Statistics
| Branch: | Tag: | Revision:

root / lib / backend.py @ 19e1b715

History | View | Annotate | Download (122.8 kB)

1
#
2
#
3

    
4
# Copyright (C) 2006, 2007, 2008, 2009, 2010, 2011, 2012 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 import bdev
58
from ganeti import objects
59
from ganeti import ssconf
60
from ganeti import serializer
61
from ganeti import netutils
62
from ganeti import runtime
63
from ganeti import mcpu
64
from ganeti import compat
65
from ganeti import pathutils
66
from ganeti import vcluster
67
from ganeti import ht
68

    
69

    
70
_BOOT_ID_PATH = "/proc/sys/kernel/random/boot_id"
71
_ALLOWED_CLEAN_DIRS = compat.UniqueFrozenset([
72
  pathutils.DATA_DIR,
73
  pathutils.JOB_QUEUE_ARCHIVE_DIR,
74
  pathutils.QUEUE_DIR,
75
  pathutils.CRYPTO_KEYS_DIR,
76
  ])
77
_MAX_SSL_CERT_VALIDITY = 7 * 24 * 60 * 60
78
_X509_KEY_FILE = "key"
79
_X509_CERT_FILE = "cert"
80
_IES_STATUS_FILE = "status"
81
_IES_PID_FILE = "pid"
82
_IES_CA_FILE = "ca"
83

    
84
#: Valid LVS output line regex
85
_LVSLINE_REGEX = re.compile("^ *([^|]+)\|([^|]+)\|([0-9.]+)\|([^|]{6,})\|?$")
86

    
87
# Actions for the master setup script
88
_MASTER_START = "start"
89
_MASTER_STOP = "stop"
90

    
91
#: Maximum file permissions for remote command directory and executables
92
_RCMD_MAX_MODE = (stat.S_IRWXU |
93
                  stat.S_IRGRP | stat.S_IXGRP |
94
                  stat.S_IROTH | stat.S_IXOTH)
95

    
96
#: Delay before returning an error for remote commands
97
_RCMD_INVALID_DELAY = 10
98

    
99
#: How long to wait to acquire lock for remote commands (shorter than
100
#: L{_RCMD_INVALID_DELAY}) to reduce blockage of noded forks when many
101
#: command requests arrive
102
_RCMD_LOCK_TIMEOUT = _RCMD_INVALID_DELAY * 0.8
103

    
104

    
105
class RPCFail(Exception):
106
  """Class denoting RPC failure.
107

108
  Its argument is the error message.
109

110
  """
111

    
112

    
113
def _Fail(msg, *args, **kwargs):
114
  """Log an error and the raise an RPCFail exception.
115

116
  This exception is then handled specially in the ganeti daemon and
117
  turned into a 'failed' return type. As such, this function is a
118
  useful shortcut for logging the error and returning it to the master
119
  daemon.
120

121
  @type msg: string
122
  @param msg: the text of the exception
123
  @raise RPCFail
124

125
  """
126
  if args:
127
    msg = msg % args
128
  if "log" not in kwargs or kwargs["log"]: # if we should log this error
129
    if "exc" in kwargs and kwargs["exc"]:
130
      logging.exception(msg)
131
    else:
132
      logging.error(msg)
133
  raise RPCFail(msg)
134

    
135

    
136
def _GetConfig():
137
  """Simple wrapper to return a SimpleStore.
138

139
  @rtype: L{ssconf.SimpleStore}
140
  @return: a SimpleStore instance
141

142
  """
143
  return ssconf.SimpleStore()
144

    
145

    
146
def _GetSshRunner(cluster_name):
147
  """Simple wrapper to return an SshRunner.
148

149
  @type cluster_name: str
150
  @param cluster_name: the cluster name, which is needed
151
      by the SshRunner constructor
152
  @rtype: L{ssh.SshRunner}
153
  @return: an SshRunner instance
154

155
  """
156
  return ssh.SshRunner(cluster_name)
157

    
158

    
159
def _Decompress(data):
160
  """Unpacks data compressed by the RPC client.
161

162
  @type data: list or tuple
163
  @param data: Data sent by RPC client
164
  @rtype: str
165
  @return: Decompressed data
166

167
  """
168
  assert isinstance(data, (list, tuple))
169
  assert len(data) == 2
170
  (encoding, content) = data
171
  if encoding == constants.RPC_ENCODING_NONE:
172
    return content
173
  elif encoding == constants.RPC_ENCODING_ZLIB_BASE64:
174
    return zlib.decompress(base64.b64decode(content))
175
  else:
176
    raise AssertionError("Unknown data encoding")
177

    
178

    
179
def _CleanDirectory(path, exclude=None):
180
  """Removes all regular files in a directory.
181

182
  @type path: str
183
  @param path: the directory to clean
184
  @type exclude: list
185
  @param exclude: list of files to be excluded, defaults
186
      to the empty list
187

188
  """
189
  if path not in _ALLOWED_CLEAN_DIRS:
190
    _Fail("Path passed to _CleanDirectory not in allowed clean targets: '%s'",
191
          path)
192

    
193
  if not os.path.isdir(path):
194
    return
195
  if exclude is None:
196
    exclude = []
197
  else:
198
    # Normalize excluded paths
199
    exclude = [os.path.normpath(i) for i in exclude]
200

    
201
  for rel_name in utils.ListVisibleFiles(path):
202
    full_name = utils.PathJoin(path, rel_name)
203
    if full_name in exclude:
204
      continue
205
    if os.path.isfile(full_name) and not os.path.islink(full_name):
206
      utils.RemoveFile(full_name)
207

    
208

    
209
def _BuildUploadFileList():
210
  """Build the list of allowed upload files.
211

212
  This is abstracted so that it's built only once at module import time.
213

214
  """
215
  allowed_files = set([
216
    pathutils.CLUSTER_CONF_FILE,
217
    pathutils.ETC_HOSTS,
218
    pathutils.SSH_KNOWN_HOSTS_FILE,
219
    pathutils.VNC_PASSWORD_FILE,
220
    pathutils.RAPI_CERT_FILE,
221
    pathutils.SPICE_CERT_FILE,
222
    pathutils.SPICE_CACERT_FILE,
223
    pathutils.RAPI_USERS_FILE,
224
    pathutils.CONFD_HMAC_KEY,
225
    pathutils.CLUSTER_DOMAIN_SECRET_FILE,
226
    ])
227

    
228
  for hv_name in constants.HYPER_TYPES:
229
    hv_class = hypervisor.GetHypervisorClass(hv_name)
230
    allowed_files.update(hv_class.GetAncillaryFiles()[0])
231

    
232
  assert pathutils.FILE_STORAGE_PATHS_FILE not in allowed_files, \
233
    "Allowed file storage paths should never be uploaded via RPC"
234

    
235
  return frozenset(allowed_files)
236

    
237

    
238
_ALLOWED_UPLOAD_FILES = _BuildUploadFileList()
239

    
240

    
241
def JobQueuePurge():
242
  """Removes job queue files and archived jobs.
243

244
  @rtype: tuple
245
  @return: True, None
246

247
  """
248
  _CleanDirectory(pathutils.QUEUE_DIR, exclude=[pathutils.JOB_QUEUE_LOCK_FILE])
249
  _CleanDirectory(pathutils.JOB_QUEUE_ARCHIVE_DIR)
250

    
251

    
252
def GetMasterInfo():
253
  """Returns master information.
254

255
  This is an utility function to compute master information, either
256
  for consumption here or from the node daemon.
257

258
  @rtype: tuple
259
  @return: master_netdev, master_ip, master_name, primary_ip_family,
260
    master_netmask
261
  @raise RPCFail: in case of errors
262

263
  """
264
  try:
265
    cfg = _GetConfig()
266
    master_netdev = cfg.GetMasterNetdev()
267
    master_ip = cfg.GetMasterIP()
268
    master_netmask = cfg.GetMasterNetmask()
269
    master_node = cfg.GetMasterNode()
270
    primary_ip_family = cfg.GetPrimaryIPFamily()
271
  except errors.ConfigurationError, err:
272
    _Fail("Cluster configuration incomplete: %s", err, exc=True)
273
  return (master_netdev, master_ip, master_node, primary_ip_family,
274
          master_netmask)
275

    
276

    
277
def RunLocalHooks(hook_opcode, hooks_path, env_builder_fn):
278
  """Decorator that runs hooks before and after the decorated function.
279

280
  @type hook_opcode: string
281
  @param hook_opcode: opcode of the hook
282
  @type hooks_path: string
283
  @param hooks_path: path of the hooks
284
  @type env_builder_fn: function
285
  @param env_builder_fn: function that returns a dictionary containing the
286
    environment variables for the hooks. Will get all the parameters of the
287
    decorated function.
288
  @raise RPCFail: in case of pre-hook failure
289

290
  """
291
  def decorator(fn):
292
    def wrapper(*args, **kwargs):
293
      _, myself = ssconf.GetMasterAndMyself()
294
      nodes = ([myself], [myself])  # these hooks run locally
295

    
296
      env_fn = compat.partial(env_builder_fn, *args, **kwargs)
297

    
298
      cfg = _GetConfig()
299
      hr = HooksRunner()
300
      hm = mcpu.HooksMaster(hook_opcode, hooks_path, nodes, hr.RunLocalHooks,
301
                            None, env_fn, logging.warning, cfg.GetClusterName(),
302
                            cfg.GetMasterNode())
303

    
304
      hm.RunPhase(constants.HOOKS_PHASE_PRE)
305
      result = fn(*args, **kwargs)
306
      hm.RunPhase(constants.HOOKS_PHASE_POST)
307

    
308
      return result
309
    return wrapper
310
  return decorator
311

    
312

    
313
def _BuildMasterIpEnv(master_params, use_external_mip_script=None):
314
  """Builds environment variables for master IP hooks.
315

316
  @type master_params: L{objects.MasterNetworkParameters}
317
  @param master_params: network parameters of the master
318
  @type use_external_mip_script: boolean
319
  @param use_external_mip_script: whether to use an external master IP
320
    address setup script (unused, but necessary per the implementation of the
321
    _RunLocalHooks decorator)
322

323
  """
324
  # pylint: disable=W0613
325
  ver = netutils.IPAddress.GetVersionFromAddressFamily(master_params.ip_family)
326
  env = {
327
    "MASTER_NETDEV": master_params.netdev,
328
    "MASTER_IP": master_params.ip,
329
    "MASTER_NETMASK": str(master_params.netmask),
330
    "CLUSTER_IP_VERSION": str(ver),
331
  }
332

    
333
  return env
334

    
335

    
336
def _RunMasterSetupScript(master_params, action, use_external_mip_script):
337
  """Execute the master IP address setup script.
338

339
  @type master_params: L{objects.MasterNetworkParameters}
340
  @param master_params: network parameters of the master
341
  @type action: string
342
  @param action: action to pass to the script. Must be one of
343
    L{backend._MASTER_START} or L{backend._MASTER_STOP}
344
  @type use_external_mip_script: boolean
345
  @param use_external_mip_script: whether to use an external master IP
346
    address setup script
347
  @raise backend.RPCFail: if there are errors during the execution of the
348
    script
349

350
  """
351
  env = _BuildMasterIpEnv(master_params)
352

    
353
  if use_external_mip_script:
354
    setup_script = pathutils.EXTERNAL_MASTER_SETUP_SCRIPT
355
  else:
356
    setup_script = pathutils.DEFAULT_MASTER_SETUP_SCRIPT
357

    
358
  result = utils.RunCmd([setup_script, action], env=env, reset_env=True)
359

    
360
  if result.failed:
361
    _Fail("Failed to %s the master IP. Script return value: %s, output: '%s'" %
362
          (action, result.exit_code, result.output), log=True)
363

    
364

    
365
@RunLocalHooks(constants.FAKE_OP_MASTER_TURNUP, "master-ip-turnup",
366
               _BuildMasterIpEnv)
367
def ActivateMasterIp(master_params, use_external_mip_script):
368
  """Activate the IP address of the master daemon.
369

370
  @type master_params: L{objects.MasterNetworkParameters}
371
  @param master_params: network parameters of the master
372
  @type use_external_mip_script: boolean
373
  @param use_external_mip_script: whether to use an external master IP
374
    address setup script
375
  @raise RPCFail: in case of errors during the IP startup
376

377
  """
378
  _RunMasterSetupScript(master_params, _MASTER_START,
379
                        use_external_mip_script)
380

    
381

    
382
def StartMasterDaemons(no_voting):
383
  """Activate local node as master node.
384

385
  The function will start the master daemons (ganeti-masterd and ganeti-rapi).
386

387
  @type no_voting: boolean
388
  @param no_voting: whether to start ganeti-masterd without a node vote
389
      but still non-interactively
390
  @rtype: None
391

392
  """
393

    
394
  if no_voting:
395
    masterd_args = "--no-voting --yes-do-it"
396
  else:
397
    masterd_args = ""
398

    
399
  env = {
400
    "EXTRA_MASTERD_ARGS": masterd_args,
401
    }
402

    
403
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "start-master"], env=env)
404
  if result.failed:
405
    msg = "Can't start Ganeti master: %s" % result.output
406
    logging.error(msg)
407
    _Fail(msg)
408

    
409

    
410
@RunLocalHooks(constants.FAKE_OP_MASTER_TURNDOWN, "master-ip-turndown",
411
               _BuildMasterIpEnv)
412
def DeactivateMasterIp(master_params, use_external_mip_script):
413
  """Deactivate the master IP on this node.
414

415
  @type master_params: L{objects.MasterNetworkParameters}
416
  @param master_params: network parameters of the master
417
  @type use_external_mip_script: boolean
418
  @param use_external_mip_script: whether to use an external master IP
419
    address setup script
420
  @raise RPCFail: in case of errors during the IP turndown
421

422
  """
423
  _RunMasterSetupScript(master_params, _MASTER_STOP,
424
                        use_external_mip_script)
425

    
426

    
427
def StopMasterDaemons():
428
  """Stop the master daemons on this node.
429

430
  Stop the master daemons (ganeti-masterd and ganeti-rapi) on this node.
431

432
  @rtype: None
433

434
  """
435
  # TODO: log and report back to the caller the error failures; we
436
  # need to decide in which case we fail the RPC for this
437

    
438
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop-master"])
439
  if result.failed:
440
    logging.error("Could not stop Ganeti master, command %s had exitcode %s"
441
                  " and error %s",
442
                  result.cmd, result.exit_code, result.output)
443

    
444

    
445
def ChangeMasterNetmask(old_netmask, netmask, master_ip, master_netdev):
446
  """Change the netmask of the master IP.
447

448
  @param old_netmask: the old value of the netmask
449
  @param netmask: the new value of the netmask
450
  @param master_ip: the master IP
451
  @param master_netdev: the master network device
452

453
  """
454
  if old_netmask == netmask:
455
    return
456

    
457
  if not netutils.IPAddress.Own(master_ip):
458
    _Fail("The master IP address is not up, not attempting to change its"
459
          " netmask")
460

    
461
  result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "add",
462
                         "%s/%s" % (master_ip, netmask),
463
                         "dev", master_netdev, "label",
464
                         "%s:0" % master_netdev])
465
  if result.failed:
466
    _Fail("Could not set the new netmask on the master IP address")
467

    
468
  result = utils.RunCmd([constants.IP_COMMAND_PATH, "address", "del",
469
                         "%s/%s" % (master_ip, old_netmask),
470
                         "dev", master_netdev, "label",
471
                         "%s:0" % master_netdev])
472
  if result.failed:
473
    _Fail("Could not bring down the master IP address with the old netmask")
474

    
475

    
476
def EtcHostsModify(mode, host, ip):
477
  """Modify a host entry in /etc/hosts.
478

479
  @param mode: The mode to operate. Either add or remove entry
480
  @param host: The host to operate on
481
  @param ip: The ip associated with the entry
482

483
  """
484
  if mode == constants.ETC_HOSTS_ADD:
485
    if not ip:
486
      RPCFail("Mode 'add' needs 'ip' parameter, but parameter not"
487
              " present")
488
    utils.AddHostToEtcHosts(host, ip)
489
  elif mode == constants.ETC_HOSTS_REMOVE:
490
    if ip:
491
      RPCFail("Mode 'remove' does not allow 'ip' parameter, but"
492
              " parameter is present")
493
    utils.RemoveHostFromEtcHosts(host)
494
  else:
495
    RPCFail("Mode not supported")
496

    
497

    
498
def LeaveCluster(modify_ssh_setup):
499
  """Cleans up and remove the current node.
500

501
  This function cleans up and prepares the current node to be removed
502
  from the cluster.
503

504
  If processing is successful, then it raises an
505
  L{errors.QuitGanetiException} which is used as a special case to
506
  shutdown the node daemon.
507

508
  @param modify_ssh_setup: boolean
509

510
  """
511
  _CleanDirectory(pathutils.DATA_DIR)
512
  _CleanDirectory(pathutils.CRYPTO_KEYS_DIR)
513
  JobQueuePurge()
514

    
515
  if modify_ssh_setup:
516
    try:
517
      priv_key, pub_key, auth_keys = ssh.GetUserFiles(constants.SSH_LOGIN_USER)
518

    
519
      utils.RemoveAuthorizedKey(auth_keys, utils.ReadFile(pub_key))
520

    
521
      utils.RemoveFile(priv_key)
522
      utils.RemoveFile(pub_key)
523
    except errors.OpExecError:
524
      logging.exception("Error while processing ssh files")
525

    
526
  try:
527
    utils.RemoveFile(pathutils.CONFD_HMAC_KEY)
528
    utils.RemoveFile(pathutils.RAPI_CERT_FILE)
529
    utils.RemoveFile(pathutils.SPICE_CERT_FILE)
530
    utils.RemoveFile(pathutils.SPICE_CACERT_FILE)
531
    utils.RemoveFile(pathutils.NODED_CERT_FILE)
532
  except: # pylint: disable=W0702
533
    logging.exception("Error while removing cluster secrets")
534

    
535
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "stop", constants.CONFD])
536
  if result.failed:
537
    logging.error("Command %s failed with exitcode %s and error %s",
538
                  result.cmd, result.exit_code, result.output)
539

    
540
  # Raise a custom exception (handled in ganeti-noded)
541
  raise errors.QuitGanetiException(True, "Shutdown scheduled")
542

    
543

    
544
def _GetVgInfo(name, excl_stor):
545
  """Retrieves information about a LVM volume group.
546

547
  """
548
  # TODO: GetVGInfo supports returning information for multiple VGs at once
549
  vginfo = bdev.LogicalVolume.GetVGInfo([name], excl_stor)
550
  if vginfo:
551
    vg_free = int(round(vginfo[0][0], 0))
552
    vg_size = int(round(vginfo[0][1], 0))
553
  else:
554
    vg_free = None
555
    vg_size = None
556

    
557
  return {
558
    "name": name,
559
    "vg_free": vg_free,
560
    "vg_size": vg_size,
561
    }
562

    
563

    
564
def _GetHvInfo(name):
565
  """Retrieves node information from a hypervisor.
566

567
  The information returned depends on the hypervisor. Common items:
568

569
    - vg_size is the size of the configured volume group in MiB
570
    - vg_free is the free size of the volume group in MiB
571
    - memory_dom0 is the memory allocated for domain0 in MiB
572
    - memory_free is the currently available (free) ram in MiB
573
    - memory_total is the total number of ram in MiB
574
    - hv_version: the hypervisor version, if available
575

576
  """
577
  return hypervisor.GetHypervisor(name).GetNodeInfo()
578

    
579

    
580
def _GetNamedNodeInfo(names, fn):
581
  """Calls C{fn} for all names in C{names} and returns a dictionary.
582

583
  @rtype: None or dict
584

585
  """
586
  if names is None:
587
    return None
588
  else:
589
    return map(fn, names)
590

    
591

    
592
def GetNodeInfo(vg_names, hv_names, excl_stor):
593
  """Gives back a hash with different information about the node.
594

595
  @type vg_names: list of string
596
  @param vg_names: Names of the volume groups to ask for disk space information
597
  @type hv_names: list of string
598
  @param hv_names: Names of the hypervisors to ask for node information
599
  @type excl_stor: boolean
600
  @param excl_stor: Whether exclusive_storage is active
601
  @rtype: tuple; (string, None/dict, None/dict)
602
  @return: Tuple containing boot ID, volume group information and hypervisor
603
    information
604

605
  """
606
  bootid = utils.ReadFile(_BOOT_ID_PATH, size=128).rstrip("\n")
607
  vg_info = _GetNamedNodeInfo(vg_names, (lambda vg: _GetVgInfo(vg, excl_stor)))
608
  hv_info = _GetNamedNodeInfo(hv_names, _GetHvInfo)
609

    
610
  return (bootid, vg_info, hv_info)
611

    
612

    
613
def VerifyNode(what, cluster_name):
614
  """Verify the status of the local node.
615

616
  Based on the input L{what} parameter, various checks are done on the
617
  local node.
618

619
  If the I{filelist} key is present, this list of
620
  files is checksummed and the file/checksum pairs are returned.
621

622
  If the I{nodelist} key is present, we check that we have
623
  connectivity via ssh with the target nodes (and check the hostname
624
  report).
625

626
  If the I{node-net-test} key is present, we check that we have
627
  connectivity to the given nodes via both primary IP and, if
628
  applicable, secondary IPs.
629

630
  @type what: C{dict}
631
  @param what: a dictionary of things to check:
632
      - filelist: list of files for which to compute checksums
633
      - nodelist: list of nodes we should check ssh communication with
634
      - node-net-test: list of nodes we should check node daemon port
635
        connectivity with
636
      - hypervisor: list with hypervisors to run the verify for
637
  @rtype: dict
638
  @return: a dictionary with the same keys as the input dict, and
639
      values representing the result of the checks
640

641
  """
642
  result = {}
643
  my_name = netutils.Hostname.GetSysName()
644
  port = netutils.GetDaemonPort(constants.NODED)
645
  vm_capable = my_name not in what.get(constants.NV_VMNODES, [])
646

    
647
  if constants.NV_HYPERVISOR in what and vm_capable:
648
    result[constants.NV_HYPERVISOR] = tmp = {}
649
    for hv_name in what[constants.NV_HYPERVISOR]:
650
      try:
651
        val = hypervisor.GetHypervisor(hv_name).Verify()
652
      except errors.HypervisorError, err:
653
        val = "Error while checking hypervisor: %s" % str(err)
654
      tmp[hv_name] = val
655

    
656
  if constants.NV_HVPARAMS in what and vm_capable:
657
    result[constants.NV_HVPARAMS] = tmp = []
658
    for source, hv_name, hvparms in what[constants.NV_HVPARAMS]:
659
      try:
660
        logging.info("Validating hv %s, %s", hv_name, hvparms)
661
        hypervisor.GetHypervisor(hv_name).ValidateParameters(hvparms)
662
      except errors.HypervisorError, err:
663
        tmp.append((source, hv_name, str(err)))
664

    
665
  if constants.NV_FILELIST in what:
666
    fingerprints = utils.FingerprintFiles(map(vcluster.LocalizeVirtualPath,
667
                                              what[constants.NV_FILELIST]))
668
    result[constants.NV_FILELIST] = \
669
      dict((vcluster.MakeVirtualPath(key), value)
670
           for (key, value) in fingerprints.items())
671

    
672
  if constants.NV_NODELIST in what:
673
    (nodes, bynode) = what[constants.NV_NODELIST]
674

    
675
    # Add nodes from other groups (different for each node)
676
    try:
677
      nodes.extend(bynode[my_name])
678
    except KeyError:
679
      pass
680

    
681
    # Use a random order
682
    random.shuffle(nodes)
683

    
684
    # Try to contact all nodes
685
    val = {}
686
    for node in nodes:
687
      success, message = _GetSshRunner(cluster_name).VerifyNodeHostname(node)
688
      if not success:
689
        val[node] = message
690

    
691
    result[constants.NV_NODELIST] = val
692

    
693
  if constants.NV_NODENETTEST in what:
694
    result[constants.NV_NODENETTEST] = tmp = {}
695
    my_pip = my_sip = None
696
    for name, pip, sip in what[constants.NV_NODENETTEST]:
697
      if name == my_name:
698
        my_pip = pip
699
        my_sip = sip
700
        break
701
    if not my_pip:
702
      tmp[my_name] = ("Can't find my own primary/secondary IP"
703
                      " in the node list")
704
    else:
705
      for name, pip, sip in what[constants.NV_NODENETTEST]:
706
        fail = []
707
        if not netutils.TcpPing(pip, port, source=my_pip):
708
          fail.append("primary")
709
        if sip != pip:
710
          if not netutils.TcpPing(sip, port, source=my_sip):
711
            fail.append("secondary")
712
        if fail:
713
          tmp[name] = ("failure using the %s interface(s)" %
714
                       " and ".join(fail))
715

    
716
  if constants.NV_MASTERIP in what:
717
    # FIXME: add checks on incoming data structures (here and in the
718
    # rest of the function)
719
    master_name, master_ip = what[constants.NV_MASTERIP]
720
    if master_name == my_name:
721
      source = constants.IP4_ADDRESS_LOCALHOST
722
    else:
723
      source = None
724
    result[constants.NV_MASTERIP] = netutils.TcpPing(master_ip, port,
725
                                                     source=source)
726

    
727
  if constants.NV_USERSCRIPTS in what:
728
    result[constants.NV_USERSCRIPTS] = \
729
      [script for script in what[constants.NV_USERSCRIPTS]
730
       if not utils.IsExecutable(script)]
731

    
732
  if constants.NV_OOB_PATHS in what:
733
    result[constants.NV_OOB_PATHS] = tmp = []
734
    for path in what[constants.NV_OOB_PATHS]:
735
      try:
736
        st = os.stat(path)
737
      except OSError, err:
738
        tmp.append("error stating out of band helper: %s" % err)
739
      else:
740
        if stat.S_ISREG(st.st_mode):
741
          if stat.S_IMODE(st.st_mode) & stat.S_IXUSR:
742
            tmp.append(None)
743
          else:
744
            tmp.append("out of band helper %s is not executable" % path)
745
        else:
746
          tmp.append("out of band helper %s is not a file" % path)
747

    
748
  if constants.NV_LVLIST in what and vm_capable:
749
    try:
750
      val = GetVolumeList(utils.ListVolumeGroups().keys())
751
    except RPCFail, err:
752
      val = str(err)
753
    result[constants.NV_LVLIST] = val
754

    
755
  if constants.NV_INSTANCELIST in what and vm_capable:
756
    # GetInstanceList can fail
757
    try:
758
      val = GetInstanceList(what[constants.NV_INSTANCELIST])
759
    except RPCFail, err:
760
      val = str(err)
761
    result[constants.NV_INSTANCELIST] = val
762

    
763
  if constants.NV_VGLIST in what and vm_capable:
764
    result[constants.NV_VGLIST] = utils.ListVolumeGroups()
765

    
766
  if constants.NV_PVLIST in what and vm_capable:
767
    val = bdev.LogicalVolume.GetPVInfo(what[constants.NV_PVLIST],
768
                                       filter_allocatable=False)
769
    result[constants.NV_PVLIST] = map(objects.LvmPvInfo.ToDict, val)
770

    
771
  if constants.NV_VERSION in what:
772
    result[constants.NV_VERSION] = (constants.PROTOCOL_VERSION,
773
                                    constants.RELEASE_VERSION)
774

    
775
  if constants.NV_HVINFO in what and vm_capable:
776
    hyper = hypervisor.GetHypervisor(what[constants.NV_HVINFO])
777
    result[constants.NV_HVINFO] = hyper.GetNodeInfo()
778

    
779
  if constants.NV_DRBDLIST in what and vm_capable:
780
    try:
781
      used_minors = bdev.DRBD8.GetUsedDevs().keys()
782
    except errors.BlockDeviceError, err:
783
      logging.warning("Can't get used minors list", exc_info=True)
784
      used_minors = str(err)
785
    result[constants.NV_DRBDLIST] = used_minors
786

    
787
  if constants.NV_DRBDHELPER in what and vm_capable:
788
    status = True
789
    try:
790
      payload = bdev.BaseDRBD.GetUsermodeHelper()
791
    except errors.BlockDeviceError, err:
792
      logging.error("Can't get DRBD usermode helper: %s", str(err))
793
      status = False
794
      payload = str(err)
795
    result[constants.NV_DRBDHELPER] = (status, payload)
796

    
797
  if constants.NV_NODESETUP in what:
798
    result[constants.NV_NODESETUP] = tmpr = []
799
    if not os.path.isdir("/sys/block") or not os.path.isdir("/sys/class/net"):
800
      tmpr.append("The sysfs filesytem doesn't seem to be mounted"
801
                  " under /sys, missing required directories /sys/block"
802
                  " and /sys/class/net")
803
    if (not os.path.isdir("/proc/sys") or
804
        not os.path.isfile("/proc/sysrq-trigger")):
805
      tmpr.append("The procfs filesystem doesn't seem to be mounted"
806
                  " under /proc, missing required directory /proc/sys and"
807
                  " the file /proc/sysrq-trigger")
808

    
809
  if constants.NV_TIME in what:
810
    result[constants.NV_TIME] = utils.SplitTime(time.time())
811

    
812
  if constants.NV_OSLIST in what and vm_capable:
813
    result[constants.NV_OSLIST] = DiagnoseOS()
814

    
815
  if constants.NV_BRIDGES in what and vm_capable:
816
    result[constants.NV_BRIDGES] = [bridge
817
                                    for bridge in what[constants.NV_BRIDGES]
818
                                    if not utils.BridgeExists(bridge)]
819

    
820
  if what.get(constants.NV_FILE_STORAGE_PATHS) == my_name:
821
    result[constants.NV_FILE_STORAGE_PATHS] = \
822
      bdev.ComputeWrongFileStoragePaths()
823

    
824
  return result
825

    
826

    
827
def GetBlockDevSizes(devices):
828
  """Return the size of the given block devices
829

830
  @type devices: list
831
  @param devices: list of block device nodes to query
832
  @rtype: dict
833
  @return:
834
    dictionary of all block devices under /dev (key). The value is their
835
    size in MiB.
836

837
    {'/dev/disk/by-uuid/123456-12321231-312312-312': 124}
838

839
  """
840
  DEV_PREFIX = "/dev/"
841
  blockdevs = {}
842

    
843
  for devpath in devices:
844
    if not utils.IsBelowDir(DEV_PREFIX, devpath):
845
      continue
846

    
847
    try:
848
      st = os.stat(devpath)
849
    except EnvironmentError, err:
850
      logging.warning("Error stat()'ing device %s: %s", devpath, str(err))
851
      continue
852

    
853
    if stat.S_ISBLK(st.st_mode):
854
      result = utils.RunCmd(["blockdev", "--getsize64", devpath])
855
      if result.failed:
856
        # We don't want to fail, just do not list this device as available
857
        logging.warning("Cannot get size for block device %s", devpath)
858
        continue
859

    
860
      size = int(result.stdout) / (1024 * 1024)
861
      blockdevs[devpath] = size
862
  return blockdevs
863

    
864

    
865
def GetVolumeList(vg_names):
866
  """Compute list of logical volumes and their size.
867

868
  @type vg_names: list
869
  @param vg_names: the volume groups whose LVs we should list, or
870
      empty for all volume groups
871
  @rtype: dict
872
  @return:
873
      dictionary of all partions (key) with value being a tuple of
874
      their size (in MiB), inactive and online status::
875

876
        {'xenvg/test1': ('20.06', True, True)}
877

878
      in case of errors, a string is returned with the error
879
      details.
880

881
  """
882
  lvs = {}
883
  sep = "|"
884
  if not vg_names:
885
    vg_names = []
886
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
887
                         "--separator=%s" % sep,
888
                         "-ovg_name,lv_name,lv_size,lv_attr"] + vg_names)
889
  if result.failed:
890
    _Fail("Failed to list logical volumes, lvs output: %s", result.output)
891

    
892
  for line in result.stdout.splitlines():
893
    line = line.strip()
894
    match = _LVSLINE_REGEX.match(line)
895
    if not match:
896
      logging.error("Invalid line returned from lvs output: '%s'", line)
897
      continue
898
    vg_name, name, size, attr = match.groups()
899
    inactive = attr[4] == "-"
900
    online = attr[5] == "o"
901
    virtual = attr[0] == "v"
902
    if virtual:
903
      # we don't want to report such volumes as existing, since they
904
      # don't really hold data
905
      continue
906
    lvs[vg_name + "/" + name] = (size, inactive, online)
907

    
908
  return lvs
909

    
910

    
911
def ListVolumeGroups():
912
  """List the volume groups and their size.
913

914
  @rtype: dict
915
  @return: dictionary with keys volume name and values the
916
      size of the volume
917

918
  """
919
  return utils.ListVolumeGroups()
920

    
921

    
922
def NodeVolumes():
923
  """List all volumes on this node.
924

925
  @rtype: list
926
  @return:
927
    A list of dictionaries, each having four keys:
928
      - name: the logical volume name,
929
      - size: the size of the logical volume
930
      - dev: the physical device on which the LV lives
931
      - vg: the volume group to which it belongs
932

933
    In case of errors, we return an empty list and log the
934
    error.
935

936
    Note that since a logical volume can live on multiple physical
937
    volumes, the resulting list might include a logical volume
938
    multiple times.
939

940
  """
941
  result = utils.RunCmd(["lvs", "--noheadings", "--units=m", "--nosuffix",
942
                         "--separator=|",
943
                         "--options=lv_name,lv_size,devices,vg_name"])
944
  if result.failed:
945
    _Fail("Failed to list logical volumes, lvs output: %s",
946
          result.output)
947

    
948
  def parse_dev(dev):
949
    return dev.split("(")[0]
950

    
951
  def handle_dev(dev):
952
    return [parse_dev(x) for x in dev.split(",")]
953

    
954
  def map_line(line):
955
    line = [v.strip() for v in line]
956
    return [{"name": line[0], "size": line[1],
957
             "dev": dev, "vg": line[3]} for dev in handle_dev(line[2])]
958

    
959
  all_devs = []
960
  for line in result.stdout.splitlines():
961
    if line.count("|") >= 3:
962
      all_devs.extend(map_line(line.split("|")))
963
    else:
964
      logging.warning("Strange line in the output from lvs: '%s'", line)
965
  return all_devs
966

    
967

    
968
def BridgesExist(bridges_list):
969
  """Check if a list of bridges exist on the current node.
970

971
  @rtype: boolean
972
  @return: C{True} if all of them exist, C{False} otherwise
973

974
  """
975
  missing = []
976
  for bridge in bridges_list:
977
    if not utils.BridgeExists(bridge):
978
      missing.append(bridge)
979

    
980
  if missing:
981
    _Fail("Missing bridges %s", utils.CommaJoin(missing))
982

    
983

    
984
def GetInstanceList(hypervisor_list):
985
  """Provides a list of instances.
986

987
  @type hypervisor_list: list
988
  @param hypervisor_list: the list of hypervisors to query information
989

990
  @rtype: list
991
  @return: a list of all running instances on the current node
992
    - instance1.example.com
993
    - instance2.example.com
994

995
  """
996
  results = []
997
  for hname in hypervisor_list:
998
    try:
999
      names = hypervisor.GetHypervisor(hname).ListInstances()
1000
      results.extend(names)
1001
    except errors.HypervisorError, err:
1002
      _Fail("Error enumerating instances (hypervisor %s): %s",
1003
            hname, err, exc=True)
1004

    
1005
  return results
1006

    
1007

    
1008
def GetInstanceInfo(instance, hname):
1009
  """Gives back the information about an instance as a dictionary.
1010

1011
  @type instance: string
1012
  @param instance: the instance name
1013
  @type hname: string
1014
  @param hname: the hypervisor type of the instance
1015

1016
  @rtype: dict
1017
  @return: dictionary with the following keys:
1018
      - memory: memory size of instance (int)
1019
      - state: xen state of instance (string)
1020
      - time: cpu time of instance (float)
1021
      - vcpus: the number of vcpus (int)
1022

1023
  """
1024
  output = {}
1025

    
1026
  iinfo = hypervisor.GetHypervisor(hname).GetInstanceInfo(instance)
1027
  if iinfo is not None:
1028
    output["memory"] = iinfo[2]
1029
    output["vcpus"] = iinfo[3]
1030
    output["state"] = iinfo[4]
1031
    output["time"] = iinfo[5]
1032

    
1033
  return output
1034

    
1035

    
1036
def GetInstanceMigratable(instance):
1037
  """Gives whether an instance can be migrated.
1038

1039
  @type instance: L{objects.Instance}
1040
  @param instance: object representing the instance to be checked.
1041

1042
  @rtype: tuple
1043
  @return: tuple of (result, description) where:
1044
      - result: whether the instance can be migrated or not
1045
      - description: a description of the issue, if relevant
1046

1047
  """
1048
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1049
  iname = instance.name
1050
  if iname not in hyper.ListInstances():
1051
    _Fail("Instance %s is not running", iname)
1052

    
1053
  for idx in range(len(instance.disks)):
1054
    link_name = _GetBlockDevSymlinkPath(iname, idx)
1055
    if not os.path.islink(link_name):
1056
      logging.warning("Instance %s is missing symlink %s for disk %d",
1057
                      iname, link_name, idx)
1058

    
1059

    
1060
def GetAllInstancesInfo(hypervisor_list):
1061
  """Gather data about all instances.
1062

1063
  This is the equivalent of L{GetInstanceInfo}, except that it
1064
  computes data for all instances at once, thus being faster if one
1065
  needs data about more than one instance.
1066

1067
  @type hypervisor_list: list
1068
  @param hypervisor_list: list of hypervisors to query for instance data
1069

1070
  @rtype: dict
1071
  @return: dictionary of instance: data, with data having the following keys:
1072
      - memory: memory size of instance (int)
1073
      - state: xen state of instance (string)
1074
      - time: cpu time of instance (float)
1075
      - vcpus: the number of vcpus
1076

1077
  """
1078
  output = {}
1079

    
1080
  for hname in hypervisor_list:
1081
    iinfo = hypervisor.GetHypervisor(hname).GetAllInstancesInfo()
1082
    if iinfo:
1083
      for name, _, memory, vcpus, state, times in iinfo:
1084
        value = {
1085
          "memory": memory,
1086
          "vcpus": vcpus,
1087
          "state": state,
1088
          "time": times,
1089
          }
1090
        if name in output:
1091
          # we only check static parameters, like memory and vcpus,
1092
          # and not state and time which can change between the
1093
          # invocations of the different hypervisors
1094
          for key in "memory", "vcpus":
1095
            if value[key] != output[name][key]:
1096
              _Fail("Instance %s is running twice"
1097
                    " with different parameters", name)
1098
        output[name] = value
1099

    
1100
  return output
1101

    
1102

    
1103
def _InstanceLogName(kind, os_name, instance, component):
1104
  """Compute the OS log filename for a given instance and operation.
1105

1106
  The instance name and os name are passed in as strings since not all
1107
  operations have these as part of an instance object.
1108

1109
  @type kind: string
1110
  @param kind: the operation type (e.g. add, import, etc.)
1111
  @type os_name: string
1112
  @param os_name: the os name
1113
  @type instance: string
1114
  @param instance: the name of the instance being imported/added/etc.
1115
  @type component: string or None
1116
  @param component: the name of the component of the instance being
1117
      transferred
1118

1119
  """
1120
  # TODO: Use tempfile.mkstemp to create unique filename
1121
  if component:
1122
    assert "/" not in component
1123
    c_msg = "-%s" % component
1124
  else:
1125
    c_msg = ""
1126
  base = ("%s-%s-%s%s-%s.log" %
1127
          (kind, os_name, instance, c_msg, utils.TimestampForFilename()))
1128
  return utils.PathJoin(pathutils.LOG_OS_DIR, base)
1129

    
1130

    
1131
def InstanceOsAdd(instance, reinstall, debug):
1132
  """Add an OS to an instance.
1133

1134
  @type instance: L{objects.Instance}
1135
  @param instance: Instance whose OS is to be installed
1136
  @type reinstall: boolean
1137
  @param reinstall: whether this is an instance reinstall
1138
  @type debug: integer
1139
  @param debug: debug level, passed to the OS scripts
1140
  @rtype: None
1141

1142
  """
1143
  inst_os = OSFromDisk(instance.os)
1144

    
1145
  create_env = OSEnvironment(instance, inst_os, debug)
1146
  if reinstall:
1147
    create_env["INSTANCE_REINSTALL"] = "1"
1148

    
1149
  logfile = _InstanceLogName("add", instance.os, instance.name, None)
1150

    
1151
  result = utils.RunCmd([inst_os.create_script], env=create_env,
1152
                        cwd=inst_os.path, output=logfile, reset_env=True)
1153
  if result.failed:
1154
    logging.error("os create command '%s' returned error: %s, logfile: %s,"
1155
                  " output: %s", result.cmd, result.fail_reason, logfile,
1156
                  result.output)
1157
    lines = [utils.SafeEncode(val)
1158
             for val in utils.TailFile(logfile, lines=20)]
1159
    _Fail("OS create script failed (%s), last lines in the"
1160
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1161

    
1162

    
1163
def RunRenameInstance(instance, old_name, debug):
1164
  """Run the OS rename script for an instance.
1165

1166
  @type instance: L{objects.Instance}
1167
  @param instance: Instance whose OS is to be installed
1168
  @type old_name: string
1169
  @param old_name: previous instance name
1170
  @type debug: integer
1171
  @param debug: debug level, passed to the OS scripts
1172
  @rtype: boolean
1173
  @return: the success of the operation
1174

1175
  """
1176
  inst_os = OSFromDisk(instance.os)
1177

    
1178
  rename_env = OSEnvironment(instance, inst_os, debug)
1179
  rename_env["OLD_INSTANCE_NAME"] = old_name
1180

    
1181
  logfile = _InstanceLogName("rename", instance.os,
1182
                             "%s-%s" % (old_name, instance.name), None)
1183

    
1184
  result = utils.RunCmd([inst_os.rename_script], env=rename_env,
1185
                        cwd=inst_os.path, output=logfile, reset_env=True)
1186

    
1187
  if result.failed:
1188
    logging.error("os create command '%s' returned error: %s output: %s",
1189
                  result.cmd, result.fail_reason, result.output)
1190
    lines = [utils.SafeEncode(val)
1191
             for val in utils.TailFile(logfile, lines=20)]
1192
    _Fail("OS rename script failed (%s), last lines in the"
1193
          " log file:\n%s", result.fail_reason, "\n".join(lines), log=False)
1194

    
1195

    
1196
def _GetBlockDevSymlinkPath(instance_name, idx):
1197
  return utils.PathJoin(pathutils.DISK_LINKS_DIR, "%s%s%d" %
1198
                        (instance_name, constants.DISK_SEPARATOR, idx))
1199

    
1200

    
1201
def _SymlinkBlockDev(instance_name, device_path, idx):
1202
  """Set up symlinks to a instance's block device.
1203

1204
  This is an auxiliary function run when an instance is start (on the primary
1205
  node) or when an instance is migrated (on the target node).
1206

1207

1208
  @param instance_name: the name of the target instance
1209
  @param device_path: path of the physical block device, on the node
1210
  @param idx: the disk index
1211
  @return: absolute path to the disk's symlink
1212

1213
  """
1214
  link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1215
  try:
1216
    os.symlink(device_path, link_name)
1217
  except OSError, err:
1218
    if err.errno == errno.EEXIST:
1219
      if (not os.path.islink(link_name) or
1220
          os.readlink(link_name) != device_path):
1221
        os.remove(link_name)
1222
        os.symlink(device_path, link_name)
1223
    else:
1224
      raise
1225

    
1226
  return link_name
1227

    
1228

    
1229
def _RemoveBlockDevLinks(instance_name, disks):
1230
  """Remove the block device symlinks belonging to the given instance.
1231

1232
  """
1233
  for idx, _ in enumerate(disks):
1234
    link_name = _GetBlockDevSymlinkPath(instance_name, idx)
1235
    if os.path.islink(link_name):
1236
      try:
1237
        os.remove(link_name)
1238
      except OSError:
1239
        logging.exception("Can't remove symlink '%s'", link_name)
1240

    
1241

    
1242
def _GatherAndLinkBlockDevs(instance):
1243
  """Set up an instance's block device(s).
1244

1245
  This is run on the primary node at instance startup. The block
1246
  devices must be already assembled.
1247

1248
  @type instance: L{objects.Instance}
1249
  @param instance: the instance whose disks we shoul assemble
1250
  @rtype: list
1251
  @return: list of (disk_object, device_path)
1252

1253
  """
1254
  block_devices = []
1255
  for idx, disk in enumerate(instance.disks):
1256
    device = _RecursiveFindBD(disk)
1257
    if device is None:
1258
      raise errors.BlockDeviceError("Block device '%s' is not set up." %
1259
                                    str(disk))
1260
    device.Open()
1261
    try:
1262
      link_name = _SymlinkBlockDev(instance.name, device.dev_path, idx)
1263
    except OSError, e:
1264
      raise errors.BlockDeviceError("Cannot create block device symlink: %s" %
1265
                                    e.strerror)
1266

    
1267
    block_devices.append((disk, link_name))
1268

    
1269
  return block_devices
1270

    
1271

    
1272
def StartInstance(instance, startup_paused):
1273
  """Start an instance.
1274

1275
  @type instance: L{objects.Instance}
1276
  @param instance: the instance object
1277
  @type startup_paused: bool
1278
  @param instance: pause instance at startup?
1279
  @rtype: None
1280

1281
  """
1282
  running_instances = GetInstanceList([instance.hypervisor])
1283

    
1284
  if instance.name in running_instances:
1285
    logging.info("Instance %s already running, not starting", instance.name)
1286
    return
1287

    
1288
  try:
1289
    block_devices = _GatherAndLinkBlockDevs(instance)
1290
    hyper = hypervisor.GetHypervisor(instance.hypervisor)
1291
    hyper.StartInstance(instance, block_devices, startup_paused)
1292
  except errors.BlockDeviceError, err:
1293
    _Fail("Block device error: %s", err, exc=True)
1294
  except errors.HypervisorError, err:
1295
    _RemoveBlockDevLinks(instance.name, instance.disks)
1296
    _Fail("Hypervisor error: %s", err, exc=True)
1297

    
1298

    
1299
def InstanceShutdown(instance, timeout):
1300
  """Shut an instance down.
1301

1302
  @note: this functions uses polling with a hardcoded timeout.
1303

1304
  @type instance: L{objects.Instance}
1305
  @param instance: the instance object
1306
  @type timeout: integer
1307
  @param timeout: maximum timeout for soft shutdown
1308
  @rtype: None
1309

1310
  """
1311
  hv_name = instance.hypervisor
1312
  hyper = hypervisor.GetHypervisor(hv_name)
1313
  iname = instance.name
1314

    
1315
  if instance.name not in hyper.ListInstances():
1316
    logging.info("Instance %s not running, doing nothing", iname)
1317
    return
1318

    
1319
  class _TryShutdown:
1320
    def __init__(self):
1321
      self.tried_once = False
1322

    
1323
    def __call__(self):
1324
      if iname not in hyper.ListInstances():
1325
        return
1326

    
1327
      try:
1328
        hyper.StopInstance(instance, retry=self.tried_once)
1329
      except errors.HypervisorError, err:
1330
        if iname not in hyper.ListInstances():
1331
          # if the instance is no longer existing, consider this a
1332
          # success and go to cleanup
1333
          return
1334

    
1335
        _Fail("Failed to stop instance %s: %s", iname, err)
1336

    
1337
      self.tried_once = True
1338

    
1339
      raise utils.RetryAgain()
1340

    
1341
  try:
1342
    utils.Retry(_TryShutdown(), 5, timeout)
1343
  except utils.RetryTimeout:
1344
    # the shutdown did not succeed
1345
    logging.error("Shutdown of '%s' unsuccessful, forcing", iname)
1346

    
1347
    try:
1348
      hyper.StopInstance(instance, force=True)
1349
    except errors.HypervisorError, err:
1350
      if iname in hyper.ListInstances():
1351
        # only raise an error if the instance still exists, otherwise
1352
        # the error could simply be "instance ... unknown"!
1353
        _Fail("Failed to force stop instance %s: %s", iname, err)
1354

    
1355
    time.sleep(1)
1356

    
1357
    if iname in hyper.ListInstances():
1358
      _Fail("Could not shutdown instance %s even by destroy", iname)
1359

    
1360
  try:
1361
    hyper.CleanupInstance(instance.name)
1362
  except errors.HypervisorError, err:
1363
    logging.warning("Failed to execute post-shutdown cleanup step: %s", err)
1364

    
1365
  _RemoveBlockDevLinks(iname, instance.disks)
1366

    
1367

    
1368
def InstanceReboot(instance, reboot_type, shutdown_timeout):
1369
  """Reboot an instance.
1370

1371
  @type instance: L{objects.Instance}
1372
  @param instance: the instance object to reboot
1373
  @type reboot_type: str
1374
  @param reboot_type: the type of reboot, one the following
1375
    constants:
1376
      - L{constants.INSTANCE_REBOOT_SOFT}: only reboot the
1377
        instance OS, do not recreate the VM
1378
      - L{constants.INSTANCE_REBOOT_HARD}: tear down and
1379
        restart the VM (at the hypervisor level)
1380
      - the other reboot type (L{constants.INSTANCE_REBOOT_FULL}) is
1381
        not accepted here, since that mode is handled differently, in
1382
        cmdlib, and translates into full stop and start of the
1383
        instance (instead of a call_instance_reboot RPC)
1384
  @type shutdown_timeout: integer
1385
  @param shutdown_timeout: maximum timeout for soft shutdown
1386
  @rtype: None
1387

1388
  """
1389
  running_instances = GetInstanceList([instance.hypervisor])
1390

    
1391
  if instance.name not in running_instances:
1392
    _Fail("Cannot reboot instance %s that is not running", instance.name)
1393

    
1394
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1395
  if reboot_type == constants.INSTANCE_REBOOT_SOFT:
1396
    try:
1397
      hyper.RebootInstance(instance)
1398
    except errors.HypervisorError, err:
1399
      _Fail("Failed to soft reboot instance %s: %s", instance.name, err)
1400
  elif reboot_type == constants.INSTANCE_REBOOT_HARD:
1401
    try:
1402
      InstanceShutdown(instance, shutdown_timeout)
1403
      return StartInstance(instance, False)
1404
    except errors.HypervisorError, err:
1405
      _Fail("Failed to hard reboot instance %s: %s", instance.name, err)
1406
  else:
1407
    _Fail("Invalid reboot_type received: %s", reboot_type)
1408

    
1409

    
1410
def InstanceBalloonMemory(instance, memory):
1411
  """Resize an instance's memory.
1412

1413
  @type instance: L{objects.Instance}
1414
  @param instance: the instance object
1415
  @type memory: int
1416
  @param memory: new memory amount in MB
1417
  @rtype: None
1418

1419
  """
1420
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1421
  running = hyper.ListInstances()
1422
  if instance.name not in running:
1423
    logging.info("Instance %s is not running, cannot balloon", instance.name)
1424
    return
1425
  try:
1426
    hyper.BalloonInstanceMemory(instance, memory)
1427
  except errors.HypervisorError, err:
1428
    _Fail("Failed to balloon instance memory: %s", err, exc=True)
1429

    
1430

    
1431
def MigrationInfo(instance):
1432
  """Gather information about an instance to be migrated.
1433

1434
  @type instance: L{objects.Instance}
1435
  @param instance: the instance definition
1436

1437
  """
1438
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1439
  try:
1440
    info = hyper.MigrationInfo(instance)
1441
  except errors.HypervisorError, err:
1442
    _Fail("Failed to fetch migration information: %s", err, exc=True)
1443
  return info
1444

    
1445

    
1446
def AcceptInstance(instance, info, target):
1447
  """Prepare the node to accept an instance.
1448

1449
  @type instance: L{objects.Instance}
1450
  @param instance: the instance definition
1451
  @type info: string/data (opaque)
1452
  @param info: migration information, from the source node
1453
  @type target: string
1454
  @param target: target host (usually ip), on this node
1455

1456
  """
1457
  # TODO: why is this required only for DTS_EXT_MIRROR?
1458
  if instance.disk_template in constants.DTS_EXT_MIRROR:
1459
    # Create the symlinks, as the disks are not active
1460
    # in any way
1461
    try:
1462
      _GatherAndLinkBlockDevs(instance)
1463
    except errors.BlockDeviceError, err:
1464
      _Fail("Block device error: %s", err, exc=True)
1465

    
1466
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1467
  try:
1468
    hyper.AcceptInstance(instance, info, target)
1469
  except errors.HypervisorError, err:
1470
    if instance.disk_template in constants.DTS_EXT_MIRROR:
1471
      _RemoveBlockDevLinks(instance.name, instance.disks)
1472
    _Fail("Failed to accept instance: %s", err, exc=True)
1473

    
1474

    
1475
def FinalizeMigrationDst(instance, info, success):
1476
  """Finalize any preparation to accept an instance.
1477

1478
  @type instance: L{objects.Instance}
1479
  @param instance: the instance definition
1480
  @type info: string/data (opaque)
1481
  @param info: migration information, from the source node
1482
  @type success: boolean
1483
  @param success: whether the migration was a success or a failure
1484

1485
  """
1486
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1487
  try:
1488
    hyper.FinalizeMigrationDst(instance, info, success)
1489
  except errors.HypervisorError, err:
1490
    _Fail("Failed to finalize migration on the target node: %s", err, exc=True)
1491

    
1492

    
1493
def MigrateInstance(instance, target, live):
1494
  """Migrates an instance to another node.
1495

1496
  @type instance: L{objects.Instance}
1497
  @param instance: the instance definition
1498
  @type target: string
1499
  @param target: the target node name
1500
  @type live: boolean
1501
  @param live: whether the migration should be done live or not (the
1502
      interpretation of this parameter is left to the hypervisor)
1503
  @raise RPCFail: if migration fails for some reason
1504

1505
  """
1506
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1507

    
1508
  try:
1509
    hyper.MigrateInstance(instance, target, live)
1510
  except errors.HypervisorError, err:
1511
    _Fail("Failed to migrate instance: %s", err, exc=True)
1512

    
1513

    
1514
def FinalizeMigrationSource(instance, success, live):
1515
  """Finalize the instance migration on the source node.
1516

1517
  @type instance: L{objects.Instance}
1518
  @param instance: the instance definition of the migrated instance
1519
  @type success: bool
1520
  @param success: whether the migration succeeded or not
1521
  @type live: bool
1522
  @param live: whether the user requested a live migration or not
1523
  @raise RPCFail: If the execution fails for some reason
1524

1525
  """
1526
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1527

    
1528
  try:
1529
    hyper.FinalizeMigrationSource(instance, success, live)
1530
  except Exception, err:  # pylint: disable=W0703
1531
    _Fail("Failed to finalize the migration on the source node: %s", err,
1532
          exc=True)
1533

    
1534

    
1535
def GetMigrationStatus(instance):
1536
  """Get the migration status
1537

1538
  @type instance: L{objects.Instance}
1539
  @param instance: the instance that is being migrated
1540
  @rtype: L{objects.MigrationStatus}
1541
  @return: the status of the current migration (one of
1542
           L{constants.HV_MIGRATION_VALID_STATUSES}), plus any additional
1543
           progress info that can be retrieved from the hypervisor
1544
  @raise RPCFail: If the migration status cannot be retrieved
1545

1546
  """
1547
  hyper = hypervisor.GetHypervisor(instance.hypervisor)
1548
  try:
1549
    return hyper.GetMigrationStatus(instance)
1550
  except Exception, err:  # pylint: disable=W0703
1551
    _Fail("Failed to get migration status: %s", err, exc=True)
1552

    
1553

    
1554
def BlockdevCreate(disk, size, owner, on_primary, info, excl_stor):
1555
  """Creates a block device for an instance.
1556

1557
  @type disk: L{objects.Disk}
1558
  @param disk: the object describing the disk we should create
1559
  @type size: int
1560
  @param size: the size of the physical underlying device, in MiB
1561
  @type owner: str
1562
  @param owner: the name of the instance for which disk is created,
1563
      used for device cache data
1564
  @type on_primary: boolean
1565
  @param on_primary:  indicates if it is the primary node or not
1566
  @type info: string
1567
  @param info: string that will be sent to the physical device
1568
      creation, used for example to set (LVM) tags on LVs
1569
  @type excl_stor: boolean
1570
  @param excl_stor: Whether exclusive_storage is active
1571

1572
  @return: the new unique_id of the device (this can sometime be
1573
      computed only after creation), or None. On secondary nodes,
1574
      it's not required to return anything.
1575

1576
  """
1577
  # TODO: remove the obsolete "size" argument
1578
  # pylint: disable=W0613
1579
  clist = []
1580
  if disk.children:
1581
    for child in disk.children:
1582
      try:
1583
        crdev = _RecursiveAssembleBD(child, owner, on_primary)
1584
      except errors.BlockDeviceError, err:
1585
        _Fail("Can't assemble device %s: %s", child, err)
1586
      if on_primary or disk.AssembleOnSecondary():
1587
        # we need the children open in case the device itself has to
1588
        # be assembled
1589
        try:
1590
          # pylint: disable=E1103
1591
          crdev.Open()
1592
        except errors.BlockDeviceError, err:
1593
          _Fail("Can't make child '%s' read-write: %s", child, err)
1594
      clist.append(crdev)
1595

    
1596
  try:
1597
    device = bdev.Create(disk, clist, excl_stor)
1598
  except errors.BlockDeviceError, err:
1599
    _Fail("Can't create block device: %s", err)
1600

    
1601
  if on_primary or disk.AssembleOnSecondary():
1602
    try:
1603
      device.Assemble()
1604
    except errors.BlockDeviceError, err:
1605
      _Fail("Can't assemble device after creation, unusual event: %s", err)
1606
    if on_primary or disk.OpenOnSecondary():
1607
      try:
1608
        device.Open(force=True)
1609
      except errors.BlockDeviceError, err:
1610
        _Fail("Can't make device r/w after creation, unusual event: %s", err)
1611
    DevCacheManager.UpdateCache(device.dev_path, owner,
1612
                                on_primary, disk.iv_name)
1613

    
1614
  device.SetInfo(info)
1615

    
1616
  return device.unique_id
1617

    
1618

    
1619
def _WipeDevice(path, offset, size):
1620
  """This function actually wipes the device.
1621

1622
  @param path: The path to the device to wipe
1623
  @param offset: The offset in MiB in the file
1624
  @param size: The size in MiB to write
1625

1626
  """
1627
  # Internal sizes are always in Mebibytes; if the following "dd" command
1628
  # should use a different block size the offset and size given to this
1629
  # function must be adjusted accordingly before being passed to "dd".
1630
  block_size = 1024 * 1024
1631

    
1632
  cmd = [constants.DD_CMD, "if=/dev/zero", "seek=%d" % offset,
1633
         "bs=%s" % block_size, "oflag=direct", "of=%s" % path,
1634
         "count=%d" % size]
1635
  result = utils.RunCmd(cmd)
1636

    
1637
  if result.failed:
1638
    _Fail("Wipe command '%s' exited with error: %s; output: %s", result.cmd,
1639
          result.fail_reason, result.output)
1640

    
1641

    
1642
def BlockdevWipe(disk, offset, size):
1643
  """Wipes a block device.
1644

1645
  @type disk: L{objects.Disk}
1646
  @param disk: the disk object we want to wipe
1647
  @type offset: int
1648
  @param offset: The offset in MiB in the file
1649
  @type size: int
1650
  @param size: The size in MiB to write
1651

1652
  """
1653
  try:
1654
    rdev = _RecursiveFindBD(disk)
1655
  except errors.BlockDeviceError:
1656
    rdev = None
1657

    
1658
  if not rdev:
1659
    _Fail("Cannot execute wipe for device %s: device not found", disk.iv_name)
1660

    
1661
  # Do cross verify some of the parameters
1662
  if offset < 0:
1663
    _Fail("Negative offset")
1664
  if size < 0:
1665
    _Fail("Negative size")
1666
  if offset > rdev.size:
1667
    _Fail("Offset is bigger than device size")
1668
  if (offset + size) > rdev.size:
1669
    _Fail("The provided offset and size to wipe is bigger than device size")
1670

    
1671
  _WipeDevice(rdev.dev_path, offset, size)
1672

    
1673

    
1674
def BlockdevPauseResumeSync(disks, pause):
1675
  """Pause or resume the sync of the block device.
1676

1677
  @type disks: list of L{objects.Disk}
1678
  @param disks: the disks object we want to pause/resume
1679
  @type pause: bool
1680
  @param pause: Wheater to pause or resume
1681

1682
  """
1683
  success = []
1684
  for disk in disks:
1685
    try:
1686
      rdev = _RecursiveFindBD(disk)
1687
    except errors.BlockDeviceError:
1688
      rdev = None
1689

    
1690
    if not rdev:
1691
      success.append((False, ("Cannot change sync for device %s:"
1692
                              " device not found" % disk.iv_name)))
1693
      continue
1694

    
1695
    result = rdev.PauseResumeSync(pause)
1696

    
1697
    if result:
1698
      success.append((result, None))
1699
    else:
1700
      if pause:
1701
        msg = "Pause"
1702
      else:
1703
        msg = "Resume"
1704
      success.append((result, "%s for device %s failed" % (msg, disk.iv_name)))
1705

    
1706
  return success
1707

    
1708

    
1709
def BlockdevRemove(disk):
1710
  """Remove a block device.
1711

1712
  @note: This is intended to be called recursively.
1713

1714
  @type disk: L{objects.Disk}
1715
  @param disk: the disk object we should remove
1716
  @rtype: boolean
1717
  @return: the success of the operation
1718

1719
  """
1720
  msgs = []
1721
  try:
1722
    rdev = _RecursiveFindBD(disk)
1723
  except errors.BlockDeviceError, err:
1724
    # probably can't attach
1725
    logging.info("Can't attach to device %s in remove", disk)
1726
    rdev = None
1727
  if rdev is not None:
1728
    r_path = rdev.dev_path
1729
    try:
1730
      rdev.Remove()
1731
    except errors.BlockDeviceError, err:
1732
      msgs.append(str(err))
1733
    if not msgs:
1734
      DevCacheManager.RemoveCache(r_path)
1735

    
1736
  if disk.children:
1737
    for child in disk.children:
1738
      try:
1739
        BlockdevRemove(child)
1740
      except RPCFail, err:
1741
        msgs.append(str(err))
1742

    
1743
  if msgs:
1744
    _Fail("; ".join(msgs))
1745

    
1746

    
1747
def _RecursiveAssembleBD(disk, owner, as_primary):
1748
  """Activate a block device for an instance.
1749

1750
  This is run on the primary and secondary nodes for an instance.
1751

1752
  @note: this function is called recursively.
1753

1754
  @type disk: L{objects.Disk}
1755
  @param disk: the disk we try to assemble
1756
  @type owner: str
1757
  @param owner: the name of the instance which owns the disk
1758
  @type as_primary: boolean
1759
  @param as_primary: if we should make the block device
1760
      read/write
1761

1762
  @return: the assembled device or None (in case no device
1763
      was assembled)
1764
  @raise errors.BlockDeviceError: in case there is an error
1765
      during the activation of the children or the device
1766
      itself
1767

1768
  """
1769
  children = []
1770
  if disk.children:
1771
    mcn = disk.ChildrenNeeded()
1772
    if mcn == -1:
1773
      mcn = 0 # max number of Nones allowed
1774
    else:
1775
      mcn = len(disk.children) - mcn # max number of Nones
1776
    for chld_disk in disk.children:
1777
      try:
1778
        cdev = _RecursiveAssembleBD(chld_disk, owner, as_primary)
1779
      except errors.BlockDeviceError, err:
1780
        if children.count(None) >= mcn:
1781
          raise
1782
        cdev = None
1783
        logging.error("Error in child activation (but continuing): %s",
1784
                      str(err))
1785
      children.append(cdev)
1786

    
1787
  if as_primary or disk.AssembleOnSecondary():
1788
    r_dev = bdev.Assemble(disk, children)
1789
    result = r_dev
1790
    if as_primary or disk.OpenOnSecondary():
1791
      r_dev.Open()
1792
    DevCacheManager.UpdateCache(r_dev.dev_path, owner,
1793
                                as_primary, disk.iv_name)
1794

    
1795
  else:
1796
    result = True
1797
  return result
1798

    
1799

    
1800
def BlockdevAssemble(disk, owner, as_primary, idx):
1801
  """Activate a block device for an instance.
1802

1803
  This is a wrapper over _RecursiveAssembleBD.
1804

1805
  @rtype: str or boolean
1806
  @return: a C{/dev/...} path for primary nodes, and
1807
      C{True} for secondary nodes
1808

1809
  """
1810
  try:
1811
    result = _RecursiveAssembleBD(disk, owner, as_primary)
1812
    if isinstance(result, bdev.BlockDev):
1813
      # pylint: disable=E1103
1814
      result = result.dev_path
1815
      if as_primary:
1816
        _SymlinkBlockDev(owner, result, idx)
1817
  except errors.BlockDeviceError, err:
1818
    _Fail("Error while assembling disk: %s", err, exc=True)
1819
  except OSError, err:
1820
    _Fail("Error while symlinking disk: %s", err, exc=True)
1821

    
1822
  return result
1823

    
1824

    
1825
def BlockdevShutdown(disk):
1826
  """Shut down a block device.
1827

1828
  First, if the device is assembled (Attach() is successful), then
1829
  the device is shutdown. Then the children of the device are
1830
  shutdown.
1831

1832
  This function is called recursively. Note that we don't cache the
1833
  children or such, as oppossed to assemble, shutdown of different
1834
  devices doesn't require that the upper device was active.
1835

1836
  @type disk: L{objects.Disk}
1837
  @param disk: the description of the disk we should
1838
      shutdown
1839
  @rtype: None
1840

1841
  """
1842
  msgs = []
1843
  r_dev = _RecursiveFindBD(disk)
1844
  if r_dev is not None:
1845
    r_path = r_dev.dev_path
1846
    try:
1847
      r_dev.Shutdown()
1848
      DevCacheManager.RemoveCache(r_path)
1849
    except errors.BlockDeviceError, err:
1850
      msgs.append(str(err))
1851

    
1852
  if disk.children:
1853
    for child in disk.children:
1854
      try:
1855
        BlockdevShutdown(child)
1856
      except RPCFail, err:
1857
        msgs.append(str(err))
1858

    
1859
  if msgs:
1860
    _Fail("; ".join(msgs))
1861

    
1862

    
1863
def BlockdevAddchildren(parent_cdev, new_cdevs):
1864
  """Extend a mirrored block device.
1865

1866
  @type parent_cdev: L{objects.Disk}
1867
  @param parent_cdev: the disk to which we should add children
1868
  @type new_cdevs: list of L{objects.Disk}
1869
  @param new_cdevs: the list of children which we should add
1870
  @rtype: None
1871

1872
  """
1873
  parent_bdev = _RecursiveFindBD(parent_cdev)
1874
  if parent_bdev is None:
1875
    _Fail("Can't find parent device '%s' in add children", parent_cdev)
1876
  new_bdevs = [_RecursiveFindBD(disk) for disk in new_cdevs]
1877
  if new_bdevs.count(None) > 0:
1878
    _Fail("Can't find new device(s) to add: %s:%s", new_bdevs, new_cdevs)
1879
  parent_bdev.AddChildren(new_bdevs)
1880

    
1881

    
1882
def BlockdevRemovechildren(parent_cdev, new_cdevs):
1883
  """Shrink a mirrored block device.
1884

1885
  @type parent_cdev: L{objects.Disk}
1886
  @param parent_cdev: the disk from which we should remove children
1887
  @type new_cdevs: list of L{objects.Disk}
1888
  @param new_cdevs: the list of children which we should remove
1889
  @rtype: None
1890

1891
  """
1892
  parent_bdev = _RecursiveFindBD(parent_cdev)
1893
  if parent_bdev is None:
1894
    _Fail("Can't find parent device '%s' in remove children", parent_cdev)
1895
  devs = []
1896
  for disk in new_cdevs:
1897
    rpath = disk.StaticDevPath()
1898
    if rpath is None:
1899
      bd = _RecursiveFindBD(disk)
1900
      if bd is None:
1901
        _Fail("Can't find device %s while removing children", disk)
1902
      else:
1903
        devs.append(bd.dev_path)
1904
    else:
1905
      if not utils.IsNormAbsPath(rpath):
1906
        _Fail("Strange path returned from StaticDevPath: '%s'", rpath)
1907
      devs.append(rpath)
1908
  parent_bdev.RemoveChildren(devs)
1909

    
1910

    
1911
def BlockdevGetmirrorstatus(disks):
1912
  """Get the mirroring status of a list of devices.
1913

1914
  @type disks: list of L{objects.Disk}
1915
  @param disks: the list of disks which we should query
1916
  @rtype: disk
1917
  @return: List of L{objects.BlockDevStatus}, one for each disk
1918
  @raise errors.BlockDeviceError: if any of the disks cannot be
1919
      found
1920

1921
  """
1922
  stats = []
1923
  for dsk in disks:
1924
    rbd = _RecursiveFindBD(dsk)
1925
    if rbd is None:
1926
      _Fail("Can't find device %s", dsk)
1927

    
1928
    stats.append(rbd.CombinedSyncStatus())
1929

    
1930
  return stats
1931

    
1932

    
1933
def BlockdevGetmirrorstatusMulti(disks):
1934
  """Get the mirroring status of a list of devices.
1935

1936
  @type disks: list of L{objects.Disk}
1937
  @param disks: the list of disks which we should query
1938
  @rtype: disk
1939
  @return: List of tuples, (bool, status), one for each disk; bool denotes
1940
    success/failure, status is L{objects.BlockDevStatus} on success, string
1941
    otherwise
1942

1943
  """
1944
  result = []
1945
  for disk in disks:
1946
    try:
1947
      rbd = _RecursiveFindBD(disk)
1948
      if rbd is None:
1949
        result.append((False, "Can't find device %s" % disk))
1950
        continue
1951

    
1952
      status = rbd.CombinedSyncStatus()
1953
    except errors.BlockDeviceError, err:
1954
      logging.exception("Error while getting disk status")
1955
      result.append((False, str(err)))
1956
    else:
1957
      result.append((True, status))
1958

    
1959
  assert len(disks) == len(result)
1960

    
1961
  return result
1962

    
1963

    
1964
def _RecursiveFindBD(disk):
1965
  """Check if a device is activated.
1966

1967
  If so, return information about the real device.
1968

1969
  @type disk: L{objects.Disk}
1970
  @param disk: the disk object we need to find
1971

1972
  @return: None if the device can't be found,
1973
      otherwise the device instance
1974

1975
  """
1976
  children = []
1977
  if disk.children:
1978
    for chdisk in disk.children:
1979
      children.append(_RecursiveFindBD(chdisk))
1980

    
1981
  return bdev.FindDevice(disk, children)
1982

    
1983

    
1984
def _OpenRealBD(disk):
1985
  """Opens the underlying block device of a disk.
1986

1987
  @type disk: L{objects.Disk}
1988
  @param disk: the disk object we want to open
1989

1990
  """
1991
  real_disk = _RecursiveFindBD(disk)
1992
  if real_disk is None:
1993
    _Fail("Block device '%s' is not set up", disk)
1994

    
1995
  real_disk.Open()
1996

    
1997
  return real_disk
1998

    
1999

    
2000
def BlockdevFind(disk):
2001
  """Check if a device is activated.
2002

2003
  If it is, return information about the real device.
2004

2005
  @type disk: L{objects.Disk}
2006
  @param disk: the disk to find
2007
  @rtype: None or objects.BlockDevStatus
2008
  @return: None if the disk cannot be found, otherwise a the current
2009
           information
2010

2011
  """
2012
  try:
2013
    rbd = _RecursiveFindBD(disk)
2014
  except errors.BlockDeviceError, err:
2015
    _Fail("Failed to find device: %s", err, exc=True)
2016

    
2017
  if rbd is None:
2018
    return None
2019

    
2020
  return rbd.GetSyncStatus()
2021

    
2022

    
2023
def BlockdevGetsize(disks):
2024
  """Computes the size of the given disks.
2025

2026
  If a disk is not found, returns None instead.
2027

2028
  @type disks: list of L{objects.Disk}
2029
  @param disks: the list of disk to compute the size for
2030
  @rtype: list
2031
  @return: list with elements None if the disk cannot be found,
2032
      otherwise the size
2033

2034
  """
2035
  result = []
2036
  for cf in disks:
2037
    try:
2038
      rbd = _RecursiveFindBD(cf)
2039
    except errors.BlockDeviceError:
2040
      result.append(None)
2041
      continue
2042
    if rbd is None:
2043
      result.append(None)
2044
    else:
2045
      result.append(rbd.GetActualSize())
2046
  return result
2047

    
2048

    
2049
def BlockdevExport(disk, dest_node, dest_path, cluster_name):
2050
  """Export a block device to a remote node.
2051

2052
  @type disk: L{objects.Disk}
2053
  @param disk: the description of the disk to export
2054
  @type dest_node: str
2055
  @param dest_node: the destination node to export to
2056
  @type dest_path: str
2057
  @param dest_path: the destination path on the target node
2058
  @type cluster_name: str
2059
  @param cluster_name: the cluster name, needed for SSH hostalias
2060
  @rtype: None
2061

2062
  """
2063
  real_disk = _OpenRealBD(disk)
2064

    
2065
  # the block size on the read dd is 1MiB to match our units
2066
  expcmd = utils.BuildShellCmd("set -e; set -o pipefail; "
2067
                               "dd if=%s bs=1048576 count=%s",
2068
                               real_disk.dev_path, str(disk.size))
2069

    
2070
  # we set here a smaller block size as, due to ssh buffering, more
2071
  # than 64-128k will mostly ignored; we use nocreat to fail if the
2072
  # device is not already there or we pass a wrong path; we use
2073
  # notrunc to no attempt truncate on an LV device; we use oflag=dsync
2074
  # to not buffer too much memory; this means that at best, we flush
2075
  # every 64k, which will not be very fast
2076
  destcmd = utils.BuildShellCmd("dd of=%s conv=nocreat,notrunc bs=65536"
2077
                                " oflag=dsync", dest_path)
2078

    
2079
  remotecmd = _GetSshRunner(cluster_name).BuildCmd(dest_node,
2080
                                                   constants.SSH_LOGIN_USER,
2081
                                                   destcmd)
2082

    
2083
  # all commands have been checked, so we're safe to combine them
2084
  command = "|".join([expcmd, utils.ShellQuoteArgs(remotecmd)])
2085

    
2086
  result = utils.RunCmd(["bash", "-c", command])
2087

    
2088
  if result.failed:
2089
    _Fail("Disk copy command '%s' returned error: %s"
2090
          " output: %s", command, result.fail_reason, result.output)
2091

    
2092

    
2093
def UploadFile(file_name, data, mode, uid, gid, atime, mtime):
2094
  """Write a file to the filesystem.
2095

2096
  This allows the master to overwrite(!) a file. It will only perform
2097
  the operation if the file belongs to a list of configuration files.
2098

2099
  @type file_name: str
2100
  @param file_name: the target file name
2101
  @type data: str
2102
  @param data: the new contents of the file
2103
  @type mode: int
2104
  @param mode: the mode to give the file (can be None)
2105
  @type uid: string
2106
  @param uid: the owner of the file
2107
  @type gid: string
2108
  @param gid: the group of the file
2109
  @type atime: float
2110
  @param atime: the atime to set on the file (can be None)
2111
  @type mtime: float
2112
  @param mtime: the mtime to set on the file (can be None)
2113
  @rtype: None
2114

2115
  """
2116
  file_name = vcluster.LocalizeVirtualPath(file_name)
2117

    
2118
  if not os.path.isabs(file_name):
2119
    _Fail("Filename passed to UploadFile is not absolute: '%s'", file_name)
2120

    
2121
  if file_name not in _ALLOWED_UPLOAD_FILES:
2122
    _Fail("Filename passed to UploadFile not in allowed upload targets: '%s'",
2123
          file_name)
2124

    
2125
  raw_data = _Decompress(data)
2126

    
2127
  if not (isinstance(uid, basestring) and isinstance(gid, basestring)):
2128
    _Fail("Invalid username/groupname type")
2129

    
2130
  getents = runtime.GetEnts()
2131
  uid = getents.LookupUser(uid)
2132
  gid = getents.LookupGroup(gid)
2133

    
2134
  utils.SafeWriteFile(file_name, None,
2135
                      data=raw_data, mode=mode, uid=uid, gid=gid,
2136
                      atime=atime, mtime=mtime)
2137

    
2138

    
2139
def RunOob(oob_program, command, node, timeout):
2140
  """Executes oob_program with given command on given node.
2141

2142
  @param oob_program: The path to the executable oob_program
2143
  @param command: The command to invoke on oob_program
2144
  @param node: The node given as an argument to the program
2145
  @param timeout: Timeout after which we kill the oob program
2146

2147
  @return: stdout
2148
  @raise RPCFail: If execution fails for some reason
2149

2150
  """
2151
  result = utils.RunCmd([oob_program, command, node], timeout=timeout)
2152

    
2153
  if result.failed:
2154
    _Fail("'%s' failed with reason '%s'; output: %s", result.cmd,
2155
          result.fail_reason, result.output)
2156

    
2157
  return result.stdout
2158

    
2159

    
2160
def _OSOndiskAPIVersion(os_dir):
2161
  """Compute and return the API version of a given OS.
2162

2163
  This function will try to read the API version of the OS residing in
2164
  the 'os_dir' directory.
2165

2166
  @type os_dir: str
2167
  @param os_dir: the directory in which we should look for the OS
2168
  @rtype: tuple
2169
  @return: tuple (status, data) with status denoting the validity and
2170
      data holding either the vaid versions or an error message
2171

2172
  """
2173
  api_file = utils.PathJoin(os_dir, constants.OS_API_FILE)
2174

    
2175
  try:
2176
    st = os.stat(api_file)
2177
  except EnvironmentError, err:
2178
    return False, ("Required file '%s' not found under path %s: %s" %
2179
                   (constants.OS_API_FILE, os_dir, utils.ErrnoOrStr(err)))
2180

    
2181
  if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2182
    return False, ("File '%s' in %s is not a regular file" %
2183
                   (constants.OS_API_FILE, os_dir))
2184

    
2185
  try:
2186
    api_versions = utils.ReadFile(api_file).splitlines()
2187
  except EnvironmentError, err:
2188
    return False, ("Error while reading the API version file at %s: %s" %
2189
                   (api_file, utils.ErrnoOrStr(err)))
2190

    
2191
  try:
2192
    api_versions = [int(version.strip()) for version in api_versions]
2193
  except (TypeError, ValueError), err:
2194
    return False, ("API version(s) can't be converted to integer: %s" %
2195
                   str(err))
2196

    
2197
  return True, api_versions
2198

    
2199

    
2200
def DiagnoseOS(top_dirs=None):
2201
  """Compute the validity for all OSes.
2202

2203
  @type top_dirs: list
2204
  @param top_dirs: the list of directories in which to
2205
      search (if not given defaults to
2206
      L{pathutils.OS_SEARCH_PATH})
2207
  @rtype: list of L{objects.OS}
2208
  @return: a list of tuples (name, path, status, diagnose, variants,
2209
      parameters, api_version) for all (potential) OSes under all
2210
      search paths, where:
2211
          - name is the (potential) OS name
2212
          - path is the full path to the OS
2213
          - status True/False is the validity of the OS
2214
          - diagnose is the error message for an invalid OS, otherwise empty
2215
          - variants is a list of supported OS variants, if any
2216
          - parameters is a list of (name, help) parameters, if any
2217
          - api_version is a list of support OS API versions
2218

2219
  """
2220
  if top_dirs is None:
2221
    top_dirs = pathutils.OS_SEARCH_PATH
2222

    
2223
  result = []
2224
  for dir_name in top_dirs:
2225
    if os.path.isdir(dir_name):
2226
      try:
2227
        f_names = utils.ListVisibleFiles(dir_name)
2228
      except EnvironmentError, err:
2229
        logging.exception("Can't list the OS directory %s: %s", dir_name, err)
2230
        break
2231
      for name in f_names:
2232
        os_path = utils.PathJoin(dir_name, name)
2233
        status, os_inst = _TryOSFromDisk(name, base_dir=dir_name)
2234
        if status:
2235
          diagnose = ""
2236
          variants = os_inst.supported_variants
2237
          parameters = os_inst.supported_parameters
2238
          api_versions = os_inst.api_versions
2239
        else:
2240
          diagnose = os_inst
2241
          variants = parameters = api_versions = []
2242
        result.append((name, os_path, status, diagnose, variants,
2243
                       parameters, api_versions))
2244

    
2245
  return result
2246

    
2247

    
2248
def _TryOSFromDisk(name, base_dir=None):
2249
  """Create an OS instance from disk.
2250

2251
  This function will return an OS instance if the given name is a
2252
  valid OS name.
2253

2254
  @type base_dir: string
2255
  @keyword base_dir: Base directory containing OS installations.
2256
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2257
  @rtype: tuple
2258
  @return: success and either the OS instance if we find a valid one,
2259
      or error message
2260

2261
  """
2262
  if base_dir is None:
2263
    os_dir = utils.FindFile(name, pathutils.OS_SEARCH_PATH, os.path.isdir)
2264
  else:
2265
    os_dir = utils.FindFile(name, [base_dir], os.path.isdir)
2266

    
2267
  if os_dir is None:
2268
    return False, "Directory for OS %s not found in search path" % name
2269

    
2270
  status, api_versions = _OSOndiskAPIVersion(os_dir)
2271
  if not status:
2272
    # push the error up
2273
    return status, api_versions
2274

    
2275
  if not constants.OS_API_VERSIONS.intersection(api_versions):
2276
    return False, ("API version mismatch for path '%s': found %s, want %s." %
2277
                   (os_dir, api_versions, constants.OS_API_VERSIONS))
2278

    
2279
  # OS Files dictionary, we will populate it with the absolute path
2280
  # names; if the value is True, then it is a required file, otherwise
2281
  # an optional one
2282
  os_files = dict.fromkeys(constants.OS_SCRIPTS, True)
2283

    
2284
  if max(api_versions) >= constants.OS_API_V15:
2285
    os_files[constants.OS_VARIANTS_FILE] = False
2286

    
2287
  if max(api_versions) >= constants.OS_API_V20:
2288
    os_files[constants.OS_PARAMETERS_FILE] = True
2289
  else:
2290
    del os_files[constants.OS_SCRIPT_VERIFY]
2291

    
2292
  for (filename, required) in os_files.items():
2293
    os_files[filename] = utils.PathJoin(os_dir, filename)
2294

    
2295
    try:
2296
      st = os.stat(os_files[filename])
2297
    except EnvironmentError, err:
2298
      if err.errno == errno.ENOENT and not required:
2299
        del os_files[filename]
2300
        continue
2301
      return False, ("File '%s' under path '%s' is missing (%s)" %
2302
                     (filename, os_dir, utils.ErrnoOrStr(err)))
2303

    
2304
    if not stat.S_ISREG(stat.S_IFMT(st.st_mode)):
2305
      return False, ("File '%s' under path '%s' is not a regular file" %
2306
                     (filename, os_dir))
2307

    
2308
    if filename in constants.OS_SCRIPTS:
2309
      if stat.S_IMODE(st.st_mode) & stat.S_IXUSR != stat.S_IXUSR:
2310
        return False, ("File '%s' under path '%s' is not executable" %
2311
                       (filename, os_dir))
2312

    
2313
  variants = []
2314
  if constants.OS_VARIANTS_FILE in os_files:
2315
    variants_file = os_files[constants.OS_VARIANTS_FILE]
2316
    try:
2317
      variants = \
2318
        utils.FilterEmptyLinesAndComments(utils.ReadFile(variants_file))
2319
    except EnvironmentError, err:
2320
      # we accept missing files, but not other errors
2321
      if err.errno != errno.ENOENT:
2322
        return False, ("Error while reading the OS variants file at %s: %s" %
2323
                       (variants_file, utils.ErrnoOrStr(err)))
2324

    
2325
  parameters = []
2326
  if constants.OS_PARAMETERS_FILE in os_files:
2327
    parameters_file = os_files[constants.OS_PARAMETERS_FILE]
2328
    try:
2329
      parameters = utils.ReadFile(parameters_file).splitlines()
2330
    except EnvironmentError, err:
2331
      return False, ("Error while reading the OS parameters file at %s: %s" %
2332
                     (parameters_file, utils.ErrnoOrStr(err)))
2333
    parameters = [v.split(None, 1) for v in parameters]
2334

    
2335
  os_obj = objects.OS(name=name, path=os_dir,
2336
                      create_script=os_files[constants.OS_SCRIPT_CREATE],
2337
                      export_script=os_files[constants.OS_SCRIPT_EXPORT],
2338
                      import_script=os_files[constants.OS_SCRIPT_IMPORT],
2339
                      rename_script=os_files[constants.OS_SCRIPT_RENAME],
2340
                      verify_script=os_files.get(constants.OS_SCRIPT_VERIFY,
2341
                                                 None),
2342
                      supported_variants=variants,
2343
                      supported_parameters=parameters,
2344
                      api_versions=api_versions)
2345
  return True, os_obj
2346

    
2347

    
2348
def OSFromDisk(name, base_dir=None):
2349
  """Create an OS instance from disk.
2350

2351
  This function will return an OS instance if the given name is a
2352
  valid OS name. Otherwise, it will raise an appropriate
2353
  L{RPCFail} exception, detailing why this is not a valid OS.
2354

2355
  This is just a wrapper over L{_TryOSFromDisk}, which doesn't raise
2356
  an exception but returns true/false status data.
2357

2358
  @type base_dir: string
2359
  @keyword base_dir: Base directory containing OS installations.
2360
                     Defaults to a search in all the OS_SEARCH_PATH dirs.
2361
  @rtype: L{objects.OS}
2362
  @return: the OS instance if we find a valid one
2363
  @raise RPCFail: if we don't find a valid OS
2364

2365
  """
2366
  name_only = objects.OS.GetName(name)
2367
  status, payload = _TryOSFromDisk(name_only, base_dir)
2368

    
2369
  if not status:
2370
    _Fail(payload)
2371

    
2372
  return payload
2373

    
2374

    
2375
def OSCoreEnv(os_name, inst_os, os_params, debug=0):
2376
  """Calculate the basic environment for an os script.
2377

2378
  @type os_name: str
2379
  @param os_name: full operating system name (including variant)
2380
  @type inst_os: L{objects.OS}
2381
  @param inst_os: operating system for which the environment is being built
2382
  @type os_params: dict
2383
  @param os_params: the OS parameters
2384
  @type debug: integer
2385
  @param debug: debug level (0 or 1, for OS Api 10)
2386
  @rtype: dict
2387
  @return: dict of environment variables
2388
  @raise errors.BlockDeviceError: if the block device
2389
      cannot be found
2390

2391
  """
2392
  result = {}
2393
  api_version = \
2394
    max(constants.OS_API_VERSIONS.intersection(inst_os.api_versions))
2395
  result["OS_API_VERSION"] = "%d" % api_version
2396
  result["OS_NAME"] = inst_os.name
2397
  result["DEBUG_LEVEL"] = "%d" % debug
2398

    
2399
  # OS variants
2400
  if api_version >= constants.OS_API_V15 and inst_os.supported_variants:
2401
    variant = objects.OS.GetVariant(os_name)
2402
    if not variant:
2403
      variant = inst_os.supported_variants[0]
2404
  else:
2405
    variant = ""
2406
  result["OS_VARIANT"] = variant
2407

    
2408
  # OS params
2409
  for pname, pvalue in os_params.items():
2410
    result["OSP_%s" % pname.upper()] = pvalue
2411

    
2412
  # Set a default path otherwise programs called by OS scripts (or
2413
  # even hooks called from OS scripts) might break, and we don't want
2414
  # to have each script require setting a PATH variable
2415
  result["PATH"] = constants.HOOKS_PATH
2416

    
2417
  return result
2418

    
2419

    
2420
def OSEnvironment(instance, inst_os, debug=0):
2421
  """Calculate the environment for an os script.
2422

2423
  @type instance: L{objects.Instance}
2424
  @param instance: target instance for the os script run
2425
  @type inst_os: L{objects.OS}
2426
  @param inst_os: operating system for which the environment is being built
2427
  @type debug: integer
2428
  @param debug: debug level (0 or 1, for OS Api 10)
2429
  @rtype: dict
2430
  @return: dict of environment variables
2431
  @raise errors.BlockDeviceError: if the block device
2432
      cannot be found
2433

2434
  """
2435
  result = OSCoreEnv(instance.os, inst_os, instance.osparams, debug=debug)
2436

    
2437
  for attr in ["name", "os", "uuid", "ctime", "mtime", "primary_node"]:
2438
    result["INSTANCE_%s" % attr.upper()] = str(getattr(instance, attr))
2439

    
2440
  result["HYPERVISOR"] = instance.hypervisor
2441
  result["DISK_COUNT"] = "%d" % len(instance.disks)
2442
  result["NIC_COUNT"] = "%d" % len(instance.nics)
2443
  result["INSTANCE_SECONDARY_NODES"] = \
2444
      ("%s" % " ".join(instance.secondary_nodes))
2445

    
2446
  # Disks
2447
  for idx, disk in enumerate(instance.disks):
2448
    real_disk = _OpenRealBD(disk)
2449
    result["DISK_%d_PATH" % idx] = real_disk.dev_path
2450
    result["DISK_%d_ACCESS" % idx] = disk.mode
2451
    if constants.HV_DISK_TYPE in instance.hvparams:
2452
      result["DISK_%d_FRONTEND_TYPE" % idx] = \
2453
        instance.hvparams[constants.HV_DISK_TYPE]
2454
    if disk.dev_type in constants.LDS_BLOCK:
2455
      result["DISK_%d_BACKEND_TYPE" % idx] = "block"
2456
    elif disk.dev_type == constants.LD_FILE:
2457
      result["DISK_%d_BACKEND_TYPE" % idx] = \
2458
        "file:%s" % disk.physical_id[0]
2459

    
2460
  # NICs
2461
  for idx, nic in enumerate(instance.nics):
2462
    result["NIC_%d_MAC" % idx] = nic.mac
2463
    if nic.ip:
2464
      result["NIC_%d_IP" % idx] = nic.ip
2465
    result["NIC_%d_MODE" % idx] = nic.nicparams[constants.NIC_MODE]
2466
    if nic.nicparams[constants.NIC_MODE] == constants.NIC_MODE_BRIDGED:
2467
      result["NIC_%d_BRIDGE" % idx] = nic.nicparams[constants.NIC_LINK]
2468
    if nic.nicparams[constants.NIC_LINK]:
2469
      result["NIC_%d_LINK" % idx] = nic.nicparams[constants.NIC_LINK]
2470
    if nic.network:
2471
      result["NIC_%d_NETWORK" % idx] = nic.network
2472
    if constants.HV_NIC_TYPE in instance.hvparams:
2473
      result["NIC_%d_FRONTEND_TYPE" % idx] = \
2474
        instance.hvparams[constants.HV_NIC_TYPE]
2475

    
2476
  # HV/BE params
2477
  for source, kind in [(instance.beparams, "BE"), (instance.hvparams, "HV")]:
2478
    for key, value in source.items():
2479
      result["INSTANCE_%s_%s" % (kind, key)] = str(value)
2480

    
2481
  return result
2482

    
2483

    
2484
def DiagnoseExtStorage(top_dirs=None):
2485
  """Compute the validity for all ExtStorage Providers.
2486

2487
  @type top_dirs: list
2488
  @param top_dirs: the list of directories in which to
2489
      search (if not given defaults to
2490
      L{pathutils.ES_SEARCH_PATH})
2491
  @rtype: list of L{objects.ExtStorage}
2492
  @return: a list of tuples (name, path, status, diagnose, parameters)
2493
      for all (potential) ExtStorage Providers under all
2494
      search paths, where:
2495
          - name is the (potential) ExtStorage Provider
2496
          - path is the full path to the ExtStorage Provider
2497
          - status True/False is the validity of the ExtStorage Provider
2498
          - diagnose is the error message for an invalid ExtStorage Provider,
2499
            otherwise empty
2500
          - parameters is a list of (name, help) parameters, if any
2501

2502
  """
2503
  if top_dirs is None:
2504
    top_dirs = pathutils.ES_SEARCH_PATH
2505

    
2506
  result = []
2507
  for dir_name in top_dirs:
2508
    if os.path.isdir(dir_name):
2509
      try:
2510
        f_names = utils.ListVisibleFiles(dir_name)
2511
      except EnvironmentError, err:
2512
        logging.exception("Can't list the ExtStorage directory %s: %s",
2513
                          dir_name, err)
2514
        break
2515
      for name in f_names:
2516
        es_path = utils.PathJoin(dir_name, name)
2517
        status, es_inst = bdev.ExtStorageFromDisk(name, base_dir=dir_name)
2518
        if status:
2519
          diagnose = ""
2520
          parameters = es_inst.supported_parameters
2521
        else:
2522
          diagnose = es_inst
2523
          parameters = []
2524
        result.append((name, es_path, status, diagnose, parameters))
2525

    
2526
  return result
2527

    
2528

    
2529
def BlockdevGrow(disk, amount, dryrun, backingstore):
2530
  """Grow a stack of block devices.
2531

2532
  This function is called recursively, with the childrens being the
2533
  first ones to resize.
2534

2535
  @type disk: L{objects.Disk}
2536
  @param disk: the disk to be grown
2537
  @type amount: integer
2538
  @param amount: the amount (in mebibytes) to grow with
2539
  @type dryrun: boolean
2540
  @param dryrun: whether to execute the operation in simulation mode
2541
      only, without actually increasing the size
2542
  @param backingstore: whether to execute the operation on backing storage
2543
      only, or on "logical" storage only; e.g. DRBD is logical storage,
2544
      whereas LVM, file, RBD are backing storage
2545
  @rtype: (status, result)
2546
  @return: a tuple with the status of the operation (True/False), and
2547
      the errors message if status is False
2548

2549
  """
2550
  r_dev = _RecursiveFindBD(disk)
2551
  if r_dev is None:
2552
    _Fail("Cannot find block device %s", disk)
2553

    
2554
  try:
2555
    r_dev.Grow(amount, dryrun, backingstore)
2556
  except errors.BlockDeviceError, err:
2557
    _Fail("Failed to grow block device: %s", err, exc=True)
2558

    
2559

    
2560
def BlockdevSnapshot(disk):
2561
  """Create a snapshot copy of a block device.
2562

2563
  This function is called recursively, and the snapshot is actually created
2564
  just for the leaf lvm backend device.
2565

2566
  @type disk: L{objects.Disk}
2567
  @param disk: the disk to be snapshotted
2568
  @rtype: string
2569
  @return: snapshot disk ID as (vg, lv)
2570

2571
  """
2572
  if disk.dev_type == constants.LD_DRBD8:
2573
    if not disk.children:
2574
      _Fail("DRBD device '%s' without backing storage cannot be snapshotted",
2575
            disk.unique_id)
2576
    return BlockdevSnapshot(disk.children[0])
2577
  elif disk.dev_type == constants.LD_LV:
2578
    r_dev = _RecursiveFindBD(disk)
2579
    if r_dev is not None:
2580
      # FIXME: choose a saner value for the snapshot size
2581
      # let's stay on the safe side and ask for the full size, for now
2582
      return r_dev.Snapshot(disk.size)
2583
    else:
2584
      _Fail("Cannot find block device %s", disk)
2585
  else:
2586
    _Fail("Cannot snapshot non-lvm block device '%s' of type '%s'",
2587
          disk.unique_id, disk.dev_type)
2588

    
2589

    
2590
def BlockdevSetInfo(disk, info):
2591
  """Sets 'metadata' information on block devices.
2592

2593
  This function sets 'info' metadata on block devices. Initial
2594
  information is set at device creation; this function should be used
2595
  for example after renames.
2596

2597
  @type disk: L{objects.Disk}
2598
  @param disk: the disk to be grown
2599
  @type info: string
2600
  @param info: new 'info' metadata
2601
  @rtype: (status, result)
2602
  @return: a tuple with the status of the operation (True/False), and
2603
      the errors message if status is False
2604

2605
  """
2606
  r_dev = _RecursiveFindBD(disk)
2607
  if r_dev is None:
2608
    _Fail("Cannot find block device %s", disk)
2609

    
2610
  try:
2611
    r_dev.SetInfo(info)
2612
  except errors.BlockDeviceError, err:
2613
    _Fail("Failed to set information on block device: %s", err, exc=True)
2614

    
2615

    
2616
def FinalizeExport(instance, snap_disks):
2617
  """Write out the export configuration information.
2618

2619
  @type instance: L{objects.Instance}
2620
  @param instance: the instance which we export, used for
2621
      saving configuration
2622
  @type snap_disks: list of L{objects.Disk}
2623
  @param snap_disks: list of snapshot block devices, which
2624
      will be used to get the actual name of the dump file
2625

2626
  @rtype: None
2627

2628
  """
2629
  destdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name + ".new")
2630
  finaldestdir = utils.PathJoin(pathutils.EXPORT_DIR, instance.name)
2631

    
2632
  config = objects.SerializableConfigParser()
2633

    
2634
  config.add_section(constants.INISECT_EXP)
2635
  config.set(constants.INISECT_EXP, "version", "0")
2636
  config.set(constants.INISECT_EXP, "timestamp", "%d" % int(time.time()))
2637
  config.set(constants.INISECT_EXP, "source", instance.primary_node)
2638
  config.set(constants.INISECT_EXP, "os", instance.os)
2639
  config.set(constants.INISECT_EXP, "compression", "none")
2640

    
2641
  config.add_section(constants.INISECT_INS)
2642
  config.set(constants.INISECT_INS, "name", instance.name)
2643
  config.set(constants.INISECT_INS, "maxmem", "%d" %
2644
             instance.beparams[constants.BE_MAXMEM])
2645
  config.set(constants.INISECT_INS, "minmem", "%d" %
2646
             instance.beparams[constants.BE_MINMEM])
2647
  # "memory" is deprecated, but useful for exporting to old ganeti versions
2648
  config.set(constants.INISECT_INS, "memory", "%d" %
2649
             instance.beparams[constants.BE_MAXMEM])
2650
  config.set(constants.INISECT_INS, "vcpus", "%d" %
2651
             instance.beparams[constants.BE_VCPUS])
2652
  config.set(constants.INISECT_INS, "disk_template", instance.disk_template)
2653
  config.set(constants.INISECT_INS, "hypervisor", instance.hypervisor)
2654
  config.set(constants.INISECT_INS, "tags", " ".join(instance.GetTags()))
2655

    
2656
  nic_total = 0
2657
  for nic_count, nic in enumerate(instance.nics):
2658
    nic_total += 1
2659
    config.set(constants.INISECT_INS, "nic%d_mac" %
2660
               nic_count, "%s" % nic.mac)
2661
    config.set(constants.INISECT_INS, "nic%d_ip" % nic_count, "%s" % nic.ip)
2662
    config.set(constants.INISECT_INS, "nic%d_network" % nic_count,
2663
               "%s" % nic.network)
2664
    for param in constants.NICS_PARAMETER_TYPES:
2665
      config.set(constants.INISECT_INS, "nic%d_%s" % (nic_count, param),
2666
                 "%s" % nic.nicparams.get(param, None))
2667
  # TODO: redundant: on load can read nics until it doesn't exist
2668
  config.set(constants.INISECT_INS, "nic_count", "%d" % nic_total)
2669

    
2670
  disk_total = 0
2671
  for disk_count, disk in enumerate(snap_disks):
2672
    if disk:
2673
      disk_total += 1
2674
      config.set(constants.INISECT_INS, "disk%d_ivname" % disk_count,
2675
                 ("%s" % disk.iv_name))
2676
      config.set(constants.INISECT_INS, "disk%d_dump" % disk_count,
2677
                 ("%s" % disk.physical_id[1]))
2678
      config.set(constants.INISECT_INS, "disk%d_size" % disk_count,
2679
                 ("%d" % disk.size))
2680

    
2681
  config.set(constants.INISECT_INS, "disk_count", "%d" % disk_total)
2682

    
2683
  # New-style hypervisor/backend parameters
2684

    
2685
  config.add_section(constants.INISECT_HYP)
2686
  for name, value in instance.hvparams.items():
2687
    if name not in constants.HVC_GLOBALS:
2688
      config.set(constants.INISECT_HYP, name, str(value))
2689

    
2690
  config.add_section(constants.INISECT_BEP)
2691
  for name, value in instance.beparams.items():
2692
    config.set(constants.INISECT_BEP, name, str(value))
2693

    
2694
  config.add_section(constants.INISECT_OSP)
2695
  for name, value in instance.osparams.items():
2696
    config.set(constants.INISECT_OSP, name, str(value))
2697

    
2698
  utils.WriteFile(utils.PathJoin(destdir, constants.EXPORT_CONF_FILE),
2699
                  data=config.Dumps())
2700
  shutil.rmtree(finaldestdir, ignore_errors=True)
2701
  shutil.move(destdir, finaldestdir)
2702

    
2703

    
2704
def ExportInfo(dest):
2705
  """Get export configuration information.
2706

2707
  @type dest: str
2708
  @param dest: directory containing the export
2709

2710
  @rtype: L{objects.SerializableConfigParser}
2711
  @return: a serializable config file containing the
2712
      export info
2713

2714
  """
2715
  cff = utils.PathJoin(dest, constants.EXPORT_CONF_FILE)
2716

    
2717
  config = objects.SerializableConfigParser()
2718
  config.read(cff)
2719

    
2720
  if (not config.has_section(constants.INISECT_EXP) or
2721
      not config.has_section(constants.INISECT_INS)):
2722
    _Fail("Export info file doesn't have the required fields")
2723

    
2724
  return config.Dumps()
2725

    
2726

    
2727
def ListExports():
2728
  """Return a list of exports currently available on this machine.
2729

2730
  @rtype: list
2731
  @return: list of the exports
2732

2733
  """
2734
  if os.path.isdir(pathutils.EXPORT_DIR):
2735
    return sorted(utils.ListVisibleFiles(pathutils.EXPORT_DIR))
2736
  else:
2737
    _Fail("No exports directory")
2738

    
2739

    
2740
def RemoveExport(export):
2741
  """Remove an existing export from the node.
2742

2743
  @type export: str
2744
  @param export: the name of the export to remove
2745
  @rtype: None
2746

2747
  """
2748
  target = utils.PathJoin(pathutils.EXPORT_DIR, export)
2749

    
2750
  try:
2751
    shutil.rmtree(target)
2752
  except EnvironmentError, err:
2753
    _Fail("Error while removing the export: %s", err, exc=True)
2754

    
2755

    
2756
def BlockdevRename(devlist):
2757
  """Rename a list of block devices.
2758

2759
  @type devlist: list of tuples
2760
  @param devlist: list of tuples of the form  (disk,
2761
      new_logical_id, new_physical_id); disk is an
2762
      L{objects.Disk} object describing the current disk,
2763
      and new logical_id/physical_id is the name we
2764
      rename it to
2765
  @rtype: boolean
2766
  @return: True if all renames succeeded, False otherwise
2767

2768
  """
2769
  msgs = []
2770
  result = True
2771
  for disk, unique_id in devlist:
2772
    dev = _RecursiveFindBD(disk)
2773
    if dev is None:
2774
      msgs.append("Can't find device %s in rename" % str(disk))
2775
      result = False
2776
      continue
2777
    try:
2778
      old_rpath = dev.dev_path
2779
      dev.Rename(unique_id)
2780
      new_rpath = dev.dev_path
2781
      if old_rpath != new_rpath:
2782
        DevCacheManager.RemoveCache(old_rpath)
2783
        # FIXME: we should add the new cache information here, like:
2784
        # DevCacheManager.UpdateCache(new_rpath, owner, ...)
2785
        # but we don't have the owner here - maybe parse from existing
2786
        # cache? for now, we only lose lvm data when we rename, which
2787
        # is less critical than DRBD or MD
2788
    except errors.BlockDeviceError, err:
2789
      msgs.append("Can't rename device '%s' to '%s': %s" %
2790
                  (dev, unique_id, err))
2791
      logging.exception("Can't rename device '%s' to '%s'", dev, unique_id)
2792
      result = False
2793
  if not result:
2794
    _Fail("; ".join(msgs))
2795

    
2796

    
2797
def _TransformFileStorageDir(fs_dir):
2798
  """Checks whether given file_storage_dir is valid.
2799

2800
  Checks wheter the given fs_dir is within the cluster-wide default
2801
  file_storage_dir or the shared_file_storage_dir, which are stored in
2802
  SimpleStore. Only paths under those directories are allowed.
2803

2804
  @type fs_dir: str
2805
  @param fs_dir: the path to check
2806

2807
  @return: the normalized path if valid, None otherwise
2808

2809
  """
2810
  if not (constants.ENABLE_FILE_STORAGE or
2811
          constants.ENABLE_SHARED_FILE_STORAGE):
2812
    _Fail("File storage disabled at configure time")
2813

    
2814
  bdev.CheckFileStoragePath(fs_dir)
2815

    
2816
  return os.path.normpath(fs_dir)
2817

    
2818

    
2819
def CreateFileStorageDir(file_storage_dir):
2820
  """Create file storage directory.
2821

2822
  @type file_storage_dir: str
2823
  @param file_storage_dir: directory to create
2824

2825
  @rtype: tuple
2826
  @return: tuple with first element a boolean indicating wheter dir
2827
      creation was successful or not
2828

2829
  """
2830
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2831
  if os.path.exists(file_storage_dir):
2832
    if not os.path.isdir(file_storage_dir):
2833
      _Fail("Specified storage dir '%s' is not a directory",
2834
            file_storage_dir)
2835
  else:
2836
    try:
2837
      os.makedirs(file_storage_dir, 0750)
2838
    except OSError, err:
2839
      _Fail("Cannot create file storage directory '%s': %s",
2840
            file_storage_dir, err, exc=True)
2841

    
2842

    
2843
def RemoveFileStorageDir(file_storage_dir):
2844
  """Remove file storage directory.
2845

2846
  Remove it only if it's empty. If not log an error and return.
2847

2848
  @type file_storage_dir: str
2849
  @param file_storage_dir: the directory we should cleanup
2850
  @rtype: tuple (success,)
2851
  @return: tuple of one element, C{success}, denoting
2852
      whether the operation was successful
2853

2854
  """
2855
  file_storage_dir = _TransformFileStorageDir(file_storage_dir)
2856
  if os.path.exists(file_storage_dir):
2857
    if not os.path.isdir(file_storage_dir):
2858
      _Fail("Specified Storage directory '%s' is not a directory",
2859
            file_storage_dir)
2860
    # deletes dir only if empty, otherwise we want to fail the rpc call
2861
    try:
2862
      os.rmdir(file_storage_dir)
2863
    except OSError, err:
2864
      _Fail("Cannot remove file storage directory '%s': %s",
2865
            file_storage_dir, err)
2866

    
2867

    
2868
def RenameFileStorageDir(old_file_storage_dir, new_file_storage_dir):
2869
  """Rename the file storage directory.
2870

2871
  @type old_file_storage_dir: str
2872
  @param old_file_storage_dir: the current path
2873
  @type new_file_storage_dir: str
2874
  @param new_file_storage_dir: the name we should rename to
2875
  @rtype: tuple (success,)
2876
  @return: tuple of one element, C{success}, denoting
2877
      whether the operation was successful
2878

2879
  """
2880
  old_file_storage_dir = _TransformFileStorageDir(old_file_storage_dir)
2881
  new_file_storage_dir = _TransformFileStorageDir(new_file_storage_dir)
2882
  if not os.path.exists(new_file_storage_dir):
2883
    if os.path.isdir(old_file_storage_dir):
2884
      try:
2885
        os.rename(old_file_storage_dir, new_file_storage_dir)
2886
      except OSError, err:
2887
        _Fail("Cannot rename '%s' to '%s': %s",
2888
              old_file_storage_dir, new_file_storage_dir, err)
2889
    else:
2890
      _Fail("Specified storage dir '%s' is not a directory",
2891
            old_file_storage_dir)
2892
  else:
2893
    if os.path.exists(old_file_storage_dir):
2894
      _Fail("Cannot rename '%s' to '%s': both locations exist",
2895
            old_file_storage_dir, new_file_storage_dir)
2896

    
2897

    
2898
def _EnsureJobQueueFile(file_name):
2899
  """Checks whether the given filename is in the queue directory.
2900

2901
  @type file_name: str
2902
  @param file_name: the file name we should check
2903
  @rtype: None
2904
  @raises RPCFail: if the file is not valid
2905

2906
  """
2907
  if not utils.IsBelowDir(pathutils.QUEUE_DIR, file_name):
2908
    _Fail("Passed job queue file '%s' does not belong to"
2909
          " the queue directory '%s'", file_name, pathutils.QUEUE_DIR)
2910

    
2911

    
2912
def JobQueueUpdate(file_name, content):
2913
  """Updates a file in the queue directory.
2914

2915
  This is just a wrapper over L{utils.io.WriteFile}, with proper
2916
  checking.
2917

2918
  @type file_name: str
2919
  @param file_name: the job file name
2920
  @type content: str
2921
  @param content: the new job contents
2922
  @rtype: boolean
2923
  @return: the success of the operation
2924

2925
  """
2926
  file_name = vcluster.LocalizeVirtualPath(file_name)
2927

    
2928
  _EnsureJobQueueFile(file_name)
2929
  getents = runtime.GetEnts()
2930

    
2931
  # Write and replace the file atomically
2932
  utils.WriteFile(file_name, data=_Decompress(content), uid=getents.masterd_uid,
2933
                  gid=getents.masterd_gid)
2934

    
2935

    
2936
def JobQueueRename(old, new):
2937
  """Renames a job queue file.
2938

2939
  This is just a wrapper over os.rename with proper checking.
2940

2941
  @type old: str
2942
  @param old: the old (actual) file name
2943
  @type new: str
2944
  @param new: the desired file name
2945
  @rtype: tuple
2946
  @return: the success of the operation and payload
2947

2948
  """
2949
  old = vcluster.LocalizeVirtualPath(old)
2950
  new = vcluster.LocalizeVirtualPath(new)
2951

    
2952
  _EnsureJobQueueFile(old)
2953
  _EnsureJobQueueFile(new)
2954

    
2955
  getents = runtime.GetEnts()
2956

    
2957
  utils.RenameFile(old, new, mkdir=True, mkdir_mode=0700,
2958
                   dir_uid=getents.masterd_uid, dir_gid=getents.masterd_gid)
2959

    
2960

    
2961
def BlockdevClose(instance_name, disks):
2962
  """Closes the given block devices.
2963

2964
  This means they will be switched to secondary mode (in case of
2965
  DRBD).
2966

2967
  @param instance_name: if the argument is not empty, the symlinks
2968
      of this instance will be removed
2969
  @type disks: list of L{objects.Disk}
2970
  @param disks: the list of disks to be closed
2971
  @rtype: tuple (success, message)
2972
  @return: a tuple of success and message, where success
2973
      indicates the succes of the operation, and message
2974
      which will contain the error details in case we
2975
      failed
2976

2977
  """
2978
  bdevs = []
2979
  for cf in disks:
2980
    rd = _RecursiveFindBD(cf)
2981
    if rd is None:
2982
      _Fail("Can't find device %s", cf)
2983
    bdevs.append(rd)
2984

    
2985
  msg = []
2986
  for rd in bdevs:
2987
    try:
2988
      rd.Close()
2989
    except errors.BlockDeviceError, err:
2990
      msg.append(str(err))
2991
  if msg:
2992
    _Fail("Can't make devices secondary: %s", ",".join(msg))
2993
  else:
2994
    if instance_name:
2995
      _RemoveBlockDevLinks(instance_name, disks)
2996

    
2997

    
2998
def ValidateHVParams(hvname, hvparams):
2999
  """Validates the given hypervisor parameters.
3000

3001
  @type hvname: string
3002
  @param hvname: the hypervisor name
3003
  @type hvparams: dict
3004
  @param hvparams: the hypervisor parameters to be validated
3005
  @rtype: None
3006

3007
  """
3008
  try:
3009
    hv_type = hypervisor.GetHypervisor(hvname)
3010
    hv_type.ValidateParameters(hvparams)
3011
  except errors.HypervisorError, err:
3012
    _Fail(str(err), log=False)
3013

    
3014

    
3015
def _CheckOSPList(os_obj, parameters):
3016
  """Check whether a list of parameters is supported by the OS.
3017

3018
  @type os_obj: L{objects.OS}
3019
  @param os_obj: OS object to check
3020
  @type parameters: list
3021
  @param parameters: the list of parameters to check
3022

3023
  """
3024
  supported = [v[0] for v in os_obj.supported_parameters]
3025
  delta = frozenset(parameters).difference(supported)
3026
  if delta:
3027
    _Fail("The following parameters are not supported"
3028
          " by the OS %s: %s" % (os_obj.name, utils.CommaJoin(delta)))
3029

    
3030

    
3031
def ValidateOS(required, osname, checks, osparams):
3032
  """Validate the given OS' parameters.
3033

3034
  @type required: boolean
3035
  @param required: whether absence of the OS should translate into
3036
      failure or not
3037
  @type osname: string
3038
  @param osname: the OS to be validated
3039
  @type checks: list
3040
  @param checks: list of the checks to run (currently only 'parameters')
3041
  @type osparams: dict
3042
  @param osparams: dictionary with OS parameters
3043
  @rtype: boolean
3044
  @return: True if the validation passed, or False if the OS was not
3045
      found and L{required} was false
3046

3047
  """
3048
  if not constants.OS_VALIDATE_CALLS.issuperset(checks):
3049
    _Fail("Unknown checks required for OS %s: %s", osname,
3050
          set(checks).difference(constants.OS_VALIDATE_CALLS))
3051

    
3052
  name_only = objects.OS.GetName(osname)
3053
  status, tbv = _TryOSFromDisk(name_only, None)
3054

    
3055
  if not status:
3056
    if required:
3057
      _Fail(tbv)
3058
    else:
3059
      return False
3060

    
3061
  if max(tbv.api_versions) < constants.OS_API_V20:
3062
    return True
3063

    
3064
  if constants.OS_VALIDATE_PARAMETERS in checks:
3065
    _CheckOSPList(tbv, osparams.keys())
3066

    
3067
  validate_env = OSCoreEnv(osname, tbv, osparams)
3068
  result = utils.RunCmd([tbv.verify_script] + checks, env=validate_env,
3069
                        cwd=tbv.path, reset_env=True)
3070
  if result.failed:
3071
    logging.error("os validate command '%s' returned error: %s output: %s",
3072
                  result.cmd, result.fail_reason, result.output)
3073
    _Fail("OS validation script failed (%s), output: %s",
3074
          result.fail_reason, result.output, log=False)
3075

    
3076
  return True
3077

    
3078

    
3079
def DemoteFromMC():
3080
  """Demotes the current node from master candidate role.
3081

3082
  """
3083
  # try to ensure we're not the master by mistake
3084
  master, myself = ssconf.GetMasterAndMyself()
3085
  if master == myself:
3086
    _Fail("ssconf status shows I'm the master node, will not demote")
3087

    
3088
  result = utils.RunCmd([pathutils.DAEMON_UTIL, "check", constants.MASTERD])
3089
  if not result.failed:
3090
    _Fail("The master daemon is running, will not demote")
3091

    
3092
  try:
3093
    if os.path.isfile(pathutils.CLUSTER_CONF_FILE):
3094
      utils.CreateBackup(pathutils.CLUSTER_CONF_FILE)
3095
  except EnvironmentError, err:
3096
    if err.errno != errno.ENOENT:
3097
      _Fail("Error while backing up cluster file: %s", err, exc=True)
3098

    
3099
  utils.RemoveFile(pathutils.CLUSTER_CONF_FILE)
3100

    
3101

    
3102
def _GetX509Filenames(cryptodir, name):
3103
  """Returns the full paths for the private key and certificate.
3104

3105
  """
3106
  return (utils.PathJoin(cryptodir, name),
3107
          utils.PathJoin(cryptodir, name, _X509_KEY_FILE),
3108
          utils.PathJoin(cryptodir, name, _X509_CERT_FILE))
3109

    
3110

    
3111
def CreateX509Certificate(validity, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3112
  """Creates a new X509 certificate for SSL/TLS.
3113

3114
  @type validity: int
3115
  @param validity: Validity in seconds
3116
  @rtype: tuple; (string, string)
3117
  @return: Certificate name and public part
3118

3119
  """
3120
  (key_pem, cert_pem) = \
3121
    utils.GenerateSelfSignedX509Cert(netutils.Hostname.GetSysName(),
3122
                                     min(validity, _MAX_SSL_CERT_VALIDITY))
3123

    
3124
  cert_dir = tempfile.mkdtemp(dir=cryptodir,
3125
                              prefix="x509-%s-" % utils.TimestampForFilename())
3126
  try:
3127
    name = os.path.basename(cert_dir)
3128
    assert len(name) > 5
3129

    
3130
    (_, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3131

    
3132
    utils.WriteFile(key_file, mode=0400, data=key_pem)
3133
    utils.WriteFile(cert_file, mode=0400, data=cert_pem)
3134

    
3135
    # Never return private key as it shouldn't leave the node
3136
    return (name, cert_pem)
3137
  except Exception:
3138
    shutil.rmtree(cert_dir, ignore_errors=True)
3139
    raise
3140

    
3141

    
3142
def RemoveX509Certificate(name, cryptodir=pathutils.CRYPTO_KEYS_DIR):
3143
  """Removes a X509 certificate.
3144

3145
  @type name: string
3146
  @param name: Certificate name
3147

3148
  """
3149
  (cert_dir, key_file, cert_file) = _GetX509Filenames(cryptodir, name)
3150

    
3151
  utils.RemoveFile(key_file)
3152
  utils.RemoveFile(cert_file)
3153

    
3154
  try:
3155
    os.rmdir(cert_dir)
3156
  except EnvironmentError, err:
3157
    _Fail("Cannot remove certificate directory '%s': %s",
3158
          cert_dir, err)
3159

    
3160

    
3161
def _GetImportExportIoCommand(instance, mode, ieio, ieargs):
3162
  """Returns the command for the requested input/output.
3163

3164
  @type instance: L{objects.Instance}
3165
  @param instance: The instance object
3166
  @param mode: Import/export mode
3167
  @param ieio: Input/output type
3168
  @param ieargs: Input/output arguments
3169

3170
  """
3171
  assert mode in (constants.IEM_IMPORT, constants.IEM_EXPORT)
3172

    
3173
  env = None
3174
  prefix = None
3175
  suffix = None
3176
  exp_size = None
3177

    
3178
  if ieio == constants.IEIO_FILE:
3179
    (filename, ) = ieargs
3180

    
3181
    if not utils.IsNormAbsPath(filename):
3182
      _Fail("Path '%s' is not normalized or absolute", filename)
3183

    
3184
    real_filename = os.path.realpath(filename)
3185
    directory = os.path.dirname(real_filename)
3186

    
3187
    if not utils.IsBelowDir(pathutils.EXPORT_DIR, real_filename):
3188
      _Fail("File '%s' is not under exports directory '%s': %s",
3189
            filename, pathutils.EXPORT_DIR, real_filename)
3190

    
3191
    # Create directory
3192
    utils.Makedirs(directory, mode=0750)
3193

    
3194
    quoted_filename = utils.ShellQuote(filename)
3195

    
3196
    if mode == constants.IEM_IMPORT:
3197
      suffix = "> %s" % quoted_filename
3198
    elif mode == constants.IEM_EXPORT:
3199
      suffix = "< %s" % quoted_filename
3200

    
3201
      # Retrieve file size
3202
      try:
3203
        st = os.stat(filename)
3204
      except EnvironmentError, err:
3205
        logging.error("Can't stat(2) %s: %s", filename, err)
3206
      else:
3207
        exp_size = utils.BytesToMebibyte(st.st_size)
3208

    
3209
  elif ieio == constants.IEIO_RAW_DISK:
3210
    (disk, ) = ieargs
3211

    
3212
    real_disk = _OpenRealBD(disk)
3213

    
3214
    if mode == constants.IEM_IMPORT:
3215
      # we set here a smaller block size as, due to transport buffering, more
3216
      # than 64-128k will mostly ignored; we use nocreat to fail if the device
3217
      # is not already there or we pass a wrong path; we use notrunc to no
3218
      # attempt truncate on an LV device; we use oflag=dsync to not buffer too
3219
      # much memory; this means that at best, we flush every 64k, which will
3220
      # not be very fast
3221
      suffix = utils.BuildShellCmd(("| dd of=%s conv=nocreat,notrunc"
3222
                                    " bs=%s oflag=dsync"),
3223
                                    real_disk.dev_path,
3224
                                    str(64 * 1024))
3225

    
3226
    elif mode == constants.IEM_EXPORT:
3227
      # the block size on the read dd is 1MiB to match our units
3228
      prefix = utils.BuildShellCmd("dd if=%s bs=%s count=%s |",
3229
                                   real_disk.dev_path,
3230
                                   str(1024 * 1024), # 1 MB
3231
                                   str(disk.size))
3232
      exp_size = disk.size
3233

    
3234
  elif ieio == constants.IEIO_SCRIPT:
3235
    (disk, disk_index, ) = ieargs
3236

    
3237
    assert isinstance(disk_index, (int, long))
3238

    
3239
    real_disk = _OpenRealBD(disk)
3240

    
3241
    inst_os = OSFromDisk(instance.os)
3242
    env = OSEnvironment(instance, inst_os)
3243

    
3244
    if mode == constants.IEM_IMPORT:
3245
      env["IMPORT_DEVICE"] = env["DISK_%d_PATH" % disk_index]
3246
      env["IMPORT_INDEX"] = str(disk_index)
3247
      script = inst_os.import_script
3248

    
3249
    elif mode == constants.IEM_EXPORT:
3250
      env["EXPORT_DEVICE"] = real_disk.dev_path
3251
      env["EXPORT_INDEX"] = str(disk_index)
3252
      script = inst_os.export_script
3253

    
3254
    # TODO: Pass special environment only to script
3255
    script_cmd = utils.BuildShellCmd("( cd %s && %s; )", inst_os.path, script)
3256

    
3257
    if mode == constants.IEM_IMPORT:
3258
      suffix = "| %s" % script_cmd
3259

    
3260
    elif mode == constants.IEM_EXPORT:
3261
      prefix = "%s |" % script_cmd
3262

    
3263
    # Let script predict size
3264
    exp_size = constants.IE_CUSTOM_SIZE
3265

    
3266
  else:
3267
    _Fail("Invalid %s I/O mode %r", mode, ieio)
3268

    
3269
  return (env, prefix, suffix, exp_size)
3270

    
3271

    
3272
def _CreateImportExportStatusDir(prefix):
3273
  """Creates status directory for import/export.
3274

3275
  """
3276
  return tempfile.mkdtemp(dir=pathutils.IMPORT_EXPORT_DIR,
3277
                          prefix=("%s-%s-" %
3278
                                  (prefix, utils.TimestampForFilename())))
3279

    
3280

    
3281
def StartImportExportDaemon(mode, opts, host, port, instance, component,
3282
                            ieio, ieioargs):
3283
  """Starts an import or export daemon.
3284

3285
  @param mode: Import/output mode
3286
  @type opts: L{objects.ImportExportOptions}
3287
  @param opts: Daemon options
3288
  @type host: string
3289
  @param host: Remote host for export (None for import)
3290
  @type port: int
3291
  @param port: Remote port for export (None for import)
3292
  @type instance: L{objects.Instance}
3293
  @param instance: Instance object
3294
  @type component: string
3295
  @param component: which part of the instance is transferred now,
3296
      e.g. 'disk/0'
3297
  @param ieio: Input/output type
3298
  @param ieioargs: Input/output arguments
3299

3300
  """
3301
  if mode == constants.IEM_IMPORT:
3302
    prefix = "import"
3303

    
3304
    if not (host is None and port is None):
3305
      _Fail("Can not specify host or port on import")
3306

    
3307
  elif mode == constants.IEM_EXPORT:
3308
    prefix = "export"
3309

    
3310
    if host is None or port is None:
3311
      _Fail("Host and port must be specified for an export")
3312

    
3313
  else:
3314
    _Fail("Invalid mode %r", mode)
3315

    
3316
  if (opts.key_name is None) ^ (opts.ca_pem is None):
3317
    _Fail("Cluster certificate can only be used for both key and CA")
3318

    
3319
  (cmd_env, cmd_prefix, cmd_suffix, exp_size) = \
3320
    _GetImportExportIoCommand(instance, mode, ieio, ieioargs)
3321

    
3322
  if opts.key_name is None:
3323
    # Use server.pem
3324
    key_path = pathutils.NODED_CERT_FILE
3325
    cert_path = pathutils.NODED_CERT_FILE
3326
    assert opts.ca_pem is None
3327
  else:
3328
    (_, key_path, cert_path) = _GetX509Filenames(pathutils.CRYPTO_KEYS_DIR,
3329
                                                 opts.key_name)
3330
    assert opts.ca_pem is not None
3331

    
3332
  for i in [key_path, cert_path]:
3333
    if not os.path.exists(i):
3334
      _Fail("File '%s' does not exist" % i)
3335

    
3336
  status_dir = _CreateImportExportStatusDir("%s-%s" % (prefix, component))
3337
  try:
3338
    status_file = utils.PathJoin(status_dir, _IES_STATUS_FILE)
3339
    pid_file = utils.PathJoin(status_dir, _IES_PID_FILE)
3340
    ca_file = utils.PathJoin(status_dir, _IES_CA_FILE)
3341

    
3342
    if opts.ca_pem is None:
3343
      # Use server.pem
3344
      ca = utils.ReadFile(pathutils.NODED_CERT_FILE)
3345
    else:
3346
      ca = opts.ca_pem
3347

    
3348
    # Write CA file
3349
    utils.WriteFile(ca_file, data=ca, mode=0400)
3350

    
3351
    cmd = [
3352
      pathutils.IMPORT_EXPORT_DAEMON,
3353
      status_file, mode,
3354
      "--key=%s" % key_path,
3355
      "--cert=%s" % cert_path,
3356
      "--ca=%s" % ca_file,
3357
      ]
3358

    
3359
    if host:
3360
      cmd.append("--host=%s" % host)
3361

    
3362
    if port:
3363
      cmd.append("--port=%s" % port)
3364

    
3365
    if opts.ipv6:
3366
      cmd.append("--ipv6")
3367
    else:
3368
      cmd.append("--ipv4")
3369

    
3370
    if opts.compress:
3371
      cmd.append("--compress=%s" % opts.compress)
3372

    
3373
    if opts.magic:
3374
      cmd.append("--magic=%s" % opts.magic)
3375

    
3376
    if exp_size is not None:
3377
      cmd.append("--expected-size=%s" % exp_size)
3378

    
3379
    if cmd_prefix:
3380
      cmd.append("--cmd-prefix=%s" % cmd_prefix)
3381

    
3382
    if cmd_suffix:
3383
      cmd.append("--cmd-suffix=%s" % cmd_suffix)
3384

    
3385
    if mode == constants.IEM_EXPORT:
3386
      # Retry connection a few times when connecting to remote peer
3387
      cmd.append("--connect-retries=%s" % constants.RIE_CONNECT_RETRIES)
3388
      cmd.append("--connect-timeout=%s" % constants.RIE_CONNECT_ATTEMPT_TIMEOUT)
3389
    elif opts.connect_timeout is not None:
3390
      assert mode == constants.IEM_IMPORT
3391
      # Overall timeout for establishing connection while listening
3392
      cmd.append("--connect-timeout=%s" % opts.connect_timeout)
3393

    
3394
    logfile = _InstanceLogName(prefix, instance.os, instance.name, component)
3395

    
3396
    # TODO: Once _InstanceLogName uses tempfile.mkstemp, StartDaemon has
3397
    # support for receiving a file descriptor for output
3398
    utils.StartDaemon(cmd, env=cmd_env, pidfile=pid_file,
3399
                      output=logfile)
3400

    
3401
    # The import/export name is simply the status directory name
3402
    return os.path.basename(status_dir)
3403

    
3404
  except Exception:
3405
    shutil.rmtree(status_dir, ignore_errors=True)
3406
    raise
3407

    
3408

    
3409
def GetImportExportStatus(names):
3410
  """Returns import/export daemon status.
3411

3412
  @type names: sequence
3413
  @param names: List of names
3414
  @rtype: List of dicts
3415
  @return: Returns a list of the state of each named import/export or None if a
3416
           status couldn't be read
3417

3418
  """
3419
  result = []
3420

    
3421
  for name in names:
3422
    status_file = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name,
3423
                                 _IES_STATUS_FILE)
3424

    
3425
    try:
3426
      data = utils.ReadFile(status_file)
3427
    except EnvironmentError, err:
3428
      if err.errno != errno.ENOENT:
3429
        raise
3430
      data = None
3431

    
3432
    if not data:
3433
      result.append(None)
3434
      continue
3435

    
3436
    result.append(serializer.LoadJson(data))
3437

    
3438
  return result
3439

    
3440

    
3441
def AbortImportExport(name):
3442
  """Sends SIGTERM to a running import/export daemon.
3443

3444
  """
3445
  logging.info("Abort import/export %s", name)
3446

    
3447
  status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3448
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3449

    
3450
  if pid:
3451
    logging.info("Import/export %s is running with PID %s, sending SIGTERM",
3452
                 name, pid)
3453
    utils.IgnoreProcessNotFound(os.kill, pid, signal.SIGTERM)
3454

    
3455

    
3456
def CleanupImportExport(name):
3457
  """Cleanup after an import or export.
3458

3459
  If the import/export daemon is still running it's killed. Afterwards the
3460
  whole status directory is removed.
3461

3462
  """
3463
  logging.info("Finalizing import/export %s", name)
3464

    
3465
  status_dir = utils.PathJoin(pathutils.IMPORT_EXPORT_DIR, name)
3466

    
3467
  pid = utils.ReadLockedPidFile(utils.PathJoin(status_dir, _IES_PID_FILE))
3468

    
3469
  if pid:
3470
    logging.info("Import/export %s is still running with PID %s",
3471
                 name, pid)
3472
    utils.KillProcess(pid, waitpid=False)
3473

    
3474
  shutil.rmtree(status_dir, ignore_errors=True)
3475

    
3476

    
3477
def _FindDisks(nodes_ip, disks):
3478
  """Sets the physical ID on disks and returns the block devices.
3479

3480
  """
3481
  # set the correct physical ID
3482
  my_name = netutils.Hostname.GetSysName()
3483
  for cf in disks:
3484
    cf.SetPhysicalID(my_name, nodes_ip)
3485

    
3486
  bdevs = []
3487

    
3488
  for cf in disks:
3489
    rd = _RecursiveFindBD(cf)
3490
    if rd is None:
3491
      _Fail("Can't find device %s", cf)
3492
    bdevs.append(rd)
3493
  return bdevs
3494

    
3495

    
3496
def DrbdDisconnectNet(nodes_ip, disks):
3497
  """Disconnects the network on a list of drbd devices.
3498

3499
  """
3500
  bdevs = _FindDisks(nodes_ip, disks)
3501

    
3502
  # disconnect disks
3503
  for rd in bdevs:
3504
    try:
3505
      rd.DisconnectNet()
3506
    except errors.BlockDeviceError, err:
3507
      _Fail("Can't change network configuration to standalone mode: %s",
3508
            err, exc=True)
3509

    
3510

    
3511
def DrbdAttachNet(nodes_ip, disks, instance_name, multimaster):
3512
  """Attaches the network on a list of drbd devices.
3513

3514
  """
3515
  bdevs = _FindDisks(nodes_ip, disks)
3516

    
3517
  if multimaster:
3518
    for idx, rd in enumerate(bdevs):
3519
      try:
3520
        _SymlinkBlockDev(instance_name, rd.dev_path, idx)
3521
      except EnvironmentError, err:
3522
        _Fail("Can't create symlink: %s", err)
3523
  # reconnect disks, switch to new master configuration and if
3524
  # needed primary mode
3525
  for rd in bdevs:
3526
    try:
3527
      rd.AttachNet(multimaster)
3528
    except errors.BlockDeviceError, err:
3529
      _Fail("Can't change network configuration: %s", err)
3530

    
3531
  # wait until the disks are connected; we need to retry the re-attach
3532
  # if the device becomes standalone, as this might happen if the one
3533
  # node disconnects and reconnects in a different mode before the
3534
  # other node reconnects; in this case, one or both of the nodes will
3535
  # decide it has wrong configuration and switch to standalone
3536

    
3537
  def _Attach():
3538
    all_connected = True
3539

    
3540
    for rd in bdevs:
3541
      stats = rd.GetProcStatus()
3542

    
3543
      all_connected = (all_connected and
3544
                       (stats.is_connected or stats.is_in_resync))
3545

    
3546
      if stats.is_standalone:
3547
        # peer had different config info and this node became
3548
        # standalone, even though this should not happen with the
3549
        # new staged way of changing disk configs
3550
        try:
3551
          rd.AttachNet(multimaster)
3552
        except errors.BlockDeviceError, err:
3553
          _Fail("Can't change network configuration: %s", err)
3554

    
3555
    if not all_connected:
3556
      raise utils.RetryAgain()
3557

    
3558
  try:
3559
    # Start with a delay of 100 miliseconds and go up to 5 seconds
3560
    utils.Retry(_Attach, (0.1, 1.5, 5.0), 2 * 60)
3561
  except utils.RetryTimeout:
3562
    _Fail("Timeout in disk reconnecting")
3563

    
3564
  if multimaster:
3565
    # change to primary mode
3566
    for rd in bdevs:
3567
      try:
3568
        rd.Open()
3569
      except errors.BlockDeviceError, err:
3570
        _Fail("Can't change to primary mode: %s", err)
3571

    
3572

    
3573
def DrbdWaitSync(nodes_ip, disks):
3574
  """Wait until DRBDs have synchronized.
3575

3576
  """
3577
  def _helper(rd):
3578
    stats = rd.GetProcStatus()
3579
    if not (stats.is_connected or stats.is_in_resync):
3580
      raise utils.RetryAgain()
3581
    return stats
3582

    
3583
  bdevs = _FindDisks(nodes_ip, disks)
3584

    
3585
  min_resync = 100
3586
  alldone = True
3587
  for rd in bdevs:
3588
    try:
3589
      # poll each second for 15 seconds
3590
      stats = utils.Retry(_helper, 1, 15, args=[rd])
3591
    except utils.RetryTimeout:
3592
      stats = rd.GetProcStatus()
3593
      # last check
3594
      if not (stats.is_connected or stats.is_in_resync):
3595
        _Fail("DRBD device %s is not in sync: stats=%s", rd, stats)
3596
    alldone = alldone and (not stats.is_in_resync)
3597
    if stats.sync_percent is not None:
3598
      min_resync = min(min_resync, stats.sync_percent)
3599

    
3600
  return (alldone, min_resync)
3601

    
3602

    
3603
def GetDrbdUsermodeHelper():
3604
  """Returns DRBD usermode helper currently configured.
3605

3606
  """
3607
  try:
3608
    return bdev.BaseDRBD.GetUsermodeHelper()
3609
  except errors.BlockDeviceError, err:
3610
    _Fail(str(err))
3611

    
3612

    
3613
def PowercycleNode(hypervisor_type):
3614
  """Hard-powercycle the node.
3615

3616
  Because we need to return first, and schedule the powercycle in the
3617
  background, we won't be able to report failures nicely.
3618

3619
  """
3620
  hyper = hypervisor.GetHypervisor(hypervisor_type)
3621
  try:
3622
    pid = os.fork()
3623
  except OSError:
3624
    # if we can't fork, we'll pretend that we're in the child process
3625
    pid = 0
3626
  if pid > 0:
3627
    return "Reboot scheduled in 5 seconds"
3628
  # ensure the child is running on ram
3629
  try:
3630
    utils.Mlockall()
3631
  except Exception: # pylint: disable=W0703
3632
    pass
3633
  time.sleep(5)
3634
  hyper.PowercycleNode()
3635

    
3636

    
3637
def _VerifyRestrictedCmdName(cmd):
3638
  """Verifies a remote command name.
3639

3640
  @type cmd: string
3641
  @param cmd: Command name
3642
  @rtype: tuple; (boolean, string or None)
3643
  @return: The tuple's first element is the status; if C{False}, the second
3644
    element is an error message string, otherwise it's C{None}
3645

3646
  """
3647
  if not cmd.strip():
3648
    return (False, "Missing command name")
3649

    
3650
  if os.path.basename(cmd) != cmd:
3651
    return (False, "Invalid command name")
3652

    
3653
  if not constants.EXT_PLUGIN_MASK.match(cmd):
3654
    return (False, "Command name contains forbidden characters")
3655

    
3656
  return (True, None)
3657

    
3658

    
3659
def _CommonRestrictedCmdCheck(path, owner):
3660
  """Common checks for remote command file system directories and files.
3661

3662
  @type path: string
3663
  @param path: Path to check
3664
  @param owner: C{None} or tuple containing UID and GID
3665
  @rtype: tuple; (boolean, string or C{os.stat} result)
3666
  @return: The tuple's first element is the status; if C{False}, the second
3667
    element is an error message string, otherwise it's the result of C{os.stat}
3668

3669
  """
3670
  if owner is None:
3671
    # Default to root as owner
3672
    owner = (0, 0)
3673

    
3674
  try:
3675
    st = os.stat(path)
3676
  except EnvironmentError, err:
3677
    return (False, "Can't stat(2) '%s': %s" % (path, err))
3678

    
3679
  if stat.S_IMODE(st.st_mode) & (~_RCMD_MAX_MODE):
3680
    return (False, "Permissions on '%s' are too permissive" % path)
3681

    
3682
  if (st.st_uid, st.st_gid) != owner:
3683
    (owner_uid, owner_gid) = owner
3684
    return (False, "'%s' is not owned by %s:%s" % (path, owner_uid, owner_gid))
3685

    
3686
  return (True, st)
3687

    
3688

    
3689
def _VerifyRestrictedCmdDirectory(path, _owner=None):
3690
  """Verifies remote command directory.
3691

3692
  @type path: string
3693
  @param path: Path to check
3694
  @rtype: tuple; (boolean, string or None)
3695
  @return: The tuple's first element is the status; if C{False}, the second
3696
    element is an error message string, otherwise it's C{None}
3697

3698
  """
3699
  (status, value) = _CommonRestrictedCmdCheck(path, _owner)
3700

    
3701
  if not status:
3702
    return (False, value)
3703

    
3704
  if not stat.S_ISDIR(value.st_mode):
3705
    return (False, "Path '%s' is not a directory" % path)
3706

    
3707
  return (True, None)
3708

    
3709

    
3710
def _VerifyRestrictedCmd(path, cmd, _owner=None):
3711
  """Verifies a whole remote command and returns its executable filename.
3712

3713
  @type path: string
3714
  @param path: Directory containing remote commands
3715
  @type cmd: string
3716
  @param cmd: Command name
3717
  @rtype: tuple; (boolean, string)
3718
  @return: The tuple's first element is the status; if C{False}, the second
3719
    element is an error message string, otherwise the second element is the
3720
    absolute path to the executable
3721

3722
  """
3723
  executable = utils.PathJoin(path, cmd)
3724

    
3725
  (status, msg) = _CommonRestrictedCmdCheck(executable, _owner)
3726

    
3727
  if not status:
3728
    return (False, msg)
3729

    
3730
  if not utils.IsExecutable(executable):
3731
    return (False, "access(2) thinks '%s' can't be executed" % executable)
3732

    
3733
  return (True, executable)
3734

    
3735

    
3736
def _PrepareRestrictedCmd(path, cmd,
3737
                          _verify_dir=_VerifyRestrictedCmdDirectory,
3738
                          _verify_name=_VerifyRestrictedCmdName,
3739
                          _verify_cmd=_VerifyRestrictedCmd):
3740
  """Performs a number of tests on a remote command.
3741

3742
  @type path: string
3743
  @param path: Directory containing remote commands
3744
  @type cmd: string
3745
  @param cmd: Command name
3746
  @return: Same as L{_VerifyRestrictedCmd}
3747

3748
  """
3749
  # Verify the directory first
3750
  (status, msg) = _verify_dir(path)
3751
  if status:
3752
    # Check command if everything was alright
3753
    (status, msg) = _verify_name(cmd)
3754

    
3755
  if not status:
3756
    return (False, msg)
3757

    
3758
  # Check actual executable
3759
  return _verify_cmd(path, cmd)
3760

    
3761

    
3762
def RunRestrictedCmd(cmd,
3763
                     _lock_timeout=_RCMD_LOCK_TIMEOUT,
3764
                     _lock_file=pathutils.RESTRICTED_COMMANDS_LOCK_FILE,
3765
                     _path=pathutils.RESTRICTED_COMMANDS_DIR,
3766
                     _sleep_fn=time.sleep,
3767
                     _prepare_fn=_PrepareRestrictedCmd,
3768
                     _runcmd_fn=utils.RunCmd,
3769
                     _enabled=constants.ENABLE_RESTRICTED_COMMANDS):
3770
  """Executes a remote command after performing strict tests.
3771

3772
  @type cmd: string
3773
  @param cmd: Command name
3774
  @rtype: string
3775
  @return: Command output
3776
  @raise RPCFail: In case of an error
3777

3778
  """
3779
  logging.info("Preparing to run remote command '%s'", cmd)
3780

    
3781
  if not _enabled:
3782
    _Fail("Remote commands disabled at configure time")
3783

    
3784
  lock = None
3785
  try:
3786
    cmdresult = None
3787
    try:
3788
      lock = utils.FileLock.Open(_lock_file)
3789
      lock.Exclusive(blocking=True, timeout=_lock_timeout)
3790

    
3791
      (status, value) = _prepare_fn(_path, cmd)
3792

    
3793
      if status:
3794
        cmdresult = _runcmd_fn([value], env={}, reset_env=True,
3795
                               postfork_fn=lambda _: lock.Unlock())
3796
      else:
3797
        logging.error(value)
3798
    except Exception: # pylint: disable=W0703
3799
      # Keep original error in log
3800
      logging.exception("Caught exception")
3801

    
3802
    if cmdresult is None:
3803
      logging.info("Sleeping for %0.1f seconds before returning",
3804
                   _RCMD_INVALID_DELAY)
3805
      _sleep_fn(_RCMD_INVALID_DELAY)
3806

    
3807
      # Do not include original error message in returned error
3808
      _Fail("Executing command '%s' failed" % cmd)
3809
    elif cmdresult.failed or cmdresult.fail_reason:
3810
      _Fail("Remote command '%s' failed: %s; output: %s",
3811
            cmd, cmdresult.fail_reason, cmdresult.output)
3812
    else:
3813
      return cmdresult.output
3814
  finally:
3815
    if lock is not None:
3816
      # Release lock at last
3817
      lock.Close()
3818
      lock = None
3819

    
3820

    
3821
def SetWatcherPause(until, _filename=pathutils.WATCHER_PAUSEFILE):
3822
  """Creates or removes the watcher pause file.
3823

3824
  @type until: None or number
3825
  @param until: Unix timestamp saying until when the watcher shouldn't run
3826

3827
  """
3828
  if until is None:
3829
    logging.info("Received request to no longer pause watcher")
3830
    utils.RemoveFile(_filename)
3831
  else:
3832
    logging.info("Received request to pause watcher until %s", until)
3833

    
3834
    if not ht.TNumber(until):
3835
      _Fail("Duration must be numeric")
3836

    
3837
    utils.WriteFile(_filename, data="%d\n" % (until, ), mode=0644)
3838

    
3839

    
3840
class HooksRunner(object):
3841
  """Hook runner.
3842

3843
  This class is instantiated on the node side (ganeti-noded) and not
3844
  on the master side.
3845

3846
  """
3847
  def __init__(self, hooks_base_dir=None):
3848
    """Constructor for hooks runner.
3849

3850
    @type hooks_base_dir: str or None
3851
    @param hooks_base_dir: if not None, this overrides the
3852
        L{pathutils.HOOKS_BASE_DIR} (useful for unittests)
3853

3854
    """
3855
    if hooks_base_dir is None:
3856
      hooks_base_dir = pathutils.HOOKS_BASE_DIR
3857
    # yeah, _BASE_DIR is not valid for attributes, we use it like a
3858
    # constant
3859
    self._BASE_DIR = hooks_base_dir # pylint: disable=C0103
3860

    
3861
  def RunLocalHooks(self, node_list, hpath, phase, env):
3862
    """Check that the hooks will be run only locally and then run them.
3863

3864
    """
3865
    assert len(node_list) == 1
3866
    node = node_list[0]
3867
    _, myself = ssconf.GetMasterAndMyself()
3868
    assert node == myself
3869

    
3870
    results = self.RunHooks(hpath, phase, env)
3871

    
3872
    # Return values in the form expected by HooksMaster
3873
    return {node: (None, False, results)}
3874

    
3875
  def RunHooks(self, hpath, phase, env):
3876
    """Run the scripts in the hooks directory.
3877

3878
    @type hpath: str
3879
    @param hpath: the path to the hooks directory which
3880
        holds the scripts
3881
    @type phase: str
3882
    @param phase: either L{constants.HOOKS_PHASE_PRE} or
3883
        L{constants.HOOKS_PHASE_POST}
3884
    @type env: dict
3885
    @param env: dictionary with the environment for the hook
3886
    @rtype: list
3887
    @return: list of 3-element tuples:
3888
      - script path
3889
      - script result, either L{constants.HKR_SUCCESS} or
3890
        L{constants.HKR_FAIL}
3891
      - output of the script
3892

3893
    @raise errors.ProgrammerError: for invalid input
3894
        parameters
3895

3896
    """
3897
    if phase == constants.HOOKS_PHASE_PRE:
3898
      suffix = "pre"
3899
    elif phase == constants.HOOKS_PHASE_POST:
3900
      suffix = "post"
3901
    else:
3902
      _Fail("Unknown hooks phase '%s'", phase)
3903

    
3904
    subdir = "%s-%s.d" % (hpath, suffix)
3905
    dir_name = utils.PathJoin(self._BASE_DIR, subdir)
3906

    
3907
    results = []
3908

    
3909
    if not os.path.isdir(dir_name):
3910
      # for non-existing/non-dirs, we simply exit instead of logging a
3911
      # warning at every operation
3912
      return results
3913

    
3914
    runparts_results = utils.RunParts(dir_name, env=env, reset_env=True)
3915

    
3916
    for (relname, relstatus, runresult) in runparts_results:
3917
      if relstatus == constants.RUNPARTS_SKIP:
3918
        rrval = constants.HKR_SKIP
3919
        output = ""
3920
      elif relstatus == constants.RUNPARTS_ERR:
3921
        rrval = constants.HKR_FAIL
3922
        output = "Hook script execution error: %s" % runresult
3923
      elif relstatus == constants.RUNPARTS_RUN:
3924
        if runresult.failed:
3925
          rrval = constants.HKR_FAIL
3926
        else:
3927
          rrval = constants.HKR_SUCCESS
3928
        output = utils.SafeEncode(runresult.output.strip())
3929
      results.append(("%s/%s" % (subdir, relname), rrval, output))
3930

    
3931
    return results
3932

    
3933

    
3934
class IAllocatorRunner(object):
3935
  """IAllocator runner.
3936

3937
  This class is instantiated on the node side (ganeti-noded) and not on
3938
  the master side.
3939

3940
  """
3941
  @staticmethod
3942
  def Run(name, idata):
3943
    """Run an iallocator script.
3944

3945
    @type name: str
3946
    @param name: the iallocator script name
3947
    @type idata: str
3948
    @param idata: the allocator input data
3949

3950
    @rtype: tuple
3951
    @return: two element tuple of:
3952
       - status
3953
       - either error message or stdout of allocator (for success)
3954

3955
    """
3956
    alloc_script = utils.FindFile(name, constants.IALLOCATOR_SEARCH_PATH,
3957
                                  os.path.isfile)
3958
    if alloc_script is None:
3959
      _Fail("iallocator module '%s' not found in the search path", name)
3960

    
3961
    fd, fin_name = tempfile.mkstemp(prefix="ganeti-iallocator.")
3962
    try:
3963
      os.write(fd, idata)
3964
      os.close(fd)
3965
      result = utils.RunCmd([alloc_script, fin_name])
3966
      if result.failed:
3967
        _Fail("iallocator module '%s' failed: %s, output '%s'",
3968
              name, result.fail_reason, result.output)
3969
    finally:
3970
      os.unlink(fin_name)
3971

    
3972
    return result.stdout
3973

    
3974

    
3975
class DevCacheManager(object):
3976
  """Simple class for managing a cache of block device information.
3977

3978
  """
3979
  _DEV_PREFIX = "/dev/"
3980
  _ROOT_DIR = pathutils.BDEV_CACHE_DIR
3981

    
3982
  @classmethod
3983
  def _ConvertPath(cls, dev_path):
3984
    """Converts a /dev/name path to the cache file name.
3985

3986
    This replaces slashes with underscores and strips the /dev
3987
    prefix. It then returns the full path to the cache file.
3988

3989
    @type dev_path: str
3990
    @param dev_path: the C{/dev/} path name
3991
    @rtype: str
3992
    @return: the converted path name
3993

3994
    """
3995
    if dev_path.startswith(cls._DEV_PREFIX):
3996
      dev_path = dev_path[len(cls._DEV_PREFIX):]
3997
    dev_path = dev_path.replace("/", "_")
3998
    fpath = utils.PathJoin(cls._ROOT_DIR, "bdev_%s" % dev_path)
3999
    return fpath
4000

    
4001
  @classmethod
4002
  def UpdateCache(cls, dev_path, owner, on_primary, iv_name):
4003
    """Updates the cache information for a given device.
4004

4005
    @type dev_path: str
4006
    @param dev_path: the pathname of the device
4007
    @type owner: str
4008
    @param owner: the owner (instance name) of the device
4009
    @type on_primary: bool
4010
    @param on_primary: whether this is the primary
4011
        node nor not
4012
    @type iv_name: str
4013
    @param iv_name: the instance-visible name of the
4014
        device, as in objects.Disk.iv_name
4015

4016
    @rtype: None
4017

4018
    """
4019
    if dev_path is None:
4020
      logging.error("DevCacheManager.UpdateCache got a None dev_path")
4021
      return
4022
    fpath = cls._ConvertPath(dev_path)
4023
    if on_primary:
4024
      state = "primary"
4025
    else:
4026
      state = "secondary"
4027
    if iv_name is None:
4028
      iv_name = "not_visible"
4029
    fdata = "%s %s %s\n" % (str(owner), state, iv_name)
4030
    try:
4031
      utils.WriteFile(fpath, data=fdata)
4032
    except EnvironmentError, err:
4033
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)
4034

    
4035
  @classmethod
4036
  def RemoveCache(cls, dev_path):
4037
    """Remove data for a dev_path.
4038

4039
    This is just a wrapper over L{utils.io.RemoveFile} with a converted
4040
    path name and logging.
4041

4042
    @type dev_path: str
4043
    @param dev_path: the pathname of the device
4044

4045
    @rtype: None
4046

4047
    """
4048
    if dev_path is None:
4049
      logging.error("DevCacheManager.RemoveCache got a None dev_path")
4050
      return
4051
    fpath = cls._ConvertPath(dev_path)
4052
    try:
4053
      utils.RemoveFile(fpath)
4054
    except EnvironmentError, err:
4055
      logging.exception("Can't update bdev cache for %s: %s", dev_path, err)