hr.RunLocalHooks, None, env_fn,
333
                                   logging.warning, cfg.GetClusterName(),
334
                                   cfg.GetMasterNode())
335
      hm.RunPhase(constants.HOOKS_PHASE_PRE)
336
      result = fn(*args, **kwargs)
337
      hm.RunPhase(constants.HOOKS_PHASE_POST)
338

    
339
      return result
340
    return wrapper
341
  return decorator
342

    
343

    
344
def _BuildMasterIpEnv(master_params, use_external_mip_script=None):
345
  """Builds environment variables for master IP hooks.
346

347
  @type master_params: L{objects.MasterNetworkParameters}
348
  @param master_params: network parameters of the master
349
  @type use_external_mip_script: boolean
350
  @param use_external_mip_script: whether to use an external master IP
351
    address setup script (unused, but necessary per the implementation of the
352
    _RunLocalHooks decorator)
353

354
  """
355
  # pylint: disable=W0613
356
  ver = netutils.IPAddress.GetVersionFromAddressFamily(master_params.ip_family)
357
  env = {
358
    "MASTER_NETDEV": master_params.netdev,
359
    "MASTER_IP": master_params.ip,
360
    "MASTER_NETMASK": str(master_params.netmask),
361
    "CLUSTER_IP_VERSION": str(ver),
362
  }
363

    
364
  return env
365

    
366

    
367
def _RunMasterSetupScript(master_params, action, use_external_mip_script):
368
  """Execute the master IP address setup script.
369

370
  @type master_params: L{objects.MasterNetworkParameters}
371
  @param master_params: network parameters of the master
372
  @type action: string
373
  @param action: action to pass to the script. Must be one of
374
    L{backend._MASTER_START} or L{backend._MASTER_STOP}
375
  @type use_external_mip_script: boolean
376
  @param use_external_mip_script: whether to use an external master IP
377
    address setup script
378
  @raise backend.RPCFail: if there are errors during the execution of the
379
    script
380

381
  """
382
  env = _BuildMasterIpEnv(master_params)
383

    
384
  if use_external_mip_script:
385
    setup_script = pathutils.EXTERNAL_MASTER_SETUP_SCRIPT
386
  else:
387
    setup_script = pathutils.DEFAULT_MASTER_SETUP_SCRIPT
388

    
389
  result = utils.RunCmd([setup_script, action], env=env, reset_env=True)
390

    
391
  if result.failed:
392
    _Fail("Failed to %s the master IP. Script return value: %s, output: '%s'" %
393
          (action, result.exit_code, result.output), log=True)
394

    
395

    
396
@RunLocalHooks(constants.FAKE_OP_MASTER_TURNUP, "master-ip-turnup",
397
               _BuildMasterIpEnv)
398
def ActivateMasterIp(master_params, use_external_mip_script):
399
  """Activate the IP address of the master daemon.
400

401
  @type master_params: L{objects.MasterNetworkParameters}
402
  @param master_params: network parameters of the master
403
  @type use_external_mip_script: boolean
404
  @param use_external_mip_script: whether to use an external master IP
405
    address setup script
406
  @raise RPCFail: in case of errors during the IP startup
407

408
  """
409
  _RunMasterSetupScript(master_params, _MASTER_START,
410
                        use_external_mip_script)
411

    
412

    
413
def StartMasterDaemons(no_voting):
414
  """Activate local node as master node.
415

416
  The function will start the master daemons (ganeti-masterd and ganeti-rapi).
417

418
  @type no_voting: boolean
419
  @param no_voting: whether to start ganeti-masterd without a node vote
420
      but still non-interactively
421
  @rtype: None
422

423
  """
424

    
425
  if no_voting:
426
    masterd_args = "--no-voting --yes-do-it"
427
  else:
428
    masterd_args = ""
429

    
430
  env = {
431
    "EXTRA_MASTERD_ARGS": masterd_args,
432
    }
433

    
434
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "start-master"], env=env)
435
  if result.failed:
436
    msg = "Can't start Ganeti master: %s" % result.output
437
    logging.error(msg)
438
    _Fail(msg)
439

    
440

    
441
@RunLocalHooks(constants.FAKE_OP_MASTER_TURNDOWN, "master-ip-turndown",
442
               _BuildMasterIpEnv)
443
def DeactivateMasterIp(master_params, use_external_mip_script):
444
  """Deactivate the master IP on this node.
445

446
  @type master_params: L{objects.MasterNetworkParameters}
447
  @param master_params: network parameters of the master
448
  @type use_external_mip_script: boolean
449
  @param use_external_mip_script: whether to use an external master IP
450
    address setup script
451
  @raise RPCFail: in case of errors during the IP turndown
452

453
  """
454
  _RunMasterSetupScript(master_params, _MASTER_STOP,
455
                        use_external_mip_script)
456

    
457

    
458
def StopMasterDaemons():
459
  """Stop the master daemons on this node.
460

461
  Stop the master daemons (ganeti-masterd and ganeti-rapi) on this node.
462

463
  @rtype: None
464

465
  """
466
  # TODO: log and report back to the caller the error failures; we
467
  # need to decide in which case we fail the RPC for this
468

    
469
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop-master"])
470
  if result.failed:
471
    logging.error("Could not stop Ganeti master, command %s had exitcode %s"
472
                  " and error %s",
473
                  result.cmd, result.exit_code, result.output)
474

    
475

    
476
def ChangeMasterNetmask(old_netmask, netmask, master_ip, master_netdev):
477
  """Change the netmask of the master IP.
478

479
  @param old_netmask: the old value of the netmask
480
  @param netmask: the new value of the netmask
481
  @param master_ip: the master IP
482
  @param master_netdev: the master network device
483

484
  """
485
  if old_netmask == netmask:
486
    return
487

    
488
  if not netutils.IPAddress.Own(master_ip):
489
    _Fail("The master IP address is not up, not attempting to change its"
490
          " netmask")
491

    
492
  result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "add",
493
                         "%s/%s" % (master_ip, netmask),
494
                         "dev", master_netdev, "label",
495
                         "%s:0" % master_netdev])
496
  if result.failed:
497
    _Fail("Could not set the new netmask on the master IP address")
498

    
499
  result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "del",
500
                         "%s/%s" % (master_ip, old_netmask),
501
                         "dev", master_netdev, "label",
502
                         "%s:0" % master_netdev])
503
  if result.failed:
504
    _Fail("Could not bring down the master IP address with the old netmask")
505

    
506

    
507
def EtcHostsModify(mode, host, ip):
508
  """Modify a host entry in /etc/hosts.
509

510
  @param mode: The mode to operate. Either add or remove entry
511
  @param host: The host to operate on
512
  @param ip: The ip associated with the entry
513

514
  """
515
  if mode == constants.ETC_HOSTS_ADD:
516
    if not ip:
517
      RPCFail("Mode 'add' needs 'ip' parameter, but parameter not"
518
              " present")
519
    utils.AddHostToEtcHosts(host, ip)
520
  elif mode == constants.ETC_HOSTS_REMOVE:
521
    if ip:
522
      RPCFail("Mode 'remove' does not allow 'ip' parameter, but"
523
              " parameter is present")
524
    utils.RemoveHostFromEtcHosts(host)
525
  else:
526
    RPCFail("Mode not supported")
527

    
528

    
529
def LeaveCluster(modify_ssh_setup):
530
  """Cleans up and remove the current node.
531

532
  This function cleans up and prepares the current node to be removed
533
  from the cluster.
534

535
  If processing is successful, then it raises an
536
  L{errors.QuitGanetiException} which is used as a special case to
537
  shutdown the node daemon.
538

539
  @param modify_ssh_setup: boolean
540

541
  """
542
  _CleanDirectory(pathutils.DATA_DIR)
543
  _CleanDirectory(pathutils.CRYPTO_KEYS_DIR)
544
  JobQueuePurge()
545

    
546
  if modify_ssh_setup:
547
    try:
548
      priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.SSH_LOGIN_USER)
549

    
550
      utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
551

    
552
      utils.RemoveFile(priv_key)
553
      utils.RemoveFile(pub_key)
554
    except errors.OpExecError:
555
      logging.exception("Error while processing ssh files")
556

    
557
  try:
558
    utils.RemoveFile(pathutils.CONFD_HMAC_KEY)
559
    utils.RemoveFile(pathutils.RAPI_CERT_FILE)
560
    utils.RemoveFile(pathutils.SPICE_CERT_FILE)
561
    utils.RemoveFile(pathutils.SPICE_CACERT_FILE)
562
    utils.RemoveFile(pathutils.NODED_CERT_FILE)
563
  except: # pylint: disable=W0702
564
    logging.exception("Error while removing cluster secrets")
565

    
566
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop", constants.CONFD])
567
  if result.failed:
568
    logging.error("Command %s failed with exitcode %s and error %s",
569
                  result.cmd, result.exit_code, result.output)
570

    
571
  # Raise a custom exception (handled in ganeti-noded)
572
  raise errors.QuitGanetiException(True, "Shutdown scheduled")
573

    
574

    
575
def _GetVgInfo(name, excl_stor):
576
  """Retrieves information about a LVM volume group.
577

578
  """
579
  # TODO: GetVGInfo supports returning information for multiple VGs at once
580
  vginfo = bdev.LogicalVolume.GetVGInfo([name], excl_stor)
581
  if vginfo:
582
    vg_free = int(round(vginfo[0][0], 0))
583
    vg_size = int(round(vginfo[0][1], 0))
584
  else:
585
    vg_free = None
586
    vg_size = None
587

    
588
  return {
589
    "name": name,
590
    "vg_free": vg_free,
591
    "vg_size": vg_size,
592
    }
593

    
594

    
595
def _GetVgSpindlesInfo(name, excl_stor):
596
  """Retrieves information about spindles in an LVM volume group.
597

598
  @type name: string
599
  @param name: VG name
600
  @type excl_stor: bool
601
  @param excl_stor: exclusive storage
602
  @rtype: dict
603
  @return: dictionary whose keys are "name", "vg_free", "vg_size" for VG name,
604
      free spindles, total spindles respectively
605

606
  """
607
  if excl_stor:
608
    (vg_free, vg_size) = bdev.LogicalVolume.GetVgSpindlesInfo(name)
609
  else:
610
    vg_free = 0
611
    vg_size = 0
612
  return {
613
    "name": name,
614
    "vg_free": vg_free,
615
    "vg_size": vg_size,
616
    }
617

    
618

    
619
def _GetHvInfo(name):
620
  """Retrieves node information from a hypervisor.
621

622
  The information returned depends on the hypervisor. Common items:
623

624
    - vg_size is the size of the configured volume group in MiB
625
    - vg_free is the free size of the volume group in MiB
626
    - memory_dom0 is the memory allocated for domain0 in MiB
627
    - memory_free is the currently available (free) ram in MiB
628
    - memory_total is the total number of ram in MiB
629
    - hv_version: the hypervisor version, if available
630

631
  """
632
  return hypervisor.GetHypervisor(name).GetNodeInfo()
633

    
634

    
635
def _GetNamedNodeInfo(names, fn):
636
  """Calls C{fn} for all names in C{names} and returns a dictionary.
637

638
  @rtype: None or dict
639

640
  """
641
  if names is None:
642
    return None
643
  else:
644
    return map(fn, names)
645

    
646

    
647
def GetNodeInfo(storage_units, hv_names, excl_stor):
648
  """Gives back a hash with different information about the node.
649

650
  @type storage_units: list of pairs (string, string)
651
  @param storage_units: List of pairs (storage unit, identifier) to ask for disk
652
                        space information. In case of lvm-vg, the identifier is
653
                        the VG name.
654
  @type hv_names: list of string
655
  @param hv_names: Names of the hypervisors to ask for node information
656
  @type excl_stor: boolean
657
  @param excl_stor: Whether exclusive_storage is active
658
  @rtype: tuple; (string, None/dict, None/dict)
659
  @return: Tuple containing boot ID, volume group information and hypervisor
660
    information
661

662
  """
663
  bootid = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
664
  storage_info = _GetNamedNodeInfo(
665
    storage_units,
666
    (lambda storage_unit: _ApplyStorageInfoFunction(storage_unit[0],
667
                                                    storage_unit[1],
668
                                                    excl_stor)))
669
  hv_info = _GetNamedNodeInfo(hv_names, _GetHvInfo)
670

    
671
  return (bootid, storage_info, hv_info)
672

    
673

    
674
# FIXME: implement storage reporting for all missing storage types.
675
_STORAGE_TYPE_INFO_FN = {
676
  constants.ST_BLOCK: None,
677
  constants.ST_DISKLESS: None,
678
  constants.ST_EXT: None,
679
  constants.ST_FILE: None,
680
  constants.ST_LVM_PV: _GetVgSpindlesInfo,
681
  constants.ST_LVM_VG: _GetVgInfo,
682
  constants.ST_RADOS: None,
683
}
684

    
685

    
686
def _ApplyStorageInfoFunction(storage_type, storage_key, *args):
687
  """Looks up and applies the correct function to calculate free and total
688
  storage for the given storage type.
689

690
  @type storage_type: string
691
  @param storage_type: the storage type for which the storage shall be reported.
692
  @type storage_key: string
693
  @param storage_key: identifier of a storage unit, e.g. the volume group name
694
    of an LVM storage unit
695
  @type args: any
696
  @param args: various parameters that can be used for storage reporting. These
697
    parameters and their semantics vary from storage type to storage type and
698
    are just propagated in this function.
699
  @return: the results of the application of the storage space function (see
700
    _STORAGE_TYPE_INFO_FN) if storage space reporting is implemented for that
701
    storage type
702
  @raises NotImplementedError: for storage types who don't support space
703
    reporting yet
704
  """
705
  fn = _STORAGE_TYPE_INFO_FN[storage_type]
706
  if fn is not None:
707
    return fn(storage_key, *args)
708
  else:
709
    raise NotImplementedError
710

    
711

    
712
def _CheckExclusivePvs(pvi_list):
713
  """Check that PVs are not shared among LVs
714

715
  @type pvi_list: list of L{objects.LvmPvInfo} objects
716
  @param pvi_list: information about the PVs
717

718
  @rtype: list of tuples (string, list of strings)
719
  @return: offending volumes, as tuples: (pv_name, [lv1_name, lv2_name...])
720

721
  """
722
  res = []
723
  for pvi in pvi_list:
724
    if len(pvi.lv_list) > 1:
725
      res.append((pvi.name, pvi.lv_list))
726
  return res
727

    
728

    
729
def VerifyNode(what, cluster_name):
730
  """Verify the status of the local node.
731

732
  Based on the input L{what} parameter, various checks are done on the
733
  local node.
734

735
  If the I{filelist} key is present, this list of
736
  files is checksummed and the file/checksum pairs are returned.
737

738
  If the I{nodelist} key is present, we check that we have
739
  connectivity via ssh with the target nodes (and check the hostname
740
  report).
741

742
  If the I{node-net-test} key is present, we check that we have
743
  connectivity to the given nodes via both primary IP and, if
744
  applicable, secondary IPs.
745

746
  @type what: C{dict}
747
  @param what: a dictionary of things to check:
748
      - filelist: list of files for which to compute checksums
749
      - nodelist: list of nodes we should check ssh communication with
750
      - node-net-test: list of nodes we should check node daemon port
751
        connectivity with
752
      - hypervisor: list with hypervisors to run the verify for
753
  @rtype: dict
754
  @return: a dictionary with the same keys as the input dict, and
755
      values representing the result of the checks
756

757
  """
758
  result = {}
759
  my_name = netutils.Hostname.GetSysName()
760
  port = netutils.GetDaemonPort(constants.NODED)
761
  vm_capable = my_name not in what.get(constants.NV_VMNODES, [])
762

    
763
  if constants.NV_HYPERVISOR in what and vm_capable:
764
    result[constants.NV_HYPERVISOR] = tmp = {}
765
    for hv_name in what[constants.NV_HYPERVISOR]:
766
      try:
767
        val = hypervisor.GetHypervisor(hv_name).Verify()
768
      except errors.HypervisorError, err:
769
        val = "Error while checking hypervisor: %s" % str(err)
770
      tmp[hv_name] = val
771

    
772
  if constants.NV_HVPARAMS in what and vm_capable:
773
    result[constants.NV_HVPARAMS] = tmp = []
774
    for source, hv_name, hvparms in what[constants.NV_HVPARAMS]:
775
      try:
776
        logging.info("Validating hv %s, %s", hv_name, hvparms)
777
        hypervisor.GetHypervisor(hv_name).ValidateParameters(hvparms)
778
      except errors.HypervisorError, err:
779
        tmp.append((source, hv_name, str(err)))
780

    
781
  if constants.NV_FILELIST in what:
782
    fingerprints = utils.FingerprintFiles(map(vcluster.LocalizeVirtualPath,
783
                                              what[constants.NV_FILELIST]))
784
    result[constants.NV_FILELIST] = \
785
      dict((vcluster.MakeVirtualPath(key), value)
786
           for (key, value) in fingerprints.items())
787

    
788
  if constants.NV_NODELIST in what:
789
    (nodes, bynode) = what[constants.NV_NODELIST]
790

    
791
    # Add nodes from other groups (different for each node)
792
    try:
793
      nodes.extend(bynode[my_name])
794
    except KeyError:
795
      pass
796

    
797
    # Use a random order
798
    random.shuffle(nodes)
799

    
800
    # Try to contact all nodes
801
    val = {}
802
    for node in nodes:
803
      success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
804
      if not success:
805
        val[node] = message
806

    
807
    result[constants.NV_NODELIST] = val
808

    
809
  if constants.NV_NODENETTEST in what:
810
    result[constants.NV_NODENETTEST] = tmp = {}
811
    my_pip = my_sip = None
812
    for name, pip, sip in what[constants.NV_NODENETTEST]:
813
      if name == my_name:
814
        my_pip = pip
815
        my_sip = sip
816
        break
817
    if not my_pip:
818
      tmp[my_name] = ("Can't find my own primary/secondary IP"
819
                      " in the node list")
820
    else:
821
      for name, pip, sip in what[constants.NV_NODENETTEST]:
822
        fail = []
823
        if not netutils.TcpPing(pip, port, source=my_pip):
824
          fail.append("primary")
825
        if sip != pip:
826
          if not netutils.TcpPing(sip, port, source=my_sip):
827
            fail.append("secondary")
828
        if fail:
829
          tmp[name] = ("failure using the %s interface(s)" %
830
                       " and ".join(fail))
831

    
832
  if constants.NV_MASTERIP in what:
833
    # FIXME: add checks on incoming data structures (here and in the
834
    # rest of the function)
835
    master_name, master_ip = what[constants.NV_MASTERIP]
836
    if master_name == my_name:
837
      source = constants.IP4_ADDRESS_LOCALHOST
838
    else:
839
      source = None
840
    result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port,
841
                                                     source=source)
842

    
843
  if constants.NV_USERSCRIPTS in what:
844
    result[constants.NV_USERSCRIPTS] = \
845
      [script for script in what[constants.NV_USERSCRIPTS]
846
       if not utils.IsExecutable(script)]
847

    
848
  if constants.NV_OOB_PATHS in what:
849
    result[constants.NV_OOB_PATHS] = tmp = []
850
    for path in what[constants.NV_OOB_PATHS]:
851
      try:
852
        st = os.stat(path)
853
      except OSError, err:
854
        tmp.append("error stating out of band helper: %s" % err)
855
      else:
856
        if stat.S_ISREG(st.st_mode):
857
          if stat.S_IMODE(st.st_mode) & stat.S_IXUSR:
858
            tmp.append(None)
859
          else:
860
            tmp.append("out of band helper %s is not executable" % path)
861
        else:
862
          tmp.append("out of band helper %s is not a file" % path)
863

    
864
  if constants.NV_LVLIST in what and vm_capable:
865
    try:
866
      val = GetVolumeList(utils.ListVolumeGroups().keys())
867
    except RPCFail, err:
868
      val = str(err)
869
    result[constants.NV_LVLIST] = val
870

    
871
  if constants.NV_INSTANCELIST in what and vm_capable:
872
    # GetInstanceList can fail
873
    try:
874
      val = GetInstanceList(what[constants.NV_INSTANCELIST])
875
    except RPCFail, err:
876
      val = str(err)
877
    result[constants.NV_INSTANCELIST] = val
878

    
879
  if constants.NV_VGLIST in what and vm_capable:
880
    result[constants.NV_VGLIST] = utils.ListVolumeGroups()
881

    
882
  if constants.NV_PVLIST in what and vm_capable:
883
    check_exclusive_pvs = constants.NV_EXCLUSIVEPVS in what
884
    val = bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
885
                                       filter_allocatable=False,
886
                                       include_lvs=check_exclusive_pvs)
887
    if check_exclusive_pvs:
888
      result[constants.NV_EXCLUSIVEPVS] = _CheckExclusivePvs(val)
889
      for pvi in val:
890
        # Avoid sending useless data on the wire
891
        pvi.lv_list = []
892
    result[constants.NV_PVLIST] = map(objects.LvmPvInfo.ToDict, val)
893

    
894
  if constants.NV_VERSION in what:
895
    result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
896
                                    constants.RELEASE_VERSION)
897

    
898
  if constants.NV_HVINFO in what and vm_capable:
899
    hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
900
    result[constants.NV_HVINFO] = hyper.GetNodeInfo()
901

    
902
  if constants.NV_DRBDVERSION in what and vm_capable:
903
    try:
904
      drbd_version = DRBD8.GetProcInfo().GetVersionString()
905
    except errors.BlockDeviceError, err:
906
      logging.warning("Can't get DRBD version", exc_info=True)
907
      drbd_version = str(err)
908
    result[constants.NV_DRBDVERSION] = drbd_version
909

    
910
  if constants.NV_DRBDLIST in what and vm_capable:
911
    try:
912
      used_minors = drbd.DRBD8.GetUsedDevs()
913
    except errors.BlockDeviceError, err:
914
      logging.warning("Can't get used minors list", exc_info=True)
915
      used_minors = str(err)
916
    result[constants.NV_DRBDLIST] = used_minors
917

    
918
  if constants.NV_DRBDHELPER in what and vm_capable:
919
    status = True
920
    try:
921
      payload = drbd.DRBD8.GetUsermodeHelper()
922
    except errors.BlockDeviceError, err:
923
      logging.error("Can't get DRBD usermode helper: %s", str(err))
924
      status = False
925
      payload = str(err)
926
    result[constants.NV_DRBDHELPER] = (status, payload)
927

    
928
  if constants.NV_NODESETUP in what:
929
    result[constants.NV_NODESETUP] = tmpr = []
930
    if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
931
      tmpr.append("The sysfs filesytem doesn't seem to be mounted"
932
                  " under /sys, missing required directories /sys/block"
933
                  " and /sys/class/net")
934
    if (not os.path.isdir("/proc/sys") or
935
        not os.path.isfile("/proc/sysrq-trigger")):
936
      tmpr.append("The procfs filesystem doesn't seem to be mounted"
937
                  " under /proc, missing required directory /proc/sys and"
938
                  " the file /proc/sysrq-trigger")
939

    
940
  if constants.NV_TIME in what:
941
    result[constants.NV_TIME] = utils.SplitTime(time.time())
942

    
943
  if constants.NV_OSLIST in what and vm_capable:
944
    result[constants.NV_OSLIST] = DiagnoseOS()
945

    
946
  if constants.NV_BRIDGES in what and vm_capable:
947
    result[constants.NV_BRIDGES] = [bridge
948
                                    for bridge in what[constants.NV_BRIDGES]
949
                                    if not utils.BridgeExists(bridge)]
950

    
951
  if what.get(constants.NV_FILE_STORAGE_PATHS) == my_name:
952
    result[constants.NV_FILE_STORAGE_PATHS] = \
953
      bdev.ComputeWrongFileStoragePaths()
954

    
955
  return result
956

    
957

    
958
def GetBlockDevSizes(devices):
959
  """Return the size of the given block devices
960

961
  @type devices: list
962
  @param devices: list of block device nodes to query
963
  @rtype: dict
964
  @return:
965
    dictionary of all block devices under /dev (key). The value is their
966
    size in MiB.
967

968
    {'/dev/disk/by-uuid/123456-12321231-312312-312': 124}
969

970
  """
971
  DEV_PREFIX = "/dev/"
972
  blockdevs = {}
973

    
974
  for devpath in devices:
975
    if not utils.IsBelowDir(DEV_PREFIX, devpath):
976
      continue
977

    
978
    try:
979
      st = os.stat(devpath)
980
    except EnvironmentError, err:
981
      logging.warning("Error stat()'ing device %s: %s", devpath, str(err))
982
      continue
983

    
984
    if stat.S_ISBLK(st.st_mode):
985
      result = utils.RunCmd(["blockdev", "--getsize64", devpath])
986
      if result.failed:
987
        # We don't want to fail, just do not list this device as available
988
        logging.warning("Cannot get size for block device %s", devpath)
989
        continue
990

    
991
      size = int(result.stdout) / (1024 * 1024)
992
      blockdevs[devpath] = size
993
  return blockdevs
994

    
995

    
996
def GetVolumeList(vg_names):
997
  """Compute list of logical volumes and their size.
998

999
  @type vg_names: list
1000
  @param vg_names: the volume groups whose LVs we should list, or
1001
      empty for all volume groups
1002
  @rtype: dict
1003
  @return:
1004
      dictionary of all partions (key) with value being a tuple of
1005
      their size (in MiB), inactive and online status::
1006

1007
        {'xenvg/test1': ('20.06', True, True)}
1008

1009
      in case of errors, a string is returned with the error
1010
      details.
1011

1012
  """
1013
  lvs = {}
1014
  sep = "|"
1015
  if not vg_names:
1016
    vg_names = []
1017
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
1018
                         "--separator=%s" % sep,
1019
                         "-ovg_name,lv_name,lv_size,lv_attr"] + vg_names)
1020
  if result.failed:
1021
    _Fail("Failed to list logical volumes, lvs output: %s", result.output)
1022

    
1023
  for line in result.stdout.splitlines():
1024
    line = line.strip()
1025
    match = _LVSLINE_REGEX.match(line)
1026
    if not match:
1027
      logging.error("Invalid line returned from lvs output: '%s'", line)
1028
      continue
1029
    vg_name, name, size, attr = match.groups()
1030
    inactive = attr[4] == "-"
1031
    online = attr[5] == "o"
1032
    virtual = attr[0] == "v"
1033
    if virtual:
1034
      # we don't want to report such volumes as existing, since they
1035
      # don't really hold data
1036
      continue
1037
    lvs[vg_name + "/" + name] = (size, inactive, online)
1038

    
1039
  return lvs
1040

    
1041

    
1042
def ListVolumeGroups():
1043
  """List the volume groups and their size.
1044

1045
  @rtype: dict
1046
  @return: dictionary with keys volume name and values the
1047
      size of the volume
1048

1049
  """
1050
  return utils.ListVolumeGroups()
1051

    
1052

    
1053
def NodeVolumes():
1054
  """List all volumes on this node.
1055

1056
  @rtype: list
1057
  @return:
1058
    A list of dictionaries, each having four keys:
1059
      - name: the logical volume name,
1060
      - size: the size of the logical volume
1061
      - dev: the physical device on which the LV lives
1062
      - vg: the volume group to which it belongs
1063

1064
    In case of errors, we return an empty list and log the
1065
    error.
1066

1067
    Note that since a logical volume can live on multiple physical
1068
    volumes, the resulting list might include a logical volume
1069
    multiple times.
1070

1071
  """
1072
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
1073
                         "--separator=|",
1074
                         "--options=lv_name,lv_size,devices,vg_name"])
1075
  if result.failed:
1076
    _Fail("Failed to list logical volumes, lvs output: %s",
1077
          result.output)
1078

    
1079
  def parse_dev(dev):
1080
    return dev.split("(")[0]
1081

    
1082
  def handle_dev(dev):
1083
    return [parse_dev(x) for x in dev.split(",")]
1084

    
1085
  def map_line(line):
1086
    line = [v.strip() for v in line]
1087
    return [{"name": line[0], "size": line[1],
1088
             "dev": dev, "vg": line[3]} for dev in handle_dev(line[2])]
1089

    
1090
  all_devs = []
1091
  for line in result.stdout.splitlines():
1092
    if line.count("|") >= 3:
1093
      all_devs.extend(map_line(line.split("|")))
1094
    else:
1095
      logging.warning("Strange line in the output from lvs: '%s'", line)
1096
  return all_devs
1097

    
1098

    
1099
def BridgesExist(bridges_list):
1100
  """Check if a list of bridges exist on the current node.
1101

1102
  @rtype: boolean
1103
  @return: C{True} if all of them exist, C{False} otherwise
1104

1105
  """
1106
  missing = []
1107
  for bridge in bridges_list:
1108
    if not utils.BridgeExists(bridge):
1109
      missing.append(bridge)
1110

    
1111
  if missing:
1112
    _Fail("Missing bridges %s", utils.CommaJoin(missing))
1113

    
1114

    
1115
def GetInstanceListForHypervisor(hname, hvparams=None,
1116
                                 get_hv_fn=hypervisor.GetHypervisor):
1117
  """Provides a list of instances of the given hypervisor.
1118

1119
  @type hname: string
1120
  @param hname: name of the hypervisor
1121
  @type hvparams: dict of strings
1122
  @param hvparams: hypervisor parameters for the given hypervisor
1123
  @type get_hv_fn: function
1124
  @param get_hv_fn: function that returns a hypervisor for the given hypervisor
1125
    name; optional parameter to increase testability
1126

1127
  @rtype: list
1128
  @return: a list of all running instances on the current node
1129
    - instance1.example.com
1130
    - instance2.example.com
1131

1132
  """
1133
  results = []
1134
  try:
1135
    hv = get_hv_fn(hname)
1136
    names = hv.ListInstances(hvparams)
1137
    results.extend(names)
1138
  except errors.HypervisorError, err:
1139
    _Fail("Error enumerating instances (hypervisor %s): %s",
1140
          hname, err, exc=True)
1141
  return results
1142

    
1143

    
1144
def GetInstanceList(hypervisor_list, all_hvparams=None,
1145
                    get_hv_fn=hypervisor.GetHypervisor):
1146
  """Provides a list of instances.
1147

1148
  @type hypervisor_list: list
1149
  @param hypervisor_list: the list of hypervisors to query information
1150
  @type all_hvparams: dict of dict of strings
1151
  @param all_hvparams: a dictionary mapping hypervisor types to respective
1152
    cluster-wide hypervisor parameters
1153
  @type get_hv_fn: function
1154
  @param get_hv_fn: function that returns a hypervisor for the given hypervisor
1155
    name; optional parameter to increase testability
1156

1157
  @rtype: list
1158
  @return: a list of all running instances on the current node
1159
    - instance1.example.com
1160
    - instance2.example.com
1161

1162
  """
1163
  results = []
1164
  for hname in hypervisor_list:
1165
    hvparams = None
1166
    if all_hvparams is not None:
1167
      hvparams = all_hvparams[hname]
1168
    results.extend(GetInstanceListForHypervisor(hname, hvparams,
1169
                                                get_hv_fn=get_hv_fn))
1170
  return results
1171

    
1172

    
1173
def GetInstanceInfo(instance, hname):
1174
  """Gives back the information about an instance as a dictionary.
1175

1176
  @type instance: string
1177
  @param instance: the instance name
1178
  @type hname: string
1179
  @param hname: the hypervisor type of the instance
1180

1181
  @rtype: dict
1182
  @return: dictionary with the following keys:
1183
      - memory: memory size of instance (int)
1184
      - state: xen state of instance (string)
1185
      - time: cpu time of instance (float)
1186
      - vcpus: the number of vcpus (int)
1187

1188
  """
1189
  output = {}
1190

    
1191
  iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
1192
  if iinfo is not None:
1193
    output["memory"] = iinfo[2]
1194
    output["vcpus"] = iinfo[3]
1195
    output["state"] = iinfo[4]
1196
    output["time"] = iinfo[5]
1197

    
1198
  return output
1199

    
1200

    
1201
def GetInstanceMigratable(instance):
1202
  """Computes whether an instance can be migrated.
1203

1204
  @type instance: L{objects.Instance}
1205
  @param instance: object representing the instance to be checked.
1206

1207
  @rtype: tuple
1208
  @return: tuple of (result, description) where:
1209
      - result: whether the instance can be migrated or not
1210
      - description: a description of the issue, if relevant
1211

1212
  """
1213
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1214
  iname = instance.name
1215
  if iname not in hyper.ListInstances(instance.hvparams):
1216
    _Fail("Instance %s is not running", iname)
1217

    
1218
  for idx in range(len(instance.disks)):
1219
    link_name = _GetBlockDevSymlinkPath(iname, idx)
1220
    if not os.path.islink(link_name):
1221
      logging.warning("Instance %s is missing symlink %s for disk %d",
1222
                      iname, link_name, idx)
1223

    
1224

    
1225
def GetAllInstancesInfo(hypervisor_list):
1226
  """Gather data about all instances.
1227

1228
  This is the equivalent of L{GetInstanceInfo}, except that it
1229
  computes data for all instances at once, thus being faster if one
1230
  needs data about more than one instance.
1231

1232
  @type hypervisor_list: list
1233
  @param hypervisor_list: list of hypervisors to query for instance data
1234

1235
  @rtype: dict
1236
  @return: dictionary of instance: data, with data having the following keys:
1237
      - memory: memory size of instance (int)
1238
      - state: xen state of instance (string)
1239
      - time: cpu time of instance (float)
1240
      - vcpus: the number of vcpus
1241

1242
  """
1243
  output = {}
1244

    
1245
  for hname in hypervisor_list:
1246
    iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
1247
    if iinfo:
1248
      for name, _, memory, vcpus, state, times in iinfo:
1249
        value = {
1250
          "memory": memory,
1251
          "vcpus": vcpus,
1252
          "state": state,
1253
          "time": times,
1254
          }
1255
        if name in output:
1256
          # we only check static parameters, like memory and vcpus,
1257
          # and not state and time which can change between the
1258
          # invocations of the different hypervisors
1259
          for key in "memory", "vcpus":
1260
            if value[key] != output[name][key]:
1261
              _Fail("Instance %s is running twice"
1262
                    " with different parameters", name)
1263
        output[name] = value
1264

    
1265
  return output
1266

    
1267

    
1268
def _InstanceLogName(kind, os_name, instance, component):
1269
  """Compute the OS log filename for a given instance and operation.
1270

1271
  The instance name and os name are passed in as strings since not all
1272
  operations have these as part of an instance object.
1273

1274
  @type kind: string
1275
  @param kind: the operation type (e.g. add, import, etc.)
1276
  @type os_name: string
1277
  @param os_name: the os name
1278
  @type instance: string
1279
  @param instance: the name of the instance being imported/added/etc.
1280
  @type component: string or None
1281
  @param component: the name of the component of the instance being
1282
      transferred
1283

1284
  """
1285
  # TODO: Use tempfile.mkstemp to create unique filename
1286
  if component:
1287
    assert "/" not in component
1288
    c_msg = "-%s" % component
1289
  else:
1290
    c_msg = ""
1291
  base = ("%s-%s-%s%s-%s.log" %
1292
          (kind, os_name, instance, c_msg, utils.TimestampForFilename()))
1293
  return utils.PathJoin(pathutils.LOG_OS_DIR, base)
1294

    
1295

    
1296
def InstanceOsAdd(instance, reinstall, debug):
1297
  """Add an OS to an instance.
1298

1299
  @type instance: L{objects.Instance}
1300
  @param instance: Instance whose OS is to be installed
1301
  @type reinstall: boolean
1302
  @param reinstall: whether this is an instance reinstall
1303
  @type debug: integer
1304
  @param debug: debug level, passed to the OS scripts
1305
  @rtype: None
1306

1307
  """
1308
  inst_os = OSFromDisk(instance.os)
1309

    
1310
  create_env = OSEnvironment(instance, inst_os, debug)
1311
  if reinstall:
1312
    create_env["INSTANCE_REINSTALL"] = "1"
1313

    
1314
  logfile = _InstanceLogName("add", instance.os, instance.name, None)
1315

    
1316
  result = utils.RunCmd([inst_os.create_script], env=create_env,
1317
                        cwd=inst_os.path, output=logfile, reset_env=True)
1318
  if result.failed:
1319
    logging.error("os create command '%s' returned error: %s, logfile: %s,"
1320
                  " output: %s", result.cmd, result.fail_reason, logfile,
1321
                  result.output)
1322
    lines = [utils.SafeEncode(val)
1323
             for val in utils.TailFile(logfile, lines=20)]
1324
    _Fail("OS create script failed (%s), last lines in the"
1325
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1326

    
1327

    
1328
def RunRenameInstance(instance, old_name, debug):
1329
  """Run the OS rename script for an instance.
1330

1331
  @type instance: L{objects.Instance}
1332
  @param instance: Instance whose OS is to be installed
1333
  @type old_name: string
1334
  @param old_name: previous instance name
1335
  @type debug: integer
1336
  @param debug: debug level, passed to the OS scripts
1337
  @rtype: boolean
1338
  @return: the success of the operation
1339

1340
  """
1341
  inst_os = OSFromDisk(instance.os)
1342

    
1343
  rename_env = OSEnvironment(instance, inst_os, debug)
1344
  rename_env["OLD_INSTANCE_NAME"] = old_name
1345

    
1346
  logfile = _InstanceLogName("rename", instance.os,
1347
                             "%s-%s" % (old_name, instance.name), None)
1348

    
1349
  result = utils.RunCmd([inst_os.rename_script], env=rename_env,
1350
                        cwd=inst_os.path, output=logfile, reset_env=True)
1351

    
1352
  if result.failed:
1353
    logging.error("os create command '%s' returned error: %s output: %s",
1354
                  result.cmd, result.fail_reason, result.output)
1355
    lines = [utils.SafeEncode(val)
1356
             for val in utils.TailFile(logfile, lines=20)]
1357
    _Fail("OS rename script failed (%s), last lines in the"
1358
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1359

    
1360

    
1361
def _GetBlockDevSymlinkPath(instance_name, idx, _dir=None):
1362
  """Returns symlink path for block device.
1363

1364
  """
1365
  if _dir is None:
1366
    _dir = pathutils.DISK_LINKS_DIR
1367

    
1368
  return utils.PathJoin(_dir,
1369
                        ("%s%s%s" %
1370
                         (instance_name, constants.DISK_SEPARATOR, idx)))
1371

    
1372

    
1373
def _SymlinkBlockDev(instance_name, device_path, idx):
1374
  """Set up symlinks to a instance's block device.
1375

1376
  This is an auxiliary function run when an instance is start (on the primary
1377
  node) or when an instance is migrated (on the target node).
1378

1379

1380
  @param instance_name: the name of the target instance
1381
  @param device_path: path of the physical block device, on the node
1382
  @param idx: the disk index
1383
  @return: absolute path to the disk's symlink
1384

1385
  """
1386
  link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1387
  try:
1388
    os.symlink(device_path, link_name)
1389
  except OSError, err:
1390
    if err.errno == errno.EEXIST:
1391
      if (not os.path.islink(link_name) or
1392
          os.readlink(link_name) != device_path):
1393
        os.remove(link_name)
1394
        os.symlink(device_path, link_name)
1395
    else:
1396
      raise
1397

    
1398
  return link_name
1399

    
1400

    
1401
def _RemoveBlockDevLinks(instance_name, disks):
1402
  """Remove the block device symlinks belonging to the given instance.
1403

1404
  """
1405
  for idx, _ in enumerate(disks):
1406
    link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1407
    if os.path.islink(link_name):
1408
      try:
1409
        os.remove(link_name)
1410
      except OSError:
1411
        logging.exception("Can't remove symlink '%s'", link_name)
1412

    
1413

    
1414
def _GatherAndLinkBlockDevs(instance):
1415
  """Set up an instance's block device(s).
1416

1417
  This is run on the primary node at instance startup. The block
1418
  devices must be already assembled.
1419

1420
  @type instance: L{objects.Instance}
1421
  @param instance: the instance whose disks we shoul assemble
1422
  @rtype: list
1423
  @return: list of (disk_object, device_path)
1424

1425
  """
1426
  block_devices = []
1427
  for idx, disk in enumerate(instance.disks):
1428
    device = _RecursiveFindBD(disk)
1429
    if device is None:
1430
      raise errors.BlockDeviceError("Block device '%s' is not set up." %
1431
                                    str(disk))
1432
    device.Open()
1433
    try:
1434
      link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1435
    except OSError, e:
1436
      raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1437
                                    e.strerror)
1438

    
1439
    block_devices.append((disk, link_name))
1440

    
1441
  return block_devices
1442

    
1443

    
1444
def StartInstance(instance, startup_paused, reason, store_reason=True):
1445
  """Start an instance.
1446

1447
  @type instance: L{objects.Instance}
1448
  @param instance: the instance object
1449
  @type startup_paused: bool
1450
  @param instance: pause instance at startup?
1451
  @type reason: list of reasons
1452
  @param reason: the reason trail for this startup
1453
  @type store_reason: boolean
1454
  @param store_reason: whether to store the shutdown reason trail on file
1455
  @rtype: None
1456

1457
  """
1458
  running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1459
                                                   instance.hvparams)
1460

    
1461
  if instance.name in running_instances:
1462
    logging.info("Instance %s already running, not starting", instance.name)
1463
    return
1464

    
1465
  try:
1466
    block_devices = _GatherAndLinkBlockDevs(instance)
1467
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
1468
    hyper.StartInstance(instance, block_devices, startup_paused)
1469
    if store_reason:
1470
      _StoreInstReasonTrail(instance.name, reason)
1471
  except errors.BlockDeviceError, err:
1472
    _Fail("Block device error: %s", err, exc=True)
1473
  except errors.HypervisorError, err:
1474
    _RemoveBlockDevLinks(instance.name, instance.disks)
1475
    _Fail("Hypervisor error: %s", err, exc=True)
1476

    
1477

    
1478
def InstanceShutdown(instance, timeout, reason, store_reason=True):
1479
  """Shut an instance down.
1480

1481
  @note: this functions uses polling with a hardcoded timeout.
1482

1483
  @type instance: L{objects.Instance}
1484
  @param instance: the instance object
1485
  @type timeout: integer
1486
  @param timeout: maximum timeout for soft shutdown
1487
  @type reason: list of reasons
1488
  @param reason: the reason trail for this shutdown
1489
  @type store_reason: boolean
1490
  @param store_reason: whether to store the shutdown reason trail on file
1491
  @rtype: None
1492

1493
  """
1494
  hv_name = instance.hypervisor
1495
  hyper = hypervisor.GetHypervisor(hv_name)
1496
  iname = instance.name
1497

    
1498
  if instance.name not in hyper.ListInstances(instance.hvparams):
1499
    logging.info("Instance %s not running, doing nothing", iname)
1500
    return
1501

    
1502
  class _TryShutdown:
1503
    def __init__(self):
1504
      self.tried_once = False
1505

    
1506
    def __call__(self):
1507
      if iname not in hyper.ListInstances(instance.hvparams):
1508
        return
1509

    
1510
      try:
1511
        hyper.StopInstance(instance, retry=self.tried_once)
1512
        if store_reason:
1513
          _StoreInstReasonTrail(instance.name, reason)
1514
      except errors.HypervisorError, err:
1515
        if iname not in hyper.ListInstances(instance.hvparams):
1516
          # if the instance is no longer existing, consider this a
1517
          # success and go to cleanup
1518
          return
1519

    
1520
        _Fail("Failed to stop instance %s: %s", iname, err)
1521

    
1522
      self.tried_once = True
1523

    
1524
      raise utils.RetryAgain()
1525

    
1526
  try:
1527
    utils.Retry(_TryShutdown(), 5, timeout)
1528
  except utils.RetryTimeout:
1529
    # the shutdown did not succeed
1530
    logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1531

    
1532
    try:
1533
      hyper.StopInstance(instance, force=True)
1534
    except errors.HypervisorError, err:
1535
      if iname in hyper.ListInstances(instance.hvparams):
1536
        # only raise an error if the instance still exists, otherwise
1537
        # the error could simply be "instance ... unknown"!
1538
        _Fail("Failed to force stop instance %s: %s", iname, err)
1539

    
1540
    time.sleep(1)
1541

    
1542
    if iname in hyper.ListInstances(instance.hvparams):
1543
      _Fail("Could not shutdown instance %s even by destroy", iname)
1544

    
1545
  try:
1546
    hyper.CleanupInstance(instance.name)
1547
  except errors.HypervisorError, err:
1548
    logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1549

    
1550
  _RemoveBlockDevLinks(iname, instance.disks)
1551

    
1552

    
1553
def InstanceReboot(instance, reboot_type, shutdown_timeout, reason):
1554
  """Reboot an instance.
1555

1556
  @type instance: L{objects.Instance}
1557
  @param instance: the instance object to reboot
1558
  @type reboot_type: str
1559
  @param reboot_type: the type of reboot, one the following
1560
    constants:
1561
      - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1562
        instance OS, do not recreate the VM
1563
      - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1564
        restart the VM (at the hypervisor level)
1565
      - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1566
        not accepted here, since that mode is handled differently, in
1567
        cmdlib, and translates into full stop and start of the
1568
        instance (instead of a call_instance_reboot RPC)
1569
  @type shutdown_timeout: integer
1570
  @param shutdown_timeout: maximum timeout for soft shutdown
1571
  @type reason: list of reasons
1572
  @param reason: the reason trail for this reboot
1573
  @rtype: None
1574

1575
  """
1576
  running_instances = GetInstanceListForHypervisor(instance.hypervisor,
1577
                                                   instance.hvparams)
1578

    
1579
  if instance.name not in running_instances:
1580
    _Fail("Cannot reboot instance %s that is not running", instance.name)
1581

    
1582
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1583
  if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1584
    try:
1585
      hyper.RebootInstance(instance)
1586
    except errors.HypervisorError, err:
1587
      _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1588
  elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1589
    try:
1590
      InstanceShutdown(instance, shutdown_timeout, reason, store_reason=False)
1591
      result = StartInstance(instance, False, reason, store_reason=False)
1592
      _StoreInstReasonTrail(instance.name, reason)
1593
      return result
1594
    except errors.HypervisorError, err:
1595
      _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1596
  else:
1597
    _Fail("Invalid reboot_type received: %s", reboot_type)
1598

    
1599

    
1600
def InstanceBalloonMemory(instance, memory):
1601
  """Resize an instance's memory.
1602

1603
  @type instance: L{objects.Instance}
1604
  @param instance: the instance object
1605
  @type memory: int
1606
  @param memory: new memory amount in MB
1607
  @rtype: None
1608

1609
  """
1610
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1611
  running = hyper.ListInstances(instance.hvparams)
1612
  if instance.name not in running:
1613
    logging.info("Instance %s is not running, cannot balloon", instance.name)
1614
    return
1615
  try:
1616
    hyper.BalloonInstanceMemory(instance, memory)
1617
  except errors.HypervisorError, err:
1618
    _Fail("Failed to balloon instance memory: %s", err, exc=True)
1619

    
1620

    
1621
def MigrationInfo(instance):
1622
  """Gather information about an instance to be migrated.
1623

1624
  @type instance: L{objects.Instance}
1625
  @param instance: the instance definition
1626

1627
  """
1628
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1629
  try:
1630
    info = hyper.MigrationInfo(instance)
1631
  except errors.HypervisorError, err:
1632
    _Fail("Failed to fetch migration information: %s", err, exc=True)
1633
  return info
1634

    
1635

    
1636
def AcceptInstance(instance, info, target):
1637
  """Prepare the node to accept an instance.
1638

1639
  @type instance: L{objects.Instance}
1640
  @param instance: the instance definition
1641
  @type info: string/data (opaque)
1642
  @param info: migration information, from the source node
1643
  @type target: string
1644
  @param target: target host (usually ip), on this node
1645

1646
  """
1647
  # TODO: why is this required only for DTS_EXT_MIRROR?
1648
  if instance.disk_template in constants.DTS_EXT_MIRROR:
1649
    # Create the symlinks, as the disks are not active
1650
    # in any way
1651
    try:
1652
      _GatherAndLinkBlockDevs(instance)
1653
    except errors.BlockDeviceError, err:
1654
      _Fail("Block device error: %s", err, exc=True)
1655

    
1656
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1657
  try:
1658
    hyper.AcceptInstance(instance, info, target)
1659
  except errors.HypervisorError, err:
1660
    if instance.disk_template in constants.DTS_EXT_MIRROR:
1661
      _RemoveBlockDevLinks(instance.name, instance.disks)
1662
    _Fail("Failed to accept instance: %s", err, exc=True)
1663

    
1664

    
1665
def FinalizeMigrationDst(instance, info, success):
1666
  """Finalize any preparation to accept an instance.
1667

1668
  @type instance: L{objects.Instance}
1669
  @param instance: the instance definition
1670
  @type info: string/data (opaque)
1671
  @param info: migration information, from the source node
1672
  @type success: boolean
1673
  @param success: whether the migration was a success or a failure
1674

1675
  """
1676
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1677
  try:
1678
    hyper.FinalizeMigrationDst(instance, info, success)
1679
  except errors.HypervisorError, err:
1680
    _Fail("Failed to finalize migration on the target node: %s", err, exc=True)
1681

    
1682

    
1683
def MigrateInstance(instance, target, live):
1684
  """Migrates an instance to another node.
1685

1686
  @type instance: L{objects.Instance}
1687
  @param instance: the instance definition
1688
  @type target: string
1689
  @param target: the target node name
1690
  @type live: boolean
1691
  @param live: whether the migration should be done live or not (the
1692
      interpretation of this parameter is left to the hypervisor)
1693
  @raise RPCFail: if migration fails for some reason
1694

1695
  """
1696
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1697

    
1698
  try:
1699
    hyper.MigrateInstance(instance, target, live)
1700
  except errors.HypervisorError, err:
1701
    _Fail("Failed to migrate instance: %s", err, exc=True)
1702

    
1703

    
1704
def FinalizeMigrationSource(instance, success, live):
1705
  """Finalize the instance migration on the source node.
1706

1707
  @type instance: L{objects.Instance}
1708
  @param instance: the instance definition of the migrated instance
1709
  @type success: bool
1710
  @param success: whether the migration succeeded or not
1711
  @type live: bool
1712
  @param live: whether the user requested a live migration or not
1713
  @raise RPCFail: If the execution fails for some reason
1714

1715
  """
1716
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1717

    
1718
  try:
1719
    hyper.FinalizeMigrationSource(instance, success, live)
1720
  except Exception, err:  # pylint: disable=W0703
1721
    _Fail("Failed to finalize the migration on the source node: %s", err,
1722
          exc=True)
1723

    
1724

    
1725
def GetMigrationStatus(instance):
1726
  """Get the migration status
1727

1728
  @type instance: L{objects.Instance}
1729
  @param instance: the instance that is being migrated
1730
  @rtype: L{objects.MigrationStatus}
1731
  @return: the status of the current migration (one of
1732
           L{constants.HV_MIGRATION_VALID_STATUSES}), plus any additional
1733
           progress info that can be retrieved from the hypervisor
1734
  @raise RPCFail: If the migration status cannot be retrieved
1735

1736
  """
1737
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1738
  try:
1739
    return hyper.GetMigrationStatus(instance)
1740
  except Exception, err:  # pylint: disable=W0703
1741
    _Fail("Failed to get migration status: %s", err, exc=True)
1742

    
1743

    
1744
def BlockdevCreate(disk, size, owner, on_primary, info, excl_stor):
1745
  """Creates a block device for an instance.
1746

1747
  @type disk: L{objects.Disk}
1748
  @param disk: the object describing the disk we should create
1749
  @type size: int
1750
  @param size: the size of the physical underlying device, in MiB
1751
  @type owner: str
1752
  @param owner: the name of the instance for which disk is created,
1753
      used for device cache data
1754
  @type on_primary: boolean
1755
  @param on_primary:  indicates if it is the primary node or not
1756
  @type info: string
1757
  @param info: string that will be sent to the physical device
1758
      creation, used for example to set (LVM) tags on LVs
1759
  @type excl_stor: boolean
1760
  @param excl_stor: Whether exclusive_storage is active
1761

1762
  @return: the new unique_id of the device (this can sometime be
1763
      computed only after creation), or None. On secondary nodes,
1764
      it's not required to return anything.
1765

1766
  """
1767
  # TODO: remove the obsolete "size" argument
1768
  # pylint: disable=W0613
1769
  clist = []
1770
  if disk.children:
1771
    for child in disk.children:
1772
      try:
1773
        crdev = _RecursiveAssembleBD(child, owner, on_primary)
1774
      except errors.BlockDeviceError, err:
1775
        _Fail("Can't assemble device %s: %s", child, err)
1776
      if on_primary or disk.AssembleOnSecondary():
1777
        # we need the children open in case the device itself has to
1778
        # be assembled
1779
        try:
1780
          # pylint: disable=E1103
1781
          crdev.Open()
1782
        except errors.BlockDeviceError, err:
1783
          _Fail("Can't make child '%s' read-write: %s", child, err)
1784
      clist.append(crdev)
1785

    
1786
  try:
1787
    device = bdev.Create(disk, clist, excl_stor)
1788
  except errors.BlockDeviceError, err:
1789
    _Fail("Can't create block device: %s", err)
1790

    
1791
  if on_primary or disk.AssembleOnSecondary():
1792
    try:
1793
      device.Assemble()
1794
    except errors.BlockDeviceError, err:
1795
      _Fail("Can't assemble device after creation, unusual event: %s", err)
1796
    if on_primary or disk.OpenOnSecondary():
1797
      try:
1798
        device.Open(force=True)
1799
      except errors.BlockDeviceError, err:
1800
        _Fail("Can't make device r/w after creation, unusual event: %s", err)
1801
    DevCacheManager.UpdateCache(device.dev_path, owner,
1802
                                on_primary, disk.iv_name)
1803

    
1804
  device.SetInfo(info)
1805

    
1806
  return device.unique_id
1807

    
1808

    
1809
def _WipeDevice(path, offset, size):
1810
  """This function actually wipes the device.
1811

1812
  @param path: The path to the device to wipe
1813
  @param offset: The offset in MiB in the file
1814
  @param size: The size in MiB to write
1815

1816
  """
1817
  # Internal sizes are always in Mebibytes; if the following "dd" command
1818
  # should use a different block size the offset and size given to this
1819
  # function must be adjusted accordingly before being passed to "dd".
1820
  block_size = 1024 * 1024
1821

    
1822
  cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
1823
         "bs=%s" % block_size, "oflag=direct", "of=%s" % path,
1824
         "count=%d" % size]
1825
  result = utils.RunCmd(cmd)
1826

    
1827
  if result.failed:
1828
    _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
1829
          result.fail_reason, result.output)
1830

    
1831

    
1832
def BlockdevWipe(disk, offset, size):
1833
  """Wipes a block device.
1834

1835
  @type disk: L{objects.Disk}
1836
  @param disk: the disk object we want to wipe
1837
  @type offset: int
1838
  @param offset: The offset in MiB in the file
1839
  @type size: int
1840
  @param size: The size in MiB to write
1841

1842
  """
1843
  try:
1844
    rdev = _RecursiveFindBD(disk)
1845
  except errors.BlockDeviceError:
1846
    rdev = None
1847

    
1848
  if not rdev:
1849
    _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
1850

    
1851
  # Do cross verify some of the parameters
1852
  if offset < 0:
1853
    _Fail("Negative offset")
1854
  if size < 0:
1855
    _Fail("Negative size")
1856
  if offset > rdev.size:
1857
    _Fail("Offset is bigger than device size")
1858
  if (offset + size) > rdev.size:
1859
    _Fail("The provided offset and size to wipe is bigger than device size")
1860

    
1861
  _WipeDevice(rdev.dev_path, offset, size)
1862

    
1863

    
1864
def BlockdevPauseResumeSync(disks, pause):
1865
  """Pause or resume the sync of the block device.
1866

1867
  @type disks: list of L{objects.Disk}
1868
  @param disks: the disks object we want to pause/resume
1869
  @type pause: bool
1870
  @param pause: Wheater to pause or resume
1871

1872
  """
1873
  success = []
1874
  for disk in disks:
1875
    try:
1876
      rdev = _RecursiveFindBD(disk)
1877
    except errors.BlockDeviceError:
1878
      rdev = None
1879

    
1880
    if not rdev:
1881
      success.append((False, ("Cannot change sync for device %s:"
1882
                              " device not found" % disk.iv_name)))
1883
      continue
1884

    
1885
    result = rdev.PauseResumeSync(pause)
1886

    
1887
    if result:
1888
      success.append((result, None))
1889
    else:
1890
      if pause:
1891
        msg = "Pause"
1892
      else:
1893
        msg = "Resume"
1894
      success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
1895

    
1896
  return success
1897

    
1898

    
1899
def BlockdevRemove(disk):
1900
  """Remove a block device.
1901

1902
  @note: This is intended to be called recursively.
1903

1904
  @type disk: L{objects.Disk}
1905
  @param disk: the disk object we should remove
1906
  @rtype: boolean
1907
  @return: the success of the operation
1908

1909
  """
1910
  msgs = []
1911
  try:
1912
    rdev = _RecursiveFindBD(disk)
1913
  except errors.BlockDeviceError, err:
1914
    # probably can't attach
1915
    logging.info("Can't attach to device %s in remove", disk)
1916
    rdev = None
1917
  if rdev is not None:
1918
    r_path = rdev.dev_path
1919
    try:
1920
      rdev.Remove()
1921
    except errors.BlockDeviceError, err:
1922
      msgs.append(str(err))
1923
    if not msgs:
1924
      DevCacheManager.RemoveCache(r_path)
1925

    
1926
  if disk.children:
1927
    for child in disk.children:
1928
      try:
1929
        BlockdevRemove(child)
1930
      except RPCFail, err:
1931
        msgs.append(str(err))
1932

    
1933
  if msgs:
1934
    _Fail("; ".join(msgs))
1935

    
1936

    
1937
def _RecursiveAssembleBD(disk, owner, as_primary):
1938
  """Activate a block device for an instance.
1939

1940
  This is run on the primary and secondary nodes for an instance.
1941

1942
  @note: this function is called recursively.
1943

1944
  @type disk: L{objects.Disk}
1945
  @param disk: the disk we try to assemble
1946
  @type owner: str
1947
  @param owner: the name of the instance which owns the disk
1948
  @type as_primary: boolean
1949
  @param as_primary: if we should make the block device
1950
      read/write
1951

1952
  @return: the assembled device or None (in case no device
1953
      was assembled)
1954
  @raise errors.BlockDeviceError: in case there is an error
1955
      during the activation of the children or the device
1956
      itself
1957

1958
  """
1959
  children = []
1960
  if disk.children:
1961
    mcn = disk.ChildrenNeeded()
1962
    if mcn == -1:
1963
      mcn = 0 # max number of Nones allowed
1964
    else:
1965
      mcn = len(disk.children) - mcn # max number of Nones
1966
    for chld_disk in disk.children:
1967
      try:
1968
        cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1969
      except errors.BlockDeviceError, err:
1970
        if children.count(None) >= mcn:
1971
          raise
1972
        cdev = None
1973
        logging.error("Error in child activation (but continuing): %s",
1974
                      str(err))
1975
      children.append(cdev)
1976

    
1977
  if as_primary or disk.AssembleOnSecondary():
1978
    r_dev = bdev.Assemble(disk, children)
1979
    result = r_dev
1980
    if as_primary or disk.OpenOnSecondary():
1981
      r_dev.Open()
1982
    DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1983
                                as_primary, disk.iv_name)
1984

    
1985
  else:
1986
    result = True
1987
  return result
1988

    
1989

    
1990
def BlockdevAssemble(disk, owner, as_primary, idx):
1991
  """Activate a block device for an instance.
1992

1993
  This is a wrapper over _RecursiveAssembleBD.
1994

1995
  @rtype: str or boolean
1996
  @return: a C{/dev/...} path for primary nodes, and
1997
      C{True} for secondary nodes
1998

1999
  """
2000
  try:
2001
    result = _RecursiveAssembleBD(disk, owner, as_primary)
2002
    if isinstance(result, BlockDev):
2003
      # pylint: disable=E1103
2004
      result = result.dev_path
2005
      if as_primary:
2006
        _SymlinkBlockDev(owner, result, idx)
2007
  except errors.BlockDeviceError, err:
2008
    _Fail("Error while assembling disk: %s", err, exc=True)
2009
  except OSError, err:
2010
    _Fail("Error while symlinking disk: %s", err, exc=True)
2011

    
2012
  return result
2013

    
2014

    
2015
def BlockdevShutdown(disk):
2016
  """Shut down a block device.
2017

2018
  First, if the device is assembled (Attach() is successful), then
2019
  the device is shutdown. Then the children of the device are
2020
  shutdown.
2021

2022
  This function is called recursively. Note that we don't cache the
2023
  children or such, as oppossed to assemble, shutdown of different
2024
  devices doesn't require that the upper device was active.
2025

2026
  @type disk: L{objects.Disk}
2027
  @param disk: the description of the disk we should
2028
      shutdown
2029
  @rtype: None
2030

2031
  """
2032
  msgs = []
2033
  r_dev = _RecursiveFindBD(disk)
2034
  if r_dev is not None:
2035
    r_path = r_dev.dev_path
2036
    try:
2037
      r_dev.Shutdown()
2038
      DevCacheManager.RemoveCache(r_path)
2039
    except errors.BlockDeviceError, err:
2040
      msgs.append(str(err))
2041

    
2042
  if disk.children:
2043
    for child in disk.children:
2044
      try:
2045
        BlockdevShutdown(child)
2046
      except RPCFail, err:
2047
        msgs.append(str(err))
2048

    
2049
  if msgs:
2050
    _Fail("; ".join(msgs))
2051

    
2052

    
2053
def BlockdevAddchildren(parent_cdev, new_cdevs):
2054
  """Extend a mirrored block device.
2055

2056
  @type parent_cdev: L{objects.Disk}
2057
  @param parent_cdev: the disk to which we should add children
2058
  @type new_cdevs: list of L{objects.Disk}
2059
  @param new_cdevs: the list of children which we should add
2060
  @rtype: None
2061

2062
  """
2063
  parent_bdev = _RecursiveFindBD(parent_cdev)
2064
  if parent_bdev is None:
2065
    _Fail("Can't find parent device '%s' in add children", parent_cdev)
2066
  new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
2067
  if new_bdevs.count(None) > 0:
2068
    _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
2069
  parent_bdev.AddChildren(new_bdevs)
2070

    
2071

    
2072
def BlockdevRemovechildren(parent_cdev, new_cdevs):
2073
  """Shrink a mirrored block device.
2074

2075
  @type parent_cdev: L{objects.Disk}
2076
  @param parent_cdev: the disk from which we should remove children
2077
  @type new_cdevs: list of L{objects.Disk}
2078
  @param new_cdevs: the list of children which we should remove
2079
  @rtype: None
2080

2081
  """
2082
  parent_bdev = _RecursiveFindBD(parent_cdev)
2083
  if parent_bdev is None:
2084
    _Fail("Can't find parent device '%s' in remove children", parent_cdev)
2085
  devs = []
2086
  for disk in new_cdevs:
2087
    rpath = disk.StaticDevPath()
2088
    if rpath is None:
2089
      bd = _RecursiveFindBD(disk)
2090
      if bd is None:
2091
        _Fail("Can't find device %s while removing children", disk)
2092
      else:
2093
        devs.append(bd.dev_path)
2094
    else:
2095
      if not utils.IsNormAbsPath(rpath):
2096
        _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
2097
      devs.append(rpath)
2098
  parent_bdev.RemoveChildren(devs)
2099

    
2100

    
2101
def BlockdevGetmirrorstatus(disks):
2102
  """Get the mirroring status of a list of devices.
2103

2104
  @type disks: list of L{objects.Disk}
2105
  @param disks: the list of disks which we should query
2106
  @rtype: disk
2107
  @return: List of L{objects.BlockDevStatus}, one for each disk
2108
  @raise errors.BlockDeviceError: if any of the disks cannot be
2109
      found
2110

2111
  """
2112
  stats = []
2113
  for dsk in disks:
2114
    rbd = _RecursiveFindBD(dsk)
2115
    if rbd is None:
2116
      _Fail("Can't find device %s", dsk)
2117

    
2118
    stats.append(rbd.CombinedSyncStatus())
2119

    
2120
  return stats
2121

    
2122

    
2123
def BlockdevGetmirrorstatusMulti(disks):
2124
  """Get the mirroring status of a list of devices.
2125

2126
  @type disks: list of L{objects.Disk}
2127
  @param disks: the list of disks which we should query
2128
  @rtype: disk
2129
  @return: List of tuples, (bool, status), one for each disk; bool denotes
2130
    success/failure, status is L{objects.BlockDevStatus} on success, string
2131
    otherwise
2132

2133
  """
2134
  result = []
2135
  for disk in disks:
2136
    try:
2137
      rbd = _RecursiveFindBD(disk)
2138
      if rbd is None:
2139
        result.append((False, "Can't find device %s" % disk))
2140
        continue
2141

    
2142
      status = rbd.CombinedSyncStatus()
2143
    except errors.BlockDeviceError, err:
2144
      logging.exception("Error while getting disk status")
2145
      result.append((False, str(err)))
2146
    else:
2147
      result.append((True, status))
2148

    
2149
  assert len(disks) == len(result)
2150

    
2151
  return result
2152

    
2153

    
2154
def _RecursiveFindBD(disk):
2155
  """Check if a device is activated.
2156

2157
  If so, return information about the real device.
2158

2159
  @type disk: L{objects.Disk}
2160
  @param disk: the disk object we need to find
2161

2162
  @return: None if the device can't be found,
2163
      otherwise the device instance
2164

2165
  """
2166
  children = []
2167
  if disk.children:
2168
    for chdisk in disk.children:
2169
      children.append(_RecursiveFindBD(chdisk))
2170

    
2171
  return bdev.FindDevice(disk, children)
2172

    
2173

    
2174
def _OpenRealBD(disk):
2175
  """Opens the underlying block device of a disk.
2176

2177
  @type disk: L{objects.Disk}
2178
  @param disk: the disk object we want to open
2179

2180
  """
2181
  real_disk = _RecursiveFindBD(disk)
2182
  if real_disk is None:
2183
    _Fail("Block device '%s' is not set up", disk)
2184

    
2185
  real_disk.Open()
2186

    
2187
  return real_disk
2188

    
2189

    
2190
def BlockdevFind(disk):
2191
  """Check if a device is activated.
2192

2193
  If it is, return information about the real device.
2194

2195
  @type disk: L{objects.Disk}
2196
  @param disk: the disk to find
2197
  @rtype: None or objects.BlockDevStatus
2198
  @return: None if the disk cannot be found, otherwise a the current
2199
           information
2200

2201
  """
2202
  try:
2203
    rbd = _RecursiveFindBD(disk)
2204
  except errors.BlockDeviceError, err:
2205
    _Fail("Failed to find device: %s", err, exc=True)
2206

    
2207
  if rbd is None:
2208
    return None
2209

    
2210
  return rbd.GetSyncStatus()
2211

    
2212

    
2213
def BlockdevGetdimensions(disks):
2214
  """Computes the size of the given disks.
2215

2216
  If a disk is not found, returns None instead.
2217

2218
  @type disks: list of L{objects.Disk}
2219
  @param disks: the list of disk to compute the size for
2220
  @rtype: list
2221
  @return: list with elements None if the disk cannot be found,
2222
      otherwise the pair (size, spindles), where spindles is None if the
2223
      device doesn't support that
2224

2225
  """
2226
  result = []
2227
  for cf in disks:
2228
    try:
2229
      rbd = _RecursiveFindBD(cf)
2230
    except errors.BlockDeviceError:
2231
      result.append(None)
2232
      continue
2233
    if rbd is None:
2234
      result.append(None)
2235
    else:
2236
      result.append(rbd.GetActualDimensions())
2237
  return result
2238

    
2239

    
2240
def BlockdevExport(disk, dest_node, dest_path, cluster_name):
2241
  """Export a block device to a remote node.
2242

2243
  @type disk: L{objects.Disk}
2244
  @param disk: the description of the disk to export
2245
  @type dest_node: str
2246
  @param dest_node: the destination node to export to
2247
  @type dest_path: str
2248
  @param dest_path: the destination path on the target node
2249
  @type cluster_name: str
2250
  @param cluster_name: the cluster name, needed for SSH hostalias
2251
  @rtype: None
2252

2253
  """
2254
  real_disk = _OpenRealBD(disk)
2255

    
2256
  # the block size on the read dd is 1MiB to match our units
2257
  expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
2258
                               "dd if=%s bs=1048576 count=%s",
2259
                               real_disk.dev_path, str(disk.size))
2260

    
2261
  # we set here a smaller block size as, due to ssh buffering, more
2262
  # than 64-128k will mostly ignored; we use nocreat to fail if the
2263
  # device is not already there or we pass a wrong path; we use
2264
  # notrunc to no attempt truncate on an LV device; we use oflag=dsync
2265
  # to not buffer too much memory; this means that at best, we flush
2266
  # every 64k, which will not be very fast
2267
  destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
2268
                                " oflag=dsync", dest_path)
2269

    
2270
  remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
2271
                                                   constants.SSH_LOGIN_USER,
2272
                                                   destcmd)
2273

    
2274
  # all commands have been checked, so we're safe to combine them
2275
  command = "|".join([expcmd, utils.ShellQuoteArgs(remotecmd)])
2276

    
2277
  result = utils.RunCmd(["bash", "-c", command])
2278

    
2279
  if result.failed:
2280
    _Fail("Disk copy command '%s' returned error: %s"
2281
          " output: %s", command, result.fail_reason, result.output)
2282

    
2283

    
2284
def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
2285
  """Write a file to the filesystem.
2286

2287
  This allows the master to overwrite(!) a file. It will only perform
2288
  the operation if the file belongs to a list of configuration files.
2289

2290
  @type file_name: str
2291
  @param file_name: the target file name
2292
  @type data: str
2293
  @param data: the new contents of the file
2294
  @type mode: int
2295
  @param mode: the mode to give the file (can be None)
2296
  @type uid: string
2297
  @param uid: the owner of the file
2298
  @type gid: string
2299
  @param gid: the group of the file
2300
  @type atime: float
2301
  @param atime: the atime to set on the file (can be None)
2302
  @type mtime: float
2303
  @param mtime: the mtime to set on the file (can be None)
2304
  @rtype: None
2305

2306
  """
2307
  file_name = vcluster.LocalizeVirtualPath(file_name)
2308

    
2309
  if not os.path.isabs(file_name):
2310
    _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
2311

    
2312
  if file_name not in _ALLOWED_UPLOAD_FILES:
2313
    _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
2314
          file_name)
2315

    
2316
  raw_data = _Decompress(data)
2317

    
2318
  if not (isinstance(uid, basestring) and isinstance(gid, basestring)):
2319
    _Fail("Invalid username/groupname type")
2320

    
2321
  getents = runtime.GetEnts()
2322
  uid = getents.LookupUser(uid)
2323
  gid = getents.LookupGroup(gid)
2324

    
2325
  utils.SafeWriteFile(file_name, None,
2326
                      data=raw_data, mode=mode, uid=uid, gid=gid,
2327
                      atime=atime, mtime=mtime)
2328

    
2329

    
2330
def RunOob(oob_program, command, node, timeout):
2331
  """Executes oob_program with given command on given node.
2332

2333
  @param oob_program: The path to the executable oob_program
2334
  @param command: The command to invoke on oob_program
2335
  @param node: The node given as an argument to the program
2336
  @param timeout: Timeout after which we kill the oob program
2337

2338
  @return: stdout
2339
  @raise RPCFail: If execution fails for some reason
2340

2341
  """
2342
  result = utils.RunCmd([oob_program, command, node], timeout=timeout)
2343

    
2344
  if result.failed:
2345
    _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
2346
          result.fail_reason, result.output)
2347

    
2348
  return result.stdout
2349

    
2350

    
2351
def _OSOndiskAPIVersion(os_dir):
2352
  """Compute and return the API version of a given OS.
2353

2354
  This function will try to read the API version of the OS residing in
2355
  the 'os_dir' directory.
2356

2357
  @type os_dir: str
2358
  @param os_dir: the directory in which we should look for the OS
2359
  @rtype: tuple
2360
  @return: tuple (status, data) with status denoting the validity and
2361
      data holding either the vaid versions or an error message
2362

2363
  """
2364
  api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
2365

    
2366
  try:
2367
    st = os.stat(api_file)
2368
  except EnvironmentError, err:
2369
    return False, ("Required file '%s' not found under path %s: %s" %
2370
                   (constants.OS_API_FILE, os_dir, utils.ErrnoOrStr(err)))
2371

    
2372
  if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2373
    return False, ("File '%s' in %s is not a regular file" %
2374
                   (constants.OS_API_FILE, os_dir))
2375

    
2376
  try:
2377
    api_versions = utils.ReadFile(api_file).splitlines()
2378
  except EnvironmentError, err:
2379
    return False, ("Error while reading the API version file at %s: %s" %
2380
                   (api_file, utils.ErrnoOrStr(err)))
2381

    
2382
  try:
2383
    api_versions = [int(version.strip()) for version in api_versions]
2384
  except (TypeError, ValueError), err:
2385
    return False, ("API version(s) can't be converted to integer: %s" %
2386
                   str(err))
2387

    
2388
  return True, api_versions
2389

    
2390

    
2391
def DiagnoseOS(top_dirs=None):
2392
  """Compute the validity for all OSes.
2393

2394
  @type top_dirs: list
2395
  @param top_dirs: the list of directories in which to
2396
      search (if not given defaults to
2397
      L{pathutils.OS_SEARCH_PATH})
2398
  @rtype: list of L{objects.OS}
2399
  @return: a list of tuples (name, path, status, diagnose, variants,
2400
      parameters, api_version) for all (potential) OSes under all
2401
      search paths, where:
2402
          - name is the (potential) OS name
2403
          - path is the full path to the OS
2404
          - status True/False is the validity of the OS
2405
          - diagnose is the error message for an invalid OS, otherwise empty
2406
          - variants is a list of supported OS variants, if any
2407
          - parameters is a list of (name, help) parameters, if any
2408
          - api_version is a list of support OS API versions
2409

2410
  """
2411
  if top_dirs is None:
2412
    top_dirs = pathutils.OS_SEARCH_PATH
2413

    
2414
  result = []
2415
  for dir_name in top_dirs:
2416
    if os.path.isdir(dir_name):
2417
      try:
2418
        f_names = utils.ListVisibleFiles(dir_name)
2419
      except EnvironmentError, err:
2420
        logging.exception("Can't list the OS directory %s: %s", dir_name, err)
2421
        break
2422
      for name in f_names:
2423
        os_path = utils.PathJoin(dir_name, name)
2424
        status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
2425
        if status:
2426
          diagnose = ""
2427
          variants = os_inst.supported_variants
2428
          parameters = os_inst.supported_parameters
2429
          api_versions = os_inst.api_versions
2430
        else:
2431
          diagnose = os_inst
2432
          variants = parameters = api_versions = []
2433
        result.append((name, os_path, status, diagnose, variants,
2434
                       parameters, api_versions))
2435

    
2436
  return result
2437

    
2438

    
2439
def _TryOSFromDisk(name, base_dir=None):
2440
  """Create an OS instance from disk.
2441

2442
  This function will return an OS instance if the given name is a
2443
  valid OS name.
2444

2445
  @type base_dir: string
2446
  @keyword base_dir: Base directory containing OS installations.
2447
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2448
  @rtype: tuple
2449
  @return: success and either the OS instance if we find a valid one,
2450
      or error message
2451

2452
  """
2453
  if base_dir is None:
2454
    os_dir = utils.FindFile(name, pathutils.OS_SEARCH_PATH, os.path.isdir)
2455
  else:
2456
    os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2457

    
2458
  if os_dir is None:
2459
    return False, "Directory for OS %s not found in search path" % name
2460

    
2461
  status, api_versions = _OSOndiskAPIVersion(os_dir)
2462
  if not status:
2463
    # push the error up
2464
    return status, api_versions
2465

    
2466
  if not constants.OS_API_VERSIONS.intersection(api_versions):
2467
    return False, ("API version mismatch for path '%s': found %s, want %s." %
2468
                   (os_dir, api_versions, constants.OS_API_VERSIONS))
2469

    
2470
  # OS Files dictionary, we will populate it with the absolute path
2471
  # names; if the value is True, then it is a required file, otherwise
2472
  # an optional one
2473
  os_files = dict.fromkeys(constants.OS_SCRIPTS, True)
2474

    
2475
  if max(api_versions) >= constants.OS_API_V15:
2476
    os_files[constants.OS_VARIANTS_FILE] = False
2477

    
2478
  if max(api_versions) >= constants.OS_API_V20:
2479
    os_files[constants.OS_PARAMETERS_FILE] = True
2480
  else:
2481
    del os_files[constants.OS_SCRIPT_VERIFY]
2482

    
2483
  for (filename, required) in os_files.items():
2484
    os_files[filename] = utils.PathJoin(os_dir, filename)
2485

    
2486
    try:
2487
      st = os.stat(os_files[filename])
2488
    except EnvironmentError, err:
2489
      if err.errno == errno.ENOENT and not required:
2490
        del os_files[filename]
2491
        continue
2492
      return False, ("File '%s' under path '%s' is missing (%s)" %
2493
                     (filename, os_dir, utils.ErrnoOrStr(err)))
2494

    
2495
    if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2496
      return False, ("File '%s' under path '%s' is not a regular file" %
2497
                     (filename, os_dir))
2498

    
2499
    if filename in constants.OS_SCRIPTS:
2500
      if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2501
        return False, ("File '%s' under path '%s' is not executable" %
2502
                       (filename, os_dir))
2503

    
2504
  variants = []
2505
  if constants.OS_VARIANTS_FILE in os_files:
2506
    variants_file = os_files[constants.OS_VARIANTS_FILE]
2507
    try:
2508
      variants = \
2509
        utils.FilterEmptyLinesAndComments(utils.ReadFile(variants_file))
2510
    except EnvironmentError, err:
2511
      # we accept missing files, but not other errors
2512
      if err.errno != errno.ENOENT:
2513
        return False, ("Error while reading the OS variants file at %s: %s" %
2514
                       (variants_file, utils.ErrnoOrStr(err)))
2515

    
2516
  parameters = []
2517
  if constants.OS_PARAMETERS_FILE in os_files:
2518
    parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2519
    try:
2520
      parameters = utils.ReadFile(parameters_file).splitlines()
2521
    except EnvironmentError, err:
2522
      return False, ("Error while reading the OS parameters file at %s: %s" %
2523
                     (parameters_file, utils.ErrnoOrStr(err)))
2524
    parameters = [v.split(None, 1) for v in parameters]
2525

    
2526
  os_obj = objects.OS(name=name, path=os_dir,
2527
                      create_script=os_files[constants.OS_SCRIPT_CREATE],
2528
                      export_script=os_files[constants.OS_SCRIPT_EXPORT],
2529
                      import_script=os_files[constants.OS_SCRIPT_IMPORT],
2530
                      rename_script=os_files[constants.OS_SCRIPT_RENAME],
2531
                      verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2532
                                                 None),
2533
                      supported_variants=variants,
2534
                      supported_parameters=parameters,
2535
                      api_versions=api_versions)
2536
  return True, os_obj
2537

    
2538

    
2539
def OSFromDisk(name, base_dir=None):
2540
  """Create an OS instance from disk.
2541

2542
  This function will return an OS instance if the given name is a
2543
  valid OS name. Otherwise, it will raise an appropriate
2544
  L{RPCFail} exception, detailing why this is not a valid OS.
2545

2546
  This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2547
  an exception but returns true/false status data.
2548

2549
  @type base_dir: string
2550
  @keyword base_dir: Base directory containing OS installations.
2551
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2552
  @rtype: L{objects.OS}
2553
  @return: the OS instance if we find a valid one
2554
  @raise RPCFail: if we don't find a valid OS
2555

2556
  """
2557
  name_only = objects.OS.GetName(name)
2558
  status, payload = _TryOSFromDisk(name_only, base_dir)
2559

    
2560
  if not status:
2561
    _Fail(payload)
2562

    
2563
  return payload
2564

    
2565

    
2566
def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2567
  """Calculate the basic environment for an os script.
2568

2569
  @type os_name: str
2570
  @param os_name: full operating system name (including variant)
2571
  @type inst_os: L{objects.OS}
2572
  @param inst_os: operating system for which the environment is being built
2573
  @type os_params: dict
2574
  @param os_params: the OS parameters
2575
  @type debug: integer
2576
  @param debug: debug level (0 or 1, for OS Api 10)
2577
  @rtype: dict
2578
  @return: dict of environment variables
2579
  @raise errors.BlockDeviceError: if the block device
2580
      cannot be found
2581

2582
  """
2583
  result = {}
2584
  api_version = \
2585
    max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2586
  result["OS_API_VERSION"] = "%d" % api_version
2587
  result["OS_NAME"] = inst_os.name
2588
  result["DEBUG_LEVEL"] = "%d" % debug
2589

    
2590
  # OS variants
2591
  if api_version >= constants.OS_API_V15 and inst_os.supported_variants:
2592
    variant = objects.OS.GetVariant(os_name)
2593
    if not variant:
2594
      variant = inst_os.supported_variants[0]
2595
  else:
2596
    variant = ""
2597
  result["OS_VARIANT"] = variant
2598

    
2599
  # OS params
2600
  for pname, pvalue in os_params.items():
2601
    result["OSP_%s" % pname.upper()] = pvalue
2602

    
2603
  # Set a default path otherwise programs called by OS scripts (or
2604
  # even hooks called from OS scripts) might break, and we don't want
2605
  # to have each script require setting a PATH variable
2606
  result["PATH"] = constants.HOOKS_PATH
2607

    
2608
  return result
2609

    
2610

    
2611
def OSEnvironment(instance, inst_os, debug=0):
2612
  """Calculate the environment for an os script.
2613

2614
  @type instance: L{objects.Instance}
2615
  @param instance: target instance for the os script run
2616
  @type inst_os: L{objects.OS}
2617
  @param inst_os: operating system for which the environment is being built
2618
  @type debug: integer
2619
  @param debug: debug level (0 or 1, for OS Api 10)
2620
  @rtype: dict
2621
  @return: dict of environment variables
2622
  @raise errors.BlockDeviceError: if the block device
2623
      cannot be found
2624

2625
  """
2626
  result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2627

    
2628
  for attr in ["name", "os", "uuid", "ctime", "mtime", "primary_node"]:
2629
    result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2630

    
2631
  result["HYPERVISOR"] = instance.hypervisor
2632
  result["DISK_COUNT"] = "%d" % len(instance.disks)
2633
  result["NIC_COUNT"] = "%d" % len(instance.nics)
2634
  result["INSTANCE_SECONDARY_NODES"] = \
2635
      ("%s" % " ".join(instance.secondary_nodes))
2636

    
2637
  # Disks
2638
  for idx, disk in enumerate(instance.disks):
2639
    real_disk = _OpenRealBD(disk)
2640
    result["DISK_%d_PATH" % idx] = real_disk.dev_path
2641
    result["DISK_%d_ACCESS" % idx] = disk.mode
2642
    if constants.HV_DISK_TYPE in instance.hvparams:
2643
      result["DISK_%d_FRONTEND_TYPE" % idx] = \
2644
        instance.hvparams[constants.HV_DISK_TYPE]
2645
    if disk.dev_type in constants.LDS_BLOCK:
2646
      result["DISK_%d_BACKEND_TYPE" % idx] = "block"
2647
    elif disk.dev_type == constants.LD_FILE:
2648
      result["DISK_%d_BACKEND_TYPE" % idx] = \
2649
        "file:%s" % disk.physical_id[0]
2650

    
2651
  # NICs
2652
  for idx, nic in enumerate(instance.nics):
2653
    result["NIC_%d_MAC" % idx] = nic.mac
2654
    if nic.ip:
2655
      result["NIC_%d_IP" % idx] = nic.ip
2656
    result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE]
2657
    if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2658
      result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK]
2659
    if nic.nicparams[constants.NIC_LINK]:
2660
      result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK]
2661
    if nic.netinfo:
2662
      nobj = objects.Network.FromDict(nic.netinfo)
2663
      result.update(nobj.HooksDict("NIC_%d_" % idx))
2664
    if constants.HV_NIC_TYPE in instance.hvparams:
2665
      result["NIC_%d_FRONTEND_TYPE" % idx] = \
2666
        instance.hvparams[constants.HV_NIC_TYPE]
2667

    
2668
  # HV/BE params
2669
  for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2670
    for key, value in source.items():
2671
      result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2672

    
2673
  return result
2674

    
2675

    
2676
def DiagnoseExtStorage(top_dirs=None):
2677
  """Compute the validity for all ExtStorage Providers.
2678

2679
  @type top_dirs: list
2680
  @param top_dirs: the list of directories in which to
2681
      search (if not given defaults to
2682
      L{pathutils.ES_SEARCH_PATH})
2683
  @rtype: list of L{objects.ExtStorage}
2684
  @return: a list of tuples (name, path, status, diagnose, parameters)
2685
      for all (potential) ExtStorage Providers under all
2686
      search paths, where:
2687
          - name is the (potential) ExtStorage Provider
2688
          - path is the full path to the ExtStorage Provider
2689
          - status True/False is the validity of the ExtStorage Provider
2690
          - diagnose is the error message for an invalid ExtStorage Provider,
2691
            otherwise empty
2692
          - parameters is a list of (name, help) parameters, if any
2693

2694
  """
2695
  if top_dirs is None:
2696
    top_dirs = pathutils.ES_SEARCH_PATH
2697

    
2698
  result = []
2699
  for dir_name in top_dirs:
2700
    if os.path.isdir(dir_name):
2701
      try:
2702
        f_names = utils.ListVisibleFiles(dir_name)
2703
      except EnvironmentError, err:
2704
        logging.exception("Can't list the ExtStorage directory %s: %s",
2705
                          dir_name, err)
2706
        break
2707
      for name in f_names:
2708
        es_path = utils.PathJoin(dir_name, name)
2709
        status, es_inst = bdev.ExtStorageFromDisk(name, base_dir=dir_name)
2710
        if status:
2711
          diagnose = ""
2712
          parameters = es_inst.supported_parameters
2713
        else:
2714
          diagnose = es_inst
2715
          parameters = []
2716
        result.append((name, es_path, status, diagnose, parameters))
2717

    
2718
  return result
2719

    
2720

    
2721
def BlockdevGrow(disk, amount, dryrun, backingstore):
2722
  """Grow a stack of block devices.
2723

2724
  This function is called recursively, with the childrens being the
2725
  first ones to resize.
2726

2727
  @type disk: L{objects.Disk}
2728
  @param disk: the disk to be grown
2729
  @type amount: integer
2730
  @param amount: the amount (in mebibytes) to grow with
2731
  @type dryrun: boolean
2732
  @param dryrun: whether to execute the operation in simulation mode
2733
      only, without actually increasing the size
2734
  @param backingstore: whether to execute the operation on backing storage
2735
      only, or on "logical" storage only; e.g. DRBD is logical storage,
2736
      whereas LVM, file, RBD are backing storage
2737
  @rtype: (status, result)
2738
  @return: a tuple with the status of the operation (True/False), and
2739
      the errors message if status is False
2740

2741
  """
2742
  r_dev = _RecursiveFindBD(disk)
2743
  if r_dev is None:
2744
    _Fail("Cannot find block device %s", disk)
2745

    
2746
  try:
2747
    r_dev.Grow(amount, dryrun, backingstore)
2748
  except errors.BlockDeviceError, err:
2749
    _Fail("Failed to grow block device: %s", err, exc=True)
2750

    
2751

    
2752
def BlockdevSnapshot(disk):
2753
  """Create a snapshot copy of a block device.
2754

2755
  This function is called recursively, and the snapshot is actually created
2756
  just for the leaf lvm backend device.
2757

2758
  @type disk: L{objects.Disk}
2759
  @param disk: the disk to be snapshotted
2760
  @rtype: string
2761
  @return: snapshot disk ID as (vg, lv)
2762

2763
  """
2764
  if disk.dev_type == constants.LD_DRBD8:
2765
    if not disk.children:
2766
      _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2767
            disk.unique_id)
2768
    return BlockdevSnapshot(disk.children[0])
2769
  elif disk.dev_type == constants.LD_LV:
2770
    r_dev = _RecursiveFindBD(disk)
2771
    if r_dev is not None:
2772
      # FIXME: choose a saner value for the snapshot size
2773
      # let's stay on the safe side and ask for the full size, for now
2774
      return r_dev.Snapshot(disk.size)
2775
    else:
2776
      _Fail("Cannot find block device %s", disk)
2777
  else:
2778
    _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2779
          disk.unique_id, disk.dev_type)
2780

    
2781

    
2782
def BlockdevSetInfo(disk, info):
2783
  """Sets 'metadata' information on block devices.
2784

2785
  This function sets 'info' metadata on block devices. Initial
2786
  information is set at device creation; this function should be used
2787
  for example after renames.
2788

2789
  @type disk: L{objects.Disk}
2790
  @param disk: the disk to be grown
2791
  @type info: string
2792
  @param info: new 'info' metadata
2793
  @rtype: (status, result)
2794
  @return: a tuple with the status of the operation (True/False), and
2795
      the errors message if status is False
2796

2797
  """
2798
  r_dev = _RecursiveFindBD(disk)
2799
  if r_dev is None:
2800
    _Fail("Cannot find block device %s", disk)
2801

    
2802
  try:
2803
    r_dev.SetInfo(info)
2804
  except errors.BlockDeviceError, err:
2805
    _Fail("Failed to set information on block device: %s", err, exc=True)
2806

    
2807

    
2808
def FinalizeExport(instance, snap_disks):
2809
  """Write out the export configuration information.
2810

2811
  @type instance: L{objects.Instance}
2812
  @param instance: the instance which we export, used for
2813
      saving configuration
2814
  @type snap_disks: list of L{objects.Disk}
2815
  @param snap_disks: list of snapshot block devices, which
2816
      will be used to get the actual name of the dump file
2817

2818
  @rtype: None
2819

2820
  """
2821
  destdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name + ".new")
2822
  finaldestdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name)
2823

    
2824
  config = objects.SerializableConfigParser()
2825

    
2826
  config.add_section(constants.INISECT_EXP)
2827
  config.set(constants.INISECT_EXP, "version", "0")
2828
  config.set(constants.INISECT_EXP, "timestamp", "%d" % int(time.time()))
2829
  config.set(constants.INISECT_EXP, "source", instance.primary_node)
2830
  config.set(constants.INISECT_EXP, "os", instance.os)
2831
  config.set(constants.INISECT_EXP, "compression", "none")
2832

    
2833
  config.add_section(constants.INISECT_INS)
2834
  config.set(constants.INISECT_INS, "name", instance.name)
2835
  config.set(constants.INISECT_INS, "maxmem", "%d" %
2836
             instance.beparams[constants.BE_MAXMEM])
2837
  config.set(constants.INISECT_INS, "minmem", "%d" %
2838
             instance.beparams[constants.BE_MINMEM])
2839
  # "memory" is deprecated, but useful for exporting to old ganeti versions
2840
  config.set(constants.INISECT_INS, "memory", "%d" %
2841
             instance.beparams[constants.BE_MAXMEM])
2842
  config.set(constants.INISECT_INS, "vcpus", "%d" %
2843
             instance.beparams[constants.BE_VCPUS])
2844
  config.set(constants.INISECT_INS, "disk_template", instance.disk_template)
2845
  config.set(constants.INISECT_INS, "hypervisor", instance.hypervisor)
2846
  config.set(constants.INISECT_INS, "tags", " ".join(instance.GetTags()))
2847

    
2848
  nic_total = 0
2849
  for nic_count, nic in enumerate(instance.nics):
2850
    nic_total += 1
2851
    config.set(constants.INISECT_INS, "nic%d_mac" %
2852
               nic_count, "%s" % nic.mac)
2853
    config.set(constants.INISECT_INS, "nic%d_ip" % nic_count, "%s" % nic.ip)
2854
    config.set(constants.INISECT_INS, "nic%d_network" % nic_count,
2855
               "%s" % nic.network)
2856
    for param in constants.NICS_PARAMETER_TYPES:
2857
      config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param),
2858
                 "%s" % nic.nicparams.get(param, None))
2859
  # TODO: redundant: on load can read nics until it doesn't exist
2860
  config.set(constants.INISECT_INS, "nic_count", "%d" % nic_total)
2861

    
2862
  disk_total = 0
2863
  for disk_count, disk in enumerate(snap_disks):
2864
    if disk:
2865
      disk_total += 1
2866
      config.set(constants.INISECT_INS, "disk%d_ivname" % disk_count,
2867
                 ("%s" % disk.iv_name))
2868
      config.set(constants.INISECT_INS, "disk%d_dump" % disk_count,
2869
                 ("%s" % disk.physical_id[1]))
2870
      config.set(constants.INISECT_INS, "disk%d_size" % disk_count,
2871
                 ("%d" % disk.size))
2872

    
2873
  config.set(constants.INISECT_INS, "disk_count", "%d" % disk_total)
2874

    
2875
  # New-style hypervisor/backend parameters
2876

    
2877
  config.add_section(constants.INISECT_HYP)
2878
  for name, value in instance.hvparams.items():
2879
    if name not in constants.HVC_GLOBALS:
2880
      config.set(constants.INISECT_HYP, name, str(value))
2881

    
2882
  config.add_section(constants.INISECT_BEP)
2883
  for name, value in instance.beparams.items():
2884
    config.set(constants.INISECT_BEP, name, str(value))
2885

    
2886
  config.add_section(constants.INISECT_OSP)
2887
  for name, value in instance.osparams.items():
2888
    config.set(constants.INISECT_OSP, name, str(value))
2889

    
2890
  utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
2891
                  data=config.Dumps())
2892
  shutil.rmtree(finaldestdir, ignore_errors=True)
2893
  shutil.move(destdir, finaldestdir)
2894

    
2895

    
2896
def ExportInfo(dest):
2897
  """Get export configuration information.
2898

2899
  @type dest: str
2900
  @param dest: directory containing the export
2901

2902
  @rtype: L{objects.SerializableConfigParser}
2903
  @return: a serializable config file containing the
2904
      export info
2905

2906
  """
2907
  cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
2908

    
2909
  config = objects.SerializableConfigParser()
2910
  config.read(cff)
2911

    
2912
  if (not config.has_section(constants.INISECT_EXP) or
2913
      not config.has_section(constants.INISECT_INS)):
2914
    _Fail("Export info file doesn't have the required fields")
2915

    
2916
  return config.Dumps()
2917

    
2918

    
2919
def ListExports():
2920
  """Return a list of exports currently available on this machine.
2921

2922
  @rtype: list
2923
  @return: list of the exports
2924

2925
  """
2926
  if os.path.isdir(pathutils.EXPORT_DIR):
2927
    return sorted(utils.ListVisibleFiles(pathutils.EXPORT_DIR))
2928
  else:
2929
    _Fail("No exports directory")
2930

    
2931

    
2932
def RemoveExport(export):
2933
  """Remove an existing export from the node.
2934

2935
  @type export: str
2936
  @param export: the name of the export to remove
2937
  @rtype: None
2938

2939
  """
2940
  target = utils.PathJoin(pathutils.EXPORT_DIR, export)
2941

    
2942
  try:
2943
    shutil.rmtree(target)
2944
  except EnvironmentError, err:
2945
    _Fail("Error while removing the export: %s", err, exc=True)
2946

    
2947

    
2948
def BlockdevRename(devlist):
2949
  """Rename a list of block devices.
2950

2951
  @type devlist: list of tuples
2952
  @param devlist: list of tuples of the form  (disk,
2953
      new_logical_id, new_physical_id); disk is an
2954
      L{objects.Disk} object describing the current disk,
2955
      and new logical_id/physical_id is the name we
2956
      rename it to
2957
  @rtype: boolean
2958
  @return: True if all renames succeeded, False otherwise
2959

2960
  """
2961
  msgs = []
2962
  result = True
2963
  for disk, unique_id in devlist:
2964
    dev = _RecursiveFindBD(disk)
2965
    if dev is None:
2966
      msgs.append("Can't find device %s in rename" % str(disk))
2967
      result = False
2968
      continue
2969
    try:
2970
      old_rpath = dev.dev_path
2971
      dev.Rename(unique_id)
2972
      new_rpath = dev.dev_path
2973
      if old_rpath != new_rpath:
2974
        DevCacheManager.RemoveCache(old_rpath)
2975
        # FIXME: we should add the new cache information here, like:
2976
        # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2977
        # but we don't have the owner here - maybe parse from existing
2978
        # cache? for now, we only lose lvm data when we rename, which
2979
        # is less critical than DRBD or MD
2980
    except errors.BlockDeviceError, err:
2981
      msgs.append("Can't rename device '%s' to '%s': %s" %
2982
                  (dev, unique_id, err))
2983
      logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2984
      result = False
2985
  if not result:
2986
    _Fail("; ".join(msgs))
2987

    
2988

    
2989
def _TransformFileStorageDir(fs_dir):
2990
  """Checks whether given file_storage_dir is valid.
2991

2992
  Checks wheter the given fs_dir is within the cluster-wide default
2993
  file_storage_dir or the shared_file_storage_dir, which are stored in
2994
  SimpleStore. Only paths under those directories are allowed.
2995

2996
  @type fs_dir: str
2997
  @param fs_dir: the path to check
2998

2999
  @return: the normalized path if valid, None otherwise
3000

3001
  """
3002
  if not (constants.ENABLE_FILE_STORAGE or
3003
          constants.ENABLE_SHARED_FILE_STORAGE):
3004
    _Fail("File storage disabled at configure time")
3005

    
3006
  bdev.CheckFileStoragePath(fs_dir)
3007

    
3008
  return os.path.normpath(fs_dir)
3009

    
3010

    
3011
def CreateFileStorageDir(file_storage_dir):
3012
  """Create file storage directory.
3013

3014
  @type file_storage_dir: str
3015
  @param file_storage_dir: directory to create
3016

3017
  @rtype: tuple
3018
  @return: tuple with first element a boolean indicating wheter dir
3019
      creation was successful or not
3020

3021
  """
3022
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3023
  if os.path.exists(file_storage_dir):
3024
    if not os.path.isdir(file_storage_dir):
3025
      _Fail("Specified storage dir '%s' is not a directory",
3026
            file_storage_dir)
3027
  else:
3028
    try:
3029
      os.makedirs(file_storage_dir, 0750)
3030
    except OSError, err:
3031
      _Fail("Cannot create file storage directory '%s': %s",
3032
            file_storage_dir, err, exc=True)
3033

    
3034

    
3035
def RemoveFileStorageDir(file_storage_dir):
3036
  """Remove file storage directory.
3037

3038
  Remove it only if it's empty. If not log an error and return.
3039

3040
  @type file_storage_dir: str
3041
  @param file_storage_dir: the directory we should cleanup
3042
  @rtype: tuple (success,)
3043
  @return: tuple of one element, C{success}, denoting
3044
      whether the operation was successful
3045

3046
  """
3047
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
3048
  if os.path.exists(file_storage_dir):
3049
    if not os.path.isdir(file_storage_dir):
3050
      _Fail("Specified Storage directory '%s' is not a directory",
3051
            file_storage_dir)
3052
    # deletes dir only if empty, otherwise we want to fail the rpc call
3053
    try:
3054
      os.rmdir(file_storage_dir)
3055
    except OSError, err:
3056
      _Fail("Cannot remove file storage directory '%s': %s",
3057
            file_storage_dir, err)
3058

    
3059

    
3060
def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
3061
  """Rename the file storage directory.
3062

3063
  @type old_file_storage_dir: str
3064
  @param old_file_storage_dir: the current path
3065
  @type new_file_storage_dir: str
3066
  @param new_file_storage_dir: the name we should rename to
3067
  @rtype: tuple (success,)
3068
  @return: tuple of one element, C{success}, denoting
3069
      whether the operation was successful
3070

3071
  """
3072
  old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
3073
  new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
3074
  if not os.path.exists(new_file_storage_dir):
3075
    if os.path.isdir(old_file_storage_dir):
3076
      try:
3077
        os.rename(old_file_storage_dir, new_file_storage_dir)
3078
      except OSError, err:
3079
        _Fail("Cannot rename '%s' to '%s': %s",
3080
              old_file_storage_dir, new_file_storage_dir, err)
3081
    else:
3082
      _Fail("Specified storage dir '%s' is not a directory",
3083
            old_file_storage_dir)
3084
  else:
3085
    if os.path.exists(old_file_storage_dir):
3086
      _Fail("Cannot rename '%s' to '%s': both locations exist",
3087
            old_file_storage_dir, new_file_storage_dir)
3088

    
3089

    
3090
def _EnsureJobQueueFile(file_name):
3091
  """Checks whether the given filename is in the queue directory.
3092

3093
  @type file_name: str
3094
  @param file_name: the file name we should check
3095
  @rtype: None
3096
  @raises RPCFail: if the file is not valid
3097

3098
  """
3099
  if not utils.IsBelowDir(pathutils.QUEUE_DIR, file_name):
3100
    _Fail("Passed job queue file '%s' does not belong to"
3101
          " the queue directory '%s'", file_name, pathutils.QUEUE_DIR)
3102

    
3103

    
3104
def JobQueueUpdate(file_name, content):
3105
  """Updates a file in the queue directory.
3106

3107
  This is just a wrapper over L{utils.io.WriteFile}, with proper
3108
  checking.
3109

3110
  @type file_name: str
3111
  @param file_name: the job file name
3112
  @type content: str
3113
  @param content: the new job contents
3114
  @rtype: boolean
3115
  @return: the success of the operation
3116

3117
  """
3118
  file_name = vcluster.LocalizeVirtualPath(file_name)
3119

    
3120
  _EnsureJobQueueFile(file_name)
3121
  getents = runtime.GetEnts()
3122

    
3123
  # Write and replace the file atomically
3124
  utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
3125
                  gid=getents.daemons_gid, mode=constants.JOB_QUEUE_FILES_PERMS)
3126

    
3127

    
3128
def JobQueueRename(old, new):
3129
  """Renames a job queue file.
3130

3131
  This is just a wrapper over os.rename with proper checking.
3132

3133
  @type old: str
3134
  @param old: the old (actual) file name
3135
  @type new: str
3136
  @param new: the desired file name
3137
  @rtype: tuple
3138
  @return: the success of the operation and payload
3139

3140
  """
3141
  old = vcluster.LocalizeVirtualPath(old)
3142
  new = vcluster.LocalizeVirtualPath(new)
3143

    
3144
  _EnsureJobQueueFile(old)
3145
  _EnsureJobQueueFile(new)
3146

    
3147
  getents = runtime.GetEnts()
3148

    
3149
  utils.RenameFile(old, new, mkdir=True, mkdir_mode=0750,
3150
                   dir_uid=getents.masterd_uid, dir_gid=getents.daemons_gid)
3151

    
3152

    
3153
def BlockdevClose(instance_name, disks):
3154
  """Closes the given block devices.
3155

3156
  This means they will be switched to secondary mode (in case of
3157
  DRBD).
3158

3159
  @param instance_name: if the argument is not empty, the symlinks
3160
      of this instance will be removed
3161
  @type disks: list of L{objects.Disk}
3162
  @param disks: the list of disks to be closed
3163
  @rtype: tuple (success, message)
3164
  @return: a tuple of success and message, where success
3165
      indicates the succes of the operation, and message
3166
      which will contain the error details in case we
3167
      failed
3168

3169
  """
3170
  bdevs = []
3171
  for cf in disks:
3172
    rd = _RecursiveFindBD(cf)
3173
    if rd is None:
3174
      _Fail("Can't find device %s", cf)
3175
    bdevs.append(rd)
3176

    
3177
  msg = []
3178
  for rd in bdevs:
3179
    try:
3180
      rd.Close()
3181
    except errors.BlockDeviceError, err:
3182
      msg.append(str(err))
3183
  if msg:
3184
    _Fail("Can't make devices secondary: %s", ",".join(msg))
3185
  else:
3186
    if instance_name:
3187
      _RemoveBlockDevLinks(instance_name, disks)
3188

    
3189

    
3190
def ValidateHVParams(hvname, hvparams):
3191
  """Validates the given hypervisor parameters.
3192

3193
  @type hvname: string
3194
  @param hvname: the hypervisor name
3195
  @type hvparams: dict
3196
  @param hvparams: the hypervisor parameters to be validated
3197
  @rtype: None
3198

3199
  """
3200
  try:
3201
    hv_type = hypervisor.GetHypervisor(hvname)
3202
    hv_type.ValidateParameters(hvparams)
3203
  except errors.HypervisorError, err:
3204
    _Fail(str(err), log=False)
3205

    
3206

    
3207
def _CheckOSPList(os_obj, parameters):
3208
  """Check whether a list of parameters is supported by the OS.
3209

3210
  @type os_obj: L{objects.OS}
3211
  @param os_obj: OS object to check
3212
  @type parameters: list
3213
  @param parameters: the list of parameters to check
3214

3215
  """
3216
  supported = [v[0] for v in os_obj.supported_parameters]
3217
  delta = frozenset(parameters).difference(supported)
3218
  if delta:
3219
    _Fail("The following parameters are not supported"
3220
          " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
3221

    
3222

    
3223
def ValidateOS(required, osname, checks, osparams):
3224
  """Validate the given OS' parameters.
3225

3226
  @type required: boolean
3227
  @param required: whether absence of the OS should translate into
3228
      failure or not
3229
  @type osname: string
3230
  @param osname: the OS to be validated
3231
  @type checks: list
3232
  @param checks: list of the checks to run (currently only 'parameters')
3233
  @type osparams: dict
3234
  @param osparams: dictionary with OS parameters
3235
  @rtype: boolean
3236
  @return: True if the validation passed, or False if the OS was not
3237
      found and L{required} was false
3238

3239
  """
3240
  if not constants.OS_VALIDATE_CALLS.issuperset(checks):
3241
    _Fail("Unknown checks required for OS %s: %s", osname,
3242
          set(checks).difference(constants.OS_VALIDATE_CALLS))
3243

    
3244
  name_only = objects.OS.GetName(osname)
3245
  status, tbv = _TryOSFromDisk(name_only, None)
3246

    
3247
  if not status:
3248
    if required:
3249
      _Fail(tbv)
3250
    else:
3251
      return False
3252

    
3253
  if max(tbv.api_versions) < constants.OS_API_V20:
3254
    return True
3255

    
3256
  if constants.OS_VALIDATE_PARAMETERS in checks:
3257
    _CheckOSPList(tbv, osparams.keys())
3258

    
3259
  validate_env = OSCoreEnv(osname, tbv, osparams)
3260
  result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
3261
                        cwd=tbv.path, reset_env=True)
3262
  if result.failed:
3263
    logging.error("os validate command '%s' returned error: %s output: %s",
3264
                  result.cmd, result.fail_reason, result.output)
3265
    _Fail("OS validation script failed (%s), output: %s",
3266
          result.fail_reason, result.output, log=False)
3267

    
3268
  return True
3269

    
3270

    
3271
def DemoteFromMC():
3272
  """Demotes the current node from master candidate role.
3273

3274
  """
3275
  # try to ensure we're not the master by mistake
3276
  master, myself = ssconf.GetMasterAndMyself()
3277
  if master == myself:
3278
    _Fail("ssconf status shows I'm the master node, will not demote")
3279

    
3280
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "check", constants.MASTERD])
3281
  if not result.failed:
3282
    _Fail("The master daemon is running, will not demote")
3283

    
3284
  try:
3285
    if os.path.isfile(pathutils.CLUSTER_CONF_FILE):
3286
      utils.CreateBackup(pathutils.CLUSTER_CONF_FILE)
3287
  except EnvironmentError, err:
3288
    if err.errno != errno.ENOENT:
3289
      _Fail("Error while backing up cluster file: %s", err, exc=True)
3290

    
3291
  utils.RemoveFile(pathutils.CLUSTER_CONF_FILE)
3292

    
3293

    
3294
def _GetX509Filenames(cryptodir, name):
3295
  """Returns the full paths for the private key and certificate.
3296

3297
  """
3298
  return (utils.PathJoin(cryptodir, name),
3299
          utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
3300
          utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
3301

    
3302

    
3303
def CreateX509Certificate(validity, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3304
  """Creates a new X509 certificate for SSL/TLS.
3305

3306
  @type validity: int
3307
  @param validity: Validity in seconds
3308
  @rtype: tuple; (string, string)
3309
  @return: Certificate name and public part
3310

3311
  """
3312
  (key_pem, cert_pem) = \
3313
    utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
3314
                                     min(validity, _MAX_SSL_CERT_VALIDITY))
3315

    
3316
  cert_dir = tempfile.mkdtemp(dir=cryptodir,
3317
                              prefix="x509-%s-" % utils.TimestampForFilename())
3318
  try:
3319
    name = os.path.basename(cert_dir)
3320
    assert len(name) > 5
3321

    
3322
    (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3323

    
3324
    utils.WriteFile(key_file, mode=0400, data=key_pem)
3325
    utils.WriteFile(cert_file, mode=0400, data=cert_pem)
3326

    
3327
    # Never return private key as it shouldn't leave the node
3328
    return (name, cert_pem)
3329
  except Exception:
3330
    shutil.rmtree(cert_dir, ignore_errors=True)
3331
    raise
3332

    
3333

    
3334
def RemoveX509Certificate(name, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3335
  """Removes a X509 certificate.
3336

3337
  @type name: string
3338
  @param name: Certificate name
3339

3340
  """
3341
  (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3342

    
3343
  utils.RemoveFile(key_file)
3344
  utils.RemoveFile(cert_file)
3345

    
3346
  try:
3347
    os.rmdir(cert_dir)
3348
  except EnvironmentError, err:
3349
    _Fail("Cannot remove certificate directory '%s': %s",
3350
          cert_dir, err)
3351

    
3352

    
3353
def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
3354
  """Returns the command for the requested input/output.
3355

3356
  @type instance: L{objects.Instance}
3357
  @param instance: The instance object
3358
  @param mode: Import/export mode
3359
  @param ieio: Input/output type
3360
  @param ieargs: Input/output arguments
3361

3362
  """
3363
  assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
3364

    
3365
  env = None
3366
  prefix = None
3367
  suffix = None
3368
  exp_size = None
3369

    
3370
  if ieio == constants.IEIO_FILE:
3371
    (filename, ) = ieargs
3372

    
3373
    if not utils.IsNormAbsPath(filename):
3374
      _Fail("Path '%s' is not normalized or absolute", filename)
3375

    
3376
    real_filename = os.path.realpath(filename)
3377
    directory = os.path.dirname(real_filename)
3378

    
3379
    if not utils.IsBelowDir(pathutils.EXPORT_DIR, real_filename):
3380
      _Fail("File '%s' is not under exports directory '%s': %s",
3381
            filename, pathutils.EXPORT_DIR, real_filename)
3382

    
3383
    # Create directory
3384
    utils.Makedirs(directory, mode=0750)
3385

    
3386
    quoted_filename = utils.ShellQuote(filename)
3387

    
3388
    if mode == constants.IEM_IMPORT:
3389
      suffix = "> %s" % quoted_filename
3390
    elif mode == constants.IEM_EXPORT:
3391
      suffix = "< %s" % quoted_filename
3392

    
3393
      # Retrieve file size
3394
      try:
3395
        st = os.stat(filename)
3396
      except EnvironmentError, err:
3397
        logging.error("Can't stat(2) %s: %s", filename, err)
3398
      else:
3399
        exp_size = utils.BytesToMebibyte(st.st_size)
3400

    
3401
  elif ieio == constants.IEIO_RAW_DISK:
3402
    (disk, ) = ieargs
3403

    
3404
    real_disk = _OpenRealBD(disk)
3405

    
3406
    if mode == constants.IEM_IMPORT:
3407
      # we set here a smaller block size as, due to transport buffering, more
3408
      # than 64-128k will mostly ignored; we use nocreat to fail if the device
3409
      # is not already there or we pass a wrong path; we use notrunc to no
3410
      # attempt truncate on an LV device; we use oflag=dsync to not buffer too
3411
      # much memory; this means that at best, we flush every 64k, which will
3412
      # not be very fast
3413
      suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
3414
                                    " bs=%s oflag=dsync"),
3415
                                    real_disk.dev_path,
3416
                                    str(64 * 1024))
3417

    
3418
    elif mode == constants.IEM_EXPORT:
3419
      # the block size on the read dd is 1MiB to match our units
3420
      prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
3421
                                   real_disk.dev_path,
3422
                                   str(1024 * 1024), # 1 MB
3423
                                   str(disk.size))
3424
      exp_size = disk.size
3425

    
3426
  elif ieio == constants.IEIO_SCRIPT:
3427
    (disk, disk_index, ) = ieargs
3428

    
3429
    assert isinstance(disk_index, (int, long))
3430

    
3431
    real_disk = _OpenRealBD(disk)
3432

    
3433
    inst_os = OSFromDisk(instance.os)
3434
    env = OSEnvironment(instance, inst_os)
3435

    
3436
    if mode == constants.IEM_IMPORT:
3437
      env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
3438
      env["IMPORT_INDEX"] = str(disk_index)
3439
      script = inst_os.import_script
3440

    
3441
    elif mode == constants.IEM_EXPORT:
3442
      env["EXPORT_DEVICE"] = real_disk.dev_path
3443
      env["EXPORT_INDEX"] = str(disk_index)
3444
      script = inst_os.export_script
3445

    
3446
    # TODO: Pass special environment only to script
3447
    script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
3448

    
3449
    if mode == constants.IEM_IMPORT:
3450
      suffix = "| %s" % script_cmd
3451

    
3452
    elif mode == constants.IEM_EXPORT:
3453
      prefix = "%s |" % script_cmd
3454

    
3455
    # Let script predict size
3456
    exp_size = constants.IE_CUSTOM_SIZE
3457

    
3458
  else:
3459
    _Fail("Invalid %s I/O mode %r", mode, ieio)
3460

    
3461
  return (env, prefix, suffix, exp_size)
3462

    
3463

    
3464
def _CreateImportExportStatusDir(prefix):
3465
  """Creates status directory for import/export.
3466

3467
  """
3468
  return tempfile.mkdtemp(dir=pathutils.IMPORT_EXPORT_DIR,
3469
                          prefix=("%s-%s-" %
3470
                                  (prefix, utils.TimestampForFilename())))
3471

    
3472

    
3473
def StartImportExportDaemon(mode, opts, host, port, instance, component,
3474
                            ieio, ieioargs):
3475
  """Starts an import or export daemon.
3476

3477
  @param mode: Import/output mode
3478
  @type opts: L{objects.ImportExportOptions}
3479
  @param opts: Daemon options
3480
  @type host: string
3481
  @param host: Remote host for export (None for import)
3482
  @type port: int
3483
  @param port: Remote port for export (None for import)
3484
  @type instance: L{objects.Instance}
3485
  @param instance: Instance object
3486
  @type component: string
3487
  @param component: which part of the instance is transferred now,
3488
      e.g. 'disk/0'
3489
  @param ieio: Input/output type
3490
  @param ieioargs: Input/output arguments
3491

3492
  """
3493
  if mode == constants.IEM_IMPORT:
3494
    prefix = "import"
3495

    
3496
    if not (host is None and port is None):
3497
      _Fail("Can not specify host or port on import")
3498

    
3499
  elif mode == constants.IEM_EXPORT:
3500
    prefix = "export"
3501

    
3502
    if host is None or port is None:
3503
      _Fail("Host and port must be specified for an export")
3504

    
3505
  else:
3506
    _Fail("Invalid mode %r", mode)
3507

    
3508
  if (opts.key_name is None) ^ (opts.ca_pem is None):
3509
    _Fail("Cluster certificate can only be used for both key and CA")
3510

    
3511
  (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
3512
    _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
3513

    
3514
  if opts.key_name is None:
3515
    # Use server.pem
3516
    key_path = pathutils.NODED_CERT_FILE
3517
    cert_path = pathutils.NODED_CERT_FILE
3518
    assert opts.ca_pem is None
3519
  else:
3520
    (_, key_path, cert_path) = _GetX509Filenames(pathutils.CRYPTO_KEYS_DIR,
3521
                                                 opts.key_name)
3522
    assert opts.ca_pem is not None
3523

    
3524
  for i in [key_path, cert_path]:
3525
    if not os.path.exists(i):
3526
      _Fail("File '%s' does not exist" % i)
3527

    
3528
  status_dir = _CreateImportExportStatusDir("%s-%s" % (prefix, component))
3529
  try:
3530
    status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
3531
    pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
3532
    ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
3533

    
3534
    if opts.ca_pem is None:
3535
      # Use server.pem
3536
      ca = utils.ReadFile(pathutils.NODED_CERT_FILE)
3537
    else:
3538
      ca = opts.ca_pem
3539

    
3540
    # Write CA file
3541
    utils.WriteFile(ca_file, data=ca, mode=0400)
3542

    
3543
    cmd = [
3544
      pathutils.IMPORT_EXPORT_DAEMON,
3545
      status_file, mode,
3546
      "--key=%s" % key_path,
3547
      "--cert=%s" % cert_path,
3548
      "--ca=%s" % ca_file,
3549
      ]
3550

    
3551
    if host:
3552
      cmd.append("--host=%s" % host)
3553

    
3554
    if port:
3555
      cmd.append("--port=%s" % port)
3556

    
3557
    if opts.ipv6:
3558
      cmd.append("--ipv6")
3559
    else:
3560
      cmd.append("--ipv4")
3561

    
3562
    if opts.compress:
3563
      cmd.append("--compress=%s" % opts.compress)
3564

    
3565
    if opts.magic:
3566
      cmd.append("--magic=%s" % opts.magic)
3567

    
3568
    if exp_size is not None:
3569
      cmd.append("--expected-size=%s" % exp_size)
3570

    
3571
    if cmd_prefix:
3572
      cmd.append("--cmd-prefix=%s" % cmd_prefix)
3573

    
3574
    if cmd_suffix:
3575
      cmd.append("--cmd-suffix=%s" % cmd_suffix)
3576

    
3577
    if mode == constants.IEM_EXPORT:
3578
      # Retry connection a few times when connecting to remote peer
3579
      cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3580
      cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3581
    elif opts.connect_timeout is not None:
3582
      assert mode == constants.IEM_IMPORT
3583
      # Overall timeout for establishing connection while listening
3584
      cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3585

    
3586
    logfile = _InstanceLogName(prefix, instance.os, instance.name, component)
3587

    
3588
    # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3589
    # support for receiving a file descriptor for output
3590
    utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3591
                      output=logfile)
3592

    
3593
    # The import/export name is simply the status directory name
3594
    return os.path.basename(status_dir)
3595

    
3596
  except Exception:
3597
    shutil.rmtree(status_dir, ignore_errors=True)
3598
    raise
3599

    
3600

    
3601
def GetImportExportStatus(names):
3602
  """Returns import/export daemon status.
3603

3604
  @type names: sequence
3605
  @param names: List of names
3606
  @rtype: List of dicts
3607
  @return: Returns a list of the state of each named import/export or None if a
3608
           status couldn't be read
3609

3610
  """
3611
  result = []
3612

    
3613
  for name in names:
3614
    status_file = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name,
3615
                                 _IES_STATUS_FILE)
3616

    
3617
    try:
3618
      data = utils.ReadFile(status_file)
3619
    except EnvironmentError, err:
3620
      if err.errno != errno.ENOENT:
3621
        raise
3622
      data = None
3623

    
3624
    if not data:
3625
      result.append(None)
3626
      continue
3627

    
3628
    result.append(serializer.LoadJson(data))
3629

    
3630
  return result
3631

    
3632

    
3633
def AbortImportExport(name):
3634
  """Sends SIGTERM to a running import/export daemon.
3635

3636
  """
3637
  logging.info("Abort import/export %s", name)
3638

    
3639
  status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3640
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3641

    
3642
  if pid:
3643
    logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3644
                 name, pid)
3645
    utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3646

    
3647

    
3648
def CleanupImportExport(name):
3649
  """Cleanup after an import or export.
3650

3651
  If the import/export daemon is still running it's killed. Afterwards the
3652
  whole status directory is removed.
3653

3654
  """
3655
  logging.info("Finalizing import/export %s", name)
3656

    
3657
  status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3658

    
3659
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3660

    
3661
  if pid:
3662
    logging.info("Import/export %s is still running with PID %s",
3663
                 name, pid)
3664
    utils.KillProcess(pid, waitpid=False)
3665

    
3666
  shutil.rmtree(status_dir, ignore_errors=True)
3667

    
3668

    
3669
def _FindDisks(nodes_ip, disks):
3670
  """Sets the physical ID on disks and returns the block devices.
3671

3672
  """
3673
  # set the correct physical ID
3674
  my_name = netutils.Hostname.GetSysName()
3675
  for cf in disks:
3676
    cf.SetPhysicalID(my_name, nodes_ip)
3677

    
3678
  bdevs = []
3679

    
3680
  for cf in disks:
3681
    rd = _RecursiveFindBD(cf)
3682
    if rd is None:
3683
      _Fail("Can't find device %s", cf)
3684
    bdevs.append(rd)
3685
  return bdevs
3686

    
3687

    
3688
def DrbdDisconnectNet(nodes_ip, disks):
3689
  """Disconnects the network on a list of drbd devices.
3690

3691
  """
3692
  bdevs = _FindDisks(nodes_ip, disks)
3693

    
3694
  # disconnect disks
3695
  for rd in bdevs:
3696
    try:
3697
      rd.DisconnectNet()
3698
    except errors.BlockDeviceError, err:
3699
      _Fail("Can't change network configuration to standalone mode: %s",
3700
            err, exc=True)
3701

    
3702

    
3703
def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
3704
  """Attaches the network on a list of drbd devices.
3705

3706
  """
3707
  bdevs = _FindDisks(nodes_ip, disks)
3708

    
3709
  if multimaster:
3710
    for idx, rd in enumerate(bdevs):
3711
      try:
3712
        _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3713
      except EnvironmentError, err:
3714
        _Fail("Can't create symlink: %s", err)
3715
  # reconnect disks, switch to new master configuration and if
3716
  # needed primary mode
3717
  for rd in bdevs:
3718
    try:
3719
      rd.AttachNet(multimaster)
3720
    except errors.BlockDeviceError, err:
3721
      _Fail("Can't change network configuration: %s", err)
3722

    
3723
  # wait until the disks are connected; we need to retry the re-attach
3724
  # if the device becomes standalone, as this might happen if the one
3725
  # node disconnects and reconnects in a different mode before the
3726
  # other node reconnects; in this case, one or both of the nodes will
3727
  # decide it has wrong configuration and switch to standalone
3728

    
3729
  def _Attach():
3730
    all_connected = True
3731

    
3732
    for rd in bdevs:
3733
      stats = rd.GetProcStatus()
3734

    
3735
      all_connected = (all_connected and
3736
                       (stats.is_connected or stats.is_in_resync))
3737

    
3738
      if stats.is_standalone:
3739
        # peer had different config info and this node became
3740
        # standalone, even though this should not happen with the
3741
        # new staged way of changing disk configs
3742
        try:
3743
          rd.AttachNet(multimaster)
3744
        except errors.BlockDeviceError, err:
3745
          _Fail("Can't change network configuration: %s", err)
3746

    
3747
    if not all_connected:
3748
      raise utils.RetryAgain()
3749

    
3750
  try:
3751
    # Start with a delay of 100 miliseconds and go up to 5 seconds
3752
    utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
3753
  except utils.RetryTimeout:
3754
    _Fail("Timeout in disk reconnecting")
3755

    
3756
  if multimaster:
3757
    # change to primary mode
3758
    for rd in bdevs:
3759
      try:
3760
        rd.Open()
3761
      except errors.BlockDeviceError, err:
3762
        _Fail("Can't change to primary mode: %s", err)
3763

    
3764

    
3765
def DrbdWaitSync(nodes_ip, disks):
3766
  """Wait until DRBDs have synchronized.
3767

3768
  """
3769
  def _helper(rd):
3770
    stats = rd.GetProcStatus()
3771
    if not (stats.is_connected or stats.is_in_resync):
3772
      raise utils.RetryAgain()
3773
    return stats
3774

    
3775
  bdevs = _FindDisks(nodes_ip, disks)
3776

    
3777
  min_resync = 100
3778
  alldone = True
3779
  for rd in bdevs:
3780
    try:
3781
      # poll each second for 15 seconds
3782
      stats = utils.Retry(_helper, 1, 15, args=[rd])
3783
    except utils.RetryTimeout:
3784
      stats = rd.GetProcStatus()
3785
      # last check
3786
      if not (stats.is_connected or stats.is_in_resync):
3787
        _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
3788
    alldone = alldone and (not stats.is_in_resync)
3789
    if stats.sync_percent is not None:
3790
      min_resync = min(min_resync, stats.sync_percent)
3791

    
3792
  return (alldone, min_resync)
3793

    
3794

    
3795
def GetDrbdUsermodeHelper():
3796
  """Returns DRBD usermode helper currently configured.
3797

3798
  """
3799
  try:
3800
    return drbd.DRBD8.GetUsermodeHelper()
3801
  except errors.BlockDeviceError, err:
3802
    _Fail(str(err))
3803

    
3804

    
3805
def PowercycleNode(hypervisor_type):
3806
  """Hard-powercycle the node.
3807

3808
  Because we need to return first, and schedule the powercycle in the
3809
  background, we won't be able to report failures nicely.
3810

3811
  """
3812
  hyper = hypervisor.GetHypervisor(hypervisor_type)
3813
  try:
3814
    pid = os.fork()
3815
  except OSError:
3816
    # if we can't fork, we'll pretend that we're in the child process
3817
    pid = 0
3818
  if pid > 0:
3819
    return "Reboot scheduled in 5 seconds"
3820
  # ensure the child is running on ram
3821
  try:
3822
    utils.Mlockall()
3823
  except Exception: # pylint: disable=W0703
3824
    pass
3825
  time.sleep(5)
3826
  hyper.PowercycleNode()
3827

    
3828

    
3829
def _VerifyRestrictedCmdName(cmd):
3830
  """Verifies a restricted command name.
3831

3832
  @type cmd: string
3833
  @param cmd: Command name
3834
  @rtype: tuple; (boolean, string or None)
3835
  @return: The tuple's first element is the status; if C{False}, the second
3836
    element is an error message string, otherwise it's C{None}
3837

3838
  """
3839
  if not cmd.strip():
3840
    return (False, "Missing command name")
3841

    
3842
  if os.path.basename(cmd) != cmd:
3843
    return (False, "Invalid command name")
3844

    
3845
  if not constants.EXT_PLUGIN_MASK.match(cmd):
3846
    return (False, "Command name contains forbidden characters")
3847

    
3848
  return (True, None)
3849

    
3850

    
3851
def _CommonRestrictedCmdCheck(path, owner):
3852
  """Common checks for restricted command file system directories and files.
3853

3854
  @type path: string
3855
  @param path: Path to check
3856
  @param owner: C{None} or tuple containing UID and GID
3857
  @rtype: tuple; (boolean, string or C{os.stat} result)
3858
  @return: The tuple's first element is the status; if C{False}, the second
3859
    element is an error message string, otherwise it's the result of C{os.stat}
3860

3861
  """
3862
  if owner is None:
3863
    # Default to root as owner
3864
    owner = (0, 0)
3865

    
3866
  try:
3867
    st = os.stat(path)
3868
  except EnvironmentError, err:
3869
    return (False, "Can't stat(2) '%s': %s" % (path, err))
3870

    
3871
  if stat.S_IMODE(st.st_mode) & (~_RCMD_MAX_MODE):
3872
    return (False, "Permissions on '%s' are too permissive" % path)
3873

    
3874
  if (st.st_uid, st.st_gid) != owner:
3875
    (owner_uid, owner_gid) = owner
3876
    return (False, "'%s' is not owned by %s:%s" % (path, owner_uid, owner_gid))
3877

    
3878
  return (True, st)
3879

    
3880

    
3881
def _VerifyRestrictedCmdDirectory(path, _owner=None):
3882
  """Verifies restricted command directory.
3883

3884
  @type path: string
3885
  @param path: Path to check
3886
  @rtype: tuple; (boolean, string or None)
3887
  @return: The tuple's first element is the status; if C{False}, the second
3888
    element is an error message string, otherwise it's C{None}
3889

3890
  """
3891
  (status, value) = _CommonRestrictedCmdCheck(path, _owner)
3892

    
3893
  if not status:
3894
    return (False, value)
3895

    
3896
  if not stat.S_ISDIR(value.st_mode):
3897
    return (False, "Path '%s' is not a directory" % path)
3898

    
3899
  return (True, None)
3900

    
3901

    
3902
def _VerifyRestrictedCmd(path, cmd, _owner=None):
3903
  """Verifies a whole restricted command and returns its executable filename.
3904

3905
  @type path: string
3906
  @param path: Directory containing restricted commands
3907
  @type cmd: string
3908
  @param cmd: Command name
3909
  @rtype: tuple; (boolean, string)
3910
  @return: The tuple's first element is the status; if C{False}, the second
3911
    element is an error message string, otherwise the second element is the
3912
    absolute path to the executable
3913

3914
  """
3915
  executable = utils.PathJoin(path, cmd)
3916

    
3917
  (status, msg) = _CommonRestrictedCmdCheck(executable, _owner)
3918

    
3919
  if not status:
3920
    return (False, msg)
3921

    
3922
  if not utils.IsExecutable(executable):
3923
    return (False, "access(2) thinks '%s' can't be executed" % executable)
3924

    
3925
  return (True, executable)
3926

    
3927

    
3928
def _PrepareRestrictedCmd(path, cmd,
3929
                          _verify_dir=_VerifyRestrictedCmdDirectory,
3930
                          _verify_name=_VerifyRestrictedCmdName,
3931
                          _verify_cmd=_VerifyRestrictedCmd):
3932
  """Performs a number of tests on a restricted command.
3933

3934
  @type path: string
3935
  @param path: Directory containing restricted commands
3936
  @type cmd: string
3937
  @param cmd: Command name
3938
  @return: Same as L{_VerifyRestrictedCmd}
3939

3940
  """
3941
  # Verify the directory first
3942
  (status, msg) = _verify_dir(path)
3943
  if status:
3944
    # Check command if everything was alright
3945
    (status, msg) = _verify_name(cmd)
3946

    
3947
  if not status:
3948
    return (False, msg)
3949

    
3950
  # Check actual executable
3951
  return _verify_cmd(path, cmd)
3952

    
3953

    
3954
def RunRestrictedCmd(cmd,
3955
                     _lock_timeout=_RCMD_LOCK_TIMEOUT,
3956
                     _lock_file=pathutils.RESTRICTED_COMMANDS_LOCK_FILE,
3957
                     _path=pathutils.RESTRICTED_COMMANDS_DIR,
3958
                     _sleep_fn=time.sleep,
3959
                     _prepare_fn=_PrepareRestrictedCmd,
3960
                     _runcmd_fn=utils.RunCmd,
3961
                     _enabled=constants.ENABLE_RESTRICTED_COMMANDS):
3962
  """Executes a restricted command after performing strict tests.
3963

3964
  @type cmd: string
3965
  @param cmd: Command name
3966
  @rtype: string
3967
  @return: Command output
3968
  @raise RPCFail: In case of an error
3969

3970
  """
3971
  logging.info("Preparing to run restricted command '%s'", cmd)
3972

    
3973
  if not _enabled:
3974
    _Fail("Restricted commands disabled at configure time")
3975

    
3976
  lock = None
3977
  try:
3978
    cmdresult = None
3979
    try:
3980
      lock = utils.FileLock.Open(_lock_file)
3981
      lock.Exclusive(blocking=True, timeout=_lock_timeout)
3982

    
3983
      (status, value) = _prepare_fn(_path, cmd)
3984

    
3985
      if status:
3986
        cmdresult = _runcmd_fn([value], env={}, reset_env=True,
3987
                               postfork_fn=lambda _: lock.Unlock())
3988
      else:
3989
        logging.error(value)
3990
    except Exception: # pylint: disable=W0703
3991
      # Keep original error in log
3992
      logging.exception("Caught exception")
3993

    
3994
    if cmdresult is None:
3995
      logging.info("Sleeping for %0.1f seconds before returning",
3996
                   _RCMD_INVALID_DELAY)
3997
      _sleep_fn(_RCMD_INVALID_DELAY)
3998

    
3999
      # Do not include original error message in returned error
4000
      _Fail("Executing command '%s' failed" % cmd)
4001
    elif cmdresult.failed or cmdresult.fail_reason:
4002
      _Fail("Restricted command '%s' failed: %s; output: %s",
4003
            cmd, cmdresult.fail_reason, cmdresult.output)
4004
    else:
4005
      return cmdresult.output
4006
  finally:
4007
    if lock is not None:
4008
      # Release lock at last
4009
      lock.Close()
4010
      lock = None
4011

    
4012

    
4013
def SetWatcherPause(until, _filename=pathutils.WATCHER_PAUSEFILE):
4014
  """Creates or removes the watcher pause file.
4015

4016
  @type until: None or number
4017
  @param until: Unix timestamp saying until when the watcher shouldn't run
4018

4019
  """
4020
  if until is None:
4021
    logging.info("Received request to no longer pause watcher")
4022
    utils.RemoveFile(_filename)
4023
  else:
4024
    logging.info("Received request to pause watcher until %s", until)
4025

    
4026
    if not ht.TNumber(until):
4027
      _Fail("Duration must be numeric")
4028

    
4029
    utils.WriteFile(_filename, data="%d\n" % (until, ), mode=0644)
4030

    
4031

    
4032
class HooksRunner(object):
4033
  """Hook runner.
4034

4035
  This class is instantiated on the node side (ganeti-noded) and not
4036
  on the master side.
4037

4038
  """
4039
  def __init__(self, hooks_base_dir=None):
4040
    """Constructor for hooks runner.
4041

4042
    @type hooks_base_dir: str or None
4043
    @param hooks_base_dir: if not None, this overrides the
4044
        L{pathutils.HOOKS_BASE_DIR} (useful for unittests)
4045

4046
    """
4047
    if hooks_base_dir is None:
4048
      hooks_base_dir = pathutils.HOOKS_BASE_DIR
4049
    # yeah, _BASE_DIR is not valid for attributes, we use it like a
4050
    # constant
4051
    self._BASE_DIR = hooks_base_dir # pylint: disable=C0103
4052

    
4053
  def RunLocalHooks(self, node_list, hpath, phase, env):
4054
    """Check that the hooks will be run only locally and then run them.
4055

4056
    """
4057
    assert len(node_list) == 1
4058
    node = node_list[0]
4059
    _, myself = ssconf.GetMasterAndMyself()
4060
    assert node == myself
4061

    
4062
    results = self.RunHooks(hpath, phase, env)
4063

    
4064
    # Return values in the form expected by HooksMaster
4065
    return {node: (None, False, results)}
4066

    
4067
  def RunHooks(self, hpath, phase, env):
4068
    """Run the scripts in the hooks directory.
4069

4070
    @type hpath: str
4071
    @param hpath: the path to the hooks directory which
4072
        holds the scripts
4073
    @type phase: str
4074
    @param phase: either L{constants.HOOKS_PHASE_PRE} or
4075
        L{constants.HOOKS_PHASE_POST}
4076
    @type env: dict
4077
    @param env: dictionary with the environment for the hook
4078
    @rtype: list
4079
    @return: list of 3-element tuples:
4080
      - script path
4081
      - script result, either L{constants.HKR_SUCCESS} or
4082
        L{constants.HKR_FAIL}
4083
      - output of the script
4084

4085
    @raise errors.ProgrammerError: for invalid input
4086
        parameters
4087

4088
    """
4089
    if phase == constants.HOOKS_PHASE_PRE:
4090
      suffix = "pre"
4091
    elif phase == constants.HOOKS_PHASE_POST:
4092
      suffix = "post"
4093
    else:
4094
      _Fail("Unknown hooks phase '%s'", phase)
4095

    
4096
    subdir = "%s-%s.d" % (hpath, suffix)
4097
    dir_name = utils.PathJoin(self._BASE_DIR, subdir)
4098

    
4099
    results = []
4100

    
4101
    if not os.path.isdir(dir_name):
4102
      # for non-existing/non-dirs, we simply exit instead of logging a
4103
      # warning at every operation
4104
      return results
4105

    
4106
    runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
4107

    
4108
    for (relname, relstatus, runresult) in runparts_results:
4109
      if relstatus == constants.RUNPARTS_SKIP:
4110
        rrval = constants.HKR_SKIP
4111
        output = ""
4112
      elif relstatus == constants.RUNPARTS_ERR:
4113
        rrval = constants.HKR_FAIL
4114
        output = "Hook script execution error: %s" % runresult
4115
      elif relstatus == constants.RUNPARTS_RUN:
4116
        if runresult.failed:
4117
          rrval = constants.HKR_FAIL